Prevalence and associated factors of mental and substance use problems among adults in Kenya: A community-based cross-sectional study

Abstract

Data on the burden and determinants of mental and substance use problems among women in urban and rural informal settlements in Kenya is sparse, thus limiting preventive and treatment efforts in these areas. To bridge the gap, we (a) determined the prevalence of depressive, anxiety and post-traumatic stress disorder (PTSD) symptoms and alcohol and drug use problems among women compared to their spouses and (b) examined the correlates of these outcomes. Data collection for this cross-sectional survey was conducted in 2022 in Mombasa, Kwale and Nairobi counties in Kenya. A total of 1528 adults (1048 women) took part. The 9-Item Patient Health Questionnaire, 7-Item Generalized Anxiety Disorder Scale, Primary Care PTSD Screen for DSM-5, Alcohol Use Disorders Identification Test, and Drug Use Disorders Identification Test were administered alongside other measures. Logistic regression was used to examine the correlates of mental and substance use problems. Overall, the prevalence of mental and substance use problems among women and men respectively was 28% (95% CI 25, 31%) vs 22% (95% CI 18, 26%) [depressive symptoms], 12% (95% CI 10, 14%) vs 8% (95% CI 6, 11%) [general anxiety symptoms], 22% (95% CI 19, 24%) vs 21% (95% CI 17, 25%) [PTSD symptoms], 4% (95% CI 3, 5%) vs 15% (95% CI 12, 19%) [alcohol use problems], and 2% (95% CI 1, 3%) vs 12% (95% CI 9, 15%) [drug use problems]. The prevalence of depressive and anxiety symptoms was significantly higher among women compared to their male counterparts. On the other hand, both current and past-year alcohol and drug use were significantly higher in men than women. Among women, stressful life events, urban residence, food insecurity, family debt, unemployment, poor self-rated health, poor eyesight, and higher educational level were the correlates for elevated depressive, anxiety and PTSD symptoms. Conversely, sexual abuse, living in rented houses, urban residence, verbal abuse, stressful life events, and somatic complaints were the correlates for depressive, anxiety and PTSD symptoms in men. Correlates against mental health problems included social support, higher subjective wellbeing, older age (>50 years), increased vigorous exercise and higher household income (in both sexes). Correlates for current alcohol use in women included stressful life events, urban residence, being sexually active, and living in a single family. Among men, higher household income was associated with current alcohol use. Correlates against current alcohol use included being married, living in a larger household (>5), being a Muslim and having multimorbidity (in both sexes). Correlates for current drug use included unemployment and sexual abuse. Female sex was associated with reduced odds of current drug use. The prevalence of depressive, anxiety and PTSD symptoms, and alcohol use problems is high in the study setting. However, needs vary by gender and study location, highlighting the importance of targeted approaches in mental health services. Our results also highlight the need for integrating mental health services into existing primary care as well as testing and scaling multi-component mental health interventions in this population.

Background

The importance of mental health as an integral part of general health and sustainable development has long been recognized [1–3]. In itself, mental health is essential for physical health and is closely connected with development factors such as work, economic growth, poverty, peace and justice. Mental health also plays a crucial role in achieving social inclusion, equity, universal health coverage, access to justice and human rights. Since the first Global Burden of Disease Study findings were published in the 1990s, there has been increasing evidence that mental and substance use disorders constitute a leading cause of disease burden around the world and about three-quarters of this burden lies in low-and-middle-income countries (LMICs) [4,5]. Predictive models also indicate a 130% rise in the burden of mental and substance disorders in sub-Saharan Africa (SSA) by 2050 because of population growth and ageing [6], further highlighting the urgent need for meaningful investments on mental health in the region.

The Eastern SSA region is undergoing a rapid epidemiological transition from communicable to non-communicable disease burden [7]. Kenya is one such country that is witnessing this transition – with improving life expectancy, reduced malnutrition, and a concomitant rise in the burden of non-communicable diseases [8]. In recent years, the conversation around mental and substance use disorders has increasingly taken centre stage in Kenya, given the realization that mental health issues are pervasive and impactful. For instance, data from the Kenya National Commission on Human Rights estimates that 25% and 40% of outpatients and inpatients suffer from mental health problems, respectively [9]. In response, the government of Kenya has progressively made efforts to improve mental health services in the country. These efforts include the launch of several policy and legislative frameworks, including the Kenya Mental Health Policy 2015–2030, Kenya Mental Health Action Plan 2021–2025, Kenya Mental Health Investment Case 2021, Mental Health (Amendment) Act of 2022, Suicide Prevention Strategy 2021–2026, the 2018 National Framework for Implementation of Problem management Plus and the 2019 presidential taskforce on Mental Health in Kenya [9]. While this is laudable, like other SSA countries, Kenya still struggles with the potent implementation of the existing mental health policies, leading to a continually widening mental health treatment gap [10,11]. Currently, about 75% of Kenyans are not able to access mental healthcare, nearly half of the counties do not have psychiatric units, and the country still does not have a separate mental health budget, and government expenditure on mental health is 0.01% of the health expenditure.

Generally, the available data on the burden and determinants of mental and substance use disorders in Kenya is sparse. For example, the national-level survey of non-communicable diseases in Kenya did not include mental health [12]. Epidemiological research is crucial to understand better the differential risk factors and burden of mental and substance use disorders across different regions and social contexts. Data on adolescent mental health is building up, thanks to the establishment of nationally representative surveys [13]. While this is encouraging, more data is needed to generate more accurate estimates of health burdens and salient determinants from which local health policymakers can draw for other populations. Such research will inform optimal strategies for prevention, treatment and follow-up for people with mental and substance use disorders across the life course. Most of the existing studies have been collected from small samples of diverse sets of populations, including people living with HIV, youth, and pregnant women, although not necessarily representative of the population. There is especially a dearth of research on the burden and determinants of mental and substance use disorders among adults in the general population.

The current study addresses this gap by a) documenting the prevalence of mental health problems (symptoms of depression, anxiety and post-traumatic stress disorder) and substance use problems (alcohol and drug use problems) in a community sample of adults from Nairobi, Mombasa and Kwale counties in Kenya, and b) investigate the factors associated with these problems.

Methods

Study design

This was a community-based cross-sectional study assessing the prevalence and associated factors of mental health problems (symptoms of depression, anxiety and PTSD) and substance use problems (alcohol and drug use) among adults in Kenya.

Setting

The study was conducted in 2022 in Nairobi, Mombasa and Kwale counties in Kenya. The 3 counties are part of the 47 devolved (semiautonomous) units of government in Kenya which were introduced in the year 2013. According to World Bank data, Kenya is classified as a lower middle-income country and has a youthful population where more than 80% of its citizens are aged 35 years and below. According to the most recent national census (2019), Kenya had a population of 47 million people (51% women). Out of the 47 million people, 19.5 million people were living below the poverty line and 60% of them were living in slums and informal settlements. Most households in informal settlements face chronic poverty and are also linked to gendered disadvantage where low income and female-headed households record elevated levels of food insecurity.

Nairobi and Mombasa counties are predominantly urban, while Kwale is largely rural. In Nairobi, the study was conducted in Westlands, one of the county’s seventeen sub-counties. In Mombasa, the study was carried out in Changamwe, one of the county’s six sub-counties. In Kwale, the study was conducted in Matuga, one of the county’s five sub-counties. Changamwe and Westlands are both urban informal settlements. They are both densely populated, and most of the inhabitants confront poor housing conditions, overcrowding, poor infrastructure, inadequate access to social amenities, water and sanitation services. Kwale on the other hand, is a rural setting, experiencing a high level of food insecurity and facing the highest poverty index in the country (approximately 67%) of the residents are living below the poverty line, higher than the national estimate.

The data used in this study was obtained from the formative phase of the ‘Advancing Gender Equality through Civil Society’ (AGECS) mental health research project being implemented at the Aga Khan University, Institute for Human Development in Kenya. AGECS is a mixed methods research project that seeks to understand women’s mental health needs and tailor programmes that address these needs. Further details of the project have been described elsewhere [14].

Participants

Our target population in the cross-sectional survey was adult men and women aged 18 years and older living in the three counties of interest. In each of the three study locations, we liaised with the sub-county community health focal persons to create a list of all the administrative wards within the sub-county and used simple random sampling to select three wards. A similar process was used in selecting the subsequent community units and villages from where participants recruitment was conducted. Recruitment was stratified by study site, and gender. Sequential sampling was used to recruit participants at the household level, i.e., from among those available at the time of the household visit and this process continued until the desired sample size was attained. In cases where more than one person in a household was eligible for recruitment, only one of them was included in the study – and this was decided by the respective participants. Recruitment of participants was carried out by community health volunteers (CHVs) at the household level. We drew on the CHVs’ knowledge of both the geography and local population of the study sites to identify potential clients. To be included in the survey, participants had to be at least 18 years old, be able to provide informed consent, and be able to speak Swahili or English. The majority of the assessments were taken in Swahili. Participants recruitment began on 27^th September 2022 and the last person was recruited on 8^th December 2022.

Once recruited by CHVs, the participants were booked for assessment, usually the following day, at a central venue in the community, which included social halls. All participants who turned up on the day of the assessment were consented to the study by a team of 22 research assistants (10 women). The team of research assistants received a 1-week training in both sites, Nairobi and the Coast of Kenya (Mombasa and Kwale). Topics covered during the training sessions included an overview of psychological constructs, a review of measures, psychological first aid, the referral pathway for clients who would require this service, research ethics, data management and how to conduct face-to-face interviews. During the training period, the enumerators were given a chance to do role plays under the observations of a counsellor, data manager, and a global mental health practitioner. Pre-testing of tools was also done through a pilot study before the eventual implementation of the study.

All the data was collected face-to-face utilizing the Open Data Kit through tablets that were password-protected and encrypted to avoid data loss. The assessments took approximately an hour. Thereafter, a small refreshment was given to the participants. Participants were also reimbursed for their transport costs. A data manager double-checked any inconsistencies in the data before uploading it to the server on a daily basis.

The final sample included in the analysis comprised of 1048 women and 480 men. 353 were from Kwale County, 583 from Mombasa County and 592 from Nairobi County. Further recruitment details are captured in Fig 1.

Fig 1. Participant recruitment flowchart.

Variables

Outcome (dependent) variables included: depressive symptoms, anxiety symptoms, PTSD symptoms, alcohol and drug use problems. We also collected several independent variables including sociodemographic characteristics, lifestyle and health history information, and psychosocial information.

Data sources/measurement

Outcome measures.

9-item patient health questionnaire (PHQ-9): We used the PHQ-9 to screen for depressive symptoms among participants. The tool has 9 items that are scored on a 4-point Likert scale (0 = not at all, 3 = nearly every day). The total scores range from 0 to 27. Total scores of 5–9, 10–14, and 15–27 indicate mild, moderate and severe depressive symptoms, respectively. The Swahili version of PHQ-9 has been validated extensively in Kenya, and there is empirical evidence that it is suitable for use among different populations, including adolescents [15], adults [16] and healthcare providers [17]. In the current study, the tool registered good internal consistency (Cronbach’s alpha = 0.83).

7-item generalized anxiety disorder questionnaire (GAD-7): GAD-7 was used to screen for anxiety among respondents. The items are scored on a 4-point Likert scale ranging from 0 (not at all) to 3 (more than half the days). The total scores range from 0 to 21, with cut-offs of 5–9, 10–14 and 15–21 representing mild, moderate and severe anxiety symptoms, respectively. The tool has previously been validated in Kenya, maintaining its unidimensional latent structure [18]. In the present study, the tool demonstrated good internal reliability with an alpha value of 0.87.

5-item primary care PTSD screen (PC-PTSD-5): We used the PC-PTSD-5 to assess for PTSD. Participants first complete a screening question to determine lifetime exposure to trauma. Respondents who have not experienced any exposure to trauma are assigned a score of zero. Those who acknowledge exposure to trauma in the past month are asked five questions, which are scored dichotomously (0 = no; 1 = yes). The total scores range from 0 to 5. We used a cut-off of ≥3 to indicate a positive screen for PTSD in the current study, similar to previously published studies [19–21]. PC-PTSD-5 was validated in the current study, yielding good psychometric properties [14].

10-item Alcohol Use Disorders Identification Test (AUDIT): AUDIT was used to screen for patterns of hazardous/harmful alcohol use and alcohol dependence. All participants first responded to a screening question (yes/no) assessing whether they currently used any alcoholic drink. Participants responding yes went ahead and answered the subsequent AUDIT items. The tool has a total score of 40. A total score of ≥8 and ≥6 for males and females, respectively, was used to define a positive screen for hazardous drinking, similar to previously published studies [22]. In the current study, the tool demonstrated excellent internal reliability, with an alpha value of 0.91.

11-item Drug Use Disorders Identification Test (DUDIT): We used DUDIT to screen for hazardous drug use. Like the AUDIT, items are summated with a maximum score of 44. We used a cut-off score of ≥6 and ≥2 for males and females, respectively, to define a probable positive screen for problematic drug use, similar to previous research in the study setting [22]. Similarly, the tool demonstrated excellent internal consistency, with an alpha value of 0.98

Independent variables.

Sociodemographic characteristics: We collected data on the participant’s age, sex, marital status, educational level, religion, participant’s occupation, household size, household income, type of family, living arrangements, family debt, and caregiving status. Participants’ socioeconomic status was assessed using an asset index that has been previously used in the study setting [23]. The tool screened for the ownership of a list of disposable assets by participants (or their families) such as radio, television, bicycle, and motorbike. A single unweighted score of socioeconomic status was generated, with a higher score translating to a higher socioeconomic status.

Lifestyle and health history information: Information collected included participant’s self-reported health status, lifetime experience of verbal abuse (e.g., belittling, humiliation, manipulation, intimidation and control tactics), physical abuse, sexual abuse (e.g., rape, coercion, unwanted acts, withholding sex) from a spouse, self-reported hearing and vision, history of any chronic condition (e.g., diabetes, hypertension, cancer, HIV), somatic complaints (e.g., migraine, fatigue, pain), and self-reported physical activity. Additional information sought from women included pregnancy status, age at first pregnancy, pregnancy in the previous year, number of live pregnancies, sexual activity, and history of infertility.

Psychosocial measures.

The 17-item Intimate Partner Violence Attitude Scale – Revised (IPVAS—R) was used to assess participants’ attitudes towards intimate partner violence [24]. Higher scores indicate a greater level of negative attitudes towards IPV. The tool registered acceptable internal reliability in the current study, with an alpha value of 0.70.

The 12-item Multidimensional Scale of Perceived Social Support (MSPSS) [25] was used to assess participants’ social support. A higher score indicates greater support perceived by an individual. The measure exhibited good internal consistency in the current study, with an alpha value of 0.82.

The 5-item WHO Wellbeing Index [26] was used to assess participants’ subjective wellbeing. A higher score indicates greater wellbeing by a participant. The tool has previously been validated in the study setting [27]. In the current study, the tool yielded good internal consistency, with an alpha value of 0.83.

The 13-item Stressful Life Events Screening Questionnaire (SLESQ) [28] was used to assess traumatic event exposure among participants. A summated score was generated to show the total number of life events reported. In the current study, the tool demonstrated acceptable internal reliability, with an alpha value of 0.70.

Referral and management of participants at risk of suicide and those experiencing severe symptoms of mental health problems: In 2013, Kenya transitioned into a devolved system of governance consisting of two levels: the national government and 47 semiautonomous county governments. Under the devolution system, healthcare service delivery function was transferred to county governments while the national government retained policy and regulatory functions. Mental health service in Kenya is provided by public health hospitals (1 national referral hospital and 25 regional/county psychiatric units within primary care facilities), private-for-profit facilities and non-governmental organizations. Both public and private facilities charge user fees to clients, however, the fee is subsidized in public facilities) with private facilities generally serving the wealthier individuals in the country.

Suicide risk among participants in this study was assessed using the PHQ-9 instrument (the 9^th Item). All participants who screened positive were connected to the relevant sub-county mental health team that had a standby psychologist for further assessment and management. Additionally, participants presenting moderate to severe symptoms of mental health and substance use problems (as per the tools cut-offs) were also referred using the same pathway. This referral approach was adopted after careful consideration with the relevant county and sub-county health management teams during the community engagement period before data collection began.

Study size

We estimated the sample size using power analyses in Stata based on previous studies [29,30]. An overall sample of at least 1000 participants was required to detect a significant difference in mental and substance use problems between males and females at 80% power and a 5% level of statistical significance. This was sufficient to detect a prevalence difference of 10%. A final sample of 1528 was considered sufficient, allowing for refusals/non-response (10%), missing or incomplete records (10%) and conducting stable psychometric models for some of the measures. A one-group sample (n = 400) from the above sample calculation was > 90% powered to conduct a logistic regression analysis of the correlates of mental health problems.

Statistical methods

We analyzed our data using STATA version 17.0 (StataCorp LP, College Station, TX, United States). Sociodemographic and health history variables were summarized using descriptive statistics: mean and standard deviation for continuous variables and frequency and proportion for categorical variables. We used proportions to estimate the prevalence of mental and substance use problems among participants. We used logistic regression models to explore univariate associations between binary outcome variables and various exposure variables. Independent variables with a p-value <0.15 in the univariate analysis were then fitted into the multivariable models using forward selection. In all the multivariable models, we performed collinearity diagnostics using STATA’s ‘collin’ syntax and no multicollinearity problems were identified based on an interpretation of the variance inflation factor. A two-tailed p-value of <0.05 was deemed statistically significant for all hypothesis tests. The overall fit of the final models was assessed using Hosmer and Lemeshow’s goodness of fit, and a p-value of >0.05 was considered a good fit.

Ethics

The study was approved by the Aga Khan, Nairobi Institutional Scientific and Ethics Review Committee (Ref: 2022/ISERC_44(V2)). Permission to carry out the study in Kenya was sought and granted by the National Commission for Science, Technology and Innovation (Ref: 346643). Respective County departments of health gave local permits to conduct the study: Nairobi (Ref: NCCG/DHS/REC/240), Mombasa (Ref: MCG/COPH/RCH./111) and Kwale (Ref: CG/KWL/6/6/1/CECM/39/VOL.1/34). All respondents provided written informed consent for their participation in the study. The consenting officer used follow up questions (when undertaking the informed consenting process with the client) to assess subject understanding. We also used wait periods (a day) between recruitment and actual assessment to allow more time for potential subject to consider the information that had been presented before verbalizing/communicating their participation decision. On the day of assessment, we also conducted a quick medical check-up to make sure the client had the capacity to undertake the assessments.

Results

Sample characteristics

The characteristics of the included participants are presented in Table 1. The overall sample comprised 1528 participants (68.6% women), with a mean age of 39.1 years (SD = 13.1). The vast majority of the participants were from urban informal settlements (77.6%), had attained basic education or more (94.7%), were married (75.4%), Christians (64.1%), living in nuclear families (74.8%), living in rented houses (67.8%), in debt (83.1%) and experiencing food insecurity (71.7%). Women were likely to be unemployed, have lower monthly household income, have lower socioeconomic status, report poor self-rated health status, experience physical abuse from a spouse, and have chronic conditions. Further details are highlighted in Table 1.

Table 1. Characteristics of the study population by gender, n = 1528.

Characteristic	Total sample n = 1528	Gender
		Females, n = 1048	Males, n = 480	p-value
Sociodemographic
Area of residence
Kwale (rural)	343 (22.4%)	254 (24.2%)	89 (18.5%)	0.01
Mombasa & Nairobi (Urban)	1185 (77.6%)	794 (75.8%)	391 (81.5%)
Age in years, missing = 2				<0.001
18–34	657 (43.1%)	484 (46.3%)	173 (36.0%)
35–49	536 (35.1%)	359 (34.3%)	177 (36.9%)
≥50	333 (21.8%)	203 (19.4%)	130 (27.1%)
Level of education
Tertiary	196 (12.8%)	130 (12.4%)	66 (13.8%)	0.001
Secondary	521 (34.1%)	328 (31.3%)	193 (40.2%)
Primary	731 (47.8%)	525 (50.1%)	206 (42.9%)
None	80 (5.3%)	65 (6.2%)	15 (3.1%)
Employment
Skilled/Professional	315 (20.6%)	165 (15.7%)	150 (31.3%)	<0.001
Unskilled/casual	705 (46.1%)	457 (43.6%)	248 (51.7%)
Unemployed	508 (33.3%)	426 (40.75)	82 (17.0%)
Marital status, missing = 5
Never married	103 (6.8%)	97 (9.3%)	6 (1.3%)	<0.001
Separated/Divorced/Widowed	271 (17.8%)	269 (25.7%)	2 (0.4%)
Married/cohabiting	1149 (75.4%)	680 (65.0%)	469 (98.3%)
Religion
Christian	980 (64.1%)	670 (63.9%)	310 (64.6%)	0.37§
Muslim	545 (35.7%)	377 (36.0%)	168 (35.0%)
No religion	3 (0.2%)	1 (0.1%)	2 (0.4%)
Type of family
Nuclear	1143 (74.8%)	722 (68.9%)	421 (87.7%)	<0.001
Extended	245 (16.0%)	187 (17.8%)	58 (12.1%)
Single	140 (9.2%)	139 (13.3%)	1 (0.2%)
Living arrangements
Own house (not paying rent)	492 (32.2%)	344 (32.8%)	148 (30.8%)	0.44
In a rented house	1036 (67.8%)	704 (67.2%)	332 (69.2%)
Size of household
1–2	193 (12.6%)	143 (13.7%)	50 (10.4%)	0.19
3–4	640 (41.9%)	430 (41.0%)	210 (43.8%)
≥5	695 (45.5%)	475 (45.3%)	220 (45.8%)
Monthly household income (Ksh)
<10000	976 (63.9%)	727 (69.4%)	249 (51.9%)	<0.001
10000–29999	490 (32.1%)	281 (26.8%)	209 (43.5%)
>30000	62 (4.0%)	40 (3.8%)	22 (4.6%)
Household in debt, missing = 20
Yes	1253 (83.1%)	848 (82.3%)	405 (84.7%)	0.25
No	255 (16.9%)	182 (17.7%)	73 (15.3%)
Household food insecurity (last 3 months)
Yes	1095 (71.7%)	759 (72.4%)	336 (70%)	0.33
No	433 (28.3%)	289 (27.6%)	144 (30%)
Family caring for a child living with a disability or chronic illness
Yes	88 (5.8%)	71 (6.8%)	17 (3.5%)	0.01
No	1440 (94.2%)	977 (93.2%)	463 (96.5%)
Asset index score – mean (SD)	2.9 (1.4)	2.8 (1.3)	3.2 (1.4)	<0.001
Health history
Self-reported health status
Good	411 (26.9%)	264 (25.2%)	147 (30.6%)	0.01
Average	968 (63.3%)	667 (63.7%)	301 (62.7%)
Poor	149 (9.8)	117 (11.1%)	32 (6.7%)
Lifetime experience of verbal abuse from a spouse
Yes	830 (54.3%)	550 (52.5%)	280 (58.3%)	0.03
No	698 (45.7%)	498 (47.5%)	200 (41.7%)
Lifetime experience of physical abuse from a spouse
Yes	430 (28.1%)	385 (36.7%)	45 (9.4%)	<0.001
No	1098 (71.9%)	663 (63.3%)	435 (90.6%)
Lifetime experience of sexual abuse from a spouse
Yes	216 (14.1%)	140 (13.4%)	76 (15.8%)	0.20
No	1312 (85.9%)	908 (86.6%)	404 (84.2%)
Self-reported vision
Good	867 (56.7%)	567 (54.1%)	300 (62.5%)	0.01
Average	561 (36.7%)	406 (38.7%)	155 (32.3%)
Poor	100 (6.6%)	75 (7.2%)	25 (5.2%)
Self-reported hearing
Good	1347 (88.1%)	920 (87.8%)	427 (89.0%)	0.57§
Average	157 (10.3%)	109 (10.4%)	48 (10.0%)
Poor	24 (1.6%)	19 (1.8%)	5 (1.0%)
Participant living with a chronic condition
Yes	549 (35.9%)	405 (38.7%)	144 (30%)	0.001
No	979 (64.1%)	643 (61.3%)	336 (70%)
Presence of somatic complaints
Yes	812 (53.1%)	593 (56.6%)	219 (45.6%)	<0.001
No	716 (46.9%)	455 (43.4%)	261 (54.4%)
Average number of days engaged in vigorous intensity physical activity (in the past 7 days), Mean (SD)	1.5 (2.3)	1.2 (2.1)	2.3 (2.6)	<0.001
Average number of days engaged in moderate intensity physical activity (in the past 7 days), Mean (SD)	3.7 (2.6)	3.9 (2.5)	3.2 (2.8)	<0.001
Average number of days engaged in light intensity physical activity (in the past 7 days), Mean (SD)	4.9 (2.5)	4.7 (2.6)	5.4 (2.3)	<0.001
Reproductive information for females
Ever pregnant before
Yes	–	984 (93.9%)	–	N/A
No	–	64 (6.1%)	–
Age at first pregnancy
<18 years	–	407 (41.4%)	–	N/A
Above 18 years	–	577 (58.6%)	–
Pregnant in the previous year
Yes	–	174 (17.7%)	–	N/A
No	–	810 (82.3%)	–
Difficulty conceiving for the past 1 year
Yes	–	71 (6.8%)	–	N/A
No	–	977 (93.2%)	–
Currently sexually active
Yes	–	695 (66.3%)	–	N/A
No	–	353 (33.7%)	–
WHO-5 Wellbeing Index – Mean (SD)	12.3 (6.5)	12.0 (6.5)	12.9 (6.5)	0.01
Multidimensional Scale of Perceived Social Support – mean (SD)	61.0 (12.0)	60.0 (12.3)	63.2 (10.8)	<0.001
Stressful life events scale – mean (SD)	3.0 (2.4)	2.9 (2.4)	3.4 (2.4)	<0.001
Intimate Partner Violence Attitude Scale – Mean (SD)	36.0 (6.8)	35.9 (6.8)	36.1 (6.9)	0.61

Note: P-values are for differences between women and men. The P-values have been derived from the Chi-square test (or Fisher’s exact test) and independent Student’s t-test for categorical and continuous variables, respectively.

Abbreviations: Ksh, Kenya shillings; SD, standard deviation.

^§ Based on Fisher’s exact test

Prevalence of mental and substance use problems

Table 2 presents the prevalence estimates for mental, and substance use problems across the sample and disaggregated estimates by gender (women vs. men). Group differences in mental and substance use problems are also compared statistically and presented in this table.

Table 2. Prevalence of mental and substance use problems among participants.

	Whole sample, n = 1528		Females, n = 1048		Males, n = 480		P-value
	Freq.	Prevalence (95% CI)	Freq.	Prevalence (95% CI)	Freq.	Prevalence (95% CI)
Common mental disorders
Depressive symptoms	395	25.9 (23.7, 28.1)	290	27.7 (25.0, 30.5)	105	21.9 (18.3, 25.8)	0.02
Anxiety symptoms	167	10.9 (9.4, 12.6)	128	12.2 (10.3, 14.4)	39	8.1 (5.8, 10.9)	0.02
PTSD symptoms	327	21.4 (19.4, 23.5)	228	21.8 (19.3, 24.4)	99	20.6 (17.1, 24.5)	0.62
Alcohol and drug use problems
Any current alcohol use	191	12.5 (10.9, 14.3)	62	5.9 (4.6, 7.5)	129	26.9 (23.0, 31.1)	<0.001
Any current drug use	81	5.3 (4.2, 6.5)	21	2.0 (1.2, 3.0)	60	12.5 (9.7, 15.8)	<0.001
Hazardous alcohol use	110	7.2 (6.0, 8.6)	38	3.6 (2.6, 4.9)	72	15 (11.9, 18.5)	<0.001
Hazardous drug use	77	5.0 (4.0, 6.3)	22	2.1 (1.3, 3.2)	55	11.5 (8.7, 14.7)	<0.001

Abbreviation: 95% CI, 95% confidence interval

Mental health problems

The overall prevalence of depressive symptoms was 25.9% (95% CI 23.7, 28.1%), and that for anxiety symptoms was 10.9% (95% CI 9.4, 12.6%). The overall prevalence of symptoms of PTSD was 21.4% (95% CI 19.4, 23.5%). Notably, women reported significantly higher frequency of depressive and anxiety symptoms than men (Table 2). A total of 333 (21.8%) of the participants screened positive for suicide ideation in the sample.

Substance use problems

The overall point prevalence of any current alcohol use was 12.5% (95% CI 10.9, 14.3%) and that for any current drug use was 5.3% (95% CI 4.2, 6.5%). Men reported significantly higher prevalence of any current alcohol use and drug use than women (Table 2). The overall past-year prevalence of hazardous alcohol and drug use was 7.2% (95% CI 6.0, 8.6%) and 5.0% (95% CI 4.0, 6.3%), respectively. Similarly, men reported significantly higher frequency of past-year hazardous alcohol and drug use than women (Table 2).

Correlates of mental and substance use problems

Table 3 summarizes the correlates of mental health problems among women, while Table 4 summarizes the results for men. Table 5 presents the correlates of any current substance in both men and women. We have also provided simplified tables of the correlates of mental and substance use problems among participants in S1 and S2 Tables.

Table 3. Univariate and multivariable analysis of correlates of mental health problems among women, n = 1048.

Covariate	Positive screen for depressive symptoms		Positive screen for anxiety symptoms				Positive screen for PTSD symptoms
	Univariate analysis OR (95% CI)	Multivariable analysis aOR (95% CI)	Univariate analysis OR (95% CI)	Multivariable analysis aOR (95% CI)		Univariate analysis OR (95% CI)		Multivariable analysis aOR (95% CI)
Residence
Rural setting	Ref	Ref	Ref		Ref		Ref	Ref
Urban setting	3.64 (2.42, 5.47)**	3.35** (2.08, 5.39)	3.14** (1.74, 5.68)		2.76* (1.41, 5.40)		2.18** (1.46, 3.25)	1.72* (1.11, 2.65)
Age in years
18–34	Ref	Ref	–		–		Ref	Ref
35–49	1.91** (1.41, 2.58)	1.24 (0.85, 1.81)	–		–		1.27 (0.92, 1.76)	0.89 (0.63, 1.28)
≥50	1.11 (0.76, 1.63)	0.71 (0.42, 1.23)	–		–		0.72 (0.47, 1.11)	0.55* (0.33, 0.90)
Educational level
None	Ref	Ref	Ref		Ref		Ref	Ref
Primary	3.97* (1.68, 9.39)	3.46* (1.32, 9.10)	9.20* (1.26, 67.46)		8.79* (1.16, 66.79)		4.67* (1.66, 13.09)	3.58* (1.22, 10.54)
Secondary	4.31** (1.80, 10.32)	4.21* (1.55, 11.45)	9.66* (1.31, 71.43)		11.11* (1.43, 86.30)		4.92* (1.73, 13.95)	3.33* (1.10, 10.05)
Tertiary	3.35* (1.32, 8.46)	3.54* (1.20, 10.46)	10.29* (1.34,78.86)		15.15* (1.85, 123.97)		2.94 (0.96, 8.95)	2.06 (0.63, 6.72)
Occupation
Skilled/Professional	–	–	Ref		Ref		–	–
Unskilled/casual labourer	–	–	2.04* (1.10, 3.81)		2.02* (1.02, 4.01)		–	–
Unemployed	–	–	1.45 (0.76, 2.76)		2.13* (1.05, 4.33)		–	–
Type of family
Nuclear	–	–	–		–		Ref	Ref
Extended	–	–	–		–		0.65 (0.42, 1.01)	0.76 (0.47, 1.22)
Single	–	–	–		–		1.82* (1.23, 2.71)	1.52* (1.00, 2.32)
Self-reported vision
Good	Ref	Ref	–		–		–	–
Average	1.07 (0.80, 1.43)	0.82 (0.57, 1.19)	–		–		–	–
Poor	4.06** (2.47, 6.66)	2.03* (1.07, 3.83)	–		–		–	–
Stressful life events
None	Ref	Ref	Ref		Ref		Ref	Ref
>5 events	1.62 (0.98, 2.66)	1.14 (0.65, 1.97)	1.96 (0.87, 4.44)		1.42 (0.61, 3.32)		2.78* (1.45, 5.34)	2.65* (1.36, 5.14)
≥5 events	4.83** (2.94, 7.95)	2.42* (1.38, 4.23)	5.86** (2.63,13.03)		3.04* (1.31, 7.04)		7.16** (3.73, 13.71)	5.95** (3.06,11.60)
Self-perceived wellbeing	0.87** (0.85, 0.89)	0.90** (0.87, 0.92)	0.84** (0.81, 0.88)		0.87** (0.83, 0.91)		–	–
Sexually active
Yes	Ref	Ref	–		–		–	–
No	1.73** (1.30, 2.28)	1.40 (0.99, 1.99)	–		–		–	–
Self-rated health status
Good	Ref	Ref	Ref		Ref		Ref	Ref
Average	2.33** (1.59, 3.42)	1.74* (1.12, 2.71)	2.05* (1.17, 3.59)		1.33 (0.73, 2.43)		2.07** (1.38, 3.09)	2.10** (1.37, 3.22)
Poor	7.18** (4.35, 11.85)	4.36** (2.30, 8.24)	6.35** (3.33, 12.09)		2.86* (1.37, 6.00)		3.25** (1.92, 5.52)	3.32** (1.85, 5.94)
Food insecurity	1.85** (1.33, 2.58)	1.70* (1.14, 2.53)	1.75* (1.10, 2.79)		1.56 (0.93, 2.61)		1.61* (1.13, 2.30)	1.47* (1.01, 2.15)
Family debt	2.78** (1.78, 4.34)	1.68* (1.01, 2.79)	–		–		–	–
Perceived social support	0.41** (0.31, 0.54)	0.61* (0.44, 0.85)	0.42** (0.29, 0.62)		0.63* (0.41, 0.96)		–	–
n for the final model		1028			1048			1046
Variance explained		24.2%			21.1%			11.1%
Hosmer-Lemeshow test		X2 = 994.32; p-value = 0.37			X2 = 830.93; p-value = 0.91			X2 = 392.84; p-value = 0.04
cvMean AUC (95% CI)		0.82 (0.79, 0.85)			0.82 (0.79, 0.86)			0.73 (0.70, 0.77)

Abbreviations: aOR, adjusted odds ratio; cvMean AUC, cross-validated mean area under the curve for the multivariable model; OR, odds ratio; Ref, reference group

^*p-value <0.05,

^**p-value ≤0.001

Table 4. Univariate and multivariable analysis of correlates of mental health problems among men, n = 480.

Covariate	Positive screen for depressive symptoms		Positive screen for anxiety symptoms		Positive screen for PTSD
	Univariate analysis OR (95% CI)	Multivariable analysis aOR (95% CI)	Univariate analysis OR (95% CI)	Multivariable analysis aOR (95% CI)	Univariate analysis OR (95% CI)	Multivariable analysis aOR (95% CI)
Residence
Rural setting	–	–	Ref	Ref	–	–
Urban setting	–	–	9.47* (1.28, 69.95)	8.75* (1.17, 65.42)	–	–
Monthly household Income (Ksh)
<10,000	–	–	–	–	Ref	Ref
10000–29999	–	–	–	–	0.52* (0.33, 0.84)	0.49* (0.30, 0.82)
>30,000	–	–	–	–	0.46 (0.13, 1.59)	0.40 (0.11, 1.48)
Somatic symptoms	–	–	–	–	2.27** (1.44, 3.57)	1.74* (1.08, 2.83)
Stressful life events
None	Ref	Ref	Ref	Ref	Ref	Ref
>5 events	1.45 (0.54, 3.91)	1.47 (0.52, 4.17)	1.00 (0.22, 4.66)	1.16 (0.24, 5.48)	5.58 (0.74, 42.11)	5.00 (0.66, 38.07)
≥5 events	3.15* (1.18, 8.42)	2.57 (0.90, 7.30)	2.88 (0.66, 12.69)	3.65 (0.81, 16.50)	21.59* (2.90, 160.49)	17.69* (2.34, 133.46)
Self-perceived wellbeing	0.87** (0.83, 0.90)	0.88** (0.84, 0.92)	–	–	–	–
Sexual abuse	3.07** (1.82, 5.18)	2.29* (1.27, 4.15)	–	–	–	–
Verbal abuse	–	–	–	–	2.36** (1.44, 3.86)	1.71* (1.01, 2.90)
Living in rented houses	2.20* (1.29, 3.75)	2.50* (1.39, 4.48)	–	–	–	–
Days spent on light activities in the previous week	1.14* (1.02, 1.27)	1.16* (1.03, 1.31)	–	–	–	–
Days spent on vigorous activities in the previous week	–	–	0.85* (0.73, 0.99)	0.83* (0.71, 0.98)	–	–
Perceived social support	0.46** (0.30, 0.72)	0.58* (0.36, 0.96)	0.42** (0.29, 0.62)	0.39* (0.20, 0.78)
n for the final model		480		480		480
Variance explained		18%		12%		13.5%
Hosmer-Lemeshow test		X2 = 341.32; p-value = 0.38		X2 = 26.37; p-value = 1.00		X2 = 26.75; p-value = 0.32
cvMean AUC (95% CI)		0.79 (0.74, 0.83)		0.75 (0.68, 0.83)		0.76 (0.71, 0.81)

Abbreviations: aOR, adjusted odds ratio; cvMean AUC, cross-validated mean area under the curve for the multivariable model; OR, odds ratio; Ref, reference group

^*p-value <0.05,

^**p-value ≤0.001

Table 5. Univariate and multivariable analysis of correlates of current substance use among men and women.

Covariate	Current alcohol use in women, n = 1048		Current alcohol use in men, n = 480		Current drug use, n = 1528
	Univariate analysis OR (95% CI)	Multivariable analysis aOR (95% CI)	Univariate analysis OR (95% CI)	Multivariable analysis aOR (95% CI)	Univariate analysis OR (95% CI)	Multivariable analysis aOR (95% CI)
Residence
Rural setting	Ref	Ref	–	–	–	–
Urban setting	21.05* (2.90, 152.66)	15.78* (2.14, 116.54)	–	–	–	–
Female sex	–	–	–	–	0.14** (0.09, 0.24)	0.13** (0.08, 0.22)
Marital status
Never married	Ref	Ref	–	–	–	–
Separated/divorced/widowed	0.80 (0.38, 1.70)	1.16 (0.49, 2.72)	–	–	–	–
Married/cohabiting	0.31* (0.15, 0.65)	0.34* (0.14, 0.84)	–	–	–	–
Religion
Christian	–	–	Ref	Ref	–	–
Muslim	–	–	0.14** (0.08, 0.26)	0.19** (0.10, 0.36)	–	–
Occupation
Skilled/Professional	–	–	–	–	Ref	Ref
Unskilled/casual labourer	–	–	–	–	1.62 (0.86, 3.05)	2.05* (1.07, 3.93)
Unemployed	–	–	–	–	1.05 (0.52, 2.12)	2.48* (1.18, 5.22)
Household size
1–2	–	–	Ref	Ref	–	–
3–4	–	–	1.38 (0.71, 2.69)	1.12 (0.55, 2.30)	–	–
>5	–	–	0.46* (0.23, 0.92)	0.49 (0.23, 1.05)	–	–
Monthly household income (Ksh)
<10000	–	–	Ref	Ref	–	–
10000–29999	–	–	1.79* (1.17, 2.74)	1.36 (0.85, 2.18)	–	–
>30000	–	–	4.66** (1.91, 11.39)	3.32* (1.21, 9.13)	–	–
Type of family
Nuclear	Ref	Ref	–	–	–	–
Extended	0.65 (0.27, 1.57)	0.69 (0.26, 1.79)	–	–	–	–
Single	3.49** (1.97, 6.21)	2.79* (1.27, 6.10)	–	–	–	–
Verbal abuse	–	–	–	–	2.07* (1.27, 3.37)	1.53 (0.89, 2.61)
Sexual abuse	–	–	–	–	2.41** (1.44, 4.02)	1.78* (1.01, 3.14)
Self-reported vision
Good	Ref	Ref	–	–	–	–
Average	0.84 (0.47, 1.48)	0.87 (0.47, 1.61)	–	–	–	–
Poor	2.21* (1.01, 4.81)	2.24 (0.94, 5.33)	–	–	–	–
Stressful life events
None	Ref	Ref	Ref	Ref	Ref	Ref
>5 events	1.99 (0.58, 6.77)	1.36 (0.39, 4.76)	1.03 (0.46, 2.29)	1.21 (0.51, 2.83)	2.27 (0.80, 6.47)	1.68 (0.58, 4.89)
≥5 events	6.36* (1.93, 20.89)	3.53* (1.03, 12.04)	1.79 (0.81, 3.98)	2.05 (0.87, 4.83)	4.03* (1.42, 11.38)	2.38 (0.81, 6.98)
Multimorbidity	Ref	Ref	0.41** (0.25, 0.68)	0.56* (0.32, 0.97)
Sexually active
Yes	Ref	Ref	–	–	–	–
No	1.80 (0.98, 3.31)	6.68** (3.10, 14.40)
n for the final model		1046		478		1528
Variance explained		19.1%		15.5%		14%
Hosmer-Lemeshow test		X2 = 148.67; p-value = 0.73		X2 = 47.45; p-value = 0.84		X2 = 106.4; p-value = < 0.001
cvMean AUC (95% CI)		0.82 (0.77, 0.87)		0.76 (0.72, 0.81)		0.80 (0.76, 0.84)

Abbreviations: aOR, adjusted odds ratio; cvMean AUC, cross-validated mean area under the curve for the multivariable model; OR, odds ratio; Ref, reference group

^*p-value <0.05,

^**p-value ≤0.001

Correlates of mental health problems among women.

In the final multivariable analysis model (Table 3), urban residence, higher educational status, household food insecurity, household debt, poor self-reported eyesight, experiencing more than five stressful life events, and declining self-rated health status were significantly associated with higher odds of depressive symptoms in women. On the other hand, urban residence, higher educational status, poor self-reported health, experiencing more than five stressful life events, and being unemployed or casual labourer were associated with higher odds of anxiety symptoms among women in the final multivariable model. Increasing self-reported wellbeing and perceived social support were associated with reduced odds of depressive and anxiety symptoms among women. Factors associated with higher odds of PTSD symptoms in the final multivariable analysis included primary/secondary level of education, urban residence, household food insecurity, declining self-rated health status, and experiencing stressful life events. Older age (>50 years) was associated with reduced odds of PTSD symptoms among women.

Correlates of mental health problems among men.

In the final multivariable analysis (Table 4), sexual abuse from a spouse (e.g., coercion, assault, withholding sex, undesired sexual acts) living in rented houses, and light physical activity were associated with higher odds of depressive symptoms. Conversely, urban residence was associated with increased odds of anxiety symptoms. On the other hand, verbal abuse, somatic complaints, and stressful life events were associated with higher odds of PTSD symptoms among men. Factors associated with reduced odds of mental health problems among men in the final multivariable analysis included increased vigorous exercise, increasing household income, perceived social support, and increasing self-reported wellbeing.

Correlates of substance use problems across the sample.

In the final multivariable analysis (Table 5), urban residence, experiencing more than five stressful life events, being sexually active, and living in a single-parent family were significantly associated with higher odds of any current alcohol use in women. On the other hand, higher monthly household income was significantly associated with higher odds of any current alcohol use among men. Protective factors of any current alcohol use included being married/cohabiting (in women), larger household size, being a Muslim, and multimorbidity (among men). Correlates for any current drug use across the sample in the final multivariable analysis (Table 5) included unemployment/casual work and lifetime experience of sexual abuse from a spouse. Female sex was significantly associated with reduced odds of any current drug use across the sample.

Discussion

Key findings

In this study, we documented the prevalence and associated factors of mental and substance use problems in a community sample of adults drawn from Nairobi, Mombasa and Kwale counties in Kenya. Our study is unique in that it encompasses a range of key mental and substance use problems. It gives a comprehensive account of the mental, general health and psychosocial correlates of these conditions. Based on a sample of 1528 adults (69% women), the present study found a relatively high prevalence of depressive symptoms (26%), symptoms of PTSD (21%), anxiety symptoms (11%), and current alcohol use (13%). However, the prevalence of current drug use and past year’s prevalence of hazardous alcohol and drug use were low (<10%) across the sample. Notably, the prevalence of depressive and anxiety symptoms was significantly higher among women compared to their male counterparts. On the other hand, both current and past-year alcohol and drug use problems were significantly higher in men than women. We found no significant differences between women and men in the prevalence of PTSD symptoms. In our second objective, we identified several factors (mostly demographic and psychosocial) that were associated with increased odds of mental and substance use problems, some overlapping between men and women (e.g., urban residence and experiencing stressful life events), others overlapping across conditions (e.g., urban residence, experiencing stressful life events, household food insecurity, and declining self-rated health) while others were unique across sex and conditions. Protective factors against mental health problems included higher perceived social support, higher subjective wellbeing, older age (>50 years) and higher household income (in both sexes), whereas those of substance use problems included being married/cohabiting (in women) and larger household size, being a Muslim and multimorbidity (among men).

Prevalence of mental and substance use problems

The prevalence of mental and substance use problems has been shown to vary widely in SSA, partly due to differences in study populations/participants, measures and cut-offs used, as well as study designs [31–35]. In this study, the reported prevalence estimates were calculated based on recommended cut-offs (for moderate levels of symptoms) of the screening measures used and these estimates were not confirmed through clinical diagnoses. Our approach takes into account the limited infrastructure and human resource capacity to address the burden of mental health problems in LMICs. Our finding of a 26% prevalence of depressive symptoms is comparable to earlier estimates of depressive symptoms in SSA [36,37]. This figure is also within the estimates of 4.2 –50.4%, which has been reported among general populations from several studies in Kenya [9,29,30] and similar settings in East Africa [38–42]. The high rates of depressive symptoms in our study could be due to several biological, demographic and psychosocial vulnerabilities, some of which have been confirmed through this study. Generally, fewer studies have examined the prevalence of anxiety and PTSD relative to other CMDs in Kenya and the wider SSA region. Our study directly addresses this gap in Kenya and the region. The reported prevalence of anxiety in our study (11%) is lower than that reported in a previous study (20%), which was conducted in a similar setting but during the COVID-19 pandemic in Kenya [29]. On the other hand, the reported prevalence of PTSD symptoms in this study (21%) is comparable to the prevalence of 22% reported in a systematic review and meta-analysis of studies conducted in SSA [34]. The observed prevalence is also higher than previously reported estimates in Kenya: 11% in a cross-sectional household survey in 2015 [43] and 16% among university students in 2015 [44]. However, our prevalence is lower compared to some previous reports in Kenya: 65% among healthcare providers in the early phase of COVID-19 [45], 62% among internally displaced persons [46] and 65% among patients [47]. This is expected given that the previous studies assessed high-risk populations. Our observations also confirm previous findings that the prevalence estimates of probable PTSD are highly variable across settings partly because of differences in study populations, contexts/regions, study designs, etc., thus highlighting the need for context-specific studies.

The prevalence estimates for current substance use observed in this study are lower than what has been previously reported in the literature in Kenya [48–50]; however, our estimates (for alcohol use) are comparable to the latest national estimates as per the national survey on the status of drugs and substance use in Kenya (NACADA) reports [51]. Similarly, the observed past-year prevalence estimates of hazardous drug use are lower than what has been reported previously in Kenya [51]. Variations in study populations (e.g., age), source of information (self-report versus other formats), study setting, and assessment instruments could partly explain the variances observed in the literature. The observations could also be due to the community sensitization on drugs and alcohol across the counties.

In our sex-stratified analysis, women presented with significantly higher prevalence estimates for depressive and anxiety symptoms than men; however, men presented with significantly higher levels of both current and past-year prevalence of alcohol and drug use problems. Our observation is consistent with previous findings in the literature [52]. These differences can partly be accounted for by biological and psychosocial vulnerabilities in the respective genders. The observed differences could also be explained by the gender-roles hypothesis which asserts that the variation in prevalence estimates could be due to differences in the typical stressors, coping resources, and opportunity structures for expressing psychological distress made available differentially to men and women. Despite the observed differences in depressive and anxiety symptoms between men and women in this study, our study points to a narrowing in the sex-stratified prevalence of these problems in this setting – emphasizing the importance of addressing these conditions in both men and women. Few studies have aggregated their results by sex in Kenya; hence, more studies will provide more insights into this finding.

Correlates of mental and substance use problems

Assessing sex disparity in mental health research is crucial for targeted support and intervention, yet often overlooked in Kenya and other settings in SSA. In addressing this gap, we stratified our analysis of the correlates of mental and substance use problems by sex. Urban residence and experiencing stressful life events were the only correlates common to both men and women for mental and substance use problems. Generally, the influence of urban environments on mental health is not well understood. There is some evidence showing that urban residence is associated with higher risks of mental health problems – especially common mental disorders and substance use disorders [53], although results are contradicting [54]. The finding that experiencing stressful life events was associated with higher odds of mental and substance use problems is supported by results from previous research across different settings [55–57]. Our findings also revealed an overlap of some of the correlates (e.g., urban residence, stressful live events, food insecurity and declining self-rated health) across mental and substance use problems. This potentially suggests that individuals who experience these shared cumulative correlates are more likely to face multiple mental health challenges. Furthermore, it also implies that actions taken to address any of these shared risk factors are likely to have a promotive spillover effect across the different mental health problems.

Correlates unique to women’s mental health problems included household debt, poor self-reported eyesight, being unemployed or casual labourer, and higher educational status (relative to none). On the other hand, being sexually active and single-parent households were uniquely associated with higher odds of current alcohol use in women. These findings are supported by previous research in different contexts [58–66]. Among men, the unique correlates of mental health problems included sexual and verbal abuse from a spouse, living in rented houses, somatic complaints and taking part in light physical activity. Higher monthly household income was significantly associated with higher odds of current alcohol use in men. Previous studies have suggested that income and education are differentially associated with drinking patterns [67]; hazardous drinking is greater for individuals with low-educational levels (presumably because of frequent exposures to social stress and low health literacy) while people with higher income tend to have frequent light consumption presumably because they can afford it, have more social opportunities that include alcohol such as networking events, social norms supportive of light and frequent drinking. However, these associations may be complicated by multiple moderating factors including ethnicity, household arrangements, gender, and race. Future studies are needed to more fully explore the underlying mechanism of this association.

The finding that sexual and verbal abuse was a significant correlate for mental health problems in men but not women was a surprise finding in this study. Emerging data from empirical research and anecdotal reports shows that men in SSA are experiencing increasing ‘hidden’ reports of sexual and other forms of violence [68,69]. Sexual violence against men is a significant problem that has largely been ignored by different practitioners and policy makers. Besides, knowledge of its prevalence, nature and characteristics remains scarce in many settings in SSA and available data is likely underreported given the patriarchal nature of these societies and the gendered stereotypes associated with male experiences of sexual abuses: e.g., it is ‘unmanly’ for men to experience abuse and as such, their experiences may be interpreted as a sign of weakness. This notion of masculinity undermines help-seeking. While the barriers may be diverse, gendered socio-cultural views are important and sexual abuse services also fail to cater for the needs of men – making male victims seek for help from families and friends rather than formal sources. Our findings support the need for awareness or sensitization programs to educate the people on abuse against men – and the need for male victims to be informed and encouraged to seek help, and service providers be trained to articulate the needs of male survivors. In the present study, men reported slightly higher frequencies of both sexual abuse (16% vs 13%) and verbal abuse (58% vs 53%) incidences than women (though not statistically significant). Our study used a single item (yes/no) to document any lifetime experience of sexual abuse among the participants (prompts such as coercion, unwanted acts and withholding sex were provided). Wide-ranging conceptualizations of sexual abuse used in research can contribute to confusion in understanding the prevalence estimates and experiences of sexual abuse among men [70]. When sexual abuse is loosely defined as ‘forced sex’ heterosexual men in relationships with women may be less likely to pick themselves as victims. Among men, forced sex is more prevalent in relationships with other men than in heterosexual relationships – and gay men are more likely to report sexual abuse than heterosexual men. We propose further studies to explore this finding, e.g., the context of the abuse (heterosexual or gay relationships), nature of the abuse (e.g., assault or unwanted sexual acts). The association of somatic complaints and living in rented houses with higher odds of mental health problems is consistent with previous research findings [71–73]. The association between light-intensity physical activity and mental health problems is generally mixed in literature [74]. Additional studies are needed to understand the underlying mechanisms of this relationship.

Given the low prevalence estimates of drug use problems – we did not disaggregate the sample by sex when investigating the correlates of drug use in the study. Identified correlates for any current drug use included unemployment/casual work and sexual abuse – all of which are consistent with previous research [75,76]. A follow-up study with a larger sample size is encouraged to fully explore the determinants of drug use problems in this setting.

Identifying protective factors for mental and substance use problems is essential for the development of interventions/programs designed to prevent or lower the burden that these conditions place on individuals and provide future directions for public health policy. Higher perceived social support, older age (>50 years), increasing self-reported wellbeing, higher household income and vigorous physical activity were significantly associated with reduced odds of mental health problems. On the other hand, being married/cohabiting, larger household size, being a Muslim and having multimorbidity were significantly associated with reduced odds of current alcohol use. Our findings are consistent with prior research.

Strengths and limitations

To the best of our knowledge, this is among the few studies that have comprehensively assessed and documented the prevalence and correlates of mental and substance use problems among adults in rural and urban informal settlements in Kenya. Our approach in stratifying the results by gender has provided in-depth understanding of the unique issues affecting each gender, which is important in the design and implementation of appropriate preventive and treatment programs. We also utilised adapted and/or validated mental health tools and recruited experienced research assistants (both males and females) who were locals or could identify/relate or understand the local culture dynamics. All the research assistants completed a one-week training (full days) with a research psychologist, mental health practitioner, counsellor and data scientist and carefully covered different aspects of the research including understanding the different concepts, detailed review of individual tool’s items/questions, supervised administration of all the measures (in role plays) and finally, supervised practice administration to clients during a pilot phase of the study. During the actual data collection, the mental health researcher conducted spot checks.

Nonetheless, our findings should be interpreted in light of certain limitations. We utilised self-report screening measures to assess the prevalence and correlates of mental and substance use problems. Previous research has shown that screening tools have the potential of over-reporting the burden of disorders partly because of using unvalidated instruments. Relatedly, self-report measures are susceptible to social desirability bias. To minimize this, all the measures used have been adequately adapted and/or validated in similar settings in Kenya. Relatedly, we did not use audio-computer-assisted self-interview (ACASI) when assessing substance use problems in the study, and this may have introduced social desirability bias when participants were responding to the substance use questions. Moreover, the tool we utilised to assess drug use problems (DUDIT) does not highlight individual drug types; thus, we could not provide individual drug use prevalence estimates. We recommend a follow-up study to provide clarifications on this account. Additionally, given the low prevalence estimates of drug use, we could not offer sex-stratified correlates of drug use in this sample – future studies should utilize larger sample sizes to provide clarity on this. Given the cross-sectional nature of this study, it was not possible to make conclusions on causality or the direction of associations for the observed correlates. Minimal dataset was not available for the eligible participants who did not turn up for assessment, and this may also have introduced non-response bias.

Implications

The findings of this study have crucial implications for mental health policy, practice and further research relating to adults living in rural and urban informal settlements in Kenya and similar settings. The World Health Organization has called for countries to fill gaps in knowledge about mental health, and the present study addresses this call. Our study found high prevalence estimates of common mental disorders and significant levels of substance use problems in this sample (with distinct patterns across sex), which was partly driven by multiple overlapping and unique correlates. These findings emphasize the urgent need for preventive, promotive and curative mental health services to address the existing mental health burden – this will, however, need significant funds to support these efforts [77]. Low-intensity lay-administered psychological interventions such as Problem Management Plus intervention, which have shown promise in feasibility studies in this setting, can be scaled up to maximize their reach among the population [78,79], given the huge gap in trained mental health specialists. A task-shifting framework for the delivery of comprehensive, collaborative, and community-based mental health care in LMICs has been proposed to expand mental health services [80]. Pathways of proper and timely management of these problems are also very important in ensuring continuity of care. The Kenya National Mental Health Plan 2021–2025 provides a framework for the national and county governments to begin taking steps in timely assessment and management, including preventive and promotive mental health care, for different populations in the country. Our study also points out the significant disparities in the prevalence and correlates of mental and substance use problems in this setting, thus emphasizing the need for tailored and gender-sensitive mental health strategies to address the unique issues faced by men and women in this setting.

Kenya’s mental healthcare system remains underdeveloped in terms of numbers of skilled personnel, distribution of personnel, and functional care and treatment facilities. Urban informal settlements are particularly vulnerable and may experience disproportionate levels of mental health and substance use problems partly because of overcrowding, unemployment, increased exposure to stressful events, inadequate access to basic amenities, e.g., food and other vital services. Indeed, some of these vulnerabilities were corroborated in the current study including urban residence, food insecurity, exposure to stressful events, living in rented houses, and household debt. Our findings highlight potential points for research, intervention and policy including a) providing appropriate tangible assistance, e.g., cash and food transfers, and financial and economic empowerment programs to reduce economic stress/strain; b) expand access to basic amenities, e.g., water and sanitation and affordable housing; c) provide and expand healthcare services to include mental health and enhance access to existing services, and strengthening the referral system. This can be done by integrating mental health services to the existing primary care facilities; d) need to capacity-build grassroots organizations to provide low-intensity psychological services and also integrate these programs into their existing health services for sustainability, and e) more research, policy and intervention efforts focusing on addressing all kinds of abuses, e.g., physical and sexual and investigating bolstering support for families and survivors of these abuses in these settings.

Our study identified several factors that were significantly associated with mental and substance use problems among the participants. The majority of the factors were psychosocial and potentially amenable with appropriate psychosocial interventions [81,82], and they can also be useful targets in preventive and promotive mental health strategies. Given the cross-sectional design of our study, our findings correlates preclude any conclusions on causality. Further studies with appropriate designs that can allow causal inferences are needed to corroborate our findings.

Conclusions

Understanding the gender-stratified prevalence and correlates of mental and substance use problems is relevant in guiding researchers, policymakers and practitioners to adequately put in place gender-specific interventions. The current study highlights important gender disparities in the burden and determinants of mental and substance use problems among adults in rural and urban informal settlements in Kenya. Women presented with significantly higher levels of depressive and anxiety symptoms than their male counterparts, although the magnitude was smaller. On the other hand, men presented with significantly higher levels of both current and past-year alcohol and substance use compared to women. We also identified several correlates for these conditions, the majority of which were demographic and psychosocial factors. Some of these factors overlapped across sex and studied conditions, while others were unique. These results underscore the urgent need for tailored and gender-distinct mental health strategies in addressing these issues in the study population and also the need to address underlying societal and structural issues that drive the observed gender disparities.

Supporting information

S1 Table. Summary of the risk and protective indicators of mental health problems among women and men.

(DOCX)

S2 Table. Summary of the risk and protective indicators of substance use problems among women and men.

(DOCX)

Data Availability

The de-identified dataset used and analysed during this current study will be made available upon reasonable request, in due consideration of Aga Khan University’s data sharing policies. Requests to access the datasets should be directed to the Research Office Aga Khan University Kenya through research.supportea@aku.edu or akukenya.researchoffice@aku.edu.

Funding Statement

This publication was produced with the financial support of Global Affairs Canada (grant number P_007597). AA and PNM are also supported by the Science for Africa Foundation to [Ref: Del-22-002] with support from Wellcome Trust and the UK Foreign, Commonwealth & Development Office and is part of the EDCTP 2 programme supported by the European Union. For purposes of open access, the author has applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission. The funders had no role in the study’s design, in the collection, analyses, or interpretation of data, in the writing of the manuscript, or in the decision to publish the results.