The age of opportunity: prevalence of key risk factors among adolescents 10–19 years of age in nine communities in sub‐Saharan Africa

To measure health‐related behaviours and risk factors among sub‐Saharan African adolescents.


Introduction
The relative importance of mortality and morbidity among adolescents is increasing as the burden of disease among young children has fallen rapidly over the past two decades. The risk of mortality is greatest among young children, and significant research has focused on improving child survival. However, adolescents have been considered a relatively healthy population and there is scarce evidence on adolescent health until recently. Information on adolescent morbidity and mortality has been particularly lacking in low-and middle-income countries (LMICs), including in sub-Saharan Africa [1].
Evidence on adolescent health depicts a complex picture of diverse health needs that vary greatly by region.
There are approximately 1.2 billion adolescents 10-19 years of age globally, with roughly 90% in LMICs [2]. Over 1.2 million adolescents die each year. The top causes of death for male and female adolescents globally include respiratory infections, diarrhoeal diseases, road injury, interpersonal violence, self-harm, drowning and meningitis [2]. The leading cause of death for girls 15-19 years of age globally is complications from pregnancy and childbirth [2]. Iron-deficiency anaemia is the leading cause of global adolescent disability-adjusted life years (DALYs), followed by road injury, depressive disorders, lower respiratory infections and diarrhoeal diseases. Poor diet is among the most important risk factors for disease and mortality for male and female adolescents globally. Other important risk factors include water, sanitation and hygiene, alcohol and substance use, air pollution and unsafe sex [2]. Nineteen per cent of the world's adolescents live in African LMICs where 45% of adolescent deaths and 35% of DALYs occur [2]. In sub-Saharan Africa, the greatest share of the population is adolescents (23%) and the number of adolescents is projected to continue increasing. Half of the population of 15 sub-Saharan African countries is under age 18 [3]. The leading causes of death and DALYs in African LMICs are lower respiratory infections, diarrhoeal diseases, meningitis, HIV/AIDS, road injury and iron-deficiency anaemia [2].
A time of profound change for physical, cognitive and social development, adolescence marks a critical period of growth in the life course. Adolescent health is impacted by childhood well-being and sets a trajectory for maternal and adult health status [4]. For example, adolescents have unique nutritional needs due to increased growth and development. Improving and correcting nutritional deficiencies persisting from childhood may help with catch-up growth during this critical period [5]. Health behaviours adopted in adolescence have implications that can persist throughout the life course, including diet and physical activity, alcohol and substance use, and sexual behaviour. Physical activity and eating habits are formed or solidified in adolescence. Risk for non-communicable diseases such as diabetes, heart disease and certain cancers increases with poor diet and limited physical activity. 15% of adolescents in Africa are overweight or obese with increased risk for diabetes and other related noncommunicable diseases later in life [6]. The health status of girls and women when they enter pregnancy also effects maternal health and birth outcomes [7]. Therefore, health risk and health-seeking behaviours during adolescence have important implications for women's reproductive health later in life.
Most data on adolescent health come from school-based settings, however, which may underestimate morbidity and mortality at population levels because adolescents who are not in school may be at higher risk for certain health behaviours and outcomes. Literature on adolescents in sub-Saharan Africa also tends to focus on HIV-positive populations. There is a significant gap in public health research among adolescent populations [8]. We conducted community-based surveys of adolescents in nine sub-Saharan African communities. These exploratory risk factor data are meant to inform the development of community-tailored interventions to address adolescent health needs.

Setting and study population
The African Research, Implementation Science, and Education (ARISE) Network Adolescent Health Study was conducted in nine communities selected by the ARISE Network in seven sub-Saharan African countries (Burkina Faso, Ethiopia, Ghana, Nigeria, Eswatini, Tanzania and Uganda) based on feasibility of study implementation. The full methods are described in detail elsewhere [9]. Briefly, at all communities, male and female residents aged 10-19 years were recruited for the study through household-based sampling. In Burkina Faso, two communities in Ethiopia, two communities in Tanzania, and Uganda, the study was nested within existing health and demographic surveillance systems (HDSSs), allowing for potential future long-term follow-up.
Community-representative samples of adolescents in each of the study communities were selected in the following way: households with at least one adolescent resident were randomly selected from sampling frames derived from HDSSs or other recent census records. Person-to-person interviews were carried out at various times between July 2015 and December 2017 by trained research assistants with prior data collection experience and knowledge of the local language. Both male and female research assistants collected data using a standardised questionnaire translated into the local language. The questionnaire was largely based on the WHO Global School-based Student Health Survey (GSHS) [10] with additional sections added from other validated instruments. Pilot testing was conducted at the Ethiopian sites.

Ethics statement
Written informed consent was obtained from all adolescents aged 18 and 19 years. Written parental consent and adolescent assent were obtained from adolescents younger than 18

Variables
From the survey data, several dichotomous (yes/no) indicators of adolescent health were selected within the following thematic domains: mental and behavioural health, violence and injury, sexual and reproductive health, nutrition and physical activity, and health services utilisation.
An indicator was constructed for low mood 'much of the time', 'most of the time' or 'all of the time' during the past week on a 5-point scale that also included 'hardly ever' and 'sometimes'. Similarly, an indicator was constructed for anxiousness 'much of the time', 'most of the time' or 'all of the time' during the past week. An indicator for suicidal ideation was created from responses to the question 'During the past 12 months, did you ever seriously consider attempting suicide?' Suicide attempts were defined as responses of one or more to the question 'During the past 12 months, how many times did you actually attempt suicide?' Violence and injury indicators included having been bullied one or more times within the past 30 days, having been in a physical fight during the past 12 months, having been physically attacked during the past 12 months and having been seriously injured during the past 12 months. Bullying was described in the questionnaire as when a person or group of people say or do bad and unpleasant things to another person, tease them a lot in an unpleasant way or leave them out of things on purpose. Physical fighting was described as two people of about the same strength or power choosing to fight each other. Physical attacks were described as when one or more people hit or strike someone, or when one or more people hurt another person with a weapon (such as a stick, knife or gun). Serious injuries were described as those that make one miss at least one full day of usual activities (such as school, sports or a job) or require treatment by a doctor or nurse.
Participants were asked several questions concerning reproductive health. Among those aged 15-19 that reported ever having sex, indicators were constructed for ever having been pregnant (females) or made someone pregnant (males), unprotected last sex (reporting not having used a condom or other contraceptive method including pill, injection, withdrawal or others at last sex), initiation of sexual activity before age 15 and having had a sexually transmitted infection (STI).
Reports of fruit or vegetable consumption less than once per day were considered indicators for inadequate intake of these foods in accordance with previous studies of adolescent diet in the region [11,12]. An indicator for frequency of carbonated soft drink consumption one or more times per day during the past 30 days was generated from responses to this questionnaire item. A 'low physical activity' indicator was defined as not having been physically active for at least 1 h per day during the past seven days. Reporting more than 4 h per day sitting and watching television, playing computer games, talking with friends or doing other sitting activities on a usual day was used as an indicator for sedentary behaviour. Substance use indicators included alcohol use during the past 30 days, ever use of tobacco, ever use of illicit drugs and ever use of khat (Ethiopia and Uganda communities).
Height and weight were measured using digital scales and stadiometry by trained field staff, except in Dar es Salaam where self-reported height and weight were collected. Height and weight were not collected in Uganda. Body mass index (BMI) was calculated as weight in kilograms divided by height in metres squared and rounded to one decimal place. BMI and height z-scores were calculated using the WHO standard for children aged 5-19 [13]. Following WHO recommendations for children and adolescents, underweight was defined as a BMI z-score for age and sex of À2 or lower. Overweight was defined as a BMI z-score for age and sex equal to or greater than 1 and below 2. Obesity was defined as a BMI z-score for age and Tropical Medicine and International Health volume 25 no 1 pp 15-32 january 2020 sex equal to or greater than 2. Stunting was defined as a height z-score for age and sex of À2 or lower. The WHO standard was chosen for the assessment of undernutrition and overweight/obesity to enable comparisons with the wide range of studies in which it has been used previously.

Statistical analysis
Weighted prevalence estimates and 95% confidence intervals were calculated for each indicator within four strata of age (10-14 and 15-19) and sex (male and female) using the PROC SURVEYFREQ command in SAS -version 9.4. Post-stratification survey weights were applied where possible to ensure that the estimates were representative of the geographic clusters from which the samples were drawn. These geographic clusters corresponded to kebeles (Ethiopia), villages (Burkina Faso, Dodoma and Uganda), streets (Dar es Salaam) and census-defined enumeration areas (Ningo Prampram). Survey weights were equal to the population proportion of each geographic cluster divided by the sample proportion of each geographic cluster. Sampling weights of one were utilised for Nigeria and Eswatini, because sampling at the Nigerian community was proportional to cluster size and because weighting information was unavailable for Eswatini. All prevalence estimates were then pooled across communities through random-effects meta-analyses in Stata version 14.

Results
Overall, 8075 adolescents participated in the study; 50.63% (4088) were female. Four communities-Dodoma (Tanzania), Kersa (Ethiopia), Nouna (Burkina Faso) and Iganga/Mayuge (Uganda)-were predominantly rural, while the others were urban centres. Demographic characteristics of participants in each community are reported in Table 1. School enrolment was lowest in two rural communities, Kersa and Nouna. Harar in Ethiopia and Ningo Prampram in Ghana, both urban centres, had the highest school enrolment. The proportion of adolescents living with both parents was highest in Kersa and Nouna; the lowest was in Ningo Prampram. Adolescents' employment level in the last 12 months was highest in Nouna. The proportion of adolescents living in households with five or more members was highest in Nouna and lowest in Harar. Prevalence of all mental health indicators ( Figure 1) was low. Older females (15-19 years of age) had the highest prevalence for all indicators ranging from 3% (95% CI 1.0, 6.0) for suicide attempts to 11% (95% CI 7.0, 16.0) for low mood much, most or all of the time. The prevalence of low mood was generally higher among older adolescents (15)(16)(17)(18)(19) years) compared to younger adolescents (10-14 years) and females compared to males. Harar had the lowest prevalence of low mood for younger adolescents and older male adolescents, and Lubombo, Eswatini, had the highest prevalence across all age and sex strata. We find that male and female adolescents in Lubombo, Eswatini, and Nouna, Burkina Faso, report relatively high rates of low mood most or all of the time, while female adolescents in Ibadan, Nigeria, have relatively high rates of anxious feelings and suicidal ideation, compared to other communities.
The pooled prevalence of violence and injury indicators among all adolescents ( Figure 2) ranged from low to moderate. Bullying was moderate (17-23%) in all study communities, with the highest rate among younger male adolescents in Ghana (36.1%, 95% CI 28.3, 43.9) and the lowest rate among younger females in Kersa (6.4%, 95% CI 4.1, 10.0). Prevalence of physical fights in the past 12 months was also moderate (12-35%), with higher rates among males than females in all communities and the highest rates among younger male adolescents in all communities except Uganda. Among females, prevalence of physical fighting is also higher among the younger age group in all communities except Kersa and Uganda. Prevalence of physical attack and serious injury in the past 12 months was lower than bullying and physical fight. The highest rates of physical attack and serious injury were observed in Dodoma and Ibadan and the lowest rates in Harar (Figure 2). Prevalence values and 95% confidence intervals disaggregated by age categories and sex for all violence and injury indicators are shown in Appendix Table A2.
The pooled prevalence of sexual and reproductive health indicators was low to moderate. A weighted total of 999 (25.9%) adolescents 15-19 years of age reported ever having sex. In all communities, the proportion of sexually active adolescents increased with age ( Figure 3). Among sexually active 15-to 19-year-olds, 37% of females reported ever being pregnant and 8% of males reported to have made someone pregnant, with rates above 50% among females in Burkina Faso, Eswatini and Ghana (Figure 4). Overall, unprotected last sex was reported by 40% of sexually active males aged 15-19% and 46% of sexually active females aged 15-19. A high prevalence (>50%) of unprotected last sex was reported among females in Burkina Faso and both sexes in Kersa, Ghana and Dodoma. Age at first sex below 15 years of age was 28% among males and 21% among females. Self-reported STI was 3% among males and 1% among females ( Figure 4). Prevalence values and 95% confidence intervals disaggregated by age categories and sex for all reproductive health indicators are shown in Appendix Table A3.
We observe a high pooled prevalence of inadequate fruit intake (57-63%) and moderate prevalence of inadequate vegetable intake (38-44%) ( Figure 5). Rates of inadequate fruit intake were above 70% for all adolescents in Kersa, Ghana and Uganda, and for male adolescents in Harar. High rates above 50% of inadequate vegetable intake were observed among all adolescents in Ibadan, male adolescents in Kersa and Harar, and younger male adolescents in Burkina Faso, with a very high prevalence among all adolescents in Uganda (75.8-85%). The consumption of soft drinks at least once per day was moderate, reported by about one-quarter of adolescents. The highest prevalence of soft drink consumption was observed among older male and female adolescents in Dodoma (51.8% and 60.9%, respectively). Soft drink consumption was most common among older females in six of the nine communities ( Figure 5).
The pooled prevalence of less than 1 h of physical activity per day was high (82-90%), while prevalence of sedentary activity five or more hours per day was low to moderate (8-13%) ( Figure 6).
Low physical activity was greater than 90% for all age and sex categories in six of the nine communities. Moderate rates of sedentary activity (10-50%) were reported among all adolescents in Eswatini, Ghana, Ibadan and Dodoma. Prevalence values and 95% confidence intervals disaggregated by age categories and sex for all nutrition and physical activity indicators are shown in Appendix Table A4. Of six communities with BMI data, the prevalence of underweight exceeded the prevalence of overweight/obese among females in four communities and among males in five communities (Figure 7).
The prevalence of alcohol and substance use indicators ( Figure 8) was very low. The total weighted number of participants who reported any use of alcohol, drugs, tobacco or khat across all study communities was 782 (9.4%). While rates are generally low for both age groups and sexes, slightly higher rates were reported among older male adolescents for alcohol use in Ibadan and Harar; ever use of drugs in Eswatini; and ever use of cigarettes in Burkina Faso. However, khat use was common among older male adolescents in Kersa (75.6%) with moderate rates among younger male and older female adolescents in Kersa (13.6% and 41.7%) and older male adolescents in Harar (23.4%) (Figure 8). Prevalence values and 95% confidence intervals  Appendix Table A5.

Discussion
Overall, across communities, we find very low rates of alcohol and substance use (with the exception of khat use in Ethiopia) and self-reported STI. We also observed a low prevalence of mental health risk factors (with the exception of low mood among older females, which is moderate). A low prevalence was also observed for ever made someone pregnant among males; physical attack among older female adolescents and serious injury for female adolescents in both age categories; and more than 5 h of sedentary activity among younger male adolescents. Inadequate fruit intake and less than 1 h of physical activity were common. All other risk factors had a moderate prevalence between 10% and 50%. The study communities differ in many basic characteristics mainly reflecting on the type of residence and associated socioeconomic conditions. Male adolescents in our study communities reported fewer symptoms of depression and more risk factors for violence than females. Older adolescents and especially females are at highest risk for depression and mental health risk factors. However, overall, prevalence of mental health risk factors (low mood, anxious feelings, suicidal ideation and attempts) was low and there was significant variation between communities. A systematic review of mental health among children 0-16 years in sub-Saharan Africa found a prevalence of general psychopathology of 14.5%, generally higher than in our study populations, suggesting that mental health is an important concern for children and adolescents in this region [14].
Mental health disorders are generally given low priority due to a lack of information on the extent of mental health problems as well as associated stigma and perceptions that mental illnesses cannot be treated; thus, many people living with mental illness suffer in silence [15,16]. We find lower rates of suicidal ideation compared to findings from the WHO GSHS in Ghana, Eswatini, Tanzania and Uganda [10]. Overall, suicidal ideation was more common among older females in this study. A study among 10-to 19-yearolds in Uganda also found higher rates of suicidality among females and increasing with age [17].
Emotional disturbances and substance abuse have been correlated with bullying in African settings [18,19]. Prevalence of bullying and physical fighting in our study was high compared to mental health, especially among male adolescents. Young male adolescents reported higher rates of violence in most communities. We find lower rates of serious injury compared to WHO GSHS data from Ghana, Eswatini, Tanzania and Uganda as well as DHS data from Ethiopia, Nigeria, Tanzania and Uganda [10,[20][21][22][23]. However, a study among adolescents 13-15 years of age in school in six sub-Saharan African countries found an average of 68.2% of participants reporting one or more serious injuries in the past 12 months and males in lower age groups were also more likely to experience injury in this study [24], consistent with our findings.
Overall, rates of SRH risk factors are low to moderate in our study populations. However, it is likely that there is significant underreporting in all population groups [25]. Notably, those who report ever having been pregnant or having made someone pregnant vary from 3% to 62%, with higher rates among females compared to males (16-62% compared to 3-16%). In all communities except for Harar, males are more likely than females to report age at first sex less than 15 years, which could be related to differential underreporting [26]. A paper examining DHS data from Burkina Faso, Ethiopia and Nigeria found a prevalence of modern contractive use among sexually active adolescents between 7.8% and 24%, with much higher rates among non-married adolescents, generally lower than our findings. This paper also finds an increase in contraceptive use with higher education levels [27]. Our findings should be interpreted with caution given small numbers and likely underreporting. Future SRH research among adolescents should address best practices for collecting sensitive information among this age group.
We find a moderate to high risk of diet and physical activity risk factors. Relatively high rates of soft drink consumption were reported across age groups, genders, and rural and urban residents. This finding is consistent with increases in soft drink and other processed food consumption in LMICs [28] as well as increasing rates of non-communicable diseases in these countries. At the same time, inadequate fruit and vegetable consumption are relatively high, also consistent with regional trends. A systematic review of diets of adolescent females in LMICs also found inadequate fruit and vegetable consumption [29].
Prevalence of at least 1 h of physical activity per day is also low. There may be underreporting due to different interpretations of what constitutes physical activity. Low rates in our populations may reflect differences between in-school and out-of-school adolescents, where those inschool adolescents are more likely to engage in physical education. Physical activity interventions should therefore be tailored to urban and rural settings and should aim to educate adolescents on the recommended amount of physical activity per day and context-specific ways to reach this target.
The low prevalence of substance use likely indicates underreporting of these risk factors. The WHO 2018 Global Status Report on Alcohol and Health finds current use of alcohol among 21.4% of adolescents 15-17 years in the African region [30], compared to 1-6% in our study population. We find no use of cigarettes in our populations. Although underreporting of cigarette use is likely, we would expect a low prevalence among these age groups. A school-based study of 13-to 15-year-olds in West Africa found that 95% and 91% of adolescents in Ghana and Nigeria, respectively, had never smoked  tobacco. The exception in our study population is khat use in Kersa and Harar, Ethiopia, where older male and female adolescents in both communities report relatively high rates. Khat is grown as a cash crop in Ethiopia. Our findings are consistent with other literature from Ethiopia: a school-based survey of 15-to 25-year-olds in Harar, Ethiopia, found 24% of participants used khat, with higher odds for male gender and older age [31].  While our findings are generally consistent with existing literature, prevalence of risk factors varied widely, and there is limited evidence on adolescents from community-based settings. A strength of this research is the inclusion of both in-and out-of-school adolescents, but there are several limitations. First, this study used nonpopulation-representative samples. Results are representative of the individual communities where this study was conducted and may be valuable for similar communities in the region; however, they cannot be used to make generalisations about the broader population of adolescents in the seven countries represented. Second, while the survey was largely based on the extensively used GSHS, the tool was modified and was not piloted in all sites. Third, quantitative data alone may not provide detailed information on certain subjects and the use of qualitative methods is warranted. Finally, self-reported data are subject to participant recall and bias and underreporting of sensitive subjects, including SRH and substance use, is likely. The difficulty of collecting sensitive subject matter is a limitation of these data. There is a need to carry out comprehensive studies involving sufficiently large sample sizes, improved data collection methods and longitudinal follow-up to gain a better understanding of adolescents' health status and risk factors in the region.
Intervention studies are needed to test the effectiveness of various approaches towards addressing these risk factors. To target nutritional status and BMI associated with conditions of underweight and stunting as well as overweight and obesity, interventions should focus on improving dietary quality of adolescents and increasing physical activity. Such programs might include multiple micronutrient and macronutrient supplementation, improvement of school food environments, and nutrition and physical activity education. To target SRH outcomes, interventions focused on knowledge and awareness for both in-and out-of-school adolescents should be tailored to local settings and implemented for both younger and older population groups. Delaying pregnancy for adolescent females could help improve retention in school, pregnancy and birth outcomes, and empowerment of women. Our findings also suggest that mental health interventions are important and should include targets for reducing violence. School-based interventions that incorporate lay counsellors have been effective at reducing bullying and violence among students [32]. Stand-alone and integrated intervention approaches need to be carefully designed for the local context and to reach adolescents of all ages in school and in the community.

Conclusion
Overall, risk factors for diet and physical activity among adolescents were high, with moderate risk of violence and bullying and SRH indicators, while risk factors for mental health and alcohol and substance use were low. This research suggests that important risk factors for the nine communities studied in sub-Saharan Africa include depression, violence, sexual behaviour, low physical activity, low fruit and vegetable intake, and consumption of soft drinks. These findings suggest more evidence is needed including tracking of risk factors over time.
Where there is sufficient evidence, there is a need to move towards community-tailored interventions to reach adolescent populations with varying needs and approaches between genders and age groups. Specific attention in the development of future research studies should be given to collecting sensitive information from adolescents.