Prevalence, comorbidity and predictors of anxiety disorders in children and adolescents in rural north-eastern Uganda

The methods used in this study are described in detail elsewhere [13, 14]. In summary, this study was conducted in the four districts of Lira, Tororo, Kaberamaido and Gulu in rural north-eastern Uganda. The study districts were selected from a list of eight districts where UNICEF was carrying out child directed medical and psychosocial interventions. In order to draw the sample of four study districts, the eight districts where UNICEF was undertaking child and adolescent directed activities were subdivided into two categories; those experiencing war conflict and those not experiencing such conflict at the time of the study. Two study districts were then randomly selected from each of these two categories. In the category of war affected districts Gulu and Lira were selected, while in the category of non-war affected districts Tororo and Kaberamaido were selected.

Using Kish’s (1965) formula for cross-sectional studies and an average district population figure based on the Uganda Housing and Population Census of 2002, a 95% confidence interval, a precision of 4% and prevalence for emotional and behavioural problems of 15% [17, 18], a sample size for each district of 420 households was estimated. To obtain this sample from each of the study districts, a multistage sampling procedure was used. During the first stage of sampling 2 sub-counties were randomly selected from a list of all sub-counties in each of the study districts. Where the district was war affected and had part of its population living in internally displaced persons camps (IDPs), the sub-counties in that district were initially divided into two groups, those that had IDPs and those that did not, then from each of these two groups a sub-county was randomly selected.

At the next stage, all the parishes in the selected sub-counties were listed and a parish randomly selected. All households in the selected parish were then listed and households with children and adolescents aged 3–19 years were consecutively enrolled into the study until the sample of 210 households per sub-county was attained. If the sample size of 210 households with children aged 3–19 years could not all be obtained from a single parish, a second parish was then randomly selected from the list of parishes in that study and sub-county and households were recruited from there until the required sample was obtained.Where a selected household had more than one child or adolescent who was less than 19 years of age, only one study respondent was selected by simple random sampling.

A generic survey instrument was compiled and translated into the main dialects spoken in the selected sub-counties. To ensure semantic equivalence between English and the local dialects, a process of forward and back translation was undertaken. For each of the 4 main dialects spoken in the study sub-counties, two teams of mental health professionals were constituted. The first team translated these two psychological assessment tools into the local dialect and the second team (which was blind to the initial English version) translated the local dialect version into English. A consensus meeting with the two teams was then held and any major differences in the two versions resolved by discussion.

The translated survey instrument was then administered by trained psychiatric nurses for each selected child or adolescent aged 3–19 years. The trained psychiatric nurses interviewed the children and adolescents themselves (for those who were 10 years or older and capable of responding verbally) or their mothers (for those who were aged less than 10 years or not capable of responding verbally).

The survey instrument contained the following sections:

The Strengths and Difficulties questionnaire (SDQ) [19], was used to assess emotional and behavioural problems. This is a 25- item questionnaire that can be administered to parents or teachers of 3–16 year olds or directly to 11–16 year olds to screen for psychological distress. It covers common areas of emotional and behavioural difficulties and has been validated in both western and developing country settings. The 25 items of the SDQ are divided into 5 subscales of 5 items each, which measure emotional symptoms, conduct problems, hyperactivity/inattention, peer relationship problems and prosocial behaviour, and which taken together comparise a total difficulties score [19].

The SDQ is scored using a Likert scale with the following scores; 0 = not true, 1 = somewhat true and 2 = certainly true. On the basis of an ROC analysis restricted to children and adolescents aged 3–16 using having ‘at least one DSM-IV psychiatric diagnosis’ as a ‘gold standard’, a score of at least 16 was chosen as indicative of psychological distress in children and adolescents. This score ensured a sensitivity of above 60% while keeping adequate specificity [13, 14].

The MINI International Neuropsychiatric Interview for children and adolescents (M.I.N.I.-KID) [20, 21], which embodies DSM-IV-TR criteria for various psychiatric disorders in children and adolescents was used to make specific psychiatric diagnoses. The MINI-KID screens for 23 axis 1 diagnoses. For most modules of the MINI-KID, two to four screening questions are used at the beginning of each module [20, 21]. Further diagnostic questions are asked if the response to screening questions is positive [21]. For each diagnostic category, DSM-IV-TR has a specific number of symptoms, often a duration of disturbance and a distress or impairment criterion [20, 21].

To construct syndrome categories for analysis, these psychiatric disorders were grouped as follows: depressive disorder syndromes (major depressive episode, dysthymia); psychotic disorder syndromes (manic episode, psychotic disorder); anxiety disorder syndromes (panic disorder, agoraphobia, separation anxiety disorder, social phobia, specific phobia, obsessive-compulsive disorder, PTSD, generalized anxiety disorder, adjustment disorder); alcohol and substance abuse disorder syndromes (alcohol abuse and dependency, non-alcohol psychoactive substance use disorder); neurodevelopmental disorders (conduct disorder, oppositional deficit disorder, pervasive development disorder, attention deficit hyperactivity disorder (ADHD) combined disorder, ADHD hyperactive/ impulsive disorder, and ADHD inattentive disorder); eating disorders (anorexia nervosa, bulimia nervosa) and tic disorders (motor tic disorder, vocal tic disorder, Tourette’s disorder, transient tic disorder). Suicidality was defined as meeting any of the three criteria for past suicidality provided in the MINI International Neuropsychiatric Interview for children and adolescents: i) have you ever felt so bad that you wished you were dead ? ii) have you ever tried to hurt yourself? iii) have you ever tried to kill yourself? [21].

A socio-demographic questionnaire included the following variables: a) the subject’s age, gender, tribe, resident district, highest level of education attained and history of exposure to war trauma ; b) previous history of mental illness (psychosis) and attendance at a mental health facility and c) current living arrangement (living with both parents, mother alone, father alone, friends, adopted parents, grandparents and other relatives), orphanhood status, number of siblings, parents’/ guardians’ employment status, family’s total income per month (in Uganda shillings), parents’ highest educational attainment, exposure to domestic violence in the home, nature of housing (permanent or hut and others) and family history of severe mental illness (psychosis).

Additional variables considered in this study included assessment for exposure to war trauma (by asking the question: ‘have you been involved in a situation of war trauma [lived in an IDP, witnessed the torture/ killing of someone, suffered physical or sexual violence as a results of war, been abducted or threatened with violence as a result of war]).

The study obtained Ethical Clearance from the Ministry of Health and the Uganda National Council of Science and Technology. Respondents 18 years and above were required to provide informed consent, while respondents below the age of 18 years were required to provide assent as well as the consent of a parent/guardian.

The prevalence of anxiety disorders was estimated. In order to assess factors associated with anxiety disorders the approach of Victoria and others was followed [22]. Firstly the association of socio-demographic factors was investigated using a backward elimination regression model, choosing the candidate variables based on prior knowledge and plausibility, and using a liberal p-value (15%) to ensure that all variables with a possible confounding effect on the ultimate risk factors were included [23].The socio-demographic factors selected were then all included in a second stage model in which candidate predictors were added and removed using a backward elimination algorithm with a stricter 5% p-value. The results were checked by carrying out forward selection with all selected socio-demographic variables and the same candidate predictors. All analyses were carried out using Stata release 11.2 (StataCorp., College Station, Texas).

The overall prevalence of anxiety disorders in this study was 26.6%, which was higher in females (29.7%) than in males (23.1%). The prevalence of specific anxiety disorders is given in Tables 1 and 2. The most common anxiety disorder in both males and females was specific phobia (15.8% ) followed by PTSD (6.6%) and separation anxiety disorder (5.8%). The prevalence of every disorder was higher among females than among males. Younger children (aged below 5 years) were significantly more likely to have separation anxiety disorder (7.7%,) and specific phobias (20.3%), while those aged 14–19 were significantly more likely to have PTSD (12.8%).

The association of anxiety disorders with socio-demographic factors is summarized in Table 3. The prevalence of anxiety disorders is lowest in Gulu and highest in Lira, the two of the districts that had IDP camps. Anxiety disorders are more common in participants who are older, have some secondary education, live with their father only or with grandparents, have 7 or more siblings, live in permanent housing, have parents with no formal education, or guardians who are unemployed and lowest amongst those whose parents had secondary or higher education.

Table 4 shows the association of anxiety disorders with psychiatric and psycho-social variables. Anxiety disorders were more prevalent among respondents with other psychiatric disorders and for the 66 subjects (4.1%) who had two or more co-morbidities the prevalence of anxiety disorders was 62.1%. The prevalence of anxiety disorders was higher among subjects whose parents were not both alive, those with a history of serious mental illness, those with emotional and behavioural problems as measured by an SDQ score of 16 or higher, and those with abnormal or borderline scores on the emotional symptoms scale.

The results of a multiple logistic regression model including the variables identified as potential socio-demographic determinants of anxiety is given in Table 5. Adjusting for district, gender, employment status of guardian, living arrangements, nature of housing, subjects’ education, parents’ education and number of siblings, the factors found to be significantly associated with anxiety disorders were experience of trauma (OR = 1.93, 95% p < 0.001), score on the emotional symptom scale (OR = 2.58, p < 0.001), and presence of DSM disorders (OR = 3.06, p = 0.001). Significant associations of anxiety disorders were found for female gender (OR = 1.38, p = 0.016), guardian unemployment (OR = 2.16, p < 0.001), living with father only (OR = 2.22, p = 0.005), living in permanent housing, and having parents without education(OR = 0.60,p = 0.049). At both model selection stages (choosing the socio-demographic factors and choosing the psycho-social factors) forward selection confirmed the factors chosen by backward elimination. There was no evidence of any significant interaction effects between sociodemographic and other variables.