Using Patient-Reported Data to Identify the Effect of Prospective Payment on Upcoding by Providers

Tuesday, June 14, 2016
Lobby (Annenberg Center)

Author(s): Philip Britteon; Matt Sutton

Discussant: Leemore Dafny


Under prospective payment systems, providers can increase their revenue by upcoding patients into higher paid Diagnostic-Related Groups. Empirical examination of this upcoding effect is limited by a lack of information on the true conditions of patients. Studies often rely on provider-reported data and must overcome the demanding task of identifying an exogenous change in the incentive to upcode without a change in the incentive for patient selection. To overcome this problem, we exploit a unique linked dataset in which patients and providers report information on co-morbidities and complications.


To compare findings on provider responses to a reduction in the incentive to upcode, using patient-reported and provider-reported data on co-morbidities and complications.

Payment Change

We exploit an exogenous payment change within the English ‘Payment by Results’ system whereby the incentive to record a co-morbidity or complication for patients undergoing knee revision replacements was removed. Prior to April 2010, providers were reimbursed up to US$5,468 more per patient coded with a co-morbidity or complication. After April 2010, a flat rate was paid regardless of the recorded patient diagnoses.


We utilise Patient Reported Outcome Measures (PROMs) data recording the pre-operative co-morbidities and post-operative complications of 114,277 patients undergoing specific elective procedures between April 2009 and March 2011. We link this with the data reported for payment by providers from Hospital Episode Statistics (HES).


We estimate and compare responses to the announcement and introduction of the payment change within a trend adjusted difference-in-difference model on:

  1. Provider-reported co-morbidity and complications
  2. Patient-reported co-morbidity and complications
  3. Other aspects of care, including waiting times and monthly volumes.


Once the payment change was announced, providers reported a 10.2 percentage point reduction in the proportion of patients having a co-morbidity or complication. This reflected a reduction in the average monthly volume of patients with a co-morbidity or complication and no change in volume for those without. This was not matched by a reduction in the proportion of patients reporting a comorbidity or complication. When the payment change was introduced, there was no change in the comorbidity and complication rate reported by providers but an increase of 3.9 percentage points in the comorbidity and complication rate reported by patients. There was also a 4.5 day reduction in waiting times.


The availability of provider-reported and patient-reported data allows for improved inference on the nominal incentives associated with a payment design change and highlights the pitfalls of conclusions based solely on provider-reported data. Our analysis of provider-reported data suggests that refining DRG payments based on secondary diagnoses has no unintended effect on hospitals. Comparisons to patient-reported data, however, show that hospitals responded to the removal of the upcoding incentives as expected by reducing their coding of patients with co-morbidities or complications.