As healthcare reform moves toward value based care, hospitals must reduce costs. As a first step, here we developed a predictive model to identify high-cost patients on admission.Methods:
We performed a retrospective observational study of 7,571 adults admitted to internal medicine services from July 1, 2013 to June 30, 2014. We compared the top 10% highest cost patients to other patients (controls) and identified clinical variables associated with high inpatient costs. Using logistic regression analyses, we developed a predictive model that could be used on admission to identify potential high utilization patients.Results:
In the 757 high utilizer patients, the median total hospital cost was $53,430 ± 60,679 compared to $8,431 ± 7,245 in the control group (P < 0.0001). The median length of stay for high utilization patients was 19.5 ± 32.5 days compared to 3.8 ± 3.9 days in the control group (P < 0.001). Variables associated with high utilization included transfer from an outside hospital (odds ratio [OR] = 1.6), admission to the pulmonary or medical intensive care unit (OR = 2.4), admission to cardiology (OR = 1.8), coagulopathy (OR = 2.6) and fluid and electrolyte disorders (OR = 2.1). A multivariate logistic regression model was used to fit a predictive model for high utilizers. The receiver operating characteristics curve of this prediction model yielded an area under the curve of 0.80.Conclusions:
High resource utilization patients appear to have a specific phenotype that can be predicted with commonly available clinical variables. Our predictive formula holds promise as a tool that may help ultimately reduce hospital costs.