Suicide behaviour among adolescents in a high HIV prevalence region of western Kenya: A mixed methods study

Abstract

The study purpose was to determine the prevalence and determinants of suicidal thoughts and behaviours in a rural community sample of 15–19-year-old Kenyan adolescents in a region with high HIV burden. Data were from an observational study examining ethical issues in adolescent HIV research (N=4,084). Participants reporting suicidal ideation were assessed for suicide risk. Directed content analyses were conducted using assessment reports. Logistic regression was used to identify factors associated with suicide outcomes. Prevalence of suicidal ideation was 16%. Of these, 38% were low risk and 12% were moderate/high-risk. Females and sexually active adolescents had higher odds of suicidal ideation and being categorized as moderate/high-risk. Adolescents with higher depression scores had higher odds of reporting ideation. Pregnancy was protective for females while impregnating a partner was a risk factor for males. Abuse from a family member, financial stress and health concerns were the most frequently mentioned precipitants of ideation. However, only abuse increased odds of suicide behaviour. Effective programs to identify and support sexually active, pregnant, and distressed adolescents at risk for suicide are needed. Approaches involving families, schools, health facilities, and community gatekeepers may have the most promise in sub-Saharan African rural areas with limited mental health services.

Introduction

Suicide is the third leading cause of death among adolescents aged 15–19, globally (World Health Organization [WHO], 2019). Suicidal thoughts and behaviours (STBs) are on a broad spectrum of self-destructive behaviour ranging from experience of fleeting death wishes; suicidal thoughts, feelings and urges; making specific plans; and acting on one’s suicide plans. Suicide ideation forms the lower end of the spectrum and is comprised of death wishes and suicidal feelings, thoughts and urges (Posner et al., 2014). Prior STBs such as suicide attempt are predictive of completed suicide (Franklin et al., 2017). In this paper, we investigate STBs and identify important precipitants of suicidal behaviour among a key at-risk population group: adolescents residing in a region of Kenya with high HIV burden. This examination is an important step to developing an effective national suicide prevention strategy.

Although not disaggregated by age, available data highlight the importance of STBs and death by suicide as a public health concern in Kenya. For example, a population-based study conducted among adults in Kisumu County, located in the country’s high HIV prevalence Nyanza region, found that at one point in their lifetime, 24.1% of participants had thought that life was not worth living, 19.2% had wished they were dead, 7.9% had suicidal thoughts, and 1.9% had made a suicidal attempt (Jenkins et al., 2015). Moreover, according to national estimates, 421 individuals died by suicide in 2017 in Kenya; a 39.4% increase from 2016 (Kenya National Bureau of Statistics [KNBS], 2018). Additionally, WHO data suggest that the age-standardized suicide rate in Kenya was 5.6 per 100,000 population in 2016 (WHO, 2019). Similarly, a recent population-based epidemiological study in a rural coastal region of Kenya found an annual incidence rate of suicide of 4.61 per 100,000 person years, which accounted for 0.78% of all deaths in the study area between 2008 and 2016 (Bitta et al., 2018). Although these data suggest that suicide rates are lower in Kenya than the global average of 10.5 per 100,000 population, a key concern about them is their reliability and accuracy.

A key barrier to studying suicide and related thoughts and behaviours in sub-Saharan African (SSA) countries, like Kenya, that were former British colonies is their legacy anti-suicide laws. For example, Chapter 63, Section 226 of the Kenyan Penal Code stipulates that ‘[a]ny person who attempts to kill himself is guilty of a misdemeanour’ (Penal Code, 2009). Additionally, Kenyan media have reported criminal prosecution of suicide attempt survivors (Adinkrah, 2016). The criminalization of suicide attempt in Kenya has a deterrent effect on reporting of suicide and related behaviours. Other factors that may contribute to the lack of reliable data include underreporting of suicide due to culturally associated stigma, lack of suicide surveillance systems, and poor quality vital registration across the country, for example, poor reporting on causes of death and frequent misclassification of deaths as accidents in reported cases (Bitta, 2018; Shiundu, 2019). Thus, death by suicide may be higher than indicated by available data.

HIV-related factors have been shown to increase risk for poor mental health and STBs among adolescents (Betancourt et al., 2014; Casale et al., 2019; Louw et al., 2016; Menon et al., 2007; Musisi & Kinyanda, 2009). For example, a cross-sectional study among children and adolescents living with HIV in Kenya found that 79% had a psychiatric disorder, including major depression (17.8%) and suicidality (10.3%). In this study, suicidality increased with age from a low of 4.6% among children aged 6–10 to a high of 25% among adolescents aged 11–18; suicidality was significantly associated with older age (Kamau et al., 2012). Additionally, in a longitudinal study among South African adolescents aged 10–18, Cluver et al. (2015) found that adverse childhood experiences (ACEs), including orphanhood by AIDS and parental AIDS illness, significantly predicted STBs. Other risk factors for STBs among adolescents include female gender, familial suicide and other family factors, emotional abuse, sexual abuse, loneliness, depression, lower quality of life, illicit drug use, low socio-economic status, and having more lifetime sexual partners (Cluver et al., 2015; Jewkes et al., 2010; Jina et al., 2012; Makame et al., 2002; Musyimi et al., 2017; Ng et al., 2015; Wilson & Dunlavy, 2012). However, except for Cluver et al. (2015), most of these studies used convenience samples from health facilities or urban schools. Additionally, none were conducted specifically with rural community-based adolescent samples nor examined contextual information pertaining to determinants of STBs.

The aims for the present study were to determine the prevalence of STBs and identify associated demographic, psychosocial, and socio-environmental factors among community-based Kenyan adolescents. The study was conducted in a rural county in the Nyanza region of western Kenya. In 2018, HIV prevalence among adolescents in this region was estimated to be 7.0% and 4.1% among young women and men aged 15–24, respectively; the national average in that age group was 2.5% among young women and 1.5% among young men (Joint United Nations Programme on HIV/AIDS [UNAIDS], 2019). We used a mixed methods approach incorporating survey data, quantified qualitative data, and qualitative interview data.

Methods

Sample

Data were from the baseline of a large observational cohort study examining ethical issues in adolescent HIV research (Luseno et al., 2020). As part of study procedures, research staff conducted suicide risk assessments (SRAs) to ensure participant safety. For this paper, we examine SRA data in conjunction with survey data.

Trained research staff screened and enrolled eligible participants during 2016–2017. Three sub-counties comprised strata, and villages or wards within sub-counties were randomly selected to participate. Eligible participants were between 15–19 years old and had never tested HIV-positive nor been tested in the previous 6 months. These inclusion criteria were for the observational study, which had the aim of examining the effects of HIV testing and results disclosure on adolescent behaviour and well-being in the context of a research study. Research staff carried out community awareness activities about the study and recruitment and information meetings with potential adolescent participants and their parent/guardians in addition to going door-to-door. The recruitment and screening activities occurred over an 18-month period. Research staff screened potential participants for study eligibility by asking a brief set of questions using computer-assisted techniques. Additional information about the study design and procedures is available elsewhere (Luseno et al., 2020).

The study design was such that half of the participants completed the baseline activities (survey and HIV testing) in a local clinic and the other half in their home. Although the cohort study followed a sub-sample of participants at additional time-points (those testing HIV-positive or who had indeterminate results and a similar number of those who had HIV-negative results), the purpose of present analyses was to examine STBs among the large baseline sample and thus only baseline data were used for this paper. The baseline data include responses collected during individual interviews to assess suicide risk conducted with participants who had answered the survey questions in a way that indicated potential risk (details provided in the section below). Of 6,726 individuals screened, 4,799 met eligibility criteria of whom 85% completed baseline procedures (N=4,096). All participants received a t-shirt; those who underwent the research procedures in a clinic also received 300KSH (~3 USD) as reimbursement for travel costs.

Surveys were administered individually using audio computer-assisted self-interview (ACASI) in Luo, English, or Swahili. They included questions about depression, quality of life, sexual behaviour, and socio-demographics. Following the survey, HIV counselling and testing was conducted. Counsellors referred participants with positive or indeterminate results to services as appropriate.

Suicide Risk Assessment (SRA)

SRAs were conducted with participants based on their responses to items about suicidal ideation which were asked in the survey. These items are from the Centre for Epidemiologic Studies Depression Scale Revised (CESD-R), which measures nine major depression symptoms (Eaton et al., 2004). Suicidal ideation, one of the depression symptoms, is assessed by two questions about frequency of suicidal thoughts and/or thoughts of self-harm: ‘During the past week or so, I wished I were dead’ and ‘During the past week or so, I wanted to hurt myself.’ Response options ranged from ‘not at all or less than 1-day last week’ to ‘nearly every day for two weeks.’

To protect participant safety, the research team developed SRA procedures that would facilitate field staff to assess the adolescents’ suicide risk, provide a compassionate way for the staff to respond to the adolescent, respect the adolescent’s privacy, and be appropriate and feasible in the local context. The procedures were reviewed and approved by the institutional review boards (IRBs) for the study. In the SRA procedures, a research staff member conducted an in-person interview with a participant if the adolescent responded to either of the suicidal ideation questions from the CESD-R with an affirmative answer of ‘one- or two-days last week’ or more frequently. The SRA interview was conducted on the same day as the survey and after the HIV testing had been conducted and test result provided. The research staff person asked assessment questions by following a prepared script. Questions covered topics about suicide attempts and plans, family suicide history, access to suicide methods, a strong wish to die, substance use, risk-taking, and social connections.

Research staff were trained to ask the adolescents for permission to engage a parent or other supportive adult in the conversation when appropriate and to provide referrals for community services when applicable and realistically available. Access to mental health services is very limited in Kenya, particularly in rural areas and impoverished communities (Bitta et al., 2017). Due to this, in-country investigators (including a clinical psychologist and a physician) were consulted regarding what service referrals field staff could practicably provide and recommendations about cases. For adolescents who were assessed as being high-risk, the staff conducted a two-week follow-up contact with the adolescent and/or supportive adult. A research staff person who conducted the SRA interview prepared a written report for each assessment interview summarizing the participant’s responses to questions, the field’s staff’s observations and assessment, any response the staff provided, and, if applicable, any plans for follow-up.

Ethical Approval

Study protocols were approved by the ethics review boards of the Pacific Institute for Research and Evaluation and the Kenya Medical Research Institute. Adolescents all provided informed verbal consent themselves to participate in the eligibility screening. After determining study eligibility, adolescents 18 years and older and emancipated minors provided written informed consent prior to participating in baseline research activities; for 15–17-year-old adolescents, we obtained written consent from their parent/guardian and child assent. Employing the in-country institution’s translation compliance and certification procedures, English versions of consent forms and questionnaires were translated into Luo and Swahili by one staff member and were then reviewed for accuracy, consistency and clarity with original documents by a different staff person. A summary of all SRA interview reports was provided regularly to the Principal Investigator’s IRB.

Measures

Dependent Variables

Suicidal ideation was coded as 1 if a participant responded ‘one- to two-days’ or more frequently to either of the two CESD-R suicidal ideation items and 0 otherwise. Project staff coded each SRA participant’s suicide risk level as High (thought of suicide in last two weeks and has another risk factor such as suicide plan or previous attempt), Moderate (thought of suicide in the last 2 weeks but does not have another risk factor or has not thought of suicide in the last two weeks but interviewer thinks that the risk is more than minimal), Low (has not thought of suicide in the last two weeks and interviewer’s impression is that risk is minimal or has thought of suicide but it was over a year ago), or Wrong Button (reported having pressed the wrong button, not understanding the question, or not intending the answer given). Moderate and high-risk categories were combined due to low frequencies. For suicide behaviour, research staff used the SRA reports and coded participants as having had suicidal ideation if they reported having suicidal thoughts; as having had a suicide plan if they mentioned specific plans to die by suicide; and as having attempted suicide if they acted on their plan. Due to low frequencies, we combined the coded behaviours as one outcome, reflecting whether participants fit criteria for at least one.

Independent Variables

Sociodemographic variables included age, gender, orphan status, school enrolment, education level, sub-county, and assignment to clinic versus home. A sum score of CESD-R items was used, excluding the two on suicidal ideation and self-harm (Eaton et al., 2004). Internal consistency for this measure was α =0.91. Quality of Life (QOL) was measured with the World Health Organization Quality of Life Questionnaire (WHOQOL-BREF) (Skevington et al., 2004). Exploratory factor analysis identified two domains: a 10-item QOL social and physical health (sp-QOL; α = 0.84) measure and a 9-item QOL psychological and environmental (pe-QOL; α =0.76) measure. We used mean scores for each domain. The HIV test results were coded as negative, positive, or indeterminate.

Quantitative Analysis

Quantitative analyses were carried out using Stata version 13 (StataCorp, 2013). Descriptive statistics were calculated and stratified by gender for the full sample and sub-sample of participants who had triggered the SRA (the ‘SRA sub-sample’). For bivariate analyses, we conducted chi-square tests and analysis of variance to examine differences in demographic, sexual behaviour, and well-being factors between the full-sample and SRA sub-sample.

Logistic regression was used for three sets of multivariable models. Model A identified risk factors from the survey for suicidal ideation among the full sample. Model B identified risk factors from the survey associated with having a moderate or high-risk level among those in the SRA sub-sample. Model C used data extracted from the SRA reports (details on method of extraction provided below) to examine risk factors for suicide behaviour for at-risk participants in the SRA. Models B and C excluded those in the SRA who said that they had pressed the wrong button. All analyses were conducted for the full sample and stratified by gender.

Qualitative Analysis

Content analysis methods were used to extract and quantify suicide behaviour precipitants from the SRA reports (Miles & Huberman, 1994). SRA reports were coded in ATLAS.ti using a directed content analysis technique, whereby the first author developed a codebook based on literature-derived risk factors (Friese, 2013; Hsieh & Shannon, 2005). Suicidal ideation was defined as a specific reference to wanting to kill oneself (e.g., ‘About two weeks ago, I got so upset when my brother said that my son and I were a burden to [him]. I got so upset that I thought of killing myself.’) Suicide plan was defined as a reference to a specific plan and method (e.g., ‘…I decided to commit suicide using Rat and Rat [brand of rat poison]. The Rat and Rat was in the house and I wanted to take it.’) Suicide attempt was defined as a reference to a specific event when the participant attempted to end their own life (e.g., ‘I attempted to commit suicide by taking an overdose of malaria medication.’) These were especially important in determining suicide risk, because, while expression of current suicidality may be absent in an interview, a history of suicidal behaviour is the ‘most reliably replicated risk factor for future suicide attempt or completion’ (Oquendo et al., 2008, p. 1383). The first author met with the co-authors to finalize the codebook and then coded each SRA report. She also immersed herself in reading and rereading the SRA reports and writing in-depth memos in order to identify emerging themes. She then compared these emerging themes to the codes based on the literature-derived risk factors and identified illustrative quotes to present in the findings (Patton et al., 2002).

Results

Quantitative Analysis

The analytical sample included 4,084 adolescents who participated in baseline activities and were not missing data on both CESD-R suicidal ideation items. About half of the participants in the analytical sample were female, 62% were ages 15–17, and 52% had completed primary school (Table 1). Compared to the full sample, the SRA sub-sample had a higher proportion of participants who were female, had ever had sex, ever been pregnant or impregnated someone, had a lower mean QOL, and had a higher CESD-R score. Seventeen participants (0.42%) had a positive HIV test result at baseline (four male and 13 female) and 26 (0.64%) had indeterminate results (14 male and 12 female). Five of the 17 HIV-positive participants and nine of the 26 participants with indeterminate results participated in the SRA.

Table 1:

Characteristics of Participants for the Full Sample and SRA Sub-Sample

Variable	Mean (std. dev)/Percentage (n)
	Full Sample (N=4,084)	Male (N=2,063)	Female (N=2,021)	SRA Sub-Sample (N=652)	Male (N=269)	Female (N=383)
Demographics
Gender
Male	50.51% (2,063)	N/A	N/A	41.26% (269)	N/A	N/A
Female	49.49% (2,021)	N/A	N/A	58.74% (383)	N/A	N/A
Age
15–17	61.63% (2,517)	56.62% (1,168)	66.75% (1,349)	61.81% (403)	53.90% (145)	67.36% (258)
18–19	38.37% (1,567)	43.38% (895)	33.25% (672)	38.19% (249)	46.10% (124)	32.64% (125)
Testing Venue
Clinic	50.61% (2,072)	49.73% (1,027)	51.50% (1,045)	59.97% (391)	55.76% (150)	62.92% (241)
Home	49.39% (2,022)	50.27% (1,038)	48.50% (984)	40.03% (261)	44.24% (119)	37.08% (142)
Sub-county
A	26.89% (1,098)	28.02% (578)	25.73% (520)	22.70% (148)	20.82% (56	24.02% (92)
B	35.65% (1,456)	35.72% (737)	35.58% (719)	33.28% (217)	37.92% (102)	30.03% (115)
C	37.46% (1,530)	36.26% (748)	38.69% (782)	44.02% (287)	41.26% (111)	45.95% (176)
Orphan	41.36% (1,689)	42.37% (874)	40.33% (815)	42.18% (275)	43.12% (116)	41.51% (159)
Missing	0.83% (34)	0.78% (16)	0.89% (18)	0.61% (4)	0.37% (1)	0.78% (3)
Currently Enrolled in School	78.92% (3,233)	78.33% (1,616)	79.52% (1,607)	77.15% (503)	74.72% (201)	78.85% (302)
Highest Level of School Completed
Did not finish primary	42.02% (1,716)	43.04% (888)	40.97% (828)	45.01% (300)	48.33% (130)	44.39% (170)
Finished Primary	52.03% (2,125)	51.14% (1,055)	52.94% (1,070)	48.31% (315)	46.10% (124)	49.87% (191)
Finished Secondary	5.90% (241)	5.82% (120)	5.99% (121)	5.67% (37)	5.58% (15)	5.74% (22)
Missing	0.05% (2)	0% (0)	0.10% (2)	0% (0)	0% (0)	0% (0)
Sexual behaviour
Ever Had Sex	41.94% (1,713)	45.27% (934)	38.55% (779)	51.84% (338)	54.28% (146)	50.13% (192)
Missing	3.70% (151)	2.91% (60)	4.50% (91)	1.07% (7)	0.74% (2)	1.31% (5)
You or Partner Ever Pregnant	9.40% (384)	4.12% (85)	14.79% (299)	11.81% (77)	8.18% (22)	14.36% (55)
Missing	0.83% (34)	0.82% (17)	0.84% (17)	0.61% (4)	0.37% (1)	0.78% (3)
HIV Status at Baseline
Positive	0.42% (17)	0.19% (4)	0.64% (13)	0.77% (5)	1.11% (3)	0.52% (2)
Indeterminate	0.64% (26)	0.68% (14)	0.59% (12)	1.38% (9)	1.86% (5)	1.04% (4)
Well-being
CESD-R Suicidal Ideation
Never or Less than One day	76.88% (3,149)	80.10% (1,654)	73.61% (1,495)	21.13% (138)	25.28% (68)	18.23% (70)
One day plus	12.52% (513)	9.54% (197)	15.56% (316)	77.18% (504)	72.49% (195)	80.47% (309)
Missing	10.60% (434)	10.36% (214)	10.83% (220)	1.68% (11)	2.23% (6)	1.30% (5)
CESD-R Self-Harm
Never or Less than One day	79.54% (3,258)	82.28% (1,699)	76.76% (1,559)	32.62% (213)	30.11% (81)	34.38% (132)
One day plus	10.69% (438)	9.01% (186)	12.41% (252)	65.54% (428)	68.77% (185)	63.28% (243)
Missing	9.77% (400)	8.72% (180)	10.83% (220)	1.84% (12)	1.12% (3)	2.34% (9)
Quality of Life- Social and Physical Health (Range 1–5)	3.79 (0.76)	3.79 (0.78)	3.81 (0.74)	3.49 (0.72)	3.47 (0.76)	3.50 (0.83)
Quality of Life Psychological and Environmental (Range 1–5)	3.09 (0.70)	3.10 (0.70)	3.07 (0.69)	2.94 (0.68)	2.96 (0.70)	2.94 (0.66)
Composite CESD-R Score (without two suicide items)	13.87 (11.79)	13.41 (11.54)	14.42 (12.02)	24.60 (11.98)	24.65 (11.98)	24.51 (11.99)

About 16% (n=652) of the sample reported suicidal ideation. Of these, 50% (n=327) were classified as ‘wrong button,’ 38% (n=247) as low risk, and 12% (n=78) as moderate/high-risk. Content analysis of the SRA reports indicated that about 59% of the SRA sub-sample with low or moderate/high-risk (n=191) reported suicide behaviour (Table 2; not statistically different by gender). Of the five participants in the SRA sub-sample who tested positive for HIV, four were deemed low risk for STBs based on their SRA reports, while one claimed to have pressed the wrong button (data not shown). Of the nine participants in the SRA sub-sample who had an indeterminate HIV test result, three were deemed low risk and two were deemed moderate risk for STBs based on their SRA reports; four said they had pressed the wrong button.

Table 2:

SRA Content Analysis and Comparison by Gender

Variable	Percentage			N=325
	SRA sub-Sample Low or Moderate/High	Male	Female	Chi Square Test Statistic and P-Value
SRA Risk-Level
Low Risk	76.00% (247)	82.58% (109)	71.50% (138)	5.27 (p=0.02)
Moderate/High Risk	24.00% (78)	17.42% (23)	28.50% (55)
Attempt, Plan, or Suicide Ideation	59.87% (191)	55.38% (72)	62.96% (119)	1.84 (p=0.18)
Abuse (Emotional, Physical, and/or Sexual)	28.21 % (90)	23.08% (30)	31.75% (60)	2.86 (p=0.09)
Financial Stress	32.60% (104)	33.08% (43)	32.28% (61)	0.02 (p=0.88)
Health Concern	11.91% (38)	13.08% (17)	11.11% (21)	0.28 (p=0.59)

Models A and B

From multivariable analysis, suicidal ideation was associated with gender, secondary school completion, ever been pregnant/impregnated someone, ever had sex, sp-QOL, and CESD-R score (Table 3, Panel 1, Model A). Stratifying the analysis revealed differences by gender. Among females, ever having sex was a risk factor for ideation, while ever pregnant and sp-QOL were protective. Analysis among only those who had ever had sex (Supplemental Table 1; n=1,373) indicated that having impregnated someone was a risk factor for suicidal ideation among males (AOR 2.85; 95% CI 1.22, 6.70), while having been pregnant was protective among females (AOR 0.55; 95% CI 0.31, 0.98).

Table 3:

Models A and B- Adjusted Odds Ratios for Suicidal Ideation in the Full Sample and Risk Level in the SRA Sub-Sample

	Model A- Suicidal Ideation			Model B- Risk-Level (Ref. Low)
Variable	Full Sample (n=3461)	Male Only (n=1,765)	Female Only (n=1,696)	SRA Sub-Sample with Any Risk (n=319)c	Male Only (n=130)	Female Only (n=189)
Gender (Ref: Male)	1.69 (1.37, 2.07)b	NA	NA	2.23 (1.23, 4.03)a	NA	NA
Age (Ref: 18–19)	1.16 (0.90, 1.48)	1.15 (0.81, 1.63)	1.15 (0.81, 1.63)	0.72 (0.38, 1.37)	0.64 (0.20, 2.13)	0.78 (0.34, 1.78)
Sub-County (Ref: A)
B	0.96 (0.74, 1.25)	1.27 (0.85, 1.88)	0.75 (0.53, 1.08)	0.67 (0.32, 1.38)	0.62 (0.15, 2.60)	0.61 (0.25, 1.52)
C	1.37 (1.06, 1.76)a	1.47 (1.0, 2.17)	1.26 (0.90, 1.77)	0.97 (0.51, 1.82)	1.57 (0.47, 5.30)	0.72 (0.33, 1.56)
Orphan (Ref: Non-Orphan)	0.91 (0.75, 1.11)	0.98 (0.72, 1.31)	0.95 (0.65, 1.11)	1.27 (0.73, 2.19)	2.04 (0.69, 6.02)	1.14 (0.59, 2.23)
Highest level of School Completed (reference: did not finish primary)
Finished Primary	1.14 (0.91, 1.43)	1.19 (0.85, 1.68)	1.09 (0.80, 1.48)	0.85 (0.43, 1.68)	1.10 (0.29, 4.16)	0.70 (0.30, 1.60)
Finished Secondary	1.67 (1.04, 2.70)a	1.79 (0.89, 3.60)	1.45 (0.75, 2.81)	0.37 (0.11, 1.27)	0.31 (0.03, 3.83)	0.38 (0.09, 1.61)
Ever Been Pregnant or Impregnated Partner (Ref: Never)	0.70 (0.50, 0.98)a	1.60 (0.88, 2.91)	0.49 (0.32, 0.75)b	0.81 (0.36, 1.82)	0.95 (0.20, 4.45)	0.91 (0.35, 2.39)
Ever Had Sex (Ref: Never)	1.32 (1.07, 1.65)a	1.11 (0.81, 1.53)	1.95 (1.22, 2.24)b	2.0 (1.05, 3.79)a	4.12 (1.03, 16.42)a	1.65 (0.75, 3.63)
Quality of Life- Social and Physical Health	0.82 (0.71, 0.96)a	0.84 (0.68, 1.05)	0.80 (0.64, 0.98)a	0.78 (0.53, 1.16)	0.75 (0.35, 1.61)	0.80 (0.49, 1.31)
Quality of Life Psychological and Environmental	0.89 (0.75, 1.04)	0.83 (0.64, 1.07)	0.96 (0.76, 1.21)	0.84 (0.52, 1.35)	0.60 (0.23, 1.53)	0.91 (0.51, 1.61)
CESD-R Score	1.10 (1.09, 1.11)b	1.09 (1.08, 1.11)b	1.10 (1.09, 1.11)b	0.99 (0.97, 1.02)	1.02 (0.97, 1.07)	0.98 (0.95, 1.01)
Cons	0.09 (0.05, 0.19)b	0.09 (0.03, 0.24)b	0.16 (0.06, 0.42)b	0.97 (0.13, 7.00)	0.35 (0.01, 10.52)	3.30 (0.25, 44.12)

^aStatistically significant at the 5% level.

^bStatistically significant at the 1% level.

^cFor these analyses, 6 cases were dropped from the SRA sub-sample due to missing data.

Table 3, Panel 2 (Model B) shows that gender and ever having sex were associated with a moderate/high suicide risk level. In analyses stratified by gender, ever having sex was only significant among males.

Model C

Experiencing abuse was associated with higher odds of suicide behaviour (Table 4). Stratifying analyses by gender, suicide behaviour was associated with endorsement of the CESD-R suicidal ideation item among only females.

Table 4:

Model C- Adjusted Odds Ratios of STBs among Individuals with Suicide Risk in SRA Sub-Sample

Variable
	SRA sub-Sample with Risk (N=304)c	Male-Only (N=125)	Female-Only (N=179)
Gender (Ref: Male)	1.17 (0.68, 2.01)	NA	NA
Age (Ref: 18–19)	0.81 (0.44, 1.48)	0.63 (0.26, 1.55)	1.10 (0.45, 2.71)
Sub-County (Ref: A)
B	0.26 (0.13, 0.52)b	0.33 (0.11, 0.98)a	0.20 (0.07, 0.59)a
C	0.42 (0.21, 0.81)a	0.78 (0.28, 2.13)	0.23 (0.08, 0.62)a
Orphan (Ref: Non-Orphan)	0.82 (0.48, 1.40)	0.58 (0.25, 1.34)	1.05 (0.50, 2.24)
Highest level of School Completed (reference: did not finish primary)
Finished Primary	1.06 (0.56, 2.02)	1.07 (0.39, 2.93)	1.08 (0.44, 2.69)
Finished Secondary	0.53 (0.18, 1.54)	0.27 (0.25, 1.34)	0.95 (0.22, 4.05)
Ever Been Pregnant or Impregnated Partner (Ref: Never)	1.39 (0.63, 3.07)	1.17 (0.30, 4.58)	1.85 (0.65, 5.26)
Ever Had Sex (Ref: Never)	0.84 (0.46, 1.52)	1.13 (0.46, 2.79)	0.55 (0.23, 1.30)
Quality of Life- Social and Physical Health	1.08 (0.74, 1.58)	1.10 (0.59, 2.04)	1.03 (0.61, 1.74)
Quality of Life Psychological and Environmental	0.75 (0.47, 1.19)	0.84 (0.40, 1.78)	0.63 (0.32, 1.21)
CESD-R Score	1.01 (0.98, 1.03)	1.01 (0.98, 1.05)	1.0 (0.96, 1.03)
Variable
Triggered SRA with Suicidal Ideation	1.38 (0.65, 2.94)	0.79 (0.26, 2.45)	3.25 (1.02, 10.29)a
Triggered SRA with Self-Harm	1.14 (0.63, 2.04)	1.19 (0.48, 2.96)	1.08 (0.48, 2.12)
Abuse (Emotional, Physical, and/or Sexual)	5.92 (3.02, 11.61)b	3.89 (1.31, 11.49)a	10.57 (2.78, 29.57)b
Financial Stress	1.10 (0.62, 1.94)	1.19 (0.49, 2.92)	0.88 (0.39, 1.95)
Health Concern	1.80 (0.78, 4.18)	2.34 (0.63, 8.60)	1.12 (0.35, 2.60)
Cons	3.14 (0.40, 24.60)	2.45 (0.13, 47.83)	5.24 (0.26, 104.59)

^aStatistically significant at the 5% level.

^bStatistically significant at the 1% level.

^cFor these analyses, 21 cases were dropped from the SRA sub-sample due to missing data.

Qualitative Analysis

Content analysis revealed abuse (including emotional, physical, and sexual), financial stress and health concerns as the most frequently mentioned precipitants of ideation. Of the three, quantitative analysis found only abuse had a significant association with suicide behaviour for both genders (stronger for females). Qualitative examination of the SRA reports indicated that the three were often interconnected.

Abuse

Unstable home environments were reported by both genders, but there were differences in how abuse manifested. Among girls, abuse was frequently a result of violating gendered expectations, particularly regarding socializing with boys. A 17-year-old female said:

‘[My grandmother] likes abusing me and calling me a prostitute whenever she sees me standing with a boy. She used to abuse us saying that there is no need of educating a girl who is promiscuous… She even called my father and [he] stopped paying my fees… I mixed water and [the contents of a] battery. My mother found me and immediately gave me some milk and I was rushed to the hospital.’

Among boys, abuse appeared to be related to general discord in the family as illustrated by a 16-year-old male:

‘My father came home drunk and started abusing me and scolding me. I was so annoyed and run from the house and went to stay with my uncle. …my uncle spoke with my father and my father called me and told me to go back home and settle the matter…. I did not want to kill myself, but I was so frustrated that I was forced to run away from home.’

Financial stress

Lack of money also was frequently mentioned, including the stress of being sent home because of unpaid school fees or a family member threatening to stop paying school fees. Often, the participants were able to find financial support, usually from extended family. However, participants also worried about paying next year’s fees. A 17-year-old male said:

‘Sometimes you can spend days at home because of lack of fees. I want to be a teacher and change [the] situation at home and that is what gives me emotions when I see my fellow students and friends going to school and am just there at home because of fees. All my elder siblings dropped out of school but me… It reached a point when I felt like dying but my father sat me down and we spoke about it… My father decided to sell his parcel of land… and he cleared my school fees for this year. I never wanted to kill myself, but I felt like dying.’

Concerns about health

The third most frequently mentioned stressor was concern about one’s own or a family member’s health, which was often associated with financial problems and schooling:

‘[L]ife has been very hard. I [have] sickle cell disease and depend on my grandmother for food and my medications. [M]y uncle pays my school fees, but I have to hustle by doing work like packing dried small fish into sacks to be transported to other places to be sold…the little money I get I use for my basic need like buying sanitary pads, books etc. [M]y mother is also suffering from a certain condition that can’t allow her to work.’

(18-year-old female)

Pregnancy

Our quantitative result that pregnancy was protective against suicidal ideation for girls who had ever been pregnant was unexpected. The SRA reports indicated that seven female participants with moderate/high-risk mentioned pregnancy as a risk factor. As described by the following 18-year-old, this was mainly due to family discord and dropping out of school. After childbirth, family tensions subsided, and new responsibilities were apportioned.

‘My principal told my mother that he does not want to see a pregnant student in his school. Whenever I saw my friends going to school, I felt bad and I could cry…. My mother kept yelling at me and telling me that she can throw me out of her house… I had nowhere to go and decided to take Jik (a household detergent containing bleach). [M]y mother took me to hospital and spoke with me. When the child [w]as born [my mother] turned me into a prisoner. I could not go out or even spend more than five minutes in the shop… It reached a point that I escaped from home and stayed with my friend…. [My uncle] told them that they have to accept me and the baby… My mother is staying with the baby and my father is paying my school fees. I am staying with my grandmother and she is very supportive and understanding.’

HIV

None of the participants in the SRA sub-sample that had a positive or indeterminate HIV test result mentioned their results during their interview. However, ten participants with HIV-negative test results mentioned they were fearful of receiving a positive test result as illustrated by the following quote from a 19-year-old woman who was deemed at high risk for STBs:

‘I was having stress. I was coming for this test and that made me think a lot. I thought I may be found with HIV because I have a boyfriend and it has been long since we went for the HIV testing. That was why I was thinking a lot. It gave me stress.’

Discussion

Although still sizeable, findings from our rural community sample suggest a lower prevalence of suicide ideation (16%) compared to school- and hospital-based Kenyan youth samples (27–28% and 82%, respectively; Khasakhala et al., 2013; McKinnon et al., 2016). Increase in CESD-R scores was associated with higher odds of reporting suicide ideation, with no difference by gender. Females had higher odds of reporting suicide ideation and having moderate/high suicide risk than males. Sexually active females had twice the odds of abstaining peers of reporting ideation, and females who reported ideation had three times the odds of suicide behaviour. While these findings are consistent with previous studies (Bridge et al., 2006; Cash & Bridge, 2009; Cluver et al., 2015; Hallfors et al., 2004; Kinyanda et al., 2011), our study adds new information about gender differences in factors associated with suicide ideation, risk, and behaviour.

Our mixed methods analyses suggest a temporal relationship between pregnancy and suicide risk among females. While the quantitative data indicate that pregnancy is protective against ideation, participants who had ever been pregnant and underwent a risk assessment described suicide behaviour either when they found out about or during their pregnancy because of family conflict and having to drop out of school. It is possible that the reference period for the suicide items (past two weeks) did not encompass the highly stressful period experienced by these young women during pregnancy. As has been found in high-income countries (Denney, 2010; Qin et al., 2003) and illustrated in our study, during the latter phases of pregnancy or after delivery, family tensions subside and social support for the adolescent mother increases, lowering suicide risk. Longitudinal studies are needed in sub-Saharan African contexts to further explore these dynamics.

Sexually active males who had impregnated a partner had over three times the odds of reporting suicide ideation as those who had not impregnated someone. Among males assessed for suicide risk, those who were sexually active had four times the odds as abstaining peers of being in the moderate/high suicide risk group. To better interpret our data, we presented these findings to our study’s community advisory boards in meetings of youth, parents, and professionals working with youth. Advisory board members suggested males may be aware and fearful of new Kenyan laws making males liable for impregnating young school-going girls (e.g., school expulsion, legal consequences, or required financial support). However, further research is needed to elucidate the relationship between sexual activity, impregnating a partner, and suicide risk in this context.

Findings on sexual behaviour and suicide risk among both genders are particularly relevant in the context of sexual norms in Africa. Despite high rates of early sexual debut and adolescent pregnancy, early sexual involvement by both males and females and sex outside of marriage is culturally proscribed (Adaji et al., 2010; Kumar et al., 2018). This could increase risk for STBs among sexually active adolescents. Additionally, while pregnancy may be a risk factor for STBs for female adolescents, having a child could be protective against suicide. Conversely, for school-going adolescent boys, pre-marital pregnancy may be an economic and social stressor. Interventions that delay sexual debut, promote safe sex, and reduce unintended pregnancy may lower suicide risk.

Emotional, physical, and sexual abuse were important risk factors for suicide behaviour. These findings are consistent with others from previous studies indicating that adverse childhood experiences, including maltreatment and violence victimization, are risk factors for suicide ideation and attempts (Bridge et al., 2006; Cash & Bridge, 2009; Cluver et al., 2015; Devries et al., 2013). Our qualitative results suggest that financial stress, particularly lack of school fees, is another important concern among adolescents. Conversely, greater quality of life in aspects of social and physical health were associated with lower odds of ideation, especially for females. Our findings suggest a role for cash transfer and school support programs (Angeles et al., 2019; Green et al., 2019) and interventions for improving family relationships and parent-adolescent communication to mitigate suicide risk.

Among those who reported suicidal ideation in the survey, about half indicated not meaning to respond as they did. Ambivalence and change of mind are common among adolescents with suicidal ideation (Czyz et al., 2019). Our advisory board members suggested that, given that suicide attempt is illegal in Kenya, these participants may have been reticent to disclose their experiences to research staff, fearing repercussions for themselves or their families. Moreover, some youth may not have expected to be assessed for suicide risk. Concerns about confidentiality and stigma are barriers to adolescents seeking help for STBs (Hawton et al., 2012). Our findings highlight a need to revise Kenyan laws that may undermine the identification of high-risk adolescents and effectiveness of suicide-related interventions. There have also been calls for the reform of similar laws in other sub-Saharan African countries (Adinkrah, 2016).

In our study, a positive or indeterminate HIV test result was not associated with triggering the SRA. Previous research has found suicidal behaviour higher among already-diagnosed HIV-positive children ages 10–17 compared to HIV-unaffected children (Ng et al., 2015). Our results, in contrast, were obtained from study participants who were unaware of their HIV status when answering questions about STBs. Given that large numbers of HIV-infected adolescents are unaware of their infection (UNAIDS, 2017), our findings imply that research on their mental health status is important to investigate further. Because HIV prevalence in our study was low, we were unable to examine whether HIV status was associated with suicide risk and behaviours. Further research is needed on the relationship between STBs and HIV among rural residing adolescents in high HIV prevalence regions.

Study limitations include the use of a non-probability sample; findings may not be generalizable to all adolescents in the region or to low HIV-prevalence regions. However, the study benefited from a large community sample and included adolescents not attending school. Another limitation is that we did not inquire about sexual orientation; lesbian, gay, bisexual, transgender, and queer (LGBTQ) youth are more vulnerable to suicide risk than their heterosexual counterparts (di Giacomo et al., 2018). Although the survey was administered via ACASI, which offers privacy, we do not know whether responses to the survey or SRA questions were more accurate for the ‘wrong button’ group. However, our findings are similar to a study in Zimbabwe where adolescents self-reported suicidal ideation twice as often when answering by ACASI versus interviewer-administration (Langhaug et al., 2009). Finally, our SRA protocol employed an open-ended interview format and inquired about risk factors documented in the literature. We did not, however, ask about HIV or participants’ HIV test results.

While using mixed methods provided an opportunity to triangulate data and understand study findings in greater depth, the variables we extracted from SRA reports were not validated measures (either clinically or statistically). For instance, the coding of SRAs as low, medium, or high was not based on a validated process. Also, our risk assessment based on the past two weeks might have underestimated the prevalence of suicidal ideation. Additionally, because we did not follow up with SRA participants over a longer period, we do not know if our protocols led to use of services or prevented adverse outcomes.

Despite these limitations, our study contributes new information about adolescent suicide risk, suggesting a need to develop effective programs to identify and support at-risk sexually active and pregnant youth. Further research is needed to revise policies that are obstacles to help-seeking and to identify how to best intervene and support families in rural communities with adolescents with suicidal ideation. Given the limited availability of mental health services in sub-Saharan rural areas, culturally-appropriate and adolescent-friendly approaches involving families, schools, health facilities, and community gatekeepers (e.g., teachers, youth leaders) may be most promising.