Hospitalization Patterns over 30 Years Across a Statewide System of Public Mental Health Hospitals: Readmission Predictors, Optimal Follow-Up Period, Readmission Clusters and Individuals with Statistically Significant High Healthcare Utilization

Abstract

Four related hospital utilization questions (optimal follow-up period, predictors of readmission, definition of individuals with statistically significant high healthcare utilization, and patterns of readmissions) were examined using data for 491,094 hospital discharges for 250,091 patients across a statewide public mental health hospital system for 30 years (1987 to 2016). Using survival analysis, the first quartile of the survival time, the time when 25% of the entire population of discharges had a readmission was 229 days. Using observed readmissions, rather than the population as in survival analysis, revealed that 50% of all observed readmissions occurred by 222 days. Both suggest that using a one year observation period for determining high utilization may be reasonable. Major predictors of readmission were diagnoses of schizophrenia (OR = 2.11) or bipolar disorder (OR = 1.57) as well as total number of previous discharges (OR = 1.23). Statistically significant z scores (p < .01) were used to determine annual (3 or more discharges) and lifetime (7 or more discharges) criteria for individuals with statistically significant high healthcare utilization that were somewhat lower than in previous research. Cluster analysis of all readmissions revealed four relatively distinct clusters of patients: short stay-quick readmission, extremely long stay, long time in community between readmissions and frequent readmissions. While no cluster corresponded exactly with the annual statistically significant high healthcare utilization criteria, the frequent readmission cluster was somewhat similar to the lifetime statistically significant high healthcare utilization criteria with 46% of this cluster's patients having 7 or more discharges.

Keywords: High healthcare utilization; Predictors of hospital readmission; Psychiatric hospital readmission; Readmission patient clusters.