Mood-, Anxiety-, and Substance Use Disorders and Suicide Risk in a Military Population Cohort

Abstract

There are meager prospective data from nonclinical samples on the link between anxiety disorders and suicide or the extent to which the association varies over time. We examined these issues in a cohort of 309,861 United States Air Force service members, with 227 suicides over follow-up. Mental disorder diagnoses including anxiety, mood, and substance use disorders (SUD) were based on treatment encounters. Risk for suicide associated with anxiety disorders were lower compared to mood disorders and similar to SUD. Moreover, the associations between mood and anxiety disorders with suicide were greatest within a year of treatment presentation.

Introduction

Suicide rates in the U.S. Department of Defense rose from 2005 until 2009 and have remained relatively flat since that time (Kinn et al., 2010). The U.S. Air Force (USAF) which had shown a decrease in rates since implementation of its population based suicide prevention program in 1996, also has observed an upswing in suicide rates (Kinn et al., 2010). Prior to this time, except for a brief period during 2004, suicide rates in the USAF remained low, with evidence of that effect being due to a suicide prevention program based on a public health strategy (Knox, Litts, Talcott, Feig, & Caine, 2003; Knox et al., 2010; McCarthy, Thompson, & Knox, 2012).

Mental disorders, most often mood and substance use disorders (SUD), are prevalent among suicide decedents and confer risk (Cavanagh, Carson, Sharpe, & Lawrie, 2003; Yoshimasu, Kiyohara, Miyashita, & The Stress Research Group of the Japanese Society for Hygiene, 2008). Indeed, mental disorders are associated with increased suicides in populations on par or greater than common social conditions such as unemployment and low educational achievement (Li, Page, Martin, & Taylor, 2011). Similarly, data from western military populations consistently show that mental disorders are associated with increased risk for suicide including studies of U.S. army service members (Bell, Harford, Amoroso, Hollander, & Kay, 2010) and Veterans of the armed services in the U.S. (Ilgen et al., 2010) and U.K. (Kapur, While, Blatchley, Bray, & Harrison, 2009). Accordingly, major western reports on suicide have noted the importance of preventing, detecting, and treating mental disorders as part of a comprehensive strategy for suicide prevention (U.S. Public Health Service, 1999; Institute of Medicine, 2002; U.S. Department of Defense, 2010).

Studies using large nonclinical population cohorts linked to mortality data are especially valuable for explicating the role of mental disorders in suicide. Advantages of use of this study design include: yielding large N's that are advantageous for statistical power, circumventing selection biases that often accompany sampling in suicide research, and avoidance of retrospective biases that are a challenge to the interpretation of postmortem studies. Data on mental disorders and suicide based on analyses of nonclinical population cohorts remains limited. However, recent years have been several informative reports from Denmark (Qin, Agerbo, & Mortensen, 2003; Qin & Nordentoft, 2005; Qin et al., 2006; Qin, 2011; Stenager & Qin, 2008) and Sweden (Reutfors et al., 2010). In recent years there have also been cohortstudies of suicide in military populations including U.S. army service members (Bell et al,. 2010) and Veterans of the U.K. armed services (Kapur et al., 2009).

Some important themes emerge from the database on mental disorders and suicide based on studies of population cohorts. First, suicide risk varies among mental disorders, with an empirical review concluding that mood disorders confer higher risk, SUD intermediate risk, and anxiety disorders lower risk in males (Li et al., 2011). Second, suicide risk is generally greatest within a year of diagnosis (or treatment presentation). Along these lines, a Danish study showing that mood disorders and SUD conferred greater risk for suicide within a year of treatment presentation than in subsequent years (Qin et al., 2006). A study of U.S. army service members showed that alcohol use disorders and non-alcohol use mental disorders conferred greater risk for suicide within a year of treatment presentation (Bell et al., 2010). Third, patterns related to timing may differ among mental disorders. A Danish study showed that mood disorders conferred greater risk for suicide than SUD within a year of hospitalization but not after one year (Qin & Nordentoft, 2005). A Swedish study demonstrated that risk for mood disorder peaked immediately following hospitalization and then lessened over the year whereas risk for suicide associated with SUD was fairly constant over the year (Reutfors et al., 2010). These data suggest that mood disorders may confer greater risk for suicide than SUD earlier but not later following a treatment episode.

We conducted analyses of USAF service members in order to compare the associations of mood, anxiety, and substance use disorders with suicide while considering time from initial treatment presentation to suicide. We tested three hypotheses. 1) We hypothesized a hierarchy of associations between mental disorders and suicide with mood disorders showing the greatest association, SUD intermediate, and anxiety disorders lowest. 2) Associations between each disorder and suicide are greater within a year of initial treatment presentation than in more than one year following initial presentation. 3) Mood disorders show a higher association with suicide than SUD or anxiety disorders within a year of presentation for treatment but a difference would not be observed at more than one year.

The current study breaks important new ground in three respects. First, there have been no systematic examinations of timing of mental disorders and suicide in a military cohort. Second, with rare exception, anxiety disorders have not been examined in nonclinical population cohort analyses of suicide (Qin, 2011) and there remains controversy about the role of anxiety disorders in suicide (Sareen, 2011). Third, analyses from Denmark and Sweden have provided most of the population level cohort data on mental disorders, timing, and suicide (Qin et al., 2003; Qin & Nordentoft, 2005; Qin et al., 2006; Qin, 2011; Reutfors et al., 2010; Qin & Mortensen, 2003). A limitation of these studies is that mental disorder diagnoses were based exclusively on psychiatric hospitalizations, creating a bias toward identification of the most severe disorders.

Method

Procedure

The study cohort included USAF personnel during October 2003-September 2009 time period. The data were derived from a distinctive longitudinal database that links individual level characteristics of all active duty USAF personnel with mental health service utilization and suicide. All data are de-identified prior to analysis and reside behind a secure firewall. In order to determine the state of the disorder episode (see the definition below) for every individual in the cohort that started in October 2003 we used utilization files going three years back, that is, October 2000-September 2009. An exception was the purchased outpatient care files which utilized files from October 2001 to September 2009. To find treatments for selected disorders we used USAF military medical care system databases for outpatient and inpatient care as well as USAF civilian inpatient and outpatient medical care system databases. We linked treatment data to the personnel file using individual identifiers that are de-identified prior to analyses. Airmen (includes both genders) are identified using a database for suicide in all active duty personnel obtained through the Program Manager of the USAF Suicide Prevention Program. This study received approval from the institutional review board at the University of Rochester and the Wilford Hall institutional review board for the USAF.

Measures

We used treatment encounters to identify events of mood, anxiety and SUD. We defined the beginning of a new episode of the disorder as a treatment encounter preceded by at least a two-year period of no treatment for this disorder. We defined the end of the episode as two years after the last documented treatment. A person's condition, relative to a specific episode, was characterized by one of three exclusive states: Someone was (1) not in an ongoing episode, (2) in the first year after onset of the episode, or (3) in the second or subsequent years. We used all sources of available information (i.e., outpatient and inpatient encounters in either military or civilian settings) to establish the presence and timing of a particular episode.

ICD-9 codes were used to identify treatments for the three mental disorders:

• mood disorder identified by codes for major depressive disorder, bipolar 1 disorder, dysthymic disorder, adjustment disorder with depressed mood, and depressive disorder not elsewhere classified (ICD-9 codes 296.0, 296.2-296.8, 300.4, 309.0, 311)

• • anxiety disorder identified by codes for anxiety states, phobic disorders, obsess-compulsive disorder, unspecified nonpsychotic disorders, post-traumatic stress disorder, and acute reactions to stress (ICD-9 codes 300.0, 300.2, 300.3, 300.9, 308.0-308.4, 308.9)

• • SUD identified by codes for alcohol or other substance intoxication, abuse or dependence (ICD-9 codes 303.0, 303.9, 304.0-304.9, 305.0, 305.2-305.9)

Analyses

To account for varying time of exposure and censoring we performed Cox proportional hazard survival model with three predictors including mood disorders, anxiety disorders, and SUD. We entered descriptive characteristics as additional explanatory time-varying (except gender) variables: age, gender, marital status, officer/enlisted status, duration of service in USAF, and USAF career group. USAF personnel serve one of nine major commands and we modeled major commands as fixed effects and used robust standard errors to account for clustering within bases. The beginning of the study was the analytical time origin in the Cox model, allowing the non-parametric baseline hazard function to account for temporal (calendar-time specific conditions) fluctuation in hazard of suicide. The three mental disorders variables were created as time-varying categorical variables with values 0: no ongoing episode, 1: first year of the episode, 2: second or subsequent years of episode (i.e., more than one year after the beginning of the episode). We also calculated unadjusted suicide rates and rate ratios by mood and anxiety status.

RESULTS

Descriptive Data

The cohort is described in Table 1 and included 309,861 USAF service members including 19.2% women with mean age 30.5 ± 7.7 years and mean duration of service 9.5 ± 7.7 years. Most service members were married (61.0%) and enlisted (80.0%) and worked in logistics (35.2%), support (24.9%), or operations (18.4%). Major commands ranged in size from 9,245 (3.0%) to 65,964 (21.3%) individuals.

Table 1.

U.S. Air Force Cohort Description at Study Midpoint (Oct 2006)

Age and Service (Years)	Mean (SD)
Age	30.5 (7.7)
Duration of service	9.5 (7.0)
Characteristics	N (%)
Marital status
Single or annulled	101,157 (32.6)
Married	188,868 (61.0)
Divorced, separated, or widowed	19,836 (6.4)
Gender
Men	250,256 (80.8)
Women	59,605 (19.2)
Rank
Enlisted	247,896 (80.0)
Officer	61,965 (20.0)
Career Group
Operations	57,107 (18.4)
Logistics	109,085 (35.2)
Support	77,022 (24.9)
Medical	28,080 (9.1)
Professional, acquisition, special investigations	14,338 (4.6)
Special duty identifiers and reporting identifiers	22,405 (7.2)
Other	1,824 (0.6)
All individuals	309,861 (100.0)

Service members were in one of nine major commands ranging in size from 9,245 (3.0%) to 65,964 (21.3%) individuals (not shown).

Descriptive data on suicide decedents (at time of death) and non-suicides (as of September, 2009) are presented in Table 2. There were 227 suicides over follow-up, with mood disorders most common (N=66, 29.1%) followed by anxiety disorders (N=42, 18.5%) and SUD (N=22, 9.7%). Most suicides with mood disorders and anxiety disorders occurred within a year of treatment presentation: Mood (N=49, 74.2%); Anxiety (N=32, 76.2%). There were an insufficient number of suicides with SUD to analyze them according to timing of presentation for treatment.

Table 2.

Descriptive Data on Mood-, Anxiety-, and Substance Use Disorders and Suicides in U.S. Air Force Cohort

	Suicides At Time of Death N (%)	Non Suicides As of Sep 2009 N (%)
Mood Disorder
Year 1	49 (21.6)	7,617 (2.7)
Year 2-plus	17 (7.5)	10,257 (3.7)
No Mood Disorder	161 (70.9)	261,541 (93.6)
Anxiety Disorder
Year 1	32 (14.1)	6,583 (2.7)
Year 2-plus	10 (4.4)	7,675 (2.8)
No Anxiety Disorder	185 (81.5)	265,157 (94.9)
Substance Use Disorder
Present (any year)	22 (9.7)	4,159 (1.5)
No Sub Use Disorder	205 (90.3)	275,256 (98.5)
Total	227 (100)	279,415 (100)

Mental Disorders and Suicide

We examined risk for suicide associated with mood, anxiety, and SUD in analyses that did not take into consideration the timing of presentation for these disorders. These results are shown in Table 3. Adjusted results (adjusted hazard ratio, 95% confidence interval, p-value) show that each mental disorder was associated with increased risk for suicide including mood disorder (AHR= 4.85, 95% CI= 3.43-6.85, p=0.000), anxiety disorder (AHR= 2.48, 95% CI = 1.70-3.62, p= 0.000), and SUD (AHR= 2.76, 95% CI=1.73-4.40, p=0.000).

Table 3.

Risk for Suicide Associated with Mental Disorders regardless of Timing of Treatment Presentation

Predictor		Unadjusted Hazard Ratios			Adjusted Hazard Ratios
		HR	95% CI	p-value	AHR	95% CI	p-value
Mood Disorder	Any year	4.035	2.856,5.700	<0.001	4.849	3.434,6.846	<0.001
	None (ref)	1.00	--	--	1.00	--	--
Anxiety Disorder	Any year	2.265	1.520,3.376	<0.001	2.479	1.695,3.624	<0.001
	None (ref)	1.00	--	--	1.00	--	--
Substance Use Disorder	Any year	3.554	2.234,5.655	<0.001	2.760	1.733, 4.395	<0.001
	None (ref)	1.00	--		1.00	--	--

Comparison		Unadjusted Hazard Ratios			Adjusted Hazard Ratios
		HR	95% CI	p-value	AHR	95% CI	p-value
Mood vs. Anxiety Disorder (ref)		1.781	0.936,3.391	0.079	1.956	1.024,3.737	0.042
Mood vs. Sub Use Disorder (ref)		1.135	0.603,2.137	0.694	1.757	0.904,3.416	0.097
Anxiety vs. Sub Use Disorder (ref)		0.637	0.334,1.217	0.173	0.898	0.502,1.606	0.717

Note. The top of the table shows risk associated with each mental disorder compared to not having the disorder; the bottom of the table compares risk associated with one mental disorder (e.g., mood disorder) compared to the risk associated with another mental disorder (e.g., substance use disorder). Unadjusted Hazard Ratios are derived from a model that includes all three mental disorders without covariates. Adjusted Hazard Ratios are also adjusted for covariates of sex, age, marital status, service duration, career group, officer-enlisted status, and major command.

We also compared risk for suicide associated with mood-, anxiety-, and SUD with one another (i.e., mood vs. anxiety, mood vs. SUD, anxiety vs. SUD). These results are also presented in Table 3 (bottom half of table). Adjusted results (adjusted hazard ratio, 95% confidence interval, p-value) show that mood disorder was associated with greater risk than anxiety disorder (AHR= 1.96, 95% CI= 1.02-3.74, p=0.042), mood disorder showed a nonsignificant trend (p<0.10) for greater association compared to SUD (AHR= 1.76, 95% CI = 0.90-3.42, p= 0.097), and risk associated with anxiety disorder did not differ from SUD at a statistically significant level (AHR= 0.90, 95% CI= 0.50-1.61, p=0.727).

Mental Disorders and Suicide Considering Treatment Timing

The remaining analyses took into consideration the timing of presentation for mood and anxiety disorders. These results are shown in Table 4. In the adjusted results, mood disorders were associated with increased risk for suicide within 1 year of treatment presentation (“mood disorder year 1”) and after 1 year (“mood disorder year 2+”). Specifically, adjusted results (adjusted hazard ratio, 95% confidence interval, p-value) showed mood disorder were associated with increased risk in year 1 (AHR= 7.26, 95%CI= 4.83, 10.90, p< 0.001) and year 2+ (AHR= 2.53, 95%CI= 1.43-4.47, p< 0.001). In adjusted results, anxiety disorder year 1 was associated with increased risk for suicide (AHR= 3.32, 95% CI= 2.06-5.37, p< 0.001) and anxiety disorder year 2+ was not associated with an increased risk at a statistically significant level (AHR= 1.50, 95% CI = 0.80-2.80, p= 0.207). In adjusted results, SUD in any year was associated with increased risk for suicide (AHR= 2.76, 95% CI= 1.74-4.39, p< 0.001).

Table 4.

Risk for Suicide Associated with Mental Disorders considering Timing of Treatment Presentation.

Predictor		Unadjusted Hazard Ratios			Adjusted Hazard Ratios
		HR	95% CI	p-value	AHR	95% CI	p-value
Mood Disorder	Year 1	6.40	4.36,9.41	<0.001	7.26	4.83, 10.90	<0.001
	Year 2-plus	2.04	1.15,3.60	0.014	2.53	1.43, 4.47	0.001
	None (ref)	1.00	--	--	1.00	--	--
Anxiety Disorder	Year 1	3.15	1.97,5.05	<0.001	3.32	2.06, 5.37	<0.001
	Year 2-plus	1.35	0.73,2.50	0.338	1.50	0.80, 2.80	0.207
	None (ref)	1.00	--	--	1.00	--	--
Substance Use Disorder	Any year	3.55	2.25, 5.60	<0.001	2.76	1.74, 4.39	<0.001
	None (ref)	1.00	--	--	1.00	--	--

Comparison		Unadjusted Hazard Ratios			Adjusted Hazard Ratios
		HR	95% CI	p-value	AHR	95% CI	p-value
Mood Disorder Yr 1 vs. Mood Yr 2+		3.14	1.65,6.00	0.001	2.87	1.48, 5.58	0.002
Anxiety Disorder Yr 1 vs. Anxiety Yr 2+		2.33	1.07,5.08	0.033	2.22	0.97, 4.95	0.051
Mood Disorder Yr 1 vs. Anxiety Yr 1		2.03	0.93,4.45	0.076	2.18	0.97, 4.93	0.060
Mood Disorder Yr 2+ vs. Anxiety Yr 2+		1.51	0.56,4.08	0.419	1.69	0.61, 4.71	0.317

Note. The top half of the table shows risk for suicide associated with each mental disorder within a year of treatment presentation (Year 1) and more than a year after treatment presentation (Year 2+) compared to not having the disorder. The bottom half of the table compares risk associated with a mental disorder within a year of treatment presentation compared to the risk associated with a mental disorder more than a year after treatment presentation (e.g., Mood Disorder Yr 1 vs. Mood disorder Yr2+) or risk associated with mood vs. anxiety disorders within the same time period (e.g., Mood Disorder Yr 1 vs. Anxiety Disorder Yr 1). Unadjusted Hazard Ratios are derived from a model that includes all three mental disorders without covariates. Adjusted Hazard Ratios are also adjusted for covariates of sex, age, marital status, service duration, career group, officer-enlisted status, and major command.

We also examined impact of timing on risk for suicide associated with mood and anxiety disorders through comparisons that took into consideration the time since presentation for each disorder (e.g., mood yr 1 vs. mood yr 2+, mood yr 1 vs. anxiety yr 1). These results are also shown in Table 4 (bottom half of table). Adjusted results show that mood disorder in year 1 was associated with greater risk for suicide than mood disorder in 2-plus years (AHR= 2.87, 95%CI= 1.48-5.58, p=0.002) and a nearly significant result (p=0.051) for anxiety disorder in year 1 to be associated with greater risk than anxiety disorder in 2-plus years (AHR= 2.22, 95% CI= 0.97-4.95, p= 0.051). There was a statistically nonsignificant trend (p<0.10) for mood disorder in year 1 to be associated with greater risk for suicide than anxiety disorder in year 1 (AHR= 2.18, 95% CI= 0.97-4.93, p=0.060). Results of the comparison of mood disorder in year 2+ vs. anxiety disorder in year 2+ are nonsignificant (AHR=1.69, 95%CI= 0.61-4.71, p= 0.317).

Covariates

Finally, in adjusted results (not shown in a table), the following covariates were associated with increased risk for suicide: male sex (AHR= 6.24, 95% CI= 3.43-11.35, p< 0.001); two youngest age groups, age quartile 1, <23.4 years (AHR= 3.19, 95% CI= 1.49-6.84, p= 0.003) and quartile 2, 23.4-27.9 years (AHR= 2.22, 95% CI= 1.98-4.54, p= 0.030) compared to quartile 4, >35.5 years of age (ref); divorced/separated/widow marital status (AHR= 2.09, 95% CI= 1.42-3.08, p <0.001) compared to married (ref). The two groups with shortest service duration were at lower risk for suicide based on the adjusted results, quartile 1, <2.6 years (AHR= 0.48, 95% CI= 0.24-0.98, p=0.042) and quartile 2, 2.6-6.8 years (AHR=0.59, 95% CI= 0.28-0.90, p=0.022), compared to quartile 4, >14.0 years of service. Career group and enlisted vs. officer status were not associated with suicide at a statistically significant level in adjusted analysis.

DISCUSSION

An integrated public health approach to suicide prevention (Knox, Conwell, & Caine, 2004) embodies the principles of population level approaches to reducing deaths from suicide and the importance of identifying vulnerable subpopulations for early clinical intervention. In the military in general, these subpopulations are likely to be service members during transition following a failed relationship or a deployment (Kinn et al., 2010). Other specific subpopulations identified during the mid 1990s by the USAF include individuals with financial and/or legal difficulties. Further assessment of service members who screen positive for depression and anxiety disorders including PTSD may provide an important opportunity for early intervention to reduce suicidal behavior (McCarthy et al., 2004). In this way, a clinical assessment becomes an important instrument for a public health portfolio. These observations suggest that using a similar line of reasoning, significant sub-populations in the military may be represented by active duty service men and women who seek mental health treatment for depression, anxiety, and/or SUD

In analyses of the USAF service member population, we examined associations of mood, anxiety, and SUD with suicide over time. Our first hypothesis was that there is a hierarchy of risk conferred by these disorders. As we predicted, the comparisons showed that mood disorders was associated with higher risk than anxiety disorders (AHR= 1.96, 95% CI= 1.02-3.74, p=0.042) and there was a trend for mood disorders to be associated with greater risk than SUD (AHR= 1.76, 95% CI = 0.90-3.42, p= 0.097). Our other prediction, that SUD is associated with greater risk than anxiety disorders, was not supported.

Our second hypothesis was that risk associated with each disorder was greater within a year of initial treatment presentation than in more than one year following initial presentation. The hypothesis was clearly supported in analyses of mood disorder that adjusted for potential confounders (p=0.002) and the adjusted comparison in anxiety disorder was nearly statistically significant (p= 0.051). Overall, the data support the notion that risk associated with mood and anxiety disorders is highest within the first year after coming to clinical attention. These findings replicate prior results that mood disorders are associated with higher risk in the initial year following treatment than in subsequent years (Qin et al., 2006), and was a novel study of this phenomenon in anxiety disorders. We were not able to test the hypothesis in SUD because there were too few cases to examine the data by time since clinical presentation.

Our third hypothesis was that mood disorders are associated with greater risk for suicide than SUD or anxiety disorders within a year of presentation for treatment but a difference would not be observed at more than one year. We were not able to test the hypothesis concerning SUD. Adjusted comparisons showed a nonsignificant trend that mood disorders are associated with greater risk for suicide than anxiety disorders within the first year after presentation for treatment (p=0.060) and no such trend was observed when risk associated with mood and anxiety disorders were compared at year 2-plus (p=0.317). Although such results seem generally consistent with the hypothesis, cautious interpretation is needed as the former result is a trend and the latter result is not akin to proof of the null hypothesis.

There were limitations of the study. We analyzed USAF service members, with unclear generalizability to other populations. For example, in its initial screening of applicants and its need to assure high readiness of its units, the USAF tends to exclude before entrance or later discharge individuals who suffer significant psychotic disorders. It also excludes or discharges many with SUD. Another limitation was the use of a one year vs. more than one year cutoff to examine the association of timing of diagnoses and suicide was necessitated by the limited number of suicides with each type of diagnosis. However, the broad categories may obscure changes in risk that occur over a shorter term within the first year after treatment (Qin & Nordentoft, 2005; Stenager & Qin, 2008; Valenstein et al., 2009; Reutfors et al., 2010) or after one year (Qin & Nordentoft, 2005; Stenager & Qin, 2008). We used broad definitions of mood and anxiety disorders, and it was not practical to examine specific diagnoses within categories (e.g., panic disorder) to generate more refined estimates of risk (Qin, 2011). Typical of a military cohort, the population was predominantly male (80%), and typical of a study of suicide in the west, suicide decedents were predominantly male. As a result, there were too few female suicides to test potential sex differences (Qin et al., 2003; Stenager & Qin, 2008; Ilgen et al., 2010). The analysis did not follow USAF service members after separation from the service (Kapur et al., 2009). Given sample size limitations, we did not distinguish diagnoses made as inpatients vs. outpatients which are known to affect risk for suicide (Valenstein et al., 2009; Bostwick & Pankratz, 2000). We also did not have sufficient power to examine the combined association of anxiety and mood disorders versus either of these disorders alone (Sareen et al., 2005) or to examine interactions between these disorders (Conner et al., accepted for publication). Diagnoses were based on treatment records and, accordingly, the results pertain only to mental disorders that come to clinical attention.

We conducted a novel study of mental disorders and suicide risk in the USAF. To our knowledge, the current study was the first systematic examination of timing of mental disorders in suicide in a nonclinical population military cohort. The study also offered a rare examination of anxiety disorders and suicide in a population cohort and a rare test of risk associated with diagnoses that were not based exclusively on psychiatric admissions (i.e., severe cases). Results support that risk for suicide associated with mood and anxiety disorders is greatest within one year of diagnoses. These findings underscore the importance of monitoring risk and providing adequate dosage of treatment during the first year following treatment presentation in particular, and of working collaboratively with service personnel and their families during this critical period.