Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Healthcare systems around the world are striving to deliver high quality care while controlling costs. One compelling strategy is the use of penalties for low-value care.1 ,2 The US federal government has made significant efforts to shift towards value-based payments for hospitals by introducing three national pay-for-performance (P4P) schemes which employ penalties: Hospital Readmission Reduction Program (HRRP), Hospital Value-Based-Purchasing (VBP) and, more recently, Hospital-Acquired Condition Reduction (HACR) Program. HRRP penalises hospitals with higher-than-expected readmissions; VBP adjusts hospital payments (either a bonus or penalty) based on performance on clinical measures and patient experience and HACR penalises the worst quartile of hospitals on HAC metrics.3 Fiscal year 2015 marks the first time hospitals may be penalised by all three programmes, with Medicare reimbursement rates potentially cut by 5.5%. Although prior work has raised concerns that hospitals serving medically complex or socioeconomically vulnerable populations are at higher risk for penalties by individual programmes,4–7 to our knowledge, there is no study that has examined the characteristics of hospitals that received the most substantial penalties across all three programmes. As more countries move towards national penalties to incentivise quality,1 it is important for policymakers to understand if there are certain hospital characteristics that make them fare worse.
Using publicly available HRRP, VBP and HACR files,8 we ranked eligible acute care hospitals by the size of the combined penalty in fiscal year 2015. We categorised hospitals into quartiles and defined three groups: most penalised (quartile with highest combined penalty), moderately penalised (quartiles 2 and 3) and least penalised (quartile of hospitals with the smallest combined penalty). Maryland and critical access hospitals, which are generally small rural hospitals, were excluded since they were ineligible for at least one programme. We used the American Hospital Association Annual Survey to identify hospital characteristics and the Centers for Medicare and Medicaid Services (CMS) Impact File9 to define safety-net hospitals (SNHs, hospitals with highest quartile of disproportionate share hospital index, which reflects proportion of care provided to patients of lower socioeconomic status). We compared the penalty groups by size, teaching, safety-net status, region, rural-urban commuting area (RUCA), whether they have a medical intensive care unit (ICU) and ownership status. We used multinomial logistic regression analyses to calculate the odds of being the most penalised or moderately penalised versus the least penalised group.
We had complete data on 3052 hospitals (table 1). Large hospitals were more likely to be in the most penalised group than in the least penalised group (19.8% vs 7.7%) as were major teaching hospitals (14.0% vs 3.4%). Similarly, SNHs were twice as likely to be in the most penalised group compared with the least penalised group (32.8% vs 16.9%). The opposite was true for small, non-teaching and non-SNHs.
In multivariable analyses, the adjusted odds of being the most penalised for large hospitals was 3.71 (95% CI 2.23 to 6.17), major teaching was 2.17 (95% CI 1.23 to 3.83) and SNHs was 1.96 (95% CI 1.46 to 2.63) (table 2). Large hospitals and SNHs were also generally more likely to be moderately penalised group compared with small, non-teaching and non-SNHs. Hospitals with a medical ICU were also more likely to be penalised while non-profit hospitals were less likely to be so. These patterns of penalties across hospitals were similar when we examined risk of penalty by individual programmes (see online supplementary appendix table 1).
We found that large hospitals, major teaching and safety-net hospitals were far more likely to be penalised the most by all three national pay-for-performance programmes in the USA. As the intent of these programmes is to penalise low-value healthcare, the current message from US policymakers suggests that quality of care is worse at these types of hospitals compared with small, non-teaching hospitals.
Although it is possible that these hospitals provide poorer quality of care, recent evidence suggests that much of the observed difference in outcomes like readmissions between high performers and low performers is driven by factors largely outside of a hospitals’ control.7 These include clinical and social factors like patient's marital status, education level, annual income and patient's baseline functional status.7 ,10 ,11 Critics have therefore called into question the extent to which these P4P programmes are targeting variations in quality of care versus variations in patient case-mix.12 ,13 Consequently, multiple organisations have called for legislation that better accounts for appropriate case-mix adjustment to help isolate between hospital variation that actually reflects quality of care.14
Our study has limitations. There is no universal approach to identify the hospitals that care for the sickest or poorest patients, although our approach has been widely used.4 ,6 Furthermore, the degree to which profit status is associated with medical complexity and socioeconomic vulnerability after accounting for size, teaching status and SNH is unclear. We also did not have patient-level data to adjust for case-mix severity; however, we used publicly available hospital-level data already risk-adjusted by CMS.15
These findings hold important lessons for policymakers regarding adequate risk-adjustment to protect hospitals from being potentially penalised for simply taking care of more poor, complex patients.
Contributors DEW completed statistical analysis and drafting of the manuscript. JF and AJ provided supervision and guidance of the concept.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.