Statistical Analyses

For aim 1, we used Cox proportional hazards regression, with age in years as the underlying time scale, to estimate the association of TS or CTD with assault and criminal conviction outcomes. We compared the risk of any experienced assault and, separately, sexual and nonsexual violent assault, as well as violent and nonviolent criminal convictions, in individuals with TS or CTD with that of unaffected individuals in the general population. We present estimates both for the full sample and separately for men and women. For aim 2, we repeated the main analyses excluding different groups of comorbid psychiatric disorders one at a time to establish whether the association of TS or CTD with the experience of violent assault and convictions for criminal offenses was explained by psychiatric comorbidities.

For aim 3, we investigated the association of TS or CTD with the experience of assault and criminal convictions in a subsample of full siblings within the cohort, identified as individuals who share both biological parents. We used Cox proportional hazards regression models stratified by sibling clusters, which rule out all factors constant within full siblings (ie, on average 50% of genetic factors and all shared environmental influences, eg, parental socioeconomic status). Such stratified models use information from clusters with discordance in the exposure to estimate the associations of TS or CTD with the assault and criminality outcomes within families. We excluded monozygotic twins and twins with unknown zygosity from the analyses.

To ensure that the observed associations were not biased owing to lack of data coverage (eg, exposure/outcome occurring before the start of the follow-up), we repeated the analyses for aim 1 in a subcohort of individuals born in Sweden after 1972 (ie, those who had complete follow-up from birth). Within this cohort, we also estimated the cumulative incidence of any assault experience and crime convictions for individuals with and without TS or CTD using Kaplan-Meier survival estimates under the assumption of no competing risks (estimated as 1 minus the Kaplan-Meier estimate of survival function). To investigate the potential overlap in assault experience and violent offenses, we calculated the proportion of those who experienced violent assault (both sexual and nonsexual) with those who also had a violent crime conviction, and vice versa, among people with TS or CTD and in the unaffected general population.

Because the National Patient Register does not contain information on TS or CTD severity, we repeated the analyses for aim 1 in a subcohort of individuals who had been diagnosed with TS or CTD at least twice after the age of 18 years, used as a proxy for severity and long-term persistence of TS or CTD into adulthood, and compared them with individuals without a TS or CTD diagnosis. This analysis excluded individuals diagnosed with TS or CTD who were no longer seen by a specialist after age 18 years as well as those who were evaluated only once after age 18 years. All models were adjusted for sex and birth year. Data management and analyses were performed between September 1 and October 22, 2021, using SAS, version 9.4 (SAS Institute) and Stata, version 15.1 (StataCorp), respectively.