PREDICTING PSYCHIATRIC EMERGENCY ADMISSIONS AND HOSPITAL OUTCOME

OBJECTIVES. A decision support tool for psychiatric hospital admission s was developed and validated to provide reliable, clinically relevant information to providers and case managers. METHODS. Using the Severi ty of Psychiatric Illness rating system, an empirical model of psychia tric emergency decision-making was constructed and validated on a spil t sample of 254 crisis cases. RESULTS. Three dimensions of the Severit y of Psychiatric Illness system-Suicide Potential, Danger to Others, a nd Severity of Symptoms-were used to construct a model that successful ly predicted 73% of decisions about level of care (inpatient or outpat ient). Clear misses, patients with a 0.20 probability of being hospita lized who were admitted, and patients with an 0.80 probability or grea ter of being hospitalized who were not admitted were reviewed to allow for utilization review. This decision support tool then was validated by predicting hospital outcomes in two additional samples. First, a r andom sample of consecutive admissions to a not-for-profit psychiatric hospital were studied. Second, a panel of admissions from a large man aged care firm were evaluated. CONCLUSIONS. Results demonstrate that t he decision to hospitalize patients in psychiatric hospitals is ration al and that models predicting admission also can predict in-hospital o utcomes.