Carve-out diagnoses. Picking managed care's "bad apples"

Under capitated managed care, patients with chronic or catastrophic illnesses may have difficulty enrolling and remaining in health plans for which participation is voluntary; and providers who care for them may incur substantial financial risk. Some plans exclude providers who attract such high-cost patients. A potential solution is to identify (or carve out) diagnoses which are relatively clinically distinct and likely to be umfbmuy high cost, paying for them outside of the usual capitated rate. We developed a method for identifying potential carve-outs from administrative data using ICD-9 diagnosis and procedure codes. Two non-elderly populations, one a targe fee-for-service (FFS) plan covering greater than 1 million lives in both FY1992 and 1993, and the other a medium-sized managed care organization (MCO) with over 300,000 slightly younger (59% vs 72% age > 18 years) members in FY 1990 and 1991 were used. Clinical experts prospectively identified a potential list of carve-out diagnoses that, based on their experience, WOE likely to be high-cost We then examined whether patients with these diagnoses met the following criteria for carve-outs: mean annual costs (inpatient and outpatient) over $25,000 and 50% of cases in more than one population with annual costs exceeding $25,000. Only 15% of the 3 digit ICD-9 codes (93 diagnoses) clinically expected by the experts to be high cost met both carve-out criteria. Approximately 75% of individuals with at least one carve-out condition were adults (age > 18 yrs). In the FFS plan, the carve-out list identified 2% of enrollees who represented over 30% of total costs; in the MCO, the lia identified less than 0.5% of members who were responsible for 8% of total costs. Approximately 70% of individuals with total costs exceeding $25,000 in the FFS plan carried at least one carve-out diagnosis, however only 25% of high-cost MCO members had diagnoses which were included in the carve-out list. We conclude that it is possible to predict some, but not all, high-cost patients. This may be easier to do in a FFS than in an MCO environment. Further testing of carve-outs is needed to judge their ability to prevent discrimination against patients and physicians and evaluate their feasibility in clinical practice.