Intimate Partnerships, Suicidal Ideation and Suicide-Related Hospitalization Among Young Kenyan Men

Abstract

The first study focused on a three-month chart review containing information on suicide attempts (n = 34) admitted to a local mission hospital in Meru County, Kenya. The second study utilized a cross-sectional survey administered to men 18–34 years old (n = 532) residing in rural Kenya. Data posit intimate partnership discord as salient to suicide ideation and behavior. Men who reported their partner status was “divorced” had four-times the odds of reporting suicide ideation than other partnership states, an association significantly mediated by loneliness. Violent conflict tactics predicted suicidal ideation, mediated by loneliness and decreased marital satisfaction.

Introduction

Deaths due to suicide are a global public health concern claiming an estimated 804,000 lives annually (WHO 2014). The World Health Organization (WHO) estimates that 8% of completed suicides occur within Africa (WHO 2014), while Africa makes up approximately 16.7% of the world’s population (United Nations 2019). Though it would appear that suicides are disproportionately lower compared to else-where, data on suicides in Africa (and worldwide) are not reliable due to underreporting, misreporting, and inconsistent forms of mortality surveillance (Mars et al. 2014). For example, only 10% of African countries report information about suicide mortality to the World Health Organization, accounting for only 15% of the continent’s population (Mars et al. 2014). Country-specific estimates of deaths due to suicide suffers from the same limitations. Kenya does not have an active suicide surveillance system. The WHO (2014) estimates that the rate of death due to suicide among males in Kenya is 24.4 per 100,000 (compared to 15.0 per 100,000 among males worldwide), and the Global Burden of Disease Study (Naghavi 2019) has estimated 15.4 per 100,000 among males (compared to 18.7 per 100,000 in eastern sub-Saharan Africa. Inconsistent reporting across Africa inhibits research, and thus little is understood about the pattern of suicidal behavior in these countries, and in particular, Kenya. Smaller observational studies provide some indication of the number of deaths by suicide and the incidence of suicide attempts and suicide ideation (both risk factors for later suicide completion). For example, it was estimated that 9% of cases of fatal intentional injuries (i.e. homicide and suicide) from two areas of Nairobi were due to suicide (Ziraba et al. 2011). A study of 1158 households from a rural area of Kenya found that approximately 7.9% of respondents had experienced suicide ideation and 1.9% had attempted suicide at some point in their lives; the odds of suicide ideation were significantly higher for women than for men (Jenkins et al. 2015). A study of 2800 patients at hospitals (psychiatric units excluded) found that 10.5% reported suicide ideation, the highest percentage (14.5%) being among young adults 18–20 years of age (Ndetei et al. 2010). Improving surveillance and the reporting of vital registration data for the second leading cause of death for young adults globally (15–29 years) has become an imperative in the global call to preventing suicides in recent years (WHO 2014).

Suicidal ideation (including thoughts and plans) is a significant longitudinal predictor of later suicide attempts and of later suicide deaths (Ribeiro et al. 2016). Therefore, understanding risk factors for suicidal ideation in a population could shed light on potential targets for community-level suicide prevention activities.

The Interpersonal Theory of Suicide predicts suicidal desire is more probable if an individual experiences thwarted belongingness, perceived burdensomeness (the person feels like they are a burden to family and friends), and hopelessness about these states (Van Orden et al. 2010). Belonging is a basic human need that underlies human motivation (Baumeister and Leary 1995) and is the foundation of long-term, mutually supporting, relationships (Van Orden et al. 2010). Previous large cohort studies conducted in Australia have shown a positive association between thwarted belongingness and suicidal ideation (Christensen et al. 2013).

Intimate partnerships are one potential source of mutual care and sense of belonging (Kissane and McLaren 2006). Being married has been found to be protective against suicide behavior and completed suicide in a wide variety of settings (Hawton 2000; Qin et al. 2000). Multiple investigations have observed that separation, divorce, and widowhood are associated with completed suicide in men (Kposowa 2000; Wyder et al. 2009). Three recently completed systematic reviews identify associations between intimate partner violence or divorce and suicide ideation and suicide attempts across many different research settings, though none of these reviews report studies conducted in sub-Saharan Africa associating relationship status or presence of violent conflict and suicide ideation (McLaughlin et al. 2012; Devries et al. 2013; Kazan et al. 2016). Conflicting observations have been noted from investigations conducted in Sub-Saharan Africa when examining a possible relationship between interpersonal relationship dynamics and suicidal ideation, suicide attempts, or completed suicides, often with differences between males and females. In Kampala, Uganda, deliberate self-harm in individuals at least 15 years old was associated with partner-related negative life events, such as emotional mistreatment by their partner and loneliness in the past year (Kinyanda et al. 2005). Responses collected from male students in Uganda indicated that problems with relationships with friends, family and partner were reasons for attempting suicide (Knizek et al. 2011). A case–control psychological autopsy study examining risk factors for completed suicide in Kampala, Uganda did not observe a relationship between divorce and partner conflict and suicide deaths (Kinyanda et al. 2011); however, a similar study of 104 cases of suicide in a rural area of Kenya found that 9.4% of cases reported stress related to marital conflicts and family disputes (Bitta et al. 2018). A 2016 study in Kampala found that women who had experienced physical or sexual IPV were significantly more likely to report an attempted suicide in the past year; however, there was no association between IPV and suicide attempts in males (Kinyanda et al. 2016). A recent study from South Africa found an association between depressive symptoms among men and intimate partner violence, though the study did not consider suicide ideation or attempts as an outcome (Nduna et al. 2010). Lastly, in a study of general medical outpatients, there was significant bivariate association between perpetrating or experiencing IPV in the past week and SI in the past month among women only (Ongeri et al. 2018).

There are theoretical reasons to evaluate whether intimate partnership conflict and dissolution through death or divorce predict suicide ideation or suicide attempts, and why this association would be mediated by loneliness and meaning in life, specifically in men. Marital discord, divorce and partner death have been associated with loneliness and thwarted belongingness (Robustelli et al. 2015; Woodward et al. 1981; Lester and Gunn 2012). McLaren et al. (2015) demonstrated that the death of a spouse for Australian men was associated with lower sense of belonging. Loneliness is the emotional pain associated with an unmet need to belong, which is not necessarily associated with social network size (Mellor et al. 2008). As meaning in life is supported by social cohesion and integration, and intimate partnerships may provide this sense of belonging, the loss of or threat to intimate partnerships may decrease meaning in life (Lambert et al. 2013; Durkheim 1979, pp. 171–216). Further investigation into a possible relationship between intimate partner conflicts, dissolution of intimate partnerships, meaning in life, loneliness and suicide ideation and suicide attempts is therefore needed.

Study Aims

In these two studies, we aim to characterize suicide ideation and suicide attempts by intimate partnership status among young Kenyan men.

Study 1

Data for the first study come from social narratives of attempted suicide cases admitting to a private mission hospital.

Hypothesis 1

We hypothesized that narratives explaining suicide attempt cases in a private mission hospital would make substantial reference to intimate partner conflict, divorce or partner death.

Study 2

Data for the second study come from a community-based survey using a semi-structured questionnaire with open line-listing of reasons to live and reasons to die among respondents who prescreened positive for suicide ideation.

Hypothesis 2

We hypothesized that reasons to die among young men in a community-based sample who screened positive for suicide ideation would make substantial reference to intimate partner conflict, divorce or partner death.

Hypothesis 3a

We hypothesized that statistical analyses of young men reporting suicide ideation would show a positive association between divorce or partner death and suicide ideation, and intimate partner conflict and suicide ideation among those reporting current intimate relationship status.

Hypothesis 3b

We further hypothesized that the association between divorce or widow status and suicide ideation was mediated by increased loneliness and the association between intimate partner conflict and suicide ideation was mediated by decreased marital satisfaction and increased loneliness.

Finally, we assessed meaning in life as a potential mediating variable in both divorce and intimate partner conflict models.

Methods

Methods for each study are presented individually.

Study 1: Chart Review of Patients Admitted for Attempted Suicide

Participants

For the first study, a three-month chart review was conducted at the host hospital—Maua Methodist Hospital in Meru County, Kenya. Hospital patients are admitted from across Meru County, a largely rural area with many small townships ranging in population size from 10,000 to 50,000. All attempted suicide cases admitted to the hospital were culled and reviewed by nursing staff. Chart information contained data on age, gender, housing location (village, name of tribal chief, nearest primary school), occupation, date of arrival, marital status presenting condition, and additional notes on patient social histories recorded by the clinician. All available data were tabulated, and the additional notes were coded and tabulated.

Procedure

Study 1 utilized a chart review that sorted through paper-based charts and hand-written notes for a three-month period, spanning February 2015 through April 2015. All reported suicide attempts were reviewed. Psychiatric services are limited across sub-Saharan Africa and securing an appointment with the psychiatrist who attends at this hospital requires advanced notice. Suicidal patients are admitted through the emergency department. All notes were generated by primary care nurses following hospital admission through the emergency department. Any report that included patient- or family-provided social history information was coded and tabulated. Social history information included narratives provided by patient or patient families regarding reported or assumed explanations for suicidal behavior.

Measures

Social history data in study 1 were obtained from patient charts recorded by primary care nurses. The social history data provided information on reported situation prior to suicide attempt. Demographic data were also obtained from patient charts.

Outcome Variable

All respondents in study 1 shared the common outcome of having previously attempted suicide.

Exposure Variables

Study 1 exposures were those supplied on the patient chart—age, gender, marital status presenting condition, treatment given and a space for additional notes on patient social histories recorded by the clinician.

Analysis Plan

Study 1 relied on content analysis to test the hypothesis that narratives explaining suicide attempt cases in a private mission hospital make substantial reference to intimate partner conflict and divorce or death. Data were coded by independent coders, tabulated and ranked based on frequency of appearance (Hsieh and Shannon 2005). Two coders read and coded responses into response categories, and differences were resolved by a third coder. Primary care nurses recorded the social histories after intake at the hospital, and the coders were independent of the intake and hospitalization process.

Study 2: Semi-structured Interview of a Community Based Sample of Kenyan Men 18–34 Years

Participants

Data for the second study came from a semi-structured interview administered to a community-based sample. Survey participants were selected from 11 townships around Maua Methodist Hospital in Meru County, Kenya. Townships were selected based on community-based clinics provided by the hospital and were selected to provide greater epidemiological context to suicide prevention efforts offered by the hospital. As the majority of attempted suicide cases reported to the hospital were males (65%) and under the age of 35 (67%), consistent with the global suicide epidemic that disproportionately affects young males, study 2 focused exclusively on young males, aged 18–34. Survey participants were included if they met these characteristics and consented to be interviewed. Participants were told in advance they would receive no direct benefit, but the study results would enable a more informed response to issues related to mental health affecting young men in the community. A total of 532 participants were interviewed.

Procedure

In study 2, trained interviewers were paid to administer the semi-structure interview, which began from the three most populated centers in each township, and followed predetermined routes based on random number tables. Interviewers approached every other house to assess for eligibility. Interviewers continued at each location, following each pre-specified path, until reaching the end of the interview day. This design can be considered a uniform walk method, which is a validated method of conducting household surveys when full housing registers are unavailable (Thompson 2006). Selected households were assessed for the presence of an eligible male (aged 18–34 years). Where more than one eligible male was available, one male was randomly selected. A total of 772 houses were visited and found eligible. Of these, 240 (31%) refused and 532 (69%) completed an interview.

Interviewers were male nursing students from a local university who had been trained in conducting standardized questionnaires. The surveys were conducted in the local language—Kimeru. Interview questionnaires were created in English, translated into Kimeru and back translated for refinement and validation.

Measures

Data in study 2 were generated by both quantitative scales recording psychosocial variables and an open line-listing supplying reasons to live and reasons to die from survey respondents who prescreened positive for suicide ideation on the open response section of the Modified Scale of Suicide Ideation (MSSI; Miller et al. 1986). Responses to the two prompts were tabulated and considered separately from quantified data collected by the rest of the survey. All other quantitative scales in the survey were closed-ended.

Outcome Variable

Study 2 measured suicide ideation using the Modified Scale for Suicide Ideation, an 18-item 4-point Likert type scale (MSSI; Miller et al. 1986). The MSSI provides a summative measure of desire, ideation, plans and preparation for suicide (Pettit et al. 2009) and has been validated across multiple cultures (Yang and Clum 1994). Respondents answer four initial screening questions (scored 0–3) in the MSSI: Over the past day or two (1) Have you thought about wanting to die?; (2) Have you thought that you want to live?; (3) When you have thought about wanting to die, did you want to kill yourself? Do you want to kill yourself now?; (4) Right now would you deliberately ignore taking care of your health? Have you thought it might be good to leave life/death to chance? The first two questions align with the concept of passive suicide ideation, as they refer to desires for death or life that are not accompanied by a plan to carry out death. The first three questions are each followed up by asking about frequency of these type of thoughts. Respondents must either have strong (2 or 3) passive ideation on the first two questions or have the presence of any active ideation on the second two questions in order to be prompted to take the remainder of the scale. Scale developers proposed three categories for severity of suicide ideation: low (range 0–8), mild/moderate (range 9–20), and severe (range 21–54) (Miller et al. 1991).

Passive and active suicide ideation may be distinct or coexisting in an individual, and the presence of either in a clinical setting indicates risk for a suicide attempt. Suicide attempts do not necessarily follow suicide ideation but they are more likely to follow severe ideation compared to mild or no suicide ideation. In this study, to enable the most strategic allocation of resources, we compare respondents who took the full scale and scored in the “severe” category to those who took the full scale and scored in the lower two categories using a binary variable. Following the four screening questions, the MSSI includes a line-listing activity where respondents who have prescreened positive for suicide ideation are prompted to provide as many reasons as they can to live and as many reasons as they can to die.

Exposure Variables

In study 2, two measurements served as primary exposures. Intimate partner status was measured using the single item “what is your current relationship status,” with “multiple partners,” “single partner/monogamous,” “divorce/separated,” “never partnered,” and “widowed” as response categories. For the purposes of this investigation, respondents who reported they were “divorced/separated” or “widowed” were considered in one category and compared with respondents who were married/partnered with one or multiple partners or were never partnered. Interpersonal theory of suicide posits that thwarted belongingness is one predisposing factor for suicide ideation. Men who had previously attempted to partner but were currently unpartnered were considered more likely to experience thwarted belongingness. Men who had never been partnered may or may not have desired an intimate partnership, and thus were not included with those who had been divorced or widowed.

The second exposure was intimate partner conflict, measured using the short-form of the Conflict Tactics Scale (CTS2S; Straus and Douglas 2004). The CTS2S uses an 8-point scale to evaluate the presence of different conflict tactics within a romantic relationship: negotiation, physical assault and psychological aggression. Factor analysis of the present data revealed a two-factor solution – aggression and negotiation (α = 0.93 and 0.84, respectively). The scree plot for the CTS2S is included in the supplemental materials. As the CTS2S required respondents to report on active intimate partner conflict tactics, we only delivered the scale to respondents who indicated being in an active partnership (n = 263/532, 49.4%).

Mediating Variables

Loneliness was measured using the 15-item Social and Emotional Loneliness Scale for Adults short-form (DiTommaso et al. 2004; SELSA-s). Items on the SELSA-s are measured on a 7-point scale ranging from strongly disagree (1) to strongly agree (7). Items include statements like “I feel alone when I am with my family,” “I feel part of a group of friends,” and “I have a romantic partner with whom I share my most intimate thoughts and feelings.” The SELSA-s measures social and emotional isolation on three subscales: romantic, family and friends. Validated previously in multiples languages and cultures (Adamczyk 2013; Çeçen 2007), the scale had very good internal consistency in the present sample (α = 0.82).

Marital satisfaction was measured using the 3-item Kansas Marital Satisfaction scale (KMS; Schumm et al. 1986). Items include statements like “how satisfied are you with your wife as a spouse” on a 7-point scale (α = 0.94). Similar to the CTS2, the marital satisfaction survey required respondents to be in a committed partnership. Intimate relationships in Kenyan culture may be considered as “marriage” by partners even if there has been no formal ceremony (Hakansson 1994). Therefore, consistent with the conventional definition in the Demographic and Health Survey, we included formal marriages and consensual unions in the definition of marriage for this scale (Bongaarts 2007).

Meaning in life was measured using the Meaning in Life Questionnaire (MLQ; Steger et al. 2006). The scale uses 10-items on a 7-point Likert-type scale to measure the presence of, and search for, meaning in life (α = 0.89). Items include statements like “I understand my life’s meaning.” Previous research on the MLQ in Kenya, Turkey, China and the United States have found the measure to be reliable and valid (Goodman et al. 2017; Boyraz et al. 2013; Wang and Dai 2008).

Control Measures

Control variables included age, years of completed schooling, two measures of wealth, number of children and number of school-enrolled children. One measure of wealth reflected the total number of cows, chickens, goats and sheep owned by the household. The other measure of wealth added the number of household assets owned from the following list: electricity, radio, television, telephone, refrigerator, land and rooms used for sleeping. Age was included as a continuous measure of years reported. These two asset-based measures of wealth have been broadly used and validated in rural Africa (Morris et al. 2000).

Analysis Plan

Study 2 relied on quantitative survey data and line-listing data. Respondents who screened positive for suicide ideation were prompted to provide as many reasons to live and die as they could. These responses were written down by the interviewer, and later coded by independent coders. The codes were then counted and ordered by frequency.

Statistical analysis of the structured portion of the survey utilized regression-based path analysis. Controlling for wealth, age, and years of formal education, we regressed suicide ideation on divorce status in the first set of models and, in the second set of models, the two manifest factors from the conflict tactics scale (negotiation and aggression). Mediation analysis was conducted using bootstrapped error terms (1000 replicates), and initially considered loneliness in the divorce set of models and marital satisfaction and loneliness in the conflict tactics set of models. As theoretical and empirical literature has established meaning in life as a consequence of social connection, we wanted to first consider the contributions of measures of social connection before including meaning in life. We do not know of a program that can facilitate significance testing of serial pathways with hierarchical data and a binary outcome. Thus, path analyses and significance testing considered only loneliness and marital satisfaction, though the regression analyses include meaning in life in the third model in both sets.

STATA v.14 was used to conduct all analyses. Path analysis was conducted using generalized structural equation modelling (GSEM command in STATA). Significance of indirect pathways was conducted using the “binary_mediation” command (Ender 2010).

Ethical Consideration

The ethics committee at Maua Methodist Hospital in Kenya reviewed and approved data for study 1 and study 2 prior to data collection. The Institutional Review Board at the University of Texas Medical Branch approved the publication of data from both studies. All subject data were anonymized prior to analysis, and all subjects provided informed consent prior to participation in the study.

Results

Study 1

The results of the chart review are displayed in Table 1. Males comprised 64.7%, or 22 of the 34 recorded suicide attempts in the hospital during the three-month review period. Because of this study’s focus on men, demographic characteristics and reasons for attempt are only displayed for men. Of the 22 men, over a quarter were married and another quarter were single. The leading cause recorded for suicide attempts among men was domestic quarrel, comprising 31.8% of responses. More than half of men (59%) did not list a known cause for suicide behavior. The mean age of male respondents was 30.3 years.

Table 1.

Mission hospital chart review of suicide attempts during a 3-month period

	N	Percent
Total population (male and female) (n = 34)
Male	22	64.70%
Female	12	35.30%
Characteristics of men only (n = 22)
Marital status
Married	9	40.9%
Widowed	1	4.5%
Single	8	36.4%
Relationship status not known	4	18.2%
Age (mean, SD)a	30.3 (12.7)
Reason for attempting suicide
Domestic quarrelb	7	31.82%
Fight with other family member	1	4.55%
Illness	1	4.55%
Not known	13	59.09%
Married men who attempted suicide because of a domestic quarrel	5	22.73%

^an = 18

^bDomestic quarrel indicates quarrel with spouse or partner

Study 2

Figure 1 shows the tabulated codes of reasons to live, as reported in the semi-structured interview by 68 men who initially screened positive for presence of suicide ideation. There were six main themes for the 32 open-response reasons to live, with the most commonly given reason related to friends and family (n = 21, 65.6%). These responses ranged from “I do not want to leave my spouse/single mother/children/friends alone” to “I have a family that loves me.” Providing care for dependent family supplied some respondents with meaning, as did receiving care and value from family. The second most commonly stated reason for living related to spirituality and religious practice (n = 6, 18.8%), with statements including “I am feeling hope after attending church yesterday,” “God created me to be successful,” and “I have peace due to my faith in God.” There were 37 respondents who did not provide a reason to live, and one respondent who gave 2 reasons to live.

Fig. 1.

Reasons to live among Kenyan men (18–34 years)

Figure 2 shows the tabulated codes of reasons to die for the 68 men who screened positive for suicide ideation. There were 52 open-responses coded into 12 themes. The most common response related to poverty, with responses relating to a loss of wealth (e.g. stolen land or possessions) or an absence of wealth due to harsh economic conditions, including food shortages, crop failure, lack of work opportunities, and high costs of living (18; 37%). The second most coded response related to intimate partner conflict (n = 8, 15.1%) and included statements such as “my wife is beating me even in front of my kids, without any apparent reason,” “my wife left without any reason and she had already changed family properties… now I am left with nothing to finance my family since I am the bread winner in the family,” and “my wife is giving me stress in marriage life.” There were 33 respondents who did not provide a reason to die, and 14 respondents who gave more than one reason.

Fig. 2.

Reasons to die among Kenyan men (18–34 years)

More than 60% of the 13 respondents who mentioned poverty in conjunction with another reason to die mentioned poverty and some kind of marital or family conflict or poverty and lack of social connection; however, of the 18 respondents citing poverty, only 4 cited poverty alone was a reason to die. Two-thirds of those citing marital conflict as a reason to die mentioned it alone (i.e. without citing any other reason).

Table 2 shows the univariate characteristics and bivariate analysis of all 532 respondents, with measures assessed by presence of suicide ideation in bivariate analysis. Twelve percent of all respondents indicated they were divorced, and the percent of all respondents with suicide ideation who were divorced was 30%. The percentage of cases of suicide ideation in the sample population attributable to divorce was 20% (PAR% not shown in Table). Respondents who reported suicide ideation had significantly more aggressive conflicts, but there was no statistical difference with respect to negotiation-based conflicts. Respondents with suicide ideation reported significantly less marital satisfaction, collective self-esteem and meaning in life, and more loneliness. Though not significant differences, respondents with suicide ideation were on average two years older, owned fewer non-livestock assets but more livestock assets, and had one less year of schooling.

Table 2.

Univariate and bivariate characteristics of respondents

	Total sample (n = 532)		No suicide ideation (n = 469)		Suicide ideation present (n = 63)		P-value
	Mean, SD or n (%)		Mean, SD or n (%)		Mean, SD or n (%)
Divorced	65 (12.08%)		46 (9.81%)		19 (30.16%)		< 0.001
Aggressive conflict tactics, SD	0	1	− 0.07	0.98	0.85	0.75	< 0.001
Negotiation conflict tactics, SD	0	1	− 0.01	1.03	− 0.01	0.72	0.88
Marital satisfaction, SD	0	1	0.1	0.95	− 1.02	0.75	< 0.001
Loneliness, SD	0	1	− 0.1	0.99	0.73	0.73	< 0.001
Collective self-esteem, SD	0	1	0.09	1.02	− 0.67	0.44	< 0.001
Meaning in life, SD	0	1	0.21	0.79	− 1.51	1.1	< 0.001
Age, years	25.99	4.99	25.75	5.02	27.62	4.38	0.01
Non-livestock assets	3.11	1.29	3.16	1.26	2.69	1.43	0.02
Livestock assets	19.67	18.49	19.42	17.04	21.48	26.77	0.32
Years of completed school	8.97	3.77	9.09	3.65	8.02	4.54	0.11
Number of household children	1.85	1.64	1.77	1.61	2.4	1.76	0.003
Number of children in school	1.14	1.7	1.16	1.8	0.95	0.96	0.83
Suicide ideation present	63 (11.84%)

Divorced status compared to other marital states (single, married—monogamous or polygamous, and widowed). Aggressive conflict tactics and negotiation conflict tactics are two factors of the CTS2S, each with good internal consistency (α = 0.93 and 0.84, respectively). Loneliness was measured using the SELSA-s scale (α = 0.82). Meaning in life was measured using the MLQ (α = 0.89). Marital satisfaction was measured using the KMS scale (α = 0.94). All scales standardized. P-value for χ² or ranksum test

Table 3 shows two sets of three models. The first set, the “a” set, shows the association of divorce/widowed status and suicide ideation, adding loneliness in the second model and meaning in life and loneliness in the third model. Respondents who reported being currently divorced or widowed had odds of reporting suicide ideation that were 4 times higher than other respondents (OR 4.05, 95% CI 1.2, 13.66). After including loneliness in the model (2a), the association between divorce and suicide ideation attenuates, and attenuates even further after including meaning in life and loneliness in the model (3a). The second set of models, the “b” set, shows the results of the serial mediation of conflict tactics and suicide ideation by loneliness, marital satisfaction, and meaning in life. For each standard deviation increase in aggressive tactics, the odds of reporting suicide ideation increase 2.3 times (OR: 2.32, 95% CI 1.32, 4.09). There is no significant association between negotiation tactics and suicide ideation, and the variable is dropped in the latter two models (2b and 3b). After including marital satisfaction and loneliness, the association between aggressive tactics and suicide ideation attenuates, and does so completely after including meaning in life. Loneliness did not significantly increase the odds of reporting suicide ideation in either of the final models (3a and 3b). In both of the final models, an increase in one standard deviation of meaning in life decreased the odds of reporting suicide ideation by 80%. In models 3a and 3b, for each standard deviation increase in marital satisfaction, the odds of reporting suicide ideation decrease by 63% and 68%, respectively.

Table 3.

Survey-adjusted logistic regression models depicting mediation of marital factors and suicide ideation

	Model 1a			Model 2a			Model 3a
	OR	95%CI		OR	95% CI		OR	95% CI
Currently divorced or widowed	4.05*	1.2	13.66	3.22t	0.9	11.53	2.55	0.73	8.91
Mediating factors
Loneliness, SD				2.17***	1.58	2.96	1.27	0.8	2.04
Meaning in life, SD							0.2***	0.12	0.34
	Model 1b			Model 2b			Model 3b
	OR	95% CI		OR	95% CI		OR	95% CI
Aggressive tactics, SD	2.32**	1.32	4.09	1.56	0.87	2.8	0.68	0.26	1.76
Negotiation tactics, SD	1.04	0.66	1.66
Mediating factors
Loneliness, SD				2.72*	1.14	6.52	2.85	0.57	14.17
Marital satisfaction, SD				0.37***	0.2	0.69	0.31*	0.11	0.89
Meaning in life, SD							0.2***	0.1	0.4

Two sets (series “a” and “b”) of three survey-adjusted logistic regression models assess associations between marital conditions and suicide ideation. The “a” series assesses associations between divorce and suicide ideation. The “b” series assesses associations between conflict tactics and suicide ideation. As the negotiation tactic was not associated with suicide ideation in Model 1b, it was excluded from subsequent models. In the first model of both series (“a” and “b”), the models establish the association between exposure and suicide ideation. In the second model of both series, relevant mediator variables are included. In the third model of both series, meaning in life is included. All models control for two forms of wealth, age and education

^tindicates p < 0.07,

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

Figures 3 and 4 depict the path analyses of divorce, aggressive conflict tactics and suicide ideation. In the situation of divorce, risk of suicide ideation is significantly higher. Roughly one-fourth of this association is mediated by loneliness. That is, men who are divorced report significantly higher levels of loneliness, and this association explains a substantial portion of the divorce-suicide ideation association. In the situation of aggressive conflict tactics, suicide ideation is also significantly higher. Respondents who used aggressive tactics to resolve conflicts reported significantly less marital satisfaction and significantly more loneliness. Lower marital satisfaction (27.5%) and more loneliness (39%) significantly mediated associations between aggressive conflict tactics and suicide ideation. Bootstrapped analyses showed each pathway to be statistically significant (p < 0.05).

Fig. 3.

Path analysis of divorce status and suicide ideation, mediated by loneliness

Notes: Path analysis of current divorce status and suicide ideation. Lonelincss mcasurcd using SELSA-s. Suicide ideation measured using MSSI. * indicates p<0.05; *** p<0.001. Decomposition of effects analyzed using 1000 bootstrap replicates.

Fig. 4.

Path analysis of aggressive conflict tactics and suicide ideation, mediated by loneliness and marital satisfaction

Notes: Path analysis of aggressive conflict tactics and suicide ideation. Conflict tactics measured using CTS2s. Loneliness measured using SELSA-s. Marital satisfaction measured using KMS. All scales standardized. Suicide ideation measured using MSSI. * indicates p<0.05; ** p<0.01. *** p<0.001. Decomposition of effects analyzed using 1000 bootstrap replicates.

Discussion

Intimate relationships have been explored primarily with reference to sexual and reproductive health and HIV transmission within sub-Saharan Africa (e.g. Eyawo et al. 2010). Without debate, HIV transmission, intimate partner violence, child marriage, and unintended pregnancies are major contributors to morbidity and mortality in sub-Saharan Africa (Eyawo et al. 2010; Raj 2010; Alio et al. 2009). Less investigated is how intimate relationship circumstances relate to mental and social health. This study is one of few investigating intimate relationship factors and suicide ideation and suicide attempts in Low- and Middle-Income Countries (Shibre et al. 2014). Previous research has oscillated between whether divorce is a risk factor for suicide ideation, though intimate partner violence has been established as a risk factor for suicide for women (Shibre et al. 2014).

Data from two studies show here support for a cluster of related hypotheses. First, narratives from clinical reports (study 1) explain suicide attempts in a private mission hospital with substantial reference to intimate partner conflict, divorce or partner death. The line-listing of reasons to live or die from those who prescreened positive for suicidal ideation in the community-based sample of 532 respondents likewise makes substantial reference to intimate partner conflict, divorce or partner death. Additionally, statistical analyses of survey data show a positive association between divorce or partner death and suicidal ideation which was mediated by loneliness. Lastly, among those in current intimate partnerships, intimate partner conflict predicted more suicidal ideation and was statistically mediated by loneliness and marital satisfaction.

These multiple types of data highlight the importance of intimate partner relationships to suicide ideation in this study population. Social histories recorded in patient records (study 1), though inconsistently coded due to absence of standardized social history indicators, suggest that domestic quarrels are important to community perceptions of motivations behind suicide attempts. Over a quarter (31.8%) of the attempted suicide cases among males observed during the 3-month period in the hospital record were men who attempted suicide due to domestic quarrels. These findings are consistent with reported reasons to live and die among men who screened positive for the presence of suicide ideation in study 2. The most commonly stated reason to live relates to meaningful relationships with family or friends. The need to belong is a fundamental motivation, and belongingness theoretically strengthens the will to live (Baumeister and Leary 1995; Wong 2012). While this is the first study to establish close social relationships as a potential source of will to live among young men in sub-Saharan Africa, these findings are consistent with standing theory tested in other settings. It is interesting that the most commonly given reason to live is closely connected to the second most commonly reason to die among those screened for suicide. While poverty was the most commonly given reason to die, intimate partner conflict was often stated as a factor intertwined with poverty. For example, one man who was unable to financially provide for his children following a divorce gave this struggle as a reason to die; another man said his wife left him due to poverty. Important to note is that more than three quarters of respondents citing poverty as a reason to die mentioned it in conjunction with some other reason, e.g. stress, marital or family conflict, sickness, or lack of social connection; whereas less than a quarter cited poverty alone as a reason to die. This could indicate that poverty alone may not be a sufficient reason to consider suicide, but that other stresses or hardships, e.g. those relating to interpersonal relationships, magnify or are magnified by the effects of poverty. In contrast, the majority of those citing marital conflict as a reason to die did not provide another reason, suggesting that marital conflict is a substantial source of stress and suffering to those contemplating suicide.

Still, the fact that poverty was the most commonly cited reason to die is consistent with studies on the positive association between different dimensions of poverty and suicide ideation and behavior (e.g. Whitley et al. 1999; Iemmi et al. 2016). Poverty places heavy burdens on individuals across all facets of life make it; however, it is important to remember that perceptions of poverty are dependent on geographical, social, familial, and cultural factors of individuals’ daily lives (Savadogo et al. 2015) and that poverty cannot only be defined solely in financial or economic terms (Alkire and Santos 2013). There were no significant bivariate or multivariate associations between the measures of wealth and suicide ideation in the second study. However, objective measures of wealth/poverty do not account for the subjective measure or perception of one’s own economic standing or wellbeing. Furthermore, they offer little sensitivity in communities where poverty is widespread and income disparity marginal. The free response data indicate that poverty does play a role in suicide ideation in this population; however, how respondents defined and perceived poverty in the free response question may not have been captured by the wealth indices used in the quantitative analysis.

The fact that a large percentage of respondents did not provide a reason to live or to die in the open-response question in the second study is disappointing in that we did not gain insight into these men’s perceived motivations for suicide; however, it could indicate one of the following. Men could have death wishes without clear or defined reasons for feeling this way; some men could have been unwilling to reveal reasons for their ideation due to internalized stigma or shame.

Just over 12% of respondents in the second study indicated they were divorced, which increased the odds of suicide ideation by over 4 times after controlling for wealth, education and age. Mediation analysis showed that divorced men were more likely to be lonely, which mediated approximately one-fourth of the association between divorce and suicide ideation. To our knowledge, there is no literature about support groups specifically for divorced men in sub-Saharan Africa. In high-income settings, divorce support groups provide a place for men and women who have experienced divorce to find social companionship and support in the transition to a new life status (Oygard 2001). Such support groups have been based in churches in high-income settings (Saunders et al. 2013), and our finding that church and spiritual reasons were the second most commonly stated reason for living suggests churches as potential partners in suicide prevention. However, further research will have to explore the cultural acceptability of church-based, and other location-based, support groups in Kenya and other sub-Saharan African settings. Stigma has been associated with divorce status in high-income countries (Hoffman and Willers 1996). Divorce-related stigma may be operative in Kimeru culture, and more broadly across Kenya and sub-Saharan Africa; however, there is little to no published research. Stigma would likely exacerbate feelings of loneliness (Grov et al. 2010) and could decrease meaning in life, especially if stigma is internalized (Or et al 2013). This suggests an opportunity for churches to promote acceptance of people in the midst of life challenges. The percent of suicide ideation in the sample population that is attributable to divorce is 22.5%. Taken with the current limited budget for mental health promotion and suicide prevention (WHO 2014), identifying culturally relevant and appropriate avenues to decrease loneliness and sources of suicidality among divorced men is imperative. The strength of association between suicide ideation and divorce, as with suicide ideation and loneliness, attenuated with the inclusion of meaning in life. Secondary analysis showed that meaning in life mediated the association between divorce and suicide ideation, as it did between loneliness and suicide ideation. It is clear that these two constructs play an integral role in how conflict within intimate partnerships influences suicide ideation. Prevention work should consider the role that churches and support groups in the community or church could play to impact meaning in life and loneliness for those experiencing marital and family conflict.

More aggressive conflict within the context of romantic relationships significantly predicted higher odds of suicide ideation. This finding, as with the association between divorce and suicide ideation, was supported by recorded social histories from clinical records and given reasons to die provided by survey respondents who screened positive for potential suicide ideation. It is worth noting that aggression between partners was reported as two-directional, both in the conflict tactics scale and in the open-ended list provided by survey participants. Statistical modeling established loneliness and marital satisfaction as significant mediators of the association between aggressive conflict and suicide ideation. The association between aggressive conflict and suicide ideation further attenuated when meaning in life was included in the model, suggesting further mediation by meaning in life. Meaning in life and satisfaction with intimate relationships appear as important contributions to the mental health of young men. Neither construct—meaning in life nor marital satisfaction—has received much attention among researchers or policy developers in sub-Saharan Africa. One avenue for future research may address the question of whether marital satisfaction among men improves sexual fidelity and decreases aggressive conflicts within romantic partnerships. Cross-sectional data, as those analyzed in the present study, does not afford opportunity to investigate sequence of events. We have suggested a causal framework that may not be accurate and which requires further investigation to validate or invalidate. That framework, positing aggressive conflict between romantic partners as a source of loneliness and marital satisfaction, applies findings from previous research that romantic intimacy can be a source of belonging and satisfaction (Bradbury et al. 2000; Laurenceau et al. 2005).

The most important implication of the current study positions romantic relationships as important contributors to the mental well-being of young men in Kenya. While other health concerns have driven important research into sexual and romantic relationships in sub-Saharan Africa, current study data demonstrate the concurrent importance of secure, non-violent romantic relationships to young Kenyan men. Three different types of data, self-report in an open list format, social histories recorded in clinical records, and standardized measures in a community-based survey, indicate the will to live and die among young Kenyan men is influenced by the nature of their intimate partnerships. While most research has investigated the pernicious effects of intimate partner violence to women’s health, this study points to pernicious effects on men’s health as well. Research should clarify how aggressive conflict predicts poor mental health among men, and why aggressive behaviors persist when they are associated with poor mental and social health. Interventions to involve men in intimate relationship and family strengthening programs should consider and promote potential benefits to men, and identify whether suicide ideation can be reduced through the promotion of non-aggressive conflict resolution strategies.

Limitations

The current study faces multiple limitations. The reliance on multiple types of data is a strength in the present study, though each data type has unique limitations. Social history for suicide cases in the hospital was not recorded for each patient. It is possible that clinicians recorded only histories that concurred with established perspectives on social causes of suicide behavior. If this occurred, it would lead to over-representation of marital conflict in social factors associated with suicide behaviors appearing at the hospital. Though the community survey of men 18–34 years yielded a large sample, 31% of households approached chose not to participate in the survey. This may have been due to perceived stigma (i.e. family- or community-level) or internalized stigma and shame surrounding suicide ideation and suicide behavior. We may have missed valuable insights from individuals who had experienced suicide ideation; however, we would hypothesize that their responses would largely support our bivariate and multivariate findings. There was a large percentage of respondents who provided no comment or reason to the open-list response question of reasons to live and reasons to die. We may therefore have under- or over-estimated the importance of various reasons to live or die among young men in this population. The source of the open-list responses indicating reasons to live and die is the same as the source of data for the closed-ended portion of the survey. Selection bias affecting one type of data would affect the other type of data. Respondents were young men who could be found by their respective homes on the interview day. Over 50% of the population indicated farming was their primary source of income. While this is expected among young men in a mostly rural environment, those who have office jobs were more likely to be excluded from the study. It is not clear what association exists between marital status, conflict tactics and primary income source, nor how these might ultimately influence suicide ideation. A secondary analysis showed no meaningful association between income source and divorce status, nor between income source and conflict tactics. It is likely that income- and employment-based stress compounds stress incurred in marital relationships, potentially increasing suicide ideation (Turner et al. 1995). Secondary, but important, findings indicate that loneliness, relationship satisfaction and meaning in life may mediate the association between intimate partner discord and suicide ideation. While what we have interpreted as mediation may actually be confounding between these variables (MacKinnon et al. 2000), we ventured to posit testable hypotheses in the absence of better data from this context. The posited pathways are consistent with more established research from high-income settings, supporting the plausibility of suggested pathways.

Future research should investigate the social and legal patterns of marriage more consistent with patterns identified here. For example, such research may ask whether legally binding marriages provide greater protection against future divorce and loneliness than informal marriages or cohabitation without formal marriage (Hattori and Dodoo 2007). Additional research may clarify how various stressors interact with marital status to yield loneliness, reduce satisfaction, and predispose towards suicide ideation and behavior. As research on suicide ideation is limited in sub-Saharan Africa, we endeavored to identify factors associated with most severe suicide ideation and modeled the most severe category of suicide ideation against mild or no suicide ideation. While this provided more conservative results than comparing severe to no suicide ideation, it may have masked important differences between mild and no suicide ideation. Future research should investigate whether patterns identified here are consistent across the spectrum of suicide ideation intensity, duration and variability. Other factors beyond intimate conflict emerged from both the chart reviews in study 1 and the open-response question in study 2 and should be considered by future investigation—e.g. the interaction of child burden and income limitations, and spirituality as a potential resilience factor preventing suicide ideation.

Conclusions

In this study, secure romantic attachments with less violent conflict appear as a resilience factor against engaging in suicide ideation among males in Meru County, Kenya. While intimate relationships have been investigated in the context of intimate partner violence, child marriage and HIV transmission, very limited research has focused on intimate relationships in sub-Saharan Africa. This study shows strong associations between intimate partner factors and suicide ideation. Men who are currently divorced have odds of engaging in suicide ideation that are 4 times higher than men are not currently divorced. Men who report more aggressive conflict tactics with their romantic partners are significantly more likely to engage in suicide ideation. Loneliness, loss of meaning in life and lower marital satisfaction appear to explain some of these associations. While further research into romantic attachment, conflict resolution and dissolution is required, these findings suggest intimate partnerships provide a source of companionship, satisfaction and meaning to young men. For their own sake, as well as for their romantic counterparts, further research should investigate how to strengthen men’s participation in romantic partnerships and provide them with social support during relationship transitions like divorce.