Prevalence of suicidal behaviour among nursing and midwifery college students in Ghana

Abstract Aim To provide exploratory and descriptive evidence on the prevalence estimate and some demographic correlates of suicidal behaviour among nursing and midwifery college students in Ghana. Design We used a cross‐sectional survey design. Method An anonymous survey involving a randomly selected sample of 305 nursing and midwifery college students was conducted in March–May 2017. The Suicide Behavior Questionnaire‐Revised was used to assess suicidal behaviour (i.e., ideation, planning, threat and attempt) and suicidal behaviour risk. Results The lifetime prevalence of suicide ideations (15.4%; 95% confidence interval [CI] = 0.11–0.20), plans (6.6%; 95% CI = 0.04–0.10), attempts (2.3%; 95% CI = 0.01–0.05), threats (13.4%; 95% CI = 0.10–0.18) and 12‐month prevalence of ideations (21.3%; 95% CI = 0.17–0.26) are comparable to estimates found in both high‐income and low‐ and middle‐income countries. However, associations between the demographic variables studied and suicidal behaviour risk were not statistically significant.

official database or national formal statistics on the prevalence, trends and characteristics of the phenomenon in the country (Osafo, Akotia, Andoh-Arthur, & Quarshie, 2015). Although some previous studies from Ghana (Eshun, 2003;Osafo, Hjelmeland, Akotia, & Knizek, 2011) have explored the socio-cultural determinants of suicidal ideation and the attitudes of university students towards suicide, generally, primary research examining the prevalence, correlates, risks and protective factors related to suicidal behaviour among adolescents and youth in Ghana remains scant (Asante, Kugbey, Osafo, Quarshie, & Sarfo, 2017;Quarshie et al., 2015). More pointedly, primary studies with focus on the phenomenon in college nursing and midwifery students in Ghana are unavailable, although there are local media reports of suicidality among this category of students (Frimpong, 2017).
Globally, recent studies have widely reported mental disorders and suicidal behaviours among university and college students in general, with suicidal ideations and other related mental health challenges being more frequent in females and adolescents (Amanya, Nakitende, & Ngabirano, 2018;Auerbach et al., 2016;Eskin et al., 2016;Han et al., 2016;Li et al., 2014;Mortier et al., 2015Mortier et al., ,2017Okwaraji & En, 2014;Yang, Zhang, Sun, Sun, & Ye, 2015). However, studies with focus on estimating the prevalence and determining the socio-demographic factors that present as correlates, risks or protective factors related to suicidal behaviours specifically among nursing (and midwifery) students are generally inadequate. Across the continent of Africa, there is a lack of published studies on the prevalence of suicidal behaviours among college students (Cipriano, Cella, & Cotrufo, 2017;Eshun, 2003;Franklin et al., 2017;. Therefore, several previous studies conducted mainly in high-income countries and a few studies among students from Africa have been reviewed to contextualize the present study. Eshun (2003)

compared samples of college students in Ghana and
America on the socio-cultural determinants of self-reported suicidal ideation. The findings showed among other things that no statistically significant relationship exists between religiosity and suicidal ideation in both samples, even though female gender showed a statistically significant association with suicidal ideation in the Ghanaian sample (Eshun, 2003). In South Africa, Van Niekerk, Scribante, and Raubenheimer (2012) surveyed 874 medical students from three universities. In terms of lifetime prevalence estimates, 32.3% of the students reported suicidal ideation, while 6.9% reported suicidal attempt. Van Niekerk et al. (2012) found no statistically significant evidence in terms of the association between suicidal attempt and age, gender, relationship status, place of study or year of study. In a cross-sectional survey of 1,130 Portuguese nursing students, Leal and Santos (2016) found a lifetime prevalence of suicidal behaviour of 5.22% (female = 5.43%; male = 4.21%), with no significant statistical difference across the years of study. Similarly, Aradilla-Herrero, Tomás-Sábado, and Gómez-Benito (2014) observed an overall suicide risk of 14% among nursing students in Spain. Of the 93 students surveyed, 14% had thought about suicide at some point in their lives and 6.5% had previously made a suicide attempt, with a statistically significant negative association observed between self-esteem and suicide risk (Aradilla-Herrero et al., 2014). Among nursing students in Greece, Melissa-Halikiopoulou, Tsiga, Khachatryan, and Papazisis (2011) reported a lifetime suicide ideation prevalence estimate of 10.6% (females = 8%; males = 25%), with 1.7% female participants expressing the desire to actually die by suicide if they had the chance. Also, a statistically significant positive association between suicidal ideations and depression was found (Melissa-Halikiopoulou et al., 2011). In a longitudinal study of the predictors of suicidal ideation among college students, Nam, Hilimire, Jahn, Lehmann, and DeVylder (2018) reported that age, gender and sexual orientation showed no statistically significant association with suicidal ideation, whereas race/ethnicity showed significant association with suicidal ideation intensity. Also, truancy has been found as a strong correlate of suicidal behaviour among students (Asante et al., 2017).

More recently, in the WHO World Mental Health International
College Student Project, cross-sectional web-based self-report questionnaires were administered to 13,984 first-year students across 19 colleges in eight countries: Australia, Belgium, Germany, Mexico, Northern Ireland, South Africa, Spain and the United States (Mortier, Auerbach, Alonso, Bantjes et al., 2018). In terms of lifetime prevalence, 32.7% of the students reported suicidal ideation, 17.5% reported suicidal plans and 4.3% reported suicide attempts. About 12-month prevalence, 17.2% of the students reported suicidal ideation, 8.8% reported suicidal plans and 1.0% reported suicide attempts. The study further showed non-heterosexual orientation and heterosexual orientation with some same-sex attraction as the strongest correlates of suicidal thoughts and behaviours across the overall sample (Mortier, Auerbach, Alonso, Bantjes et al., 2018).
Thus far in Ghana, only three published primary studies are available on the prevalence and correlates of suicidal behaviour among senior high school (SHS) students (Asante et al., 2017;Baiden et al., 2019;Liu, Huang, & Liu, 2018), with no published study on the phenomenon among (nursing and midwifery) college students, although media reports are showing worrying trends of suicidal deaths among students across the country . A college in Ghana is an immediate post-senior high school level/stage of education. Hence, a potential contribution of the present study is the provision of a comparative evidence to help assess the trends of the prevalence of suicidal behaviour as young people move from SHS to college.
The present study is thus born out of the need for systematic evidence on the prevalence and (demographic) correlates of suicidal behaviour among college trainee nurses and midwives in Ghana for a pioneering understanding of the extent of the problem among this population and to potentially inform intervention and prevention efforts and programmes in the country. Specifically, this study seeks to provide exploratory, descriptive evidence on the prevalence estimates of suicidal behaviour (ideations, planning, threats and attempts) and the associations between some socio-demographic factors and suicidal behaviour risk among nursing and midwifery college students in Ghana.

| Structure of nursing education in Ghana
Ghana is a West African country with a nurse-population ratio of 1:542 (Ghana Health Service, 2017). Being the first to achieve independence from colonial rule, Ghana led the development of nurses' education, with the establishment of the first universitybased diploma programme to train nurses in tropical Africa (Opare & Mill, 2000). Detailed historical facts and contemporary development of nursing and midwifery training in Ghana have been provided elsewhere (see Adu-Gyamfi & Brenya, 2016;Bell, Rominski, Bam, Donkor, & Lori, 2013;Bvumbwe & Mtshali, 2018;Opare & Mill, 2000;Talley, 2006). The nursing and midwifery professions remain female-dominated in Ghana and as such fewer males enrol as trainees (Adu-Gyamfi & Brenya, 2016;Tagoe & Quarshie, 2017;Talley, 2006). Although some universities in Ghana offer nursing and midwifery training programmes, most nurses and midwives in the country are educated in training colleges often with a teaching hospital affiliation (Bell et al., 2013). Currently, four basic nursing and midwifery programmes (Registered General Nursing, Registered midwifery, Registered Community Health Nursing and Registered Mental Health Nursing) are offered by universities and training colleges across the country (Nursing & Midwifery Council of Ghana -NMCG, 2018). Annually, about 400 persons graduate from nursing and midwifery training institutions in Ghana (Darko, 2015).

| Theoretical framework
The theory guiding this study is the risk-protective factor model (Monahan, Oesterle, Rhew, & Hawkins, 2014). Risk and protective factors play a critical role in suicide prevention and identifying risk, and protective factors provide critical information to assess and manage suicide risk in individuals. Socio-demographic factors such as nonheterosexual orientation, female gender, low socio-economic status, first year of college, marital status, family relationships, adolescence, among others, have been found to be associated with increased  Ten response options were provided under ethnic groups; however, each of six options had frequencies <5, hence were grouped into an "other" category. b A disproportionately higher number of the participants self-identified as heterosexual, compared to lesbian, gay, bisexual and transgender orientations, which were collapsed into "non-heterosexual." c Truancy refers to number of days absent from college without permission during the past 12 months.
the associations between some socio-demographic factors (i.e., age, gender, religious groups, ethnic groups, romantic relationship status, sexual orientation, marital status, year of study, programme of study, school residential status and truancy) and suicidal behaviour risk among nursing and midwifery college students in Ghana.

| DE S I G N AND ME THODS
The general methodological approach to this study was descriptive in nature. Specifically, we used a cross-sectional (prevalence) survey design (Lachat et al., 2016)

| Measures
The questionnaire used for this study was in English and consisted of two parts. The first part was made up of 11 items on socio-demographic characteristics of the participants: age, gender, religious group, ethnic group, romantic relationship status, sexual orientation, marital status, years of study, programme of study, school residential status and truancy (see Table 1).  (Aloba, Ojeleye, & Aloba, 2017). The SBQ-R yields a total score between 3-18 points.
For college students and undergraduates, a total score of 7 out of 18 on the SBQ-R is the cut-off (≥7) which indicates high risk of suicidal behaviour, whereas a total score between 3-6 indicates low risk of suicidal behaviour (Osman et al., 2001).

| Data analysis
The answered questionnaires were checked for completeness, and data were entered into the Statistical Package for Social Sciences (SPSS version 21.0) for analysis. The list-wise deletion of missing data approach was used, as the loss of cases due to missing data was <5% (Graham, 2009). Specifically, frequencies and cross-tabulations were used to ascertain the prevalence estimates (and their overall 95% confidence intervals) of suicidal behaviours across the sociodemographic characteristics of the participants. For further analysis, the socio-demographic variables (shown in Table 1) were included as the correlates/explanatory variables, while the total score on the SBQ-R, suicidal behaviour risk, was included as the outcome variable.
The chi-square (χ 2 ) test was used to examine possible relationships between the specified categorical explanatory variables and suicidal behaviour risk; Fisher's exact test was used where a cell expected frequency was less than five counts (Agresti, 2002;Kim, 2017). Pointbiserial correlation tests were performed to assess the associations between the continuous explanatory variables and suicidal behaviour risk (Prematunga, 2012). Given the general categorical nature of the data and the binary response format of the outcome variable (i.e., suicidal behaviour risk, scored low or high), binary logistic regression (Agresti, 2002) was used to ascertain possible associations between the correlates and the outcome variable. The candidate explanatory variables (i.e., socio-demographic variables) were entered in the multivariable logistic regression model whether or not they had a statistically significant bivariate (χ 2 ) relationship with the outcome variable (Babyak, 2004;Peduzzi, Concato, Kemper, Holford, & Feinstein, 1996). Results of the logistic regression were reported as odds ratios with 95% confidence intervals. Statistically significant results were determined using the p-value <0.05 (p < 0.05).

| Ethical statement
Prior to the administration of the questionnaire, all the students in each randomly selected class were informed about the study and their questions and concerns were addressed by the researchers.

| RE SULT
The results of the data analysis are presented in relation to the prevalence estimates of suicidal behaviour and possible demographic correlates of suicidal behaviour risk. Relative to the Christian participants, the Muslims reported lower scores on suicidal ideations and plans (with no lifetime attempted suicide), but higher on suicidal behaviour risk (Muslim = 12.0%; Christian = 8.4%). Also, although more married students reported lifetime suicide attempts (9.1%) compared students who self-identified as single (2.1%), in terms of suicidal behaviour risk, both groups were fairly similar, while students of other marital status scored the highest (33.3%) on suicidal behaviour risk. Students with non-heterosexual orientation, married students and students of registered community health nursing (RCHN) all reported no likelihood of future suicidal behaviour. Students of RGN reported high suicidal behaviour risk (10.0%), compared with the midwifery (7.2%) and RCHN students (6.5%). The risk of suicidal behaviour was relatively similar across the three levels of study (1st year = 7.3%, 2nd year = 9.9% and 3rd year = 7.8%), although the 2nd-year students were at a slightly higher risk. Similarly, students who identified as not in a romantic relationship were identified to be at a slightly higher risk of suicidal behaviour (5.0%) relative to those students who responded "yes"

| Prevalence estimates of suicidal behaviours
(3.2%) to being in a romantic relationship at the time of the study.

| Socio-demographic correlates of suicidal behaviour risk
The statistical test for possible univariate relationships (chi-square test) between the socio-demographic variables and suicidal behaviour risk found no statistically significant results (Table 2). A pointbiserial correlation was run to determine the relationship between age and suicidal behaviour risk; there was a statistically non-significant positive correlation between age and suicidal behaviour risk (r pb = 0.063, N = 305, p = 0.271).
In Table 2, several of the socio-demographic variables showed sparse data bias (Greenland, Mansournia, & Altman, 2016) in terms of suicidal behaviour risk, mostly less than five cases in the high category. Thus, to avoid an unstable logistic regression model due to the sparse data problem, we applied the relaxed rule of at least five events per variable (Vittinghoff & McCulloch, 2007) to select the socio-demographic variables included as correlates in the final logistic regression model. As shown in Table 3, the final logistic regression model was not statistically significant (χ 2 (df = 4) = 1.774, p = 0.777), accounting for 1.3% of the variance in the outcome TA B L E 3 Logistic regression assessing the association between sociodemographic variables and suicidal behaviour risk variable (Nagelkerke R 2 = 0.013). In other words, the model did not show any statistically significant distinction between participants who scored low and high on suicidal behaviour risk. Thus, the prevalence of suicidal behaviour, differences and associations observed in this study is descriptive in nature rather than statistical.

| D ISCUSS I ON
To our knowledge, this is the first study from Ghana that provides estimates of the prevalence of suicidal behaviour (ideations, plans, threats and attempts) and tries to document potential socio-demographic correlates of the behaviour among nursing and midwifery college students.

| Prevalence of suicidal behaviour
The main finding of this study is that 8.5% (female = 8.7%; male = 7.1%) of the 305 nursing and midwifery college students surveyed met the cut-off criterion for the categorization as at high risk of suicidal behaviour. The overall lifetime prevalence estimates of suicide ideation (15.4%; female = 16.2%; male = 7.1%), plans (6.6%; female = 6.9%; male = 3.6%), threats (13.4%; female = 14.1%; male = 7.1%) and attempts (2.3%; female = 2.2%; male = 3.6%) and the 12-month prevalence estimate of suicide ideation (21.3%; female = 21.7%; male = 17.9%) reported in this study are generally comparable to those found in high-income countries, with more females than males reporting suicidal behaviour.. The 12-month prevalence estimate of suicide ideation found in this study (21.3%) is comparable and quite consistent with the recent prevalence estimates among senior high school students in Ghana-18.2% (Asante et al., 2017;Baiden et al., 2019). More pointedly, the findings on the prevalence estimates of suicidal behaviour in this study are consistent with the global situation as found in the meta-analysis by . Put together, the prevalence estimates of suicidal behaviour and risk found in this study resonate with recent calls and campaigns from Ghana (Andoh-Arthur, Asante, & Osafo, 2015) and elsewhere (Mortali & Moutier, 2018;Pace, Silk, Nazione, Fournier, & Collins-Eaglin, 2018) for the improvement of on-site counselling services to facilitate mental health help-seeking by students on college campuses. To this end, emphasis has been placed on mental health literacy and stigma reduction, as doing this has the potential of encouraging students, particularly, males, to seek professional help (Rafal, Gatto, & DeBate, 2018).
It is evident in this study that more females than males reported both lifetime and 12-month suicidal ideations, even though the suicidal behaviour risk between females and males is comparable (Table 2). This observation is consistent with previous studies in the area from high-income countries (Aradilla-Herrero et al., 2014;Leal & Santos, 2016) and among young people generally (Asante et al., 2017;Hawton et al., 2012;O'Connor et al., 2018), but contradicts the finding by Melissa-Halikiopoulou et al. (2011) that lifetime suicidal ideations are higher among males than female nursing students in Greece.
Also, the students who reported as being in a romantic relationship scored lower on the suicidal behaviour risk, relative to those who reported as not in a romantic relationship. One potential explanation for this is that being in a relationship strengthens individuals' sense of belongingness and could be a protective factor against suicidal behaviour. Additionally, meaningful social relationships represent an important avenue for support, which could help individuals deal with distress (Rhoades, Kamp Dush, Atkins, Stanley, & Markman, 2011).
Generally, although the evidence presented in this study does not provide enough basis to attribute the reported prevalence estimates of suicidal behaviour to the family circumstances or even psycho-

| Demographic correlates of suicidal behaviour
This study also attempted to describe some of the possible demographic correlates of suicidal behaviour among college nursing and midwifery students. However, the analysis showed no statistically significant associations, even though some odds ratios greater one (OR > 1) were obtained. The statistically insignificant results obtained in relation to the chi-square tests and logistic regression analysis in this study may be attributable to three factors. The first plausible reason has to do with guarded responses or non-disclosure mainly due to the general negative attitudes towards suicide and suicidal behaviour in Ghana. Attempted suicide is a crime in Ghana (Mishara & Weisstub, 2016); it is religiously sinful , and morally, it is even tabooed to openly talk about suicide and suicidal behaviours (Sarpong, 2006). There is evidence to suggest that people found guilty of this code have been jailed or given hefty fines (Adinkrah, 2013). Knowledge of this law and social mores might have discouraged the participants from honestly reporting their histories of suicidal behaviours.
The second reason could be the possibility that as trainee healthcare professionals, the participants might be aware of the medical danger associated with attempting suicide, even though having suicidal ideations might seem less threatening. Previous studies have shown that among the pathways to suicidal death, the transition from suicidal ideations to death by suicide is less strong, compared with the transition from suicide attempts to death by suicide (Van Heeringen, Hawton, & Williams, 2000). Thus, our participants' possible awareness of the strong connection between suicide attempt and death might account for the low estimates of suicide attempts in this study.
The third plausible reason is related to the sample size of this study. A cursory visual inspection of Table 2 shows frequent occurrence of zero cells and a relatively wide spread of sparse data.
Future studies may consider obtaining a relatively larger sample of participants to partly avoid the problem of sparse data.

| Implications
This study has important implications for healthcare professionals and administration of nursing and midwifery training colleges (in Ghana) and research. An awareness of the suicidal behaviour risk and prevalence of suicidal behaviours among students should underscore the need for periodic screening of college students and the strengthening of on-site student mental health and counselling services. Suicidal ideation should be a major concern for (mental) healthcare providers serving students and trainers of these students in colleges. In terms of research, the findings of this study could serve a useful beginning point for more detailed and broader research on suicidal behaviours among nursing and college students in Ghana.

| Limitations
The findings of this research should not be taken without regard to the plausible limitations related to the study. First, the prevalence estimates may be an underestimation of the true extent of the problem among college students in Ghana. Previous studies have found nondisclosure of suicidal behaviours among college students (De Luca, Yan, Lytle, & Brownson, 2014). This tendency of non-disclosure, coupled with the highly stigmatized nature of suicidal behaviour in Ghana , might have precluded our participants from fully reporting their history of suicidal behaviour, thereby providing socially desirable responses. Also, given that the data are cross-sectional, causation of suicidal behaviours cannot be inferred. Additionally, the data for this study were collected from nursing and midwifery students from one training college, which limits the generalizability of the findings; hence, future studies can broaden the sample to include trainees from other nursing programmes and training colleges in the country.
Despite these limitations, this study contributes pioneering descriptive evidence to our knowledge about the prevalence estimates of suicidal behaviour among nursing and midwifery college students in Ghana.

| CON CLUS ION
This study highlights the reality of suicidal behaviours among nursing and midwifery college students in Ghana and underscores the need to sensitize all stakeholders to promote mental health intervention efforts focusing on this group of students to improve early detection and treatment and support. The findings may also point to developing strategies and initiatives for suicide prevention on the campuses of nursing and midwifery training institutions in Ghana to building support networks and increasing awareness about the signs of suicidal behaviours.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.