Mental Health, Social Functioning, and Disability in Postwar Afghanistan FREE

More than 2 decades of war and conflict and 3 years of drought have led to widespread human suffering and substantial population displacement in Afghanistan. The country's infrastructure has been destroyed or degraded and vital human resources have been depleted.

The invasion and subsequent occupation (1979-1989) of Afghanistan by Soviet forces exposed Afghans to bombardment, executions, and the mass placement of landmines. This conflict generated an estimated 6 million refugees and caused an estimated 1 million deaths.¹ The period following the Soviet withdrawal until 1996 was characterized by factional fighting among the Afghan resistance forces (mujahideen) that led to the death of tens of thousands of civilians and displacement of more than half a million people.¹ During this period, the capital city of Kabul was repeatedly rocketed and bombed, and tens of thousands of landmines were laid.¹

The emergence of the Taliban in 1994 resulted in extreme restrictions on the Afghan people.¹ Under the Taliban, women were prohibited from working outside their homes unless accompanied by a close male relative. In January 1997, the Taliban announced a policy of segregating men and women, which further impeded women's access to health care.

Coalition and Afghan opposition military action led to the fall of the Taliban government in December 2001,² and the return of an estimated 1.6 million refugees from Pakistan and Iran.³ Afghanistan is left with little capacity to support the economic, social, or health care needs of its people. The current estimate, as of July 2003, of the population of Afghanistan is 28.7 million.²

Mental health facilities throughout Afghanistan are nonexistent or in poor condition. In Kabul, the main psychiatric hospital was destroyed during the recent war.⁴ Three of 4 community mental health centers in Kabul are no longer functioning and there is a shortage of trained mental health care professionals in Afghanistan.⁴ Chronic mental illness has been left unattended in Afghanistan for decades. Previous studies have shown that 20% to 30% of a population affected by war and civil strife develop some level of mental distress.⁵ To our knowledge, the only studies that have addressed the impact of protracted armed conflict on mental health in Afghanistan were limited to select provinces in Afghanistan and to Afghan refugees in Pakistan.^{6 - 7}

Basic mental health indicators are essential for guiding efforts to provide mental health programs to the population. From August through October 2002, the Centers for Disease Control and Prevention (CDC), United Nations Children's Emergency Fund (UNICEF), and Vietnam Veterans of America Foundation, in collaboration with the Ministry of Health and the Ministry of Martyrs and Disabled of Afghanistan, and nongovernmental organizations conducted a nationwide population-based mortality, injury, disability, and mental health survey.

The objectives of this survey were to measure the mental health status of Afghans and their traumatic experiences during decades of conflict, as well as self-reported chronic mental illness and substance abuse; assess risk and mitigating factors for mental illness; compare the mental health status of the nondisabled and disabled populations; and develop recommendations for mental health priorities.

The mental health survey was a multistage cluster sample survey. The study population comprised all Afghan residents aged 15 years or older as of August 2002. Because Afghanistan is divided into 32 provinces, 321 districts, and more than 30 000 villages, a 3-stage cluster sampling was used in this survey (Figure 1). The first stage in the sampling was district. The district population frame was determined by using data from 2 sources. First, we obtained data from UNICEF that documented the number of children younger than 5 years who had been vaccinated for poliomyelitis during the national immunization days in April 2002 and applied an adjustment factor for each district to estimate the total population in each district. Second, we combined these estimates with data from the United Nations High Commissioner for Refugees,³ which documented the number of refugees who had returned to each district since April 2002.

To calculate sample size, we used the limit of statistical significance α = .05 with 95% confidence intervals (CIs), assumed a prevalence of mental illness of 25%, and an estimated design effect of 2.5. Based on these assumptions, a sample size of 720 would be necessary. We selected 50 clusters of 15 households per cluster, which yielded 750 households. This sample size was adequate for estimation of prevalence for the overall mental health status of the population. In addition, it allowed identification of risk and mitigating factors for mental illness, and vulnerable groups within the sample of nondisabled Afghan individuals.

During the first stage of sampling, 50 district-level clusters were selected using probability proportional to size method. We selected clusters using the C-Survey software package (Department of Biostatistics and Population Studies Faculty of Public Health, University of Indonesia, Indonesia and Fogarty International HIV/AIDS Training Program Department of Epidemiology, School of Public Health, University of California, Los Angeles).

For the second stage of sampling, village-level population data were not available. Therefore, we randomly selected 1 village or neighborhood from each of the 50 district-level clusters. If the village was inaccessible for any reason, we substituted the nearest accessible village.

For the third sampling stage, we asked a mullah, abob, or other village leader to provide a preexisting list of all households in each village or to help create a list. The households were numbered and 15 households were selected using a random number table. In villages of more than 200 to 300 households, village leaders were asked to help create a list of mosques, with the number of subscribing households in each mosque. As in the first sampling stage, 1 mosque was selected with the probability of selection proportional to size: all households subscribing to this mosque were listed and 15 households were selected as described above. We defined the term household as a group of people who normally lived under the same roof and shared meals. If more than 1 household resided in the same dwelling, 1 household was randomly selected.

For Kabul district, we conducted an additional sampling stage. Using data from 1999 obtained from the Afghan Information Management Service, we estimated the number of houses in each of the 14 subdistricts in Kabul district. Based on the results of the first sampling stage, 4 clusters (subdistricts) were selected from within Kabul district by using probability proportional to size method. After we selected these 4 subdistrict clusters, we selected individual houses using the methods described above.

From the list of household members in each of the households surveyed, we randomly selected 1 nondisabled person aged 15 years or older and administered the mental health questionnaire to this person by reading the instrument aloud. In addition, if a disabled person lived in the household, he or she was also given the mental health questionnaire. If more than 1 disabled person lived in the household, 1 person was randomly selected from among those persons who were disabled and aged 15 years or older.

We defined the term disability as any restriction or lack (resulting from impairment) of ability to perform an activity in the manner or within the range considered normal for a human being.⁸ Disability summarizes many different functional limitations occurring in any population. Such impairments, conditions, or illnesses may be permanent or transitory in nature.⁹

Each survey team comprised 5 persons: a team leader who had experience with surveys, a male and female interviewer, an Afghan interpreter, and a driver/logistician. Each of the 12 survey teams received 5 days of training, which included a half-day field trial.

After selecting the households, survey teams identified and attempted to meet with the head of each household. The team explained the purpose of the survey, reviewed the risks and benefits of participating, and obtained consent for participation. If the head of the household or another adult member was not available, the survey teams asked a neighbor about the whereabouts of the occupants. If the household members had departed permanently or were not expected to return before the survey team had to leave the village, the household was not replaced. If household members were expected to return, the survey team revisited the house at least twice before declaring the household missing.

Data were collected for program purposes, as part of an assessment to inform the development of programs to assist the population of Afghanistan. As such, the objectives of this survey were not primarily research and formal review by an institutional review board at CDC was not required. The data used for analyses were anonymous (contained no personal identifiers of survey participants). Representatives of the Afghan Ministry of Health, UNICEF, and the International Rescue Committee approved the protocol. In each household visited, a consent form was read and oral consent was obtained.

The mental health questionnaire was designed to provide information on the mental health status of the population in relation to war-related psychological trauma. We collected information using open-ended questions from key informants about traumatic events, culture-specific attitudes and symptoms, and mental health problems. Key informants were health care professionals, people from different socioethnic groups, and both men and women of different age groups that we interviewed. We used this qualitative information to adapt standard instruments such as the trauma events scale. In addition to the standardized screening tools mentioned, we asked a series of standard and culture-specific questions to provide data on (1) ethnicity, using a predefined list from which participants selected their preference; (2) mental health coping mechanisms, as determined by the question, "Which of the following factors would help improve your state of mind?"; (3) self-reported previous mental illness disagnosed by a physician or health care professional; (4) substance abuse; and (5) feelings of hatred and revenge and attitudes toward justice, as determined by 3 questions: "Do you feel hatred about what has happened to you and your family during the recent war?", "Do you have feelings and fantasies of taking revenge over what has happened to you and your family?", and "If the person(s) who has/have wronged you and you feel hatred for, would be tried and sentenced by a war crimes tribunal or other justice system, would you still want to take revenge?"

The questionnaire included 3 standard instruments: Medical Outcomes Study 36-Item Short-Form Health Survey (SF-36),^{10 - 12} the Hopkins Symptom Checklist-25,^{13 - 14} and the Harvard Trauma Questionnaire.¹⁵ We selected 4 scales from the SF-36 that assess self-perceived general health, bodily pain, social functioning, and role-emotional functioning. We scored the selected SF-36 questions as recommended in the user's manual; each raw score was transformed to fit a 0-to-100 scale by using a standard formula, with the higher scores on this scale representing better functioning.

The Hopkins Symptom Checklist-25, a screening tool used to detect the symptoms of anxiety and depression, is composed of a 10-item subscale for anxiety and a 15-item subscale for depression, with each item scored from 1 to 4.¹³ Mean cumulative symptom scores of more than 1.75 for each subcategory have been found to be valid in predicting clinical diagnosis of anxiety and affective disorders.¹⁴

We used an adapted version of the Harvard Trauma Questionnaire,¹⁵ which combined the measurement of trauma events (part 1) and symptoms (part 2) of posttraumatic stress disorder (PTSD) as described in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV).¹⁶ The traumatic events questions of the Harvard Trauma Questionnaire (part 2) were modified from previous versions used by the Harvard Program in Refugee Trauma in Cambodia and Bosnia, and by the CDC in Kosovo. We used the information from key informants to determine the most common traumatic events experienced in Afghanistan. We defined cases meeting PTSD symptom criteria according to a scoring algorithm proposed by the Harvard Refugee Trauma Group, on the basis of DSM-IV diagnostic criteria.¹⁷ This definition of PTSD requires a score of 3 or 4 on at least 1 of 4 reexperiencing symptoms, at least 3 of 7 avoidance and numbing symptoms, and at least 2 of 5 arousal symptoms.¹⁷ Trauma event questions were adapted for the specific events among the Afghan population, and the first 16 questions were used from the PTSD symptoms, according to DSM-IV and derived from the Harvard Trauma Questionnaire (part 1). Recall period for questions related to trauma events spanned 10 years, and the previous 4 weeks were used for questions on symptoms of depression, anxiety, and PTSD.

The questionnaires were translated into 2 local languages (Pashtun and Dari) by field staff and back-translated by other staff not involved with the initial translation. All instruments were piloted in 8 to 16 households in the Kabul area to test for ambiguous phrasing, appropriateness of language, and interview technique. The participants in the pilot test were excluded from the study.

We entered results into Epi Info version 6.04 (CDC, Atlanta, Ga). After data were entered from the survey instrument into the electronic database, all identifiers were removed from the electronic database. We calculated prevalence rates for the characteristics described and basic frequencies using Epi Info version 2002 C sample (CDC). Multivariate analyses were performed by using the SUDAAN software version 8.0 (Research Triangle Institute, Research Triangle Park, NC) to account for the complex sampling design. All presented data were adjusted for clustering and assigned a weighting factor. Bonferroni adjustment was performed for comparing traumatic events of disabled and nondisabled persons, with P<.002 considered statistically significant for this analysis.

We used multivariate linear regression to assess the effect of demographic, occupational, and exposure variables on continuous outcome variables, such as the SF-36 scales. Multivariate logistic regression was used for the dichotomous variables, depression, anxiety, and PTSD. We performed statistical comparisons using Wald F tests from multivariate linear regression for continuous variables and Wald χ² tests from multivariate logistic regression for dichotomous variables. P<.05 was considered statistically significant for these analyses.

The demographic variables used in the multivariate regression models were selected based on results from previous studies.⁵ Other variables were included in the model based on their relative importance in this study. All regression models were based on partial, not sequential analysis.

A total of 799 persons completed the mental health survey, of whom 100 were disabled. One cluster of 50 total clusters was not accessible because of security reasons. In some clusters, less than 15 households were present because of the small size of some villages. Only 4 households (0.5%) refused to participate in the survey (Figure 1). We sampled more women than men among nondisabled respondents and more men than women among disabled respondents (Table 1). Most people in both groups were in the younger age category (15-34 years) and married. All participants were Muslim. Most participants reported not having sufficient food and most reported no regular employment. Disabled respondents were mainly Tajik and half of these respondents had no education. However, we found no statistically significant differences in the demographic characteristics between nondisabled and disabled respondents. The weighted prevalence of self-reported mental illness diagnosed previously by a physician or health care professional was 19.3% (n = 144) for nondisabled respondents and 32.4% (n = 25) for disabled respondents. No respondents reported alcohol use. Diazepam (0.7% [n = 10] in nondisabled group; 0.2% [n = 2] in disabled group) and sniffing glue (4.2% [n = 20] in nondisabled group; 10.8% [n = 5] in disabled group) were the most commonly reported drugs used.

A total of 49 participants (7%) experienced at least 11 trauma events; 250 participants (38%) experienced 0-3 trauma events; 231 (35%) experienced 4-7 trauma events; 127 (19%) experienced 8-11 trauma events; and 49 (7%) experienced 12-16 trauma events. The most common trauma events experienced by nondisabled respondents were lack of food and water (56.1%), ill health without access to medical care (54.9%), the need to flee suddenly (44.1%), lack of shelter (44.1%), lost property or belongings (41.4%), and experiencing shelling or rocket attacks (40.8%) (Table 2). Approximately one third of nondisabled respondents reported experiencing bombardments by Coalition forces (34.9%), beatings by Taliban or other groups (32.4%), and living in refugee camps (31.2%).

Disabled respondents most commonly reported lack of shelter (69.7%), ill health without access to medical care (67.3%), bombardments by Coalition forces (62.6%), shelling or rocket attacks (61.6%), serious injury due to knife/gunshot or fighting (60.8%), lack of food or water (58.9%), and need to flee suddenly (55.8%). Nearly one third of disabled respondents experienced lost property or belongings (32.5%) and beatings by Taliban or other groups (31.2%). Most respondents (407 [62.0%]) had experienced multiple traumatic events (≥4 trauma events during the previous 10 years). Trauma events for nondisabled respondents differed statistically from those of disabled respondents for experiencing imprisonment, shelling or rocket attacks, and need to flee suddenly (Table 2).

Mental health coping mechanisms varied among nondisabled and disabled respondents (Table 3). Among nondisabled respondents, reading the Koran or praying (36.9%), more income (27.6%), and talking to family or friends (8.5%) were reported as coping mechanisms. Among disabled respondents, reading the Koran or praying (21.1%), more income (41.3%), medical assistance (14.0%), and talking to family or friends (5.4%) were the most frequently reported coping mechanisms. Overall coping mechanisms for nondisabled respondents differed statistically from those of disabled respondents (P = .04).

Both disabled and nondisabled respondents commonly reported feelings of hatred. Among nondisabled respondents, 84% reported "a lot" of or "extreme" hatred, 14% reported feelings of revenge, and 14% reported wanting to act on their feelings of revenge. Eighty-six percent of nondisabled participants reported not wanting to take revenge if the people who had wronged them and toward whom they felt hatred were tried and sentenced by a war crimes tribunal or other justice system. Among disabled respondents, 81% of participants reported "a lot" of or "extreme" hatred, 20% reported a desire for revenge, and 27% reported wanting to act on this desire. Sixty-five percent of disabled respondents did not want to take revenge if a justice system tried the people who had wronged them.

We compared mean scores of general health, bodily pain, social functioning, and role-emotional functioning scales of the SF-36 for nondisabled and disabled respondents (Table 4). All outcomes were adjusted for demographic characteristics. Disabled respondents had significantly lower health perception, reported more bodily pain, had lower social functioning, and had lower role-emotional functioning than did nondisabled respondents. The prevalence of symptoms of anxiety was significantly higher for disabled respondents (84.6%; 95% CI, 74.1%-95.0%) than for nondisabled respondents (72.2%; 95% CI, 63.8%-80.7%; P = .004). The prevalence of symptoms of depression did not differ among groups: 71.7% (95% CI, 65.0%-78.4%) for disabled and 67.7% (95% CI, 54.6%-80.7%) for nondisabled respondents. The prevalence of PTSD symptom criteria was similar for both disabled and nondisabled respondents (42.2% and 42.1%, respectively).

In addition, we compared mean scores of the general health, social functioning, bodily pain, and role-emotional functioning scales of the SF-36 of nondisabled respondents to the US reference population.¹¹ Higher scores represent better functioning. For nondisabled and disabled persons, the mean scores for general health (39.2 and 29.3), social functioning (57.2 and 46.4), bodily pain (49.0 and 37.9), and role-emotional functioning (56.7 and 41.6), respectively, were substantially lower for the adult population in Afghanistan than for the US reference population mean scores (72.0, 83.3, 75.2, and 81.3, respectively).¹¹

We analyzed the association between selected demographic factors, traumatic events experienced, coping mechanisms, and feelings of hatred, and the social functioning and mental health outcomes for nondisabled and disabled respondents. For nondisabled respondents (Table 5), female sex was associated with lower prevalence of social functioning and higher prevalences of symptoms of depression, anxiety, and PTSD. Older respondents had significantly poorer social functioning and higher levels of depression symptoms. Respondents with little or no education had symptoms of anxiety more often than did respondents with higher levels of education. History of mental illness was associated with higher prevalences of symptoms of depression and symptom criteria for PTSD.

Among nondisabled respondents, multiple trauma events were associated with poorer social functioning and higher levels of anxiety symptoms. Landmine injuries were associated with a higher prevalence of anxiety symptoms and the murder of family or friend was associated with poorer social functioning. Religious or spiritual practices such as reading the Koran or engaging in traditional ceremonies were associated with lower anxiety. Those respondents who focused on covering basic needs, such as more income, better housing, and more food, were less likely to have symptoms of depression and PTSD.

For disabled respondents (Table 6), the prevalences of symptoms of depression and symptom criteria for PTSD were higher for women than men. History of mental illness was associated with poorer social functioning and higher prevalences of symptoms of depression, anxiety, and PTSD. In addition, older respondents were more likely than younger respondents to have symptoms of anxiety; lack of a regular income was associated with anxiety and PTSD among disabled respondents. Married respondents were more likely than single respondents to have symptom criteria for PTSD. Less education correlated with higher prevalences of symptoms of depression (P = .01) and anxiety (P = .004).

Among disabled respondents, multiple trauma events were associated with poorer social functioning. Engagement in religious practices was associated with less depression symptoms. Respondents who sought medical assistance and who focused on covering basic needs were less likely to have symptoms of anxiety. Disabled respondents who had feelings of extreme hatred were more likely to have poorer social functioning but less likely to have depression symptoms.

This population-based mental health survey revealed high prevalence of exposure to trauma events and mental health symptoms among the conflict-affected population of Afghanistan in 2002. Prevalences of symptoms of depression and anxiety and symptom criteria for PTSD were high even when compared with those symptoms of other communities traumatized by war and conflict,⁵ and were higher for women than for men. In comparison, we previously reported PTSD symptom rates of 17% and nonspecific psychiatric morbidity of 43% shortly after the end of the war in Kosovo.⁵ In this study, social functioning was lower among disabled respondents than among nondisabled respondents. Not unexpectedly, social functioning was lower in the surveyed population in Afghanistan than in a US reference population.¹²

The significantly higher prevalences for symptoms of depression, anxiety, and PTSD, and lower social functioning for women than men is not surprising given the restrictions placed on women, especially during the Taliban regime. Although the end of the Taliban regime improved women's liberties, many restrictions remain. Studies among women conducted by the Physicians for Human Rights during the Taliban regime in 1 province in Afghanistan also showed high rates for self-reported symptoms of PTSD, depression, and anxiety.^{6 - 7}

Not unexpectedly, respondents who reported previous mental illness had worse mental health outcomes, including higher levels of symptom criteria for PTSD and symptoms of depression. Disabled respondents also had lower social functioning and higher levels of anxiety symptoms if a psychiatric illness previously had been diagnosed. These results are consistent with findings in other mental health surveys in postwar settings.⁵

In this study, 62% of Afghan respondents have experienced multiple trauma events. Multiple trauma events were associated with decreased social functioning and higher rates of anxiety symptoms for nondisabled respondents. Lower social functioning associated with multiple trauma events also was observed in a mental health study conducted in postwar Kosovo, while higher rates of symptoms of depression and PTSD were associated with trauma events among Cambodian refugees.^{5 ,17} However, we did not find a significant association between trauma events and symptom criteria for PTSD in this survey as was found in these other studies. This is surprising in view of the high prevalence of both PTSD symptoms as well as having experienced multiple trauma events. Extreme poverty and concerns for day-to-day survival caused by economic hardship commonly causes stress. In Afghanistan, socioeconomic factors may have been more important risk factors than traumatic events for PTSD.

Lack of a regular income was associated with higher rates of symptoms of anxiety and symptom criteria for PTSD among disabled but not among nondisabled respondents. Economic hardship may affect disabled persons differently; lack of resources for persons with disabilities can cause additional hardship and may be more difficult to resolve than for people without disabilities.

It became clear during interviews with key informants that the questions regarding alcohol and substance use were considered inappropriate in this Muslim culture, which forbids the use of these substances. Substance abuse was therefore probably underreported in this study because of cultural and religious taboos and the fear of legal ramifications. Use of alcohol, heroin, cocaine, or marijuana was not reported by any of the participants. Some of the participants reported using diazepam and sniffing glue. Some Islamic societies, like Afghanistan, forbid the use of substances that impair thinking and consciousness.¹⁸ However, anecdotal information and reports describe the use of amphetamines and marijuana and an increase of opium production in Afghanistan.^{19 - 20} Reporting use of diazepam may be less of a taboo than reporting use of other substances because it is often prescribed by health care practitioners and therefore accepted as a medical treatment.

Feelings of hatred were commonly reported by these Afghan respondents and prevalence levels (81%-84%) were similar to those levels observed in a postwar mental health survey in Kosovo in 1999.^{5 ,21} As in previous studies in war-affected countries, feelings of hatred and revenge may partially explain continued cycles of violence.²¹ However, in Afghanistan, a desire for revenge was less common than in Kosovo and a substantial number of respondents reported that revenge would not be necessary if a justice system existed. Establishment (or reestablishment) of a justice system may be essential for postwar societies to decrease feelings of hatred and revenge.

We identified several subpopulations with higher risk for poor mental health status and attempted to identify resilience factors. Afghan men in general, men and women younger than 35 years, and people who had no history of previous mental illness, had a high school education, engaged in religious or spiritual practices as coping mechanisms, and believed that covering basic needs was important appear to have been protected from war-related psychiatric morbidity. In general, women in Afghanistan were at higher risk for psychiatric morbidity and lower social functioning. Religious or spiritual practices were important coping mechanisms and protected against anxiety symptoms among nondisabled respondents, and depression symptoms among disabled respondents. Results were similar in a study of Cambodian refugees in the Thailand-Cambodian border camps conducted during the 1980s and 1990s; refugees who engaged in indigenous religious practices were less likely than those refugees who did not to have PTSD.²² Such protective effects of religious practices were also observed among Bhutanese refugees in Nepal and survivors of torture²³ and among Tibetan refugees residing in India.²⁴

Our study had several limitations. No village-level population data were available before the survey began. However, population data were adjusted with weighting factors after data collection using population estimates for surveyed villages. No nationwide baseline data existed on the mental health outcomes for the Afghanistan population, but a survey conducted by the Physicians for Human Rights among Afghan women in parts of Afghanistan and in refugee camps in Pakistan also showed high rates of symptoms of depression.^{6 - 7} Cross-cultural differences could have influenced the results of this study; the instruments were not specifically validated for this society. Nevertheless, in other postwar populations, these instruments have proven to have reliable internal consistency. Care was taken with translation and back-translation of the instruments, and Harvard Trauma Events Scale was adapted for the specific trauma events in Afghanistan. Because no structured clinical interviews were performed, the extent to which self-reported symptoms of PTSD, depression, and anxiety would match clinical diagnosis is unclear. Finally, for persons who were disabled, we did not collect information on whether the disability was permanent or temporary. Although it is possible that there are differences in outcomes among persons with differing durations of disability, the proportion of persons whose disability was temporary was likely to be small. Future assessments should examine these proportions and whether there are differences in outcomes based on duration of disability.

Our survey demonstrates a high prevalence of exposure to trauma and the magnitude of mental health problems among Afghan individuals in postwar Afghanistan. Prevalences of symptoms of depression, anxiety, and PTSD were high, even when compared with other communities traumatized by war and conflict. Women and disabled respondents had significantly poorer mental health status than men and nondisabled respondents. These data underscore the need for donors and health care planners to address the current lack of mental health care resources, facilities, and trained mental health care professionals in Afghanistan.