Existing Data Resources for Assessing Survey and of care problems. PROMPTS-2 samples are designed to be representative of Institutional providers of medical services participate in the Medicare and Medicaid programs through a formal certification process. Qualifications are reviewed at least yearly thereafter by state surveyors under agreements with HCFA to determine whether the facility should be recertified or terminated from participation in the Medicare or Medicaid programs. Information on the basic structural characteristics of certified facilities-size, staffing, staff qualifications, services provided, compliance with safety codes-is recorded on the standardized forms submitted to HCFA. Significant variations have been found, however, in the ways in which surveyors interpret the guidelines and criteria for identifying and reporting deficiencies in facility conditions or operations.8 As noted in chapter 2, HCFA has begun to redesign the survey and certification process, particularly as it pertains to long-term care facilities. These efforts include both greater standardization of processes and instruments and an increased emphasis on the collection of data directly related to patient care and well-being. The long-term care survey includes a patient interview component and new sampling methods for patient surveys. The system initially requires sampling 25 percent of the residents from each facility for indepth review. Those residents will be sampled randomly, but surveyors may also select additional residents for indepth review, based on their assessment of possible problems that require additional attention. Thus, the reviewed cases include both random and targeted cases. A more complex system providing for selection of cases stratified by facility size is planned. This should enable the generation of more precise national estimates of the status of nursing home residents, provided the total population and mix of randomly selected versus targeted cases within each facility is recorded. 8See, for example, the Report of the Committee on Nursing Home Regulation, Institute of Medicine, Improving the Quality of Care in Nursing Homes (Washington, D.C.: National Academy Press, 1986), pp. 108ff. Existing Data Resources for Assessing A standard worksheet is used in each indepth review of residents' care to record the information obtained by observation, interview, and record review. The worksheet documents residents' ability to perform the basic activities of daily living, skin condition, a range of physical care and therapy needs, dietary needs, mental and emotional condition, and ability to function socially. Most items listed on the form represent patient care problems (for example, presence of rashes, problems with decubitus ulcers or catheters, weight problems, dehydration, privacy not maintained, signs of mental or physical abuse). In addition to interviewing and observing patients, surveyors review residents' medical records, care plans, and evaluations. These detailed worksheets are maintained by the states. While information from the patient surveys is used to support findings of various deficiencies that are reported to HCFA, the only data relating to the patient surveys that are recorded on HCFA automated files are data on average facility-wide skilled nursing patients' disability levels, as measured by patients' ability to perform basic activities of daily living. HCFA has requested $3.5 million in fiscal year 1988 to enhance the data base on the health care and health status of nursing home residents. After the current planning stages, HCFA envisions a 5-year implementation period, including projects to identify and describe existing data resources, design a national data collection system, conduct pilot tests, train interviewers, and evaluate various program components. HCFA has also made changes in the survey and certification process for home health agencies. Inspections now include visits to patients, although these are not yet mandatory for all facilities.o Surveyors determine whether visits are desirable or necessary based on their onsite facility review. The criteria for patient selection in the home health survey guidelines are less systematic than those for nursing home review (a minimum of three home visits per agency, but recommended sample of 10-20). Observations and information obtained by interviewing patients are incorporated into survey forms and specific citations of deficiencies are noted, but surveyors do not use standard patient interview worksheets, and no patient-level information is maintained. 9The Omnibus Budget Reconciliation Act of 1987 (Public Law 100-203) revised the survey and certification requirements for home health agencies. The law contains a requirement that standard facility surveys include visits to patients' homes, using a protocol to be developed, tested, and validated no later than January 1, 1989. Existing Data Resources for Assessing Central HCFA Data Information from Medicare claims and survey and certification activities The hospital stay record file (and the medical provider and review file derived from it) contain extensive information on inpatient hospital stays, including diagnoses, surgical procedures, length of stay, and patient discharge destination. The recently developed Part B beneficiary medicare annual data (B-MAD) files include extensive patient-level information on the use of physician and other part B services for a 5-percent sample of beneficiaries. The Medicare history sample file includes longitudinal data for all covered services for a sample of Medicare beneficiaries. A new file still under development, the Medicare automated data retrieval system (MADRS) will include information on individuals' use and charges for the full range of Medicare services and charges, as well as inpatient diagnoses, in a format specifically designed to be useful to 10 There are three basic components of the Medicare statistical system: beneficiary enrollment and eligibility data, hospital insurance (part A) and supplementary medical insurance (part B) data, and provider of services data. HCFA, Program Statistics, Medicare and Medicaid Data Book, 1984 (Washington, D.C.: U.S. Government Printing Office, June 1986), p. 54. Existing Data Resources for Assessing Accuracy and Completeness of researchers. The provider of service file includes facility-level data that can be matched, by provider number, with the other files. Descriptions of the data in these files are presented in appendix V. The Medicare statistical system relies heavily on the activities of claims processors for ensuring the accuracy and completeness of data. As the previous discussion has demonstrated, the verification of billing data consists mostly of internal consistency and logic checks. Independent confirmation of the accuracy of particular data elements, based on reviews of medical records, focuses primarily on payment-related questions, for example, were diagnoses and procedures recorded correctly, so that the appropriate DRG assignment could be made? Medicare editing systems virtually ensure that submitted bills are complete, or at least contain all the information necessary for processing claims. Little is known about the volume of bills that are never submitted."1 Information on the overall accuracy of the data in the Medicare data 12 Some improvement in coding accuracy may have occurred with the introduction of the ICD-9-CM System. The emphasis on diagnosis and procedure coding in the DRG System is also believed to have led to greater 11A report prepared under contract to HCFA reviewing Medicare data from Florida and South Caro lina found, for example, that “in a substantial number of cases some expected claims appear to be missing." In some cases, there were no part B bills for patients either preceding hospital admission or after discharge. This could reflect problems in the submission of claims or problems related to continuity of care. In addition, however, there were instances where there were no physician bills associated with inpatient stays. A more precise estimate of missing bills was not developed as part of the study, but the potential problem was raised as an issue that should be pursued. Mandex, Inc., “Developing MD-DRG Algorithms," (Vienna, Va.: February 6, 1985). 12Overall, the study abstractors agreed with the determination of principal diagnosis found in the HCFA data about 60 percent of the time. But the levels of agreement varied considerably by diagnosis. For patients with chronic ischemic heart disease, the level of agreement was about 37 percent; for diabetes mellitus, about 50 percent. Institute of Medicine, Reliability of Medicare Hospital Discharge Records (Washington, D.C.: National Academy of Sciences, November 1977). Existing Data Resources for Assessing accuracy in coding. PROS Specifically examine the accuracy of diagnostic coding as part of their DRG validation function. The increasing sophistication of carrier edits and screens is also designed to improve accuracy of coding for procedures, although information on the accuracy of coding on part B records is very limited. 13 Nevertheless, there has been no systematic, national assessment of the accuracy of the key data elements in HCFA files since 1977, so the extent to which data problems undermine quality assurance activities is unknown. The accuracy of data elements that do not directly affect reimbursement amounts is especially problematic. Because DRGS often encompass broad ranges of ICD-9-CM codes, some coding errors do not affect DRG assignment, but may be important in quality and utilization review. The coding of complications and/or secondary diagnoses has been immaterial for assignment of many DRGS, particularly for patients who are over 69 years old, because age currently serves as a proxy for complicating conditions in many DRG assignments. 14 Neither PROS nor intermediaries have strong incentives to verify information on hospital discharge abstracts detailing type of admission (billing form categories are emergency, urgent, or elective newborn, and unknown) or source of admission (billing form categories are physician referral, clinical referral, HMO referral, transfer from a hospital, transfer from a skilled nursing facility, or transfer from another health care facility), because this information is usually not used for determining eligibility for services, coverage of services, or payment amounts. This type of information could, however, be useful in screening for possible quality of care problems. (See chapter 6.) Serious questions about the accuracy of discharge destination data recorded on inpatient hospital and skilled nursing facility bills have also been raised. This data element indicates whether a patient has been discharged, and if so, where to (home to self care, to a short-term hospital, a skilled nursing facility, an intermediate care nursing facility, to some 13 A recently completed study which compared 1981 part B Medicare claims data to information abstracted from medical records found that for three specific procedures studied, the claims file data was accurate in over 95 percent of the 4,988 cases reviewed. J. Kosecoff, et al., “Obtaining Clinical Data on the Appropriateness of Medical Care in Community Practice," Journal of the American Medical Association (November 13, 1987), p. 2541. 14The Prospective Payment Assessment Commission has recommended that DRGs not be defined based on age, and the Secretary of HHS has concurred with this recommendation and proposed to eliminate age over 69 as a criterion for DRG classification. Prospective Payment Assessment Commission, 1988 Adjustments to the Medicare Prospective Payment System Report to the Congress, November 1987, p. 28. |