Origins of spousal cross-concordance for psychiatric disorders: a test of the social stress theory for alcohol use disorder

Abstract

Background

The authors sought to clarify the impact of spousal psychiatric disorders of differing severity [major depression or anxiety disorders (DAD) v. bipolar disorder or nonaffective psychosis (BPN)] on proband risk for alcohol use disorder (AUD) during marriage.

Methods

In a Swedish cohort (N = 744 628), associations between spousal DAD and BPN and proband AUD were estimated with Cox proportional hazards; associations between parental AUD, proband premarital AUD, and spousal lifetime DAD and BPN were estimated with logistic regression; and whether spousal DAD or BPN causally increased risk for AUD was evaluated with frailty models.

Results

Spousal premarital DAD, spousal marital-onset DAD, and spousal BPN (premarital or marital-onset) were associated with proband AUD during marriage [hazard ratios (HR) range 1.44–3.72]. Those with a parental or premarital history of AUD (v. without) were more likely to marry a spouse with DAD or BPN (odds ratios 1.22–2.77). Moving from an unaffected first spouse to a DAD-affected second spouse increased AUD risk in males (HR 2.90). Moving from an unaffected first spouse to a BPN-affected second spouse increased AUD risk (HR_males 3.96; HR_females 5.64). Moving to an unaffected second spouse from a DAD-affected first spouse decreased AUD risk, with stronger evidence in females compared to males (HR_males 0.59; HR_females 0.28).

Conclusions

Associations between spousal DAD or BPN and proband AUD reflect both selection and causal effects. Marriage to a BPN-affected spouse has a particularly strong effect on AUD risk, with more modest effects for spousal DAD.

Spouses profoundly shape one another's alcohol use and risk for alcohol use disorder (AUD) (Kendler, Larsson Lönn, Salvatore, Sundquist, & Sundquist, 2018; Kendler, Lönn, Salvatore, Sundquist, & Sundquist, 2016; Leonard & Eiden, 2007). A great deal is known about spousal concordance for alcohol consumption and problems, with evidence for both selection and contagion effects (Kendler et al., 2018; Maes et al., 1998; Merikangas, 1982). Yet, despite evidence for spousal cross-concordance (i.e. the tendency for individuals with one disorder to have a spouse with another disorder) among AUD and other psychiatric disorders (Low, Cui, & Merikangas, 2007; Merikangas, 1982), there is little understanding of how these other forms of spousal psychiatric disorder influence AUD risk. Informed by social stress theories of AUD (Keyes, Hatzenbuehler, & Grant, 2012; Spanagel, Noori, & Heilig, 2014), which posit that experiences of psychosocial stress may precipitate the onset of disorder, we expected that the level of disruption associated with a spouse's psychiatric disorder would be directly, causally linked to AUD risk. To evaluate this hypothesis, we sought to clarify the nature of the associations between AUD during marriage and two classes of psychiatric disorders in spouses that vary in their level of impairment: major depression or anxiety disorders (DAD) and bipolar disorder or other nonaffective psychosis (BPN), where the latter is often more serious than the former.

Major depression and anxiety disorders are common and typically associated with moderate impairment (Kessler & Bromet, 2013; Ruscio et al., 2017). In contrast, bipolar disorder and nonaffective psychosis represent more severe illnesses often involving pronounced impairments, alarming behaviors, and hospitalizations (Druss et al., 2009; Harvey et al., 2012). The majority of individuals with lifetime major depression or anxiety disorders marry (Akhtar-Danesh & Landeen, 2007; Hasin et al., 2018; The ESEMeD/MHEDEA 2000 Investigators et al., 2004; Vesga-López et al., 2008). To a lesser degree (MacCabe, Koupil, & Leon, 2009), marriage is also common among those with bipolar disorder or nonaffective psychosis, with the prevalence of marriage/cohabiting ranging from 20% (schizophrenia) to 61% (bipolar I; the most severe form of bipolar disorder) (Saarni et al., 2010).

Marriage to a spouse with DAD or BPN is associated with AUD (Maes et al., 1998; Merikangas, 1982; Nordsletten et al., 2016). The degree to which these associations reflect selection effects, whereby individuals predisposed to AUD are more likely to partner with affected spouses, or causal effects, whereby these spousal psychiatric disorders have direct, pathogenic effects on proband AUD risk, is not clear. Evaluating the evidence for these non-causal and causal explanations could have implications for timing preventive interventions among those married to affected spouses.

We used epidemiological Swedish national registry data, including within-person designs, to understand the origin of spousal cross-concordance among DAD, BPN, and AUD. We asked the following questions:

1. (1) What is the association between spousal DAD, spousal BPN, and proband AUD during marriage?

2. (2) Are the associations between spousal DAD, spousal BPN, and proband AUD explained by proband premarital AUD registration, low parental education, parental history of AUD, parental separation, or parental death?

3. (3) Are those at higher risk for an AUD registration during marriage, owing to a parental history of AUD or a premarital AUD registration, more likely to marry a spouse with DAD or BPN?

4. (4) Does a premarital history of AUD modify the effects of spousal DAD or BPN?

5. (5) Using a within-person design in a sample of individuals with multiple marriages, can we determine whether spousal DAD or BPN causally increases risk for AUD?

Methods

Sample

We linked nationwide Swedish registers via the unique 10-digit identification number assigned at birth or immigration to all Swedish residents. The identification number was replaced by a serial number to ensure people's integrity. The following sources were used to create our dataset: Total Population Register (year of birth, sex, family and marital status); Multi-Generation Register (linking individuals born after 1932 to their parents); the Longitudinal Integration Database for Health Insurance and Labor Market Studies (LISA) (education and income from 1990 to 2014); the Hospital Discharge Register (hospitalizations for Swedish inhabitants from 1964 to 2017); Prescribed Drug Register (all prescriptions in Sweden picked up by patients from July 2005 to 2017); Outpatient Care Register (all outpatient clinics from 2001 to 2017); and regional Primary Health Care Registers [time-periods vary due to the region-specific differences in patient record digitalization and were as follows: Blekinge (2009–2016), Dalarna (2005–2013), Värmland (2005–2015), Kalmar Län (2007–2016), Sörmland (1992–2017), Uppsala Län (2005–2015), Västernorrland (2008–2015) Norrbotten Län (2001–2014), Gävleborg (2010–2017), Gotland (2011–2018), Halland (2007–2014), Jönköpings Län (2008–2014), Kronoberg (2006–2016), Skåne (1989–2018), Västerbotten (1992–2018), Östergötland (1990–2014), Stockholms Län (2003–2016), and Västra Götaland (2000–2013)]. In addition, we used the Crime Register (national complete data on all convictions in lower court from 1973 to 2017); the Swedish Suspicion Register (national data on individuals strongly suspected of crime from 1998 to 2015); and the Mortality Register (dates and causes of death from 1952 until 2016).

Measures

AUD was defined from medical registries by the following ICD codes: ICD8: 571.0, 291, 303; ICD9: 305A, 357F, 571A, 571B, 571C, 571D, 425F, 535D, 291, 303; and ICD 10: E24.4, G31.2, G62.1, G72.1, I42.6, K29.2, K70.0, K70.1, K70.2, K70.3, K70.4, K70.9, K85.2, K86.0, O35.4, F10.1, F10.2, F10.3, F10.4, F10.5, F10.6, F10.7, F10.8, F10.9; and from the Prescribed Drug Register if retrieved Drugs used in alcohol dependence [Anatomical Therapeutic Chemical (ATC) Classification System N07BB]; disulfiram (N07BB01), acamprosate (N07BB03), naltrexone (N07BB04), or nalmefene (N07BB05). In addition, AUD was defined by conviction for or suspicion of at least two alcohol-related crimes according to law 1951:649, paragraph 4 and 4A and law 1994:1009, Chapter 20, paragraph 4 and 5 from the Swedish Crime Register, and code 3005 and 3201 in the Suspicion register. Major depression or anxiety disorders were defined in the medical registries by the following ICD codes: ICD-8: 296.0, 296.2, 298.0, 300.4, 300.0, 300.2; ICD-9: 296B, 298A, 300A, 300C, 300E; ICD-10: F32, F33, F40, F41. Bipolar disorder or nonaffective psychosis were defined in the medical registries by the following ICD codes: ICD-8: 298.3, 298.9, 296.1, 296.3, 296.8, 296.9, 298.1; ICD-9: 298E, 298W, 298X, 296A, 296C, 296D, 296E, 296W, 298B; ICD-10 codes: F20, F22, F23, F24, F25, F26, F27, F28, F29, F30, F31.

Covariates included parental history of AUD, which was defined as any AUD registration in either biological parent. Parental separation was defined at age 16. If an individual resided with both parents, the parents were considered married/cohabiting; otherwise, the parents were defined as separated if both parents were alive. If one parent had died, we defined that as parental death, and any prior separation was ignored. Parental education was used to control for family socioeconomic background, and was assessed as the highest level of either parent and categorized into low (compulsory school only), mid (upper secondary school), and high (university level).

Statistical analyses

We first used Cox proportional hazard models to estimate the associations between proband AUD and spousal DAD (premarital and marital-onset, separately, as there was sufficient variation to examine both) and spousal BPN (premarital or marital-onset combined, as BPN are lifetime disorders and to avoid misclassifying someone who, e.g. had a premarital diagnosis of major depression that later converted to bipolar disorder during marriage) in model 1. Spousal DAD during marriage was included as a time-dependent covariate. To adjust for possible measured confounding, we added the following as covariates in model 2: proband premarital AUD, parental education, parental AUD, parental separation, and parental death. To investigate the proportionality assumption and examine the possible effect of follow-up time on the observed associations, we added a linear effect of follow-up time that interacted with spousal premarital DAD, the time-dependent covariate DAD during marriage, and spousal BPN, respectively.

We used logistic regressions to examine the associations between parental AUD, proband premarital AUD, and spousal lifetime DAD and BPN. Next, to examine whether proband premarital AUD modified the association with spousal DAD (premarital or marital-onset) and spousal BPN, we included three interaction terms: proband premarital AUD × spousal premarital DAD; proband premarital AUD × spousal marital-onset DAD; and proband premarital AUD × spousal BPN.

Finally, to evaluate whether marriage to a spouse with DAD or BPN causally increase risk for AUD, we compared first and second marriages within individuals. For these analyses we used a frailty model, which is an extension of a Cox model where clustering within individuals is accounted for by including a random effect.

Analyses were sex-stratified in view of sex differences in rates of AUD, BPN, and DAD, which can change the underlying hazard function, as well as preliminary analyses which indicated sex-specific effects. Analyses were run in SAS software (version 9.4) and R Studio (version 4.0.3).

Results

Descriptive statistics

Table 1 summarizes descriptive information for the study sample. More males than females developed AUD during marriage. Compared to females, more males were married to a spouse with DAD or BPN.

Table 1. Descriptive statistics for the study sample

AUD, alcohol use disorder; DAD, major depression or anxiety disorders disorder; BPN, bipolar disorder or nonaffective psychosis; M, mean; s.d., standard deviation.

Notes. AUD married refers to probands who developed AUD during marriage.

Associations between spousal DAD, spousal BPN, and proband AUD during marriage

We used Cox models to examine the associations between spousal DAD, spousal BPN, and proband AUD registration during marriage. As shown in Table 2, model 1, spousal premarital DAD [hazard ratio (HR_males) 1.99, 95% confidence interval (CI) 1.79–2.21; HR_females 2.45, 95% CI 2.00–3.00]; spousal marital-onset DAD (HR_males 1.44, 95% CI 1.30–1.60; HR_females 1.74, 95% CI 1.45–2.08), and spousal BPN (HR_males 2.82, 95% CI 2.30–3.45; HR_females 3.72, 95% CI 2.63–5.25) were associated with proband AUD registration during marriage. As shown in Table 2, model 2, the associations between spousal DAD, spousal BPN, and proband AUD were attenuated after controlling for a number of covariates, including whether the proband had a premarital AUD registration and a number of family-of-origin risk factors including parental AUD, lower parental educational attainment, parental separation, and parental death. In a formal test, the excess AUD risk associated with spousal BPN was approximately 1.5-fold that of spousal premarital DAD (HR_males 1.42, 95% CI 1.13–1.78; HR_females 1.52, 95% CI 1.03–2.23), and twice that of spousal marital-onset DAD (HR_males 1.96, 95% CI 1.57–2.45; HR_females 2.14, 95% CI 1.46–2.14).

Table 2. Results of Cox regression analyses predicting AUD during marriage as a function of spousal DAD and BPN

AUD, alcohol use disorder; DAD, major depression or anxiety disorders; BPN, bipolar disorder or nonaffective psychosis; HR, hazard ratio.

To better understand the temporal dynamics underlying the associations between spousal DAD, spousal BPN, and proband AUD, we then ran Cox models that included multiplicative interaction terms between spousal DAD × time and spousal BPN × time. The results of these models are depicted in Fig. 1. There was approximately 1.5–2.0-fold excess risk associated with spousal premarital or spousal marital-onset DAD, and approximately 2.5–3.5-fold excess risk associated with spousal BPN. This excess risk was not stable but rather decayed slowly over time.

Fig. 1. HR for AUD registration during marriage in males and females as a function of a spouse with (v. without) a premarital major depression or anxiety disorder (DAD) registration; DAD during marriage; or a bipolar disorder or nonaffective psychosis (BPN) registration.

Note. Time is represented as year from marriage for spousal premarital DAD and BPN, and from year of registration for spousal DAD during marriage.

Likelihood of marriage to a spouse with DAD or BPN

The top panel of online Supplementary Table S1 summarizes the percent of probands whose first spouses had a DAD or BPN diagnosis, stratified by proband parental AUD. In logistic regressions, males with v. without a parental history of AUD were more likely to have a first marriage to a spouse with a lifetime diagnosis of DAD [odds ratio (OR) 1.22, 95% CI 1.19–1.25] or BPN (OR 1.25, 95% CI 1.16–1.36). Likewise, females with v. without a parental history of AUD were more likely to marry a spouse with DAD (OR 1.27, 95% CI 1.23–1.31) or BPN (OR 1.31, 95% CI 1.18–1.46).

The bottom panel of online Supplementary Table S1 summarizes the percentage of male and female probands whose first spouses had a DAD or BPN diagnosis, stratified by proband premarital AUD registration. Males with a premarital AUD registration were more likely than those without to have a first marriage to a spouse with a lifetime diagnosis of DAD (OR 1.63, 95% CI 1.56–1.72) or BPN (OR 2.46, 95% CI 2.15–2.80). Likewise, females with v. without a premarital AUD registration were more likely to marry a spouse with DAD (OR 1.84, 95% CI 1.79–1.99) or BPN (OR 2.77, 95% CI 2.21–3.47).

Proband premarital AUD as a modifier of spousal DAD and BPN

We used multiplicative interactions to test whether proband premarital AUD registration modified the risk associated with spousal DAD or BPN. As shown in Table 3, there were interactions between proband premarital AUD and spousal premarital DAD in both sexes (HR_males 1.26, 95% CI 1.00–1.58; HR_females 2.19, 95% CI 1.44–3.31). Among those with premarital AUD, marriage to a spouse with v. without premarital DAD was associated with a relative increase in the risk of AUD registration (males: HR_{PreAUD,PreDAD} 24.87, 95% CI 20.77–29.78) v. HR_{PreAUD,NoPreDAD} 14.46, 95% CI 13.09–15.98; females: HR_{PreAUD,PreDAD} 53.93, 95% CI 39.42–73.78 v. HR_{PreAUD,NoPreDAD} 17.88, 95% CI 15.30–20.89).

Table 3. Cox regression models of proband AUD during marriage as a function of proband premarital AUD, spousal DAD, and their interaction

AUD, alcohol use disorder; DAD, major depression or anxiety disorders; BPN, bipolar disorder or nonaffective psychosis; HR, hazard ratio.

^a Derived estimates reflect the combined effects for the significant interactions. For example, in males, the joint effect of proband premarital AUD with spousal premarital DAD is the product of 14.46 (i.e. proband premarital AUD parameter) × 1.37 (i.e. spousal premarital DAD parameter) × 1.26 (i.e. interaction proband premarital AUD and spousal premarital DAD parameter).

We also found evidence for an interaction between proband premarital AUD and spousal marital-onset DAD in both sexes (HR_males 0.45, 95% CI 0.34–0.59; HR_females 0.36, 95% CI 0.21–0.63). Among probands with premarital AUD, there was a significant relative reduction in the risk of AUD registration when married to a spouse with v. without marital-onset DAD (males: HR_PreAUD,DAD 10.75, 95% CI 8.33–13.87 v. HR_PreAUD,NoDAD 14.46, 95% CI 13.09–15.98; females: HR_PreAUD,DAD 12.47, 95% CI 7.36–21.12 v. HR_PreAUD,NoDAD 17.88, 95% CI 15.30–20.89).

In contrast, as shown in online Supplementary Table S2, the interaction between proband premarital AUD and spousal BPN did not differ from unity in either sex (HR_males 1.35, 95% CI 0.88–2.06; HR_females 0.85, 95% CI 0.32–2.23).

Within-person comparisons of AUD in multiple marriages

We next examined AUD registrations among the subset of the Swedish population that married twice, and whose first and second spouses differed in their DAD and BPN status. Given a causal relationship between spousal DAD, spousal BPN, and AUD, we expected that the move to an affected second spouse from an unaffected first spouse would be associated with increased AUD risk, while the move to an unaffected second spouse from an affected first spouse would be associated with decreased AUD risk.

Table 4 summarizes these effects, always comparing the AUD risk in the second marriage compared to the first. Beginning with males, risk of AUD registration was higher during a second marriage to a spouse with a marital-onset DAD compared to a first marriage to a spouse without a marital-onset DAD (HR 2.90, 95% CI 1.64–5.11). Likewise, proband AUD risk was higher during a second marriage to a BPN-affected spouse compared to a first marriage to a spouse without BPN (HR 3.96, 95% CI 1.63–9.61). In contrast, AUD risk as a function of moving to a second spouse with premarital DAD from a first spouse without premarital DAD did not differ from unity (HR 1.25, 95% CI 0.72–2.16). The contrasts capturing transitions from an affected first spouse to an unaffected second spouse did not differ from unity; however, there was suggestive evidence that moving from a first spouse with marital-onset DAD to an unaffected second spouse decreased AUD risk (HR_DAD 0.59, 95% CI 0.32–1.06).

Table 4. Risk of proband AUD registration during marriage among individuals with multiple marriages

AUD, alcohol use disorder; DAD, major depression or anxiety disorders; BPN, bipolar disorder or nonaffective psychosis; HR, hazard ratio.

Among female probands, AUD risk was higher during a second marriage to a BPN-affected spouse compared to a first marriage to a BPN-unaffected spouse (HR 5.64, 95% CI 1.02–31.06). AUD risk as a function of moving to a second spouse with premarital or marital-onset DAD from an unaffected first spouse did not differ from unity (HR_PreDAD 2.01, 95% CI 0.85–4.75; HR_DAD 1.26, 95% CI 0.41–3.87). We observed a protective effect whereby the within-person risk of AUD was reduced when moving from a first spouse with marital-onset DAD to an unaffected second spouse (HR 0.28, 95% CI 0.09–0.87). The contrasts capturing AUD risk in second marriages to unaffected spouses compared to first marriages to spouses with premarital DAD or BPN did not differ from unity (HR_PreDAD 6.05, 95% CI 0.68–54.00; HR_BPN 0.93, 95% CI 0.21–4.09).

Discussion

In a population cohort, marriage to a spouse with a premarital major depression or anxiety disorder registration, a major depression or anxiety disorder registration during marriage, or bipolar disorder or nonaffective psychosis was associated with AUD, with effects reflecting 1.4–3.7-fold excess risk. The AUD risk associated with spousal bipolar disorder or nonaffective psychosis was approximately two-fold that of spousal major depression or anxiety disorders, which is consistent with the social stress perspective (Spanagel et al., 2014) that the level of disruption associated with the spouse's disorder directly corresponds to its risk for AUD. These effects were robust to several potential confounders.

The associations between proband AUD and spousal premarital major depression or anxiety disorders and bipolar or nonaffective psychosis were highest at the beginning of marriage, or immediately following the spouse's major depression or anxiety disorder onset during marriage, and decayed over time. This transient increase in risk is suggestive of a causal effect, and is of modest/moderate magnitude (1.5–4-fold) compared to the large (over nine-fold) increase in AUD registration following a spouse's AUD registration (Kendler et al., 2018). The high risk associated with spousal AUD likely reflects the direct socialization of alcohol use among spouses (Leonard & Das Eiden, 1999), while the more moderate risk associated with spousal bipolar disorder or nonaffective psychosis and major depression or anxiety disorders may reflect the spousal stress associated with these conditions.

We then probed the nature of the associations between these spousal psychiatric disorders and AUD. Consistent with the selection hypothesis, there were modest associations between proband parental history of AUD and likelihood of marriage to a spouse with a major depression or anxiety disorder registration, or with bipolar disorder or nonaffective psychosis. Likewise, probands with a premarital AUD registration were also somewhat more likely to marry affected partners. These findings, while of modest magnitude, add to the prior literature showing that the offspring of AUD-affected parents are at increased risk of marrying an AUD-affected spouse (Hall, Hesselbrock, & Stabenau, 1983; Salvatore et al., 2019; Schuckit, Tipp, & Kelner, 1994) and demonstrate that these offspring are also at-risk of marrying a spouse with other psychiatric disorders.

Proband premarital AUD registration modified the effects of spousal major depression or anxiety disorders, with different patterns observed for premarital and marital-onset. Among probands with a premarital AUD registration, the risk ratio for AUD during marriage was higher if married to a spouse with (v. without) a premarital major depression or anxiety disorder. This is consistent with a diathesis-stress effect (Zuckerman, 1999), whereby probands with a predisposition to AUD are sensitized to spousal major depression or anxiety disorders. In contrast, among probands with premarital AUD, the risk ratio for AUD during marriage was lower if married to a spouse with (v. without) a marital onset of major depression or anxiety disorder. This relative reduction in risk does not suggest a ‘protective’ effect of spousal major depression or anxiety disorder, but rather an attenuation of the already large risk associated with probands' premarital AUD registration.

In our final set of analyses, we used a within-person design to evaluate the causal impact of these spousal disorders on proband AUD. Within-person designs are uniquely powerful for drawing quasi-causal inferences (Schmiedek & Neubauer, 2020), as they control for familial factors (e.g. genes, rearing environment) and other time-invariant individual confounders (e.g. personality, adolescent drinking). When probands moved from an unaffected first spouse to a second spouse with bipolar disorder or nonaffective psychosis, risk for AUD increased approximately four- to five-fold. The effects for spousal major depression or anxiety disorders were more modest. Among males, moving from an unaffected first spouse to a second spouse with a major depression or anxiety disorder registration during marriage increased AUD risk nearly three-fold. However, among females, the excess AUD risk associated with moving to a second spouse with a major depression or anxiety disorder registration during marriage did not differ from unity. The excess AUD risk associated with moving to a spouse with a premarital major depression or anxiety disorder registration was in the same direction across the epidemiological and within-person models; however, the estimate did not differ from unity in the within-person models.

We also found that moving from a first spouse with a major depression or anxiety disorder registration during marriage to a second spouse without these disorders reduced risk for AUD. This protective effect was strongest in female probands, and only suggestive in males. The estimates for the hypothesized protective effects of moving from a first spouse with a premarital major depression or anxiety disorder registration, or bipolar disorder or nonaffective psychosis, to a second unaffected spouse did not differ from unity. The within-person analysis results build upon prior epidemiologic evidence that spousal DAD and BPN are associated with AUD (Galbaud du Fort, Bland, Newman, & Boothroyd, 1998; Nordsletten et al., 2016), and provide important information about the temporal dynamics underlying these cross-disorder spousal associations. Consistent with the idea that psychosocial stress may precipitate the onset of AUD (Keyes et al., 2012; Spanagel et al., 2014), we found that marriage to a DAD- or BPN-affected spouse increased risk of developing AUD, and that this risk was more pronounced for the more disruptive disorders (i.e. BPN) compared to the relatively less disruptive disorders (i.e. DAD). There was more modest evidence that reductions in psychosocial stress (i.e. divorce from an affected spouse and remarriage to an unaffected spouse) were protective against AUD. It may be the case that exposure to an affected first spouse may not end abruptly following divorce, particularly if the couple is co-parenting minor children.

Limitations

The results should be considered in the context of study limitations. First, AUD diagnoses from registers capture the more severely affected individuals. The under-diagnosis of AUD in registers may have led us to underestimate the number of individuals with AUD caused by spousal disorder. However, tempering this concern, the rates of AUD observed here closely parallel the rates observed in an epidemiological study in nearby Norway (Kringlen, Torgersen, & Cramer, 2001). Second, we included only legally married opposite-sex couples, and it is unknown whether the pattern of findings generalizes to non-marital cohabiting relationships, which are relatively common in Sweden (Organisation for Economic Co-operation and Development, 2016), or to same-sex couples.

Third, correlated ascertainment, whereby one partner seeking medical attention for a psychiatric disorder increases the likelihood that illness in the other partner will also be noticed (Penrose, 1944), could potentially explain our observation that these spousal disorders increase risk for proband AUD. To evaluate the degree to which correlated ascertainment might explain our effects, we examined whether the associations held when spousal diagnoses came from a different register than the proband AUD diagnosis (medical v. criminal). Owing to the fact that very few females receive AUD registrations through the criminal register, these supplementary analyses were conducted in males only. We found that spousal premarital DAD and spousal BPN were associated with increased risk for males' AUD identified through the criminal registry (HR_PreDAD 3.01, 95% CI 2.44–3.71; HR_BPN 1.54, 95% CI 1.61–4.02). Thus, the results from these analyses temper this concern for spousal premarital DAD and spousal BPN. However, the association between spousal marital-onset DAD and males' AUD (from the criminal register only) no longer significantly differed from unity (HR 0.99, 95% CI 0.77–1.28). Thus, on the basis of these supplementary analyses, we cannot completely rule out the possibility that some of the effect of spousal marital-onset DAD on males' AUD may be attributable to the fact that wives' illness may also bring their husbands' symptomatology to medical attention.

Conclusions

Spousal cross-concordance for major depression or anxiety disorders, bipolar disorder or nonaffective psychosis, and AUD reflect both selection and causal effects. Consistent with selection effects, those with a parental or premarital history of AUD were more likely to marry a spouse with a major depression or anxiety disorder, or bipolar disorder or nonaffective psychosis. We also found evidence for causal effects, including a transient increase in AUD risk at the time of marriage or spousal onset; and within-person changes in AUD risk when moving from an unaffected to affected spouse or the reverse. This study provides some of the strongest evidence that the social stress associated with a spouse's psychiatric disorder has direct, causal effects on risk for AUD. More broadly, the effects reported here underscore the importance of family systems approaches in clinical practice to reduce the sequelae of psychiatric disorders on other members of the family (Heru, 2006).