The utility of hospital administrative data for generating a screening program to predict adverse outcomes

A system to predict which patients will suffer medical complications or poor financial outcomes during a hospitalization would be very useful to providers of medical care. To develop such a system, we applied two previously developed indices that predict in-hospital complications to all 321,558 adult patients discharged from our hospital network. The indices identified 26,377 patients (8.2%) who experienced one or more medical complications. For these patients, high-risk admitting diagnoses were identified. We tabulated 4235 admitting diagnoses and focused on 26 (0.6%) diagnoses that were high-risk and high-volume for complications. We found that 25% of patients with these admitting diagnoses experienced complications during hospitalization. Prevention of these complications could have saved 1241 hospital days, 11 lives, and $10.5 million. Administrative data available at the time of admission can be useful in identifying the small subset of patients who are likely to experience adverse clinical outcomes during a hospitalization and those who are likely to generate adverse financial outcomes for the hospital.