Coincident Alcohol Dependence and Depression Increases Risk of Suicidal Ideation among Army National Guard Soldiers

Abstract

Purpose

Suicide rates among military service members have risen dramatically, while drivers remain poorly understood. We aimed to examine the relationship between coincident alcohol dependence and depression in shaping subsequent risk of suicidal ideation among National Guard forces.

Methods

We performed a longitudinal analysis using a randomly selected, population-based sample of Ohio Army National Guard soldiers. Telephone-based surveys of 1582 soldiers who participated in both wave 1 (data collected 2008–2009) and wave 2 (data collected 2009– 2010) were analyzed.

Results

Incident suicidal ideation was present among 2.47% of soldiers at follow-up. Odds ratios (ORs) for suicidal ideation among those with vs. without alcohol dependence were similar among non-depressed [OR=3.85 (95% Confidence Intervals(CI) = 1.18–12.52)] and depressed individuals [OR = 3.13 (95% CI = 0.88–11.14)]; a logistic model cross-product term confirmed an absence of multiplicative interaction (beta coefficient=−0.21, p=0.82). In contrast, the risk differences (RD) for suicidal ideation among those with vs. without alcohol dependence diverged for those without depression [RD = 0.04 (95% CI = 0.02–0.07)] compared to those with depression [RD 0.11(95% CI=0.06–0.18)]; strong evidence of additive interaction was observed - [Relative Excess Risk of Interaction (RERI) = 5.978(95% CI=0.364–11.591)].

Conclusions

We found that alcohol dependence and depression worked together to shape risk for incident suicidal ideation among Army National Guard service members. Because coincident alcohol dependence and depression is relatively rare, a high-risk prevention approach is recommended. Population-based screening for suicidality among patients with alcohol dependence, depression, and particularly those with both conditions is warranted in military populations.

INTRODUCTION

Suicide and suicide risk among U.S. military forces and veterans have increased substantially in the last decade [1,2]. Suicide in soldiers and veterans has historically been lower than that in the general population [3]; however, this difference has diminished substantially in recent years among service members from Operation Enduring Freedom (OEF), Operation Iraqi Freedom (OIF), and Operation New Dawn (OND) [4–7]. This increase in suicide rates has come despite the investment of tremendous financial and human resources aimed at mitigating potential suicidal behavior among military forces and veterans [8].

The suicide rate among Army National Guard soldiers has in recent years exceeded that of Active Duty Army and Army Reserve [9,10,7], representing a divergence from earlier trends. Indeed, in both 2009 and 2010, Active Duty service members had nearly twice the risk of suicide relative to those in the National Guard or Reserves (reservists) [9]. By branch, Army soldiers had the highest rate of suicide in 2010, with a rate of 21.7 per 100,000. These trends reversed between 2010 and 2013, such that the suicide rate among Army National Guard soldiers exceeded that of Active Duty Army and Army Reserve [9,10,7]. Surveillance data from 2014 suggests this excess risk among Army National Guard service members may be leveling off [7], but this divergence nonetheless raises concern about the mental health of service members in the Army National Guard.

Sociodemographic predictors of suicide risk appear to be consistent among both reservists and active duty service members [11,9,10,7]. Prior deployment in support of OEF/OIF/OND (Operation Enduring Freedom/ Operation Iraqi Freedom/Operation New Dawn) and combat exposure have been associated with suicide in some studies [10,5], but not others [12,6]. Lower rank [12,5,6] has also been associated with increased suicide risk, and while some studies have linked greater suicide risk to male sex [12,10,7], others have found increased risk among female service members [13,5].

Depression is perhaps the most common and well-recognized predictor of suicidality both in civilian [14–16] and in military populations [12,17,5,18]. Depression is relatively common among OEF/OIF/OND service members and veterans, with prevalence estimates for National Guard and reserves forces ranging from 2.7% to 14.7% [19]. Given the approximately 3-fold magnitude of associated risk for suicide [12] and suicidal ideation [5] among service members, depression remains an important target for suicide prevention.

It is highly plausible that alcohol misuse plays an important role, in its own right, in suicide and suicidality among military personnel. However, whereas a few studies have examined the influence of alcohol misuse [12,20], most literature to date on suicide and suicidality in OEF/OIF/OND service members has focused almost exclusively on depression and PTSD [21,22] as causes, largely considering alcohol misuse only peripherally, as a covariate or comorbidity. Within the general population, alcohol use disorders and acute alcohol intoxication are strongly associated with suicide attempts and suicide [23–25]. Both acute alcohol intoxication (including abuse) and chronic use of alcohol (including dependence) are associated with increased risk of suicide and attempted suicide [23,26,25]. Alcohol use disorders are among the most common diagnoses associated with suicide, second only to depression, and are implicated in approximately one quarter of suicides [23].

Of particular concern, alcohol use disorders are highly prevalent in military populations. Among Reserve and National Guard forces serving during the OEF/OIF/OND conflicts, alcohol use disorders are prevalent among approximately 14 to 15% [19]. In 2014, alcohol use was implicated in 28.8% of suicides throughout the military, and 23.4% of suicides within the Army [7], lending further plausibility to the idea that alcohol dependence may be causal in suicide among military populations. Among service members with alcohol misuse, the increased odds of suicidal ideation and suicide are approximately-3 fold [12,18].

In addition to being independent risk factors among service members, it is likely that alcohol dependence and depression work together in shaping risk of suicidal ideation. Evidence from civilian studies [27,16] suggests that comorbid alcohol dependence and depression are associated with a severity of suicidality greater than that for either condition alone. These civilian findings, along with the relatively common occurrence and comorbidity of both conditions among military service members [28,29], provide strong warrants for investigating their synergy in predicting suicidal ideation among service members. Evidence that alcohol misuse may exacerbate depression symptom trajectories among Army National Guard service members [30] adds further credence.

Accordingly, the present investigation aimed to estimate among Army National Guard forces, (1) the effect of alcohol dependence on risk of incident suicidal ideation; (2) the effect of depression on risk of incident suicidal ideation; and (3) the interaction between alcohol dependence and depression on risk of incident suicidal ideation. We utilize an Ohio Army National Guard (OHARNG) sample and adjust for the influence of sociodemographic characteristics and military service factors. In order to examine the degree to which alcohol dependence and depression act together in predicting suicidal ideation, we examine statistical interaction on both the additive and multiplicative scales.

METHODS

Population

We conducted a longitudinal investigation of a representative sample of the OHARNG. Our sample was drawn from 10,778 soldiers serving in the ONG as of June 2008, and the 1,792 who enlisted between then and February 2009. In total, 12,225 OHARNG soldiers were invited to participate through a 2-stage process including an opt-out alert letter directly from the OHARNG, and a phone call to obtain consent to participate in a phone interview. 1013 (8%) soldiers opted out, and 345 (2.8%) were excluded due to lack of a current address. After 3 weeks, the OHARNG sent us contact information for 11,212 soldiers who did not opt-out. We called 6,514 (58.1%) working numbers: 218 (3.3%) were not eligible (i.e. too young, retired, non-English speaker, hearing problems); 1,364 (20.9%) did not wish to participate; and 2316 (35.6%) were not contacted before the cohort closed. Service members deployed at the time of initial contact were called again following return from deployment. The final baseline study sample included 2616 OHARNG soldiers, men and women aged 17 years or older, who were capable of informed consent. Soldiers could have been 17 as opposed to 18 because emancipated minors are considered adults by the military. The overall cooperation rate (defined as number consented divided by number of successfully contacted numbers)[31] was 68.2%, and the response rate was 43.2% (defined as those who completed the survey plus those who consented but were ineligible, dived by the number of working numbers minus those disqualified)[31]. A second wave of interviews was conducted 1 year later and included 1,767 (67.5%) of those interviewed at baseline, of whom we excluded 183 (10.4%) with a lifetime history of suicidal ideation at baseline. Our final analytic sample included 1582 soldiers.

Participants were administered a 60-minute telephone survey, including questions on military history and experiences, health status and medical history, social support and psychosocial resources, deployment-related and civilian-related trauma and psychopathology, mental health service use, health risk behavior and demographic characteristics. Interviewers obtained consent from each participant at the start of the interview and offered financial compensation for their time. All human subjects procedures were overseen and approved by the Institutional Review Boards of Columbia University, Case Western Reserve, University of Toledo, University of Michigan and the Department of Defense.

Traumas

Deployment-related traumatic event experiences were identified with a list of 21 traumatic events specific to military service [32] and asked in reference to most recent deployment. Traumatic event experiences that may have occurred either within or outside of military deployment experiences were additionally identified using a list of 25 traumatic events [33]; if participants endorsed any of these events, they were asked whether the event(s) happened in relation to their most recent deployment. Therefore, we assessed for a total of 46 events related to most recent deployment and 25 events that were not.

Diagnoses

We used the Patient Health Questionnaire (PHQ-8 version) [34] to evaluate depression symptoms. Each of the eight questions asked about symptom frequency within the prior 2 weeks, and possible responses included 0 (not at all), 1 (several days), 2 (more than half the days) and 3 (nearly every day), with total scores ranging from 0 to 24 [35]. We added questions to assess timing, duration, severity of illness, and disability resulting from symptoms. Major depressive disorder (MDD) was assessed using DSM-IV criteria [36] and considered present if five or more of the eight depressive symptoms were present at least “more than half the days” in the past two weeks, and one of the symptoms was depressed mood or anhedonia. Other depressive disorder (ODD) was considered present if 2–4 depressive symptoms were present at least “more than half the days” in the past two weeks, and one of the symptoms was depressed mood or anhedonia. We identified depression cases as those individuals who met criteria for either MDD or ODD and reported that his/her symptoms seemed to have occurred together. We included questions to assess the duration and grouping of symptoms. Based on a validation testing of 500 individuals who completed both a telephone and clinical sample interview, the best validity for MDD and ODD excluded the impairment question [37]. Thus, we did not include impairment as a requirement for the diagnosis. Our telephone-based definition of depression is highly specific (83%) and moderately sensitive (51%) [37].

Alcohol dependence was assessed using the Mini-International Neuropsychiatric Interview (MINI) [38] based on DSM IV criteria [36]. Alcohol dependence was diagnosed among those who met at least three of seven substance dependence criteria that include tolerance (criteria 1), withdrawal (criteria 2), and compulsive use (criteria 3–7). Participants were asked whether these symptoms had occurred within the past month and year. Our telephone-based alcohol dependence definition is highly specific (81%) and moderately sensitive (60%) [37].

Suicidal ideation was assessed using a question from the PHQ-9 [35], which asks participants about “thoughts that you would be better off dead or of hurting yourself in some way, within the past 30-days?” and “within the past year?” For the purposes of this investigation, we were interested in predicting incident or new-onset suicidal ideation at wave 2, within the prior year. Accordingly, this question was omitted from our measure of depression to avoid overlapping constructs within our statistical model. Our telephone-based suicidal ideation item was highly specific (0.87), with low sensitivity (0.32) [37].

Statistical Analyses

All analyses were weighted to (1) calibrate our estimates to the distribution of the total population of the OHARNG at time of sampling, and (2) adjust for predictors of unit non-response at one-year follow-up. Chi-squared tests were used to examine the prevalence of incident suicidal ideation at one-year follow-up, within strata of our independent variables. We used logistic regression models to illustrate bivariate relationships between alcohol dependence, depression, and other putative predictors of incident suicidal ideation such as demographics, military rank, lifetime trauma, and past year trauma. Missing values are denoted with asterisks in table 1. We examined interaction between alcohol and depression in predicting suicidal ideation by examining relative measures- odds ratios and relative risks of suicidal ideation for those with alcohol dependence with and without depression, and with absolute measures - by plotting conditional risks of suicidal ideation across interaction strata. Finally, we quantitatively examined multiplicative and additive interaction between depression and alcohol dependence using a cross-product term, and by calculating the relative excess risk of interaction (RERI), the Attributable Proportion due to Synergy (AP), and the Synergy Index (SI). Statistical analyses were performed with SUDAAN 11 [39] and SAS 9.4 [40].

Table 1.

Characteristics of study sample of Ohio Army National Guard (n=1582)

CHARACTERISTICS	Unweighted Ns	Weighted % (95% CI)
Current Alcohol Dependence (n=1582)
No	1489	93.61(92.16–94.81)
Yes	93	6.39(5.19–7.84)
Current Depression (n=1582)
No	1418	89.50(87.78–91.01)
Yes	164	10.50(8.99–12.22)
Gender (n=1582)
Male	1367	86.29(84.40–87.99)
Female	215	13.71(12.01–15.60)
Age (n=1582)
18–24	490	37.26(34.64–39.96)
25–34	519	34.35(31.87–36.92)
35–44	404	20.76(18.88–22.77)
45+	169	7.63(6.55–8.87)
Race (n=1581)*
White	1417	88.85(86.85–90.58)
Black	100	9.56(7.88–11.56)
Other	64	1.59(1.24–2.03)
Education (n=1582)
High School Graduate /GED or less	387	30.20(27.69–32.84)
Some College or Technical Training	763	48.16(45.53–50.81)
College/Graduate degree	432	21.64(19.72–23.69)
Household Income (n=1540)*
<$60,000	868	62.01(59.45–64.51)
>=$60,000	672	37.99(35.49–40.55)
Marital Status (n=1581)*
Married	798	40.01(37.56–42.51)
Divorced/Separated/Widowed	146	9.40(7.97–11.06)
Never Married	637	50.59(47.95–53.23)
Rank (n=1582)
Officer	236	9.77(8.59–11.10)
Enlisted, cadets, and civilian employees	1346	90.23(88.90–91.41)
Number of lifetime deployments (n=1582)
0	539	39.16(36.53–41.84)
1–4	1005	58.86(56.17–61.49)
5+	38	1.99(1.44–2.74)
Number of lifetime traumatic events at Wave 1 (n=1582)
0–2	297	20.19(18.10–22.46)
3–6	451	29.37(27.00–31.86)
7–11	409	24.97(22.76–27.31)
12+	425	25.47(23.28–27.28)
Number of traumatic events between Wave 1 & Wave 2 (n=1582)
0	781	49.69(47.05–52.34)
1	346	21.55(19.47–23.78)
2	182	11.53(9.96–13.32)
3+	273	17.23(15.31–19.33)
Interaction Indicator Variable (n=1582)
Neither Condition	1352	84.85(82.84–86.66)
Alcohol Dependence Alone	66	4.65(3.62–5.95)
Depression Alone	137	8.76(7.38–10.38)
Alcohol Dependence and Depression	27	1.74(1.18–2.55)

^*Denotes missing values

RESULTS

As shown in Table 1, the final study population was predominantly male (86.3%) and white (88.9%). Approximately half had some college or technical training (48.2%) and nearly half were married (40%); median age of soldiers was between 18–24 (37.3%). Most soldiers were enlisted personnel (90.2%), had 1–4 deployments (58.9%), 3–6 lifetime traumatic events at baseline (29.4%), and 0 traumatic events between baseline and follow-up (49.7%). At baseline, 6.4% of our sample had current alcohol dependence, and 10.5% had current depression.

Incident suicidal ideation was present among 42 or 2.47% (1.68–3.26) of soldiers at follow-up. Table 1 shows that suicidal ideation at wave 2 was associated with alcohol dependence at baseline (8.97%; p=0.03), and depression at baseline (8.09%; p<0.001). In unadjusted logistic regression models, alcohol dependence (odds ratio [OR]=4.76; 95% confidence interval [CI] =2.05–11.06) and current depression (OR=4.77; 95% CI = 2.37–9.63) were both associated with incident suicidal ideation at a similar magnitude (table 2). None of the other predictors evaluated in table 2 were associated with our outcome at a statistical significance level of p<0.05.

Table 2.

Bivariable associations between study sample characteristics and suicidal ideation in the Ohio Army National Guard (n=1582).

CHARACTERISTICS	Unweighted Frequency of events	Weighted % (95% CI)	χ2, p-value	Crude Weighted OR (95%CI)
Current Alcohol Dependence (n=1582)			χ2 (df=1)=4.54; p=0.03
No	34	2.03(1.42–2.89)		1
Yes	8	8.97(4.41–17.38)		4.76(2.05–11.06)
Current Depression (n=1582)			χ2(df=1)=7.96; p<0.001
No	28	1.81(1.21–2.70)		1
Yes	14	8.09(4.74–13.48)		4.77(2.37–9.63)
Gender (n=1582)			χ2(df=1)=0.16, p=0.69
Male	34	2.41(1.68–3.43)		1
Female	8	2.86(1.37–5.87)		1.20(0.52–2.75)
Age (n=1582)			χ2(df=3)=0.26, p=0.86
18–24	15	2.69(1.58–4.56)		1
25–34	11	2.00(1.07–3.74)		0.74(0.32–1.71)
35–44	11	2.73(1.46–5.02)		1.01(0.44–2.34)
45+	5	2.79(1.15–6.60)		1.04(0.36–2.97)
Race (n=1581)*			χ2(df=2)=3.15, p=0.21
White	38	2.60(1.86–3.62)		1
Black	1	0.95(0.13–6.46)		0.36(0.05–2.67)
Other	3	4.57(1.47–13.35)		1.80(0.53–6.07)
Education (n=1582)			χ2(df=2)=2.24, p=0.11
High School Graduate /GED or less	6	1.34(0.59–3.01)		1
Some College or Technical Training	24	3.04(1.98–4.63)		2.31(0.91–5.90)
College/Graduate degree	12	2.78(1.53–5.03)		2.11(0.75–5.93)
Income (n=1540)*			χ2(df=3)=3.55, p=0.06
<$60,000	29	3.07(2.09–4.47)		1.94(0.93–4.06)
>=$60,000	12	1.60(0.86–2.95)		1
Marital Status (n=1581)*			χ2(df=3)=0.59, p=0.74
Married	21	2.54(1.63–3.95)		1
Divorced/Separated/Widowed	6	3.43(1.46–7.84)		1.36(0.51–3.65)
Never Married	15	2.24(1.32–3.76)		0.88(0.43–1.77)
Rank (n=1582)			χ2(df=1)=0.24, p=0.62
Officer	5	2.02(0.84–4.81)		1
Enlisted, cadets, and civilian employees	37	2.52(1.79–3.53)		1.25(0.48–3.27)
Number of lifetime deployments (n=1582)			χ2(df=2)=0.81, p=0.44
0	11	2.11(1.14–3.87)		1
1–4	28	2.53(1.71–3.74)		1.21(0.57–2.54)
5+	3	7.78(2.41–22.40)		3.92(0.99–15.58)
Number of lifetime traumatic events at Wave 1 (n=1582)			χ2(df=3)=1.02, p=0.38
0–2	7	2.10(0.91–4.80)		1
3–6	9	1.63(0.82–3.22)		0.77(0.26–2.32)
7–11	12	3.01(1.67–5.36)		1.44(0.51–4.11)
12+	14	3.21(1.85–5.51)		1.54(0.55–4.30)
Number of traumatic events between Wave 1 & Wave 2 (n=1582)			χ2(df=2)=2.49, p=0.06
0	11	1.37(0.72–2.56)		1
1	13	3.99(1.90–5.98)		2.54(1.06–6.08)
2	8	3.93(1.86–8.11)		2.96(1.09–8.04)
3+	10	3.52(1.84–6.64)		2.64(1.05–6.65)
Interaction Indicator Variable (n=1582)			χ2(df=3)=3.43, p=0.02)
Neither Condition	24	1.59(1.03–2.44)		1
Alcohol Dependence Alone	4	5.85(2.04–15.66)		3.85(1.18–12.52)
Depression Alone	10	6.26(3.30–11.56)		4.14(1.86–9.22)
Alcohol Dependence and Depression	4	17.31(6.66–38.06)		12.97(4.05–41.47)

To evaluate and compare additive vs. multiplicative interaction, we first calculated unadjusted odds ratios, risk ratios, and risk differences for those without vs. with alcohol dependence, across strata of depression status. Odds ratios (ORs) for suicidal ideation among those with vs. without alcohol dependence were similar among non-depressed [OR=3.85 (95% CI = 1.18–12.52)] and depressed individuals [OR = 3.13 (95% CI = 0.88–11.14)]. Risk ratios (RR) for suicidal ideation among those with vs. without alcohol dependence were similar among non-depressed [RR=2.68 (95% Confidence Intervals(CI) = 1.21–11.26] and depressed individuals [RR = 2.76 (95% CI = 0.93–8.22)]. Risk differences (RD) for suicidal ideation among those with vs. without alcohol dependence diverged for those without depression [RD = 0.04 (95% CI = 0.02–0.07)] compared to those with depression [RD 0.11(95% CI=0.06–0.18)].

Formal and descriptive analyses of interaction are presented in Figure 1. To descriptively characterize the risk of suicidal ideation within interaction strata, we calculated conditional risks of incident suicidal ideation in those with neither condition (0.02; 95% CI: 0.01–0.02), with alcohol dependence but no depression (0.06; 95% CI: 0.02–0.16), with depression but no alcohol dependence (0.06; 95% CI: 0.03–0.12), and with both conditions (0.17; 95% CI: 0.07–0.38). To assess multiplicative interaction, we fit a multivariable logistic model including both alcohol dependence (adjusted odds ratio [aOR]=3.51; 95% CI = 1.45–8.49) and depression (aOR=3.93; 95% CI = 1.90–8.12); we tested a cross-product term for alcohol dependence and depression in a multivariable logistic model including both conditions, and this term was statistically non-significant (beta coefficient=−0.21, p=0.82), suggesting absence of multiplicative interaction. To test additive interaction we computed the RERI, AP, and S for alcohol dependence and depression, finding strong evidence of interaction on the additive scale (RERI = 5.978(95% CI: 0.364–11.591), AP = 0.461(95% CI: 0.199–0.723), and S = 1.998(95% CI: 1.163–3.434)).

Figure 1.

Assessing Interaction Between Alcohol Dependence And Depression In Predicting Incident Suicidal Ideation

Note: All measures reported in figure 1 are weighted, but otherwise unadjusted.

DISCUSSION

Using a representative sample of Ohio Army National Guard soldiers, we found that alcohol dependence and depression work together to place a soldier at greater risk of subsequent suicidal ideation. Each condition alone was associated with risk, but the presence of both conditions together had a stronger effect than the summed risks of each condition alone, providing evidence of positive additive interaction. This has important implications for screening of soldiers and may ultimately improve our efforts to reduce suicide rates among reservists.

Although no other studies to our knowledge have examined the link between coincident alcohol dependence and depression on incident suicidal ideation in the military prospectively, our key findings are in line with findings in other military samples that have examined the links between each condition individually. Our finding of an association between alcohol dependence and suicidal ideation is consistent with that of a prior National Guard sample that demonstrated this association using crosssectional data [41]. Our finding of an association between depression and suicidal ideation is consistent with findings in mixed component samples of OEF/OIF veterans [18,20] and with findings in the 2008–2014 Department of Defense Suicide Event Reports [9,10,7]. Griffth [11], who examined suicide in a national sample of the Army National Guard using suicide data spanning from 2007–2010, found that an alcohol abuse history was prevalent among 12–17% of suicide cases, but did not specifically examine alcohol dependence.

Our results are consistent with an extensive civilian population literature demonstrating the link between alcohol dependence and suicidality, both in cross-sectional and prospective [42,26,43], and the well-established link between depression and suicidality [44]. Indeed, Kessler and colleagues [15] found that both alcohol dependence and depression were independently associated with suicidal ideation and suicide attempts in cross-sectional data from a representative sample of the general US population. Finally, our findings of positive additive interaction between alcohol dependence and depression in suicidal ideation risk are consistent with civilian studies [27,16] demonstrating an increased risk of suicidal ideation and behavior for those with both conditions. The RERI (part of the total effect due to interaction) of 5.978 confirms positive additive interaction as the point estimate and its confidence intervals exceed 0; the AP of 0.461 tells us that almost 50% of the risk of suicidal ideation among those with both conditions is attributable to the synergy between those conditions; the S (ratio of combined and individual effects) of 1.998 confirms positive additive interaction, as this point estimate and its confidence intervals exceed 1 [45,46].

Future work is needed to help understand component features of alcohol dependence that lead to suicidal thoughts among this population. For example, determining whether and how quantity of alcohol consumption or abuse modifies risk of suicidal ideation among depressed persons could inform targeted screening and assessment. Additionally, work is needed to understand the causal role of alcohol dependence in the progression from suicidal thoughts to suicide attempts and suicide. Finally, further work is needed to understand how depression and alcohol dependence work together, and how best to treat those with both conditions.

Whatever the specific mechanisms driving these associations, it is clearly important to target individuals with comorbid alcohol dependence and depression. While so-called “dual diagnosis” patients may be challenging to treat, this work suggests a need to find ways to reach and treat them. Evidence based treatments for patients with comorbid alcohol dependence and depression are available [47,48], and these findings argue that such treatments be further deployed and utilized for prevention and treatment.

Study limitations include (1) use of telephone interview techniques, a method with less diagnostic validity than clinical interviews and (2) lack of information on loss to follow-up due to suicide among wave 1 participants, which if present would lead to an underestimation of incident suicidal ideation. Despite the inherent limitations with use of telephone interview data, we are reassured about the observations drawn here based on a validation study that demonstrated acceptable sensitivity and specificity for the measures used [37]. Additionally, although loss to follow-up due to suicide is possible among some baseline respondents, this would likely account for only a small percentage of attrition given the low incidence of suicide, and any impact on our effect estimates would likely be negligible.

Notwithstanding these limitations, this study shows a temporal link between coincident alcohol dependence and depression and risk of suicidal ideation; and provides evidence of additive interaction or synergy in shaping risk of suicidal ideation. A high-risk suicide prevention strategy that targets military personnel with alcohol dependence, depression, and especially both conditions for screening of suicidal thoughts should receive special attention in suicide prevention efforts.