Tropical Medicine and International Health doi:10.1111/tmi.13327
volume 25 no 1 pp 5–14 january 2020
Design and field methods of the ARISE Network Adolescent
Health Study
Anne Marie Darling1, Nega Assefa2, Till Ba€rnighausen1,3,4, Yemane Berhane5, Chelsey R. Canavan1,
David Guwatudde6, Japhet Killewo7, Ayoade Oduola8, Mary M. Sando9, Ali Sie10, Christopher Sudfeld1,
Said Vuai11, Richard Adanu12 and Wafaie W. Fawzi1
1 Department of Global Health and Population, Harvard T.H. Chan School of Public Health, Boston, MA, USA
2 School of Public Health, College of Health and Medical Sciences, Haramaya University, Dire Dawa, Ethiopia
3 Heidelberg Institute of Global Health, University of Heidelberg, Heidelberg, Germany
4 Africa Health Research Institute, Somkhele, KwaZulu-Natal, South Africa
5 Addis Continental Institute of Public Health, Addis Ababa, Ethiopia
6 Department of Epidemiology and Biostatistics, Makerere University School of Public Health, Kampala, Uganda
7 Department of Epidemiology and Biostatistics, Muhimbili University of Health and Allied Sciences, Dar es Salaam, Tanzania
8 University of Ibadan Research Foundation, University of Ibadan, Ibadan, Nigeria
9 Africa Academy for Public Health, Dar es Salaam, Tanzania
10 Nouna Health Research Center, Nouna, Burkina Faso
11 College of Natural and Mathematical Sciences, University of Dodoma, Dodoma, Tanzania
12 Department of Population, Family and Reproductive Health, University of Ghana School of Public Health, Accra, Ghana
Summary The ARISE Network Adolescent Health Study is an exploratory, community-based survey of 8075
adolescents aged 10–19 in 9 communities in 7 countries: Burkina Faso, Eswatini, Ethiopia, Ghana,
Nigeria, Tanzania and Uganda. Communities were selected opportunistically and existing population
cohorts maintained by health and demographic surveillance systems (HDSSs). The study is intended
to serve as a first round of data collection for African adolescent cohorts, with the overarching goal
of generating community-based data on health-related behaviours and associated risk factors in
adolescents, to identify disease burdens and health intervention opportunities. Household-based
sampling frames were used in each community to randomly select eligible adolescents (aged 10–
19 years). Data were collected between July 2015 and December 2017. Consenting participants
completed face-to-face interviews with trained research assistants using a standardised questionnaire,
which covered physical activity, cigarette and tobacco use, substance and drug use, mental health,
sexual behaviours and practices, sexually transmitted infections, pregnancy, food security and food
diversity, teeth cleaning and hand washing, feelings and friendship, school and home activities,
physical attacks and injuries, health care, health status assessment and life satisfaction, as well as
media and cell phone use and socio-demographic and economic background characteristics. Results
from this multi-community study serve to identify major adolescent health risks and disease burdens,
as well as opportunities for interventions and improvements through policy changes.
keywords adolescent health, community-based survey, multi-country study, cohort study, sub-
Saharan Africa
Sustainable Development Goals (SDGs): SDG 2 (zero hunger), SDG 3 (good health and well-being),
SDG 4 (quality education), SDG 5 (gender equity), SDG 10 (reduced inequalities), SDG 17
(partnerships for the goals)
transform their lives and generate high economic and
Introduction
social returns for them and their future offspring. Popula-
Numbering 1.2 billion strong, adolescents comprise the tion projections suggest that the number of individuals
largest generation in history [1]. The vast majority (90%) aged 10–24 in sub-Saharan Africa will increase to 436
of them reside in low-and middle-income settings. Invest- million by 2025 and to 605 million by 2050.[2] Increased
ing in the health of this age group has the potential to focus on the health risks they encounter during this
© 2019 John Wiley & Sons Ltd 5
Tropical Medicine and International Health volume 25 no 1 pp 5–14 january 2020
A. M. Darling et al. ARISE Network Adolescent Health Study
transitional stage is crucial to ensuring they are able to
Setting and study population
survive, thrive and lead.
The ARISE Adolescent Health Study was conducted at
nine communities in seven sub-Saharan African coun-
Rationale for the study design
tries. It was performed in partnership between Africa
Filling the knowledge gap surrounding adolescent health Academy of Public Health (Tanzania), the Centre de
has been identified as the first step in advancing the glo- Recherche en Sante de Nouna (Burkina Faso), Hara-
bal health agenda within this important age group [3]. maya University (Ethiopia), Addis Continental Institute
The African Research, Implementation Science and Edu- of Public Health (Ethiopia), Muhimbili University of
cation (ARISE) Network undertook an exploratory com- Health and Allied Sciences (Tanzania), the University of
munity-based cross-sectional survey on adolescent health Dodoma (Tanzania), the University of Ghana, the
across nine sites in seven sub-Saharan African countries University of Ibadan (Nigeria) and the University of
with the goal of identifying priority areas for interven- Swaziland (Eswatini). Site participation was based on
tion, resource allocation and future research communi- the following factors: budgetary constraints, existing
ties. data collection infrastructure, research team capacity
Much of the evidence pertaining to adolescent health and willingness of a site leader. sites in South, West
used by decision makers has been obtained from school and East Africa.
settings, such as the Global School-based Health Survey Study communities were selected opportunistically by
(GSHS) [4] and the Health Behaviour in School-Aged investigators at each individual site. To allow for poten-
Children (HBSC) survey [5]. Although school-based tial future long-term follow-up and for selection of repre-
approaches have many advantages, they systematically sentative samples, the research team purposefully selected
exclude adolescents who do not attend school. An esti- communities within existing Health and Demographic
mated 89 million adolescents aged 12–24 are out of Surveillance Systems (HDSSs) where possible six of the
school in sub-Saharan Africa, comprising approximately nine sites. Where no HDSS existed, site leaders selected
half the population of this age group [6]. Adolescents communities based on various factors such as available
excluded from educational opportunities may have differ- and reliable population-level data for sampling frames,
ent health behaviour patterns from those attending accessibility and proximity to the research institution,
school and may be more likely to be disadvantaged, vul- community buy-in and approval for carrying out the
nerable and/or marginalized [7]. By implementing a research and demographic characteristics such as urbanic-
household-based sampling design, ARISE sought to ity and population size.
ensure that the study was all segments of the adolescent Three of the study communities were in urban areas:
population that could benefit from the findings. Dar es Salaam, Tanzania; Harar, Ethiopia and Ibadan,
Although WHO defines adolescents as individuals in Nigeria; the other six were in rural areas: Dodoma,
the 10–19 years age range, younger adolescents have not Tanzania; Iganga/Mayuge, Uganda; Kersa, Ethiopia;
consistently been included in nationally representative Lubombo/Manzini, eSwatini; Ningo Prampram, Ghana;
health surveys. The Demographic and Health Survey and Nouna, Burkina Faso (Figure 1). In all communi-
(DHS) [8], used widely in sub-Saharan Africa, collects ties, male and female residents aged 10–19 years were
data from a large number of households (approximately recruited for the study through household-based sam-
5000–30 000) on a range of demographic and health pling. In Burkina Faso, both communities in Ethiopia,
indicators, but the minimum age for participation is both communities in Tanzania and Uganda, the study
15 years and the questionnaire is not specifically tailored was nested within existing Health and Demographic
to the adolescent experience. Early adolescents (ages 10– Surveillance Sites (HDSS). Adolescents who refused to
14 years) are typically considered to be a healthy popula- participate, were too sick to be interviewed, or were
tion, even though many health behaviours become estab- absent at the time of data collection were excluded.
lished during this transitional period [9]. To address gaps Married adolescents were additionally excluded at the
in our understanding of this phase, the ARISE Adolescent Ethiopian communities. According to the 2016 Demo-
Health study enrolled adolescents starting at age 10 and graphic and Health Survey (DHS) from Ethiopia,
used a questionnaire specifically developed for adoles- 78.1% of females and 98.3% of males aged 15–19
cents, who comprise a major proportion of the rapidly reported never being married. Demographic characteris-
expanding youth population on the sub-Saharan African tics of participants enrolled in each community are
continent. shown in Table 1.
6 © 2019 John Wiley & Sons Ltd
Tropical Medicine and International Health volume 25 no 1 pp 5–14 january 2020
A. M. Darling et al. ARISE Network Adolescent Health Study
Figure 1 Map of ARISE Network Adolescent Health Study communities.
Table 1 Demographic characteristics of participants enrolled at each community in the ARISE Network Adolescent Health Study
Gender Age In School
Community N Male Female 10–14 15–19 Yes No
Nouna, Burkina Faso 1629 687 (42.2) 942 (57.8) 742 (45.5) 88.7 (54.5) 817 (50.2) 812 (49.8)
Lubombo/Manzini, eSwatini 412 161 (39.1) 251 (60.9) 124 (30.1) 288 (69.9) 303 (73.5) 109 (26.5)
Harar, Ethiopia 1059 500 (47.2) 559 (52.8) 562 (53.1) 497 (46.9) 1001 (94.5) 58 (5.5)
Kersa, Ethiopia 951 528 (55.5) 423 (44.5) 667 (70.1) 284 (29.9) 540 (56.8) 411 (43.2)
Ningo prampram, Ghana 625 281 (45.0) 344 (55.0) 339 (54.2) 286 (48.8) 496 (79.4) 129 (20.6)
Ibadan, Nigeria 750 339 (45.2) 411 (54.8) 298 (39.7) 452 (60.3) 560 (75.6) 181 (24.4)
Dar es Salaam, Tanzania 825 376 (45.6) 449 (54.4) 266 (32.3) 558 (67.7) 577 (71.3) 232 (28.7)
Dodoma, Tanzania 1226 548 (44.7) 678 (55.3) 818 (66.7) 408 (33.3) 898 (73.2) 328 (26.8)
Iganga/Mayuge, Uganda 598 312 (52.2) 286 (47.8) 320 (53.6) 277 (46.4) 521 (87.1) 77 (12.9)
one adolescent resident were randomly selected from the
Communities and sampling procedures
sampling frame. When two or more adolescents were
Neither the communities included in the study nor sam- listed in one household, one was selected randomly for
ples of adolescents selected from each community were interview. All sampled adolescents were sought for inter-
selected to be representative of the larger national or view at each community. Face-to face interviews were
regional populations. As described below, however, each carried out at various times between July 2015 and
site employed sampling methods aimed at obtaining a December 2017 by trained research assistants with prior
sample of adolescents that represented the individual data collection experience and good knowledge of the
community. Age-eligible potential participants in all com- local language. Both male and female research collected
munities were randomly selected from household and data. In all communities, research assistants underwent at
household member sampling frames derived from HDSSs least one day of intensive training on how to conduct the
or other recent census records. Households with at least survey. Each individual community is described below.
© 2019 John Wiley & Sons Ltd 7
Tropical Medicine and International Health volume 25 no 1 pp 5–14 january 2020
A. M. Darling et al. ARISE Network Adolescent Health Study
Nouna, Burkina Faso Ningo Prampram, Ghana
Nouna is a semi-urban town in the north-west of Burkina The Ghanaian study community is located in the Ningo
Faso. It serves as the administrative centre of Kossi pro- Prampram district within the greater Accra region in
vince. The Nouna HDSS, located in the town of Nouna coastal southern Ghana. The population of the district was
and surrounding villages, was established in 1992. Its estimated at approximately 70 000 by the 2010 census
catchment area corresponds to an area of 7464 km2 and [13]. In this community, a two-stage sampling technique
an estimated population of 320 232 [10]. Trained field was used. First, 25 enumeration areas (EAs), as defined by
staff visit all households within the HDSS boundaries the 2010 census, were randomly selected. The selection
every four months to record births, deaths and in- and probability was proportional to EA size. All housing units
out-migration. The Nouna community utilised a two-part were then documented within each EA to form the sam-
stratified sampling procedure aimed at ensuring commu- pling frame for the selection of adolescents in the second
nity representativeness with respect to ethnicity (and thus stage. In this stage, 25 households with at least one adoles-
roughly religion) and urbanicity, since health practices cent resident were randomly selected from each EA. When
differ systematically by these variables. First, a sample of two or more adolescents were listed in one household, one
10 (of a total of 59) villages within the HDSS census area was chosen randomly for interview. Data were collected
was selected to ensure inclusion of all five main ethnici- during the rainy season in June and July 2016.
ties. All children in these 10 villages who had been enu-
merated in the 2015 HDSS census and were between the
Ibadan, Nigeria
ages of 12–19 years on 1 October 2017 were classified
by ethnicity. A total of 1795 individuals were then sam- The Nigerian study community is located in Ibadan, the capi-
pled proportionally to the population size within each tal and most populous city of Oyo State in south-western
ethnicity stratum. Second, a simple random sample of Nigeria. The population of Ibadan has been estimated at over
749 age-eligible adolescents was selected from all those 3 million. There are 11 Local Government Areas (LGAs) in
in one sector of Nouna town. This ratio of semi-urban to Ibadan Metropolitan area: five within the city and six semi-
rural individuals matched the ratio seen in the overall urban ones. In this community, a two-stage sampling design
HDSS. Data were collected during the dry season in was used. First, two LGAs (Ibadan North and Ibadan South-
November and December 2017. West) within the Ibadan Metropolitan area were randomly
selected for inclusion in the study. Households with at least
one adolescent resident were then randomly selected from
Harar and Kersa, Ethiopia each LGA proportionally to the population size of the LGA.
Harar is located in eastern Ethiopia, 510 km from the When two or more adolescents were listed in one household,
capital city of Addis Ababa. The Harar HDSS, which one was selected randomly for interview. Data were collected
began in 2011 and is managed by Haramaya University, during the dry season in March 2017.
covers a 19.5 km2 area inhabited by approximately
60 000 individuals that encompasses 12 of the 19 sub- Lumboo and Manzini, Eswatini
districts (kebeles) in Harar City [11]. Trained field staff
visit the more than 9000 households under surveillance In Eswatini, data collection took place in the eastern
at six-monthly intervals to ascertain vital events, socio- region of Lubombo and the central-western region of
economic characteristics and physical characteristics of Manzini. Both regions are mostly rural, though the Man-
the dwellings. The Kersa HDSS, located in the east Har- zini region also includes the city of Manzini, Eswatini’s
araghe zone of the Oromia region, was established in second largest urban area. At the time of the last census
2007. It is an open cohort of all individuals permanently in 2007, the populations of Lubombo and Manzini were
living in 24 of the 35 rural sub-districts of Kersa (popula- estimated at 220 000 and 350 000, respectively [14]. In
tion approximately 129 000) that is also managed by this community, a two-stage stratified cluster random
Haramaya University [12]. Approximately 63 000 indi- sampling was used. Strata were region (Lubombo, Man-
viduals are currently under surveillance. Censuses of the zini) and urbanicity (rural, urban). In the first stage, 50
Kersa HDSS are conducted at six-monthly intervals to EAs in each of the two regions were selected. In both
record births, deaths and in- and out-migration. Other regions, 37 of the selected areas were classified as rural
information, including changes in marital status, preg- and 13 were classified as urban by the Eswatini Statistics
nancy outcomes, health status and economic status, is Office. In the second stage, 20 households in each EA
collected at longer intervals. were selected using systematic random sampling. If no
8 © 2019 John Wiley & Sons Ltd
Tropical Medicine and International Health volume 25 no 1 pp 5–14 january 2020
A. M. Darling et al. ARISE Network Adolescent Health Study
adult household member was available at the time of in 2004 and is located in the two eastern Uganda districts
visit, the household was replaced with another randomly of Iganga and Mayuge, about 120 km east of the capital
selected household. Data collection took place during the city of Kampala. The 63 villages that comprise the pre-
dry season in July and August 2015. dominantly rural IMHDSS contain a total population of
approximately 74 000 residing in approximately 16 000
households. The approximately 25 000 registered adoles-
Dar es Salaam and Dodoma, Tanzania
cents aged 10-19 make up one-third of the total popula-
The Tanzanian study communities were located in Dar es tion. Twice per year, trained field workers record births,
Salaam and Dodoma, Tanzania. Dar es Salaam lies along deaths, pregnancies and in- and out-migrations [16].
the Indian Ocean coast and is the financial capital of Tan- IMHDSS records were used to select a random sample of
zania. The Dar es Salaam HDSS, known as the Dar es Sal- adolescents stratified by urbanicity. Peri-urban residents
aam Urban Cohort Study (DUCS), was established in 2011 comprised 20% of the sample and rural residents com-
for the purposes of longitudinally monitoring of demo- prised 80% of the sample. Adolescents were eligible for
graphic events and conducting nested epidemiologic sur- selection if they had resided in their household for at
veys. DUCS covers the Ukonga and Gongo la Mboto least one year. When the selected adolescent could not be
wards of the Ilala region [15]. It encompasses seven admin- contacted, traced or declined to participate in the study,
istrative streets (Gongo la Mboto, Guluka kwa lala, he or she was replaced by another adolescent from the
Mwembe Madafu, Markaz, Mazizini, Mongo la Ndege nearest household. Data collection in this community
and Ulongoni), which span a 9.91 km2 area about 20 km took place in February and March 2016, a period that
from the city centre. All households within this area are spans the dry season and beginning of the rainy season.
visited twice per year, and any household member consid-
ered as having the primary dwelling within the household
Informed consent
and having lived in the household for the previous
3 months preceding the census was eligible for inclusion. In all communities, field staff visited households selected
Over 100 000 individuals living in 21 000 households by the sampling procedures to recruit potential partici-
have been enumerated through 30 June 2015. At the time pants. Participants and their parents were first informed
of the study, 14 920 adolescents had been registered. of the purpose and nature of the study and informed that
Births, deaths, changes in marital status and residency sta- their participation was voluntary. Written informed con-
tus data are recorded during every update. Household sent was obtained from all adolescents aged 18 and
information is updated every other year. Ten trained 19 years. Written parental consent and adolescent assent
research assistants with prior data collection experience were obtained from adolescents younger than 18 years of
within DUCS carried out the ARISE survey between Febru- age. Interviews were held in private settings within the
ary and September 2016, a period that spans both the long household or compound.
rainy season and the dry season. Dodoma is situated in the
central plateau of Tanzania. The Dodoma HDSS was
Questionnaire
launched in 2016 and is based in the rural district of
Chamwino. At the time of the study, the Dodoma HDSS Face-to-face interviews were conducted by trained inter-
has enumerated 23 785 individuals residing in 5266 house- viewers with all consenting participants using a standard-
holds across the five villages of the Mlowa barbarani and ised questionnaire translated into the local language. The
Makang’wa wards. Household members who have lived questionnaire was adapted from the widely used GSHS
continuously in the project area for 4 months or longer are [4], which has been extensively validated and applied in
eligible for inclusion. Adolescents were randomly selected 22 countries across sub-Saharan Africa including Ethio-
from the sampling frame of all households containing one pia, Ghana, Nigeria, Tanzania, Eswatini and Uganda.
or more adolescent. Only one adolescent was selected per Burkina Faso was the only country in the current study
household. Data collection took place from April to June that did not participate in the GHSS.
2017, a period that includes both rainy and dry months. Because the GSHS was designed for use among in-
school adolescents only, our research study team mem-
bers reviewed the content to ensure its appropriateness
Uganda for both in- and out-of-school adolescents. The ARISE
Adolescent Health Study team includes experts across the
The Iganga-Mayuge Health & Demographic Surveillance
domains covered in the research, including nutrition,
Site (IMHDSS) was established by Makerere University
mental health, reproductive health, substance use, HIV/
© 2019 John Wiley & Sons Ltd 9
Tropical Medicine and International Health volume 25 no 1 pp 5–14 january 2020
A. M. Darling et al. ARISE Network Adolescent Health Study
AIDS and other areas. These experts also provided Table 2 Modules included in the ARISE Network Adolescent
important local expertise on context and culture in each Health Study questionnaire
community included in the research.
Domain Subtopic Items
At an ARISE Network meeting in Dar es Salaam, Tan-
zania, in 2015, the GSHS tool was reviewed in detail. Socio-demographics Demographics • Age
Our team of experts recommended minimal revisions to • Sex
existing questions and addition of other validated instru- • Educational
status
ments where needed. A consensus vetting of changes was
Parental vital
then carried out, and a draft instrument was circulated to • status
all research team members for comment. The final instru- • Parental age
ment was agreed upon by the multidisciplinary team of • Parental
research scientists. Modules covered by the questionnaire education
included socio-demography, socio-economy, food secu- • Parental
rity, food diversity, teeth cleaning and hand washing, occupation
Household
feelings and friendship, physical activity, school and •
composition
home activities, physical attacks, injuries, health care, Household • Drinking water
health status assessment, life satisfaction, cigarette and socio-economic quality
tobacco use, substance use, drug use, sexual practice, status • Toilet facilities
pregnancy, media use and sexually transmitted infection. • Cooking fuel use
As dictated by the GSHS tool, insertions were made to • Assets
certain questions by each site to capture site-specific types • Food security
Nutrition and Diet Food group
of foods, physical activity, alcohol and illicit substances •
physical activity consumption
and site-specific slang words used to describe certain within previous
issues. Revisions were made to the GSHS question on 24 hours
bicycle riding for physical activity to encompass out-of- • Frequency of
school adolescents: “During the past 7 days, on how fruit, vegetable,
many days did you walk or ride a bicycle to or from carbonated
beverage and fast
school?” was changed to “During the past 7 days, on
food
how many days did you walk or ride a bicycle to or from consumption
school, work, store or other location?” • Height
Other validated and previously used instruments were • Weight
included for additional information on topics of impor- • Arm
tance. The GSHS core module includes four questions on circumference
HIV that are school-specific and does not include other Physical activity • Frequency of
physical activity
sexually transmitted diseases. Therefore, ARISE used the
• Frequency of
WHO “Asking young people about sexual and reproduc- walking/biking
tive behaviours” [17] questions for HIV and STIs as well for transport
as pregnancy. • Sedentary
Self-rated health was measured using the following val- behaviour
idated instruments: Kutcher Adolescent Depression Scale Hygiene Dental hygiene • Frequency of
(KADS-6) [18], health-related quality of life (HRQOL) tooth brushing
Frequency of
[19], the Health Behaviour in school-aged Children’ •
dental visits
(HBSC) symptom checklist [20], and questions on sexual Hand hygiene • Frequency of
assault adapted from Life Events Checklist (LEC) for hand washing:
DSM-5 [21]. Life satisfaction was assessed using two • Before eating
measures: Cantril’s Ladder [22] and the Student’s Life • After using the
Satisfaction Scale (SLSS) [23]. Dietary diversity was toilet
With soap
ascertained via 24-hour recall of food categories from a •
standard 16-item food category list [24].
Where the GSHS collects self-reported height and
weight, the ARISE instrument added measured height,
10 © 2019 John Wiley & Sons Ltd
Tropical Medicine and International Health volume 25 no 1 pp 5–14 january 2020
A. M. Darling et al. ARISE Network Adolescent Health Study
Table 2 (Continued) Table 2 (Continued)
Domain Subtopic Items Domain Subtopic Items
Reproductive health Menstruation • Knowledge of Injury • 12-month injury
menstruation history
• Menarche status • Type(s) of injury
• Menstrual experienced
hygiene Substance use Tobacco • Ever use of
• Menstruation- tobacco products
related school • 30-day use of
absences tobacco products
Sexual • Sexual activity • Passive exposure
behaviour status to tobacco
• Age at sexual products
initiation Alcohol • Ever use of
• Contraceptive use alcohol
• Sexting • 30-day use of
behaviour alcohol
• Sexually explicit • Source of alcohol
media exposure • Alcohol-related
Pregnancy • Ever been/made impairment
someone Drugs • Use of:
pregnant • Marijuana
• Pregnancy • Cocaine
wantedness • Amphetamines
• Pregnancy • Inhalants
outcome • Khat (Ethiopia
Socio-emotional • Experiences of: and Uganda)
health • Loneliness Health care utilisation • Type of health
• Anxiety-related service accessed
insomnia within the past
• Low mood 12 months:
• Anxiousness • Hospital
• Depression admissions
symptoms • Primary care
• Suicidal ideation visits
• Suicidal attempts • Traditional
• Social healer visits
relationships • Reason for
• School absences accessing service
• Parental • Cost of service
connection • Satisfaction with
• Parental service
regulation • Use of Internet to
• Self-rated health seek health
status information
• Self-reported
physical
symptoms weight and mid-upper arm circumference. Detailed
• Life satisfaction descriptions of each model are provided in Table 2.
Injury/Violence Violence • 12-month history Piloting was necessary despite the fact that the vast
of physical
majority of questions in the survey were from the GSHS,
attacks
• 12-month history which has been conducted and vetted previously in 97
of physical fights countries. Pilot testing of the ARISE instrument took
• Bullying history place in Ethiopian communities, allowing for feedback
• History of sexual from both a rural and urban community (two of the nine
violence sites were in Ethiopia).
© 2019 John Wiley & Sons Ltd 11
Tropical Medicine and International Health volume 25 no 1 pp 5–14 january 2020
A. M. Darling et al. ARISE Network Adolescent Health Study
must be taken by field interviewers to ensure privacy from
Data management
potential intrusions by family members or neighbours. In
Data were recorded on paper and entered electronically addition, some participants may experience discomfort
into Epi-data version 3.1. Identifiable information was when asked sensitive information. To minimise this dis-
confidentially maintained in secure databases with access comfort, some communities matched interviewers to par-
restricted to key study personnel. Standardised variable ticipants by sex. For adolescents under the age of 18,
names were used across communities. Data collection parental involvement in the informed consent process is
progress was monitored internally by each community crucial in order to ensure their protection from research-re-
and reviewed at interim meetings of the ARISE Network. lated harms. Folayan et al. [25] note that a potential trade-
Data from all communities were pooled and cleaned cen- off for this involvement, however, is that it may hinder the
trally using SAS v. 9.4 (SAS Institute). autonomous decision-making of the adolescent and com-
promise confidential information about them.
Managing social desirability bias around sensitive
Limitations
topics is another challenge particular to conducting face-
This study has three main limitations. First, as stated to-face interviews with adolescents. Prevalence estimates
above, the adolescents sampled at each site are not repre- for sexual behaviour and substance use do suggest that
sentative of the general population of adolescents within some underreporting may have occurred in this study
the seven countries or sub-Saharan Africa as a whole. [26]. The interviewer-administered survey nevertheless
Survey results therefore cannot be generalised beyond the remains the principal data collection tool for surveys in
individual communities included in the study. Neverthe- resource-limited settings, and no consensus has been
less, we believe that meaningful inferences can be drawn reached regarding alternative data collection methods
from the cross-community comparison of results across that may enhance data quality in sub-Saharan Africa.
these 9 sites that can be applied to communities with Audio computer-assisted self-interview (ACASI) is the
similar geographic and socio-demographic characteristics preferred method for collecting data on sensitive beha-
in the region. viours in the United States, but Mensch et al. [27] have
Second, this study collected solely quantitative data. A suggested that its usefulness and applicability in Africa
mixed-methods design would have provided richer infor- may be dependent on the local context. This observation
mation about attitudes and behaviours and their cultural is supported by the results of a systematic review of 15
context, but was neither feasible or preferable at this datasets from low-and middle-income countries, most of
exploratory stage. Collecting quantitative data through which compared ACASI methods to face-to-face inter-
the use of a structured questionnaire allowed for stan- views (FTFI) for reporting HIV risk behaviours [28]. The
dardisation of data collection across 9 diverse sites. These review concluded that non-FTFI methods were not con-
data can also be used as baseline measures for tracking sistently associated with a significant increase in the
progress towards quantitative goals in future rounds of reporting of all outcomes.
the survey. Furthermore, the quantitative results can Non-verbal response cards (NVRC) were developed by
inform future qualitative work. For example, since the Lindstrom et al. [29] as an affordable, user friendly alter-
results of the survey suggest that many adolescents have native to ACASI. In rural Tanzania, this data collection
an unmet need for health services, qualitative data collec- method was associated with greater reporting of sexual
tion can be targeted towards understanding the types of activity, HIV testing, a larger number of lifetime sexual
barriers to healthcare access that adolescents face.Lastly, partners and younger ages at first sex among young
only one site pilot-tested the instrument prior to field women [30]. While potentially a promising technique for
implementation. The instrument was, however, largely collecting sensitive data among sub-Saharan African ado-
based on a repeatedly validated instrument developed by lescents, more research is needed to determine its feasibil-
WHO that has been widely used around the globe with ity and acceptability in different local contexts within
the addition of other previously validated instruments. this region. At the Burkina Faso community of the ARISE
Adolescent Health Study, a module for pilot testing this
method was included in the questionnaire, which may
Challenges encountered in adolescent survey
provide additional insight regarding its ability to enhance
research
data quality in this setting.
We wish to comment on challenges encountered when con- Mobile phones are an adolescent-friendly technology
ducting household-based survey research in an adolescent that may provide another means of collecting sensitive
population. From an ethical perspective, particular care data. They have been used for phone-based surveillance
12 © 2019 John Wiley & Sons Ltd
Tropical Medicine and International Health volume 25 no 1 pp 5–14 january 2020
A. M. Darling et al. ARISE Network Adolescent Health Study
projects on a small scale in sub-Saharan Africa [31] and (HBSC) study: origins, concept, history and development
shown to increase reporting of sexual activity among 1982–2008. Int J Public Health 2009: 54: 131–139.
young women in South Africa [32]. Several concerns have 6. Inoue K, Di Gropello E, Taylor YS, Gresham J. Out-of-
been identified regarding their use in this population, School Youth in Sub-Saharan Africa: A Policy Perspective.
The World Bank 2015.
however, including lack of airtime with which to
7. Auerswald CL, Piatt AA, Mirzazadeh A.Unicef. Research
respond, device loss, swapping of phone subscriber iden-
with disadvantaged, vulnerable and/or marginalized adoles-
tity module cards, deletion of the survey application by cents.
participants, low confidence about understanding the 8. ICF. “Methodology”. The DHS Program website. Funded by
questions being asked and concerns about confidentiality USAID. (Available from: http://www.dhsprogram.com) [14
of responses [33]. More refinement of this method may Dec 2018].
be needed before it becomes widely adopted for survey 9. Blum RW, Mmari K, Moreau C. It begins at 10: How gen-
research in this setting. der expectations shape early adolescence around the world. J
Adolesc Health 2017: 61: S3–S4.
Future directions 10. Sie A, Louis VR, Gbangou A et al. The Health and Demo-
graphic Surveillance System (HDSS) in Nouna, Burkina
The ARISE Network is currently collaborating on the Faso, 1993–2007. Global Health Action 2010: 3: 1993–
development of integrated multi-community studies to 2007.
test the effectiveness and impact of interventions and 11. Assefa N, Semahegn A. Fertility is below replacement in
policies to address adolescent health needs identified in Harar Health and Demographic Surveillance System (Harar
the Adolescent Health Study. In addition, the Network is HDSS), Harar town, Eastern Ethiopia. Fertil Res Pract
working on additional rounds of longitudinal data collec- 2016: 2: 10.
12. Assefa N, Oljira L, Baraki N et al. HDSS profile: The Kersa
tion in order to evaluate trends in adolescent health indi-
health and demographic surveillance system. Int J Epidemiol
cators over time and age. Future rounds will incorporate
2016: 45: 94–101.
qualitative data collection, innovative methods for col- 13. Ghana Statistical Service. 2010 Population and Housing
lecting data to minimise reporting and social desirability Census. Republic of Ghana, 2012.
biases as well as biomarker and anthropometric metrics 14. Central Statistical Office. 2007 Population and Housing
of health risk factors and outcomes. Census. UNFPA: The Kingdom of Swaziland, 2010.
15. Leyna GH, Berkman LF, Njelekela MA et al. Profile: The
Acknowledgements Dar es Salaam Health and Demographic Surveillance System
(Dar es Salaam HDSS). Int J Epidemiol 2017: 46: 801-808.
We acknowledge the contribution of field supervisors and 16. Kadobera D, Waiswa P, Peterson S et al. Comparing perfor-
data collectors in each site as well as the contribution of mance of methods used to identify pregnant women, preg-
study participants. The following people were instrumen- nancy outcomes, and child mortality in the Iganga-Mayuge
tal to this work: Augustine Malelo, Abdallah Mtumwa, Health and Demographic Surveillance Site, Uganda. Glob
Jumanne Kisweka and Nathan Isabirye. Funding for the Health Action 2017: 10: 1356641.
ARISE Adolescent Health Study was provided by the 17. Cleland J, Ingham R, Stone N.Asking young people about
Department of Global Health and Population at Harvard sexual and reproductive behaviours, 2001 (Available from:
http://www.who.int/reproductive-health/adolescent/docs/que
T.H. Chan School of Public Health.
stionnaire_intro.pdf: UNDP/UNFPA/WHO/World Bank spe-
cial programme for research development and research
References training in human reproduction).
1. Sheehan P, Sweeny K, Rasmussen B et al. Building the foun- 18. LeBlanc JC, Almudevar A, Brooks SJ, Kutcher S. Screening
dations for sustainable development: a case for global invest- for adolescent depression: comparison of the Kutcher Ado-
ment in the capabilities of adolescents. Lancet 2017: 390: lescent Depression Scale with the Beck depression inventory.
1792–1806. J Child Adolesc Psychopharmacol 2002: 12: 113–126.
2. Hervish A, Clifton D.Status report: Adolescents and young 19. Centers for Disease Control and Prevention. Measuring
people in sub-Saharan Africa. Opportunities and challenges. Healthy Days. CDC: Atlanta, Georgia, 2000.
3. Gates M. Advancing the adolescent health agenda. Lancet 20. King A. The Health of Youth: A Cross-National Survey.
2016: 387: 2358–2359. WHO Regional Publications, European Series No. 69. Office
4. World Health Organization. Global school-based student of Publications, WHO Regional Office for Europe: Scher-
health survey (GSHS). WHO CHP;2009. figsvej 8, DK-2100, Copenhagen 0, Denmark, 1996.
5. Currie C, Gabhainn SN, Godeau E. International HBSC 21. Gray MJ, Litz BT, Hsu JL, Lombardo TW. Psychometric
Network Coordinating Committee. The Health Behaviour in properties of the Life Events Checklist. Assessment 2004:
School-aged Children: WHO Collaborative Cross-National 11: 330–341.
© 2019 John Wiley & Sons Ltd 13
Tropical Medicine and International Health volume 25 no 1 pp 5–14 january 2020
A. M. Darling et al. ARISE Network Adolescent Health Study
22. Cantril H. The pattern of human concerns. Rutgers Univer- behaviours in low-and middle-income countries. Int J Epi-
sity Press: New Brunswick, NJ, 1965. demiol 2010: 39: 1541–1555.
23. Huebner ES. Initial development of the student’s life satis- 29. Lindstrom DP, Belachew T, Hadley C, Hattori MK, Hogan
faction scale. School Psychol Int 1991: 12: 231–240. D, Tessema F. Nonmarital sex and condom knowledge
24. Smith LC, Dupriez O, Troubat N. Assessment of the relia- among Ethiopian young people: improved estimates using a
bility and relevance of the food data collected in national nonverbal response card. Stud Fam Plann 2010: 41: 251–
household consumption and expenditure surveys. Interna- 262.
tional Household Survey Network, 2014. 30. Aichele SR, Mulder MB, James S, Grimm K. Attitudinal
25. Folayan MO, Haire B, Harrison A, Fatusi O, Brown B. and behavioral characteristics predict high risk sexual
Beyond informed consent: ethical considerations in the activity in rural Tanzanian youth. PLoS ONE 2014: 9:
design and implementation of sexual and reproductive e99987.
health research among adolescents. Afr J Reprod Health 31. Brinkel J, Kra€mer A, Krumkamp R, May J, Fobil J. Mobile
2014: 18: 118–126. phone-based mHealth approaches for public health surveil-
26. Berhane Y, Canavan CR, Darling AM et al. The age of lance in sub-Saharan Africa: a systematic review. Int J Envi-
opportunity: Prevalence of key risk factors among adoles- ron Res Public Health 2014: 11: 11559–11582.
cents 10-19 years of age in nine communities in sub-Saharan 32. Dietrich JJ, Lazarus E, Andrasik M et al. Mobile phone
Africa. Tropical Med Int Health 2020: 25: 15–32. questionnaires for sexual risk data collection among young
27. Mensch BS, Hewett PC, Erulkar AS. The reporting of sensi- women in Soweto, South Africa. AIDS Behav 2018: 1.
tive behavior by adolescents: a methodological experiment 33. van Heerden AC, Norris SA, Richter LM. Using mobile
in Kenya. Demography 2003: 40: 247–268. phones for adolescent research in low and middle income
28. Phillips AE, Gomez GB, Boily MC, Garnett GP. A system- countries: preliminary findings from the birth to twenty
atic review and meta-analysis of quantitative interviewing cohort, South Africa. J Adolesc Health 2010: 46: 302–
tools to investigate self-reported HIV and STI associated 304.
Corresponding Author Anne Marie Darling, Department of Global Health and Population, Harvard T.H. Chan School of Public
Health, 677 Huntington Avenue, Boston, MA, USA. E-mail: adarling@hsph.harvard.edu
14 © 2019 John Wiley & Sons Ltd