A recent study published in JAMA Open Network found that using individual codes and present on admission designations instead of group codes can improve the predictability of patient total payment within 30 days of a hospitalization.
Yale researchers looked at nearly 2 million Medicare fee-for-service hospitalizations and compared the risk models from CMS and using present on admission codes and single diagnosis codes and separation of index admission codes. They found that changes to the candidate variables in CMS models improved the risk models predicting patient total payment for a 30-day period after hospitalization for three conditions––acute myocardial infarction (AMI), heart failure (HF) and pneumonia.
And changing the models could have bigger impacts.
“Better models can potentially improve research, benchmarking, public reporting and calculations for population-based programs,” wrote first author Harlan M. Krumholz, MD, SM, of the Department of Health Policy and Management at Yale School of Public Health, and colleagues.
The mean risk standardized payment for the regular CMS models was $23,105, compared to $23,211 for the individual-codes model. The new model better identified people at the ends of the payment spectrum, the researchers wrote.
“The individual-codes models produced better discrimination, goodness of fit, and predictive range, with similar calibration,” Krumholz et al. wrote. “…Our findings suggest that there are opportunities for improvements in the predictive performance of these models with relatively simple changes in the candidate variables.”