Does patient severity affect variation in hospital use

Objective. To determine the relative importance of practice-style versus the amount of diagnosed disease (disease prevalence) and severity of disease (severity effect) in explaining small area variations in the number of people hospitalized.;Design. I used year 2000 Medicare claims data on both inpatient admissions and outpatient visits in Massachusetts for 15 different medical conditions (diseases). Seventy (70) hospital service areas were constructed using Ward's clustering algorithm to group zip codes that displayed a similar pattern of hospital discharges. In each area, the number of people with each medical condition was calculated as the number of people hospitalized plus the number treated on an outpatient-only basis. Observed disease prevalence was calculated as the ratio of the number of people with the condition to the number expected given the age-sex distribution in the area. The observed severity effect was calculated as the average population-based severity according to the DxCG classification system, which estimates an area's overall severity based on demographic information, and a compilation of the total illness that was encountered by the residents of that area over the previous 12 months. The observed practice-style effect was calculated as the ratio of the proportion of those with the condition that were hospitalized to the proportion expected given the age-sex distribution in the area. A hierarchical Bayesian method was used to estimate the "true" underlying effect sizes in each area and to rank areas while taking into account different effects.;Findings. In earlier work, researchers have shown that the rank order of areas by hospitalization rate changes markedly after adjusting for disease prevalence. In this study, I found that adjusting for the severity effect, results in a substantial change in rank order in addition to that caused by adjusting for disease prevalence.;Implications. Previous studies of small area variations have rarely taken into account the severity of disease in an area's population. The focus of public policy interventions to reduce variations in hospitalization rates should be directed to factors giving rise to differences in disease prevalence and disease severity, as well as to practice-style.