- Using grouped diagnostic and 'present at admission' codes in CMS modeling improved predictions for total patient payments within 30 days of hospitalization for acute myocardial infarction, heart failure and pneumonia, a comparative effectiveness study released Wednesday in JAMA Network Open found.
- The study used data from Medicare fee-for-service hospitalizations for the three conditions at acute care hospitals from July 2013 to September 2015.
- Changing variables from those currently used in CMS payment prediction models has implications for research, benchmarking, public reporting and calculations for population-based programs, the researchers said.
As more than 10,000 Americans age into Medicare daily, accurately predicting beneficiary spending is becoming critical. Indeed, by 2030, the percentage of Americans 65 years and older will outnumber those under the age of 18 for the first time in the country's history, according to the U.S. Census Bureau.
And experts worry that the federal entitlement program won't be able to keep up.
In 2015, CMS developed a model using the conditions of AMI, HF and pneumonia to support public reporting of risk-standardized Medicare payments at hospitals across the U.S. These payment measures are slated to become part of the Hospital Value-Based Purchasing in 2021, and Medicare Advantage uses similar models to calculate payments.
Researchers, led by Harlan Krumholz at the Department of Internal Medicine at the Yale School of Medicine, studied whether using a new approach to predict Medicare beneficiary payment modeling could improve predictions.
Researchers found changing variables used in CMS payment prediction models improved the accuracy of the models.
Current CMS payment prediction models group ICD-9-CM diagnosis codes into clinically coherent categories. The grouping of diagnostic codes reduces the number of candidate variables but may diminish the performance of the models by combining different codes with different relationships to the outcomes, the researchers found.
But using single instead of grouped diagnostic codes has the potential to enhance payment predictions. "The use of single codes rather than condition categories could provide a better characterization of risk," the researchers said.
Using so-called present on admission codes, which have been included in claims since 2014, also has the potential to enhance risk modeling, they said.
"With POA codes, it is possible to know that the condition was not a complication during hospitalization, allowing the use of more diagnoses from the index with the possibility of model improvement," according to the study, which found the new models incorporating single and POA codes significantly improved model performance.
The upper range of predictions for patients with AMI increased from $55,387 to $113,446, the study found, while the estimate on the lower range decreased from $10,993 to $8,204.
"The difference revealed that the new models were better able to identify people at the ends of the payment spectrum," the authors wrote. "The improvements in model performance were meaningful."