Background Compulsory mental health care aims to manage risk and treatment needs among people with severe mental illness, but evidence on long-term outcomes across health, justice, and socioeconomic domains is scarce. We aimed to compare long-term outcomes across adult groups defined by community treatment orders (CTOs), compulsory hospitalisation, compulsory observation, and voluntary mental health care, and the general population. Methods In this nationwide registry cohort study, we identified all individuals aged 18-65 years with specialist mental health-care contact in Norway from Jan 1, 2015, to Dec 31, 2016 and general population comparators sampled via Statistics Norway based on age and sex. Follow-up was through 2021-23, depending on registry availability. Study groups were defined hierarchically: CTO, compulsory hospitalisation without CTO, compulsory observation without compulsory hospitalisation or CTO, voluntary mental health care, and no specialist mental health-care contact (ie, the general population). Outcomes included mortality and cause of death, self-harm treated in emergency departments or hospital, substance use disorder, victimisation, violent or sexual offence charges or imprisonment, subsequent compulsory care, acute admissions, and socioeconomic outcomes. Data for all groups were sourced from linked nationwide registries. Separate binomial regressions estimated sex-specific probabilities by group, adjusting for birth year. People with lived experience of mental health services advised study design and interpretation. Findings The final study sample comprised 321 708 individuals (mean age 36 & centerdot;5 years, range 18-65) across five mutually exclusive groups: CTO (n=3331), compulsory hospitalisation (n=2907), compulsory observation (n=2554), voluntary care (n=177 396), and general population (n=135 520), 193 473 (60 & centerdot;1%) were female and 128 235 (39 & centerdot;9%) were male; ethnicity data were unavailable. Compulsory care groups were associated with substantially worse long-term outcomes across health, criminal justice, and socioeconomic domains than voluntary patients and the general population. Across compulsory care groups, female participants had higher adjusted probabilities of self-harm or suicide than male participants, whereas male participants had higher adjusted probabilities of substance use disorder and of being charged with or imprisoned for violent or sexual offences compared with female participants. The compulsory observation group had the highest adjusted probabilities of self-harm or suicide and substance use disorder across the five groups. Self-harm or suicide was recorded in 135 (12 & centerdot;1%) of 1115 female participants (adjusted probability 12% [95% CI 10-14]) and 75 (5 & centerdot;7%) of 1308 male participants (5 & centerdot;5% [4 & centerdot;4-6 & centerdot;9]). Substance use disorder was recorded in 407 (39 & centerdot;1%) of 1041 female participants (39% [36-42]) and 639 (53 & centerdot;2%) of 1202 male participants (53% [50-56]). Charges or imprisonment for violent or sexual offences were recorded in 146 (14 & centerdot;2%) of 1030 female participants (14% [12-16]) and 446 (37 & centerdot;4% [34-40]) of 1194 male participants. During 2015-21, including the inclusion period, suicide and other external causes accounted for 132 (55 & centerdot;7%) of 237 classified deaths in compulsory observation, compared with 1818 (38 & centerdot;2%) of 4754 in voluntary care and 174 (18 & centerdot;6%) of 938 in the general population. The CTO group had the highest adjusted probability of disability benefit receipt: 89% (95% CI 87-91) in female participants and 93% (92-94) in male participants; a substantial minority of disability-benefit recipients in the CTO group also had earned income. Interpretation Compulsory care groups were associated with distinct long-term prognoses from the voluntary care and general population groups. These findings provide population-level benchmarks for service planning and policy evaluation and show that legally defined compulsory care groups should not be treated as a single prognostic population.
