Changes in Hospital Acquired Conditions and Mortality Associated with the Hospital Acquired Condition Reduction Program

Tuesday, June 25, 2019: 4:00 PM
Wilson B - Mezzanine Level (Marriott Wardman Park Hotel)

Presenter: Emily Arntson

Co-Authors: Justin Dimick; Ushapoorna Nuliyalu; Joshua Errickson; Tedi Engler; Andrew Ryan

Discussant: Grace Kim

Hospital acquired conditions (HACs) are common, costly, and deadly. Despite recent reductions, HACs still occur at a rate of 121 events per 1,000 discharges. Building on previous payment reforms to reduce HACs, the Centers for Medicare and Medicaid Services announced the Hospital Acquired Condition Reduction Program (HACRP) in August 2013. The program – largely focused on surgical admissions – reduces Medicare payments by 1% for hospitals in the worst performance quartile of targeted hospital-acquired conditions. The design of the HACRP is similar to that of the recently implemented Hospital Readmission Reduction Program: penalty criteria are based on absolute performance levels for a relatively narrow set of measures. On the other hand, measuring HACs is more challenging than measuring readmissions. Specifically, rates of HACs are highly sensitive to hospital coding practices. As a result, the HACRP’s effect on both targeted and downstream clinical outcomes is uncertain.

To examine the effect of the HACRP, we used MedPAR data for Medicare patients undergoing surgery and discharged from eligible hospitals between January 1, 2009 and August 31, 2015 (n = 8,857,877). Using an interrupted time series design, we estimated models with linear splines to test for changes in hospital-acquired conditions and 30-day mortality before the Affordable Care Act (ACA), after the ACA, and after the HACRP. To assess whether improvements may be related to coding practices, we examined changes in the percentage of hospital-acquired conditions that were coded as present-on-admission.

Patients experienced hospital-acquired conditions at a rate of 13.39 per 1,000 discharges (95% confidence interval {CI}, 13.1 to 13.68) in the pre-ACA period. This rate declined after the ACA was passed and declined further after the HACRP was announced (adjusted difference in annual slope, −1.34 [95% CI, -1.64 to -1.04]). Adjusted 30-day mortality was 3.69 (95% CI, 3.64 to 3.74) in the pre-ACA period among patients receiving surgery. 30-day mortality declined after the ACA (adjusted annual slope -0.04 [95% CI, -0.05 to -0.02]) but was flat after the HACRP (adjusted annual slope -0.01 [95% CI, -0.04 to 0.02]). While the use of coding for present-on-admission slightly increased after the HACRP was announced, coding changes did not explain reductions in hospital-acquired conditions.

Overall, hospital-acquired conditions targeted under the HACRP declined at a greater rate after the program was announced. The HACRP was not associated with declines in 30-day mortality. This suggests that either: 1) hospitals are artificially reducing reported HAC rates and not improving downstream clinical outcomes, such as mortality, or 2) hospital-acquired conditions are not sufficiently common or mechanistically related to major harm such that improvement will drive a mortality reduction. Further investigation, especially into hospitals’ HAC reduction mechanisms, is needed.

Our findings are instructive to policymakers seeking to improve patient safety. CMS should continue to invest in robust systems to capture valid HAC measures and externally audit these data in order to ensure their integrity for value-based payment programs.

Full Papers: