Method for evaluation of health care quality5544044Abstract A software-based medical information system performs a method of analyzing health care claims records for an enrolled population (e.g., HMO, Medicaid) to assess and report on quality of care based on conformance to nationally recognized medical practice guidelines or quality indicators. The system analyzes health care received by enrollees having a specified health care condition by: providing to the system health care claims records for a selected enrollee population; defining at least one health care condition in terms of health care events reportable in health care claims records; identifying in the health care claims records those enrollees meeting the definition for that health care condition; defining health care quality criteria for that health care condition in terms of health care events reportable in health care claims records; comparing the health care quality criteria for the at least one health care condition to the health care claims records for at least a portion of those enrollees meeting the definition for that health care condition; and developing and outputting from the system a health care quality report based on the comparison and formulating action recommendations to improve care. The system provides an efficient means to supplement claims data with data from patient medical records. Claims What is claimed is: Description TECHNICAL FIELD
______________________________________
Enrollment/Membership Data
1.R Unique Member Identifier
2.R Enrollment Date
3.R Expiration Date
4.R Birth Date
5.R Gender
6. Patient Name (only if a medical record review is requested)
7. Medicare Membership Indicator
8. Medicaid Membership Indicator
9. Zip code
Medical Professional Claims Data
1.R Unique Member Identifier
2.R Date of Service
3.R Provider Type Code (for physicians this must indicate
medical specialty)
4. Unique Provider Identifier (required if Provider Type is
not available on the claims record or if medical record
review is anticipated)
5.R Diagnosis Code (ICD-9)
6.R Procedure Code (CPT-4)
7.R Place of Service Code (e.g., hospital, doctor's office)
Hospital (Inpatient/Emergency Room) Claims Data
1.R Unique Member Identifier
2.R Admission Date
3. Discharge Date
4.R Provider Type Code (for in-patient claims this must indicate
type of facility. e.g., hospital, emergency room, urgi-center)
5. Unique Provider Identifier (required if Provider Type is
not available on the claims record or if in-patient medical
record review is anticipated)
6.R Place of Service Code
7. Unique Admission Identifier
8.R Diagnosis Code (ICD-9)
9. Revenue Code (i.e., from UB82 codes)
Pharmaceutical Claims Data (not used for all conditions)
1.R Unique Member Identifier
2.R Drug Codes
3. Route of Administration
4. Quantity of Drug
5.R Filled Date
______________________________________
The minimum required fields for the present implementation of the invention are marked with an "R" in the above data definitions. "CPT-4" in the medical professional data refers to an edition of the Current Procedure Terminology code, published by the American Medical Association. In the medical professional data and the hospital data, "ICD-9" refers to an edition of the International Classification of Diseases; "UB82" refers to revenue codes published by Health Care Finance Administration; and Place of Service Codes are those promulgated by the American Medical Association. Alternative coding systems can be used, based on a code conversion table associating the preceding codes with the alternative coding system. If the customer desires, provider-specific data is also extracted 16 from the customer data, permitting the later analysis to be broken down by the particular provider of services or products, which may be a particular doctor, clinic or hospital. (The provider data also can be used to identify the possible sources of medical records.) The provider data fields utilized are as follows:
______________________________________
Provider Data (e.g. Physician Hospital, Urgi-Center)
______________________________________
1. Unique Provider Identifier
2. Provider Type Code (if not in Claims Files)
3.RM Name
4.RM Street Address
5.RM City
6.RM State
7.RM Zip Code
8.RM Area Code
9.RM Phone Number
______________________________________
Additional fields required if a medical record review is to be undertaken are marked with an "RM". The resulting files are merged 19 to produce uncorrected master files 21. Because claims data are not totally "dean", it will typically be necessary to exclude duplicate claims 23 and claims that have been reversed through the claims adjudication process (coordination of benefits, subrogation, clerical errors). This produces a master file 25 of health care claims records. No adjustment is performed for claims lag. It is assumed that a customer will not forward data for analysis until the claims have been sufficiently processed for the period of study desired. At this point the health care claims records are ready to be analyzed for indicators of quality. But this analysis requires two further preliminary steps: (1) defining at least one health care condition in terms of health care events reportable in health care claims records; and (2) defining health care quality criteria for each defined health care condition, also in terms of health care events reportable in health care claims records. This must be done in advance of the analysis of the master file 25 of health care claims records. The reason that it is necessary to define a health care condition is that quality measurements are specific to discrete health care conditions: what a provider should do obviously depends on the member's condition, the symptoms, diagnosis or the health care status (e.g., pregnancy). More important for this invention, individual health care claims data items accumulated in the normal course of business by a health care management entity may or may not include a direct and specific diagnosis of the particular health care condition suffered by an enrollee or the particular health care condition that is of interest from a quality measurement viewpoint. Even if a diagnosis is present in claims records, it is not necessarily valid to take it at face value. The claims data include one or more codes identifying specific treatment events or outcomes, such as, an office visit, a blood test, a CAT scan, a surgical procedure, a drug prescription, a hospital stay with charges for drugs, X-rays or specific therapeutic regimens, a hospital readmission or a death. It is a selection of such data items and their relationships that becomes the definition of a particular health care condition. For purposes of the present invention "health care condition" is broadly defined to mean a condition in the nature of a disease or an organic dysfunction or a "condition" that might also be viewed as a status or an outcome. In its present implementation, the following conditions characterized as diseases or organic dysfunctions or outcomes are addressed: diabetes mellitus, pediatric asthma, hypertension, breast cancer, hospital readmissions and ambulatory surgical complications. The following conditions characterized as status-based are also addressed: pediatric immunizations, breast cancer screening, cervical cancer screening, caesarian section and prenatal care. The point of each condition is to provide a focus for analysis of quality of care, not for medical diagnosis. Given this orientation, the preceding conditions were selected for implementation, because each can be associated with a cluster of health care events that can be related to health care quality criteria as well as with data that can be used to identify a population having the condition. A proper course of testing and treatment is part of quality care to maximize the health of a person with diabetes mellitus. To protect the health of children with certain age status, a prescribed set of immunizations is agreed to be necessary. To protect the health of women in various age groups and/or meeting other status criteria, certain screening tests and examinations, at prescribed intervals, are agreed to be necessary. To protect the health of newborns and their mothers certain prenatal care is agreed to be advisable. As will be seen below, defining some health care conditions for purposes of quality measurement is easier than defining others. The caesarian section condition is usually easy to define, because there will often be a specific claims code when the surgical procedure is performed. However, in some circumstances, there will not be a specific claims code and the condition will be deduced from the presence of a combination of claims codes, perhaps including amniocentesis, frequency of prenatal visits and specific anesthesia or surgery events. Pediatric asthma is an example of a more subtle condition, and, as will be shown, is defined in terms of a set of medical claims codes, hospital claims codes and pharmacy claims codes with specified Boolean (logical) relationships. Selecting the codes for such conditions can be difficult, because they are not always diagnosed the same way and may be diagnosed only after repeated visits to a health care provider or may not be diagnosed at all. The code selection definitional effort resembles the diagnostic process performed by a health care professional who studies the medical record of a new patient. That is, defining such a condition is a complex logical process requiring extensive clinical and claims data knowledge to implement the definition in a software algorithm. For all of the conditions addressed by the present invention, it is necessary to define the health care condition of interest solely in terms of data elements reportable in health care claims records. This is a different task than if the entire medical record were available, with its greater range of terminology and information. As noted above, in the preferred embodiment the claims data include outpatient records, inpatient records and pharmacy records. Thus, data from all of the record sources can be used to make a health care condition definition. This is significant, because provider practice patterns and enrollee behaviors may cause a clear picture of the enrollee's health care status to emerge only when data from all of these sources is considered. Once a specific definition is created for a health care condition of interest for quality screening, this definition is implemented as a series of logically linked compare operations in the software. Applying the definition of the health care condition to the claims records becomes a matter of performing the compare operations and executing the Boolean logic implicit in the definition (and explicit in the software) to identify those enrollees having claims records in the claims data base that meet the definition. The enrollees so identified are the subset of the universe of health care system enrollees that becomes the focus of the further analysis to determine whether this subset, identified as having the defined health care condition, received quality health care for that condition. The second prerequisite for analysis of the claims master files 25 is defining health care quality criteria for each defined health care condition, also in terms of health care events reportable in health care claims records. Here, the health care system manager must rely on knowledge sources credible to practicing clinicians. Typically, study of the condition will reveal that quality of care is indicated by the absence of certain kinds of health care claims (e.g., an emergency room visit for a pediatric asthma patient, hospitalization for diabetes mellitus, reoccurrence of breast cancer or other adverse events) and the presence of other kinds of health care claims (e.g., annual inoculations, laboratory testing at specific intervals, visits to appropriate medical/surgical specialists, prescription of specific medications). Specific evaluation criteria are developed from clinical or health services literature and research, from standards developed by the professional societies and through private or public (federal government) task forces. As with the definition of the health care condition of interest, the health care quality criteria must be implemented in the software as a set of physician office claims codes, hospital claims codes and pharmacy codes with specified Boolean relationships. Returning to FIG. 2B, step 29 involves the application of the definitions for the health care condition to identify the population having that condition, followed by an analysis of the claims records for that population (a subset of the master files 25) under the defined quality care criteria. The enrollment data 11 is accessed in this process, as well. In general, the invention is implemented to examine one year of claims data, which may begin on any date in a year. Analysis of an enrollee's claims records for the following conditions occurs only when the enrollee has been continuously enrolled for one year: diabetes mellitus, pediatric asthma, hypertension, breast cancer, and breast cancer screening. There is no one year continuous enrollment requirement when the following conditions are analyzed: cervical cancer screening, pediatric immunizations (ages birth to two and four to six), hospital readmission, ambulatory surgical complications and caesarian section. For the prenatal condition, the enrollment requirement is enrollment six months prior to delivery, but enrollment past the date of delivery is not required. The result of the analysis in step 29 is a report that includes: charts and graphs 31 reporting statistically observed quality of care data in the population defined as having the health care condition of interest (e.g., a graph showing emergency room visits by patient age); a written analysis reporting, from a care quality viewpoint, statistical results considered worthy of highlighting 33, 35 (e.g., a report of use of a particular drug intervention and the occurrence of adverse events associated with the particular health care condition); and a report containing recommendations for actions to improve health care quality 37, 39 (e.g., a suggestion that physician guidelines be developed and promulgated stating that a specified laboratory test be performed annually). Action recommendations to improve health care quality are produced for each condition studied. These recommendations are selected from a menu of nine quality management action categories developed based on a synthesis of the relevant research literature and management expertise. The nine categories are: (1) reporting the results of the quality analysis to physicians, (2) disseminating practice guidelines to physicians, (3) instituting condition-specific medical record flowsheets, (4) modifying reimbursement to physicians, (5) instituting provider discipline, (6) providing patient education and reminders, (7) establishing case management programs, (8) improving patient access to care, and (9) changing the benefit design. As can be seen, these actions are directed toward affecting the behavior not only of physicians, but also patients, purchasers or providers of benefits, and health care organizations. Each condition has a set of condition-specific action recommendations defined for it. The customer's report will include those recommendations from the set that are appropriate in view of the results of the analysis performed in step 29. Analysis for multiple health care conditions takes place iteratively through the software at step 41, and the process just described, comprising steps 29, 31, 33, and 37 and producing charts and graphs 31 and reports 35, 39 is repeated, using the next health care condition definition to identify the population having that condition, followed by an analysis of the claims records for that population (a subset of the master files 25) under the defined quality care criteria for that next condition. After all the specified health care conditions have been processed in this manner, the reports for each condition are assembled 43 into a claims-based quality report 45 that is presented to the customer 47. b. Detailed Analysis with Medical Records Because it is recognized that a data base of health care claims, even if it includes data from medical professional services, hospital services and pharmacy services, may not have all the information available in other medical records maintained by the physician or other health care provider, the present invention provides as a supplement, additional methods to gather medical records and perform a more detailed analysis. In FIG. 2C at step 51, the system recognizes whether there is the need for detailed analysis. If no such need exists, no further data collection or analysis occurs. However, if a need for detailed analysis of any health care condition has been determined, then the population identified as having that condition is subjected to sampling 55 to determine for which enrollees additional medical records information will be collected. Both the master claims data files 27 and provider data 18 are used for this step. The determination of which members will have medical records collected for further analysis is based on varying criteria. For most health care conditions, medical records will be of interest for further study of the medical history if an enrollee has exhibited some unfavorable outcome or unexplained deviation from the quality of care criteria. Such enrollees are in greatest need of intervention and the quality of the care they have received or will receive is of greater significance. For instance, it may be desirable to focus detailed analysis on enrollees having more than a specified number of emergency room visits in a year or those for whom there is a deviation from the recommended drug regimen. To select these members, a further sampling definition is developed. It is implemented in software as a set of medical claims codes, hospital claims codes and pharmacy codes with specified Boolean relationships. If the member population remaining after application of these criteria is too large to be economic, it may be reduced in size by appropriate statistical sampling methods. The result of these steps is to define a subset of the population recognized as having the particular medical condition of interest to be subjected to further, more detailed study, based on the additional information available in medical records. Appendix A contains software used to set up the sample for which medical records data is collected and to identify the appropriate source for this data (i.e., provider location). To initiate the gathering of medical records data, a request letter to a physician or other provider is generated. Selecting the appropriate provider to contact may be important, because there may be multiple sources of medical records, some of which will be inadequate for evaluation. The appropriate provider must be selected to receive the letter and be the unit for analysis. In some cases, a second source may also be selected. With the provider's consent, the medical records are abstracted 59 with a particular focus on events that relate to the particular health care condition under study, resulting in a completed medical records abstract form 61. This abstracted information is then entered into the system 63, via personal computer 150 to produce a medical record abstract data file 65. Additional and more highly clinically specific quality of care criteria are then used to analyze the medical record file, in a manner similar to the way the claims data was analyzed. One difference is that the medical record may introduce additional data for medical, hospital or pharmacy activities that can be used to implement the quality care definition in the software. As can be seen, using the present invention means that only a relatively limited amount of medical records data need be assembled for health care quality analysis. Because a condition definition has been employed to identify an enrollee group of interest, only medical records for those enrollees having the condition under analysis are of interest. Moreover, those portions of the medical records relevant to the condition under analysis are abstracted. Thus, the volume of data to be analyzed is kept manageable and cost efficient. In FIG. 2D, it is shown that the result of the additional, detailed analysis is charts and graphs reporting statistically observed data in the population defined as having the health care condition of interest 69 and a report containing recommendations for actions to improve health care quality 71, 73. If detailed analysis of medical records is specified for multiple health care conditions, then the preceding steps are repeated until charts and graphs reporting statistically observed data 69 and a report containing recommendations for actions to improve health care quality 71, 73 are developed for each health care condition. (These are drawn from the same set of condition-specific quality management actions as discussed above in connection with steps 37, 39.) After all the specified health care conditions have been processed in this manner, the reports for each condition are assembled 77 into a detail level report 79 that is presented to the customer 81, and the process ends 83. c. Examples for Particular Conditions To further and more specifically explain the present invention, the data processing steps represented by steps 29 and 31 in FIG. 2B will be explained by way of two examples of specific health care conditions for which the present invention has been implemented. EXAMPLE 1 Pediatric Asthma To demonstrate further specifics about the present invention, reference is made to FIG. 3, which represents further details about the analysis and development of data for steps 29 and 31 in FIG. 2B when the defined health care condition to be studied is pediatric asthma. Still further details about the way in which pediatric asthma health care quality issues are analyzed in accordance with the present invention can be found in Appendix B, which is the SAS source code corresponding to steps 29 and 31 in FIGS. 2A-2D and to FIGS. 3A-3C. The definition of the pediatric asthma condition used by the system is embodied in steps 303, 305, 307 and 309 of FIG. 3A, which determine the presence or absence of certain data in the claims data master file 25. If the claims data are coded in ICD-9 (International Code of Diseases, 9th Edition) form, the criteria may be expressed as follows. A member is defined as having pediatric asthma if he or she is under 19 years of age, is continuously enrolled for the year under study and meets one of the following criteria: 1. At least one hospitalization for asthma (IDC-9=493.xx); or 2. At least one emergency room visit for asthma (ICD-9=493.xx); or 3. Two or more doctor visits for asthma or asthma related conditions, provided that at least one of the visits is for asthma specifically, determined based on the following codes: asthma (493.xx), croup (464.4), acute bronchitis and bronchitis (466.xx), allergic rhinitis (477.xx), unspecified bronchitis (490.xx) and chronic bronchitis (491.xx); or 4. Two or more pharmacy claims for asthma related drugs, at least 45 days apart: beta-2 adrenic drugs; methyl xanthine drugs; and selected other bronchodilator related drugs (beclomethasone, cromolyn, flunsolide) (Criteria 1, 2 and 3 show the ICD-9 code that must be found in order for the criteria to be satisfied. The "xx" following these codes means that only the first three digits of these five digit codes are compared to determine a match.) When the above pediatric asthma criteria are satisfied for any member, his or her name is added to the asthma member file list 313. The decision at step 311 ensures that the full claims data master file is searched for pediatric asthma cases. That is, the available records, which comprehensively cover claims records for medical professional services, claims records for hospital services and claims records for pharmaceutical prescriptions, are searched. It has been found that evidence of the asthma condition may appear in only one of these three categories of records. Yet a proper health care quality study of this condition must identify essentially all enrollees that have the condition, regardless of how the condition is reflected in the available claims records. For possible later analysis, a patient's record in the asthma member file list is flagged to indicate its inclusion in an adverse events and pharmaceuticals grouping. The asthma member file list 313 is then used in step 317 to build a new file that is a subset of the claims data master file and contains claims data only for members meeting the pediatric asthma definition. This asthma master file 319 is then used to calculate a variety of statistics on health care events and to develop charts and graphs to present the aggregated and calculated data. The calculations and the resulting reports and graphs are detailed at steps 321-351 in FIG. 3B. As noted previously, the asthma member list file 313 is flagged to indicate an adverse events and pharmaceutical grouping. At step 353 of FIG. 3C the pharmaceutical asthma subset file is created and subjected to further processing at steps 355, 359, 363, 367. This processing results in the calculation of aggregate statistics of various kinds having to do with prescription drug use for the members identified as pediatric asthma patients. The calculations and the resulting charts are detailed at steps 353-371 of FIG. 3C. Shown at Appendix C are examples of the kinds of reports and charts and graphs developed by the present invention when analyzing the pediatric asthma condition. These reports reflect an actual analysis from a health maintenance organization (the identity has been deleted) and include: General explanation of the pediatric asthma quality review Chart of demographics of children with asthma Pie graph of age distribution of pediatric asthma patients Bar graph of frequency of adverse events Chart of frequency of emergency room visits Chart of frequency of hospital admissions Bar graph of percent of children having an annual office visit for asthma, differentiated by adverse event occurrence Bar graph of average number of office visits Chart of pulmonary function testing Bar graph of influenza immunization Chart of serum theophyline level monitoring Charts of use of beta-agonists, cromolyn, theophylline and barbituates Summary of key statistical findings Quality management actions recommendations, including proposed medical record flow sheet for pediatric asthma patients In the present instance, the management actions recommendations include all those that are part of the condition-specific set of action recommendations for pediatric asthma. In a report for another customer, only a subset of these might appear, because fewer corrective actions are necessary. EXAMPLE 2 Caesarian Section To demonstrate further specifics about the present invention, reference is made to FIGS. 4A-4B, which represents further details about the analysis and development of data for steps 29 and 31 in FIG. 2B when the defined health care condition to be studied is caesarian sections. Still further details about the way in which caesarian section health care quality issues are analyzed in accordance with the present invention can be found in Appendix D, which is the SAS source code corresponding to steps 29 and 31 and to FIGS. 4A-4B. The definition of the caesarian section condition used by the system is embodied in steps 405-421, which test for certain data in the claims data master file 403. At steps 405 and 407 a subset of the master file is created that includes all enrollees who were pregnant and thus potential candidates for a caesarian section birth (C-section). From this subject enrollees having miscarriages are deleted 409, producing a smaller working C-section file 411. This file is divided 413 into a C-section births file 415 and a vaginal births file 417. These two files are then merged 419 to produce the births file 421. Although there is no agreement on an ideal C-section birth rate that would provide a health care quality measure, there is agreement that a C-section birth rate above the national average could indicate that health care providers are performing inappropriate C-section procedures. Accordingly, the health care quality definition implemented at steps 425-447 of FIG. 4C involves calculation of overall C-section birth rates 425, charting that rate by comparison to the national norm 427 and then charting C-section birth rates by age group 429 and by hospital 431-445. Finally, a "trombone chart" 441 is produced that shows the C-section birth rate for individual hospitals, plotted against the average for all hospitals in the analysis and a curve showing the upper 95% confidence bound (based on statistical measures adjusting for sample size). Shown at Appendix E are examples of the kinds of reports and charts and graphs developed by the present invention when analyzing the C-section condition. These reports reflect an actual analysis from a health maintenance organization (the identity has been deleted) and include: General explanation of C-section quality review Bar graph of C-section rate vs. national norm Bar graph of C-section rates by age "Trombone graph" of C-section rates Chart of C-section rates by hospital Summary of key statistical findings Quality management action recommendations C-section medical record abstraction form and instructions (for more detailed analysis) To perform a detailed analysis beyond the level of claims-based quality evaluation, the medical record may be obtained. This would usually be performed in specific populations where the overall observed C-section rate was higher than expected. Therefore, the objective would be to determine whether there were specific indications of medical need for a C-section justifying that the procedure was performed. Appendix E contains the medical record abstraction form and instructions. This data obtained from the medical record would be entered and analyzed in the software of the present invention. d. Summary As can be seen from the preceding, the present invention involves a data processing system and analytic method that utilizes multiple sources of health care data to evaluate health care quality. The invention initially utilizes the health care claims data accumulated by a health care benefit organization as its enrollees utilize health care benefits. Specific health care conditions of interest are defined in terms of the claims data and these definitions are implemented in software logic to allow a subset of enrollees to be identified as having the health care condition of interest. The health care claims data for this subset are then analyzed in terms of health care quality criteria, developed from multiple expert clinical sources. These criteria are also defined in terms of the claims data and implemented in software that develops statistical measures of health care quality. If desired, additional medical record data can be collected for the specific portion of the subset of enrollees identified as having the condition of interest. These additional medical records data supplement the claims data and allow additional, more detailed health care quality criteria to be defined. These criteria are analyzed against the additional data. Further statistical measures of health care quality can then be developed. After analysis of the original claims data and after analysis of the additional medical record data, descriptive and evaluative reports and recommendations on improving health care quality are developed and provided. The invention combines at least three primary advantages to improve over the prior art. First, it permits use to be made of the large claims data bases that are maintained by health care organizations, which in most cases are already in computerized form, permitting large scale health care quality analysis to be performed based on this claims data. Second, it limits the laborious and expensive task of assembling medical records data for quality analysis to a population that has been screened and selected as having specific potential need for quality improvement (and is thus smaller than the entire universe of enrollees). Third, by enabling evaluation of health care quality to be performed in a more standardized and efficient manner, through use of distinctly identified data elements in conjunction with discretely defined clinical algorithms, it permits quality assessment on a larger and more geographically distributed population base and the production of information sought by managers, purchasers, providers and regulators of health care. Appendixes A-E contain the program source code that implement the embodiment described above. ##SPC1##
|
Same subclass Same class Consider this |
||||||||||