Reducing Excess Readmissions: Promising Effect of Hospital Readmissions Reduction Program in US Hospitals

Ning Lu; Kuo-Cherh Huang; James A. Johnson

Disclosures

Int J Qual Health Care. 2016;28(1):53-58. 

In This Article

Methods

Sources of Data

The readmission ratios data used in this study came from the publicly available HRRP Supplemental Data Files for the fiscal years 2013, 2014 and 2015. The data files contain readmission ratios for each applicable condition, measures of a hospital's readmission performance for that condition. CMS determined for each eligible hospital the readmission ratios for PN, AMI and HF using discharges that occurred from July 2008 through June 2011 for FY 2013, July 2009 to June 2012 for FY 2014 and July 2010 to June 2013 for FY 2015.[9]

We linked readmission ratios data to the 2009 American Hospital Association (AHA) annual survey to obtain information on hospital characteristics. This included hospital size measured by the number of hospital beds (small hospitals with <200 beds, medium-sized hospitals with 200–399 beds and large hospitals with ≥400 beds); teaching status (teaching hospitals with a membership in the Council of Teaching Hospitals or non-teaching hospitals); type of ownership of a hospital (public, for-profit or not-for-profit); location (metropolitan: core urban area of population ≥50 000, micropolitan: core urban area of population ≥10 000 to <50 000 population or rural, based on OMB's CBSA classifications[15]); percentage of patients with Medicaid admitted and percentage of patients with Medicare admitted. We also used the disproportionate share hospital index that CMS utilizes to quantify hospital care provided to low-income and medically vulnerable populations to identify a hospital as a safety-net hospital or not.[16] Hospitals in the highest quartile of the disproportionate share hospital index are categorized as safety-net hospitals in our study.

Study Population

We included a total of 3395 IPPS nonfederal, short-term, acute care hospitals that are subject to the HRRP program and that reported discharge data to the HRRP for PN, AMI and HF for the calculations of readmission ratios for the fiscal years 2013, 2014 and 2015. Hospitals located in Puerto Rico were excluded from our study.

Statistical Analysis

We first conducted univariate analyses to provide a profile of the hospitals used in our study and examined whether excess hospital readmissions differed by hospital characteristics. We then performed a series of repeated-measures analysis of variance models to examine the effect of HRRP on hospital readmissions for PN, AMI and HF. The primary goal of our analyses was to determine whether the HRRP had an effect on reducing inpatient readmissions for PN, AMI and HF between FY 2013 (before the execution of up to 1% reduction in Medicare base payment) and FY 2014 (after the execution of up to 1% reduction in Medicare payment), and FY 2015 (after the execution of up to 2% reduction in Medicare payment). We focused our analyses on hospitals with excess readmissions (readmissions ratios >1) to determine the effect of HRRP on excess readmissions for each of the applicable conditions. To do so, we restricted our sample to hospitals identified as having excess readmissions in FY 2013 (n = 1457). Lastly, we assessed whether the HRRP effect on excess readmissions differed according to hospital characteristics. To address these research questions, we hypothesized:

  1. The HRRP had an effect on reducing excess readmissions for PN, AMI and HF.

  2. The effect of HRRP on reducing excess readmissions differed according to hospital characteristics. We hypothesized that the HRRP program had a greater effect on hospitals with a higher proportion of Medicare and Medicaid patients, and on safety-net hospitals.

The outcome variables used in examining the effect of HRRP are excess readmission ratios for PN, AMI and HF. Independent variables included three periods of fiscal years and hospital characteristics of size, type of ownership, location, safety-net hospitals, teaching status and percentage of patients with Medicare admitted and percentage of patients with Medicaid admitted. We included two-way interaction terms in the repeated-measures analysis of variance models to examine whether the HRRP had a differential effect on hospitals according to their characteristics. Hospitals with fewer than 25 readmission cases for a given condition were excluded from the analysis for that condition due to the fact that a readmission ratio would not be reliable and thus was not calculated by the HRRP. Statistical significance was set at P < 0.05 (two-tailed). All analyses were performed using the SAS software version 9.3 (SAS Institute, Cary, NC, USA).

processing....