Modeling Risk Adjustment in Pay-for-Performance
An article published in the upcoming August issue of Medical Care details a predictive mortality model developed that utilizes laboratory rather than administrative data. The researchers developed the model, noting the inadequacies of billing data as a predictor of mortality. These models are appropriate, necessary tools to adjust clinical data to account for patient risk in pay-for-performance programs.
In developing the predictive models, the research team evaluated a variety of datapoints, including ICD-9 codes, vital signs, demographic characteristics, labs from the time of admission, and altered mental status. Mortality from septicemia, heart failure, hemorrhagic and ischemic stroke, myocardial infarction, and pneumonia during hospitalization were the mortality causes considered.
The researchers concluded that objective data, overall, is a stronger predictor of inpatient mortality. such as pathophysiologic data. Billing and administrative data may add to the overall strength of the predictive model, but should not be the sole source of means to adjust for risk in pay-for-performance efforts. The team also noted that using such data should not be prohibitively expensive.
Related stories:
- Evaluation of Performance Incentives Published
- Rewards Hinge on Safety and Quality
- CMS Releases Measurement Criteria
- Assessment of Medicare Pay-for-Performance Incentive
July 30, 2007 Related topics: Corporate Financials, Quality, Safety, Errors