Kroch E, Brailer D, Duan M, Terasawa EL; AcademyHealth. Meeting (2003 : Nashville, Tenn.).
Abstr AcademyHealth Meet. 2003; 20: abstract no. 727.
Villanova University, Department of Economics, 800 Lancaster Avenue, 3074 Bartley Hall, Villanova, PA 19085-1678 Tel. (610) 519-6428 Fax (610) 519-6054
RESEARCH OBJECTIVE: While controversy surrounds the purported objectivity of ratings of overall hospital "excellence" among providers, outside purchasers continue to express interest in such rating systems. Using risk-adjusted clinical outcome rates and length-of-stay from inpatient all-payor data from 18 states, we evaluate whether "best performance" is a hospital attribute or a service-line effect. STUDY DESIGN: 56 disease groups composed of 41 3-digit ICD-9 diagnosis codes and 15 broad diagnosis groups and representing more than 90% of all cases in the dataset were identified. For each disease group, hospitals were ranked for efficiency and quality separately with the highest ranking assigned to hospitals with the lowest risk-adjusted length-of-stay and adverse outcome (mortality, complications, morbidity) rates. To qualify as "best performance" for a given disease, facilities needed to fall into the top two quintiles of both efficiency and quality rankings and exhibit a minimum case count of 100. Hospitals were then ranked by their number of eligible disease groups that met the "best performance" criteria. A logit model was constructed to evaluate "best performance" across service-lines. POPULATION STUDIED: 1999 inpatient all-payor data from 2,697 facilities in 18 states. (Data unavailable for remaining 32 states.) PRINCIPAL FINDINGS: Only a small positive correlation is exhibited between the risk-adjusted adverse outcomes and length-of-stay. Of the 2,697 hospitals in the dataset, 1,518 qualify as "best performance" in at least one disease area, implying that over half of the hospitals never qualify as "best performance" in any of the disease groups. Less than 8% (213) of the facilities meet the "best performance" criteria for at least 10 disease areas. On average, these 213 facilities qualify for 40% of their eligible disease groups, indicating that "best performance" is disease-specific and that no hospital is superior across all disease areas. Nevertheless, some hospitals exhibit better performance than others. 31 hospitals qualify for 25-34 "best performance" disease groups, ~63% of their eligible disease groups. By contrast, the average percentage of "best performance" disease groups across all facilities is 16%. The top 213 hospitals are more than twice the size of the average hospital in the dataset, although size is not related to ranking among the top 213 hospitals. These hospitals are also more than twice as likely to be teaching hospitals and are rarely located in rural areas. Geographically, the top 213 hospitals are disproportionately represented in California, North Carolina, Massachusetts, and western Pennsylvania.Logit analysis of "best performance" among disease groups reveals a strong service line effect among Mental Disorders, Circulatory System Diseases, Respiratory System Diseases, Musculoskeletal System Diseases, and Injury/Poisoning. CONCLUSIONS: Most hospitals qualify for "best performance" for fewer than four disease groups, suggesting that "best performance" is influenced by factors lower than hospital level. Logit analysis similarly supports the hypothesis of "best performance" as a service-line effect with individual departments and service-lines displaying superior performance within a hospital. These results suggest that general "top 100" hospital ratings may, in many instances, be misleading. IMPLICATIONS FOR POLICY, DELIVERY OR PRACTICE: Insurers may consider limiting hospital contracts to select "best performance" service-lines, forcing hospitals to specialize.
Publication Types:
Keywords:
- California
- Hospitals
- Logistic Models
- Massachusetts
- North Carolina
- Pennsylvania
- hsrmtgs
Other ID:
UI: 102275696
From Meeting Abstracts