Study design
This was a cross-sectional quantitative study among pregnant teenagers attending antenatal care in selected health facilities within Mbarara city, southwestern Uganda.
Study setting
The study was conducted from May 28, 2024 to Aug 2, 2024 at 4 primary health care facilities in Mbarara city, which offer high-risk antenatal services (ANC), as per the Ugandan Ministry of Health (MOH) guidelines [19]. Mbarara city is approximately 270 km (170 miles), by road, to the southwest of Kampala, Uganda’s capital city. The selected health facilities included Mbarara Municipal Council Health Centre IV, Kakoba Health Centre III (HC III), Biharwe HC III and Nyakayojo HC III. In Uganda, provision of antenatal services to pregnant teenagers in public institutions starts at health centre III (HC III) which is at sub-county level, with subsequent referrals to health centre IV at the county level, district hospitals, regional referral hospitals, and finally to the national referral hospital where specialized high-risk antenatal services are provided.
From the 2021/2022 annual antenatal records of Mbarara city, it was found that Mbarara Municipal Council Health Centre IV receives the highest number of teenage mothers annually (302 teenagers) followed by Kakoba HC III (180 teenagers), then Biharwe HC III (139 teenagers). The high numbers of pregnant teenagers have been attributed to cultural practices, including early marriages and related risky sexual practices in the region, compounded by low socioeconomic status and illiteracy about existing sexual and reproductive services [20].
Eligibility criteriaInclusion criteria
We enrolled pregnant teenagers aged 13 to 19 years who were receiving antenatal care (ANC) at the selected health facilities and who were willing to provide written informed consent for participation in the study.
Exclusion criteria
We excluded those who were in need of emergency intervention (e.g. severe pre-eclampsia in hypertensive crisis, per vaginal bleeding, altered level of consciousness) and those with communication impairments, as well as those unable to provide consent.
Sample size estimation
Sample size was determined using the Kish Leslie formula and based on prevalence estimates of depression rates of 32.9% among pregnant teenagers in Nairobi, which was assessed with a similar tool (Edinburgh Postnatal Depression Scale) to the one used in this study [21]
$$n =\frac{{Z}^{2}P(1-P)}{{e}^{2}}$$
$$n=\frac{{1.96}^20.329(1-0.329)}{{0.05}^2}=339+10\%\;\mathrm{non}-\text{response rate }=373\text{ participants}.$$
where; n is the sample size, Z is standard normal distribution of 1.96 which corresponds with 95% confidence level, p is 32.9%; the prevalence of depression among pregnant teenagers in Kenya [21], e is the allowable margin of error of 0.05. We added a 10% non-response rate, making the final sample size of 373 participants.
Sampling procedure
Pregnant teenagers were recruited using consecutive sampling. After their routine clinical assessment and management, participants were approached and given information about the purpose of the study. Those who provided written informed consent were included in the study. Sampling was done proportionately to the number of patients enrolled at the specific health facility.
Study variablesDependent variable
The dependent variable was depression using the Edinburgh Postnatal Depression Scale (EPDS) [22].
Independent variables
The independent variables were factors that have been identified in the literature to be associated with depression among pregnant teenagers. These included sociodemographic characteristics, pregnancy-related factors, and psychosocial factors [1, 11, 23,24,25].
Sociodemographic factors included age (in complete years), geographical location (rural, urban), level of education (none, primary, secondary, tertiary), marital status (currently in an intimate relationship, not currently in an intimate relationship), employment status (employed, unemployed), and HIV status (negative, positive). Pregnancy-related factors included gravidity, parity, gestational age (in weeks), number of antenatal care (ANC) visits, conception circumstances (consensual, commercial, or rape). The psychological factors included the family history of mental illness (yes, no), current use of alcohol use (yes, no), perceived social support, resilience, and food insecurity.
Study measures
Depression was assessed using the Edinburgh Postnatal Depression Scale (EPDS) [22]. The EPDS is a 10-item instrument originally used to screen depression among postnatal women. It has been used widely in sub-Saharan Africa for comprehensive screening of depression symptoms during pregnancy among adolescents, and has shown good and reliable psychometric properties [26]. Each item is responded to on a 4-point Likert scale ranging from 0 to 3. Items 3, 5, 6, 7, 8, 9, 10 are reverse scored (3 to 0). The total score is determined by adding together the scores for each of the 10 items, producing a sum score range from 0 to 30, with higher scores indicating increased severity of depressive symptoms. In this study, pregnant teenagers who scored 10 and above were considered depressed [25]. Depression was categorized into 4 categories depending on the total scale scores as follows: no depression (0 to 6), mild depression (7 to 13), moderate depression (14 to 19) and severe depression (20 to 30). The Cronbach alpha of the scale in the present study was 0.87.
Social support was assessed using the multi-dimensional scale of perceived social support (MDSPSS) (Zimet et al., 1988). The MDSPSS is a 12-item measure of perceived adequacy of social support from three sources: family, friends, and significant others. The MDSPSS is scored on a 7-point Likert-type scale ranging from 1(strongly disagree) to 7 (strongly agree) giving a total minimum score of 12 and a total maximum score of 84 with higher scores indicating greater perceived social support. The scale has 3 subscales: “Family subscale” (items 3, 4, 8 and 11), “Friends subscale” (items 6, 7, 9 and 12) and “significant others subscale” (items 1, 2, 5 and 10). The MDSPSS has been validated for use in Uganda where it demonstrated good internal consistency, reliability, and validity [27]. Among pregnant teenagers, the total Cronbach’s alpha of the MDSPSS reported at 0.80, while Cronbach’s alpha coefficients for family, friends and significant others subscales were 0.81, 0.88 and 0.92 respectively [28]. For the present study, the Cronbach alpha coefficients for family, friends, and significant other subscales and total scale were 0.89, 0.91, 0.80, and 0.87 respectively.
Resilience was screened using the 14-item Wagnild and Young Resilience Scale (RS) which has been used among pregnant teenagers in Africa [29]. The scale is rated on a 7-point Likert type scale (1 = strongly disagree to 7 = strongly agree) with summed scores ranging from 14 to 98 where higher scores reflect higher resilience. Based on standard classification, the summed scores were categorized into levels of resilience as follows: low (≤ 64), moderate (65–81) and high (≥ 82) [30]. A prior study in Nigeria reported good reliability of the scale among pregnant teenagers with a Cronbach’s alpha of 0.80 [29]. In this study, the scale had a Cronbach alpha of 0.89.
Food insecurity was assessed using the 9-item Household Food Insecurity Access Scale (HFIAS). This scale addresses the frequency in the previous 4 weeks with which any household member experienced anxiety and uncertainty of food access, inadequacy of food quality, insufficient food intake or hunger [31, 32]. Responses are coded based on the frequency of occurrence with 0 meaning “never (0 times)”, 1 meaning “rarely (1–2 times)”, 2 meaning “sometimes (3–10 times)”, and 3 meaning “often (11 + times)” with a total score ranging from 0 to 27. It has been used in Uganda to differentiate between food secure and food insecure households with good reliability [33]. In the present study, the scale had a Cronbach alpha of 0.85.
Data collection and management
Data collection was done using the Kobo Collect Toolbox. Data was collected by two research assistants who had received training in responsible conduct of research and data collection techniques. The study participants were approached one at a time after receiving their ANC package. Research assistants administered the interviewer-administered questionnaire to the participant. The principal investigator settled any mishaps in data at the end of each data collection day. All data entered was stored on google cloud with a backup copy on a password protected computer of the principal investigator.
Quality control
The research data collection tools and consent forms were translated from English to the local language (Runyankole/Rukiga) and back-translated to English to ensure consistency. Research assistants were fluent in both English and Runyankore, and were trained on how to administer the questionnaire, how to ask sensitive questions, and ethical conduct of research before data collection started. Participants who were found to have significant distress (those whose psychological state causes impairment in social, occupational, or other important areas of daily functioning) were referred for evaluation and management at the psychiatry department of Mbarara regional referral hospital.
Data analysis
Data were imported into STATA version 17 for data cleaning and analysis [34]. Continuous variables were summarized using means and standard deviations (if normally distributed), and median and interquartile rage (if non-normally distributed). Categorical variables were summarized as frequencies and percentages. The Shapiro-Wilks test and histograms were used to assess normality under the Gaussian assumption. Our outcome variable was depression, defined as scores of 10 and above on the Edinburgh Postnatal Depression Scale. Logistic regression models were run to determine factors associated with depression among pregnant teenagers. Following bivariate analysis, only factors that are biologically plausible (consistent with known biological mechanisms) and statistically significant were considered for the final multivariable model. This was done to avoid model overfitting and to improve model parsimony. All analysis was done at 95% confidence interval considering the level of significance to be less than 0.05.