G de Lissovoy

THE RELATIONSHIP OF PROVIDER ORGANIZATIONAL STATUS AND ERYTHROPOIETIN DOSING IN END STAGE RENAL DISEASE PATIENTS

Controversy exists as to whether provider organizational characteristics such as profit status and setting are associated with the content of medical care or efficiency with which care is rendered. Following FDA approval of human recombinant erythropoietin (EPO) for use in clinical practice, Medicare approved coverage for beneficiaries in its end stage renal disease program and established a fixed payment per dose. Because cost of EPO administration varied positively with dose, providers could realize larger profit with prescription of smaller doses. We used Medicare claims data to assess EPO use by renal dialysis providers one year after FDA approval (June 1990) as a function of provider ownership (for-profit, not-for-profit, government agency) and setting (hospital-based, free-standing). Mean dose of EPO was 236 units greater (P =0.0001) for not-for-profit freestanding facilities, 593 units greater (P =0.0001) for government facilities, and 555 units greater for not-for-profit hospitals (P =0.0001 than among for-profit freestanding providers. With fixed payment per dose of EPO, for-profit, freestanding providers prescribed EPO more often and administered smaller doses than not-for-profit or government providers, behavior that is consistent with profit maximization.

PMI7 MODELING ON THE STOCHASTIC FRONTIER: COST OF TREATMENT FOR ACUTE DECOMPENSATED HEART FAILURE

physical examinations. Under-diagnosis for PAD in practice was common and it might have under-estimated PAD prevalence. CONCLUSIONS: Previously reported PAD prevalence varies depending on clinical presentations, different screening tools, and the distribution of risks for PAD. Understanding of and effectively adjusting for these factors may be helpful to appropriately interpret and utilize the study results for future research.

PQP8: EVALUATING BIAS INTRODUCED BY ANNUALIZING UTILIZATION AND COST MEASURES

OBJECTIVES: Administrative databases are often used to investigate patterns of health care resource use and expenditure over time for particular conditions in order to project future costs and to evaluate the cost-effectiveness of interventions. Duration of the period of observation typically varies across individuals due to differences in enrollment/disenrollment dates, timing of the index event, and the database time span. Projecting expenditure on an annual or per member per month basis requires standardizing these periods of observation. We evaluate bias introduced by a simple method for annualizing utilization and cost measures. METHODS: We investigated resource use and costs surrounding an index event; hospital admission with a primary diagnosis of heart failure. For each patient, claims data were available for a pre-event and post-event period ranging from 6 to 24 months. We standardized periods to one year pre- and post-event by first tabulating counts and costs for the actual period observed. We then multiplied these values by the ratio of 365 days to actual days in the period such that values for periods shorter than 365 days were inflated and longer than 365 days deflated. To determine whether this adjustment biased the magnitude of annualized values, we estimated a regression model with annualized cost and adjustment ratio as dependent and independent variables respectively. RESULTS: For the pre-event period, there was no significant association between annualized cost and adjustment factor (p = 0.72, R2 = 0.00014, p = 0.012) indicating absence of bias. While bias was observed in the post-event adjustment factor (p = 0.005), the low model explanatory power (R2 = 0.00869) and lack of correlation (Pearson correlation coefficient 0.093) suggest minimal impact on the magnitude of annualized estimates. CONCLUSION: We find that a simple approach to standardization was reasonably robust. We compare advantages and disadvantages with more complex regression-based methods.