Mortality and causes of death among violent offenders and victims-a Swedish population based longitudinal study

From 1 July 1969 to 30 June 1970, 50465 men 18-20 years were conscripted for military service in Sweden. Only a small part (2-3%) was exempted from the military service mainly because of a congenital psychiatric or physical disorder or severe handicap. Most of the conscripts were born in 1949 (6%), 1950 (18%) and 1951 (75%) and in order to get as homogenous group as possible according to age, we included 48834 conscripts born 1949-51 and followed them until 2004. Their mean age at the end of the follow-up was about 53 years of age.

Each participant completed two non-anonymous questionnaires. The first included questions about family and social background, behavior, and health and the second included questions about cigarette smoking, and alcohol and drug use. The questionnaires have been found to have sufficient validity for epidemiological studies [23‐26]. Each participant was seen by a trained military psychologist for a structured interview and assessed on a series of variables based on questionnaire and interview data. The measures were assessed on a 1-5-point Likert scale (5 being the highest), yielding a distribution corresponding to 1 = 7, 2 = 24, 3 = 38, 4 = 25 and 5 = 7%. The ratings of the psychologists were regularly checked for inter-rater reliability in order to check good quality [26]. The conscript was referred to a psychiatrist if any psychiatric disorder was suspected, or reported by the conscript. An eventual diagnosis was coded according to ICD-8 [25, 26].

Based on items from the questionnaires and earlier studies of risk factors for violence and mortality, we included 19 variables associated with both lifetime violent offending and mortality. After bivariate Cox proportional regression analysis, we selected ten significant variables as confounders for multivariate analyses. Variables included were: parents' divorce (yes vs. no), any previous contact with the police or juvenile authorities (yes vs. no), conduct problems at school (yes vs. no), ever run away from home (yes vs. no) and smoking (≥ 10 cigarettes/day vs. < 10 cigarettes/day or none). Also included were problem drinking (yes vs. no, with 'yes' defined as consumption of ≥ 210 g pure alcohol per week, being intoxicated often, and/or having been apprehended for public drunkenness on at least one occasion) and drug misuse, defined as having used illicit drugs 10 times or more or having used drugs intravenously one or more times vs. less than ten times or no use.

Included psychological variables were emotional control (a summary of assessed mental stability, emotional capacity, and tolerance to stress and frustration) and social maturity (feeling responsibility for activities by other persons, and having a sense of independence, and a degree of social extroversion). The two variables were assessed on the five point scale, where 1 = very bad; 2 = bad; 3 = moderate; 4 = good; 5 = very good, with the categorizations emotional control 1-2 vs. 3-5, and social maturity 1-2 vs. 3-5. Psychiatric diagnosis at conscription was defined as: yes (one or more diagnoses) vs. no.

The cohort was followed through official registers and was linked at Statistics Sweden via the unique personal number for each individual in the cohort, which then was replaced with an individual serial number making the data anonymous to the research group, after approval of Karolinska Institute Research Ethics Committee (Dnr 2007/174-31, Dnr 2008/1086-31/5).

Data from the National Hospital Register were used to identify inpatient care with violence, alcohol and drug use according to ICD-8 and ICD-9 from 1987 and ICD-10 from 1997 onwards.

The National Hospital Register includes details of inpatient care, and has covered all public hospitals in Stockholm and Uppsala County since 1972, 85% of all Swedish public inpatient care stays since 1983, and about 98-99% of such care stays since 1987. The ICD classifications for hospitalization and mortality were: Violence - ICD-8 and ICD-9: E960-E969; ICD-10: X85-X99, Y00-Y09. Drug misuse - ICD-8: 304 and 965.0; ICD-9: 304, 965A, 968 F, 969 G and 969H; and ICD-10: F11-12, F14, F15, F16, F18, F19, O35.5, P04.4, T40.0-T40.3, T40.5-T40.9, T43.6, Z71.5, and X42. Alcohol misuse - ICD-8: 291, 303, 571.00, 571.01 and 980; ICD-9: 291, 303, 305A, 357 F, 425 F, 535D, 571A-571D and 980; and ICD-10: E24.4, F10, G31.2, G62.1, G72.1, I42.6, K29.2, K70, K86.0, O35.4, P04.3, Q86.0, T51, X45, Y91, Z50.2 and Z71.4. Psychiatric disorders (except alcohol and drug diagnoses): ICD8, and ICD9 290-319 and ICD10: F00-F69.

Mortality data came from the Cause of Death Register, which covers more than 99% of all deaths occurring in Sweden and is based on information from death certificates. Underlying causes of death are classified according to ICD-8, ICD-9 and ICD-10 as for the hospital register. One underlying cause of death is given on each death certificate, although contributing causes can be added.

Data from the national Crime Register were used to identify date, type, and number of criminal offense. The Crime Register contains information on all convictions in Sweden from 1966 onwards. Serious violence was defined as: homicide, manslaughter and assault. Violent offense were categorized as follows: conviction for violence (at least one vs. none) and violent recidivism (two or more convictions vs. none).

Cox proportional bivariate and multivariate regression analyses were used to calculate the hazard ratios (HRs) in the three groups of violence and all other conscripts for mortality and causes of death from 1 January 1970 up to the time of death or to the end of the observation period (December 31, 2004). The person times were calculated for all persons as mentioned above, including those with violent crime episodes and those with treatment episodes of injuries caused by violence during the time period. The HR:s for mortality could be somewhat lower than the HR:s in the tables because one or more violent convictions or hospitalizations were not usually reached at 1970, the starting point of the time period.

In order to analyze the impact of violence offence, victimization or both on mortality and causes of death, we performed multivariate Cox regression analyses adjusted for early risk factors including alcohol and drug misuse in relation to the outcomes. The early covariates are measured at the time of conscription and the exposure variables (violence, victimization and both) occurred from 1970 and onwards (see Figure 1). In the multivariate models we included only those variables which in the bivariate analyses were significant. We used SAS, version 9.2 (SAS Institute Inc., Cary, NC, USA).

The most frequent underlying or contributing cause of death was alcohol-related in all three violent groups (34.7-38.4%, compared to 13.7% in the other conscripts). Even though proportional mortality from suicide was quite similar in all four groups (Table 1), suicide mortality was higher in the violent offender group (2.6%), in victims of violence (2.2%), and in the group of both offenders and victims (3%) compared to the other conscripts (0.9%).

The bivariate analyses showed that the recidivists of violent offences had nearly 11 times higher hazard of dying with an alcohol or drug-related cause and nearly four times higher hazard of suicide compared to the other conscripts, while the victims had six and twofold higher hazards of the same outcomes (Table 2). Death with an alcohol or drug-related cause, and also other causes was high among participants who were both violent offenders and victims: HRs = 10.51 and 6.15, respectively.

Multivariate analyses showed high hazards (HR = 2.04 to HR = 3.61) of mortality due to alcohol and drug-related causes in all three groups (Table 3). Moreover, violent recidivists had more than threefold higher hazard of dying from an alcohol or drug related diagnosis and twofold higher hazards of dying from suicide, while no significant risk was found for accidents, neoplasm, other causes and circulatory disease. Having experiences of both violent offense and victimization was especially associated with mortality from other causes (HR = 4.56, 95% CI 2.19-9.47) (Table 3).

In Table 4, we stratified for alcohol, drug misuse, and psychiatric inpatient care during the follow up period. Violent offenders with alcohol or illicit drug diagnoses had nearly fourfold higher hazard to die compared to violent offenders with no such diagnoses. The hazards were much higher when comparing with all other participants: HR = 5.53 and HR = 7.67, respectively. Victims with alcohol-related diagnoses had a nearly fourfold elevated hazard of dying when compared to victims with no alcohol diagnosis, and a more than six fold higher hazard when compared to all the others. Likewise, participants with experiences of both violent victimization and offending and additionally alcohol or drug misuse diagnoses were associated with a nearly six and eightfold higher hazard of dying compared to the others.

Nearly 7% of the total cohort had been treated for a psychiatric disorder. More than one-fifth of the violent offenders and victims also had a psychiatric disorder, which led to higher hazards of dying compared to those without psychiatric disease: HR = 3.80 and HR = 2.54, and the hazards were additionally about twofold higher compared to all others in the cohort.

Considering mortality due to suicide, we also controlled for psychiatric inpatient care alone, and psychiatric inpatient care in combination with alcohol and substance misuse. We found that especially repeat violent offense were associated with higher hazards of suicide, giving an HR of 1.94 (95% CI 1.33-2.82) and 1.71 (95% CI 1.14-2.58), respectively (not shown in tables). No such association was found among victims and individuals who were both offenders and victims.

One advantage of this nationwide, representative population study is the low non-participation rate, of about 2-3%. The reasons were mainly severe disability or congenital disorders. Use of non-anonymous questionnaires may have resulted in lower reporting of sensitive issues than may have been elicited with anonymous questionnaires. In previous studies of conscripts, although limited to Stockholm, it was found that a large proportion of those participants who had not completed the questionnaire had higher rates of criminality in adulthood. This means that the prevalence of, e.g., drug use may have been underestimated, but it does not mean that the association between the variables is biased [47]. Furthermore, persons with antisocial or paranoid tendencies may be especially prone to deny their use of illicit drugs in non-anonymous questionnaires. Further, we do not know whether the participants had stopped or changed their substance use behavior after conscription, however alcohol or drug-related diagnoses derived from inpatient care registers can be considered a rather more robust measure of alcohol or drug use.

Another limitation of this study is, as has been mentioned above, that we were only able to include victims who had been hospitalized for injuries and we had no information about the perpetrators who had caused the injury. Likewise, information about crimes that had not come to the attention of the police is lacking.