Skip to main content
American Journal of Epidemiology logoLink to American Journal of Epidemiology
. 2017 Nov 16;187(5):1010–1018. doi: 10.1093/aje/kwx321

Associations Between Divorce and Onset of Drug Abuse in a Swedish National Sample

Alexis C Edwards 1,, Sara Larsson Lönn 2, Jan Sundquist 2, Kenneth S Kendler 1, Kristina Sundquist 2,3
PMCID: PMC5928459  PMID: 29155917

Abstract

Rates of drug abuse are higher among divorced individuals than among those who are married, but it is not clear whether divorce itself is a risk factor for drug abuse or whether the observed association is confounded by other factors. We examined the association between divorce and onset of drug abuse in a population-based Swedish cohort born during 1965–1975 (n = 651,092) using Cox proportional hazards methods, with marital status as a time-varying covariate. Potential confounders (e.g., demographics, adolescent deviance, and family history of drug abuse) were included as covariates. Parallel analyses were conducted for widowhood and drug-abuse onset. In models with adjustments, divorce was associated with a substantial increase in risk of drug-abuse onset in both sexes (hazard ratios > 5). Co-relative analyses (among biological relatives) were consistent with a partially causal role of divorce on drug-abuse onset. Widowhood also increased risk of drug-abuse onset, although to a lesser extent. Divorce is a potent risk factor for onset of drug abuse, even after adjusting for deviant behavior in adolescence and family history of drug abuse. The somewhat less-pronounced association with widowhood, particularly among men, suggests that the magnitude of association between divorce and drug abuse may not be generalizable to the end of a relationship.

Keywords: addiction, divorce, drug abuse, family history, widowhood


Drug abuse (DA) is more common among individuals who are divorced, separated, or widowed (14). Previous research has suggested a protective effect of marriage with respect to substance use and abuse (58), but fewer studies have explored the associations with divorce. Because many large studies are limited to cross-sectional data, it is unclear whether divorce increases the risk for onset of DA or vice versa. Given the consequences of DA with respect to physical and mental health (9) and economic burden (10), it is critical that the various pathways to abuse be clarified to improve efforts at risk reduction.

While divorce has been associated with alcohol problems and other manifestations of subsequent psychopathology (11, 12) a parallel study of illicit drug problems has not, to our knowledge, been conducted. One study of the relationship between divorce and alcohol abuse in a population-based UK sample found that divorce was associated with higher scores on the CAGE alcohol screening instrument even after controlling for other key variables (13). A study of the Swedish population found that risk of onset of alcohol-use disorder remained elevated after divorce when accounting for demographic and other risk factors, including family history (14). In both studies, the potential causal role of divorce could not be excluded. Given the stress experienced after divorce and the role of stress in DA onset (15, 16), divorce may similarly affect the risk of DA.

Critically, DA risk is multifactorial. Twin studies demonstrate that genetic factors contribute substantially, with twin- and family-based heritability estimates of 0.2–0.8 across substances (1720); environmental factors such as childhood adversity (21) and adolescent peer deviance (22) also play a role. While the age of onset of drug use is typically in late adolescence or young adulthood (23, 24), the increased rate of abuse among divorcées and widowed persons raises the possibility that environmental factors occurring in adulthood may also be influential. In considering the potential causal effect of divorce, it is important to account for the influence of genetic/biological factors, which may jointly affect risk for both DA and divorce. Indeed, prior studies have demonstrated a genetic correlation between divorce and alcohol-use disorder (25).

Prior studies provide evidence of assortative mating for substance use (2628), and partner substance use has been implicated as a risk factor for one’s own substance use (29). Spousal substance abuse has been identified as a contributor to marital dissolution (30). Other evidence suggests that divorce, although stressful, may have a positive impact: Depressive symptoms increase less among those who end an unsatisfactory marriage (31), and after-divorce adjustment is modestly higher when the divorce is due to the spouse (rather than to oneself) (30). Combined, these findings suggest a complicated relationship between substance use and marriage/divorce. Few quantitative analyses have explored the role of substance use by a former spouse with respect to future DA risk: Might ending a relationship with a drug-abusing spouse reduce risk for the nonabusing partner? Clarification of previously observed associations may improve our understanding of differential risk.

In the present study, we used survival analyses and allowed for marital status as a time-dependent covariate to investigate the association between divorce and subsequent DA onset in a population-based Swedish cohort (n = 651,092). Our research questions were as follows: 1) Is risk of DA onset increased among those who divorce relative to those who remain married? 2) Does observed risk of divorce remain after controlling for demographics, family history of DA, early-onset deviance, and spousal deviance? 3) Does divorcing a deviant spouse mitigate risk conferred by divorce itself? 4) Is the association with divorce potentially causal, or might familial confounding contribute to the association? And 5) are associations with widowhood comparable to those of divorce?

METHODS

Sample

We linked nationwide Swedish registers, described below, using unique identification numbers assigned at birth or immigration, in this case replaced with serial numbers to retain anonymity, to all Swedish residents. Only Swedish-born residents were included in the present analyses. The present analyses used data from individuals born between 1965 and 1975, a time frame that maximized the availability of information across registries and allowed a meaningful length of observation during and after marriage.

We used the Total Population Register for annual information on marital status from 1968 to 2013 and family identification from 1990; the Swedish Census with household information every 5 years from 1960 to 1990; the Longitudinal Integration Database for Health Insurance and Labor Market Studies register for information on parental education from 1990 to 2012; the Multi-generation Register, with information on sex and birth and linking children with their parents; the Swedish Hospital Discharge Register, containing data on hospitalizations from 1964 to 2012; the Outpatient Care register, containing information from all outpatient clinics between 2001 and 2012; the Swedish mortality register, including dates and causes of deaths from 1952 until 2012, the Primary Health Care Register, containing data on outpatient diagnoses from 2001 to 2013; the Swedish Crime Register, which contains complete national data an all lower-court convictions from 1973 to 2012; the Swedish Suspicion Register, which includes complete data on individuals strongly suspected of crime from 1998 to 2012. The study was approved by the ethics committee in Lund, Sweden; subject consent was waived.

Measures

Outcome variables

Our primary outcome variable was a DA registration. This could be obtained through the crime or suspicion registries via a drug-related arrest or conviction or through the medical registries based on a primary or secondary diagnosis related to DA. International Classification of Diseases codes and crime registration codes related to drug use have been previously described (32). Although cultural differences exist, registrations derived from these sources can be considered comparable to medical and criminal issues related to DA in the United States.

Independent variables

The predictor of interest was first divorce. This was ascertained through the Total Population Register, which includes timing of marriage, divorce, and widowhood. We included birth year and parental education, operationalized as the highest level of education attained by either parent and categorized into 3 levels (in years of schooling: low, ≤9; mid, 10–11; high, ≥12), as covariates in all analyses. We further controlled for early-onset (prior to age 18 years) externalizing behavior, defined here as a criminal conviction for serious infractions (for example, arson, assault, fraud; described in detail in Kendler et al. (33)) or registration for alcohol-use disorder ascertained through medical, criminal, and pharmacy records (33). Family history of DA in biological parents and siblings was included as a binary variable. Finally, a spousal lifetime history of deviant behavior was determined based on registrations for DA, alcohol use disorder, or criminal behavior (as above).

Secondary analyses examined the association between first widowhood and DA registration to assess whether results observed for divorce generalized to an independent form of the loss of a spouse. These analyses were conducted in parallel fashion to the primary divorce analyses.

Statistical methods

We used Cox proportional hazards methods to estimate the increased risk of onset of DA as a function of divorce compared with during marriage, with divorce included in the model as a time-dependent covariate. We limited our analyses to first divorce. The outcome variable was time to onset of DA, and we censored at death, migration, end of divorced period (i.e., if an individual remarried), or end of follow-up (2011). Individuals with a previous DA registration were also censored because the outcome of interest was DA onset.

We first assessed the role of divorce controlling for only demographic variables (birth year and parental education; model 1). In model 2, we introduced family history of DA and early-onset externalizing behavior. We then added spousal externalizing behavior, examining both the main (model 3a) and the interaction associations with divorce (model 3b). This interaction tests whether divorcing a spouse with a history of drug or alcohol abuse or criminal behavior mitigates the negative association with divorce itself. All models were fitted using SAS, version 9.4 (SAS Institute, Inc., Cary, North Carolina).

Co-relative analysis

We used a co-relative design using same-sex cousin, half-sibling, and full-sibling pairs. Informative monozygotic twin pairs meeting our criteria were too rare for inclusion. If the relationship between a putative risk factor and an outcome is noncausal and instead due to shared genetic liability, the observed hazard ratio of the risk factor will diminish with increasing genetic correlation within relative pairs (34). We identified unique combinations of pairs and used conditional Cox regression, where each pair is treated as a stratum, meaning the associations of the covariates are estimated within each pair, thereby controlling for confounding factors shared within that pair. Because of the low numbers of informative pairs, additional confounders were not included in these analyses. Neither member of the pair had a previous abuse registration. Divorce was included as a time-dependent covariate; either one member of the pair met our criteria for divorce over the follow-up period and one did not, or they were divorced at different ages. Sexes were combined in these analyses.

RESULTS

Descriptive statistics

Descriptive data is provided in Table 1. The sample meeting our inclusion criteria consisted of 651,092 individuals (51.82% female). Of these, the first marriage ended in divorce for 145,809 individuals (22.39%). Among individuals still in their first marriage, the rate of DA was 0.38% (n = 2,452), while among divorced individuals it was 1.66% (n = 2,416), resulting in an initial hazard ratio of 7.31 (95% confidence interval (CI): 6.91, 7.74).

Table 1.

Descriptive Statistics of Swedish Sample Born 1965–1975

Measure Women Men
No. of Participants %a No. of Participants %a
Total sample 337,398 51.82 313,694 48.18
Divorce 81,829 24.25 63,980 20.40
DA registration
 Among married individuals 1,159 0.34 1,293 0.41
 Among divorced individuals 1,161 1.42 1,255 1.96
Age at marriage, yearsb 29.95 (5.91) 32.05 (5.66)
Age at divorce, yearsb 34.77 (6.21) 36.29 (5.77)

Abbreviation: DA, drug abuse.

a For the first row (total sample), % refers to the percentage of the total sample that is female or male. For subsequent rows, % refers to the percentage of women or men who are divorced or have a DA registration.

b Values are expressed as mean (standard deviation).

Sex differences

We first assessed whether the association between divorce and subsequent DA registration differed as a function of sex by estimating a term for the divorce × sex interaction. The hazard ratio for the women was estimated at 0.87 (95% CI: 0.77, 0.97) that of men, indicating that the risk associated with divorce was significantly worse for men. Accordingly, subsequent models were stratified by sex. Figures 1 and 2 depict the increased risk of DA onset as a function of marital status among women and men, respectively.

Figure 1.

Figure 1.

Prevalence of drug-abuse registrations as a function of divorce status among Swedish women born during 1965–1975. The relationship with divorce is shown as the moving yearly prevalence (a 3-year rolling average) of drug-abuse onset. Squares represent women who divorced, while circles represent the rate of drug-abuse onset of married women whose average age at time 0 matched the age of the divorced sample.

Figure 2.

Figure 2.

Prevalence of drug-abuse registrations as a function of divorce status among Swedish men born during 1965–1975. The relationship with divorce is shown as the moving yearly prevalence (a 3-year rolling average) of drug-abuse onset. Squares represent men who divorced, while circles represent the rate of drug-abuse onset of married men whose average age at time 0 matched the age of the divorced sample.

Findings in women

Results for women are presented in Table 2. Model 1 demonstrated that baseline risk (accounting only for demographic factors) of DA onset attributable to divorce is substantial (hazard ratio (HR) = 6.80). In model 2, this is mitigated 16% by controlling for early-onset deviant behavior and a family history of DA. The association is further reduced by accounting for marriage to a deviant spouse (model 3a) to hazard ratio = 5.17 (a 9% reduction); spousal deviance itself increased risk nearly 2-fold. Finally, we introduced an interaction term (model 3b) to test whether hazard ratios differed depending on whether the former spouse had a history of deviant behavior. The resulting estimate was significantly lower than 1, indicating that divorce from a deviant spouse was less risky, with respect to DA onset, than divorce from a nondeviant spouse. The hazard ratios for divorce from a nondeviant spouse versus a deviant spouse were 6.56 (95% CI: 5.90, 7.30) and 3.54 (95% CI: 3.10, 4.04), respectively.

Table 2.

Hazard Ratios From a Cox Proportional Hazards Model Examining the Association Between Divorce and Onset of Drug Abuse Among Swedish Women Born During 1965–1975

Predictor Model 1a Model 2b Model 3ac Model 3bd
HR 95% CI HR 95% CI HR 95% CI HR 95% CI
Divorce 6.80 6.25, 7.39 5.68 5.22, 6.18 5.17 4.75, 5.63 6.56 5.90, 7.30
Birth year 1.03 1.02, 1.05 1.02 1.01, 1.04 1.03 1.01, 1.04 1.03 1.01, 1.04
Parental educatione
 Mid vs. low 0.90 0.81, 1.00 0.91 0.82, 1.01 0.92 0.83, 1.03 0.93 0.83, 1.03
 High vs. low 0.82 0.73, 0.92 0.84 0.75, 0.94 0.87 0.77, 0.98 0.88 0.78, 0.98
Early-onset deviant behavior 3.81 3.38, 4.31 3.54 3.13, 4.00 3.54 3.13, 4.00
Family history of DA 3.40 3.09, 3.73 3.21 2.93, 3.53 3.21 2.92, 3.52
Spousal history of deviant behavior 1.98 1.81, 2.15 2.74 2.43, 3.09
Interaction of divorce and spousal history of deviant behavior 0.54 0.46, 0.64

Abbreviations: CI, confidence interval; DA, drug abuse; HR, hazard ratio.

a Baseline model accounting only for demographic factors.

b Model 2 included variables in model 1 and accounted for family history of DA and early-onset externalizing behavior.

c In addition to variables in model 2, model 3a accounted for spousal externalizing behavior.

d In addition to variables in model 3a, model 3b included a term for interaction between spousal externalizing behavior and divorce.

e Parental education was categorized according to years of schooling: low, ≤9; mid, 10–11; high, ≥12.

Findings in men

Controlling only for demographic factors, the hazard ratio for divorce was higher for men than for women (HR = 8.29; Table 3, model 1). This estimate was reduced by 23% by controlling for deviant behavior prior to age 18 years and family history of DA (model 2). Further controlling for a deviant former spouse reduced the hazard ratio to 5.69 (an 11% reduction, model 3a). The hazard ratio estimate of spousal deviance was higher for men than for women (HR = 2.58 vs. 1.98). We tested whether divorce-based hazard ratios differed as a function of a wife’s history of deviant behavior, and we found that they did (model 3b): The hazard ratio for divorce from a nondeviant spouse was 7.38 (95% CI: 6.73, 8.10) and from a deviant spouse only 2.64 (95% CI: 2.26, 3.08).

Table 3.

Hazard Ratios From a Cox Proportional Hazards Model Examining the Association Between Divorce and Onset of Drug Abuse Among Swedish Men Born During 1965–1975

Predictor Model 1a Model 2b Model 3ac Model 3bd
HR 95% CI HR 95% CI HR 95% CI HR 95% CI
Divorce 8.29 7.65, 8.97 6.37 5.88, 6.91 5.69 5.24, 6.17 7.38 6.73, 8.10
Birth year 1.04 1.03, 1.06 1.04 1.03, 1.06 1.04 1.03, 1.05 1.04 1.03, 1.06
Parental educatione
 Mid vs. low 0.81 0.73, 0.89 0.83 0.75, 0.92 0.85 0.77, 0.94 0.84 0.76, 0.93
 High vs. low 0.71 0.64, 0.79 0.75 0.68, 0.84 0.78 0.70, 0.87 0.79 0.71, 0.88
Early-onset deviant behavior 4.73 4.35, 5.14 4.39 4.04, 4.78 4.35 4.00, 4.73
Family history of DA 2.87 2.62, 3.14 2.73 2.50, 3.00 2.73 2.50, 2.99
Spousal history of deviant behavior 2.58 2.35, 2.83 4.52 3.99, 5.12
Interaction of divorce and spousal history of deviant behavior 0.36 0.30, 0.43

Abbreviations: CI, confidence interval; DA, drug abuse; HR, hazard ratio.

a Baseline model accounting only for demographic factors.

b Model 2 included variables in model 1 and accounted for family history of DA and early-onset externalizing behavior.

c In addition to variables in model 2, model 3a accounted for spousal externalizing behavior.

d In addition to variables in model 3a, model 3b included a term for interaction between spousal externalizing behavior and divorce.

e Parental education was categorized according to years of schooling: low, ≤9; mid, 10–11; high, ≥12.

Proportionality assumption

To address whether the risk of divorce on DA onset is consistent across age at divorce, we conducted a post-hoc test adding a term for the divorce × age interaction to model 3. Hazard ratio estimates were 0.95 (95% CI: 0.93, 0.96) per year for women and 0.97 (95% CI: 0.96, 0.99) per year for men, indicating that the association with divorce was stronger at younger ages than at older ages (Figure 3).

Figure 3.

Figure 3.

Change in drug-abuse onset risk across age at divorce among Swedish women and men born during 1965–1975. The risk of drug-abuse onset associated with divorce diminished with age of divorce. Estimated adjusted hazard ratios (95% confidence intervals) are plotted for Swedish women and men who divorced at age 30 years versus age 40 years.

Co-relative analyses

We identified pairs of cousins, half-siblings, and full siblings, to test whether the risk of divorce was reduced with increasing genetic similarity. Familial confounding factors contributing to risk for both divorce and DA are increasingly accounted for at higher degrees of genetic relatedness. Hazard ratios were consistent across groups (for cousins, HR = 4.89 (95% CI: 4.33, 5.51); for half-siblings, HR = 4.85 (95% CI: 3.24, 6.49); and for full siblings, HR = 4.85 (95% CI: 3.63, 5.79)).

Associations with widowhood

Tables 4 and 5 present findings for the associations between widowhood and DA onset for women and men, respectively. Though the sex × widowhood interaction term was not significant (HR = 0.695, 95% CI: 0.307, 1.575), we elected to conduct analyses separately according to sex to more closely parallel the analyses of divorce. For both women and men, the association between widowhood, controlling only for demographic factors, and subsequent DA onset was lower than that of divorce (HR = 5.27 among women and HR = 4.22 among men). This risk was attenuated when further controlling for early-onset deviant behavior, family history of DA, and spousal deviance, with final hazard ratios of 4.47 (for women) and 3.04 (for men). Notably, the hazard ratio of widowhood was only 53% of that for divorce among men, while for women the corresponding figure is 86%. However, in a post-hoc test directly contrasting the associations with divorce versus widowhood, these differences were not statistically significant in either women (HR = 1.33, 95% CI: 0.73, 2.41) or men (HR = 1.86, 95% CI: 0.83, 4.15), likely due to the imprecision of the widowhood estimates.

Table 4.

Hazard Ratios From a Cox Proportional Hazards Model Examining the Association Between Widowhood and Onset of Drug Abuse Among Swedish Women Born During 1965–1975

Predictor Model 1a Model 2b Model 3ac Model 3bd
HR 95% CI HR 95% CI HR 95% CI HR 95% CI
Widowed 5.27 3.42, 8.14 5.03 3.26, 7.76 4.47 2.89, 6.89 4.30 2.30, 8.05
Birth year 1.04 1.02, 1.06 1.03 1.01, 1.05 1.03 1.01, 1.06 1.03 1.01, 1.06
Parental educatione
 Mid vs. low 0.80 0.69, 0.92 0.79 0.69, 0.92 0.81 0.70, 0.94 0.81 0.70, 0.94
 High vs. low 0.65 0.55, 0.76 0.65 0.55, 0.76 0.70 0.60, 0.82 0.70 0.60, 0.82
Early-onset deviant behavior 4.90 4.05, 5.93 4.17 3.45, 5.06 4.18 3.45, 5.06
Family history of DA 1.51 1.28, 1.79 1.48 1.25, 1.75 1.48 1.25, 1.75
Spousal history of deviant behavior 2.90 2.57, 3.27 2.90 2.57, 3.27
Interaction of widowhood and spousal history of deviant behavior 1.075 0.45, 2.55

Abbreviations: CI, confidence interval; DA, drug abuse; HR, hazard ratio.

a Baseline model accounting only for demographic factors.

b Model 2 included variables in model 1 and accounted for family history of DA and early-onset externalizing behavior.

c In addition to variables in model 2, model 3a accounted for spousal externalizing behavior.

d In addition to variables in model 3a, model 3b included a term for interaction between spousal externalizing behavior and widowhood.

e Parental education was categorized according to years of schooling: low, ≤9; mid, 10–11; high, ≥12.

Table 5.

Hazard Ratios From a Cox Proportional Hazards Model Examining the Association Between Widowhood and Onset of Drug Abuse Among Swedish Men Born During 1965–1975

Predictor Model 1a Model 2b Model 3ac Model 3bd
HR 95% CI HR 95% CI HR 95% CI HR 95% CI
Widowhood 4.22 2.10, 8.46 3.44 1.72, 6.90 3.04 1.51, 6.10 3.76 1.68, 8.41
Birth year 1.07 1.05, 1.10 1.07 1.05, 1.09 1.07 1.05, 1.09 1.07 1.05, 1.09
Parental educatione
 Mid vs. low 0.72 0.63, 0.83 0.73 0.64, 0.84 0.75 0.65, 0.86 0.75 0.65, 0.86
 High vs. low 0.57 0.49, 0.66 0.61 0.53, 0.71 0.65 0.56, 0.76 0.65 0.56, 0.76
Early-onset deviant behavior 6.66 5.94, 7.47 5.76 5.13, 6.47 5.76 5.12, 6.47
Family history of DA 1.46 1.25, 1.71 1.42 1.21, 1.66 1.42 1.21, 1.66
Spousal history of deviant behavior 4.56 4.02, 5.17 4.58 4.04, 5.20
Interaction of widowhood and spousal history of deviant behavior 0.51 0.10, 2.54

Abbreviations: CI, confidence interval; DA, drug abuse; HR, hazard ratio.

a Baseline model accounting only for demographic factors.

b Model 2 included variables in model 1 and accounted for family history of DA and early-onset externalizing behavior.

c In addition to variables in model 2, model 3a accounted for spousal externalizing behavior.

d In addition to variables in model 3a, model 3b included a term for interaction between spousal externalizing behavior and widowhood.

e Parental education was categorized according to years of schooling: low, ≤9; mid, 10–11; high, ≥12.

DISCUSSION

In this study, we examined the association between divorce status and onset of DA in the Swedish population. Our analyses accounted for early-onset deviant behavior, family history of DA, and other important covariates. Although the association of divorce was mediated by these variables, results indicated that divorce is a potent risk factor, with adjusted hazard ratios greater than 5 for both women and men. Co-relative analyses suggested that divorce may causally affect risk: Potentially confounding familial factors did not completely account for the observed association. Furthermore, the association with divorce was somewhat stronger than the association with widowhood, suggesting that while the end of a marriage in general can be pathogenic, divorce is particularly salient.

Accounting for an individual’s propensity toward deviant behavior by controlling for early-onset deviance and family history mitigated the role of divorce substantially. This suggests that, among individuals inclined toward deviance, the association between divorce and abuse is due in part to shared liability. Indeed, as demonstrated by Salvatore et al. (25), divorce and alcohol-use disorder are genetically and environmentally correlated (rG = 0.52–0.76; rE = 0.27–0.32) in the Swedish population; a similar relationship may exist for divorce and DA. However, the association with divorce was not entirely diminished in these models, indicating that the relationship may due to both shared liability and causality.

The potential causal role of divorce was further bolstered by co-relative analyses. If the association between divorce and DA were due entirely to shared genetic factors influencing both, we would expect the observed hazard ratio to decrease with increasing genetic relatedness in these analyses. However, hazard ratio estimates remained quite consistent across relatives (all between 4.8 and 4.9), suggesting that the association with divorce is robust to potentially confounding familial factors. These results differ somewhat from those observed for the association between divorce and alcohol problems: A similar study (14) reported a substantial decline in hazard ratio between cousins and full siblings in co-relative analyses, suggesting a more prominent, although not entirely explanatory, role of familial confounding. Divorce is a stressful life event that can lead to depression (35) and may contribute to self-medicating behavior such as substance use or misuse (36). Substance use may manifest when there is no longer a spouse who helps control potentially deviant behavior. However, in the absence of appropriate data, hypotheses regarding the mechanisms underlying a causal pathway from divorce to DA would be speculative.

We explored the role of spousal deviance in part to account for potential assortative mating for substance use or other externalizing behaviors (2628, 37, 38). Some evidence suggests that substance-use outcomes are poorer among concordant couples (38), which is not entirely consistent with the notion that marriage is protective against psychopathology; the implications for divorce have not previously been explored. Thus, our analysis of the interaction between spousal deviance and divorce provides novel insight to the complicated nature of DA risk factors. Divorce from a deviant spouse was not actually protective against DA onset; however, its reduced association relative to divorce from a nondeviant spouse demonstrates that the removal of one risk factor (a deviant spouse) can mitigate the association with another (divorce). These results provide support for heterogeneity of the role of divorce (31) and merit further consideration in studies designed to dissect these relationships.

Our analyses of widowhood provide important context for the divorce findings. While DA potentially leads to divorce (30), even if the level of use has not led to a drug-related registration, it is less likely that DA would lead to widowhood. Thus, testing associations with widowhood enabled us to explore whether another form of spousal loss, one with fewer confounders, is similarly associated with abuse onset. We confirmed that the end of a marriage increases risk for DA onset generally; critically, however, the estimate for widowhood is markedly lower than that of divorce, particularly among men. The difference was not statistically significant, but if it is indicative of a true discrepancy, this result suggests that women may be sensitive to loss of a spouse overall, while men’s response is more dependent on the nature of the loss. Explanations for this discrepancy are unclear: Prior studies about gender differences in the associations with divorce are inconsistent (39), although some evidence suggests that men are more likely to experience first onset of major depression in the wake of a divorce (40), have higher rates of overall morbidity (41), and adjust to divorce more poorly than women (42). Differences between men and women in post-divorce changes in social networks have been observed (43) that, if not paralleled among widowed persons, may contribute to poorer abuse-related outcomes among divorced men. Although additional research is necessary to elucidate these observations, the present study suggests that divorce is a more salient risk factor for DA onset than widowhood.

We note several limitations to the present study. First, we relied on registry-based data rather than on self-report. While this minimizes the risk of recall bias and subjectivity, it might represent a high threshold for DA and fail to capture problems that are undetected by medical professionals or through criminal registers. However, the rate of abuse observed using Swedish registry data is comparable to that in Norway (44), suggesting that no substantial bias is present. The registry data does not detect subthreshold substance problems prior to a divorce, which might contribute to the end of a marriage (45), nor does it detect problems within a marriage that might play a role in risk of DA onset prior to divorce. Being observational, registry data also preclude completely blinded conditions. Furthermore, relevant confounders not captured by registry data cannot be considered in our analyses.

Second, the DA registries are qualitatively different, representing criminal versus medical outcomes. While the correlation across registries was high (r = 0.78), it is possible that the relationship with divorce differs across crime and medical outcomes. Post-hoc sensitivity analyses indicated that divorce-based hazard ratios were higher for crime-related DA registrations, which are less common than medical-related registrations; the difference was more pronounced among women. These findings suggest that divorce may be related to the context in which individuals use substances, or may generally be associated with lifestyle in ways that correspond to different manifestations (and thus consequences) of substance use. Such questions are beyond the scope of these analyses; studies in which more personal data are available may be better suited to explore the mechanisms underlying these differences.

Third, the present analyses were restricted to the cohort born during 1965–1975. The selection of this cohort was due to the availability of data across registries, along with considerations of the average age at marriage, length of marriage, and post-divorce length of observation. Including younger and/or older cohorts might have changed results, but any changes would be based on less complete data. Results might also not be generalizable to other countries or cultures, although the lifetime rate of divorce in Sweden is comparable to that in the United States (46). Finally, our tests of the proportionality assumption suggested that younger divorcées are at higher risk for DA onset, which we are unable to further dissect due to limited registry data for older and younger cohorts.

Despite these limitations, these analyses demonstrate a robust association between divorce and DA onset, which likely is attributable to both causal and noncausal mechanisms. While other explanatory variables, including family history of DA and spousal deviance, play a role, the end of a marriage through divorce or death can have significant public health consequences. Prevention efforts, typically targeted toward adolescents and young adults, may be additionally effective when inclusive of adults whose marriage has ended.

ACKNOWLEDGMENTS

Author affiliations: Virginia Institute for Psychiatric and Behavioral Genetics, Department of Psychiatry, Virginia Commonwealth University, Richmond, Virginia (Alexis C. Edwards, Kenneth S. Kendler); Center for Primary Health Care Research, Lund University, Malmö, Sweden (Sara Larsson Lönn, Jan Sundquist, Kristina Sundquist); and Icahn School of Medicine at Mount Sinai, New York, New York (Kristina Sundquist).

This work was supported by the National Institute on Alcohol Abuse and Alcoholism (grant K01AA021399) and the National Institute on Drug Abuse (grant R01DA030005) at the National Institutes of Health; the Swedish Research Council; and the Swedish Research Council for Health, Working Life, and Welfare.

Conflict of interest: none declared.

Abbreviations

CI

confidence interval

DA

drug abuse

HR

hazard ratio

REFERENCES

  • 1. Compton WM, Thomas YF, Stinson FS, et al. Prevalence, correlates, disability, and comorbidity of DSM-IV drug abuse and dependence in the United States: results from the national epidemiologic survey on alcohol and related conditions. Arch Gen Psychiatry. 2007;64(5):566–576. [DOI] [PubMed] [Google Scholar]
  • 2. Scott KM, Wells JE, Angermeyer M, et al. Gender and the relationship between marital status and first onset of mood, anxiety and substance use disorders. Psychol Med. 2010;40(9):1495–1505. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Hall W, Teesson M, Lynskey M, et al. The 12-month prevalence of substance use and ICD-10 substance use disorders in Australian adults: findings from the National Survey of Mental Health and Well-being. Addiction. 1999;94(10):1541–1550. [PubMed] [Google Scholar]
  • 4. Lin JC, Karno MP, Grella CE, et al. Alcohol, tobacco, and nonmedical drug use disorders in US adults aged 65 years and older: data from the 2001–2002 National Epidemiologic Survey of Alcohol and Related Conditions. Am J Geriatr Psychiatry. 2011;19(3):292–299. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Kendler KS, Lönn SL, Salvatore J, et al. Effect of marriage on risk for onset of alcohol use disorder: a longitudinal and co-relative analysis in a Swedish national sample. Am J Psychiatry. 2016;173(9):911–918. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6. Curran PJ, Muthén BO, Harford TC. The influence of changes in marital status on developmental trajectories of alcohol use in young adults. J Stud Alcohol. 1998;59(6):647–658. [DOI] [PubMed] [Google Scholar]
  • 7. Duncan GJ, Wilkerson B, England P. Cleaning up their act: the effects of marriage and cohabitation on licit and illicit drug use. Demography. 2006;43(4):691–710. [DOI] [PubMed] [Google Scholar]
  • 8. Merline AC, Schulenberg JE, O’Malley PM, et al. Substance use in marital dyads: premarital assortment and change over time. J Stud Alcohol Drugs. 2008;69(3):352–361. [DOI] [PubMed] [Google Scholar]
  • 9. National Institute on Drug Abuse Medical Consequences of Drug Abuse. NIDA; 2012. https://www.drugabuse.gov/related-topics/medical-consequences-drug-abuse. Accessed September 1, 2017.
  • 10. National Drug Intelligence Center National Drug Threat Assessment. Washington, DC: United States Department of Justice; 2011.
  • 11. Kendler KS, Karkowski LM, Prescott CA. Causal relationship between stressful life events and the onset of major depression. Am J Psychiatry. 1999;156(6):837–841. [DOI] [PubMed] [Google Scholar]
  • 12. Simon RW. Revisiting the relationships among gender, marital status, and mental health. AJS. 2002;107(4):1065–1096. [DOI] [PubMed] [Google Scholar]
  • 13. Richards M, Hardy R, Wadsworth M. The effects of divorce and separation on mental health in a national UK birth cohort. Psychol Med. 1997;27(5):1121–1128. [DOI] [PubMed] [Google Scholar]
  • 14. Kendler KS, Lönn SL, Salvatore J, et al. Divorce and the onset of alcohol use disorder: a Swedish population-based longitudinal cohort and co-relative study. Am J Psychiatry. 2017;174(5):451–458. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Sinha R. How does stress increase risk of drug abuse and relapse? Psychopharmacology (Berl). 2001;158(4):343–359. [DOI] [PubMed] [Google Scholar]
  • 16. Sinha R. Chronic stress, drug use, and vulnerability to addiction. Ann NY Acad Sci. 2008;1141:105–130. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Kendler KS, Gardner C, Jacobson KC, et al. Genetic and environmental influences on illicit drug use and tobacco use across birth cohorts. Psychol Med. 2005;35(9):1349–1356. [DOI] [PubMed] [Google Scholar]
  • 18. Kendler KS, Jacobson KC, Prescott CA, et al. Specificity of genetic and environmental risk factors for use and abuse/dependence of cannabis, cocaine, hallucinogens, sedatives, stimulants, and opiates in male twins. Am J Psychiatry. 2003;160(4):687–695. [DOI] [PubMed] [Google Scholar]
  • 19. Palmer RH, Young SE, Corley RP, et al. Stability and change of genetic and environmental effects on the common liability to alcohol, tobacco, and cannabis DSM-IV dependence symptoms. Behav Genet. 2013;43(5):374–385. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Few LR, Grant JD, Trull TJ, et al. Genetic variation in personality traits explains genetic overlap between borderline personality features and substance use disorders. Addiction. 2014;109(12):2118–2127. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Kessler RC, Davis CG, Kendler KS. Childhood adversity and adult psychiatric disorder in the US National Comorbidity Survey. Psychol Med. 1997;27(5):1101–1119. [DOI] [PubMed] [Google Scholar]
  • 22. Kendler KS, Ohlsson H, Sundquist K, et al. Peer deviance, parental divorce, and genetic risk in the prediction of drug abuse in a nationwide Swedish sample: evidence of environment-environment and gene-environment interaction. JAMA Psychiatry. 2014;71(4):439–445. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Shmulewitz D, Greene ER, Hasin D. Commonalities and differences across substance use disorders: phenomenological and epidemiological aspects. Alcohol Clin Exp Res. 2015;39(10):1878–1900. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24. McGorry PD, Purcell R, Goldstone S, et al. Age of onset and timing of treatment for mental and substance use disorders: implications for preventive intervention strategies and models of care. Curr Opin Psychiatry. 2011;24(4):301–306. [DOI] [PubMed] [Google Scholar]
  • 25. Salvatore JE, Larsson Lönn S, Sundquist J, et al. Alcohol use disorder and divorce: evidence for a genetic correlation in a population-based Swedish sample. Addiction. 2017;112(4):586–593. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26. Nordsletten AE, Larsson H, Crowley JJ, et al. Patterns of nonrandom mating within and across 11 major psychiatric disorders. JAMA Psychiatry. 2016;73(4):354–361. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Maes HH, Neale MC, Kendler KS, et al. Assortative mating for major psychiatric diagnoses in two population-based samples. Psychol Med. 1998;28(6):1389–1401. [DOI] [PubMed] [Google Scholar]
  • 28. Treur JL, Vink JM, Boomsma DI, et al. Spousal resemblance for smoking: underlying mechanisms and effects of cohort and age. Drug Alcohol Depend. 2015;153:221–228. [DOI] [PubMed] [Google Scholar]
  • 29. Washburn IJ, Capaldi DM, Kim HK, et al. Alcohol and marijuana use in early adulthood for at-risk men: time-varying associations with peer and partner substance use. Drug Alcohol Depend. 2014;140:112–117. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Amato PR, Previti D. People’s reasons for divorcing: gender, social class, the life course, and adjustment. J Fam Issues. 2003;24(5):602–626. [Google Scholar]
  • 31. Kalmijn M, Monden CWS. Are the negative effects of divorce on well-being dependent on marital quality? J Marriage Fam. 2006;68(5):1197–1213. [Google Scholar]
  • 32. Giordano GN, Ohlsson H, Kendler KS, et al. Unexpected adverse childhood experiences and subsequent drug use disorder: a Swedish population study (1995–2011). Addiction. 2014;109(7):1119–1127. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33. Kendler KS, Lonn SL, Maes HH, et al. A Swedish population-based multivariate twin study of externalizing disorders. Behav Genet. 2016;46(2):183–192. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Kendler KS, Neale MC, MacLean CJ, et al. Smoking and major depression. A causal analysis. Arch Gen Psychiatry. 1993;50(1):36–43. [DOI] [PubMed] [Google Scholar]
  • 35. Kendler KS, Thornton LM, Prescott CA. Gender differences in the rates of exposure to stressful life events and sensitivity to their depressogenic effects. Am J Psychiatry. 2001;158(4):587–593. [DOI] [PubMed] [Google Scholar]
  • 36. Keyes KM, Hatzenbuehler ML, Grant BF, et al. Stress and alcohol: epidemiologic evidence. Alcohol Res. 2012;34(4):391–400. [PMC free article] [PubMed] [Google Scholar]
  • 37. Agrawal A, Heath AC, Grant JD, et al. Assortative mating for cigarette smoking and for alcohol consumption in female Australian twins and their spouses. Behav Genet. 2006;36(4):553–566. [DOI] [PubMed] [Google Scholar]
  • 38. Low N, Cui L, Merikangas KR. Spousal concordance for substance use and anxiety disorders. J Psychiatr Res. 2007;41(11):942–951. [DOI] [PubMed] [Google Scholar]
  • 39. Amato PR. The consequences of divorce for adults and children. J Marriage Fam. 2000;62(4):1269–1287. [Google Scholar]
  • 40. Bruce ML, Kim KM. Differences in the effects of divorce on major depression in men and women. Am J Psychiatry. 1992;149(7):914–917. [DOI] [PubMed] [Google Scholar]
  • 41. Riessman CK, Gerstel N. Marital dissolution and health: do males or females have greater risk? Soc Sci Med. 1985;20(6):627–635. [DOI] [PubMed] [Google Scholar]
  • 42. Diedrick P. Gender differences in divorce adjustment. J Divorce & Remarriage. 1991;14(3–4):33–46. [Google Scholar]
  • 43. Albrecht SL. Reactions and adjustments to divorce: differences in the experiences of males and females. Fam Relat. 1980;29(1):59–68. [Google Scholar]
  • 44. Kendler KS, Ohlsson H, Sundquist K, et al. The rearing environment and risk for drug abuse: a Swedish national high-risk adopted and not adopted co-sibling control study. Psychol Med. 2016;46(7):1359–1366. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45. Cranford JA. DSM-IV alcohol dependence and marital dissolution: evidence from the National Epidemiologic Survey on Alcohol and Related Conditions. J Stud Alcohol Drugs. 2014;75(3):520–529. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46. Organisation for Economic Co-operation and Development (OECD) Marriage and Divorce Rates Paris, France: Organisation for Economic Co-operation and Development; 2016.

Articles from American Journal of Epidemiology are provided here courtesy of Oxford University Press

RESOURCES