Methods: Pennsylvania Inpatient Discharge data and American Hospital Association data from 2006 were obtained. The sample consisted of the universe of 46,211 adults with serious mental illness (SMI) discharged from 111 inpatient facilities in Pennsylvania. A diagnosis of Schizophrenia or Affective Disorders using ICD-9 codes was obtained from the discharge records. All individual-level characteristics (e.g., age, sex, race, primary payer type, daily cost, etc.) were aggregated into facility-level measures in order to construct facility values (e.g., % of female, % of White, etc.). Along with the aggregated individual-level characteristics, facility-level characteristics included facility type, % of SMI, total bed count, psychiatric bed count, the total number of discharges, the number of full-time personnel, HMO status, and occupancy rate. Regional-level characteristics included population, community mental health expenditures, residential share of the mental health budget, and poverty rate. To account for the multi-level structure of the data that facilities (Level 1) are nested within regions (Level 2), hierarchical linear modeling (HLM) was employed.
Results: Patients had on average 11 days (SD = 6.4) of hospitalization. People with schizophrenia stayed longer than people with affective disorders (15.5 days vs. 8.7 days, respectively). Results from HLM showed that facilities having a greater number of discharges and Medicare recipients had on average longer lengths of stay (p<.001). Furthermore, facilities with a higher recidivism rate had a longer length of hospitalization (p<.01). Regions with larger populations and higher poverty levels had longer lengths of stay among people with SMI (p<.05).
Conclusions and Implications: Length of hospital stays varies considerably among people with SMI and is more associated with facility and regional characteristics than patient characteristics. Practice patterns of hospitals as well as alternative community resources for discharging patients are likely to be relevant factors that need to be explored in future models.