Study setting
The current study uses data collected as part of an on-going, longitudinal impact evaluation of a pilot entitled a “Cash Plus” Model for Youth Well-Being and Safe, Healthy and Productive Transitions to Adulthood. The study was implemented in four districts: Rungwe and Busokelo in Mbeya region and Mufindi and Mafinga in Iringa region. Mbeya is located in the South West Highlands, while Iringa is in the Southern Highlands zone. In these regions, agriculture is the most common sector of employment, and the populations face significant health challenges, including higher than national averages of HIV infection, high rates of stunting and low rates of access to health insurance [25].
Financing for health services in Tanzania relies on a mixture of taxes and user fees. There are three main insurance schemes introduced by the MoHSW, including the Community Health Fund, TIKA (stands for Tiba Kwa Kadi and translates to “treatment with card”), and the National Health Insurance Fund [26]. CHF requires an annual premium, even for extremely poor households, and NHIF is available only to formal sector employees. Nationally, the rates of uninsured are 91% among women and 90.5% among men. In Iringa and Mbeya, specifically, the rates are 89.7 and 90.4 (compared to 88.8 and 84% among men), respectively [27]. There are fees associated with most health services in Tanzania, however some services, such as HIV testing and treatment, are provided free to the public through government subsidization. Some facilities may opt to use a sliding scale fee and take into account ability to pay in determining fees.
Study design
This study is observational, but uses secondary data on adolescents and health facilities gathered during an impact evaluation. The evaluation utilizes cluster randomized control trial (cRCT) design to assess an intervention implemented by the Tanzania Social Action Fund (TASAF), an agency of the Government of the United Republic of Tanzania. The intervention targeted adolescents living in extremely poor households participating in the government’s flagship social protection program, the Productive Social Safety Net (PSSN). Components of the adolescent-focused intervention included: 1) livelihoods and life skills training; 2) mentoring and a productive grant; and 3) linkage to strengthened HIV, SRH and violence response services provided by government-run, primary health care facilities. More information on the overall impact evaluation study design and sampling is provided in the Appendix.
The current observational study leverages data from the second round of data collection of this on-going impact evaluation, which follows a panel sample of 2191 youth from 1779 households in 130 villages. Eligibility criteria for youth to be included in the study were as follows: 1) aged 14–19 years at baseline and 2) living in a PSSN household at baseline. Thus, all study households receive cash transfers and other components of the PSSN. Baseline surveys were conducted between April and June 2017, prior to randomization of villages into study arms. The adolescent-focused intervention was randomized at the village-level (65 randomized to treatment and 65 randomized to control). Round 2 of the youth/household surveys was conducted May – July 2018. Youth questionnaires were administered directly with youth, using same-sex enumerators. Informed consent was obtained from all youth aged 18 years and above and for married youth aged 14 to 17 years. For unmarried adolescents aged 14 to 17 years, informed consent was obtained from household head or caregiver, and informed assent was obtained from the youth. Questionnaires were administered in private areas in and around the household, where family members or others could not overhear sensitive topics of the questionnaire.
This study also includes data from 91 government-run, primary health care facilities. Eligibility criteria for inclusion of health facilities in the study were: 1) government-run primary health care facility, including dispensaries and 2) located in study area. Health facility surveys were administered to staff working at these facilities. Data used in the current analysis come from Round 2 of the health facility data collection (conducted February–March, 2018). For more information on sample selection and data collection, see Tanzania Cash Plus Evaluation Team (2018) [28].
To link the facility- and adolescent-level data, adolescents were assigned to a health facility based on their residential location. A third (33%) of the adolescents had no health facility located in their village. In these cases, the adolescent was matched to the closest health facility in another village based on GPS coordinates. About a quarter of adolescents (26%) had more than one health facility located in their village. These adolescents were assigned to the facility that reported performing the highest number of adolescent HIV tests in the past year. The rationale for this choice was that adolescents as a group had already indicated, through their choice of clinic, where they were most likely to seek HIV testing. While there were 91 health facilities in the data, only 69 resulted in the closest match for adolescents in the study sample, and thus n = 69 facilities comprises the sub-sample for analyses.
Measures
The main independent variables of interest related to availability of adolescent-friendly services. Table 2 lists the variables used to define the indicators, organized by the WHO’s five objectives for assessing quality of health care services for adolescents. To examine whether health facilities are accessible to adolescents, for example, we measured whether they are conveniently located, had adequate opening hours, offered affordable services, made efforts to inform adolescents about their health services, and tried to communicate the importance of adolescent services to the community. Health facility surveys also captured indicators of whether services were equitable (for instance, provided to all adolescents irrespective of their marital status), appropriate (e.g., offered comprehensive SRH), and effective (e.g., availability of adolescent friendly trained staff, use of evidence-based protocols).
The final WHO quality objective – the acceptability of services – can only be assessed from the perspective of the adolescent. Those adolescents who reported utilizing SRH services were thus asked about whether they felt comfortable asking SRH staff questions; whether staff answered SRH questions adequately; whether staff were friendly, and whether SRH services were adequately confidential. Since such indicators were only available for a subset of the adolescents and had high endorsement, we do not include these factors in models predicting uptake of HIV and SRH services.
The main outcome of interest was utilization of health services, categorized with two indicators: 1) HIV testing and 2) visiting a health care facility in the past year for HIV/SRH services. HIV testing was captured using the questions, “I do not want to know the result, but have you ever been tested for HIV?” If the answer was yes, a follow-up question was asked about testing behavior in the past year. While HIV testing can take place at community organizations or even at home, other SRH services are primarily accessible through health facilities. Adolescents were asked if they had visited a health facility in the past year for HIV/SRH services, defined as services related to contraception, pregnancy, or sexually-transmitted infections (STI) testing or treatment.
Control variables used in the regression analyses include basic health facility characteristics (type of clinic (health care facility v. dispensary), number of medical staff) and adolescent demographic characteristics (gender, age, education, marital status, and sexual debut).
Analysis
Frequencies and means (as appropriate) are used to characterize the adolescent sample, the health facilities, and the availability of AFHS. To estimate the association between facility-level AFHS characteristics and adolescents’ uptake of services, we calculated risk ratios using multivariate log binomial regression models. One regression was separately run for each independent variable (indicating facility-level quality), and standard errors were clustered at the village level to account for the fact that all adolescents within one village are linked to the same health facility. All models include the control variables listed above. For both past-year outcomes, we also created models stratified by whether the adolescent reported sexual debut.