Decoding the Upcoding: Have Incentives for Health Information Technology Increased Documented Severity for Hospitalized Patients under the Hospital Readmission Reduction Program?
Discussant: Jeffrey S. McCullough
In this paper, we test whether hospitals’ adoption of EHRs under the HITECH Act led to increases in patients’ documented severity. We use 100% MedPAR inpatient claims from discharges between January 1, 2011 (the beginning of incentive payments under HITECH) and September 1, 2015. To account for unobserved patient differences, we identify a cohort of patients who had two admissions for similar diagnoses at two different hospitals - one receiving incentives under the HITECH ACT, the other not receiving incentives - that occurred more than 1 month and less than 1 year apart. We exclude all patients that were transferred in or out of hospitals but otherwise use the same inclusion and exclusion criteria employed by CMS for the hospital-wide 30-day readmission measure. Our dependent variable is the Hierarchical Condition Category (HCC) risk score, calculated solely on the basis of secondary diagnoses documented during the index admission. We estimate an episode-level linear model in which HCC risk score is regressed on hospitals’ receipt of HITECH incentives, controlling for year and season dummies, admitting diagnosis, whether the patient was admitted to a HITECH hospital during the first or second admission, hospital fixed effects, and patient fixed effects.
We then test whether any changes in documented severity following HITECH represent a more, or less, accurate assessment of true severity. To do this, we compare the predictive accuracy of patient severity between hospitals that received and did not receive incentives under HITECH. Predictive accuracy is assessed using the difference in the C-statistic between two logistic models: one predicting 30-day readmission as a function of relevant hospital characteristics, demographic characteristics, and severity; the other as a function of only hospital characteristics and demographic characteristics. A larger difference in the C-statistic for hospitals receiving incentives under HITECH would indicate that true severity was more accurately assessed in these hospitals.
Our findings inform the ongoing policy debate about the effects of health information technology and payment reform in the United States.