Projected rates of psychological disorders and suicidality among soldiers based on simulations of matched general population data

Abstract

Limited data are available on lifetime prevalence and age-of-onset distributions of psychological disorders and suicidal behaviors among Army personnel. We used simulation methods to approximate such estimates based on analysis of data from a U.S. national general population survey with the socio-demographic profile of U.S. Army personnel. Estimated lifetime prevalence of any DSM-IV anxiety, mood, behavior, or substance disorder in this sample was 53.1 percent (17.7 percent for mood disorders, 27.2 percent for anxiety disorders, 22.7 percent for behavior disorders, and 14.4 percent for substance disorders). The vast majority of cases had onsets prior to the expected age-of-enlistment if they were in the Army (91.6 percent). Lifetime prevalence was 14.2 percent for suicidal ideation, 5.4 percent for suicide plans, and 4.5 percent for suicide attempts. The proportion of estimated pre-enlistment onsets was between 68.4 percent (suicide plans) and 82.4 percent (suicidal ideation). Externalizing disorders with onsets prior to expected age-of-enlistment and internalizing disorders with onsets after expected age-of-enlistment significantly predicted post-enlistment suicide attempts, with population attributable risk proportions of 41.8 percent and 38.8 percent, respectively. Implications of these findings are discussed for interventions designed to screen, detect, and treat psychological disorders and suicidality in the Army.

INTRODUCTION

Despite considerable information from ongoing surveys about current prevalence of psychological disorders among U.S. Army personnel,^1,2 little is known about lifetime prevalence or about the proportion of lifetime cases that started prior to age-of-enlistment. This information could be useful for intervention planning purposes as soldiers are in the age range of onset of many psychological disorders.^3,4 In addition, data are unavailable on associations of lifetime psychological disorders with suicidality in the Army, a topic of considerable importance.⁵ Definitive data on these issues would require surveys that assessed lifetime disorders and retrospective age-of-onset reports to differentiate onsets prior to versus after enlistment. Such surveys are currently underway in the Army Study to Assess Risk and Resilience in Service members (Army STARRS; www.armystarrs.org), but it would also be useful to obtain comparable data for general population samples that have the same socio-demographic characteristics as Army personnel in order to understand the base rates of these disorders in the population among the kinds of people who are in the Army. A simulation study of this sort was carried out by Messer and colleagues⁶ using data collected in the early 1980s in the epidemiologic catchment area study,⁷ a large community epidemiological survey of psychological disorders in five U.S. communities. The authors weighted the epidemiologic catchment area data to have the same socio-demographic distribution as the Army and estimated lifetime prevalence of selected psychological disorders in that sample.

Building on Messer and colleagues, the current report presents results of analyses using more recent general population survey data to estimate lifetime prevalence and age-of-onset of DSMIV psychological disorders and suicidal behaviors (ideation, plans, and attempts) as well as associations of temporally primary psychological disorders with subsequent suicidal behaviors among people in the general population with the same socio-demographic characteristics as Army personnel. We distinguish between psychological disorders and suicidal behaviors that began prior to and after expected age-of-enlistment in this civilian sample by developing a prediction equation that imputes expected age-of-enlistment from Army administrative data to each civilian in the sample. The survey data used to make these projections come from the National Comorbidity Survey Replication (NCS-R),⁸ a national survey of prevalence and correlates of DSM-IV psychological disorders in the U.S. household population.

METHOD

The NCS-R sample

The NCS-R was a face-to-face household survey conducted between February 2001 and April 2003 in a national sample of U.S. adults (ages 18+ years). A two-part interview was used. Part one, which assessed core DSM-IV psychological disorders, was completed by all 9,282 respondents. Part two, which assessed additional disorders and correlates, was administered to all respondents who screened positive for a part one disorder (n = 4,235) plus a probability sub-sample of other respondents (n = 1,457). The response rate was 70.9 percent. Part one was weighted to adjust for differential probabilities of selection. Part two was additionally weighted to adjust for under-sampling respondents with no part one disorder. A final post-stratification weight was used to match part two with the 2000 Census on a variety of socio-demographic and geographic variables. NCS-R sampling, field, and weighting procedures are discussed in more detail elsewhere.⁹

Sample matching

We sub-sampled part two NCS-R respondents and then weighted to create a sub-sample that matched active duty Army personnel on a range of socio-demographic variables. Sub-sampling began by limiting NCS-R respondents to those employed in the age range 18-65 years having at least a high school education or GED and health insurance to approximate the broad profile of Army personnel. We then excluded NCS-R respondents who would be ineligible for Army service due to: (1) conviction of a felony or serving at least one year in prison; (2) handicaps, including deafness, blindness, paralysis, and a missing limb; (3) chronic physical disorders, including cardiovascular disorders (heart attack, stroke, hypertension, heart disease), respiratory disorders (COPD, asthma), diabetes, ulcer, HIV or AIDS, epilepsy or seizure disorder, Crohn's disease, cancer (except skin cancer), severe migraines, and extreme obesity; and (4) severe psychological disorders, including schizophrenia, other non-affective psychoses, bipolar I disorder, and serious suicide attempts that occurred before the estimated age-of-enlistment. These exclusions are over-inclusive in that they remove people who might have entered the Army with waivers or developed chronic conditions after enlistment.

Once the sample was restricted in these ways, we selected a series of eight weighting variables available in the NCS-R as well as in the DMDC Master Personnel dataset for Army personnel who were on active duty in December 2007. This information was obtained as part of the Total Army Injury and Health Outcomes Database (TAIHOD), an integrated database constructed by the Injury Epidemiology Research Section, Military Performance Division, of the U. S. Army Research Institute of Environmental Medicine, Natick, MA. These eight variables were ones that we knew both to be significantly related to psychological disorders and to have a significantly different distribution among Army personnel than in the general population. Cross-tabulation of these variables from the TAIHOD dataset for active duty Army personnel in December 2007 was used for weighting. The variable distributions in this dataset were calculated as the person-month level for all Soldiers in the regular Army who were on active duty at any time during this five-year period. The eight weighting variables were age in years, sex, race-ethnicity (Non-Hispanic Black, Non-Hispanic White, Hispanic, and all others), education (high school graduates including those with a GED, some post-high school education without a bachelor's degree, and bachelor's degree or more education), marital status (married, never married, and previously married, where the latter included separated, divorced, and widowed), citizenship, nativity (born in the U.S. yes/no), and religion (Protestantism, Catholicism, Judaism, Eastern, Other, and Atheist/No Religion). The NCS-R weights were generated by using an exponential weighting function to make the distributions of the eight weighting variables in the adjusted NCS-R sample agree with the distributions in the TAIHOD dataset. (Detailed information is available on request.)

Measures

Psychological disorders

DSM-IV psychological disorders were assessed with the World Health Organization (WHO) Composite International Diagnostic Interview (CIDI),¹⁰ a fully-structured lay-administered interview that generates diagnoses for commonly occurring psychological disorders. The lifetime disorders considered include mood disorders (major depressive episode, dysthymic disorder, bipolar disorder), anxiety disorders (generalized anxiety disorder, panic disorder, agoraphobia without panic disorder, specific phobia, social phobia, post-traumatic stress disorder, separation anxiety disorder), behavior disorders (attention-deficit/hyperactivity disorder, oppositional-defiant disorder, conduct disorder, intermittent explosive disorder [IED]), and substance disorders (alcohol and drug abuse with and without dependence, alcohol and drug dependence with a history of abuse). Good concordance was found between CIDI diagnoses and blinded clinical assessments in an NCS-R clinical reappraisal study.¹¹ All DSM-IV/CIDI disorders were defined with organic exclusions but without diagnostic hierarchy rules. Each disorder module in the WHO-CIDI yields indices of lifetime history, age at first onset, and presence of the disorder in the past year. Diagnoses were assigned based on reports of symptoms, duration, and intensity as specified in the DSM-IV.

Suicidal behaviors

All part two NCS-R respondents were asked if they had ever seriously thought about killing themselves and, if so, the age when they first had these thoughts. Those who reported suicidal ideation were then asked if they ever made a suicide plan (and, if so, their age when they first did), and if they ever made a suicide attempt (and, if so, their age when they first did). Lifetime attempters were then asked additional questions about the lethality of their intent in their first lifetime attempt by presenting three statements about intent and asking them to say which statement most closely approximated their situation at the time of their first suicide attempt. The three statements were: (A) I made a serious attempt to kill myself and it was only luck that I did not succeed; (B) I tried to kill myself, but knew that the method was not fool-proof; and (C) My attempt was a cry for help. I did not intend to die. As noted above, all NCS-R respondents who endorsed Statement A for their first lifetime suicide attempt and reported that this attempt occurred at an earlier age than their imputed age-of-enlistment were excluded from the sample on the grounds that these attempts should be discovered in the course of recruitment screening and, if so, would be deemed ineligible for military service. We did not exclude people from the sample, though, who reported attempts and endorsed Statements B or C.

Imputed age-of-enlistment

Regression-based imputation was used to assign an expected ageof-enlistment to each NCS-R respondent. This was done by estimating a multiple regression equation using the weighting variables from the TAIHOD dataset to predict age-of-enlistment in the TAIHOD data. The regression coefficients from that equation were then applied to the NCS R dataset to impute individual-level estimates of age-of-enlistment to match the TAIHOD distribution conditional on the matching variables.

Analysis methods

We began by analyzing the weighted NCS-R data to estimate lifetime prevalence and age-of-onset of each DSM-IV/CIDI disorder. Estimated age-of-enlistment data were then used to estimate the proportion of all such disorders that began before the respondent would be expected to enlist. We then repeated this exercise for 12-month disorders to determine both the expected prevalence in the past year and the proportion of 12-month cases that first occurred prior to expected age-of-enlistment. We next repeated this exercise for lifetime and 12-month suicidal behaviors, bearing in mind that respondents with serious pre-enlistment suicide attempts were excluded from the sample by definition.

Once these descriptive analyses were completed, we carried out a discrete-time survival analysis with person-year as the unit of analysis in which prior lifetime psychological disorders were used to predict subsequent first onset of suicide attempts among respondents who never made an attempt prior to expected age-of-enlistment. In carrying out this analysis we distinguished between internalizing disorders (mood disorders, anxiety disorders, and IED) and externalizing disorders (the behavior disorders [with the exception of IED] and substance disorders), as these are the two broad groups of common mental disorders most often found to co-occur in the epidemiological literature.¹² IED was included with internalizing disorders because preliminary analysis found it to be more strongly related to phobias than other behavioral disorders among young people. For each type of disorders, we distinguished between lifetime disorders that began prior to estimated age-of-enlistment and those that only began subsequent to estimated age-of-enlistment. These four predictor variables (i.e., number of internalizing and externalizing disorders that began prior to and subsequent to enlistment) were all included in the same prediction equation to predict suicide attempts that occurred in the years since estimated age-of-enlistment.

Finally, the population attributable risk proportion (PARP) was calculated for the predictors in the survival equation. A PARP is the proportion of observed outcomes (in this case suicide attempts after enlistment) that would not have occurred if the corresponding independent variable was eliminated and assuming the coefficients in the survival equation were due to causal effects of the predictors (e.g., the proportion of suicide attempts that would be prevented if internalizing disorders were eliminated/cured, assuming a causal association between internalizing disorders and suicide attempts). PARPs were calculated using simulation methods to generate individual-level predicted probabilities of the outcome disorders twice from the coefficients in the model: the first time using all the coefficients and the second time assuming that the coefficients associated with the focal predictors were all zero. The ratio of the predicted prevalence estimates in the two specifications was then used to calculate PARPs. This ratio simulates the potential effects of thorough detection and aggressive early treatment of disorders that cause suicide attempts among Army personnel.

RESULTS

Estimated lifetime prevalence of psychological disorders

As noted above, the NCS-R sample was restricted by excluding people with severe psychological disorders or a pre-enlistment history of a serious suicide attempt. Despite these restrictions, the weighted NCS-R data show that the expected lifetime prevalence of DSM-IV/CIDI disorders among people in the general population with socio-demographic profiles of Army personnel is quite high. We estimate that 53.1 percent of the people in the sample have a lifetime psychological disorder. These include an estimated lifetime prevalence of 17.7 percent for mood disorders, 27.2 percent for anxiety disorders, 22.7 percent for behavior disorders, and 14.4 percent for substance disorders. (Table I) In addition, the sum of disorder-specific prevalence estimates (127.2 percent) is more than twice as high as 53.1 percent, indicating that comorbidity of psychological disorders would be expected to be the norm. Perhaps as striking as the high prevalence of psychological disorders is the estimate from the simulated data that the vast majority (91.6 percent) of people with a history of psychological disorders would already have these disorders at the imputed time of enlistment. This proportion is estimated to range between a low of 53.0 percent for dysthymic disorder to a high of 100 percent for the trio of behavior disorders that typically begin in childhood (oppositional-defiant disorder, conduct disorder, attention-deficit/hyperactivity disorder), and to average 69.7 percent for mood disorders, 90.7 percent for anxiety disorders, 97.5 percent for behavior disorders, and 77.7 percent for substance disorders. These results suggest that although seriously impairing illnesses are both exclusions from entry and causes for discharge from the Army, pre-enlistment history of non-serious psychological disorders such as depression, anxiety, and behavioral disorders is likely high among Army personnel.

Table I.

Simulated lifetime prevalence estimates of DSM-IV/CIDI disorders prior and subsequent to Army enlistment in the sub-sample of NCS-R respondents weighted to approximate the population of active duty Army personnel (n = 1785)^a

	Current lifetime prevalence % (SE)	Lifetime prevalence of a disorder that began prior to enlistment % (SE)	Lifetime prevalence of a disorder that began only after enlistment % (SE)	Proportion of lifetime prevalence due to a disorder that began prior to enlistment % (SE)
I. Mood disorders
Major depressive episode	16.0 (3.1)	11.1 (2.7)	5.5 (1.7)	69.4 (8.6)
Dysthymic disorder	3.6 (1.8)	1.9 (1.0)	1.7 (1.5)	53.0 (26.0)
Bipolar II disorder	0.4 (0.1)	0.3 (0.1)	0.1 (0.1)	68.6 (15.9)
Sub-threshold bipolar disorder	2.8 (1.1)	2.0 (1.1)	0.8 (0.3)	72.1 (13.4)
Any mood disorder	17.7 (2.9)	12.3 (2.5)	6.1 (1.8)	69.7 (7.4)
II. Anxiety disorders
Generalized anxiety disorder	4.3 (1.6)	3.1 (1.6)	1.2 (0.3)	72.2 (11.5)
Agoraphobia without panic disorder	0.4 (0.1)	0.4 (0.1)	0.0 (0.0)	95.5 (3.3)
Social phobia	11.1 (2.0)	10.7 (2.0)	0.5 (0.4)	95.9 (2.9)
Specific phobia	8.8 (1.5)	8.5 (1.5)	0.4 (0.3)	95.7 (2.9)
Post-traumatic stress disorder	3.7 (1.7)	3.1 (1.7)	0.6 (0.3)	83.2 (9.7)
Separation anxiety disorder	9.9 (2.3)	8.1 (2.2)	2.0 (0.6)	81.9 (6.2)
Panic disorder	3.3 (0.9)	2.5 (0.9)	0.9 (0.2)	74.9 (9.5)
Any anxiety disorder	27.2 (3.6)	24.7 (3.7)	3.4 (0.8)	90.7 (2.7)
III. Behavior disorders
Oppositional-defiant disorder	6.0 (1.1)	6.0 (1.1)	. .	100.0 (0.0)
Intermittent explosive disorder	12.2 (2.0)	11.5 (1.9)	0.8 (0.4)	94.2 (2.8)
Attention-deficit/hyperactivity disorder	6.1 (1.4)	6.1 (1.4)	. .	100.0 (0.0)
Conduct disorder	9.2 (2.7)	9.2 (2.7)	. .	100.0 (0.0)
Any behavior disorder	22.7 (3.5)	22.1 (3.4)	0.7 (0.5)	97.5 (1.5)
IV. Substance disorders
Alcohol abuse	13.4 (2.0)	9.8 (1.8)	3.9 (1.2)	73.5 (6.8)
Alcohol dependence	4.8 (1.4)	3.4 (1.2)	1.5 (0.6)	71.0 (10.4)
Drug abuse	9.0 (1.8)	7.9 (1.5)	1.1 (0.6)	88.4 (4.7)
Drug dependence	2.2 (0.7)	1.5 (0.5)	0.7 (0.5)	69.3 (17.6)
Any substance disorder	14.4 (2.2)	11.2 (1.8)	3.6 (1.1)	77.7 (5.6)
V. Any disorder
Any disorder	53.1 (4.5)	48.6 (4.4)	8.7 (1.7)	91.6 (1.5)

^aRespondents were all ages 18-65 at the time of interview, had at least a high school education, and were employed with health insurance to match the broad socio-demographic bound of Army personnel. All NCS-R respondents who were ever convicted of a felony or served at least one year in prison were excluded from the sample. All NCS-R respondents with handicaps or physical or psychological disorders that would normally lead to rejection from Army enlistment or discharge were excluded from the sample. The handicaps included deafness, blindness, paralysis (of one or both arms, legs, or sides of the body), and a missing limb (hand, foot, arm, leg). The physical disorders included cardiovascular disorders (heart attack, stroke, hypertension, heart disease), respiratory disorders (COPD, asthma), diabetes, ulcer, HIV or AIDS, epilepsy or seizure disorder, Crohn's disease, cancer (except skin cancer), severe migraines, and extreme obesity. The psychological disorders included schizophrenia and other non-affective psychoses, bipolar I disorder, and serious suicide attempts that occurred before the imputed age-of-enlistment.

Turning to psychological disorders present in the past year, we find that the estimated prevalence based on general population distributions remains high. (Table II) Indeed, 27.0 percent of Army personnel would be expected to have one or more active psychological disorders in the past year based on the prevalence among people with the same socio-demographic profile in the civilian population. We estimate 12-month prevalence to be 8.1 percent for mood disorders, 14.5 percent for anxiety disorders, 7.0 percent for behavior disorders, and 5.9 percent for substance disorders.

Table II.

Simulated 12-month prevalence estimates of DSM-IV/CIDI disorders prior and subsequent to Army enlistment in the sub-sample of NCS-R respondents weighted to approximate the population of active duty Army personnel (n = 1785)^a

	12-month prevalence % (SE)	12-month prevalence of a disorder that began prior to enlistment % (SE)	12-month prevalence of a disorder that began only after enlistment % (SE)	Proportion of 12-month prevalence due to a disorder that began prior to enlistment % (SE)
I. Mood disorders
Major depressive episode	6.2 (2.1)	4.9 (2.1)	1.3 (0.3)	79.3 (8.2)
Dysthymic disorder	1.7 (1.0)	1.7 (1.0)	0.0 (0.0)	97.6 (1.8)
Bipolar II disorder	0.3 (0.1)	0.2 (0.1)	0.1 (0.1)	69.5 (18.9)
Sub-threshold bipolar disorder	1.9 (1.0)	1.5 (1.0)	0.4 (0.2)	81.0 (13.4)
Any mood disorder	8.1 (2.2)	6.6 (2.2)	1.6 (0.4)	81.1 (6.5)
II. Anxiety disorders
Generalized anxiety disorder	3.0 (1.6)	2.3 (1.6)	0.7 (0.2)	78.6 (12.7)
Agoraphobia without panic disorder	0.3 (0.1)	0.3 (0.1)	0.0 (0.0)	98.4 (1.3)
Social phobia	6.1 (1.4)	6.1 (1.4)	0.0 (0.0)	99.9 (0.1)
Specific phobia	6.5 (1.4)	6.1 (1.4)	0.4 (0.3)	94.7 (3.7)
Post-traumatic stress disorder	2.3 (1.5)	1.8 (1.5)	0.5 (0.3)	79.1 (16.2)
Separation anxiety disorder	3.3 (1.4)	2.3 (1.2)	1.0 (0.8)	71.2 (20.1)
Panic disorder	1.5 (0.5)	1.1 (0.6)	0.4 (0.1)	77.0 (13.3)
Any anxiety disorder	14.5 (2.6)	13.2 (2.5)	1.6 (0.4)	90.7 (2.5)
III. Behavior disorders
Oppositional-defiant disorder	0.1 (0.1)	0.1 (0.1)	. .	100.0 (0.0)
Intermittent explosive disorder	5.2 (1.1)	4.7 (1.0)	0.5 (0.4)	90.8 (6.3)
Attention-deficit/hyperactivity disorder	2.2 (0.8)	2.2 (0.8)	. .	100.0 (0.0)
Conduct disorder	0.6 (0.5)	0.6 (0.5)	. .	100.0 (0.0)
Any behavior disorder	7.0 (1.4)	6.6 (1.3)	0.5 (0.4)	93.3 (4.8)
IV. Substance disorders
Alcohol abuse	4.3 (1.3)	3.0 (1.0)	1.4 (0.9)	68.4 (15.0)
Alcohol dependence	0.8 (0.3)	0.4 (0.3)	0.5 (0.2)	44.6 (19.8)
Drug abuse	2.5 (1.2)	1.9 (1.1)	0.6 (0.5)	76.0 (18.6)
Drug dependence	0.3 (0.3)	0.3 (0.3)	0.0 (0.0)	99.8 (0.2)
Any substance disorder	5.9 (1.6)	4.2 (1.3)	1.8 (1.0)	71.0 (13.1)
V. Any disorder
Any disorder	27.0 (3.1)	24.6 (3.1)	3.2 (0.9)	91.1 (2.5)

^aSee footnote 1 to Table I for a description of the sample.

As with the lifetime prevalence data, the vast majority of these past year cases (91.1 percent) would be expected to have had first onsets prior to expected age-of-enlistment. This includes an estimated 81.1 percent of those with a past-year mood disorder, 90.7 percent of those with a past-year anxiety disorder, 93.3 percent of those with a past-year behavior disorder, and 71.0 percent of those with a past-year substance disorder.

Despite the fact that the simulated sample excluded people who made a serious suicide attempt prior to their expected age-of-enlistment, 14.2 percent of the sample had a history of suicidal ideation, 5.4 percent had a history of making a suicide plan, and 4.5 percent made a suicide attempt. (Table III) Predicted prevalence in the past year is considerably lower (0.0-0.5 percent), but earlier lifetime history remains important here as well in that we estimate that 59.5 percent of all past-year ideation occurs to people who first began thinking of killing themselves prior to their expected age-of-enlistment. The parallel percentage estimates for past year suicide plans and attempts are 73.1 percent and 55.8 percent, respectively.

Table III.

Simulated lifetime and 12-month prevalence estimates of suicidal behaviors in the sub-sample of NCS-R respondents weighted to approximate the population of active duty Army personnel (n = 1785)^a

	Lifetime prevalence % (SE)	Proportion of lifetime cases with first onsets prior to enlistment % (SE)	12-month prevalence % (SE)	Proportion of 12-month cases with first onsets prior to enlistment % (SE)
Suicidal thought	14.2 (2.6)	82.4 (5.8)	0.5 (0.2)	59.5 (14.2)
Suicide plan	5.4 (1.8)	68.4 (14.0)	0.1 (0.1)	73.1 (18.8)
Suicide attempt	4.5 (1.5)	72.8b (12.7)	0.0 (0.0)	55.8 (34.3)

^aSee footnote 1 to Table 1 for a description of the sample.

^bAlthough the sample excluded respondents who made a serious suicide attempt (defined as an attempt in which the individual seriously intended to die and it was only because of luck that he or she did not die) prior to the imputed age-of-enlistment, those who had less serious attempts were not excluded. The 72.8% of lifetime attempters whose first attempt occurred prior to enlistment are made up exclusively of individuals whose first attempt was classified as not at the highest level of severity in the NCS-R classification scheme.

Survival analysis shows that both internalizing and externalizing DSM-IV/CIDI psychological disorders are powerful predictors of suicide attempts during the years subsequent to expected age-of-enlistment. (Table IV) In the case of internalizing disorders, though, it is only those that began subsequent to expected age-of-enlistment that significantly predict post-enlistment suicide attempts (odds ratio [OR] = 4.0), whereas internalizing disorders with onsets prior to expected age-of-enlistment do not significantly predict attempts that occurred in the years subsequent to expected age-of-enlistment (OR = 0.9). In the case of externalizing disorders, onsets that occurred prior to expected age-of-enlistment significantly predict subsequent suicide attempts (OR = 2.4). Subsequent onsets have an odds ratio of similar magnitude (OR = 3.0), but this association is not statistically significant due to the rarity of externalizing disorders that began after expected age-of-enlistment.

Table IV.

Associations of numbers of temporally primary DSM-IV/CIDI internalizing and externalizing disorders that occurred either before or after enlistment with subsequent post-enlistment suicide attempts in the sub-sample of NCS-R respondents weighted to approximate the population of active duty Army personnel (n = 1785)^a

	Timing of first onset of disorders
	Before enlistment OR (95% CI)	After enlistment OR (95% CI)
Internalizingb	0.9 (0.5-1.8)	4.0* (2.3-7.2)
Externalizingb	2.4* (1.6-3.8)	3.0 (0.3-26.3)

^*Significant at the .05 level, two-sided test.

^aBased on a discrete-time survival model with person-year as the unit of analysis to predict first lifetime onset of a suicide attempt subsequent to the imputed age-of-enlistment from time-varying measures of number of temporally primary lifetime DSM-IV/CIDI internalizing and externalizing disorders. Controls include age, sex, education, person-year, and imputed age-of-enlistment. See footnote 1 to Table I for a description of the sample.

^bThe measures of internalizing disorders are scored in the range 0-9 and include the count of all mood disorders (major depressive episode or dysthymia and bipolar II or sub-threshold bipolar disorder), all anxiety disorders (generalized anxiety disorder, social phobia, specific phobia, agoraphobia or panic disorder, separation anxiety disorder, posttraumatic stress disorder) and intermittent explosive disorder. The measures of externalizing disorders are scored in the range 0-5 and include the count of three behavior disorders (conduct disorder, oppositional-defiant disorder and attention-deficit/hyperactivity disorder) and substance disorders (alcohol abuse or dependence and drug abuse or dependence). Four measures were included in the equation – two for the numbers of lifetime internalizing and externalizing disorders that had ages of onset prior to the individual's imputed age-of-enlistment and two others for comparable lifetime disorders that had ages of onset subsequent to age-of-enlistment.

As shown in the PARP calculations, nearly half of all suicide attempts (47.7 percent) in the simulated sample are associated with externalizing disorders, with a much higher proportion due to externalizing disorders that began prior to estimated age-of-enlistment (41.8 percent) rather than later (8.5 percent). (Table V) Though the association between estimated post-enlistment externalizing disorders and estimated post-enlistment suicide attempts was not statistically significant, it was included in this analysis because of the large OR. Estimated post-enlistment internalizing disorders, in comparison, are associated with 38.8 percent of estimated post-enlistment suicide attempts. All three groups of these psychological disorders considered together are associated with 70.6 percent of all estimated post-enlistment suicide attempts.

DISCUSSION

It is important to recognize that the prevalence estimates reported here are projections based on data from a general population survey. We did not obtain direct estimates of the prevalence of psychological disorders or suicidal behaviors from a representative sample of Army personnel. In addition, the survey data were collected a decade ago (2001-03) and the TAIHOD data were only available to us for a more recent time period (December 2007). These temporal issues limit the extent to which the data are directly relevant to today's Army. In addition, failure to control for the many unmeasured factors that might influence enlistment and retention in the Army and the onset of psychological disorders/suicidal behaviors could lead to bias in the extent to which the results represent the segment of the population that is comparable to the personnel in today's Army. In the absence of directly measured prevalence rates, though, these projections provide a useful approximation of the expected lifetime prevalence of psychological disorders and suicidal behaviors in the segment of the U.S. household population that most closely approximates members of the U.S. Army in terms of the matching and exclusionary variables considered here. These estimates show that we would expect high prevalence of psychological disorders and suicidal behavior in the Army merely by virtue of the prevalence of these outcomes in the segment of the general population with the same matching-exclusion characteristics as Army personnel. The results also indicate that high proportions of both lifetime and past-year cases of psychological disorders would be expected to have onsets at an age earlier than the age of Army enlistment. The latter result is consistent with more general evidence from epidemiological surveys that most psychological disorders have early ages of onset.³ It is important to note that our lifetime prevalence estimates are likely conservative given our exclusion of individuals with serious psychological disorders and physical health problems given that at least some individuals with these sorts of conditions are likely to enter the Army with waivers or to develop the condition following enlistment.

The results also suggest that the vast majority of suicide attempts among people sociodemographically similar to Army personnel are expected to be associated with pre-enlistment psychological disorders. The NCS-R results are consistent with much other evidence in this regard in showing that the vast majority of suicidal behavior is linked to psychopathology.^13-17 Recent studies in military samples also indicate that psychiatric disorders are important risk factors for suicide among military personnel.^18,19 For example, Black et al.²⁰ reported that soldiers with a psychiatric diagnosis were 4.7 times more likely to commit suicide than soldiers without a psychiatric diagnosis, while Bachynski et al.²¹ reported that diagnoses of mood, anxiety, substance, and psychotic disorders were associated with more than a 5-fold increase in suicide mortality in the US Army. However, these studies did not establish the timing of disorder onset relative to age at enlistment.

To the extent that pre-enlistment mental disorders are as important for post-enlistment suicidality as suggested in our simulation, the question arises whether it might be possible to reduce Army suicides by restricting the enlistment of soldiers with a history of pre-enlistment psychological disorders. The PARP calculations suggest that the answer to this question might be a tentative “yes,” as removing externalizing disorders that began prior to enlistment would be associated with an expected 40 percent reduction of post-enlistment suicide attempts (assuming a causal relation between these constructs). However, there would be two serious practical difficulties in trying to exclude recruits with pre-enlistment externalizing disorders from Army service. The first is that it would be extraordinarily difficult to implement this policy because no archival data exist on the majority of these early-onset disorders and it is unrealistic to think that applicants would volunteer this information once it became known that it was a basis for rejection. The second is that the absolute risk of suicidal behavior is so small that the costs of this broad-gauged exclusion would far outweigh the benefits even if it were possible.

The issue of the cost-effectiveness of exclusion has been considered in recent years by corporate medical directors who are turning their attention to consideration of healthcare spending as a corporate human capital investment opportunity. It is becoming clear from these deliberations that the number of people with a history of psychological disorder is much too high to exclude all of them from the corporate workforce. However, the workplace costs of these individuals are substantial enough in terms of lost productivity that corporate medical directors are coming to believe that increased investment in detection, outreach, and best-practices treatment of these workers can have a positive return-on-investment (ROI) for the corporation.²² The same set of considerations might lead the Army to decide that targeted expansion of detection, outreach, and treatment efforts for soldiers with certain psychological disorders might have a positive ROI. Suicide prevention would be only one consideration, albeit an important one, in this line of reasoning.

Another important issue to consider is that the high rates of exposure to traumatic events experienced by Army personnel^23-26 almost certainly lead to higher rates of psychological disorders with post-enlistment onsets than would be expected in the general population. This probably means that a higher proportion of suicidality in the Army is linked to disorders with post-enlistment onsets than estimated in our simulation. This is only speculation, of course, as it is not possible to evaluate this issue in the NCS-R data. To the extent that it is true, though, efforts to limit enlistment of recruits with a history of psychological disorders would be even less productive than suggested above. This being the case, it is important to think more in terms of outreach and treatment of Soldier psychological problems than in terms of exclusion of people with psychological disorders from the Army.

When thinking about outreach and treatment of psychological disorder as a human capital investment opportunity, our results regarding post-enlistment internalizing disorders become especially important. As shown above, the vast majority of the internalizing disorders of people in the general population with socio-demographic profiles similar to those of Army personnel begin in childhood or adolescence, prior to the time soldiers enlist in the Army. Yet it is only the small proportion of internalizing disorders that begin in the years after expected age-of-enlistment that significantly predict subsequent suicide attempts. These internalizing disorders with expected post-enlistment onsets, which are estimated to be found in only 6-7 percent of the general population, are associated with 38.8 percent of all suicide attempts in the general population that occur after expected age-of-enlistment.

This is a substantially higher concentration of risk than that associated with externalizing disorders, as indicated by the higher odds ratio associated with expected post-enlistment internalizing disorders than with either expected pre- or post-enlistment externalizing disorders. Given that the simulations did not take into consideration the much higher risks of PTSD and other post-enlistment internalizing disorders associated with combat exposure² and other stressful experiences that are much more common among soldiers than the general population, it is likely that post-enlistment internalizing disorders in the Army actually account for an even higher proportion of suicide attempts among soldiers than suggested in our simulations. Reduction in the onset of these disorders through preventive interventions and increases in the rapid detection and treatment of these disorders once they occur have the potential to reduce the prevalence of Army suicidal behaviors substantially.

However, formidable challenges exist in developing preventive interventions for internalizing psychological disorders that begin post-enlistment in a population that is often suffering from comorbid conditions. One aspect of these challenges is that a substantial proportion of these disorders are likely to occur among people who have a history of at least one earlier psychological disorder that started in childhood or adolescence.⁴ In such a situation, secondary prevention – that is, the prevention of progression among people who already have a history of psychological illness – is much more the issue than primary prevention and this usually requires some form of treatment of the primary psychological disorder(s).²⁷ Treatment as a type of secondary prevention has several advantages over primary prevention, including the fact that people who already have disorders are often more willing to accept preventive interventions than people who have no disorder and that the technology of treatment interventions is more advanced than the technology of preventive interventions.²⁸ There also are barriers to treatment, though, that have to be addressed if treatment-oriented secondary preventive interventions are to succeed.²⁹ The Army has focused on addressing the problem of stigma in recent years³⁰ and is currently carrying out innovative programs to increase the quality of treatment of psychological disorders.³¹ The results of the current report suggest that these efforts will have the potential to be of greater importance in reducing Army suicidal behaviors if they focus on recent-onset internalizing disorders. However, caution is needed in interpreting this result too precisely in the absence of replication in a sample of Army personnel. As noted in the introduction, such a replication is currently underway in the Army Study to Assess Risk and Resilience in Service members (Army STARRS; www.armystarrs.org). It will be of great interest to determine from the Army STARRS data whether the patterns of lifetime prevalence, recent prevalence, and age-of-onset of psychological disorders and suicidality are similar or different among Army personnel to the patterns found in the simulated general population data.

Table V.

Population attributable risk proportion (PARP) of first suicide attempts occurring after enlistment associated with pre-enlistment and post-enlistment DSM-IV/CIDI internalizing and externalizing disorders in the sub-sample of NCS-R respondents weighted to approximate the population of active duty Army personnel (n = 1785)^a

	PARP (%) (SE)
Externalizing disorders with onset before enlistment	41.8 (10.8)
Externalizing disorders with onset after enlistment	8.5 (15.9)
Externalizing disorders with onset either before or after enlistment	47.7 (9.6)
Internalizing disorders with onset after enlistment	38.8 (15.5)
Externalizing disorders with onset before + internalizing after enlistment	67.0 (11.6)
Externalizing disorders with onset before or after + internalizing after enlistment	70.6 (11.6)

^aSee footnote 1 to Table I for a description of the sample.