Measuring quality of care in psychiatric emergencies: construction and evaluation of a Bayesian index

• Published in: Health services research 1993-06, Vol.28 (2), p.131-158
• Main Authors: Gustafson, D H, Sainfort, F, Johnson, S W, Sateia, M
• Format: Article
• Language: English
• Subjects: Bayes Theorem; Bayesian analysis; Bayesian statistical decision theory; Emergencies; Emergency medical care; Emergency Service, Hospital - standards; Emergency Services, Psychiatric - standards; Emergency Services, Psychiatric - statistics & numerical data; Evaluation; Health Services Research - methods; Hospital Bed Capacity, 300 to 499; Hospital Bed Capacity, 500 and over; Humans; Likelihood Functions; Management; Measurement; Medical care; Models, Statistical; New England; Observer Variation; Patient Compliance; Program Evaluation - methods; Psychiatric emergencies; Psychiatry; Quality of Health Care - standards; Quality of Health Care - statistics & numerical data; Quality of service; Reproducibility of Results; Studies
• ISSN: 0017-9124; 1475-6773

This study was conducted to determine whether an index for measuring quality of care for psychiatric emergencies is reliable and valid. The study used primary data collected over a 12-month period from two urban hospitals in the Northeast. One had 700 inpatient beds, an inpatient psychiatric unit, and community mental health personnel located in the emergency department. The other had 300 beds but none of the other hospital's features. The index was developed by a panel of experts in emergency psychiatry using a subjective Bayesian statistical methodology and was evaluated in terms of its ability to: (1) predict a second panel's judgments of quality; (2) predict a specific quality-related patient outcome, i.e., compliance with follow-up recommendations; (3) provide a reliable measurement procedure; and (4) detect variations in patterns of emergency department practices. Data were collected on 2,231 randomly selected emergency psychiatric patients (psychiatric diagnosis, alcohol abuse, nonverbal patients experiencing a psychiatric emergency, and patients with somatic complaints such as life crisis) treated in the emergency departments of the two hospitals. The index predicted physician judgments of quality, was reliable, exhibited sufficient variation in scores, and was strongly associated with patient compliance. The study demonstrated that a subjective Bayesian model can be used to develop a reliable and valid index for measuring quality of care, with potential for practical application in management of health services.