Date of Award
Open Access Thesis
Medical Doctor (MD)
Harlan M. Krumholz
Regional Variation in Hospitalization Rates: Causes and Implications
Sachin J. Shah, Harlan M. Krumholz. Section of Cardiology, Department of Internal Medicine, Yale University, School of Medicine, New Haven, CT.
Background: Admission rates vary by regions and states, but the extent by which variation in regional admission rates are related to variation in the medical need of populations and the association with hospital outcomes is unknown. To address these issues, we examine two cardiovascular conditions that differ in physician discretion to admit, acute myocardial infarction (AMI), less discretionary, and heart failure (HF), more discretionary. We first determined whether regional cardiovascular risk factors predict admission rates and then examined whether regional admission rates were related to 30-day risk-standardized mortality and readmission rates (RSMRs and RSRRs).
Methods: We used 2006-2008 Medicare ICD-9-CM claims data and the Medicare Denominator file to determine AMI and HF admission rates. The statewide prevalence of cardiovascular risk factors were obtained from the 2007 Behavioral Risk Factor Surveillance System. First, the relationship between statewide AMI and HF admission rates and cardiovascular risk factors was determined by a multivariate, least squares linear regression model. Second, hierarchical logistic models were used to estimate hospital RSMRs and RSRRs and then were aggregated to the level of hospital referral regions (HRRs). The correlation (R2) was obtained by linear regression to characterize the relationship between both AMI and HF admission rates and regional RSMRs and RSRRs. Where significant relationships were observed, "cross condition" analyses were performed comparing admission rates of one condition against the RSMR or RSRR of the other in an effort identify potentially confounded relationships.
Results: In the first analysis, cardiovascular risk factors explained 49% of the variation observed in statewide AMI admission rates and 50% of the variation in HF admission rates. In the second analysis, regional AMI admission rate was not correlated with AMI RSMR (R2 0.01, 95% CI 0.00-0.04). Regional HF admission rate was inversely correlated with HR RSMR (R2 0.13, 95% CI 0.07-0.21). Regional AMI hospitalization rate was weakly correlated with AMI RSRR (R2 0.05, 95% CI 0.02-0.11). Regional HF admission rate was modestly correlated with HF RSRR (R2 0.25, 95% CI 0.17-0.34). In the cross condition analyses, regional HF admission rate was not associated with AMI RSMR (R2 0.00, 95% CI 0.00-0.02) but was associated with AMI RSRR (R2 0.25, 95% CI 0.17-0.34).
Conclusion: Cardiovascular risk factors explain part, but not all, of the variation in AMI and HF admission rates. The modest association between regional HF admission rate, a more discretionary admission condition, and both AMI and HF RSRRs suggests a system propensity to patients. The same was not seen true of AMI a less discretionary admission condition. The modest inverse relationship between regional HF admission rate and HF RSMR, which was not observed with AMI RSMR, suggests an unmeasured confounder affecting the HF RSMR model.
Shah, Sachin, "Regional Variation In Hospitalization Rates: Causes And Implications" (2011). Yale Medicine Thesis Digital Library. 1595.