by Christine Woolstenhulme, QCC, QMCS, CPC, CMRS
November 21st, 2014
The outcome of care measures and the utilization measures are risk adjusted. The process of care measures aren't risk adjusted. Risk adjustment of the outcome of care and utilization measures is a multi-step process. These are the major steps in this process displayed in temporal order:
- OASIS-C assessment data collection and transmission; for utilization measures, claims data processing (on-going)
- OASIS-C or claims-based episodes of care creation (on-going)
- Development of outcome and utilization prediction models (periodic)
- Computation of risk adjusted values for home health outcome and utilization quality measures displayed on Home Health Compare (quarterly)
The purpose of risk adjusting outcome and utilization measures is to adjust reported values to account for case-mix differences among home health agencies. That is, by adjusting the observed improvement rates for home health agencies with different patient clientele (case-mix differences), the resulting displayed value more closely reflects differences in agency quality.
Prediction model development — improvement and utilization outcomes
For OASIS-based quality measures, prediction models use patient case mix information taken from OASIS-C SOC/ROC assessments to establish a relationship between these characteristics and the likelihood of a quality outcome. Each improvement and utilization outcome has its own unique prediction model. Virtually every item on OASIS-C, with the exception of a few demographic/patient tracking items and clinical record items, are potential risk factors that can be used to create this prediction equation.
For claims-based utilization quality measures, prediction models use patient case mix information taken from previous Medicare claims data across all care settings to establish the relationship between these characteristics and the likelihood of a quality outcome. Many diagnoses items from previous Medicare claims are potential risk factors.
Several changes in the prediction model building process account for the quality of these prediction models, including:
- Increased number of risk factors (now including care process information)
- More recent assessment data used to develop the models
- Elimination of length of stay as a risk factor
- More appropriate representation of baseline values
- Use of higher correlation value criteria for risk factor inclusion
The statistical and analytic methodology used to create the prediction models includes these steps:
- Acquisition of assessment data and formation of episodes of care
- Specification/calculation of risk factors and outcome quality measures
- Development of prediction model for each outcome quality measure using a standardized statistical methodology
- Review of prediction models by clinical staff to ensure that risk factors used in the models are reasonable and in the appropriate predictive direction
- Re-estimation of any prediction models questioned by clinical staff