+

WO2000077665A2 - Method and apparatus for automatically allocating staffing - Google Patents

Method and apparatus for automatically allocating staffing Download PDF

Info

Publication number
WO2000077665A2
WO2000077665A2 PCT/US2000/016032 US0016032W WO0077665A2 WO 2000077665 A2 WO2000077665 A2 WO 2000077665A2 US 0016032 W US0016032 W US 0016032W WO 0077665 A2 WO0077665 A2 WO 0077665A2
Authority
WO
WIPO (PCT)
Prior art keywords
factor information
model
obtaining
information relating
workload
Prior art date
Application number
PCT/US2000/016032
Other languages
French (fr)
Other versions
WO2000077665A8 (en
Inventor
Kelly Ray O'keefe
Original Assignee
Clinical Dynamics, Inc.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Clinical Dynamics, Inc. filed Critical Clinical Dynamics, Inc.
Priority to JP2001503075A priority Critical patent/JP2003526137A/en
Priority to AU54808/00A priority patent/AU5480800A/en
Publication of WO2000077665A2 publication Critical patent/WO2000077665A2/en
Publication of WO2000077665A8 publication Critical patent/WO2000077665A8/en

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H40/00—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices
    • G16H40/20—ICT specially adapted for the management or administration of healthcare resources or facilities; ICT specially adapted for the management or operation of medical equipment or devices for the management or administration of healthcare resources or facilities, e.g. managing hospital staff or surgery rooms

Definitions

  • the present invention relates to a method and apparatus for automatically allocating staffing and, more particularly, a method and apparatus for automatically allocating staffing m a health care environment.
  • the method and apparatus is also used for automatically seventy adjusting workload data and, more particularly, is a method and apparatus for automatically seventy adjusting workload data from a health care environment.
  • a continuing challenge m modern business is to manage and to use resources efficiently
  • One of the resources that need to be managed is staffing.
  • staffing At all different levels of staffing, such as doctors, nurses, and technicians in the health care environment, if there is too many staff, more resources are expended than necessary If, however, too few staff are present, not all the work that is required can be completed m time.
  • patient care can be compromised and quality suffers.
  • resources are not being efficiently used. This is complicated in the healthcare setting by the fact that the workload required to care for a patient is dependent upon how severely ill the patient is.
  • a traditional approach to this problem is to determine staffing levels using a tiered process.
  • a baseline is established for staffing, usually by a full time equivalent (“FTE") allocation for the work under consideration.
  • FTE full time equivalent
  • This staffing level is refined by having a knowledgeable supervisor estimate the staffing required to meet the work needs for the area in which the supervisor is responsible and then prospectively make minor changes to scheduled staffing levels
  • Final adjustments are made m real time, by contemporaneously calling additional staff in to manage unanticipated demand or by calling off staff if the work was overestimated.
  • Refinement of the baseline and final adjustments are notable for their subjective nature. The staffing decisions depend upon both the expenence of the decision- maker and upon their biases.
  • the present invention provides a method and apparatus for estimating staffing that uses a computer to estimate required staffing.
  • the computer stores in a memory vanous predictive factors regarding the work presently being performed, predictive factors regarding work that has been completed and the resources allocated to that completed work that allowed for its timely completion.
  • models are created that describe the correlation between that past work and the allocated resources. Thereafter, the models are used to determine the staffing levels required to adequately perform the present work
  • the computer stores in a memory vanous charactenstics regarding the present patient population, such as age, sex, and physical ailment, along with a representative past patient population having similar charactenstics and resources allocated to that past patient population, including staffing levels, that were required to care for the past patent population.
  • a memory vanous charactenstics regarding the present patient population such as age, sex, and physical ailment
  • a representative past patient population having similar charactenstics and resources allocated to that past patient population, including staffing levels, that were required to care for the past patent population.
  • the models are used to determine the staffing levels required to adequately care for the present patient population
  • the present invention allows for the refinement of the estimation of staffing levels based upon other factors such as local practice, adjustment for changing workload patterns, desired changes in resource utilization, and new information obtained as a result of work recently completed.
  • the present invention is able to make unbiased estimates of staffing requirements in order to more appropnately match the required staffing with the present work in a given facility.
  • the present invention is able to seventy adjust the histoncal data for patient charactenstics and provide an unbiased estimate of the productivity of the workforce caring for the historical patient population.
  • Fig. 1 illustrates a computer system capable of implementing the present invention
  • Fig. 2 illustrates a flowchart of the process according to the present invention of creating a model based upon a past patient population
  • Fig. 3 illustrates a flowchart of process used to predict the staffing required to support a present patient population according to the present invention.
  • Fig. 1 illustrates a staffing allocation system 100 according to the present invention.
  • the system is preferably configured as a distnubbed computer, having separate computers 110 for data input, processing, and output.
  • the separate computers are preferably tied together in some type of local area network, such that each computer 1 10 has access, through a server
  • the central database 130 need not be a single physical database, but rather is properly viewed as a collection of databases that the server 120 or computers
  • the staffing allocation system has three overall aspects.
  • the first aspect of the staffing allocation system is the use of completed work, predictive factors relating to such completed work, and associated resource allocations for the completed work to create models that descnbes the relationship between the past work and the resource allocations required for completion.
  • the second aspect of the present invention is the use of the created models to estimate the resource allocation required to complete a present work that needs to be completed, or alternatively, the work that can be completed for a given resource allocation.
  • the third aspect of the present mvention is the updating the created models based upon new or changing information
  • each patient can be viewed as having a certain amount of work that is associated with that patient.
  • Certain charactenstics of that patient are predictive of the amount of care ("work") that the given patient requires. Therefore, while the preferred embodiment discussed hereinafter will descnbe the present invention with regard to the staffing needed in a hospital, it will be appreciated that certain aspects of the present mvention can be applied to other work environments.
  • work or "project” is associated with task requinng completion, such as all of the work associated with admimstenng care to a given patient, this is distinguished from the term “workload,” which is used m association with the resource allocation needed to complete the work, or, in the specific embodiment, properly administer care to the patient.
  • the preferred embodiments of the present mvention will be descnbed with reference to the work that is performed in the hospital environment and the staffing associated therewith. Initially, however, background is provided which will assist in understanding the system implemented by the preferred embodiment of the present invention.
  • the patient can be viewed as requinng vanous different staff to allocate different amounts of time (workload) to the patient m order to properly administer care to the patient. Since records are conventionally kept on each patient, the present inventors have determined it useful to view the patient as requinng a certain workload from vanous different staff of the hospital, such as doctors, nurses, clerical, phlebotomist, lab tech, etc.
  • workload units Associated with each of these different staff are workload units, or workumts, required to care for the patient.
  • the workload units may be broken down mto skill levels, such as nursing, clencal, phlebotomist, lab tech, etc. or may be overall workload units m any unit desired as long as that workload reference data is captured m the reference database
  • Table 1 A typical cost center breakdown for an acute care hospital is provided in Table 1 below as an indication of the type of workload data detail that is almost universally available
  • the cost centers such as the clinical laboratory, may be further broken down into the level of detail that is generally available for each of the cost centers. This process is illustrated above for the clinical laboratory.
  • the workload units collected and predicted may be total workload for the patient, the clinical laboratory workload, the microbiology workload, or the microbiology technician workload, in increasing level of detail. Similar levels of detail will exist for each of the cost centers in Table I above. This level of detail is referred to as the granularity of the data. Finely granular data contains a great amount of detail and coarsely granular data has correspondingly less detail Finely granular data can typically be obtained for the entire facility, which allows more detailed predictions Furthermore, associated with each patient are a variety of characteristics that are potentially predictive of the workload required of the different staff.
  • charactenstics include, for example, age and sex, demographic factors such as zip code and insurance payer, operative procedures that patient is undergomg( ⁇ f any), histoncal and diagnostic information as reflected in ICD-9-CM diagnosis codes, procedure codes, laboratory test results, and physiologic measurements for the current episode of care.
  • demographic factors such as zip code and insurance payer
  • operative procedures that patient is undergomg( ⁇ f any) histoncal and diagnostic information as reflected in ICD-9-CM diagnosis codes, procedure codes, laboratory test results, and physiologic measurements for the current episode of care.
  • ICD-9-CM diagnosis codes e.g( ⁇ f any)
  • procedure codes e.g( ⁇ f any)
  • laboratory test results e.g., ambulatory care encounters, presc ⁇ ptions, laboratory test results, and so forth they will generally have predictive value for the current work requirements
  • the predictive charactenstics may be useful for predicting the workload for either an entire episode of care or on a day-by-day or shift
  • step 200 illustrates a flowchart of the process according to the present invention of creating a model based upon a past patient population.
  • the overall process will first be described, with each of the process steps elaborated upon more fully hereinafter.
  • many such created models are typically needed m order to descnbe an entire patient population.
  • many different models are created, with each created model charactenzmg the relationship between the staffing requirements and a particular type of patient.
  • the appropnate model is selected subsequently for prediction, based upon the target patient charactenstics, from among the many models available.
  • a reference data set containing case data related to a past patient population is obtained so that it can be operated upon by the processor.
  • this data set is referred to as the training set.
  • the reference data set is typically a discharge data set, such a modified UB-92 discharge data set to which workload units have been linked.
  • Other financial and clinical data elements may also be linked to this data set, as WO 00/77665 PCT/USOO/l 6032
  • the discharge data set will contain fields relating to patient demographics, and to charactenstics of this episode of care
  • the workload units linked to the discharge data set for a particular patient may include detailed workunit descnptors or summary descnptors
  • An exemplary discharge data set for a single discharged patient with exemplary linked workload units is illustrated in Table II below, with this table providing the short name for an exemplary set of fields that are associated with a single patient, and a description of that field. Appendix A provides more detailed descnptions for certain of these and other exemplary fields associated with the discharge data set.
  • the discharge data set may also contain or be linked to another data set, such as a data set that contains laboratory test results such as blood sugar, white count, or hematocnt Or a data set that contains pharmacy prescnption data such as medication, dose, and time of administration. Or a data WO 00/77665 PCT/USOO/l 6032
  • physiologic monitoring data such as blood pressure, respiratory rate or temperature
  • other data sets that contain such similar data as is collected dunng a hospital episode of care
  • the discharge data set may also contain such information for multiple episodes of care for patients who were hospitalized more than once dunng the pe ⁇ od covered by the data set and may be further linked to outpatient and ambulatory data, insurance data, and similar healthcare related data.
  • the entire population of patients discharged from a healthcare system or a hospital for a year or more is preferably used, such that the numbers of past patients are large enough to provide for adequate predictions of workloads for various types, or groups, of patients Cases may be added to the reference data set from another source if necessary
  • a reference data set is prepared from a general hospital and the predictions are to be made in a hospital with a large neurosurgical population not reflected in the prepared reference data set
  • additional discharged patients from a source with a large representation of neurosurgical cases may be added to the reference data set. This allows the reference data set more closely correlate to the typical patient population in the facility for which staffing requires estimation
  • aggregations of patients that have similar charactenstics are made.
  • groupings may be by procedure, diagnosis, or other charactenstics.
  • One grouping that is useful due to its content as well as widespread usage is that of a Diagnosis Related Group (DRG) vanant; HCFA- DRG, APR-DRG, R-DRG, APG or a similar resource-based group.
  • the processor associated with one of the computers 110 implements these groupings by using established extraction parameters that cause association of a single discharged patient and the data corresponding thereto with a particular group. It should be noted, however, that a single discharge patient can be associated with multiple groupings if the aggregating categories are different.
  • the parameters used to establish these groupings will be called "DRG" parameters in the remainder of this discussion, but they can actually be any selection cntena.
  • the DRG field is used to assign the grouping. By identifying this DRG field within the discharge data set, this allows the computer 110 to select from this discharge data set all of the cases to be associated with the DRG grouping. So, for example, all patients who are assigned to DRG 127, "Heart Failure and Shock ' may be considered as a single group to develop a model for workload prediction for patients who are admitted with heart failure and shock.
  • the computer 110 operates on all of the discharged patient cases, and identifies all of the appropriate groupings for those cases.
  • This can be implemented m many ways, such as by adding group fields to the already existing data for a particular discharged patients discharge data set or extracting all patients m a group into a separate data file. No matter how implemented, there becomes established a number of different groups, each distinguished from the other by the parameters, such as the DRG parameters mentioned above. WO 00/77665 PCT/USOO/l 6032
  • step 214 for each group, the operator or computer 110 selects a set of candidate predictive factors.
  • a person knowledgeable about the group being considered preferably performs or monitors this selection process, or evaluates the results While for most groups these factors will include sex and age, among others, any measured patient charactenstic may be a predictive factor, including insurance payer, zip code, diagnosis, admission temperature, and others
  • age and sex one of the most available and powerful predictive factors, pnma ⁇ ly due to is current widespread use as a diagnostic tool, are ICD-9-CM diagnoses. Other diagnostic categones may also be used, but standardized definitions are desirable. Any number of candidate predictive factors for a group can be used, there typically being on the order of tens or hundreds of such candidate factors.
  • step 216 follows in which certain of the candidate predictive factors may be clustered together to form a clustered candidate predictive factor.
  • This operation is such that the existence of any one of the clustered candidate predictive factors will result m an indication of the presence of the clustered candidate predictive factor.
  • ICD- 9-CM codes 412 and V45.82 are respectively, "history of an acute myocardial infarction” and “history of a percutaneous translummal coronary angiography " While these codes may be used as separate candidate predictive factors, they may also be clustered into a clustered candidate predictive factor, such as C1001 for instance.
  • the clusters may be based on clinical c ⁇ tena, as here, or may be based on any other cnte ⁇ a of interest.
  • clinical cntena corresponding to ICD-9-CM codes 402, 4020. 40200, 40201, 4021, 40210, 40211, 4029, 40290, and 40291 are candidate predictive factors that have been clustered together to form a clustered candidate predictive factor, labeled CL0011.
  • Step 218 follows thereafter, which begins the process of selecting those candidate predictive factors for the group that will be used as actual predictive factors. Since there are thousands of ICD- 9-CM codes, many codes that can be clustered together, and numerous other possible patient charactenstics, this can lead to there being an extremely large group of candidate diagnosis-related predictive factors.
  • the selection of predictive factors from these candidates may be accomplished either automatically by the computer or interactively by the operator The automatic process is similar to the interactive process that is described immediately hereafter.
  • To initiate the process of selecting the actual predictive factors the presence or absence of each candidate predictive factor, inclusive of the clustered candidate predictive factors, for each case in the discharge data set is summanzed.
  • a spreadsheet or other listing is generated by the computer 110 that provides, for each candidate predictive factor, such as individual diagnoses, or collectively diagnoses that make up a cluster, such as CL0011, the number of patients who displayed that candidate predictive factor (count), the length of stay (los) and the average workload units of each type consumed by the episodes of care that had that factor associated with them (workload units).
  • candidate predictive factor such as individual diagnoses, or collectively diagnoses that make up a cluster, such as CL0011
  • count the number of patients who displayed that candidate predictive factor
  • los the length of stay
  • workload units the average workload units of each type consumed by the episodes of care that had that factor associated with them
  • the operator can then examine the list and interactively select the candidate predictive factors to examine, up to the number available
  • the candidate predictive factors can be automatically selected based upon preset critena such as number of occurrences, weighted values such as (occurrences * workumts) to emphasize factors that have a large overall effect, or upon factors such as (average workun ⁇ ts-workun ⁇ ts) 2 to emphasize extreme values.
  • Additional predictive factors may be added automatically or based upon manual selection. Some of these factors may be selected based upon the clinical status of the patient. For example, a pattent who is assigned to a group of patients who had an infection may provoke selection of predictive factors to include time from presentation to antibiotic administration, use of intravenous antibiotic, presentation white blood cell count , and infectious organism identified. A patient who is admitted with a myocardial infarction my have time to thrombolytic therapy, peak blood creattne kmase MB band level (a laboratory test), and chronic coumadin medication as candidate predictive factors that are selected. And an obstetnc patient may have history of prenatal care, chronic anti-epileptic drug therapy, and admission hematocnt as candidate predictive factors.
  • the number of predictive factors must be large enough to allow an accurate prediction, but small enough to allow generalization of the model to data other than the training data set.
  • a rule of thumb is to start with a number of predictive factors that is no greater than one eighth the number of training WO 00/77665 PCT/USOO/l 6032
  • Step 220 follows thereafter, in which a group input table is constructed by the computer 110 based upon the actual predictive factors, the reference data set for each case in the group and the associated workload value(s) for each case.
  • the group input table is constructed and indicates whether, for each case, the actual predictive factors exist.
  • Each actual predictive factor is coded as a binary "1" if the patient under study displays that factor, and is coded as a binary "0" if that factor is not displayed.
  • Continuous factors such as age or number of previous admissions, are calculated for each case as appropnate.
  • age it has been determined that ages under a certain minimum, and over a certain maximum, do not always provide any further predictive value for purposes of determining staffing. Accordingly, ages over this maximum, typically 85 years, and under the minimum, such as 45 years, for general medical/surgical patients, are entered as the maximum and minimum
  • ages over this maximum typically 85 years, and under the minimum, such as 45 years, for general medical/surgical patients, are entered as the maximum and minimum
  • An example group-input table is illustrated m Table 5, which table includes sex, tnmmed age, and six other predictive factors, as well as the total workload units associated with each case in the group.
  • Step 222 follows in which the user selects the model type from a list of approp ⁇ ate candidates.
  • Types of models that exist include a linear model type of the form.
  • Workload l/(l+exp-(a + b*Factorl+c*Factor2 +... +z*FactorN)) where Factor 1, Factor2...FactorN are the binary factors, either "0" or "1,” corresponding to each actual predictive factor. Workload is the actual workload required in carmg for the patient and a, b, c.z are the coefficients obtained by the model as a result of operating upon the group-input table.
  • Table 6 An exemplary result for three different linear models is provided below in Table 6, which illustrates the linear model developed for a model based on DRG 080, a model based on Procedure 8151, and a model based upon DRG 127. As illustrated, each of these has a corresponding model type, and a number of workload units will result when the data of a particular patient is applied to the model, which has the coefficients (a) through (g), m this instance, determined in the manner descnbed above.
  • model types such as other mathematical models or artificial neural network (neural net) model types may be selected. While a neural net model type does not lend itself to a simple numenc notation, it is implemented for prediction m the same manner as the parametric model types.
  • the operator may choose from a number of different model types in order to select a model that provides the best fit to the relationship between the predictive factors and the actual workload data. Alternatively, this selection may be made automatically by the implementation of the invention. This is accomplished by building each of the possible model types and selecting the one that provides the best predictions, using preset cntena, such as the highest correlation coefficient for the fit of the predicted to the actual workload or the smallest total model error. Appendix B provides additional information regarding model building.
  • step 224 follows, in which the analytical tool corresponding to the model type is used to process the group-input table m order to fit the data to the model type and build the model for the group, or to determine that the type of model selected is mappropnate and to indicate that another model should be selected.
  • the model building essentially performs an analysis of the group data to determine if the selected model type that can be used to describe the relationships between the known actual predictive factors for past patients that make up the group, and the workload associated with that group.
  • This model building can be performed using a vanety of techniques. For the mathematical models curve fitting, methods such as a least-squares, simplex, Newton-Raphson or similar methods may be used.
  • the objective function may be a function of the difference between the predicted and actual workload for all of the patients for whom the model is being developed.
  • a common function that is used m this type of optimization is the sum of the squares of the differences between the actual and predicted workload values.
  • the absolute value of the differences may also be used, as may other functions. If the former function is WO 00/77665 PCT/USOO/l 6032
  • the fit is the equivalent of a so-called least-squares fit of the model to the training data set
  • Exemplary tools that may be used to implement these fitting algorithms include custom-coded software modules, an OLE server such as Microsoft Excel with its built in analytical tools, an OLE server with add-in analytical extensions, or analytical software engines such as SPSS, SAS, Mathematica, Statistica, or other stand-alone analytical modules
  • the most common modules are either Excel or Excel with an analytical add in
  • step 224 invokes the neural net trainer or other non-parametnc method and passes it the group-input table(Table 5 from step 220 to create the model[Correct]
  • the training is as is known for neural nets, with one implementation of an artificial neural network with back-propagation training being equivalent to a steepest-gradient least-squares fit of the model to the training data
  • the model is archived for subsequent use m the database 130, bemg stored in the format that corresponds to the model type, such as,
  • the model may be validated against a test discharge data set to establish its generality and accuracy.
  • the model is used to predict the workload for discharge data sets of cases of the appropriate type, for which the actual workload is known These are called the test or validation data sets.
  • the accuracy of the model is expressed as the aggregated difference between the predicted and the actual workloads in the validation sets, and if this value is acceptably small, the model is accepted and is used for actual predictions. If the model is inaccurate, a new model type is built or additional, fewer, or different predictive factors are selected and the same model is rebuilt until an acceptably accurate and general model is found
  • steps 222 and 224 can be combined, such that the program searches for different model types and attempts to fit the data, such that if one model type is not acceptable, the computer 110 automatically proceeds to the next model type until an accurate and general model is developed.
  • steps 222 and 224 may be combined to develop two or more models on a selection of patients that are initially assigned to a single group.
  • the group is subdivided into two or more groups based on either operator input, an automatic decision tree process performed by the implementation of the invention, or a combination of the two If the operator selects the subdivision, this operation is equivalent to building two separate models, as described above
  • An exemplary situation in which the automatic decision tree process occurs is when the initial group includes both patients that have only been seen for one episode of care m the training discharge data set and also patients that have been seen multiple times If the model building process is unable to develop an acceptable model, the group is automatically subdivided into single- and multiple- episode patients, and separate models are developed for each group.
  • An exemplary situation m which the combined manual and decision tree process occurs is when the initial model shows a very large sex effect on the WO 00/77665 PCT/USOO/l 6032
  • this model can then be used to practice another aspect of the invention, which is the actual estimation of workload based upon a present patient population. Or to practice yet another aspect of the invention, which is the seventy adjustment of and productivity calculation for a histoncal collection of patients. The former aspect of the invention is discussed first.
  • the staffing evaluation system 100 can use these models to predict the staffing required to support a present patient population. This process will be descnbed with reference to Fig. 3.
  • patient data regarding the present patient population is obtained This patient data corresponds to the data that was previously discussed with respect to Fig 2
  • this data is input into the prediction module of the present invention
  • the present pattent data may comprise such possible data, or may be restricted to the data elements actually required to calculate the predictive factors m the accepted model.
  • step 312 in which the computer 110 uses the patient data set and determines the appropnate group in which to list each patient. Since the groupings have already been established, it is only necessary to assign the patient to an existing group.
  • This assignment can be automatically performed by computer 110 by assigning a group based upon a predetermined decision tree or other decision structure, which operates on the patient data set, performed manually by the operator, or, with some combination in which the computer 110 uses a decision tree to determine the most likely groups for a given patient, and the operator then selects the most appropnate group Generally, this is an automatic process accomplished by the computer alone.
  • a decision tree See Table 2
  • the patient will preferably be assigned to the group corresponding to that DRG.
  • a pattent might fall into two or more possible groupings
  • the pattent may be automatically assigned to a group based on a decision tree, may be assigned manually by the operator, WO 00/77665 PCT/USOO/l 6032
  • the most common implementation is for the assignment to be automatic, with the assignment made to the group that has the most accurate model.
  • This automatic assignment is used very commonly when multiple models have been automatically built by the implementation, such as sex-specific models or number-of-episode models as described above, to improve the accuracy of the predictions.
  • those groups to which the patient could be assigned can be presented to the operator, and the group to which the operator judges is most appropnate can then be selected.
  • step 314 the correct model corresponding to that group is looked up.
  • the factors and weights are also retrieved from the archive if the model is mathematical, or the non-parametnc model is loaded in the library routine if not.
  • step 316 follows, m which the patient data is used to determine the status of the actual predictive factors used for that group, and other data used by the model is obtained.
  • the presence or absence of diagnoses or other binary factors is used to determine the state ("0" or "1") of the binary predictive factors.
  • Continuous predictive factors are calculated such as the age (with thresholds applied as discussed above if applicable the number of previous admissions, and similar required continuous factors are processed. And other predictive factors such as sex and zip code are processed, as required.
  • step 318 in which the approp ⁇ ate model is run to determine the estimated workload associated with that patient, for the workload category of interest.
  • Sample patient data, actual predictive factors, and an example calculation for a linear model are illustrated in Tables 6A-6C below.
  • Table 6A illustrates another set of linear models developed for DRG 080, Procedure 8151, and DRG 127.
  • Table 6B illustrates a partial set of patient data for three different patients,
  • clusters C1001 contains diagnoses 3451 and 5433; C1002 contains diagnoses 41400, 41401, 41402,...41499.
  • Table 6B The first row of the archived model Table 6A is for DRG 080, which is applicable, and the model type is 4, which is a linear model.
  • Table 6C the following terms are summed in the linear model: WO 00/77665 PCT/USOO/l 6032
  • the predicted workload units are 10.108 + 0 + 2.1735 + 0 - 5 4895 or 6.792 workload units.
  • steps 310-318 is then repeated for eachpatient, to determine the estimated workload for a larger group, such as a ward or an entire hospital, and the workunit predictions are aggregated according to predetermined rules. This frequently amounts to simple summation, but patient care interaction factors may also be considered Steps 310-318 are also repeated for each patient and for each different category of workload that is to be estimated and similar aggregation occurs for each different workload category
  • the interaction of the matrix elements of patient rows and workload category columns may demonstrate second order patient care interactions that are considered m the aggregations.
  • This data may be long term to reflect baseline differences in the practices of the target hospital and the reference data set or may be short term data from the hospital to reflect slightly more modern practice patterns m the hospital than in the reference data set Typically, although not necessanly, this refinement is limited to alternativeng the predictive weights of the already determined predictive factors by no more than a preset amount, say 20% based upon the goals of the adjustment and operator knowledge of the likely magnitude of the effects
  • This may be implemented m the invention by repeating step 224 using a training discharge data set that contains the required data 21 elements but that incorporates discharges from the hospital of interest over the time period of interest rather than the large scale discharge data.
  • Step 224 is constrained to allow only small adjustments of the predictive weights, as is known in the field of constrained optimization
  • the model may be further fine-tuned using contemporaneous data (i.e. yesterday's and last week's), with the added advantage that many of the patients that will receive care in the predicted period will be represented in the refinement data set.
  • contemporaneous data i.e. yesterday's and last week's
  • new models for new groups can be developed as well, and groups can be consolidated and split.
  • the prediction may combine benchmark data and the local data m a relationship that reflects the management focus. This may result in a one-time adjustment or in a phased adjustment penod. For example 80% local data and 20% benchmark data may be used this quarter, 60% local data and 40% benchmark data next quarter, and so forth, until the benchmark data is used exclusively to predict the required workload. The latter stage is the equivalent of a finely granular targeted productivity based on ob j ective norms.
  • the model used m this process is a parametric linear model
  • the coefficients of the model as estimated individually on the benchmark data and local data may be simply combined in the desired ratio to develop the new model.
  • the model is non-parametnc or non-linear
  • the objective function of the model-building step is modified to weight the contnbutions of the local data and benchmark data appropriately and a new model is constructed as descnbed above for fine-tuning an existing model.
  • the present mvention can be integrated into either a special purpose reporting tool or a more general clinical outcomes reporting tool to facilitate user predictions of workforce requirements based upon current data or to allow user analysis of workforce efficiency as compared to histoncal or benchmark norms.
  • the user analysis of workforce efficiency is based upon the seventy adjustment and productivity calculation features of this invention. If Steps 310 to 318 are applied to a histoncal population of patients rather than to a present population of patients, the expected workload that would have been required to take care of each pattent, at the level of granulanty of the model and the histoncal data, is obtained.
  • This expected workload can be thought of as the workload to care for the pattent if he/she behaved like the average pattent with his/her predictive factors, and thus the expected workload has been seventy adjusted for the patient's predictive factors.
  • the productivity is calculated as the ratio of the actual workload collected with the histoncal data and the expected workload. This productivity can be used, as is known m business practice, to evaluate the efficiency of the workforce ca ⁇ ng for the patients. In the discussion above, it was also assumed for simplification of understanding that each patient was assigned to only a single group, and, therefore, a single model. This, however, is an over- WO 00/77665 PCT/USOO/l 6032
  • Similar cntena are used to select the models for predicting workload for an entire length of stay, a single day, single shift, and so forth.
  • the automatic selection of the appropnate model based on these cntena may be embodied in the implementation of the invention.
  • the appropnate model selected will then be based upon the input specified by the user, such that if estimates related to first-day first shift nursing are desired, then that model is chosen
  • Allowable Values Any number greater than or equal to zero and less than or equal to 9999999.99 with two decimal points.
  • Missing Data Logic Field is Null.
  • Allowable Values Any number greater than or equal to zero and less than or equal to 9999999.99 with two decimal points.
  • Missing Data Logic Field is Null.
  • Data Element Name ADMISSION DATE Indicators Using: All patient episode of care records. Definition: The date the patient was admitted to the health care organization for inpatient or outpatient service.
  • Valid date must be entered (e.g.: 02/30/1993 is not valid)
  • Missing Data Logic Field is considered missing if DC_DATE or LOS is missing, as it is calculated from them.
  • Valid date must be entered (e.g.: 02/30/93 is not valid).
  • Missing Data Logic Field is blank.
  • Allowable Values Any valid ICD-9-CM Code. Missing Data Logic: None Specified.
  • Element Name MEDICAL RECORD NUMBER (EOC) Indicators Using: All patient episode of care records. Definition: A health care organization provided number used to identify a specific episode of care.
  • Missing Data Logic Field is blank.
  • Valid date must be entered (e.g.: 02/30/93 is not valid).
  • Missing Data Logic Field is blank.
  • Missing Data Logic Field is Null.
  • Data Element Name SEX Indicators Using: All patient episode of care records. Definition: The sex of the patient as recorded at date of admission, outpatient services, or start of care.
  • Missing Data Logic Field is blank, or value does not represent distinguishable gender.
  • Missing Data Logic Field is blank, or value does not represent distinguishable race.
  • Allowable Values Refer to Dynittls.dbf for an up to date list. Missing Data Logic: Field is blank.
  • HOSPITAL IDENTIFICATION NUMBER Indicators Using: All patient episode of care records. Definition: A uniquely assigned identification number, used to reference health care organization level information. (CHW hospitals use the OSHPD Hospital ID.)
  • Allowable Values Any combination of letters and/or numbers that represents a single existing facility.
  • Missing Data Logic Field is blank.
  • Definition The category that describes the patient's admission status.
  • Missing Data Logic DC DATE or ADM DATE is not blank, and field is blank.
  • HOSPITAL SERVICE Indicators Using: All patient episode of care records. Definition: Abbreviation of the service provided to the patient by the health care organization.
  • ICD-9-CM Clinical Modification
  • Allowable Values Any valid ICD-9-CM code.
  • Allowable Values Any valid ICD-9-CM code. Missing Data Logic: Field is blank.
  • Allowable Values Any valid, encrypted doctor code. Missing Data Logic: None specified.
  • Missing Data Logic Field is blank, and corresponding ICD-9-CM procedure field is not blank.
  • Allowable Values Any valid, encrypted doctor code. Missing Data Logic: Field is blank.
  • Allowable Values Any valid HCFA DRG Code. (Usually between 1 and 500.) Missing Data Logic: Field is blank.
  • Allowable Values Any valid MDC Code. (Usually between 1 and 25, or 99.)
  • Missing Data Logic Field is blank.
  • Allowable Values Any whole number greater than or equal to zero. Missing Data Logic: Field is Null.
  • Missing Data Logic Field is considered missing if any of LOS, DRG APRDRG, or SEVERITY is missing. WO 00/77665 PCT/USOO/l 6032
  • Allowable Values Any whole number between 0 and 999, that is less than or equal to the LOS.
  • Missing Data Logic Field is Null.
  • Allowable Values Any whole number between 0 and 999, that is less than or equal to the LOS.
  • Missing Data Logic Field is Null.
  • Allowable Values Any whole number between 0 and 999, that is less than or equal to the LOS.
  • Missing Data Logic Field is Null.
  • Missing Data Logic Field is blank, or field represents a non-determinable value. *** WO 00/77665 PCT/USOO/l 6032
  • Allowable Values Any valid DC_DISPO code that represents a discharge to a location.
  • Missing Data Logic Field is blank.
  • Missing Data Logic Field is blank, and DC_DISPO indicates death of patient.
  • Missing Data Logic Field is blank.
  • BIOPSY RESULTS Indicators Using: All patient episode of care records. Definition: The category that describes the results of a biopsy performed during this episode of care.
  • Missing Data Logic Field is blank or evaluates to "99” WO 00/77665 PCT/USOO/l 6032
  • Missing Data Logic Field is blank, or field evaluates to a non-determinable value.
  • Missing Data Logic Field is blank or outside the range of allowable values. *** 4/
  • Allowable Values Any whole number between 0 and 9,999.
  • Missing Data Logic Field is blank and is categorized as neonatal.
  • APGAR is the Activity, Pulse, Grimace, Appearance, and Respiration sum score given to describe neonatal condition.
  • Missing Data Logic Field is blank and DX_ADMIT denotes birth. *** WO 00/77665 PCT/USOO/l 6032
  • Allowable Values Y N. Missing Data Logic: Field is blank or outside the range of allowable values.
  • Allowable Values Y/N Missing Data Logic: Field is blank or outside the range of allowable values.
  • Missing Data Logic Field is blank or outside the range of allowable values. ***
  • Allowable Values Y/N Missing Data Logic: Field is blank or outside the range of allowable values.
  • Missing Data Logic Field is blank or outside the range of allowable values.
  • Allowable Values Any 1-3 digit alphanumeric code. Missing Data Logic: None specified.
  • CODER Indicators Using: All patient episode of care records. Definition: The ID number of the initial coder of this episode of care. Short Name: CODER Format:
  • Allowable Values Any 1-3 digit alphanumeric code. Missing Data Logic: None specified.
  • Missing Data Logic Field is Null.
  • Allowable Values Any whole number greater or equal to zero and less than the total direct cost (COSTJDIR) of this episode of care.
  • Missing Data Logic Field is Null or contains data outside the valid range. WO 00/77665 PCT/USOO/l 6032
  • Missing Data Logic Field is Null and corresponding Ancillary Cost element (ANC_COST « «) is NOT Null, or, Field does not represent a determinable ancillary description category.
  • Allowable Values Any whole number that is greater than or equal to zero, and less than the Total Costs (COST_TOT) of this episode of care.
  • Missing Data Logic Field is Null and COST TOT is NOT Null, or Field is outside the range of allowable values. *** WO 00/77665 PCT/USOO/l 6032 O
  • Allowable Values Any whole number greater than or equal to zero, and less than the Total Costs (COST_TOT) of this episode of care.
  • Missing Data Logic Field is Null and COST_TOT is NOT Null, or Field is outside the range of allowable values.
  • Missing Data Logic Field is Null and one or more of the ICD-9CM codes indicate ventilator administration.
  • Allowable Values Any valid insurer contract ID. Missing Data Logic: Field is blank or does not represent a valid insurer contract ID.
  • Allowable Values 0-4 Missing Data Logic: Field is Null or outside the range of allowable values.
  • Missing Data Logic Field is Null, or outside the range of allowable values.
  • Missing Data Logic Field is blank or
  • Missing Data Logic Field is Null or outside the range of allowable values.
  • APR-DRG Grouper.Severity is also assigned when the case is grouped by any grouper
  • Allowable Values 0-4 Missing Data Logic: Field is Null or is not within the range of allowable values.
  • Missing Data Logic Field is blank or outside the range of allowable values.
  • Allowable Values User Defined Missing Data Logic: None specified.
  • Allowable Values Any value greater than or equal to zero that is less than the severity adjust total cost (SA_COST).
  • Missing Data Logic Field is Null, or Field not within the range of allowable values. ***
  • Allowable Values Any valid HCFA DRG code. Missing Data Logic: Field is blank or does not represent a valid HCFA DRG.
  • Allowable Values Any valid RDRG code. Missing Data Logic: Field is blank or outside the range of allowable values.
  • Allowable Values Any valid SRDRG Code Missing Data Logic: Field is blank or is outside the range of allowable values.
  • Allowable Values Any valid APR-DRG code. Missing Data Logic: Field is blank or does not represent a valid APR-DRG code.
  • Allowable Values Any valid Abulatory Patient Group code. Missing Data Logic: Field is blank or outside the range of allowable values.
  • Diagnosis code was present at the time of the patient's admission to this health care organization.
  • Missing Data Logic Field is blank or does not represent an allowable value.
  • Missing Data Logic Field is Null or does not represent an allowable value.
  • Missing Data Logic Field is Null or does not represent an allowable value. ***
  • Missing Data Logic Field is Null or does not represent an allowable value.
  • Missing Data Logic Field is Null or does not represent an allowed value. ***
  • Missing Data Logic Field is Null or does not represent an allowed value.
  • Missing Data Logic Field is Null or does not represent an allowed value. ***
  • Missing Data Logic Field is Null or does not represent an allowed value.
  • Missing Data Logic Field is Null or does not represent an allowed value. ***
  • Missing Data Logic Field is Null or does not represent an allowed value.
  • NEWBORN Bitmap (Binary Flag) Indicators Using: All patient episode of care records. Definition: Boolean flag that describes whether the patient is neonatal. Short Name: NEWBORN Format:
  • Missing Data Logic Field is Null or does not represent an allowed value.
  • Missing Data Logic Field is Null or does not represent an allowed value.
  • Missing Data Logic Field is Null or does not represent an allowed value. ***
  • PSYCH Format
  • Missing Data Logic Field is Null or does not represent an allowed value.
  • Missing Data Logic Field is Null or does not represent an allowed value.
  • Missing Data Logic Field is Null or does not represent an allowed value.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • General Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Health & Medical Sciences (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Human Resources & Organizations (AREA)
  • Tourism & Hospitality (AREA)
  • Development Economics (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Game Theory and Decision Science (AREA)
  • Physics & Mathematics (AREA)
  • Educational Administration (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Marketing (AREA)
  • Biomedical Technology (AREA)
  • Epidemiology (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Primary Health Care (AREA)
  • Public Health (AREA)
  • Medical Treatment And Welfare Office Work (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

The present invention provides a method using a computer to estimate staffing required for a present patient population. The computer operates upon data related to various characteristics regarding the present patient population, such as age, sex, and physical ailment, along with data related to a representative past patient population having similar characteristics and resources allocated to that past patient population, including staffing levels, that were required to care for the past patient population. Using the past patient characteristics and associated allocated resources as a guide, models are created that describe the correlation between that past patient population and the allocated resources. Thereafter, the models are used to determine the staffing levels required to adequately care for the present patient population.

Description

METHOD AND APPARATUS FOR AUTOMATICALLY ALLOCATING STAFFING
Field of the Invention
The present invention relates to a method and apparatus for automatically allocating staffing and, more particularly, a method and apparatus for automatically allocating staffing m a health care environment. The method and apparatus is also used for automatically seventy adjusting workload data and, more particularly, is a method and apparatus for automatically seventy adjusting workload data from a health care environment.
Background of the Invention
A continuing challenge m modern business is to manage and to use resources efficiently One of the resources that need to be managed is staffing. At all different levels of staffing, such as doctors, nurses, and technicians in the health care environment, if there is too many staff, more resources are expended than necessary If, however, too few staff are present, not all the work that is required can be completed m time. Thus, in a health care setting, if there is insufficient staff, patient care can be compromised and quality suffers. However, if there are too many staff, resources are not being efficiently used. This is complicated in the healthcare setting by the fact that the workload required to care for a patient is dependent upon how severely ill the patient is.
Still further, in many business environments, the work changes. In the health care environment, for instance, a sudden influx of patients will oftenttmes dramatically change the medical staffing needs. These work changes make it difficult to determine optimum staffing at any moment in time.
A traditional approach to this problem is to determine staffing levels using a tiered process. A baseline is established for staffing, usually by a full time equivalent ("FTE") allocation for the work under consideration. This staffing level is refined by having a knowledgeable supervisor estimate the staffing required to meet the work needs for the area in which the supervisor is responsible and then prospectively make minor changes to scheduled staffing levels Final adjustments are made m real time, by contemporaneously calling additional staff in to manage unanticipated demand or by calling off staff if the work was overestimated. Refinement of the baseline and final adjustments are notable for their subjective nature. The staffing decisions depend upon both the expenence of the decision- maker and upon their biases.
What is needed, therefore, is an apparatus and method of more efficiently and objectively allocating staffing resources What is further needed is an apparatus and method for seventy adjusting patient populations so the workload that was actually used to care for them can be compared. WO 00/77665 PCT/USOO/l 6032
Summary of the Invention
It is an object of the present invention to provide an apparatus and method of more efficiently allocating staffing resources.
It is a further object of the present invention to use data regarding past workload statistics in estimating staffing required for work presently being performed.
It is another object of the present mvention to use a computer to estimate the staffing required It is an object of the present mvention to provide an apparatus and method of more efficiently allocating staffing resources m a health care environment with a changing patient population.
It is a further object of the present invention to use data regarding past patient charactenstics and workload statistics in estimating staffing required for a present patient workload.
It is another object of the present invention to use a computer to estimate the staffing required for a present patient population.
It is another object of the present invention to provide an apparatus and method of companng workload used to care for different patient populations. It is a further object of the present invention to use data regarding past workload statistics and reference workload data to determine the productivity of healthcare delivery.
It is another object of the present invention to use a computer determine the seventy adjusted workload and productivity.
It is an object of the present invention to provide an apparatus and method of more accurately analyzing productivity.
In order to attain the above objects of the present invention, among others, the present invention provides a method and apparatus for estimating staffing that uses a computer to estimate required staffing. The computer stores in a memory vanous predictive factors regarding the work presently being performed, predictive factors regarding work that has been completed and the resources allocated to that completed work that allowed for its timely completion. Using the past work predictive factors and associated allocated resources as a guide, models are created that describe the correlation between that past work and the allocated resources. Thereafter, the models are used to determine the staffing levels required to adequately perform the present work
In a preferred embodiment, the computer stores in a memory vanous charactenstics regarding the present patient population, such as age, sex, and physical ailment, along with a representative past patient population having similar charactenstics and resources allocated to that past patient population, including staffing levels, that were required to care for the past patent population. Using the past patient charactenstics and associated allocated resources as a guide, models are created that descnbe the correlation between that past patient population and the allocated resources. Thereafter, WO 00/77665 PCT/USOO/l 6032
the models are used to determine the staffing levels required to adequately care for the present patient population
Furthermore, the present invention allows for the refinement of the estimation of staffing levels based upon other factors such as local practice, adjustment for changing workload patterns, desired changes in resource utilization, and new information obtained as a result of work recently completed.
By using data obtained from past and present workload, the present invention is able to make unbiased estimates of staffing requirements in order to more appropnately match the required staffing with the present work in a given facility.
By using histoncal patient and workload data the present invention is able to seventy adjust the histoncal data for patient charactenstics and provide an unbiased estimate of the productivity of the workforce caring for the historical patient population. Brief Description of the Drawings
The above and other objects, features, and advantages of the invention are further descnbed in the detailed description which follows, with reference to the drawings by way of non-hmitmg exemplary embodiments of the present invention, wherein like reference numerals represent similar parts of the present invention throughout several views and wherein
Fig. 1 illustrates a computer system capable of implementing the present invention; Fig. 2 illustrates a flowchart of the process according to the present invention of creating a model based upon a past patient population; and Fig. 3 illustrates a flowchart of process used to predict the staffing required to support a present patient population according to the present invention.
Detailed Description of the Preferred Embodiments
Fig. 1 illustrates a staffing allocation system 100 according to the present invention. As lllus- trated, the system is preferably configured as a distnbuted computer, having separate computers 110 for data input, processing, and output. As is known, the separate computers are preferably tied together in some type of local area network, such that each computer 1 10 has access, through a server
120, to a central database 130 which contains the data used by the present invention in order to obtain the desired staffing allocation information The central database 130 need not be a single physical database, but rather is properly viewed as a collection of databases that the server 120 or computers
110 have the capability of accessing in order to obtain the necessary data. Another portion of the staffing allocation system 100 according to the present invention is an application software program resident on either certain of the computers 110 or the server 120. As is known, the processor associated with one of the computers 110 or the server 120 executes program instructions that implement the features of the invention descnbed hereinafter. It will be understood that vanous WO 00/77665 PCT/USOO/l 6032
modifications to the particular system 100 can be made without departing from the intended scope of the present invention.
The staffing allocation system according to the present invention has three overall aspects. The first aspect of the staffing allocation system is the use of completed work, predictive factors relating to such completed work, and associated resource allocations for the completed work to create models that descnbes the relationship between the past work and the resource allocations required for completion. The second aspect of the present invention is the use of the created models to estimate the resource allocation required to complete a present work that needs to be completed, or alternatively, the work that can be completed for a given resource allocation. The third aspect of the present mvention is the updating the created models based upon new or changing information Each of these aspects will be discussed in more detail hereinafter.
Before discussing a particular embodiment of the present invention directed to the health care environment, and in particular a hospital health care environment, it should be understood that the present invention has applicability to environments that require staffing that are different than the health care environment.
In the health care environment, each patient can be viewed as having a certain amount of work that is associated with that patient. Certain charactenstics of that patient are predictive of the amount of care ("work") that the given patient requires. Therefore, while the preferred embodiment discussed hereinafter will descnbe the present invention with regard to the staffing needed in a hospital, it will be appreciated that certain aspects of the present mvention can be applied to other work environments.
A few terms that are used herein should also be understood While the term "work" or "project" is associated with task requinng completion, such as all of the work associated with admimstenng care to a given patient, this is distinguished from the term "workload," which is used m association with the resource allocation needed to complete the work, or, in the specific embodiment, properly administer care to the patient.
In light of the above, the preferred embodiments of the present mvention will be descnbed with reference to the work that is performed in the hospital environment and the staffing associated therewith. Initially, however, background is provided which will assist in understanding the system implemented by the preferred embodiment of the present invention. In the hospital environment, the patient can be viewed as requinng vanous different staff to allocate different amounts of time (workload) to the patient m order to properly administer care to the patient. Since records are conventionally kept on each patient, the present inventors have determined it useful to view the patient as requinng a certain workload from vanous different staff of the hospital, such as doctors, nurses, clerical, phlebotomist, lab tech, etc. Associated with each of these different staff are workload units, or workumts, required to care for the patient. The workload units may be broken down mto skill levels, such as nursing, clencal, phlebotomist, lab tech, etc. or may be overall workload units m any unit desired as long as that workload reference data is captured m the reference database A typical cost center breakdown for an acute care hospital is provided in Table 1 below as an indication of the type of workload data detail that is almost universally available
Figure imgf000006_0001
WO 00/77665 PCT/USOO/l 6032
Figure imgf000007_0001
The cost centers, such as the clinical laboratory, may be further broken down into the level of detail that is generally available for each of the cost centers. This process is illustrated above for the clinical laboratory. Thus the workload units collected and predicted may be total workload for the patient, the clinical laboratory workload, the microbiology workload, or the microbiology technician workload, in increasing level of detail. Similar levels of detail will exist for each of the cost centers in Table I above. This level of detail is referred to as the granularity of the data. Finely granular data contains a great amount of detail and coarsely granular data has correspondingly less detail Finely granular data can typically be obtained for the entire facility, which allows more detailed predictions Furthermore, associated with each patient are a variety of characteristics that are potentially predictive of the workload required of the different staff. These potentially predictive charactenstics include, for example, age and sex, demographic factors such as zip code and insurance payer, operative procedures that patient is undergomg(ιf any), histoncal and diagnostic information as reflected in ICD-9-CM diagnosis codes, procedure codes, laboratory test results, and physiologic measurements for the current episode of care. In addition, if historical data is available for previous hospitalizations, ambulatory care encounters, prescπptions, laboratory test results, and so forth they will generally have predictive value for the current work requirements The predictive charactenstics may be useful for predicting the workload for either an entire episode of care or on a day-by-day or shift-by-shift basis dunng an episode of care. Fig. 2 illustrates a flowchart of the process according to the present invention of creating a model based upon a past patient population. The overall process will first be described, with each of the process steps elaborated upon more fully hereinafter. Before describing the process of creating this model, it should be understood that many such created models are typically needed m order to descnbe an entire patient population. Thus, many different models are created, with each created model charactenzmg the relationship between the staffing requirements and a particular type of patient. The appropnate model is selected subsequently for prediction, based upon the target patient charactenstics, from among the many models available. In step 200, a reference data set containing case data related to a past patient population is obtained so that it can be operated upon by the processor. As is known, this data set is referred to as the training set. In particular, the reference data set is typically a discharge data set, such a modified UB-92 discharge data set to which workload units have been linked. Other financial and clinical data elements may also be linked to this data set, as WO 00/77665 PCT/USOO/l 6032
7 will be discussed. The discharge data set will contain fields relating to patient demographics, and to charactenstics of this episode of care The workload units linked to the discharge data set for a particular patient may include detailed workunit descnptors or summary descnptors In fact, there will typically be multiple detailed working descnptors, such as nursing, clerical, etc. as noted above, as well as time-segmentation of them, for example, first-day first-shift, first-day second-shift, etc. Alternatively, there may only be a single aggregate value of workumts depending upon the level of granularity available in the data and that desired in the model and in the predictions.
An exemplary discharge data set for a single discharged patient with exemplary linked workload units, is illustrated in Table II below, with this table providing the short name for an exemplary set of fields that are associated with a single patient, and a description of that field. Appendix A provides more detailed descnptions for certain of these and other exemplary fields associated with the discharge data set.
Figure imgf000008_0001
WO 00/77665 PCT/USOO/l 6032
Figure imgf000009_0001
While certain fields, such as age and sex will typically be on all discharge data sets, the contents of the fields in the discharge data sets will differ considerably depending on the reason the pattent was in the hospital, which will have caused the patient episode of care characteristics to differ from one patient to another. Other fields that may exist within a patient's discharge data set are provided in the table attached at Appendix A, with an exemplary data dictionary descnbmg these example fields.
The discharge data set may also contain or be linked to another data set, such as a data set that contains laboratory test results such as blood sugar, white count, or hematocnt Or a data set that contains pharmacy prescnption data such as medication, dose, and time of administration. Or a data WO 00/77665 PCT/USOO/l 6032
set that contains physiologic monitoring data such as blood pressure, respiratory rate or temperature, and other data sets that contain such similar data as is collected dunng a hospital episode of care
The discharge data set may also contain such information for multiple episodes of care for patients who were hospitalized more than once dunng the peπod covered by the data set and may be further linked to outpatient and ambulatory data, insurance data, and similar healthcare related data.
The entire population of patients discharged from a healthcare system or a hospital for a year or more is preferably used, such that the numbers of past patients are large enough to provide for adequate predictions of workloads for various types, or groups, of patients Cases may be added to the reference data set from another source if necessary For example, if a reference data set is prepared from a general hospital and the predictions are to be made in a hospital with a large neurosurgical population not reflected in the prepared reference data set, additional discharged patients from a source with a large representation of neurosurgical cases may be added to the reference data set. This allows the reference data set more closely correlate to the typical patient population in the facility for which staffing requires estimation In step 210, aggregations of patients that have similar charactenstics are made. These groupings may be by procedure, diagnosis, or other charactenstics. One grouping that is useful due to its content as well as widespread usage is that of a Diagnosis Related Group (DRG) vanant; HCFA- DRG, APR-DRG, R-DRG, APG or a similar resource-based group. The processor associated with one of the computers 110 implements these groupings by using established extraction parameters that cause association of a single discharged patient and the data corresponding thereto with a particular group. It should be noted, however, that a single discharge patient can be associated with multiple groupings if the aggregating categories are different. For simplicity in the discussion hereinafter, the parameters used to establish these groupings will be called "DRG" parameters in the remainder of this discussion, but they can actually be any selection cntena. Thus, for example, with reference to the discharge data set of Table II, the DRG field is used to assign the grouping. By identifying this DRG field within the discharge data set, this allows the computer 110 to select from this discharge data set all of the cases to be associated with the DRG grouping. So, for example, all patients who are assigned to DRG 127, "Heart Failure and Shock ' may be considered as a single group to develop a model for workload prediction for patients who are admitted with heart failure and shock. Thereafter, m step 212, using the groupings that have been established, the computer 110 operates on all of the discharged patient cases, and identifies all of the appropriate groupings for those cases. This can be implemented m many ways, such as by adding group fields to the already existing data for a particular discharged patients discharge data set or extracting all patients m a group into a separate data file. No matter how implemented, there becomes established a number of different groups, each distinguished from the other by the parameters, such as the DRG parameters mentioned above. WO 00/77665 PCT/USOO/l 6032
10
Thereafter, in step 214, for each group, the operator or computer 110 selects a set of candidate predictive factors. A person knowledgeable about the group being considered preferably performs or monitors this selection process, or evaluates the results While for most groups these factors will include sex and age, among others, any measured patient charactenstic may be a predictive factor, including insurance payer, zip code, diagnosis, admission temperature, and others Other than age and sex, one of the most available and powerful predictive factors, pnmaπly due to is current widespread use as a diagnostic tool, are ICD-9-CM diagnoses. Other diagnostic categones may also be used, but standardized definitions are desirable. Any number of candidate predictive factors for a group can be used, there typically being on the order of tens or hundreds of such candidate factors.
As part of the selection of candidate predictive, step 216 follows in which certain of the candidate predictive factors may be clustered together to form a clustered candidate predictive factor. This operation is such that the existence of any one of the clustered candidate predictive factors will result m an indication of the presence of the clustered candidate predictive factor. For instance, ICD- 9-CM codes 412 and V45.82 are respectively, "history of an acute myocardial infarction" and "history of a percutaneous translummal coronary angiography " While these codes may be used as separate candidate predictive factors, they may also be clustered into a clustered candidate predictive factor, such as C1001 for instance. The clusters may be based on clinical cπtena, as here, or may be based on any other cnteπa of interest.
An example of clustering of candidate predictive factors for a group is shown in Table III below
TABLE III code title cluster
4011 BENIGN HYPERTENSION
4019 HYPERTENSION NOS CL0003
402 HYPERTENSIVE HEART DIS CL0011
4020 MAL HYPERTENSIVE HRT DIS CL0011
40200 MAL HYPERTEN HRT DIS NOS CL0011
40201 MAL HYPERT HRT DIS W CHF CL0011
4021 BENIGN HYPERTEN HRT DIS CL0011
40210 BEN HYPERTEN HRT DIS NOS CL0011
40211 BENIGN HYP HRT DIS W CHF CL0011
4029 HYPERTENSIVE HRT DIS NOS CL0011
40290 HYPERTENSIVE HRT DIS NOS CL0011
40291 HYPERTEN HEART DIS W CHF CL0011
403 HYPERTENSIVE RENAL DIS
4030 MAL HYPERTENS RENAL DIS
40300 HTN MALIG RENAL DIS
40301 MAL HTN RENAL W/ FAILURE CL014F
4031 BENIGN HYPERT RENAL DIS
40310 BEN HTN REN DIS W/O FAIL
40311 BEN HTN REN DIS W/O FAIL CL014F 4039 HYPERTENS RENAL DIS NOS 11
40390 UNSPC HTN RENAL W/O FAIL
40391 UNSPC HTN RENAL W FAILUR CL014F 404 HYPERTEN HEART/RENAL DIS
4040 MAL HYPERT HRT RENAL DIS
40400 HTN HRT/REN W/O CHF/FAIL
40401 HTN HEART/REN W/CHF, MAL
40402 MAL HTN HRT/REN W/FAILUR CL014F
40403 HTN HRT REN W/CHF,FAILUR CL014F
4041 BEN HYPERT HRT/RENAL DIS
40410 BEN HTN HRT/REN W/O FAIL
4041 1 BEN HTN HRT REN W/CHF
40412 BEN HTN HRT REN W/FAILUR CL014F
40413 BEN HRT REN W/CHF.F AILUR CL014F 4049 HYPERT HRT/RENAL DIS NOS 40490 UNSPC HTN HRT REN DIS
As illustrated, clinical cntena corresponding to ICD-9-CM codes 402, 4020. 40200, 40201, 4021, 40210, 40211, 4029, 40290, and 40291 are candidate predictive factors that have been clustered together to form a clustered candidate predictive factor, labeled CL0011.
Step 218 follows thereafter, which begins the process of selecting those candidate predictive factors for the group that will be used as actual predictive factors. Since there are thousands of ICD- 9-CM codes, many codes that can be clustered together, and numerous other possible patient charactenstics, this can lead to there being an extremely large group of candidate diagnosis-related predictive factors. The selection of predictive factors from these candidates may be accomplished either automatically by the computer or interactively by the operator The automatic process is similar to the interactive process that is described immediately hereafter. To initiate the process of selecting the actual predictive factors, the presence or absence of each candidate predictive factor, inclusive of the clustered candidate predictive factors, for each case in the discharge data set is summanzed. Specifically, a spreadsheet or other listing is generated by the computer 110 that provides, for each candidate predictive factor, such as individual diagnoses, or collectively diagnoses that make up a cluster, such as CL0011, the number of patients who displayed that candidate predictive factor (count), the length of stay (los) and the average workload units of each type consumed by the episodes of care that had that factor associated with them (workload units). An example of the generated predictive factor summary is illustrated m Table 4 below
TABLE 4 diagnosis title count los Workload cluster
Units
41071 SUBENDO AMI/ 1ST EPISODE 396 6 28670
41401 AMI, FIRST EPISODE, NOS 773 6 32796
2720 PURE 161 5 27470
HYPERCHOLESTEROLEM 12
496 CHR AIRWAY OBSTRUCT NEC 7 788 6 31748 CL0017
41001 AMI A/L WALL/1 ST EPISODE 6 666 6 32227
4281 LEFT HEART FAILURE 9 9 10 46131
78551 CARDIOGENIC SHOCK 5 522 6 45371
42741 VENTRICULAR FIBRILLATION 4 422 6 40744
5990 URIN TRACT INFECTION NOS 4 488 8 32272
0414 E. COLI INFECT NOS 2 233 7 29506
40291 HYPERTEN HEART DIS W CHF 1 177 7 28794 CL0011
4439 PERIPH VASCULAR DIS NOS 1 122 7 34236 CL0011
41400 AMI NOS 5 522 7 30591
41041 AMI INF WALL/ 1ST EPISODE 2 288-4 5 33062
4011 BENIGN HYPERTENSION 9 9 4 21271
4110 POST MI SYNDROME 1 100 6 28500
4254 PRIM CARDIOMYOPATHY 1 144 6 29185 NEC
42731 ATRIAL FIBRILLATION 1 1551 7 39949
4589 HYPOTENSION NOS 7 711 6 32617
9981 HEMORR COMPLIC 3 300 6 35651 PROCEDURE
2851 AC POSTHEMORRHAG 67 10 63795
ANEMIA
4275 CARDIAC ARREST 50 5 34299
4271 PAROX VENTRIC 97 6 38100 TACHYCARD
41402 83 7 39369
49390 ASTHMA W/O STATUS ASTHM 12 7 39748 CL0002
The operator can then examine the list and interactively select the candidate predictive factors to examine, up to the number available Alternatively, the candidate predictive factors can be automatically selected based upon preset critena such as number of occurrences, weighted values such as (occurrences * workumts) to emphasize factors that have a large overall effect, or upon factors such as (average workunιts-workunιts)2to emphasize extreme values.
Additional predictive factors may be added automatically or based upon manual selection. Some of these factors may be selected based upon the clinical status of the patient. For example, a pattent who is assigned to a group of patients who had an infection may provoke selection of predictive factors to include time from presentation to antibiotic administration, use of intravenous antibiotic, presentation white blood cell count , and infectious organism identified. A patient who is admitted with a myocardial infarction my have time to thrombolytic therapy, peak blood creattne kmase MB band level (a laboratory test), and chronic coumadin medication as candidate predictive factors that are selected. And an obstetnc patient may have history of prenatal care, chronic anti-epileptic drug therapy, and admission hematocnt as candidate predictive factors.
The number of predictive factors must be large enough to allow an accurate prediction, but small enough to allow generalization of the model to data other than the training data set. A rule of thumb is to start with a number of predictive factors that is no greater than one eighth the number of training WO 00/77665 PCT/USOO/l 6032
13 set cases. Then the number can be automatically adjusted by the implementation or manually adjusted by the operator to give an accurate and general model This process is discussed below. Once completed, those predictive factors that are deemed most pertinent to the group being considered will be selected, and thereby become the actual predictive factors. Step 220 follows thereafter, in which a group input table is constructed by the computer 110 based upon the actual predictive factors, the reference data set for each case in the group and the associated workload value(s) for each case. Thus, usmg this information, the group input table is constructed and indicates whether, for each case, the actual predictive factors exist. Each actual predictive factor is coded as a binary "1" if the patient under study displays that factor, and is coded as a binary "0" if that factor is not displayed. Continuous factors, such as age or number of previous admissions, are calculated for each case as appropnate. With respect to continuous factors, it may be determined that the full range of the factor is not necessary or desirable for prediction. For example, with respect to age, it has been determined that ages under a certain minimum, and over a certain maximum, do not always provide any further predictive value for purposes of determining staffing. Accordingly, ages over this maximum, typically 85 years, and under the minimum, such as 45 years, for general medical/surgical patients, are entered as the maximum and minimum These age limits may vary for each group and workload predicted. An example group-input table is illustrated m Table 5, which table includes sex, tnmmed age, and six other predictive factors, as well as the total workload units associated with each case in the group.
TABLE 5 eocid num sex ageT fl G β f4 f5 f6 Workload Units
1995011590K9CTNKK 1 49 1 0 1 0 0 0 16867 199501150TCTAATUT 0 69 0 0 1 0 0 0 19437 19950115- 1 76 0 0 0 0 0 0 21711
1995011560054DOOG 0 68 0 0 1 0 0 0 13250 19950115PW9E99E4P 0 61 1 0 0 0 1 0 8311 19950115PPJRN2NVQ 0 80 1 0 1 0 0 0 15698 19950115ZZD344Z44 1 56 0 0 0 0 1 0 14534 199501157SSU78QWW 0 67 1 0 0 0 1 0 15068 19950115VFB3BVNBJ 0 71 0 0 1 0 0 1 13876 199501157SQQK7MUQ 1 61 1 1 1 0 0 0 14245 19950115W7V1SGS7K 1 44 0 1 0 0 1 0 20886 19950115EFF4EEKFF 0 66 1 1 0 0 1 0 9519 19950115JSQ50JJJQ 0 79 1 1 1 0 0 0 19724 19950115SL18X6V8X 0 47 1 0 0 0 1 0 8695 19950115ZZEXE4URR 0 70 0 0 1 0 0 0 19934 19950115E7K08PM8Q 1 70 0 0 1 0 0 1 12908 19950115YBGHGBYHY 0 60 0 1 1 0 0 1 15938 19950115YL1WNNMM5 0 80 1 0 1 0 0 1 18784 19950115MTBATLA2M 1 80 1 0 0 0 1 0 19853 19950115GG8ESGG48 0 51 1 1 0 0 0 0 17017 19950115RURCIRUMM 0 49 1 1 0 0 0 0 13396 WO 00/77665 PCT/USOO/l 6032
14
19950115MM22MM3ZY 0 73 1 0 0 0 0 0 18836 19950115PWQWZZ03I 1 72 1 0 0 0 1 0 17186 19950115S02T1STFH 0 67 1 0 0 0 0 0 12437 19950115XRBCBBZZ9 0 74 1 1 1 0 0 0 9042
Step 222 follows in which the user selects the model type from a list of appropπate candidates. Types of models that exist, as is known, include a linear model type of the form.
Workload = a + b*Factorl+c*Factor2 +. +z*FactorN Or a nonlinear model of the form:
Workload = exp-(a + b*Factorl+c*Factor2 + .. +z*FactorN)
Or:
Workload = l/(l+exp-(a + b*Factorl+c*Factor2 +... +z*FactorN)) where Factor 1, Factor2...FactorN are the binary factors, either "0" or "1," corresponding to each actual predictive factor. Workload is the actual workload required in carmg for the patient and a, b, c.z are the coefficients obtained by the model as a result of operating upon the group-input table.
An exemplary result for three different linear models is provided below in Table 6, which illustrates the linear model developed for a model based on DRG 080, a model based on Procedure 8151, and a model based upon DRG 127. As illustrated, each of these has a corresponding model type, and a number of workload units will result when the data of a particular patient is applied to the model, which has the coefficients (a) through (g), m this instance, determined in the manner descnbed above.
TABLE 6 agg_type agg_ model_ constant terml_ terml_ term2 code type (a) code value code (b)
DRG 080 4 10.10755713 Sex 1.39123176 AgeT
Procedure 8151 2 809590.2806 Sex 486480.3032 AgeT
DRG 127 1099.4259 Sex 716.956156 AgeT
term2_ term3_ term3_ term4_ term4_ term5_ term5 value code value code value code value WO 00/77665 PCT/USOO/l 32
15
(c) (d) (e) (0
0.03149705 C1001 4.96443384 C1002 -5.98450878
53687091.2 RF211 729720.4751
-2.1674520 41401 111.90826 C1003 -986 841154 41071 3099 89923
term6_ term6_ code value (g)
PrevAdm 98.9884238
Of course, other model types, such as other mathematical models or artificial neural network (neural net) model types may be selected. While a neural net model type does not lend itself to a simple numenc notation, it is implemented for prediction m the same manner as the parametric model types. The operator may choose from a number of different model types in order to select a model that provides the best fit to the relationship between the predictive factors and the actual workload data. Alternatively, this selection may be made automatically by the implementation of the invention. This is accomplished by building each of the possible model types and selecting the one that provides the best predictions, using preset cntena, such as the highest correlation coefficient for the fit of the predicted to the actual workload or the smallest total model error. Appendix B provides additional information regarding model building.
Once the model type is chosen in step 222, then step 224 follows, in which the analytical tool corresponding to the model type is used to process the group-input table m order to fit the data to the model type and build the model for the group, or to determine that the type of model selected is mappropnate and to indicate that another model should be selected. The model building essentially performs an analysis of the group data to determine if the selected model type that can be used to describe the relationships between the known actual predictive factors for past patients that make up the group, and the workload associated with that group. This model building can be performed using a vanety of techniques. For the mathematical models curve fitting, methods such as a least-squares, simplex, Newton-Raphson or similar methods may be used. As is known, an objective function is defined and is minimized by these tools. In the preferred implementation, the objective function may be a function of the difference between the predicted and actual workload for all of the patients for whom the model is being developed. A common function that is used m this type of optimization is the sum of the squares of the differences between the actual and predicted workload values. The absolute value of the differences may also be used, as may other functions. If the former function is WO 00/77665 PCT/USOO/l 6032
16 used, the fit is the equivalent of a so-called least-squares fit of the model to the training data set Exemplary tools that may be used to implement these fitting algorithms include custom-coded software modules, an OLE server such as Microsoft Excel with its built in analytical tools, an OLE server with add-in analytical extensions, or analytical software engines such as SPSS, SAS, Mathematica, Statistica, or other stand-alone analytical modules The most common modules are either Excel or Excel with an analytical add in If the model is a neural net or other non-parametnc method, step 224 invokes the neural net trainer or other non-parametnc method and passes it the group-input table(Table 5 from step 220 to create the model[Correct] The training is as is known for neural nets, with one implementation of an artificial neural network with back-propagation training being equivalent to a steepest-gradient least-squares fit of the model to the training data The model is archived for subsequent use m the database 130, bemg stored in the format that corresponds to the model type, such as, for example, a database format for a parametric model or in a model library format for non-parametnc and neural net models
As is known, the model may be validated against a test discharge data set to establish its generality and accuracy. In this process, the model is used to predict the workload for discharge data sets of cases of the appropriate type, for which the actual workload is known These are called the test or validation data sets. The accuracy of the model is expressed as the aggregated difference between the predicted and the actual workloads in the validation sets, and if this value is acceptably small, the model is accepted and is used for actual predictions. If the model is inaccurate, a new model type is built or additional, fewer, or different predictive factors are selected and the same model is rebuilt until an acceptably accurate and general model is found
It should be noted that steps 222 and 224 can be combined, such that the program searches for different model types and attempts to fit the data, such that if one model type is not acceptable, the computer 110 automatically proceeds to the next model type until an accurate and general model is developed.
Further, steps 222 and 224 may be combined to develop two or more models on a selection of patients that are initially assigned to a single group. In this instance, the group is subdivided into two or more groups based on either operator input, an automatic decision tree process performed by the implementation of the invention, or a combination of the two If the operator selects the subdivision, this operation is equivalent to building two separate models, as described above An exemplary situation in which the automatic decision tree process occurs is when the initial group includes both patients that have only been seen for one episode of care m the training discharge data set and also patients that have been seen multiple times If the model building process is unable to develop an acceptable model, the group is automatically subdivided into single- and multiple- episode patients, and separate models are developed for each group. An exemplary situation m which the combined manual and decision tree process occurs is when the initial model shows a very large sex effect on the WO 00/77665 PCT/USOO/l 6032
17 predicted workload. In this situation, the operator may be asked if the discharge data set should be subdivided into male and female patients, and separate models developed for each group The new models developed would frequently have different high occurrence diagnoses or procedures for the two groups, and would have better predictive accuracy than any single model Similar situations where multiple models are more accurate are easily envisioned Generally, a different model is developed for each different workload unit to be predicted for a patient in the group Thus, a model is developed for total workload, a different model is developed for nursing workload, for total laboratory workload, for microbiology workload, for microbiology technician workload, and so on, for each level of granulanty and workload category of interest. Further, a different model is developed for each time scale of interest, as descnbed below.
Once the model is obtained for each group, this model can then be used to practice another aspect of the invention, which is the actual estimation of workload based upon a present patient population. Or to practice yet another aspect of the invention, which is the seventy adjustment of and productivity calculation for a histoncal collection of patients. The former aspect of the invention is discussed first.
After the models have been built for all groups, the staffing evaluation system 100 can use these models to predict the staffing required to support a present patient population. This process will be descnbed with reference to Fig. 3.
Initially, patient data regarding the present patient population is obtained This patient data corresponds to the data that was previously discussed with respect to Fig 2 In step 310, this data is input into the prediction module of the present invention Whereas the data set used for model building contained all possible data that was accessible and might have predictive value, the present pattent data may comprise such possible data, or may be restricted to the data elements actually required to calculate the predictive factors m the accepted model. Thereafter follows step 312, in which the computer 110 uses the patient data set and determines the appropnate group in which to list each patient. Since the groupings have already been established, it is only necessary to assign the patient to an existing group. This assignment can be automatically performed by computer 110 by assigning a group based upon a predetermined decision tree or other decision structure, which operates on the patient data set, performed manually by the operator, or, with some combination in which the computer 110 uses a decision tree to determine the most likely groups for a given patient, and the operator then selects the most appropnate group Generally, this is an automatic process accomplished by the computer alone. With respect to the decision tree, each of the vanous patients are uniquely assigned to a DRG (See Table 2), so the patient will preferably be assigned to the group corresponding to that DRG. For some other groupings, such as diagnosis or procedure, a pattent might fall into two or more possible groupings In this case, the pattent may be automatically assigned to a group based on a decision tree, may be assigned manually by the operator, WO 00/77665 PCT/USOO/l 6032
18 or may be assigned by a combination of the two processes. Again, the most common implementation is for the assignment to be automatic, with the assignment made to the group that has the most accurate model. This automatic assignment is used very commonly when multiple models have been automatically built by the implementation, such as sex-specific models or number-of-episode models as described above, to improve the accuracy of the predictions. For combined manual and computer operation, those groups to which the patient could be assigned can be presented to the operator, and the group to which the operator judges is most appropnate can then be selected.
Once the appropriate group is selected in step 314, then the correct model corresponding to that group is looked up. As part of this step, the factors and weights are also retrieved from the archive if the model is mathematical, or the non-parametnc model is loaded in the library routine if not.
Having determined the model to use, step 316 follows, m which the patient data is used to determine the status of the actual predictive factors used for that group, and other data used by the model is obtained. Thus, for that patient, the presence or absence of diagnoses or other binary factors is used to determine the state ("0" or "1") of the binary predictive factors. Continuous predictive factors are calculated such as the age (with thresholds applied as discussed above if applicable the number of previous admissions, and similar required continuous factors are processed. And other predictive factors such as sex and zip code are processed, as required.
Thereafter follows step 318, in which the appropπate model is run to determine the estimated workload associated with that patient, for the workload category of interest. Sample patient data, actual predictive factors, and an example calculation for a linear model are illustrated in Tables 6A-6C below. Table 6A illustrates another set of linear models developed for DRG 080, Procedure 8151, and DRG 127. Table 6B illustrates a partial set of patient data for three different patients,
Table 6A agg_type agg_ model_ const terml_ terml_ term2_ code type code value code
DRG 080 4 10.10755713 Sex 1.39123176 AgeT
Procedure 8151 2 809590.2806 Sex 486480.3032 AgeT
DRG 127 2 731832.2817 Sex 492991.6661 AgeT WO 00/77665 PCT/USOO/l 6032
19 term2 term3 term3 term4 term4_ term5_ term5 value code value code value code value
0.03149705 C1001 4.96443384 C1002 -5.98450878
53687091.2 RF211 729720.4751
53687091.2 4280 563419.0014 C1003 101047.9659 CLOU 312330.0743
Table 6B dc_date pat_id age_ sex race zipcode payorl los sa_los drg yrs
19950215 B9GBLEE3L 69 0 1 95111 08 3 5.2 080 19950715 ROJOQOIJV 44 1 1 95062 07 3 5.1 083 19950215 570H0U3E3 57 1 1 95110 07 15 11.6 080
mdc dxl dx2 dx3 dx4 dx5 dx6 dx7 dx8 dx9 dxlO dxll dxI2 04 48289 53019 412 41401 41420
04 1363
04 1 124 3592 53019 53130 9351 41401 412 44020 V1582
procl proc2 proc3
4223 3893 9915
Note that clusters C1001 contains diagnoses 3451 and 5433; C1002 contains diagnoses 41400, 41401, 41402,...41499. To calculate the predicted workload units for patient B9GBLEE3L in Table 6B, it is apparent that the record shows a DRG of 080. The first row of the archived model Table 6A is for DRG 080, which is applicable, and the model type is 4, which is a linear model. As shown in Table 6C, the following terms are summed in the linear model: WO 00/77665 PCT/USOO/l 6032
20
Table 6C
Factor Value Coefficient
Constant 10.108 Sex 0 * 1.391 AgeT 69 * 0.0315 FClOOl 0 * 4 9644 FC1002 1 * -5.4895
Accordingly, the predicted workload units are 10.108 + 0 + 2.1735 + 0 - 5 4895 or 6.792 workload units.
The sequence of steps 310-318 is then repeated for eachpatient, to determine the estimated workload for a larger group, such as a ward or an entire hospital, and the workunit predictions are aggregated according to predetermined rules. This frequently amounts to simple summation, but patient care interaction factors may also be considered Steps 310-318 are also repeated for each patient and for each different category of workload that is to be estimated and similar aggregation occurs for each different workload category The interaction of the matrix elements of patient rows and workload category columns may demonstrate second order patient care interactions that are considered m the aggregations.
Although the workload can be estimated, there are many dynamic events that may require updating of the estimated workload, or updating of the model. Each process will now be discussed.
With respect to updating of the estimated workload, after a patient has had workload estimated, that patient's condition may change. As a result of this change, the patient's workload can be estimated again, m light of the changed condition. Accordingly, the previously estimated workload will be removed and the new estimate, based upon the new conditions, used instead.
With respect to updating of the model, all of the models have to be rebuilt periodically to reflect changing national or state (reference) medical practice and hospital work patterns. In addition, the predictions of the model may themselves be benchmarked against reference data sets to expand the goals that may be realized. Generally, this means that a model is built using very large scale data, such as national or statewide, and the model may be refined for local conditions, such as staffing levels, practice patterns, etc., by modifying the model based upon histoncal data for the hospital of interest, if available. This data may be long term to reflect baseline differences in the practices of the target hospital and the reference data set or may be short term data from the hospital to reflect slightly more modern practice patterns m the hospital than in the reference data set Typically, although not necessanly, this refinement is limited to altenng the predictive weights of the already determined predictive factors by no more than a preset amount, say 20% based upon the goals of the adjustment and operator knowledge of the likely magnitude of the effects This may be implemented m the invention by repeating step 224 using a training discharge data set that contains the required data 21 elements but that incorporates discharges from the hospital of interest over the time period of interest rather than the large scale discharge data. Step 224 is constrained to allow only small adjustments of the predictive weights, as is known in the field of constrained optimization The model may be further fine-tuned using contemporaneous data (i.e. yesterday's and last week's), with the added advantage that many of the patients that will receive care in the predicted period will be represented in the refinement data set. Of course, new models for new groups can be developed as well, and groups can be consolidated and split.
If the focus of the hospital is to move towards a benchmark level of workload, then present invention can be used to estimate the desired staffing level to care for the actual mix of patients rather than the workload that is expected to be required absent any management. For such a focus, the prediction may combine benchmark data and the local data m a relationship that reflects the management focus. This may result in a one-time adjustment or in a phased adjustment penod. For example 80% local data and 20% benchmark data may be used this quarter, 60% local data and 40% benchmark data next quarter, and so forth, until the benchmark data is used exclusively to predict the required workload. The latter stage is the equivalent of a finely granular targeted productivity based on objective norms. If the model used m this process is a parametric linear model, the coefficients of the model as estimated individually on the benchmark data and local data may be simply combined in the desired ratio to develop the new model. If the model is non-parametnc or non-linear, the objective function of the model-building step is modified to weight the contnbutions of the local data and benchmark data appropriately and a new model is constructed as descnbed above for fine-tuning an existing model.
Also, the present mvention can be integrated into either a special purpose reporting tool or a more general clinical outcomes reporting tool to facilitate user predictions of workforce requirements based upon current data or to allow user analysis of workforce efficiency as compared to histoncal or benchmark norms. The user analysis of workforce efficiency is based upon the seventy adjustment and productivity calculation features of this invention. If Steps 310 to 318 are applied to a histoncal population of patients rather than to a present population of patients, the expected workload that would have been required to take care of each pattent, at the level of granulanty of the model and the histoncal data, is obtained. This expected workload can be thought of as the workload to care for the pattent if he/she behaved like the average pattent with his/her predictive factors, and thus the expected workload has been seventy adjusted for the patient's predictive factors. The productivity is calculated as the ratio of the actual workload collected with the histoncal data and the expected workload. This productivity can be used, as is known m business practice, to evaluate the efficiency of the workforce caπng for the patients. In the discussion above, it was also assumed for simplification of understanding that each patient was assigned to only a single group, and, therefore, a single model. This, however, is an over- WO 00/77665 PCT/USOO/l 6032
22 simplification that must be noted In practice, there will typically need to be many different models associated with each patient The selection of the appropπate model is based upon the level of granulanty required and the accuracy of the available models Thus, while there may be a model based upon the entire stay of a patient, there may be another model based upon the first day of care, yet another model based upon second day of care, and so forth similarly, there can be different models based upon the type of staff, such that there is one model for nurses, another model for clencal staff, and another model for technicians Accordingly, the models chosen will depend in large part upon the level of granularity that has been selected and the models that have been built upon that granularity Of course, the finer the granularity, the more choices that the operator has available, or the greater detail an automatically generated report may have If different levels of granulanty exist, it is preferable to determine relationships between the different models, so that models which are not intended to be used for a given level of granulanty are not used For example, microbiology workload could be estimated in several ways, depending upon the granulanty of the training data available. It could be estimated as a fixed fraction of the total laboratory workload if that were the finest granulanty available, where the fraction is established by examining existing reference or facility workload data. It could be a direct model of the microbiology workload if that data were available in the training data set. Or it could be a sum of the workloads for microbiology clencal staff, technician staff, technologist staff, and all other microbiology staff. In the preferred implementation of the invention, a direct model of the desired workload category is always the preferred choice. The summation of component workload elements is the secondary choice, and the estimation of allocated workload from a less detailed model is the least desirable choice, with lower levels of detail being increasingly less desirable. Similar cntena are used to select the models for predicting workload for an entire length of stay, a single day, single shift, and so forth. The automatic selection of the appropnate model based on these cntena may be embodied in the implementation of the invention. Also, for parametπc models, there will generally be a new model for each level of granulanty. So a model may be required for first-day first-shift nursing, first-day first-shift clencal, and so forth. The appropnate model selected will then be based upon the input specified by the user, such that if estimates related to first-day first shift nursing are desired, then that model is chosen
While the present invention has been descnbed herein with reference to particular embodiments thereof, a latitude of modification, vanous changes and substitutions are intended in the foregoing disclosure, and it will be appreciated that in some instances some features of the invention will be employed without a corresponding use of other features without departing from the spmt and scope of the mvention as set forth in the appended claims. WO 00/77665 PCT/USOO/l 6032
23
APPENDIX A
Data Element Name: PATIENT WORKLOAD UNITS IN CATEGORY N Indicators Using: All patient episode of care records. Definition: The units of work required for the patient in category n between the admission date and the discharge date.
Short Name: WORK Cn Format:
Length = 10.2
Type = Numeric
Occurs = 10
Allowable Values: Any number greater than or equal to zero and less than or equal to 9999999.99 with two decimal points.
Missing Data Logic: Field is Null.
***
Data Element Name: PATIENT TOTAL WORKLOAD UNITS Indicators Using: All patient episode of care records. Definition: The total units of work required for the patient between the admission date and the discharge date.
Short Name: WORK CT Format:
Length = 10.2
Type = Numeric
Occurs = 1
Allowable Values: Any number greater than or equal to zero and less than or equal to 9999999.99 with two decimal points.
Missing Data Logic: Field is Null.
*** WO 00/77665 PCT/USOO/l 6032
*?
Data Element Name: ADMISSION DATE Indicators Using: All patient episode of care records. Definition: The date the patient was admitted to the health care organization for inpatient or outpatient service.
Short Name: ADM_DATE Format: YYYYMMDD Length = 8
Type = Character
Occurs = 1
Allowable Values: MM = Month (01-12)
DD = Day (01-31)
YYYY = Year (e.g.: 1991)
Valid date must be entered (e.g.: 02/30/1993 is not valid)
Missing Data Logic: Field is considered missing if DC_DATE or LOS is missing, as it is calculated from them.
***
Data Element Name: DISCHARGE DATE Indicators Using: All patient episode of care records. Definition: The month, day, year the patient was discharged from the health care organization as an inpatient or outpatient. The month and year of this data element are used to include patients into groups for comparative analysis.
Short Name: DC DATE
Format: YYYYMMDD Length = 8
Type = Character
Occurs = 1
Allowable Values: MM = Month (01-12)
DD = Day (01-31)
YYYY = Year (e.g.: 1991)
Valid date must be entered (e.g.: 02/30/93 is not valid).
Missing Data Logic: Field is blank.
*** WO 00/77665 PCT/USOO/l 6032
2
Data Element Name: ICD-9-CM PROCEDURE CODES Indicators Using: All patient episode of care records, when applicable. Definition: The International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), (Volume 3) is the classification system used to assign the ICD-9-CM Procedure Codes for this hospitalization.
Short Name: PR( DC««
Format:
Length = 4
Type = Character
Occurs = 8
Allowable Values: Any valid ICD-9-CM Code. Missing Data Logic: None Specified.
Data Element Name: MEDICAL RECORD NUMBER (EOC) Indicators Using: All patient episode of care records. Definition: A health care organization provided number used to identify a specific episode of care.
Short Name: PAT ID Format:
Length = 12 Type = Character Occurs = 1
Allowable Values: Any.
Missing Data Logic: Field is blank.
***
WO 00/77665 PCT/USOO/l 6032
3L <-
Data Element Name: PATIENT DATE OF BIRTH Indicators Using: All patient episode of care records. Definition: The month, day, year the patient was born. Short Name: DOB Format: YYYYMMDD Length = 8
Type = Character
Occurs = 1
Allowable Values: MM = Month (01-12)
DD = Day (01-31)
YYYY = Year (e.g.: 1991)
Valid date must be entered (e.g.: 02/30/93 is not valid).
Missing Data Logic: Field is blank.
***
Data Element Name: PATIENT AGE IN YEARS Indicators Using: All patient episode of care records. Definition: The number of years since the patient's birth, rounded down. Short Name: AGE YRS Format:
Length = 3
Type = Numeric
Occurs = 1
Allowable Values: Zero to 130.
Missing Data Logic: Field is Null.
WO 00/77665 PCT/USOO/l 6032
11
Data Element Name: PATIENT AGE IN MONTHS Indicators Using: All patient episode of care records. Definition: The number of months since the patient's birth, rounded down. Short Name: AGE MOS Format:
Length = 2
Type = Numeric
Occurs = 1
Allowable Values: Zero to eleven. Missing Data Logic: Field is Null and AGE_YRS equals zero.
***
Data Element Name: PATIENT AGE IN DAYS Indicators Using: All patient episode of care records. Definition: The number of days since the patient's birth, rounded down. Short Name: AGE DAYS Format:
Length = 3
Type = Numeric
Occurs = 1
Allowable Values: Zero to thirty-one. Missing Data Logic: Field is Null and AGE_YRS equals zero. ***
WO 00/77665 PCT/USOO/l 6032
Z1
Data Element Name: SEX Indicators Using: All patient episode of care records. Definition: The sex of the patient as recorded at date of admission, outpatient services, or start of care.
Short Name: SEX Format:
Length = 1
Type = Character
Occurs = 1
Allowable Values: 1 = Male
2 = Female
3 = Other
4 = Unknown
Missing Data Logic: Field is blank, or value does not represent distinguishable gender.
***
Data Element Name: PATIENT RACE Indicators Using: All patient episode of care records. Definition: The category that describes the patient's race. Short Name: RACE Format:
Length = 1
Type = Character
Occurs = 1
Allowable Values: 1 = White
2 = Black
3 = Native American/ Eskimo/ Aleut
4 = Asian Pacific Islander
5 = Other
6 = Unknown
Missing Data Logic: Field is blank, or value does not represent distinguishable race.
WO 00/77665 PCT/USOO/l 6032
Z1
Data Element Name: PATIENT POSTAL CODE (ZIPCODE) Indicators Using: All patient episode of care records. Definition: Postal Code of the patient's primary residence, or mailing address.
Short Name: ZIPCODE Format:
Length = 5
Type = Character
Occurs = 1
Allowable Values: Any valid U.S. Zip Code. Missing Data Logic: Field is blank.
***
Data Element Name: PATIENT COUNTY Indicators Using: All patient episode of care records. Definition: The category that describes the patient's county of residence. Short Name: COUNTY Format:
Length = 5
Type = Character
Occurs = 1
Allowable Values: Refer to Dynittls.dbf for an up to date list. Missing Data Logic: Field is blank.
***
3ώ
Data Element Name: HOSPITAL IDENTIFICATION NUMBER Indicators Using: All patient episode of care records. Definition: A uniquely assigned identification number, used to reference health care organization level information. (CHW hospitals use the OSHPD Hospital ID.)
Short Name: HOSP ID Format:
Length = 6
Type = Character
Occurs = 1
Allowable Values: Any combination of letters and/or numbers that represents a single existing facility.
Missing Data Logic: Field is blank.
#**
Data Element Name: PAYOR Indicators Using: All patient episode of care records. Definition: The category that describes the payor. Short Name: PAYOR/-H Format:
Length = 2
Type = Character
Occurs = 2
Allowable Values: 01 = Medicare
02 = Medicaid
03 = Worker's Compensation
04 = County Indigent Program
05 = CHAMPUS/ CHAMPVA VA
06 = Other Government
07 = Commercial HMO
08 = Commercial PPO
09 = Private Insurance
10 = Blue Cross/ Blue Shield 11 = Self Pay
12 = Charity Care
13 = No Charge
14 = Other Non Governmental
15 = Medicare HMO
16 = Medicaid HMO
Missing Data Logic: Field is blank. WO 00/77665 PCT/USOO/l 6032
3/
Data Element Name: TYPE OF ADMISSION
Indicators Using: All patient episode of care records.
Definition: The category that describes the patient's admission status.
Short Name: ADM_TYPE
Format:
Length = 1
Type = Character
Occurs = 1
Allowable Values:
1 = Scheduled at least 24 hours before surgery.
2 = Unscheduled.
3 = Infant, under 24 hours old.
4 = Unknown
Missing Data Logic: DC DATE or ADM DATE is not blank, and field is blank.
***
Data Element Name: HOSPITAL SERVICE Indicators Using: All patient episode of care records. Definition: Abbreviation of the service provided to the patient by the health care organization.
Short Name: HOSP SRVC Format:
Length = 3
Type = Character
Occurs = 1
Allowable Values: Any valid 3-digit abbreviation. Missing Data Logic: Field is blank.
***
WO 00/77665 PCT/USOO/l 6032
3 -
Data Element Name: ICD-9-CM DIAGNOSIS CODES Indicators Using: All patient episode of care records. Definition: The International Classification of Diseases, Ninth Revision,
Clinical Modification (ICD-9-CM), (Volume 1 and 2) is the classification system used to assign diagnosis codes for this hospitalization.
Name: DXJ nn at:
Length = 5
Type = Character
Occurs = 12
Allowable Values: Any valid ICD-9-CM code.
Missing Data Logic: None specified.
Data Element Name: ADMISSION DIAGNOSIS Indicators Using: All patient episode of care records. Definition: The international Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), (Volume 1 and 2) is the classification system used to assign the admission diagnosis for this hospitalization
Short Name: DX _ADMIT
Format:
Length = 5
Type = Character
Occurs = 1
Allowable Values: Any valid ICD-9-CM code. Missing Data Logic: Field is blank.
WO 00/77665 PCT/USOO/l 6032
33
Data Element Name: CONSULTING DOCTOR Indicators Using: All patient episode of care records. Definition: Five digit health care organization assigned code which represents the doctor who consulted on this episode of care.
Field is encrypted.
Short Name: MD CONS/w
Format:
Length = 5
Type = Character
Occurs = 5
Allowable Values: Any valid, encrypted doctor code. Missing Data Logic: None specified.
***
Data Element Name: PROCEDURE DOCTOR Indicators Using: All patient episode of care records. Definition: Five digit, health care organization assigned code which represents the doctor who performed the corresponding ICD-9-
CM procedure during this episode of care.
Field is encrypted.
Short Name: MD PROC/zn Format:
Length = 5 Type Character
Occurs = 8
Allowable Values: Any valid, encrypted doctor code. Missing Data Logic: Field is blank, and corresponding ICD-9-CM procedure field is not blank.
WO 00/77665 PCT/USOO/l 6032
JV
Data Element Name: ATTENDING DOCTOR Indicators Using: All patient episode of care records. Definition: Five digit, health care organization assigned code which represents the patient's attending doctor for this episode of care.
Field is encrypted.
Short Name: MD ATTEND
Format:
Length = 5
Type = Character
Occurs = 1
Allowable Values: Any valid, encrypted doctor code. Missing Data Logic: Field is blank.
***
Data Element Name: HCFA DIAGNOSIS RELATED GROUP Indicators Using: All patient episode of care records. Definition: The HCFA classification for the admission diagnosis of this hospitalization.
Short Name: DRG Format:
Length = 3
Type = Character
Occurs = 1
Allowable Values: Any valid HCFA DRG Code. (Usually between 1 and 500.) Missing Data Logic: Field is blank.
WO 00/77665 PCT/USOO/l 6032 ss-
Data Element Name: MAJOR DIAGNOSTIC CATEGORY Indicators Using: All patient episode of care records. Definition: The Major Diagnostic Category assignment for this hospitalization.
Short Name: MDC Format:
Length = 2
Type = Character
Occurs = 1
Allowable Values: Any valid MDC Code. (Usually between 1 and 25, or 99.)
Missing Data Logic: Field is blank.
***
Data Element Name: PATIENT LENGTH OF STAY Indicators Using: All patient episode of care records. Definition: The number of days between the admission date and the discharge date.
Short Name: LOS Format:
Length = 4
Type = Numeric
Occurs = 1
Allowable Values: Any whole number greater than or equal to zero. Missing Data Logic: Field is Null.
***
Data Element Name: SEVERITY ADJUSTED LENGTH OF STAY Indicators Using: All indicators measuring risk adjusted values. Definition: The patient's Length of Stay modified by the patient's severity score and APR-DRG assignment, as determined by the 3M APR-DRG grouper commercial software that that assigns the APR-DRG based on the other data in the database.
Short Name: SA_ LOS
Format:
Length = 4
Type = Numeric
Occurs = 1
Allowable Values: Any whole number greater than or equal to zero. Missing Data Logic: Field is considered missing if any of LOS, DRG APRDRG, or SEVERITY is missing. WO 00/77665 PCT/USOO/l 6032
3<-
Data Element Name: DAYS IN INTENSIVE CARE UNIT Indicators Using: All patient episode of care records. Definition: The number of days this patient spent in the Intensive Care
Unit during this episode of care.
Short Name: ICU DAYS Format:
Length = 3
Type = Numeric
Occurs = 1
Allowable Values: Any whole number between 0 and 999, that is less than or equal to the LOS.
Missing Data Logic: Field is Null.
***
Data Element Name: DAYS IN CRITICAL CARE UNIT Indicators Using: All patient episode of care records. Definition: The number of days the patient spent in the Critical Care Unit during this episode of care.
Short Name: CCU DAYS Format:
Length = 3
Type = Numeric
Occurs = 1
Allowable Values: Any whole number between 0 and 999, that is less than or equal to the LOS.
Missing Data Logic: Field is Null.
WO 00/77665 PCT/USOO/l 6032
51
Data Element Name: DAYS IN TELEMETRY CARE UNIT Indicators Using: All patient episode of care records. Definition: The number of days the patient spent in the Telemetry Care
Unit during this episode of care.
Short Name: TCU DAYS Format:
Length = 3
Type = Numeric
Occurs = 1
Allowable Values: Any whole number between 0 and 999, that is less than or equal to the LOS.
Missing Data Logic: Field is Null.
#**
Data Element Name: ADMITTED FROM Indicators Using: All patient episode of care records. Definition: The category that describes the location from which the patient was admitted to this health care organization.
Short Name: ADM FROM Format:
Length = 2
Type = Character
Occurs = 1
Allowable Values:
1 = Home
2 = Residential Care Facility
3 = Ambulatory Surgery
4 = Long Term Care
5 = Acute Inpatient Hospital
6 = Other Inpatient Hospital
7 = Newborn
8 = Prison/Jail
9 = Other
Missing Data Logic: Field is blank, or field represents a non-determinable value. *** WO 00/77665 PCT/USOO/l 6032
3S
Data Element Name: DISCHARGED TO Indicators Using: All patient episode of care records. Definition: The category that describes the location that the patient was discharged to when leaving this health care organization.
Short Name: DC TO Format:
Length = 2
Type = Character
Occurs = 1
Allowable Values: Any valid DC_DISPO code that represents a discharge to a location.
Missing Data Logic: Field is blank.
***
Data Element Name: AUTOPSY PERFORMED
Indicators Using: All patient episode of care records. Definition: Describes whether an autopsy was performed during this episode of care.
Short Name: AUTOPSY
Format: Y/N/U Length = 1 Type Character Occurs = 1
Allowable Values:
Y = Yes
N = No
U = Unknown
Missing Data Logic: Field is blank, and DC_DISPO indicates death of patient.
WO 00/77665 PCT/USOO/l 6032
3?
Data Element Name: TISSUE/ORGAN DONATION Indicators Using: All patient episode of care records. Definition: Describes whether the patient donated any tissue or organ(s). Short Name: DONOR Format: Y/N/U
Length = 1
Type = Character
Occurs = 1
Allowable Values:
Y = Yes
N = No
U = Unknown
Missing Data Logic: Field is blank.
***
Data Element Name: BIOPSY RESULTS Indicators Using: All patient episode of care records. Definition: The category that describes the results of a biopsy performed during this episode of care.
Short Name: BIOPSY Format: Left padded to two characters with zeroes. Length = 2
Type = Character
Occurs = 1
Allowable Values:
00 = Normal Tissue
10 = Abnormal tissue, not otherwise specified
11 = Abnormal tissue, inflammation
12 = Abnormal tissue, benign neoplasm
13 = Abnormal tissue, malignant neoplasm
98 = Biopsy not performed
99 = Information not available
Missing Data Logic: Field is blank or evaluates to "99" WO 00/77665 PCT/USOO/l 6032
¥ό
Data Element Name: DISCHARGE DISPOSITION Indicators Using: All patient episode of care records. Definition: The code indicating patient status as of the ending service date of the period covered by this episode of care.
Short Name: DC_DISPO Format: Left padded to two characters with zeroes. Length = 2
Type = Character
Occurs = 1
Allowable Values:
01 = Routine (Home)
02 = Acute care within this hospital
03 = Other care within this hospital
04 = Long term care within this hospital
05 = Acute care at another hospital
06 = Other care at another hospital
07 = Long term care at another hospital
08 = Residential Care Facility
09 = Prison or Jail
10 = Against Medical Advice 11 = Died
12 = Home health service
13 = Other
Missing Data Logic: Field is blank, or field evaluates to a non-determinable value.
Data Element Name: NOSOCOMIAL
Indicators Using: Nosocomial infection rate.
Definition: Nosocomial Infection identified during admission.
Short Name: NOSOCOMIAL
Format:
Length = 2
Type = Character
Occurs = 1
Allowable Values: Y/N
Missing Data Logic: Field is blank or outside the range of allowable values. *** 4/
Data Element Name: BIRTH WEIGHT IN GRAMS Indicators Using: All patient episode of care records. Definition: The patient's weight at birth, in grams. Short Name: BIRTH WT Format:
Length = 4
Type = Numeric
Occurs = 1
Allowable Values: Any whole number between 0 and 9,999.
Missing Data Logic: Field is blank and is categorized as neonatal.
***
Data Element Name: APGAR SCORE Indicators Using: All patient episode of care records.
Definition: APGAR is the Activity, Pulse, Grimace, Appearance, and Respiration sum score given to describe neonatal condition.
Short Name: APGAR ΛMMIN Format: 0-10 Length = 2 Type = Numeric Occurs = 2
Allowable Values:
Figure imgf000042_0001
Missing Data Logic: Field is blank and DX_ADMIT denotes birth. *** WO 00/77665 PCT/USOO/l 6032
V
Data Element Name: PATIENT COMPLAINTS Indicators Using: All patient episode of care records. Definition: The category that describes a complaint the patient may have had during this hospitalization.
Short Name: COMPLAINTS Format:
Length = 2
Type = Character
Occurs = 1
Allowable Values: Y N. Missing Data Logic: Field is blank or outside the range of allowable values.
Data Element Name: INCIDENT Indicators Using: All patient episode of care records. Definition: The category that describes any incident that may have occurred during this hospitalization.
Short Name: INCIDENT Format:
Length = 2
Type Character
Occurs = 1
Allowable Values: Unknown.
Missing Data Logic: None specified.
***
WO 00/77665 PCT/USOO/l 6032
*>
Data Element Name: RISK MANAGEMENT Indicators Using: All patient episode of care records. Definition: The category that describes the level of Risk Management applied during this episode of care.
Short Name: RISK MAN Format:
Length = 2
Type = Character
Occurs = 1
Allowable Values: Y/N Missing Data Logic: Field is blank or outside the range of allowable values.
Data Element Name: UTILIZATION REVIEW Indicators Using: All patient episode of care records. Definition: The category that describes whether a utilization review took place for this episode of care.
Short Name: UTIL REV Format:
Length = 2
Type = Character
Occurs = 1
Allowable Values: Y/N
Missing Data Logic: Field is blank or outside the range of allowable values. ***
WO 00/77665 PCT/USOO/l 6032 t*
Data Element Name: MED_HX
Indicators Using: None.
Definition: Significant medical history present.
Short Name: MEDJHX
Format:
Length = 2
Type = Character
Occurs = 1
Allowable Values: Y/N Missing Data Logic: Field is blank or outside the range of allowable values.
***
Data Element Name: QUALITY ASSURANCE DEPARTMENT REVIEW Indicators Using: All patient episode of care records. Definition: The category that describes whether the episode of care was reviewed by this health care organization's quality assurance department.
Short Name: QA_REVIEW Format:
Length = 2
Type = Character
Occurs = 1
Allowable Values: Y/N
Missing Data Logic: Field is blank or outside the range of allowable values.
WO 00/77665 PCT USOO/l 6032
Vό'
Data Element Name: CASE MANAGER Indicators Using: All patient episode of care records. Definition: The ID number of the case manager responsible for this episode of care.
Short Name: CASE MANGR Format:
Length = 3
Type = Character
Occurs = 1
Allowable Values: Any 1-3 digit alphanumeric code. Missing Data Logic: None specified.
***
Data Element Name: CODER Indicators Using: All patient episode of care records. Definition: The ID number of the initial coder of this episode of care. Short Name: CODER Format:
Length = 3
Type Character
Occurs = 1
Allowable Values: Any 1-3 digit alphanumeric code. Missing Data Logic: None specified.
WO 00/77665 PCT/USOO/l 6032
4
Data Element Name: TOTAL CHARGES TO THE PATIENT Indicators Using: All patient episode of care records. Definition: The total number of dollars billed for this episode of care. Short Name: CHARGE TOT Format:
Length = 8
Type = Numeric
Occurs = 1
Allowable Values: 0-99,999,999 Missing Data Logic: Field is Null.
***
Data Element Name: SEVERITY ADJUSTED TOTAL CHARGES Indicators Using: All patient episode of care records. Definition: The total of dollars billed for this episode of care, adjusted to account for severity.
Short Name: SA CHARGE Format:
Length = 8
Type = Numeric
Occurs = 1
Allowable Values: 0-99,999,999 Missing Data Logic: Field is Null.
***
WO 00/77665 41 PCT/USOO/l 6032
Data Element Name: TOTAL COST Indicators Using: All patient episode of care records. Definition: The total amount of costs incurred by this health care organization in regards to this episode of care.
Short Name: COST TOT Format:
Length = 8
Type Numeric
Occurs = l
Allowable Values: 0-99,999,999 Missing Data Logic: Field is Null.
***
Data Element Name: SEVERITY ADJUSTED TOTAL COST Indicators Using: All patient episode of care records. Definition: The total amount of costs incurred by this health care organization in regards to this episode of care, adjusted to account for severity.
Short Name: SA COST Format:
Length = 8
Type = Numeric
Occurs = l
Allowable Values: 0-99,999,999 Missing Data Logic: Field is Null.
***
WO 00/77665 PCT/USOO/l it
Data Element Name: REVENUE
Indicators Using: All patient episoc
Definition: The total revenue attrib
Short Name: REVENUE
Format:
Length = 8
Type = Numeric
Occurs = 1
Allowable Values: 0-99,999,999
Missing Data Logic: Field is Null.
Data Element Name: ANCILLARY COST BUCKETS Indicators Using: All patient episode of care records. Definition: Ancillary cost buckets each contain a single monetary value which represents a portion the total costs incurred by the health care organization during this episode of care. The current breakdown is as follows:
Figure imgf000049_0001
Short Name: ANC COSTHM
Format:
Length = 6
Type = Numeric
Occurs = 10
Allowable Values: Any whole number greater or equal to zero and less than the total direct cost (COSTJDIR) of this episode of care.
Missing Data Logic: Field is Null or contains data outside the valid range. WO 00/77665 PCT/USOO/l 6032
If
Data Element Name: ANCILLARY COST DESCRIPTION Indicators Using: All patient episode of care records. Definition: A four-digit code that describes the category that describes the corresponding ancillary cost bucket.
Short Name: ANC ACnn Format:
Length = 4
Type = Numeric
Occurs = 10
Allowable Values:
Figure imgf000050_0001
Missing Data Logic: Field is Null and corresponding Ancillary Cost element (ANC_COST««) is NOT Null, or, Field does not represent a determinable ancillary description category.
***
Data Element Name: TOTAL DIRECT COST Indicators Using: All patient episode of care records. Definition: The monetary value that describes the total direct costs incurred by this health care organization during this episode of care.
Short Name: COST DΓR Format:
Length = 8
Type = Numeric
Occurs = 1
Allowable Values: Any whole number that is greater than or equal to zero, and less than the Total Costs (COST_TOT) of this episode of care.
Missing Data Logic: Field is Null and COST TOT is NOT Null, or Field is outside the range of allowable values. *** WO 00/77665 PCT/USOO/l 6032 O
Data Element Name: TOTAL SALARY COSTS Indicators Using: All patient episode of care records. Definition: The monetary value that describes the total salary costs incurred by this health care organization during this episode of care.
Short Name: COST SAL Format:
Length = 8
Type = Numeric
Occurs = 1
Allowable Values: Any whole number greater than or equal to zero, and less than the Total Costs (COST_TOT) of this episode of care.
Missing Data Logic: Field is Null and COST_TOT is NOT Null, or Field is outside the range of allowable values.
***
Data Element Name: TOTAL VENTILATOR DAYS Indicators Using: All patient episode of care records. Definition: The total number of days the patient was on a ventilator during this episode of care.
Short Name: DAYS VENT Format:
Length = 2
Type = Numeric
Occurs = 1
Allowable Values: 0-99
Missing Data Logic: Field is Null and one or more of the ICD-9CM codes indicate ventilator administration.
WO 00/77665 PCT/USOO/l 6032 f
Data Element Name: CONTRACT Indicators Using: None. Definition: Identification number of primary contract insurer with primary coverage for this episode of care.
Short Name: CONTRACT Format:
Length = 5
Type = Character
Occurs = 1
Allowable Values: Any valid insurer contract ID. Missing Data Logic: Field is blank or does not represent a valid insurer contract ID.
***
Data Element Name: READMIT Indicators Using: Unplanned readmission. Definition: Identifies this episode of care as an unplanned readmission related to a previous admission.
Short Name: READMIT Format:
Length = 1
Type = Character
Occurs = 1
Allowable Values: Y/N
Missing Data Logic: None specified.
***
SI
Data Element Name: HCFA DRG SEVERITY SCORE Indicators Using: All patient episode of care records. Definition: The severity score as assigned by an HCFA Grouper. Short Name: SEVERITY Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values: 0-4 Missing Data Logic: Field is Null or outside the range of allowable values.
Data Element Name: HCFA DRG SEVERITY SCORE AT ADMISSION Indicators Using: All patient episode of care records. Definition: The severity score as assigned by an HCFA Grouper at time of admission.
Short Name: SEV ADMIT Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values: 0-4
Missing Data Logic: Field is Null, or outside the range of allowable values.
WO 00/77665 PCT USOO/l 6032
SJ
Data Element Name: SRDRG SEVERITY
Indicators Using: None.
Definition: SRDRG Severity score
Short Name: SEV_SRDRG
Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values: Any valid SRDR<
Missing Data Logic: Field is blank or (
***
Data Element Name: RDRG SEVERITY
Indicators Using: None.
Definition: Refined Diagnosis Related Group severity score
Short Name: SEV_RDRG
Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values: 0-3
Missing Data Logic: Field is Null or outside the range of allowable values.
WO 00/77665 PCT/USOO/l 6032
Si
Data Element Name: ALL-PATIENT REFINED DRG SEVERITY SCORE Indicators Using: All patient episode of care records. Definition: The severity score of this episode of care as assigned by an
APR-DRG Grouper.Severity is also assigned when the case is grouped by any grouper
Short Name: SEV APRDRG Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values: 0-4 Missing Data Logic: Field is Null or is not within the range of allowable values.
Data Element Name: APACHE SEVERITY SCORE
Indicators Using: None.
Definition: Apache ICU Severity score,
Short Name: SEV_APACHE
Format:
Length = 3
Type = Numeric
Occurs = 1
Allowable Values: 0-99
Missing Data Logic: Field is blank or outside the range of allowable values.
WO 00/77665 PCT USOO/l 6032
SS
Data Element Name: CLINICAL PATHWAY Indicators Using: All patient episode of care records. Definition: The category that describes the clinical pathway of this episode of care. This element is User Defined.
Short Name: PATHWAY Format:
Length = 5
Type Character
Occurs = 1
Allowable Values: User Defined Missing Data Logic: None specified.
Data Element Name: SEVERITY ADJUSTED TOTAL DIRECT COST Indicators Using: All patient episode of care records. Definition: The value that describes the total direct costs the health care organization incurred from this episode of care, adjusted to account for severity.
Short Name: SA DCOST Format:
Length = 8
Type = Numeric
Occurs = 1
Allowable Values: Any value greater than or equal to zero that is less than the severity adjust total cost (SA_COST).
Missing Data Logic: Field is Null, or Field not within the range of allowable values. ***
WO 00/77665 PCT/USOO/l 6032
S
Data Element Name: ORIGINAL DIAGNOSIS RELATED GROUP Indicators Using: All patient episode of care records. Definition: The Diagnosis Related Group assigned by the health care organization (rather than that assigned by the present invention) to this episode of care.
Short Name: DRG ORIG Format:
Length = 3
Type = Character
Occurs = 1
Allowable Values: Any valid HCFA DRG code. Missing Data Logic: Field is blank or does not represent a valid HCFA DRG.
***
Data Element Name: RDRG DIAGNOSIS RELATED GROUP
Indicators Using: None.
Definition: Refined Diagnosis Related Group
Short Name: DRG_RDRG
Format:
Length = 3
Type = Character
Occurs = 1
Allowable Values: Any valid RDRG code. Missing Data Logic: Field is blank or outside the range of allowable values.
WO 00/77665 PCT/USOO/l 6032
51
Data Element Name: SRDRG DIAGNOSIS RELATED GROUP Indicators Using: None. Definition: SRDRG Grouper assigned refined Diagnostic Related Group code.
Short Name: DRG SRDRG Format:
Length = 3
Type = Character
Occurs = 1
Allowable Values: Any valid SRDRG Code Missing Data Logic: Field is blank or is outside the range of allowable values.
***
Data Element Name: ALL-PATIENT REFINED DRG Indicators Using: All patient episode of care records. Definition: The Diagnosis Related Group as assigned by an All-Patient
Refined DRG Grouper.
Short Name: DRG APRDRG Format:
Length = 3
Type = Character
Occurs = 1
Allowable Values: Any valid APR-DRG code. Missing Data Logic: Field is blank or does not represent a valid APR-DRG code.
WO 00/77665 PCT/USOO/l 6032
St
Data Element Name: APG Indicators Using: None. Definition: The category that describes the Abulatory Patient Group assigned to this episode of care.
Short Name: APG Format:
Length = 3
Type = Character
Occurs = 1
Allowable Values: Any valid Abulatory Patient Group code. Missing Data Logic: Field is blank or outside the range of allowable values.
***
Data Element Name: DIAGNOSIS PRESENT ON ADMISSION Indicators Using: All patient episode of care records. Definition: A value that describes whether the corresponding ICD-9-CM
Diagnosis code was present at the time of the patient's admission to this health care organization.
Short Name: Ann Format: Y/N/U
Length = 2
Type = Character
Occurs = 12
Allowable Values:
Y = DXnn present on admission N = DXnn not present on admission
U = Status of OXnn on admission is not known
Missing Data Logic: Field is blank or does not represent an allowable value.
WO 00/77665 PCT/USOO/l 6032
ST
Data Element Name: AUDIT STATUS Indicators Using: All patient episode of care records. Definition: The category that describes the audit status for this episode of care.
Short Name: AUDIT Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values:
0 = Unaudited
1 = Audited at least once
2 = Final sign off for submission to state 9 = Unknown
Missing Data Logic: Field is Null or does not represent an allowable value.
Data Element Name: PATIENT DEATH (BINARY FLAG) Indicators Using: All patient episode of care records. Definition: The boolean description of whether the patient died during this episode of care.
Short Name: DEATH Format: Boolean
Length = 1
Type = Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowable value. ***
WO 00/77665 PCT/USOO/l 6032
C O
Data Element Name: MEDICARE DRG (BINARY FLAG)
Indicators Using: None.
Definition: The Medicare-assigned Diagnosis Related Group
Short Name: MCARE_DRG
Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowable value.
***
Data Element Name: MEDICARE DRG RISK (BINARY FLAG) Indicators Using: None. Definition: Boolean value that represents whether the episode of care is classified as a risk due to the Medicare Diagnosis Related
Group.
Short Name: MCARE RISK Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value. ***
WO 00/77665 PCT/USOO/l 6032
6 /
Data Element Name: MCAID PERD (Binary Flag) Indicators Using: None. Definition: Indicates that this episode of care represents a Medicare Per
Diem Patient.
Short Name: MCAID PERD Format:
Length = 1
Type : Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value.
***
Data Element Name: MEDICAID PER DIEM RISK (Binary Flag) Indicators Using: None. Definition: A boolean value that represents whether the episode of care is classified as a financial risk due to its Medicaid Per Diem contract.
Short Name: MCAID RISK Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value. ***
WO 00/77665 PCT/USOO/l 6032
*A
Data Element Name: COMMERCIAL PER DIEM (Binary Flag)
Indicators Using: None.
Definition: Indicates that the episode of care represents a Commercial Per
Diem Patient.
Short Name: COMM PERD
Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value.
*-»-*
Data Element Name: COMMERCIAL PER DIEM RISK (Binary Flag) Indicators Using: None. Definition: A boolean value which represents whether the episode of care classifies as a risk of financial loss due to its Commercial Per
Diem contract.
Short Name: COMM RISK Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value. ***
WO 00/77665 PCT/USOO/l 6032
✓3
Data Element Name: PATIENT UNINSURED (Binary Flag) Indicators Using: All patient episode of care records. Definition: Boolean flag describing whether the patient was uninsured during this episode of care.
Short Name: UNINSURED Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value.
Data Element Name: NEWBORN (Binary Flag) Indicators Using: All patient episode of care records. Definition: Boolean flag that describes whether the patient is neonatal. Short Name: NEWBORN Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value.
WO 00/77665 PCT/USOO/l 6032 i
Data Element Name: EOC IS OBSTETRIC (Binary Flag) Indicators Using: All patient episode of care records. Definition: Boolean flag that describes whether the episode of care is
Obstetric.
Short Name: OB Format:
Length = 1
Type : Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value.
***
Data Element Name: EOC IS CARDIOVASCULAR (Binary Flag) Indicators Using: All patient episode of care records. Definition: Boolean flag that describes whether the episode of care is cardiovascular.
Short Name: CARDIAC Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value. ***
WO 00/77665 PCT/USOO/l 6032
Li
Data Element Name: EOC IS PSYCHOLOGICAL (Binary Flag) Indicators Using: All patient episode of care records. Definition: Boolean flag that describes whether the episode of care is
Psychological.
Short Name: PSYCH Format:
Length = 1
Type : Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value.
***
Data Element Name: C-SECTION (Binary Flag) Indicators Using: All patient episode of care records. Definition: Boolean flag that describes whether the patient underwent a C-
Section.
Short Name: C SECTION Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value.
WO 00/77665 PCT/USOO/l 6032
Data Element Name: DISCHARGED HOME (Binary Flag) Indicators Using: All patient episode of care records. Definition: Boolean flag that describes whether the patient was discharged to their home.
Short Name: DC HOME Format:
Length = 1
Type Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value.
***
Data Element Name: DISCHARGED TO ANOTHER DEPARTMENT
(Binary Flag)
Indicators Using: All patient episode of care records. Definition: Boolean flag that describes whether the patient was discharged to another department within this health care organization.
Short Name: DC HERE Format:
Length = 1
Type Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value. ***
WO 00/77665 PCT/USOO/l 6032 i i
Data Element Name: ADMITTED FROM THIS HCO (Binary Flag) Indicators Using: All patient episode of care records. Definition: Boolean flag that describes whether the patient was admitted from another department within this health care organization, excluding the emergency room.
Short Name: FROM HERE Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value.
Data Element Name: ADMITTED FROM EMERGENCY ROOM
(Binary Flag)
Indicators Using: All patient episode of care records. Definition: Boolean flag that describes whether the patient was admitted from the emergency room of this health care organization.
Short Name: FROM ER Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value. ***
61
Data Element Name: POST PROCEDURE MORTALITY IN 48 HOURS
(Binary Flag)
Indicators Using: All patient episode of care records. Definition: Boolean flag that describes whether the patient died within 48 hours of a major surgical procedure which took place during this episode of care.
Short Name: MORT P48 Format:
Length = 1
Type : Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value.
***
Data Element Name: PROCEDURE COMPLICATIONS (Binary Flag) Indicators Using: All patient episode of care records. Definition: Boolean flag that describes whether a complication occurred after a procedure which took place during this episode of care.
Short Name: COMPL Format:
Length = 1
Type = Numeric
Occurs = 1
Allowable Values:
0 = False
1 = True
Missing Data Logic: Field is Null or does not represent an allowed value.
WO 00/77665 PCT/USOO/l 6032 i f
Data Element Name: DIRECT VARIABLE COST Indicators Using: All patient episode of care records. Definition: The monetary value that represents the sum total of all Direct Variable costs incurred during this episode of care.
Short Name: COST DVAR Format:
Length = 8
Type = Numeric
Occurs = 1
Allowable Values: Any whole number that is greater than or equal to zero and less than the Total Direct Cost (COST_DIR) of this episode of care.
Missing Data Logic: Field is Null or outside the range of allowable values.
Data Element Name: HOSPITAL TYPE Indicators Using: All patient episode of care records. Definition: The category that describes the health care organization itself. Short Name: HOSP TYPE Format:
Length = 2
Type = Character
Occurs = 1
Allowable Values:
1 = General Acute Care
2 = Skilled Nursing or Immediate Care
3 = Psychiatric Care
4 = Alcohol Drug Rehabilitation
5 = Acute Rehabilitation Care
Missing Data Logic: Field is blank or does not represent one of the allowed values. ***
WO 00/77665 PCT/USOO/l 6032
70
Data Element Name: EXPECTED LOS Indicators Using: All patient episode of care records. Definition: The total number of days the patient was expected to stay under the care of the health care organization during this episode of care, given the assigned Diagnosis Related Group and Severity score.
Short Name: EX P_LOS
Format:
Length = 5
Type = Numeric
Occurs = 1
Allowable Values: Any whole number that is greater than or equal to zero and less than or equal to 99,999.
Missing Data Logic: Field is Null.
Data Element Name: EXPECTED CHARGES Indicators Using: All patient episode of care records. Definition: The total charges expected to be issued by the health care organization for this episode of care, given the assigned
Diagnosis Related Group and Severity score.
Short Name: EXP CHARGE Format:
Length = 8
Type = Numeric
Occurs = 1
Allowable Values: Any whole number greater than or equal to zero and less than or equal to 99,999,999.
Missing Data Logic: Field is Null.
***
WO 00/77665 PCT/USOO/l 6032
1/
Data Element Name: EXPECTED COST Indicators Using: All patient episode of care records. Definition: The sum of all expected costs incurred by the health care organization during this episode of care, given the Diagnosis
Related Group and Severity score.
Short Name: EXP COST Format:
Length = 8
Type = Numeric
Occurs = l
Allowable Values: Any whole number that is greater than or equal to zero, and less than or equal to 99,999,999.
Missing Data Logic: Field is Null.
Data Element Name: EXPECTED TOTAL DIRECT COST Indicators Using: All patient episode of care records. Definition: The expected sum of all Direct Costs incurred by this health care organization during this episode of care, given the assigned Diagnosis Related Group and Severity score.
Short Name: EXP DCOST Format:
Length = 8
Type = Numeric
Occurs = l
Allowable Values: Any whole number that is greater than or equal to zero, and is less than the expected Total Cost (EXP_COST).
Missing Data Logic: Field is Null or does not represent an allowable value.
WO 00/77665 PCT/USOO/l 6032
7
Data Element Name: EXPECTED DIRECT VARIABLE COST Indicators Using: All patient episode of care records. Definition: The expected sum of all Direct Variable costs incurred by the health care organization during this episode of care, given the assigned Diagnosis Related Group and Severity score.
Short Name: EXP DVCOST Format:
Length = 8
Type = Numeric
Occurs = 1
Allowable Values: Any whole number greater than or equal to zero, and less than the expected Total Direct Cost (EXP_DCOST).
Missing Data Logic: Field is Null, or does not represent an allowable value.
Data Element Name: PATIENT SOCIAL SECURITY NUMBER
Indicators Using: All patient episode of care records. Definition: The United States Government assigned social security number of the patient. This field is encrypted.
Short Name: PATJDS
Format:
Length = 11 Type = Character Occurs = 1
Allowable Values: Any.
Missing Data Logic: None.
73
Data Element Name: PATIENT ENCOUNTER NUMBER
Indicators Using: All patient episode of care records
Definition: The patient encounter number.
Short Name: PAT DX
Format:
Length = 20
Type = Character
Occurs = 1
Allowable Values: Any.
Missing Data Logic: : Field is blank.
***
Data Element Name: HOSPITAL SYSTEM Indicators Using: All patient episode of care records. Definition: The category that describes the system to which this health care organization belongs. Usually this evaluates to the Hospital
Association.
Short Name: HOSP SYS Format:
Length = 3
Type = Character
Occurs = 1
Allowable Values: Any valid hospital system category. Missing Data Logic: Field is blank.
***
WO 00/77665 PCT/USOO/l 6032
71
Data Element Name: HOSPITAL REGION Indicators Using: All patient episode of care records. Definition: The category that describes the hospital system assigned region in which this health care organization belongs.
Short Name: HOSP REG Format:
Length = 3
Type = Character
Occurs = 1
Allowable Values: Any category code that represents a valid region. Missing Data Logic: Field is blank.
Data Element Name: HOSPITAL COUNTY Indicators Using: All patient episode of care records. Definition: The category or abbreviation that describes the county in which the health care organization is physically located.
Short Name: HOSP CNTY Format:
Length = 4
Type = Character
Occurs = 1
Allowable Values: Along with the following codes, 4-digit abbreviations are also valid:
01 = Alameda 13 = Imperial 25 = Modoc 37 = San Diego 49 = Sonoma
02 = Alpme 14 = Inyo 26 = Mono 38 = San Fransisco 50 = Stanislaus
03 = Amador 15 = Kern 27 = Monterey 39 = San Joaquin 51 = Sutler
04 = Butte 16 = Kings 28 = Napa 40 = San Louis Obispo 52 = Tehama
05 = Calaveras 17 = Lake 29 = Nevada 41 = San Mateo 53 = Trinity
06 = Colusa 18 = Lassen 30 = Orange 42 = Santa Barbara 54 = Tulare
07 = Contra Costa 19 = Los Angeles 31 = Placer 43 = Santa Clara 55 = Tuolumne
08 = Del Norte 20 = Madera 32 = Plumas 44 = Santa Cruz 56 = Ventura
09 = El Dorado 21 = Maπn 33 = Riverside 45 = Shasta 57 = Yolo
10 = Fresno 22 = Maπposa 34 = Sacramento 46 = Sierra 58 = Yuba 11 = Glenn 23 = Mendocino 35 = San Bemto 47 = Siskiyou
12 = Humboldt 24 = Merced 36 = San Bernadino 48 = Solano
Missing Data Logic: Field is blank. WO 00/77665 PCT/USOO/l 6032
1≤
Data Element Name: ICD-9-CM PROCEDURE DAY Indicators Using: All patient episode of care records. Definition: The number of days from the admission date (ADM_DATE) that the ICD-9-CM procedure was performed upon the patient.
Short Name: PROC«« DAY Format:
Length = 3
Type = Numeric
Occurs = 8
Allowable Values: Any whole number that is greater than zero and less than or equal to the length of stay (LOS).
Missing Data Logic: Field is Null or is outside the range of allowable values.
Data Element Name: SURGICAL FLAG
Indicators Using: All patient episode of care records. Definition: A single character flag assigned during processing that recognizes major surgical procedures within this episode of care.
Short Name: SURGICAL
Format: P or S Length = 3 Type Character Occurs = l
Allowable Values:
P = Primary ICD-9-CM procedure is a major surgery.
S = Any one of the secondary ICD-9-CM procedures is a major surgery.
Missing Data Logic: None specified.
*** WO 00/77665 PCT/USOO/l 6032 *
Data Element Name: SEVERITY ADJUSTED DIRECT VARIABLE COST Indicators Using: All patient episode of care records. Definition: The sum of all direct variable costs incurred by the health care organization during this episode of care, adjusted to account for the assigned Severity score.
Short Name: SA DVCOST Format:
Length = 8
Type = Numeric
Occurs = 1
Allowable Values: Any whole number greater than or equal to zero, and less than the severity adjusted total direct cost (SA_DCOST).
Missing Data Logic: Field is Null or is outside the range of allowable values.
Data Element Name: EXPECTED REVENUE Indicators Using: All patient episode of care records. Definition: The expected revenue of this episode of care, given the assigned Diagnosis Related Group and Severity score.
Short Name: EXP REV Format:
Length = 8
Type = Numeric
Occurs = 1
Allowable Values: Any whole number greater than or equal to zero, and less than or equal to 99,999,999.
Missing Data Logic: Field is Null or outside the range of allowable values.
WO 00/77665 PCT/USOO/l 6032
71
Data Element Name: CAREUNIT Indicators Using: All patient episode of care records. Definition: The 3-digit abbreviation of the care unit most involved in this episode of care.
Short Name: CAREUNIT Format:
Length = 3
Type = Character
Occurs = 1
Allowable Values: Any valid ICU abbreviation. Missing Data Logic: Field is blank or outside the range of allowable values.
***
Data Element Name: ADMISSION DAY OF WEEK Indicators Using: All patient episode of care records. Definition: The category that describes the day of the week upon which the patient was admitted.
Short Name: ADM DOW Format:
Length = 1
Type = Character
Occurs = 1
Allowable Values:
1 = Monday
2 = Tuesday
3 = Wednesday
4 = Thursday
5 = Friday
6 = Saturday
7 = Sunday
Missing Data Logic: Field is blank, or does not contain a valid category code.
***
It
Data Element Name: EXPECTED MORTALITY Indicators Using: All patient episode of care records. Definition: A value which represents the expected possibility of mortality given the Diagnosis Related Group and Severity score assigned to this episode of care.
Short Name: EXP MORT Format:
Length = 3
Type = Numeric
Occurs = 1
Allowable Values: Any whole number between zero and 999. Missing Data Logic: Field is Null, or outside the range of allowable values.
Data Element Name: EXPECTED COMPLICATIONS Indicators Using: All patient episode of care records. Definition: A value that represents the likelihood of a complication given the Diagnosis Related Group and Severity score assigned to this episode of care.
Short Name: EXP COMP Format:
Length = 3
Type = Numeric
Occurs = 1
Allowable Values: Any whole number between 0 and 999. Missing Data Logic: Field is null, or outside the range of allowable values.
WO 00/77665 PCT/USOO/l 6032
7f
Data Element Name: DYNKEY Indicators Using: All patient episode of care records. Definition: A Clinical Dynamics assigned record identifier that is unique to the scope of all known data.
Short Name: DYNKEY Format:
Length = 10
Type : Numeric
Occurs = 1
Allowable Values: Any Missing Data Logic: Field is Null.
***
fd
Appendix B. Models and Model Building
I General
To build a model, we start with a collection of reference data that contains outcomes and factors Although there may be many other elements in the data, for the purpose of the building the model the data can be considered to be a table of the form
Yl XI, 1 Xl,2 Xl,n Y2 X2,l X2,2 X2,n
Ym Xn-,1 Xm,2 Xm,n
Here each row represents a single admission, the Yi value is the workload for the ι"' admission, and the Xι,j is the factor value for the ιth admission and the j"1 factor
II. Mathematical Models
The general formula for descπbing or predicting a dependent vaπable, Y, that is a function of n independent vaπables, Xls X2,...Xn , is
(1) Y = f( X,, X2,...X„ )
Where f is an arbitrary function For our purposes, Y is the workload in the desired units, and the X's are the factors that we will use to describe or predict the workload.
If the formula is of the form
(2) Y = A + B X, + C X2 + Z X„
The formula is said to be linear with constant coefficients This is one of the most common models used to predict workload.
If the formula is of the form ι-l Y _ (A + B Xl + C X2 + Z Xn )
The formula is said to be a model with transformation of the dependent vaπables If the formula is of the form
(4) Y = l/l+ e-<A t B X1 +C X2 + Z Xn >
The formula is said to be a logistic regression model.
These are the three most commonly used mathematical models, but the model can be any equation that relates the outcome, workload units, to the independent vaπables WO 00/77665 PCT/USOO/l 6032
H
The model building problem is to select the coefficients A, B, etc such that the model explains the outcome as well as possible given the known reference data There are numerous methods to accomplish this process These methods include multiple linear regression and curvilinear regression, and other so-called curve fitting methods Although special cases of this problem may be solved analytically, the most general solutions use multi-step processes
The multi-step curve fitting methods start with the definition of an objective function that is to be minimized This objective function is the eπor of the model, and minimizing the objective function corresponds to minimizing the eπor in the model Using the general statement of the reference data m I, and the linear model in 11(2), the observed outcome is
Figure imgf000082_0001
and the Modeled outcome is
(6) A +B Xι,l + C Xι,2 + .. .
The eπor can be defined in numerous ways, including the raw eπor
(7) Yι - A +B Xι,l + C Xι,2 + .. , The absolute eπor
(8) |Yι - A +B Xι,l + C Xι,2 + . .|, the squared eπor
(9) (Yι - A +B Xι,l + C Xι,2 + ...)Λ2, or the maximum eπor
(10) max (Yι - A +B Xι,l + C Xι,2 + . )
The squared eπor is frequently used because it does not allow cancellation of positive and negative eπors and because it magnifies the importance of very large eπors The model is built, then, by selecting coefficients that minimize
(11) Σ (Yi - A +B Xt,l + C Xι,2 + . )Λ2 m if the squared eπor is chosen, and in this case the model is called the least squares model
The selection of the coefficients is actually accomplished by an initial selection more or less at random, then by varying the coefficients m a systematic way until the eπor in (11) reaches an acceptable (predetermined) level There are numerous commercial software packages available to accomplish this task, WINNER uses a commercially available variant of GRG2, generalized reduced gradient method, version 2
III. Heuristic Models
Heuπstic models are those that do not use an explicit mathematical model as above, but use a tπal-and-eπor approach to develop a model Two of the most common heuristic models are genetic algoπthms (GA) and artificial neural nets (NN).
Neural nets typically use a network model as shown below fλ
Figure imgf000083_0001
Neural Net
Here the mputs are the factors XI, X2, etc. and the output is Y similar to our mathematical models above. But instead of a simple linear relationship between the coefficients and factors and the output, the neural net has a weight assigned to each aπow above, i.e. to each relationship between nodes. A defined mathematical operation takes place in each of the node evaluation steps, frequently the summation of the product of each mput and its coπesponding weight. The output of the node is thus determined by the outputs of the nodes in the preceding layer, the weights of the relationship between each of the preceding layer nodes and the receivmg node, and the operation that takes place in the node. This model allows a better prediction accuracy when there are lnteπelationships between the inputs as they affect the output, as might be expected with diagnoses and workload units.
The aπow weight is calculated using the training set using a specific methodology, in our case modified back propagation. The result of this training is similar to the curve fitting descnbed above, in that the weights are varied until the eπor between the predicted and measured Y-values for the training set is minimized. Then the same weights are used for prediction when the inputs are fed to the NN and the output is the predicted Y.

Claims

WO 00/77665 PCT/USOO/l 6032I claim:
1. A method of estimating workload units of staffing needed at a facility to care for a person having a predetermined charactenstic using a computer comprising the steps of: obtaining a data set corresponding to a plurality of past persons having been treated for the predetermined characteristic, each of said past persons having a corresponding record that includes workload unit information and factor information relating to the predetermined charactenstic; building a model that associates the obtained workload information and the factor information in the data set; obtaining new factor information associated with the person having the predetermined charactenstic; estimating the workload units of staffing needed to care for the person by applying the model to the new factor information associated with the person having the predetermined charactenstic
2. A method according to claim 1, wherein the step of obtaining a data set includes the step of obtaining factor information relating to age and sex; and wherein the step of building the model uses the factor information relating to age and sex.
3. A method according to claim 2 wherein the step of obtaining factor information mcludes obtaining factor information relating to diagnosis; and wherein the step of building the model uses the factor information relating to diagnosis.
4. A method according to claim 2 wherein the step of obtaining factor information includes obtaining factor information relating to pharmacy prescription data and wherein the step of building the model uses the factor information relating to pharmacy prescnption data.
5. A method according to claim 2 wherein the step of obtaining factor information includes obtaining factor information relating to physiologic monitoring; and wherein the step of building the model uses the factor information relating to physiologic monitonng.
6. A method according to claim 2 wherein the physiologic monitonng information mcludes one of blood pressure, respitory rate and temperature. WO 00/77665 PCT/USOO/l 6032
9*t
7. A method according to claim 2 wherein the step of obtaining factor information mcludes obtaining factor information relating to insurance payer; and wherein the step of building the model uses the factor information relating to insurance payer.
8. A method according to claim 2 wherein the step of obtaining factor information includes obtaining factor information relating to clinical status; and wherein the step of building the model uses the factor information relating to clinical status.
9. A method according to claim 1 wherein the model is one of a linear model, a nonlinear model and a neural network model.
10. A method according to claim 1 wherein the step of obtaining the new factor information include obtaining new factor information relating to age and sex.
11. A method according to claim 10 wherein the step of obtaining new factor information includes obtaining new factor information relating to past visits to the facility.
12. A method according to claim 11 wherein the new factor information relating to past visits to a facility is a number corresponding to the number of past visits to a facility.
13. A method according to claim 12 wherein the number corresponding to the number of past visits to the facility is obtained for a predetermined period of time.
14. A method according to claim 13 wherein the predetermined peπod of time is within the last 5 years.
15. A method according to claim 13 wherein the predetermined penod of time is within the last year.
16. A method according to claim 10 wherein the step of obtaining new factor information mcludes obtaining new factor information relating to diagnosis.
17. A method according to claim 10 wherein the step of obtaining new factor information includes obtaining new factor information relating to pharmacy prescnption data.
18. A method according to claim 10 wherein the step of obtaining new factor information includes obtaining new factor information relating to laboratory test results WO 00/77665 PCT/USOO/l 6032
9S
19 A method according to claim 10 wherein the step of obtaining new factor information includes obtaining new factor information relating to physiologic monitonng
20. A method according to claim 19 wherein the physiologic monitoring information includes one of blood pressure, respitory rate and temperature.
21. A method according to claim 10 wherein the step of obtaining new factor information includes obtaining new factor information relating to pharmacy prescnption data
22. A method according to claim 10 wherein the step of obtaining new factor information includes obtaining new factor information relating to insurance payer.
23. A method according to claim 1 wherein the predetermined characteristic is a DRG parameter.
24. A method according to claim 1, wherein the predetermined charactenstic is a cluster of diagnoses.
25. A method according to claim 1 wherein the predetermined charactenstic is a cluster of procedures.
26. A method according to claim 1 wherein the step of building the model includes the steps of identifying a plurality of candidate predictive factors usmg the factor information; determining a subset plurality of predictive factors that are capable of being used to predict the association of the obtained workload information and the factor information in the data set; selecting factor information relating to the subset plurality of predictive factors from the record of each of the past persons; and building the model using the selected factor information relating to the subset plurality of predictive factors.
27. A method of determining a model to estimate workload units of staffing needed at a facility to care for a person having a predetermined charactenstic using a computer compnsing the steps of obtaining a data set corresponding to a plurality of past persons, each of said past persons having a corresponding record that includes workload unit information and factor information; identifying a subset plurality of said plurality of past persons that have the predetermined charactenstic as a portion of the factor information; building a model that associates the obtained workload information and the factor information in the data set using the factor information associated with the subset plurality WO 00/77665 PCT/USOO/l 6032
28. A method according to claim 27 wherein the step of building the model includes the steps of identifying a plurality of candidate predictive factors using the factor information, determining a subset plurality of predictive factors that are capable of bemg used to predict the association of the obtained workload information and the factor information in the data set, selecting factor information relating to the subset plurality of predictive factors from the record of each of the subset plurality of said plurality of past persons; and building the model using the selected factor information relating to the subset plurality of predictive factors.
29. A method according to claim 27 further including the steps of: obtaining a updated data set coπesponding to another plurality of past persons, each of said past persons having a corresponding record that includes workload unit information and factor information; identifying another subset plurality of said another plurality of past persons that have the predetermined charactenstic as a portion of the factor information; updating the model that associates the obtained workload information and the factor information in the data set using the factor information associated with the another subset plurality.
30. A method according to claim 29 wherein the updated data set corresponding to the another plurality of past persons comes from persons having been cared for at the facility.
31. A method according to claim 29 wherein the updated data set corresponding to the another plurality of past persons comes from persons having the predetermined charactenstic.
32. A method of estimating workload units of staffing needed at a facility to care for a person having a predetermined charactenstic using a computer that operates upon a model comprising the steps of: obtaining new factor information associated with the person having the predetermined characteristic; estimating the workload units of staffing needed to care for the person by applying the model to the new factor information associated with the person having the predetermined charactenstic.
33. A method according to claim 32 wherein the step of obtaining the new factor information include obtaining new factor information relating to age and sex.
34. A method according to claim 33 wherein the step of obtaining new factor information includes obtaining new factor information relating to past visits to the facility. WO 00/77665 PCT USOO/l 6032
«7
35. A method according to claim 34 wherein the new factor information relating to past visits to a facility is a number corresponding to the number of past visits to a facility
36. A method according to claim 35 wherein the number corresponding to the number of past visits to the facility is obtained for a predetermined period of time.
37. A method according to claim 36 wherein the predetermined penod of time is within the last 5 years.
38. A method according to claim 36 wherein the predetermined penod of time is within the last year.
39. A method according to claim 32 wherein the step of obtaining new factor information includes obtaining new factor information relating to diagnosis.
40. A method according to claim 32 wherein the step of obtaining new factor information includes obtaining new factor information relating to pharmacy prescnption data.
41. A method according to claim 32 wherein the step of obtaining new factor information includes obtaining new factor information relating to laboratory test results.
42. A method according to claim 32 wherein the step of obtaining new factor information includes obtaining new factor information relating to physiologic monitonng
43. A method according to claim 42 wherein the physiologic monitoring information includes one of blood pressure, respitory rate and temperature.
44. A method according to claim 32 wherein the step of obtaining new factor information mcludes obtaining new factor information relating to insurance payer.
45. A method according to claim 32 wherein the predetermined characteristic is a DRG parameter.
46. A method according to claim 32, wherein the predetermined charactenstic is a cluster of diagnoses.
47. A method according to claim 32 wherein the predetermined charactenstic is a cluster of procedures. WO 00/77665 PCT/USOO/l 6032
48. A method of estimating workload units of staffing needed at a facility to care for a plurality of persons using a computer compnsing the steps of. obtaining a data set corresponding to a plurality of past persons, each of said past persons having a corresponding record that includes workload unit information and factor information relating to predetermined charactenstics; determining a plurality of groups, such that each group has a plurality of common charactenstics; identifying one of the plurality of groups with which each record should be associated to obtain corresponding records for each group; building a model for each of said groups using the obtained corresponding records for each group, each model associating the obtained workload information and the factor information m the data set; obtaining new factor information associated with each of the plurality of persons; determining the group with which each of the plurality of persons should be associated; and estimating the workload units of staffing needed to care for the plurality of persons using the group that each of the persons has been determined to be associated with, the new factor information for each of the plurality of persons, and the models corresponding to the determined groups.
49. A method according to claim 41, wherein, dunng the step of estimating, each model operates only on the new factor information of the persons that are determined to be in the same group as that of the model.
50. A method according to claim 49, wherein, each model operates to provide partial workload information and the partial workload information is summed to obtain the workload units of staffing.
51. A method of seventy adjusting workload unit information relating to a plurality of past persons using a computer compnsing the steps of: obtaining a data set corresponding to the plurality of past persons, each of said past persons having a corresponding record that includes workload unit information and factor information, building a model that associates the obtained workload information and the factor information in the data set; severity adjusting the workload unit information to obtain seventy adjusted workload unit information.
52. A method according to claim 51, wherein the step of obtaining a data set includes the step of obtaining factor information relating to age and sex; and wherein the step of building the model uses the factor information relating to age and sex. WO 00/77665 PCT/USOO/l 6032
?f
53. A method according to claim 51 wherein the step of obtaining factor information includes obtaining factor information relating to diagnosis; and wherein the step of building the model uses the factor information relating to diagnosis
54. A method according to claim 51 wherein the step of obtaining factor information includes obtaining factor information relating to pharmacy prescnption data and wherein the step of building the model uses the factor information relating to pharmacy prescnption data.
55. A method according to claim 51 wherein the step of obtaining factor information includes obtaining factor information relating to physiologic monitonng; and wherein the step of building the model uses the factor information relating to physiologic monitonng.
56. A method according to claim 51 wherein the physiologic monitonng information includes one of blood pressure, respitory rate and temperature.
57. A method according to claim 51 wherein the step of obtaining factor information includes obtaining factor information relating to insurance payer; and wherein the step of building the model uses the factor information relating to insurance payer.
58. A method according to claim 51 wherein the step of obtaining factor information includes obtaining factor information relating to clinical status; and wherein the step of building the model uses the factor information relating to clinical status.
59. A method according to claim 51 wherein the model is a plurality of different models, each different model associated with a corresponding different predetermined charactenstic, and each of said plurality of past persons are associated with one of the corresponding different charactenstics.
60. A method according to claim 59 wherein each different model is one of a linear model, a nonlinear model and a neural network model.
61. A method according to claim 60 wherein the corresponding different predetermined charactenstics are one of plurality of DRG parameters. WO 00/77665 PCT/USOO/l 6032 fύ
62. A method according to claim 60 wherein certain of the corresponding different predetermined charactenstics include different clusters of diagnoses.
63. A method according to claim 60 wherein certain of the corresponding different predetermined characteristics include different clusters of procedures.
PCT/US2000/016032 1999-06-11 2000-06-09 Method and apparatus for automatically allocating staffing WO2000077665A2 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
JP2001503075A JP2003526137A (en) 1999-06-11 2000-06-09 Method and apparatus for automatically allocating staffing
AU54808/00A AU5480800A (en) 1999-06-11 2000-06-09 Method and apparatus for automatically allocating staffing

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US33092099A 1999-06-11 1999-06-11
US09/330,920 1999-06-11

Publications (2)

Publication Number Publication Date
WO2000077665A2 true WO2000077665A2 (en) 2000-12-21
WO2000077665A8 WO2000077665A8 (en) 2001-11-15

Family

ID=23291878

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2000/016032 WO2000077665A2 (en) 1999-06-11 2000-06-09 Method and apparatus for automatically allocating staffing

Country Status (4)

Country Link
JP (1) JP2003526137A (en)
AU (1) AU5480800A (en)
TW (1) TW494331B (en)
WO (1) WO2000077665A2 (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8265978B2 (en) * 2005-09-22 2012-09-11 Siemens Aktiengesellschaft Computerized scheduling system and method for apparatus-implemented medical procedures
WO2018024672A1 (en) * 2016-08-02 2018-02-08 Koninklijke Philips N.V. Health care facility unit computer simulation system
JP2018151956A (en) * 2017-03-14 2018-09-27 株式会社富士通アドバンストエンジニアリング Person-in-charge determining method, person-in-charge determining apparatus, and person-in-charge determining program
WO2020036571A1 (en) * 2018-08-16 2020-02-20 RICHARDSON, Paul, Stephen Systems and methods for automatic bias monitoring of cohort models and un-deployment of biased models
US11062802B1 (en) * 2015-06-04 2021-07-13 Cerner Innovation, Inc. Medical resource forecasting
US11633624B2 (en) 2018-09-04 2023-04-25 Koninklijke Philips N.V. Resource scheduling in adaptive radiation therapy planning
CN117933954A (en) * 2024-03-22 2024-04-26 四川省医学科学院·四川省人民医院 Multi-hospital area manpower resource allocation evaluation method and system based on clinical specialty capability
US12148523B2 (en) 2022-09-22 2024-11-19 Teambuilder LLC Electronic interface for healthcare resource scheduling

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2002203042A (en) * 2000-12-28 2002-07-19 Oriental Yeast Co Ltd Preclinical test support system
JP2003050868A (en) * 2001-08-06 2003-02-21 Koichi Kawabuchi Medical information analysis system using drg

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
No Search *

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8265978B2 (en) * 2005-09-22 2012-09-11 Siemens Aktiengesellschaft Computerized scheduling system and method for apparatus-implemented medical procedures
US11062802B1 (en) * 2015-06-04 2021-07-13 Cerner Innovation, Inc. Medical resource forecasting
WO2018024672A1 (en) * 2016-08-02 2018-02-08 Koninklijke Philips N.V. Health care facility unit computer simulation system
US11594323B2 (en) 2016-08-02 2023-02-28 Koninklijke Philips N.V. Health care facility unit computer simulation system
JP2018151956A (en) * 2017-03-14 2018-09-27 株式会社富士通アドバンストエンジニアリング Person-in-charge determining method, person-in-charge determining apparatus, and person-in-charge determining program
WO2020036571A1 (en) * 2018-08-16 2020-02-20 RICHARDSON, Paul, Stephen Systems and methods for automatic bias monitoring of cohort models and un-deployment of biased models
US11694777B2 (en) 2018-08-16 2023-07-04 Flatiron Health, Inc. Systems and methods for automatic bias monitoring of cohort models and un-deployment of biased models
US11848081B2 (en) 2018-08-16 2023-12-19 Flatiron Health, Inc. Systems and methods for automatic bias monitoring of cohort models and un-deployment of biased models
US11633624B2 (en) 2018-09-04 2023-04-25 Koninklijke Philips N.V. Resource scheduling in adaptive radiation therapy planning
US12148523B2 (en) 2022-09-22 2024-11-19 Teambuilder LLC Electronic interface for healthcare resource scheduling
CN117933954A (en) * 2024-03-22 2024-04-26 四川省医学科学院·四川省人民医院 Multi-hospital area manpower resource allocation evaluation method and system based on clinical specialty capability
CN117933954B (en) * 2024-03-22 2024-06-04 四川省医学科学院·四川省人民医院 Multi-hospital area manpower resource allocation evaluation method and system based on clinical specialty capability

Also Published As

Publication number Publication date
JP2003526137A (en) 2003-09-02
WO2000077665A8 (en) 2001-11-15
TW494331B (en) 2002-07-11
AU5480800A (en) 2001-01-02

Similar Documents

Publication Publication Date Title
US8645158B2 (en) Displaying clinical predicted length of stay of patients for workload balancing in a healthcare environment
US8428963B2 (en) System and method for administering health care cost reduction
US7716066B2 (en) Managing patient bed assignments and bed occupancy in a health care facility
US8666787B2 (en) Method and apparatus for repricing a reimbursement claim against a contract
US20120065987A1 (en) Computer-Based Patient Management for Healthcare
US20100114607A1 (en) Method and system for providing reports and segmentation of physician activities
US20100312581A1 (en) Process and system for efficient allocation of medical resources
US20030163352A1 (en) Method and system for gainsharing of physician services
US20060149784A1 (en) System and method for operating modules of a claims adjudication engine
US20180314802A1 (en) System and method and graphical interface for performing predictive analysis and prescriptive remediation of patient flow and care delivery bottlenecks within emergency departments and hospital systems
US20140025390A1 (en) Apparatus and Method for Automated Outcome-Based Process and Reference Improvement in Healthcare
US11881292B2 (en) Clinical documentation improvement (CDI) smart scoring systems and methods
US20050278196A1 (en) System and method for developing and utilizing member condition groups
WO2000077665A2 (en) Method and apparatus for automatically allocating staffing
Thorwarth et al. Application of discrete-event simulation in health care: a review
Gibson et al. A copayment increase for prescription drugs: the long-term and short-term effects on use and expenditures
Hribar et al. Clinic workflow simulations using secondary EHR data
KR20210147329A (en) An Open-type Platform System for Preventing Spread of Infectious Disease
US20050125257A1 (en) System and method for creating data links between diagnostic information and prescription infornation records
KR102441610B1 (en) Infectious disease spread prevention open platform system
Pi et al. Using Case Mix Index within Diagnosis-Related Groups to Evaluate Variation in Hospitalization Costs at a Large Academic Medical Center
US20050197544A1 (en) System and method for indexing emergency department crowding
Woodbury et al. A model for allocating budgets in a closed system which simultaneously computes DRG allocation weights
Camgöz-Akdağ et al. Improvement of patient pathway in a breast cancer center
Gubler et al. Employee Dilemmas From Competing Organizational Objectives: Insights from Emergency Medical Services

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A2

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY CA CH CN CR CU CZ DE DK DM DZ EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT UA UG UZ VN YU ZA ZW

AL Designated countries for regional patents

Kind code of ref document: A2

Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
AK Designated states

Kind code of ref document: C1

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY CA CH CN CR CU CZ DE DK DM DZ EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT UA UG UZ VN YU ZA ZW

AL Designated countries for regional patents

Kind code of ref document: C1

Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN GW ML MR NE SN TD TG

D17 Declaration under article 17(2)a
ENP Entry into the national phase in:

Ref country code: JP

Ref document number: 2001 503075

Kind code of ref document: A

Format of ref document f/p: F

WWE Wipo information: entry into national phase

Ref document number: 54808/00

Country of ref document: AU

REG Reference to national code

Ref country code: DE

Ref legal event code: 8642

122 Ep: pct application non-entry in european phase
点击 这是indexloc提供的php浏览器服务，不要输入任何密码和下载