Factors Associated With Poor Mental Health Among Guatemalan Refugees Living in Mexico 20 Years After Civil Conflict FREE

Twenty years ago an estimated 200 000 Mayan refugees from Guatemala fled a 36-year civil conflict to cross into Mexico.¹ By 1984, 46 000 refugees were under the protection of the United Nations High Commissioner for Refugees (UNHCR) in the Mexican state of Chiapas. In 2000, 12 500 Guatemalan refugees and their Mexican-born children remained in 60 UNHCR refugee camps in Chiapas.¹ Findings from more than 10 000 interviews by 2 Guatemalan truth commissions under the separate auspices of the United Nations and the Human Rights Office of the Archdiocese of Guatemala indicated that Mayan Guatemalans experienced human rights violations, violence, and multiple traumatic events during the civil conflict, such as disappearances, the forced removal of an individual from the community that may involve detention, torture, or murder),^2- 3 massacres, rapes, and razing of villages.^3- 4 Health care professionals who worked with these refugees affirmed the need for mental health services based on clinical observation and treatment-seeking behavior at health clinics.^5- 7

There have been no previously published epidemiological mental health surveys of the Guatemalan refugees who have lived in Chiapas for the past 20 years. Several mental health pilot studies of children were conducted but were not sustained.^5,7- 9 Refugees who live long-term in refugee camps in the developing world may experience traumatic stressors related to refugee camp life that are associated with high rates of psychiatric morbidity, economic stressors, or poor health.^10- 12 In addition, Central American refugees may experience traumatic exposures that impact their mental health.^5,13- 14 Clinicians need to be aware of these risks when working with refugees from this region.

In an agreement between UNHCR and the Mexican government, the refugees who remained in Mexico were entitled to become Mexican citizens, most by the end of 2001. An additional 3400 Guatemalan refugees await naturalization, with 2500 expected to receive this status in 2003; the Mexican government currently finances naturalization.¹⁵ Other UNHCR programs have been largely discontinued, but financial assistance remains for microcredit organizations that assist the Guatemalan refugees, and some food, lodging, education, and legal support are provided on a case-by-case basis.¹⁵ This study was conducted in part to provide the UNHCR and the Mexican government's Refugee Commission for the Assistance of Refugees with rates of psychiatric morbidity and information on vulnerable groups prior to UNHCR's departure. The survey was performed in November-December 2000.

We conducted a cross-sectional, household survey in 5 refugee camps in Chiapas because random selection from the state's 60 refugee camps was not possible due to regional security concerns. Therefore, camp selection was convenience-based and derived from qualitative information obtained during a preliminary rapid assessment. A set of necessary characteristics of the overall camps was constructed with expertise from UNHCR staff. Ten potential refugee camp study sites were visited with UNHCR personnel. Five camps were selected according to the following criteria: (1) variation in size, (2) representation of the 2 main Guatemalan-Mayan subethnic groups in Chiapas (Kanjobal and Chuj), (3) difference in location to highway routes and a city, (4) difference in degree of intensity of conflict exposure in Guatemala (based on UNHCR expert opinion), (5) UNHCR need for information, and (6) safety of the survey team in regions of current conflict in Chiapas.

Qualitative fact gathering on location with residents prior to beginning the survey indicated that residents lived uniformly in poverty. This was evidenced by insufficient food and land sources, lack of sustainable employment, and poor housing to protect from rain and cold. At the time of survey, there were no mental health services being offered to Guatemalan refugees residing in refugee camps in Chiapas.

The survey was intended to be a survey of the entire adult population in each respective camp, but this proved to be logistically impossible. Therefore 1 adult per household was included in the analyses. In the 9 households in which 2 survey participants resided, we randomly deleted 1 of the participants in the final analyses by flipping a coin to select 1 from each pair. Houses were defined as the physical structures; household members were defined as any person who slept and ate meals in the same physical structure during their stay in the refugee camp.

Table 1 describes the characteristics of the 5 selected refugee camps. An overall population estimate of 1546 adults and children was derived from the Mexican Commission for the Assistance of Refugees monthly census report from September 2000. The number may be inflated due to household members who were living outside the camps in other parts of Mexico or the United States at the time of the survey. There were a total of 183 houses approached for the survey in the 5 camps. This number represents all houses in camps A, B, C, and E, and every third home in camp D. Camp D was understaffed by data collectors and every third household was surveyed due to this lack of resources. In camp D, interviewers were dispersed to different corners of the camp, instructed to count 3 households from the edge of the camp, and proceeded to every third door. Camp D did not differ statistically in variables for age, number of persons in the home, year of arrival to Mexico or for all 3 outcomes—PTSD, anxiety, and depression from the other camps.

Eligible respondents were all refugee camp residents who had left Guatemala as refugees to seek political asylum in Mexico and who were present in their respective homes at the time of the survey or who were the Mexican-born children of a Guatemalan refugee parent and who were at least 16 years of age at the time of the survey.

Oral consent was obtained due to high rates of illiteracy. Respondents were asked to make an "X" on their form to indicate consent. Interviews were conducted individually in a private location within the home. Respondents were compensated for their time with art supplies for their children; camp schools received instructional supplies. When it was possible to return to a community, data collectors returned to homes in which no one had been home initially.

Data collectors were Guatemalan refugees with a certificate in Spanish reading and writing, prior experience in field work, and fluency in 2 required Mayan languages. A 3-day training session included rehearsals of questionnaire interviews in Spanish and the 2 Mayan languages of Kanjobal and Chuj, training in the protection of human subjects, and implementation of a mental health crisis protocol. Data collectors were instructed to inform all participants of the availability of a licensed, on-site social worker (M.S.) if they wished to talk before, during, or after their participation in the survey. Participants were also informed that the results of the study would be shared with representatives of their respective communities in spring 2001. This survey was approved by the University of Georgia institutional review board and the governing representative boards of the refugee camps.

Respondents were administered demographic questions, the Harvard Trauma Questionnaire (HTQ),¹⁶ a scale that measures symptoms of posttraumatic stress disorder (PTSD) and traumatic events, and the Hopkins Symptom Checklist-25 (HSCL-25), a measure of elevated symptom scores for anxiety and depression.¹⁷ Previous studies have documented the reliability and validity of the HTQ and the HSCL-25 in Southeast Asian and Bosnian refugee populations.^12,17- 18 Because no previously established cut-off points exist for Guatemalan refugees, an algorithm based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV)¹⁹ PTSD criteria was used to determine whether the respondents met the study's case definition for symptoms of PTSD (B.L.C., unpublished data, 2001).

For this study, exposure to trauma, as measured by the HTQ, included those who responded that they experienced and/or witnessed a traumatic event. As required by the HTQ guidelines, the HTQ was modified for accuracy of traumatic event content and for language to maintain instrument validity.¹⁹ Six independent reviewers, including Guatemalan refugees who work for aid organizations, then reviewed the instrument for validity of traumatic event content and accuracy of Spanish. A Spanish version was provided by the Harvard Program on Refugee Trauma. The reviewers received this Spanish-language version and were asked to translate it to a Spanish Southern Mexico–Guatemalan dialect to increase comprehensibility for the study population. All reviewers concurred independently with the final, sixth version in back-translations into a final Spanish Southern Mexico–Guatemalan dialect version. The HSCL-25 was translated by 3 reviewers from English into the same regional Spanish dialect and back-translated into English by the 3 reviewers into a final Spanish Southern Mexico–Guatemalan version. In the HSCL-25, a score of 1.75 or higher was considered significant.¹⁷ All outcomes analyzed in this article were bivariate, based on meeting the study's symptom criteria for PTSD (HTQ), anxiety or depression (HSCL-25), or not.

The instruments were not translated into the 2 Mayan languages because they are not written languages. To compensate for variability in vocabulary during the interview process, data collectors agreed on and rehearsed vocabulary to be used in the event that Spanish was not a possibility. The study of Karenni refugees used the same technique in an indigenous population without a written language (B.L.C., unpublished data, 2001).

The HTQ and the HSCL-25 were examined to determine the internal reliability of the scales in this study population. Using the Cronbach α, the HTQ's internal reliability level for the 19-item list of experienced and observed traumatic events, was 0.93. The internal reliability level for the symptom criteria for PTSD was 0.88. There were a total of 24 items in this section. Based on the testing of the 25 items on the HSCL-25, an α level for internal reliability of 0.95 was calculated for the entire scale.

Data were entered and analyzed in SPSS for Windows, Version 10.0.²⁰ Continuous variables were tested for normality using the Levene test. Differences in select demographics and traumatic events and mental health outcomes in the 5 refugee camps were examined with analysis of variance. Variables for age, sex, marital status, number of persons living in the home, number of traumatic events, number of years as a refugee, and the same traumatic event variables from the HTQ for PTSD symptoms were examined for statistical significance. Analyses were conducted on the number of participants who completed the relevant parts of the surveys. Least-squares difference post hoc tests were used to determine significance of association. Following univariate analyses on all traumatic events and selected demographic variables, multivariate logistic regression models, using a backward-step procedure were fitted to analyze associations between factors and dichotomous outcomes with all variables associated at P<.10. Logistic regression was validated using the Hosmer and Lemeshow method and collinearity and overfitting were examined. Statistical significance in the multivariate logistic regression models was assessed at a P value of .05. Years elapsed since the traumatic events in Guatemala and years spent in the refugee camps were examined using t tests with the 3 main outcomes of interest.

Out of 183 total houses approached, 170 houses had at least 1 adult (aged ≥16 years) who agreed to participate, representing 93% of all households approached (Table 1). Of the 170 participants included in this analysis who answered the demographic questions and the HTQ, 10 dropped out before completing the HSCL-25. The reason given for refusal to complete the HSCL-25 was due to the need to look after children or the length of the survey. Therefore, results on anxiety and depression outcomes had a total of 160 respondents.

Demographic characteristics are summarized in Table 2. All participants reported Guatemala as their birth country. Fifty-eight percent of respondents were women. Sixty-five percent of interviews were conducted in indigenous languages. All continuous variables were normally distributed. Forty-five percent of the respondents were aged 16 to 35 years, with a mean age of 37.9 years (95% confidence interval [CI], 35.9-40.0 years). Fifty percent of respondents indicated they had no school education. Demographic differences among the 5 study sites were examined using 1-way analysis of variance and least-squares difference post hoc tests to determine statistical significance. Years elapsed since the traumatic events in Guatemala and years spent in the refugee camps were analyzed for the 97% of respondents who arrived in refugee camps in Mexico between 1981 and 1982 using t tests with PTSD, anxiety, and depression outcomes and were not statistically significant. There were no significant differences that warranted the stratification of data in further analyses.

All respondents reported that they experienced at least 1 traumatic event (Table 3). A total of 19 different traumatic events were examined. There were 1230 reported traumatic events with a mean of 8.3 traumatic events (95% CI, 7.7-8.9) experienced per individual, with a range of 1 to 19 events reported. The top 3 traumatic events were lack of food (94.7%), lack of water (85.9%), and lack of refuge or housing (85.3%). Overall, 55.9% reported being close to death, 43.5% reported a family member or friend massacred, and 45.9% reported witnessing the disappearance of others. Fourteen percent indicated being tortured.

Twenty of 170 respondents (11.8%) met the symptom criteria for PTSD. Women comprised 65% of all respondents who had symptoms of PTSD. An additional 20 respondents had all but 1 of the PTSD symptom criteria. Table 4 lists the participant characteristics and traumatic event variables that were statistically associated with symptom criteria for PTSD in the univariate analysis. Study participants who met symptom criteria for PTSD were more likely than those without PTSD to be between the ages of 36 and 80 years; to have been close to death; to have witnessed an assassination, the disappearance of others, a massacre, or others being sexually abused or raped; to be living with 9 or more persons; or to have lived in 3 or more other camps.

Four of the above variables remained significantly associated with PTSD symptom criteria after being fitted in the logistic regression model. Respondents who witnessed the disappearance of others (adjusted odds ratio [AOR], 4.58; 95% CI, 1.35-15.50) and who reported being close to death (AOR, 4.19; 95% CI, 1.03-17.00) were more likely to meet symptom criteria for PTSD than respondents who had not experienced these traumatic events. In addition, respondents who had more than 9 persons living in the household in the refugee camp (AOR, 3.69; 95% CI, 1.19-11.39) were more likely to meet symptom criteria for PTSD than respondents who had fewer household members. Respondents who experienced a lack of food were less likely to meet symptom criteria for PTSD (AOR, 0.08; 95% CI, 0.01-0.59) than those who did not report experiencing a lack of food.

Eighty-seven (54.4%) of 160 respondents had elevated anxiety symptom scores on the HSCL-25. Women comprised 54% of respondents with anxiety symptoms. Anxiety symptoms were associated with being between 36 and 80 years of age, having experienced between 7 and 12 traumatic events or between 13 and 19 events, and the following specific traumatic events: imprisonment, being wounded, sexual abuse or rape, forced separation from family or friends, witnessing the disappearance of others, torture, or a massacre (Table 5).

Adjusting for each of these variables, 3 variables remained statistically significant in the final logistic regression model. Respondents who had witnessed a massacre (AOR, 10.63; 95% CI, 4.31-26.22) or who had been wounded (AOR, 3.22; 95% CI 0.95-10.89) were more likely to have elevated anxiety symptom scores than those who did not experience these traumatic events. Respondents who experienced between 7 and 12 traumatic events (AOR, 2.67; 95% CI, 1.14-6.27) or between 13 and 19 traumatic events (AOR, 2.26; 95% CI, 0.65-7.89) were more likely to have elevated anxiety symptoms than those who experienced fewer traumatic events.

Sixty-two (38.8%) of 160 respondents had elevated symptom scores for depression. Women comprised 68% of the cases (Table 6). Symptoms for depression were associated with being a woman; being a widow; witnessing the disappearance of others; experiencing torture, mutilation, or between 13 and 19 traumatic events.

After adjustment, the final logistic regression model retained 4 variables. Respondents who were women (AOR, 3.64; 95% CI, 1.47-9.04) or widowed (AOR, 27.55; 95% CI, 2.54-299.27) were more likely than men to have elevated symptom scores for depression. Respondents who experienced between 13 and 19 traumatic events (AOR, 7.44; 95% CI, 2.18-25.37) or who witnessed the disappearances of others (AOR, 2.68; 95% CI, 1.16-6.19) were more likely to have elevated symptom scores for depression than those who did not experience these traumatic events.

Psychiatric morbidity was found to be highly prevalent in this study population of Guatemalan refugees living in Mexico 20 years after civil conflict: 11.8% met DSM-IV symptom criteria for PTSD, 54.4% had anxiety symptoms, and 38.8% had depression symptoms. Respondents indicated that they experienced or witnessed multiple traumatic and violent events; these events appear, in this population, to have had an effect on the estimated prevalence of poor mental health outcomes. Associations remained after adjusting for effects of demographics and other exposures.

Respondents who witnessed the disappearance of others were associated with anxiety symptoms in univariate analyses and were associated with symptoms of PTSD and depression in the final models. Although disappearances were commonly used in Latin America during periods of civil strife,^21- 22 no other published studies in peer-reviewed journals were found that examined the effects of family disappearances for refugees from Central America with poor mental health outcomes. Although the results cannot be extrapolated to other populations, it may highlight unique mental health risks to refugees with a history of a family disappearance of Latin American origin who present at mental health or health clinics. Because persons who disappeared are frequently never found, it may make finalization of the grieving process impossible or difficult for family members.²³ When family members disappeared, it can also permanently alter the structure and roles of these families who use their emotional resources searching for their lost family members.²³

In this study, 11.8% of the Guatemalan refugee respondents met symptom criteria for PTSD. Using the Cambodian version of the HTQ, Mollica et al¹² reported a 14.7% rate of symptoms of PTSD among 993 Cambodian refugees living for 10 years in refugee camps.

The rate of symptoms of depression found in this study, 38.8%, was comparable with other studies that used the HSCL-25. In the 1996 study by Allden et al,²⁴ the rate of depression for Burmese refugees was 38% 4 to 5 years after refugee flight while the 1999 Mollica et al²⁵ study of Bosnian refugees using the HSCL-25 found a 39.2% estimated prevalence rate 1 to 2 years after exposure to trauma.

Respondents who witnessed a massacre were associated with elevated anxiety symptoms. Massacres were a common occurrence during the civil strife in Guatemala as reported in the truth commission reports^3- 4 and are a commonly used form of terror in other violent global events.

A protective factor for PTSD symptoms was found for participants who indicated that they lacked sufficient food during the flight from Guatemala. Anecdotal reports collected from qualitative fact gathering indicated that while refugees hid in the highlands of Guatemala en route to Mexico, food was gathered, cooked, and shared collectively. This may have created enough sense of well-being to counteract the effects of the shortage. It should be noted that this may be an artifact of the data.

The high internal reliability levels found in the analysis of the Cronbach α on the HTQ and the HSCL-25 are promising for future use for Guatemalan refugees. These results indicate that respondents who met symptom criteria for PTSD and those who had elevated symptom scores for anxiety and depression answered questions consistently. This may also indicate that the Spanish-Southern Mexico–Guatemalan translation and on-site translation to 2 indigenous languages was consistent and accurate. Although further validation and norming tests should be conducted with Guatemalan refugees, it may indicate that the scales are applicable for use with Guatemalan refugees in clinical and other refugee settings.

This study had several limitations. This was a cross-sectional survey conducted in 5 out of 60 refugee camps in Chiapas. It was not possible to conduct a survey in all 60 camps; the results are only representative of households in the 5 refugee camps. Results from camp D were not weighted due to a lack of individual census numbers.

Other issues included an attempt to recruit an even number of male and female data collectors to reduce the possibility of interviewer bias, but only 2 men (out of 8 total interviewers) were able to be recruited. There may have been underreporting of sensitive traumatic events by women if they were interviewed by a man. Although we rehearsed correct vocabulary for translations to indigenous languages, it is possible that some items were not translated uniformly or were not understood correctly by respondents. Furthermore, the HTQ traumatic event list is limited in scope and does not measure the full range of traumatic events experienced by this study population. The HSCL-25 and HTQ provide an estimate or proxy of anxiety and depression and PTSD and cannot provide formal diagnoses. In addition, 11 persons dropped out prior to taking the HSCL-25. This may have produced a selection bias; however, none indicated that they would not complete the survey due to psychological stress. Finally, although there was an attempt to recruit Mexican-born respondents, there were none in the sample. This may have been a result of the younger age of Mexican-born respondents who were more likely to be working outside of the refugee camps at the time the surveys were conducted.

More than half the participating Guatemalan refugees reported experiencing anxiety symptoms at the time of the survey (54.3%); 38.8% reported symptoms of depression, and 11.8% met symptom criteria for PTSD. Although it is not possible to determine whether these rates of psychiatric morbidity are a result of violence or trauma incurred during the war in Guatemala or as a result of life in the refugee camps, it is clear that there was significant psychiatric morbidity in the 5 refugee camps at the time of survey. Clinicians in Mexico, Central America, the United States, and elsewhere who work with refugees from Guatemala should be aware of their unique risks of poor mental health associated with experiences of civil strife.

The phasing out of UNHCR's refugee program in Chiapas is a critical juncture in the future of the Guatemalan refugees. The Mexican state of Chiapas is already strained with poor land resources and a rebel conflict with the Mexican government. Mental health is vital to the economic, physical, and emotional survival of refugees.²⁶ Poor mental health status may result in more difficulty integrating into the larger Mexican community. There are currently no known mental health programs available to the Guatemalan refugees.²⁷ Mental health programs, such as refugee-led groups and educational campaigns, may assist with this integration and heal the wounds incurred in Guatemala 20 years ago or as a result of refugee camp life.