A dataset on adolescent mental health in Kenya

Abstract

Despite the growing recognition of the mental health crisis among adolescents in low-and middle-income countries (LMICs), there remains insufficient reliable data on youth mental health challenges in Kenya and other LMICs. This dataset presents one of the most comprehensive mental health data from N=17,089 adolescents collected in 2023 across sixty-three public secondary schools in four counties in Kenya. Data were collected through self-report paper-based questionnaires and include depression and anxiety symptoms as well as twelve other psychosocial measures capturing adverse childhood experiences, digital stressors, family and emerging stressors, and help-seeking behaviours. Additionally, the dataset includes comprehensive socio-demographic data. The dataset addresses a significant gap in contextually representative data on adolescent mental health as well as possible protective and risk factors and provides useful insights in the experience of depression and anxiety among Kenyan adolescents. The dataset supports psychometric validation of widely used mental health instruments in sub-Saharan African contexts and offers valuable resources for researchers, practitioners, and policymakers working on adolescent mental health interventions and policy development.

Specifications Table

Subject	Social Sciences
Specific subject area	Epidemiology, psychosocial functioning, psychosocial and sociodemographic determinants of mental health outcomes.
Type of data	Raw and processed (.csv format)
Data collection	Self-reported paper-based questionnaires administered to adolescents enrolled in secondary schools using validated instruments: PHQ-8 (depression), GAD-7 (anxiety), SDQ (behavioural screening), CATS (trauma), BL (bullying), ACES (adverse childhood experiences), MSPSS (social support), AI (aspirations), FSS (financial stress), CSS (COVID-19 stress), CCAS (climate anxiety), Media and Eating Image Scale (MEIS) and the Body Esteem Scale for Adolescents (BESSA), SMD (social media disorder), EPV (political violence exposure), NOK (culturally adapted African scale). Multi-stage stratified sampling across sixty-three schools in four counties was used. Sampling and data collection took place between May and June 2023. Each of the survey targeted approximately 3,500 consenting students enrolled in the participating schools. Schools were selected based on their status as public institutions within the four targeted counties and their interest in the program.
Data source location	Kenya (Nairobi, Kiambu, Machakos, Makueni counties)
Data accessibility	Publicly Available. Repository Name: Open Science Foundation (OSF) Data identification number: doi:10.17605/OSF.IO/K3XTD Direct URL to data: https://osf.io/k3xtd/files/osfstorage Instructions for accessing these data: Data available on the public repository. Individuals interested in conducted research analyses should contact the corresponding author to ensure that appropriate ethical approvals and data sharing agreements are in place.
Related research article	https://trialsjournal.biomedcentral.com/articles/10.1186/s13063-023-07539-y

1. Value of the Data

• •Addresses critical data gaps: This represents one of the largest adolescent mental health datasets from Kenya (n=17,089), addressing significant gaps in nationally representative data from sub-Saharan Africa where such comprehensive assessments are rare.
• •Enables multi-dimensional analysis: The dataset includes fourteen validated scales covering traditional mental health measures, emerging stressors (digital, climate, political), and culturally adapted assessments, allowing researchers to examine complex interactions between various risk and protective factors.
• •Supports cross-cultural validation: Researchers can use this dataset to validate Western-developed mental health instruments in African contexts and compare their performance against culturally adapted measures.

2. Background

Mental health disorders among adolescents represent a critical global public health challenge, with particularly severe impacts in low- and middle-income countries where treatment resources remain scarce [1]. In Kenya, studies suggest that up to 37% of adolescents experience depression symptoms, yet comprehensive data on youth mental health challenges remain limited [2]. The establishment of a national taskforce in 2019 [3] highlighted the urgent need for large-scale surveys to inform resource allocation and intervention planning, but few subsequent comprehensive studies have been conducted.

Kenyan adolescents face unique psychosocial challenges including poverty, educational pressures, family disruption, exposure to violence, and emerging stressors such as climate change and digital technology impacts. Traditional Western-developed mental health assessments may not adequately capture culturally specific expressions of distress, necessitating validation of both international and culturally adapted instruments in sub-Saharan African contexts.

This dataset was compiled as part of a larger research initiative evaluating scalable mental health interventions, funded by the Fund for Innovation in Development. The data collection was designed to provide comprehensive insights into mental health and psychosocial characteristics of Kenyan adolescents to support evidence-based intervention design, assessment tool validation, and policy formulation in low-resource school settings.

3. Data Description

The dataset contains comprehensive mental health and sociodemographic data from 17,089 Kenyan adolescents collected through six survey versions (Survey A-F) administered across sixty-three public secondary schools in four counties. Each survey version includes core depression and anxiety measures alongside specific additional scales to reduce participant burden while maintaining comprehensive coverage.

The data files [4] contain an R script called Analysis Script.R which outlines the analysis done on the merged dataset from all the six surveys. The six surveys were merged by appending them vertically (row-binding). Prior to merging, individual surveys were checked to ensure consistency in variable names, coding and labels that were present in all six surveys (depression, anxiety and socio-demographic variables).

Codebooks and Documentations are second in the folder, containing codebooks for each of the surveys in portable document format (.pdf) and copies of the six questionnaires used in .pdf format. The last file contains the Data in comma separated value (.csv) format, including the merged dataset containing all the six survey data, as well as individual csv files for each of the surveys. Table 1 shows the distribution of participants across the six surveys:

Table 1.

Shows the distribution of participants across the six surveys.

Survey	No of Participants
Survey A	2,843 (17%)
Survey B	2,842 (17%)
Survey C	2,905 (17%)
Survey D	2,865 (17%)
Survey E	2,727 (16%)
Survey F	2,907 (17%)

The merged dataset contains sets of measures that capture multiple dimensions of adolescent mental health and wellbeing. All surveys contain the Patient Health Questionnaire (PHQ-8) and the Generalized Anxiety Disorder Screener (GAD-7).

Depression symptoms are assessed using eight items from the Patient Health Questionnaire (PHQ-1 to PHQ-8). Items ask about symptom frequency over the past two weeks (e.g. “little interest or pleasure in doing things”), rated from 0 (not at all) to 3 (nearly every day). A total score is computed by summing all items.

Anxiety is measured through seven items from the Generalized Anxiety Disorder Screener (GAD-1 to GAD-7). Items (e.g. “feeling nervous, anxious or on edge”) use the same scoring (0-3) as the PHQ-8 scale. Composite scores reflect the sum of all items.

Survey A contains an additional measure on behavioral challenges screened using ten items from the Strengths and Difficulties Questionnaire (SDQ-1 to SDQ-10). They capture conduct problems (e.g. “I get very angry”) and hyperactivity and inattention (“I am easily distracted”) subscales. The items are rated on a 3-point scale (0 = not true, 1 = somewhat true, 2 = certainly true). The SDQ_2 (I usually do as I am told), SDQ_9 (I think before I do things), and SDQ_10 (I finish the work I am doing) items were reverse coded in the data file. The items were summed to derive the total score.

Survey B contains additional measures on experiences of bullying (BL-1 to BL-10) and adverse childhood experiences (ACES-1 to ACES-10). The ten bullying items (e.g. “How often have you been physically bullied at school?”) are rated from 0 (never) to 4 (very often). Composite bullying scores were summed for a total score. The ten items from the ACES were assessing early adversity (e.g. “Did a parent or other adult in the household often swear at you, insult you, put you down, or humiliate you? or often act in a way that made you afraid you might be physically hurt?”) and response options were yes (1) or no (0) and summed to create a total ACEs score.

Survey C contains measures on Childhood trauma captured through twenty items from the Childhood Trauma Screener (CATS-1 to CATS-20). Items (e.g. “upsetting thoughts or pictures about what happened that pop into your head”) are rated from 0 (never) to 4 (very often). A total score is derived by computing the sum of all items.

Survey C contains additional measures on aspiration, with fifteen items (AI-1 to AI-15), which asked participants to rank the level of importance based on statements (e.g. “to be rich” or “to know and accept who I really am”). The options ranged from 1 (not important at all) to 7 (very important). The same survey also contained a measure on perceived social support, measured using four items from the family subscale of the Multidimensional Scale of Perceived Social Support (MSPSS-1 to MSPSS-4). Items included statements such as “My family really tries to help me” and used a 7-point scale from 1 (very strongly disagree) to 7 (very strongly agree). The items were summed up for a total score. Financial stress was also captured in Survey C, using the four-item financial stress scale (FSS-1 to FSS-4). Items (e.g. “Are you unable to do the things you need because of shortage of money?”) are rated from 1 (not at all) to 5 (very often) and averaged.

Survey D contained measures on stressors related to climate change and the COVID-19 pandemic, assessed through ten (CCAS-1 to CCAS-10) and eleven items (CSS-1 to CSS-11), respectively. Items in the CCAS scale (eg. “Thinking about climate change makes it difficult for me to concentrate”) have options ranged from 1 (never), to 5 (almost always). The CSS scale had items relating to how stressful they felt about certain things such as “Cancellation of planned or scheduled celebrations, entertainment, vacations or trips (e.g., graduations, birthdays, concerts)”. The responses used a scale from 0 (not stressful at all) to 4 (very stressful). Composite scores are calculated as item averages.

Survey E contained measures on social media use, evaluated using items from the Media and Eating Image Scale (MEIS), with items including statements such as “I compare my body/shape with other people’s body/shape on social media.”. The response options of this scale ranged from 1 (strongly disagree) to 5 (strongly agree). The survey also included the Body Esteem Scale for Adolescents (BESSA), focusing on body comparison and body esteem. Items included “I like what I look like in pictures”, and the options ranged from 0 (strongly disagree) to 4 (strongly agree). All items are averaged to obtain a total score.

Problematic social media use is measured through nine items from the Social Media Disorder Scale (SMD-1 to SMD-9) in survey E as well, with items (e.g. “Regularly neglected other activities (i.e., hobbies, sports, homework) because you wanted to use your phone/tablet/laptop?” having response options of yes or no. The score is obtained by summing the number of positive (yes) responses.

An additional domain included exposure to political violence (EPV-1 to EPV-7) in the same survey, with questions such as “How often has a friend or acquaintance of yours been injured as a result of political or military violence?”. The options ranged from 0 (never) to 2 (many times). Scores are summed across all items.

Survey F contained additional and supplementary depression and anxiety screening items, using the Ndetei-Othieno-Kathuku Scale – depression and anxiety subscales (NOK). The items (e.g. “Feeling your heart has fallen down” and “Feeling as if you are carrying a heavy load on your head.”) had option responses ranging from 0 (not at all) to 4 (extremely) based on how distressed they were in the past week. The sum of all items provided the score for this measure.

In the same survey F, help-seeking behaviors (NFOH), and help-seeking barriers (HSB) were assessed, with items such as “Within the past six months, have you at any point felt a need for outside help (someone outside your immediate family) with your problems, feelings, behavior or emotional trouble?” and “If you have not sought help, what prevented you from doing so?”. These items use a Likert scale and are summed up depending on the particular question.

The codebooks provided can be referenced for additional clarification.

3.1. Descriptive statistics

Table 1 provides sample characteristics, with the following variables:

• -Age: Refers to the age distribution of the students within the recruited schools. The mean age was 15.89 years (SD=1.41), with most students aged between 11 and 24 years.
• -Gender: Refers to the gender of the students. Female students made up majority of the sample (53.6%). Male students made up 46.4% of the sample.
• -Religion: Indicates the students’ self-reported religious affiliation. Protestants made up majority (58.6%). Traditional African and Buddhist religions had the least percentages of students (1.9% and 0.1% respectively).
• -Boarding day: Refers to the student classification by school residency as either boarding or day (students returning home daily). In this sample, day students constituted the majority (59.3%), compared to boarders (29.4%).
• -School type: Refers to the classification of the school by the Kenyan Ministry of Education, with four-tiered levels: national, county, extra-county, and sub-county. The sample had 71.2% of students enrolled in subcounty schools while county and extra-county schools had the least enrollments (16.4% and 12.4% respectively).
• -Form: Refers to the grade level of the students, with Form 1 being the first year of high school and form four being the fourth and last year of high school. The sample was made up of Form 1s and 2s (37.7 and 32.5% respectively), while Form 4 had the least percentage of students (11.0%).
• -Parents Home: Refers to how many parents are living at home with the student at the time of the data collection. Students living with both parents had the highest percentage (65.3%) in the sample.
• -Parents Dead: Indicates whether the students’ mother or father are dead or alive. Most students in the sample have both parents alive (84.9%) while students who have lost both parents had the lowest percentage (2.2%).
• -County: Refers to Nairobi (36.6%), Kiambu (23.5%), Machakos (12.2%), Makueni (11.5%)

Table 2 shows the psychometric properties of the different measures in the dataset:

Table 2.

Psychometrics (Cronbach’s alpha).

Scale	No of Items	Cronbach Alpha
PHQ	8	0.641693
GAD	7	0.739542
SDQ	10	0.582104
CATS	20	0.857611
BL	10	0.809409
ACES	10	0.681129
MSPSS	4	0.787897
AI	15	0.781887
FSS	4	0.564968
CSS	11	0.840191
CCAS	10	0.774082
SMD	9	0.777542
EPV	7	0.54802

4. Experimental Design, Materials and Methods

Study participants (n = 17,089) were recruited from 63 public secondary schools across four counties in Kenya: Nairobi, Kiambu, Machakos and Makueni. The schools included in our study were selected from a pool of schools identified across the four counties based on several criteria: demonstrated interest in participating in the Shamiri program, availability within the academic calendar, and eligibility as a public secondary school. A purposive sampling approach was applied, informed by the Kenyan Ministry of Education’s school classifications [5]. Our aim was to capture the diversity of secondary school contexts while prioritizing youths in under-resourced settings. As such, the sample included sub-county schools, which are community-run schools serving adolescents from lower-income communities, as well as county and extra-county schools, which admit students from wider geographic areas and vary in resource availability. This approach ensured representation across socio-economic contexts and school types.

Data collection was conducted in the above four counties for several practical and strategic reasons. First, among other counties, Nairobi and Kiambu have some of the highest concentrations of public secondary schools in the country [6], making them efficient settings for school-based research with adolescents. Second, their proximity to the team’s headquarters made it feasible to coordinate and supervise data collection centrally. Additionally, our longstanding partnerships with implementing organizations operating in Machakos and Makueni provided established infrastructure, trusted school relationships, and logistical support, which helped with smooth entry, coordination, and data collection across all four counties.

Additionally, participants were not financially compensated, in line with the guidance from both the school administration and the ethics review committee. Because study activities were embedded within regular school programming – typically as after-class activities – students were not required to commit additional time or resources. However, we offered small tokens of appreciation such as pens, keychains, and wristbands to the students after data collection, as recommended by the school administrators.

The sampling and data collection were done between May and June 2023. The recruitment and selection capture the diversity of the Kenyan educational system, as classified by the Ministry of Education [5], ensuring that the sample reflects the varied educational environments. All procedures were approved by the Ministry of Education and a licensed Institutional Review Board (Kenyatta University Ethical Review Committee, approval number: PKU/2627/E1752). We also obtained a research permit in accordance with national guidelines on conducting research in Kenya (National Commission for Science, Technology, and Innovation, permit number: NACOSTI/P/23/23559).

Data collection occurred as an after-school activity in small groups of 10-15 students supervised by a trained group leader. Paper-based questionnaires with standardized measures were used. The group leader was available to support in case participants have clarifications regarding any item on the questionnaire. Each participant completed the questionnaire individually and handed it to the group leader. The study team then collected all complete questionnaires and processed them.

The measures used in this study are described below and specific questions for each measure can be found in the Supplementary Materials. The survey employed a six-survey design where participants completed one of six questionnaire batteries. Each survey targeted approximately 3,500 students who provided consent and were enrolled in the recruited schools. Schools were selected based on their status as public institutions within the four counties, and their expressed interest in the program. We estimated that each questionnaire took students approximately 20-25 minutes to complete. Each student completed the questionnaire once.

Core measures were depression and anxiety symptoms which were administered to all students:

• -Depression was measured by the Patient Health Questionnaire – 8 (PHQ-8), an 8-item validated tool used to assess the severity of depressive symptoms. It omits the item on suicidality from the PHQ-9 and has been widely validated for use among adolescent populations [7].
• -Anxiety was measured by The Generalized Anxiety Disorder Screener –7 (GAD-7) is a 7-item self-report scale used to screen for symptoms of generalized anxiety disorder. It has been validated for use among adolescents in various contexts, including Kenya and demonstrated strong internal consistency and convergent validity [8].

Additional measures were distributed across version based on thematic relevance described below:

Survey A: Trauma and behavioral problems. This survey contained measures assessing behavioral problems and exposure to trauma. The following measures were used:

• a.The Strengths and Difficulties Questionnaire (SDQ) - a behavioral screening tool for children and adolescents. This study used the 5-item conduct and 5-item hyperactivity subscales. The scale has been validated across diverse settings with adequate internal consistency [9].
• b.The Child and Adolescent Trauma Screen (CATS) - a validated instrument used to assess exposure to potentially traumatic events and post-traumatic stress disorder among children and adolescents. This study included items on trauma exposure and excluded impairment questions [10].

Survey B: Bullying and adverse childhood experiences. This survey assessed the level and experiences of bullying as well as the exposure to childhood adversity. The instruments used are as follows:

• a.A Bullying Questionnaire (BL) developed by researchers using items adapted from previously validated instruments assessing peer victimization and bullying behaviors. It captures both victim and perpetrator experiences across vernal, physical, and relational domains [11].
• b.The Adverse Childhood Experiences Scale (ACES) - a 10-item widely used tool for measuring exposure to various forms of childhood adversity, including abuse, neglect and household dysfunction [12].

Survey C: Social support and financial strain. This survey captures the levels of perceived social support and the relationship with financial stress and aspirations. The survey included the following measures:

• a.The Multidimensional Scale of Perceived Social Support (MSPSS) – a validated self-reporting tool for assessing perceived social support from friends, family and significant others using the family subscale. The tool has been approved for use among young individuals, such as adolescents and young adults [13].
• b.Aspirations Index (AI) - A 15-item measurement scale used to assess participants’ life aspirations based on the perceived relevance of their future goals. The scale has been approved for use in diverse environments and has displayed internal consistency [14].
• c.The Financial Stress Scale (FSS) – A 4-item measurement scale used to access financial stress levels among participants. The scale has shown internal consistency and has been approved for use in various contexts[15].

Survey D: Covid-19 and climate change stressors. This survey contained measures that assess the levels of stress caused by the COVID-19 pandemic and climate change, as emerging stressors. The tools used are:

• a.The COVID-19 Stressor Scale (CSS) - A 36-item tool used to assess stressors associated with the COVID-19 pandemic among individuals [16].
• b.The Climate Change Anxiety Scale (CCAS) - A 22-item tool used to measure anxiety and emotional distress related to climate change. It is used to understand how environmental issues affect mental health [17].

Survey E: Social media and political stress. This survey looked at the usage of social media and the stress caused by exposure to political violence and unrest, either directly or indirectly. The measures used are described below:

• a.The Social Media Use Scale – A 17-item scale used to measure the frequency, intensity, and patterns of social media usage in relation to its impact on mental health and wellbeing (14). The Media and Eating Image Scale (MEIS) [18]and the Body Esteem Scale for Adolescents (BESSA)[19], focusing on body comparison and body esteem were also used.
• b.The Political and Social Stress Scale (PSSS) - A scale used to measure the perceived stress of an individual in response to political and social events or conditions[20]. It is especially relevant during times of civil unrest or policy change.

Survey F: Validating the Ndetei-Othieno-Kathuku scale for depression and anxiety. This survey focused on using the Ndetei-Othieno-Kathuku scale for assessing depression and anxiety, as well as assessing help-seeking behaviors and barriers among participants. The tools captured in this survey were:

• a.The Ndetei-Othieno-Kathuku Scale (NOK) – A 15 item subscale to measure depression and anxiety that is culturally adapted for African and Kenyan individuals in the context of culturally specific idioms of distress [21].
• b.A service use questionnaire developed by the study team was administered to assess help-seeking and barriers. It included three core questions on the need to seek help, source sof help, and factors that prevented help-seeking.

In addition to the outcomes, all participants (n=17,089) in all the surveys completed a self-report socio-demographic questionnaire, which contained basic information such as age and gender, and socio-demographic information (e.g., level of parent education, religion, parents alive). For the school name and county variables, further de-identification was done to protect the identities of underage participants and maintain a negligible likelihood of re-identification. School names were masked by replacing them with alphabetical codes (A–AJ). County was replaced with random four-letter alphanumeric codes. A master key linking these codes to the original names is kept in a secure, separate file. It can be available to qualified researchers for the purpose of replication or meta-analysis upon reasonable request to the corresponding author, subject to institutional ethical approval.

Physical questionnaires were processed using Papersurvey.io software with automated scanning and data entry, followed by systematic verification against original responses. Data cleaning included range checking, consistency verification, and missing data pattern analysis.

Limitations

Several limitations should be considered when using this dataset. The sampling was restricted to public secondary school students, excluding private school attendees and out-of-school youth who may experience different mental health patterns. Geographic coverage was limited to four of Kenya's 47 counties, potentially limiting national generalizability despite diverse county selection.

All measures were self-reported, introducing potential recall bias and social desirability effects. Some contextual measures showed lower reliability coefficients, reflecting cultural adaptation challenges. Cross-sectional design prevents causal inference and assessment of mental health trajectories over time.

School-based sampling may underrepresent adolescents with the most severe mental health challenges who may have dropped out of education. Rural-urban differences within counties were not systematically examined, and seasonal variations in mental health symptoms could not be assessed due to single time-point collection.

Ethics statement

This study was approved by the Kenyatta University Ethics Review Committee (application number PKU/2627/E1752). The research was conducted in accordance with the Declaration of Helsinki. Informed consent and assent were obtained from all participants and their guardians through school administration prior to data collection. As approved by the ethics committee, the consent process followed the protocols typically used in Kenyan boarding and day secondary schools. For boarding schools where most students reside on campus for the entire term, directly obtaining signed parental consent is often not feasible because parents are only accessible during school breaks. In these cases, the ethics committee authorized the study team to obtain consent from the school administration through the established mechanisms for communicating with parents. The administration notified parents and guardians about the study and provided them with the opportunity to opt for their child to not be included in the study. For day schools where students return home after classes, the study team requested that students obtain documented parental or guardian consent prior to participation.

Across all schools, in-person information sessions with interested participants were conducted, explaining the purpose of the study, confidentiality, voluntary participation, and the right to withdraw at any time without penalty. All students were required to provide written informed consent, and underage participants provided written assent before completing any questionnaires or participating in study activities.

This approach ensured that ethical considerations for working with underage participants were met and that both guardians and students were adequately informed.

Participation was voluntary, and participants could withdraw at any time. Confidentiality was maintained throughout the study process, with all data de-identified for analysis and reporting.

CRediT Author Statement

Rosine Baseke: Writing – original draft, Writing – review & editing. Rachael Kilonzo: Writing – original draft, Writing – review & editing. Maureen Ngesa: Writing – review & editing. Purity Mwende: Writing – original draft. Tom Osborn: Conceptualization, Investigation, Writing – review & editing.

Data Availability

• Open Science FoundationA Dataset on Adolescent Mental Health in Kenya (Original data)