In the era of value-based care, caregivers and policymakers alike are intensely interested in strategies to reduce 30-day hospital readmissions. Researchers continue to offer up helpful tools in this effort. Recently published online ahead of print in Medical Care, Burke and colleagues make an important contribution with their article The Hospital Score Predicts Potentially Preventable 30-Day Readmissions in Conditions Targeted by the Hospital Readmissions Reduction Program.
In a follow-up to 2013 work deriving and validating the HOSPITAL score, a prediction model to assess a patient’s risk for avoidable readmission, Burke and colleagues report on how the score holds up when evaluating patients admitted for heart attack (MI), chronic obstructive pulmonary disease (COPD), pneumonia (PNA), and congestive heart failure (CHF). These four conditions are of particular interest because they are featured in the Center of Medicare and Medicaid Services (CMS) Hospital Readmissions Reduction Program (HRRP). The HRRP penalizes hospitals that have higher than average rates of readmissions for any of those four diagnoses. Hospitals across the US paid fines relating to readmissions totaling $420 million in fiscal year 2016, so this is both a financially and clinically important area for improving care.
The HOSPITAL score takes into account seven factors ranging from extent of anemia to number of admissions in the last 12 months and assigns points to each patient. Depending on the final score, the tool can stratify a patient’s risk of a potentially preventable 30-day readmission into low (5%), medium (10%), or high (20%).
For the most recent article, the researchers analyzed a retrospective cohort of medical patients from 6 hospitals across the United States and identified 9,181 discharges during the study period. The researchers used a multivariable logistic regression model including the HOSPITAL score and fixed-effects to account for differences across the 6 sites. Overall, the HOSPITAL score had very good accuracy, discrimination, and calibration for potentially preventable 30-day readmissions after inpatient treatment for each of the HRRP conditions. The score did similarly well even when restricting the sample to patients older than 65, who are typically sicker, more complex, and sometimes behave differently in models of clinical scores.
As a primary care provider who strives to keep her patients out of the hospital, I can attest that having such a tool to help determine which discharges are highest-risk is very helpful – and all the more so for these specific HRRP diagnoses related readmissions that I know management are keen on minimizing. The score may also facilitate implementation of team-based practice wide disease-specific interventions for high-risk patients, such as intensive coaching and nutritional support for lifestyle-driven diagnoses such as CHF or an MI.
One downside of the tool is that, as you can see in Table 1, none of the patient characteristics included in the score are easily modifiable – for example, there’s really nothing I can do as the primary care provider if the patient had a procedure or was admitted to the oncology service. Although, I suppose that this is at the very heart of the thorny problem of preventable re-admissions – patients who are discharged and then bounce back are by definition very acutely ill, or actually may not be that ill but have very poor social supports and are constantly in and out of the hospital, which can easily bump a patient into the highest risk category. Regardless, given that all primary care practices and hospital discharge planners have limited time and resources, this tool can allow us to rationally triage our patients coming in for hospital follow up in terms of urgency and intensity of support they will likely require.
Audrey Provenzano