Comparing the validity of different measures of illness severity: a hospital-level analysis for acute myocardial infarction

Abstract

The aim of the study is to assess the validity of three measures of illness severity (prior year's hospital expenditures, Charlson and Elixhauser indices), by analysing the effect of introducing report cards on hospitals treating patients with acute myocardial infarction (AMI). Medicare claims data were obtained for 1992–1997 for AMI patients aged 65+. We used differences-in-differences regression analysis to assess the impact of report cards introduced in New Jersey and Pennsylvania on the illness severity of AMI patients with and without coronary artery bypass graft (CABG) surgery (relative to states without report cards). The analysis was conducted at the hospital level. For validation we used raw mortality and re-admission trends for AMI patients. While prior hospital expenditures suggest a considerable change in the illness severity of AMI patients in Pennsylvania relative to other states, raw mortality and re-admission trends in Pennsylvania are relatively consistent with the trend in the rest of the USA. In line with raw mortality and re-admission data, the Charlson and Elixhauser indices do not imply a considerable change in the severity of AMI patients in Pennsylvania. For CABG patients, illness severity – as measured by all three severity measurement methods – decreased after introduction of report cards, particularly in Pennsylvania. In conclusion, for AMI patients the Charlson and Elixhauser indices are a more valid measure of illness severity than prior year's hospital expenditures. After report cards were introduced, healthier AMI patients were more likely to receive CABG surgery, while sicker patients were avoided.