137
Are Hospitals Coding Up or Coding Correctly? Evidence from Electronic Medical Record Adoption

Tuesday, June 14, 2016
Lobby (Annenberg Center)

Author(s): Jianjing Lin; Gautam Gowrisankaran; Keith Joiner

Discussant:

This paper seeks to evaluate whether the adoption of electronic medical records (EMRs) leads to upcoding in the market of hospitalized Medicare patients. As healthcare costs in the U.S. have continued to rise, observers have been concerned about the incentives of healthcare providers to inflate their revenues by billing for care that is not provided and/or categorizing a condition as being more serious than justified by the medical evidence. This process is called upcoding. A number of studies with older data have found evidence that hospitals upcode. A more recent literature has sought to understand the interaction between hospital EMRs and upcoding, but the literature has not reached a consensus.

We identify four different potential explanations for how EMR adoption might affect billing. First, EMRs might facilitate upcoding by hospitals. Second, EMRs might allow hospitals to code more accurately. Third, hospitals with EMRs might select different patients. Finally, EMR hospitals might provide a different amount of services to the same patients.

In order to provide evidence on upcoding, we use what we believe is a novel identification strategy that helps separate our four explanations. We use a triple difference: (1) between EMR and non-EMR hospitals; (2) before and after the 2007 Medicare payment reform; and (3) between medical and surgical admissions. The 2007 payment reform is important because it made it more difficult to obtain high reimbursements but relatively less hard for hospitals with EMRs. Why is the third difference important? Physicians who are dealing with medical admissions are trained to document conditions. Surgeons are trained to document procedures, and hence may not document all medical conditions of their patient in the same detail. EMRs can only code conditions to the extent that physicians document them. Medicare inpatient upcoding is entirely dependent on medical and not surgical secondary conditions, of the sort preferentially documented during medical admissions. To our knowledge, this third difference has not been exploited in the literature on upcoding.

We investigate the separate impact of EMR adoption and the 2007 payment reform on medical and surgical DRGs using a panel of data from 2006 to 2010. We use the universe (100% sample) of Medicare inpatient hospital claims data, and link our data with data on hospital characteristics and with EMR adoption data.

Our main findings are that there is an increase only in medical DRG top codes following the payment reform for hospitals that adopted EMRs. This effect is bigger for early adopters, who started EMRs in 2006 or earlier. The finding that there is no significant change in top codes for surgical DRGs and that the pattern does not follow financial incentives suggests that this is not done strategically but rather simply where possible. This in turn suggests that EMRs allow hospitals to code more accurately, and are not facilitating upcoding. We also examine whether EMRs might cause hospitals to provide more or less services to patients and whether EMRs might cause hospitals to select different patients. We find little evidence of these.