Reducing hospital readmissions
with data science


Hospital readmission is one the major drivers of high health-care cost in the US, and it disproportionately affects disadvantaged communities. Using data science techniques, we can identifying patients with high risk of rehospitalization. Ultimately, with pro-active interventions, we can reduce hospital readmission rates.



The United States spends more on healthcare than any other country, on a per capita basis 1. Unfortunately, there are still huge gaps in quality of care. One specific and measurable gap is the patient readmission rate: the proportion of patients who end up back in the hospital, which stands at 15% for Medicare patients nationwide 2. The readmission of patients in the Medicare system alone costs taxpayers more than $26 billion annually, and $17 billion is considered avoidable 3 . The problem is most acute for socioeconomically disadvantaged patients and those requiring complex follow-up care 4. That’s why Bayes Impact is working to identify patients most at risk of returning to the hospital, in order to match them to appropriate follow-up care.

Reducing readmissions requires evidence-based practices, and data-driven solutions, to understand the root causes of individual patients’ readmissions, and identify personalized interventions to proactively address them. This is a problem where the power of data science to transform social services becomes clear. In recent years, many academic studies have attempted to model patient readmission risk, but few have applied the advanced machine learning techniques now available. And rarely have the resulting models been fully integrated into hospitals’ technology stack (their electronic health records system), or their care management workflow (a holistic, team-based and patient-centered approach to assist patients).

Bayes Impact has partnered with Sutter Health , one of California’s largest hospital systems, to take on this challenge with the goal of helping people at scale. We are conducting cohort studies and constructing machine learning models to predict patients’ readmission risk at an individual level. The objective is to build software based on cutting edge risk prediction models that integrates seamlessly into our partner hospitals’ EHR and care management procedures. Funded by a grant from Robert Wood Johnson Foundation , we are committed to provide our final product as an open source tool. The eventual aim is to provide hospitals of different sizes with an advanced, adaptive, and integrated care management tool, which can improve the quality of care for millions of people across the country.