By Sarianne Gruber
Twitter: @subtleimpact
Reducing readmission is moving beyond the clinical metrics and starting to focus on health insurance coverage and patients’ health behaviors prior to hospitalization. I came across a research study “The Impact of Patient Health Insurance Coverage and Latent Health Status on Hospital Readmissions” by Sezgin Ayabakan from the University of Baltimore, Baltimore, Maryland and Indranil R. Bardhan and Zhiqiang (Eric) Zheng from the University of Texas, Richardson, Texas published in the Management Information Systems Research Center (MISRC) in December, 2014. What makes this study unprecedented is that they concentrate on the effect of two non-clinical factors (1) the change in a patient’s insurance policy and (2) unobserved patient health status on readmissions. Their research brings into debate whether readmission rates are valid measures of quality of care delivered at hospitals. The study monitored Congestive Heart Failure patients across 68 hospitals in North Texas from 2005 to 2011 with 305 patients whose insurance changed from private to Medicare and 676 patients whose insurance changed from self pay to Medicare.
The research team defined unobserved patient health status as unhealthy lifestyle, alcohol or drug abuse, medical factors such as lack of access to outpatient facilities and primary care providers or social supports such as family or friends. Of note, these are unobservable because providers and hospitals do not include these health statuses in discharge claims. As for readmission behavior with regard to insurance status, the team hypothesis was that lowered financial liability would increase the number of readmissions.
The results revealed that when patients switched their insurance policies from private to Medicare, the odds of readmission increases by 79% compared to patients who stayed on private insurance, and patients who switched from selfpay to Medicare had a 97% increase in the odds of readmission. They also ran a survival analysis on each group which validated their findings. In addition, they modeled unobserved patient health as a latent state using a hidden Markov Model, and were able to differentiate a bad and a good health status. A comparison of mean readmission rates showed that patients in bad health status are readmitted significantly higher than patients in a good health status.
The authors comment of the impact of their research questioning whether “policy makers should consider non-clinical factors causally linking to patients’ readmission risk” and suggest that “hospitals alone do not seem to be the only source of accountability when analyzing the outcomes of patients”. The study also suggests that with Medicaid expansion, the population of people with incomes at or below poverty level would add about 16 million persons to the insurance pool, and, in effect, increase the readmission rates. In addition, the expansion of ACOs may become a plausible system for hospitals and providers to improve upon healthy behaviors with populations and redirect the question whether readmission is a “hospital quality metric”. On a final note, the authors comment on the need for future research, particularly on the impact of telemedicine on patients’ health status and readmissions. I would suggest reading this insightful and enlightening study.