Improving retention in antenatal and postnatal care: a systematic review of evidence to inform strategies for adolescents and young women living with HIV

Abstract Introduction Young pregnant and postpartum women living with HIV (WLHIV) are at high risk of poor outcomes in prevention of mother‐to‐child transmission services. The aim of this systematic review was to collate evidence on strategies to improve retention in antenatal and/or postpartum care in this population. We also conducted a secondary review of strategies to increase attendance at antenatal care (ANC) and/or facility delivery among pregnant adolescents, regardless of HIV status, to identify approaches that could be adapted for adolescents and young WLHIV. Methods Selected databases were searched on 1 December 2020, for studies published between January 2006 and November 2020, with screening and data abstraction by two independent reviewers. We identified papers that reported age‐disaggregated results for adolescents and young WLHIV aged <25 years at the full‐text review stage. For the secondary search, we included studies among female adolescents aged 10 to 19 years. Results and discussion Of 37 papers examining approaches to increase retention among pregnant and postpartum WLHIV, only two reported age‐disaggregated results: one showed that integrated care during the postpartum period increased retention in HIV care among women aged 18 to 24 years; and another showed that a lay counsellor‐led combination intervention did not reduce attrition among women aged 16 to 24 years; one further study noted that age did not modify the effectiveness of a combination intervention. Mobile health technologies, enhanced support, active follow‐up and tracing and integrated services were commonly examined as standalone interventions or as part of combination approaches, with mixed evidence for each strategy. Of 10 papers identified in the secondary search, adolescent‐focused services and continuity of care with the same provider appeared to be effective in improving attendance at ANC and/or facility delivery, while home visits and group ANC had mixed results. Conclusions This review highlights the lack of evidence regarding effective strategies to improve retention in antenatal and/or postpartum care among adolescents and young WLHIV specifically, as well as a distinct lack of age‐disaggregated results in studies examining retention interventions for pregnant WLHIV of all ages. Identifying and prioritizing approaches to improve retention of adolescents and young WLHIV are critical for improving maternal and child health.

child transmission have been observed in younger women [8][9][10]. Pregnant adolescents and young women without HIV also have poor health outcomes compared to adult mothers, including late entry into ANC and higher risks of pre-term delivery, infants with low birthweight and maternal and infant mortality [11][12][13].
Reasons for these poor outcomes are likely multifactorial [14], and combination approaches may be needed to address the multiple overlapping risks that this group faces. Adolescence is a critical stage of biological and psychosocial development, and this transition period is further complicated by HIV and pregnancy [15]. Qualitative data suggest that barriers to care among adolescents and young women include stigma surrounding adolescent pregnancy and HIV infection, lack of social support, concerns about confidentiality and negative relationships with healthcare providers [14,16], highlighting the importance of addressing the multiple needs of this vulnerable group. With approximately 30% of new HIV infections in sub-Saharan Africa occurring in women younger than 25 years [17] and high rates of pregnancy during adolescence in this region [18], young pregnant and postpartum WLHIV are a priority population. However, standard PMTCT services have not been designed to address the unique needs of adolescents and young women [14]. For this vulnerable population to achieve optimal maternal and child health outcomes, evidence-based strategies are needed, including new models for differentiated service delivery [19].
The aim of this systematic review was to collate the available evidence on strategies to improve retention in antenatal and/or postpartum care among adolescents and young WLHIV. We also conducted a secondary review of strategies to increase attendance at ANC and/or facility delivery among pregnant adolescents, regardless of HIV status. The purpose of this secondary review was to identify approaches that have been effective in the general adolescent population and could be adapted for pregnant and postpartum adolescents and young WLHIV. Given that WLHIV are at increased risk of poor retention during pregnancy and the postpartum period in particular [4][5][6], we did not consider interventions tested among non-pregnant women.

| METHODS
This review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [20]. Search methods were discussed and finalized before beginning the searches, and the two searches were conducted on December 1, 2020. Searches were conducted in PubMed and Scopus and were restricted to English language articles published between January 2006 (corresponding with the date when lifelong ART was first recommended for pregnant and breastfeeding women based on disease staging [21]) and November 2020. No restrictions were placed on study design or geographical location for either search. The search terms were selected based on similar systematic reviews [22][23][24][25] and are presented in Table S1. Two of these reviews focused on pregnant women of all ages [22,23], and the others focused on non-pregnant adolescents and youth living with HIV [24,25]. Titles and abstracts of retrieved studies were merged and de-duplicated in Mendeley. Two independent reviewers screened titles and abstracts for relevance, followed by full-text review of all potentially relevant articles. Conference abstracts were not searched; letters, editorials, review articles and commentaries were excluded; and we did not contact authors for additional study details except for studies that included authors of this review in which case additional data were provided (where noted). Reference lists of all included studies and relevant review articles were screened for additional references. For both searches, data from eligible studies were extracted into a standardized Microsoft Excel table by the two independent reviewers.
We defined adolescents as women aged 10 to 19 years and young women as those aged 20 to 24 years. For the primary search of strategies to improve retention in pregnant adolescents and young WLHIV, the following criteria were used to screen abstracts: (1) study population included pregnant and/ or postpartum WLHIV; (2) study population included women aged <25 years; (3) study examined an approach to increase maternal retention in antenatal and/or postpartum care and (4) study included a comparison group. Although our aim was to review strategies to improve retention specifically among pregnant adolescents and young WLHIV, we reviewed full texts of all studies with any pregnant participants aged <25 years, including those with participants aged ≥25 years. This was done to identify all effective approaches to retaining WLHIV in antenatal and postnatal services, even those that have not been specifically targeted to younger women, as these approaches could potentially be adapted for younger women specifically. At the full-text review stage, we identified papers that reported age-disaggregated results (i.e. effective approaches for women aged <25 years), and we present results separately for papers with and without agedisaggregated results. Interventions that appear to be effective overall may not improve retention among younger women specifically, and identifying differences by age is critical to refine interventions for this vulnerable population.
For the secondary search of approaches aimed at retaining pregnant adolescents regardless of HIV status, the following criteria were used: (1) study population included pregnant adolescents; (2) study examined an approach to increase attendance at ANC and/or facility delivery; and (3) study included a comparison group. No restrictions were placed on study design or location. For this search, search terms such as "adolescent, " "teenager" and "youth" were used to limit the retrieved studies to adolescent populations, thereby identifying approaches that were examined specifically for adolescents. As above, we defined adolescents as women aged 10 to 19 years but included studies with women aged up to 22 years if authors identified these women as part of the adolescent population. Unlike the primary search, the secondary search was limited to studies examining outcomes among only adolescents as they are considered a high-risk population for poor health outcomes, whereas younger pregnant women (>20 years) have not been a focus.

| Studies presenting results among adolescents and young WLHIV specifically
The two studies that presented age-disaggregated results [26,27] and one that noted that age did not modify the study results [28] allow an examination of the effectiveness of these interventions among adolescents and young WLHIV specifically ( Table 1). The MCH-ART study, conducted in South Africa, examined the effectiveness of integrated care during the postpartum period, defined as the co-delivery of services at the same time and location. This randomized controlled trial was conducted among 471 women (mean age, 29 years) who initiated ART during pregnancy and opted to breastfeed [26]. During pregnancy, all women received integrated antenatal and HIV care within the Midwife Obstetric Unit, consistent with the local standard of care. Women were randomized immediately postpartum to either continued integrated care within the Midwife Obstetric Unit through the end of breastfeeding, including maternal ART care and routine infant care at the same visits, or the local standard of postpartum care (referral out to general adult ART services, with infant care provided at separate child health clinics). Women in the intervention arm were referred out to general adult ART services at a median of 32 weeks' postpartum, compared to 11 days in the control arm. The primary outcome was a composite endpoint of maternal retention in routine HIV care at 12 months' postpartum and viral suppression <50 copies/mL. Women in the intervention arm were significantly more likely to be retained in care and virally suppressed at 12 months' postpartum (77% vs. 56% in the standard of care arm; p < 0.001) [26]. Compared to women aged ≥25 years, younger women Records identified through database searching and other sources (n = 2,701) Records screened after removing duplicates (n = 1,662) Records excluded (n = 1,608) Full-text articles assessed for eligibility (n = 54) Full-text articles excluded, with reasons: -maternal retention not included as an outcome (n = 7) -no comparison (n = 7) -combined results with those from HIVnegative women (n=2) -combined results from previous articles (n=1) Studies included in qualitative synthesis (n = 37) were less likely to be retained in care and virally suppressed in both the intervention (61% vs. 82%) and the control arm (41% vs. 62%; MCH-ART study team, personal communication). In addition, the integrated care intervention effectively improved retention in care and viral suppression at 12 months' postpartum among women aged <25 years at enrolment (61% among women in the intervention arm vs. 41% among those in the control arm; MCH-ART study team, personal communication). Following completion of the MCH-ART trial, follow-up was extended to 36 to 60 months' postpartum [38]. Retention data were available for 450 women (96% of the original cohort), with no difference in retention in care by original allocation (63% of women in both arms; p = 0.885) and loss from care occurring soon after transfer out of the integrated clinic [38]. The second study, MIR4Health, was a randomized controlled trial of a lay counsellor-led combination intervention conducted among 340 Kenyan women who were aged 16 years or older (median age, 26 years; 41% <25 years of age) and had access to a cell phone [27]. Women randomized to the intervention were assigned a lay counsellor who provided individualized PMTCT health education, support for retention and adherence, telephonic and text message appointment reminders, and follow-up and tracking for missed clinic visits. The primary outcome was mother-infant attrition at six months' postpartum due to maternal or infant death or loss to follow-up. In the intervention arm, attrition of motherinfant pairs at six months' postpartum was significantly lower than in the standard of care arm (19% vs. 28% in the standard of care arm; p = 0.04). However, the intervention effect was modified by maternal age, with no significant difference in attrition across arms among women aged 16 to 24 years at enrolment (31% vs. 32% in the standard of care arm; p = 0.96) [27].
Finally, a randomized controlled trial conducted in Zambia examined the effect of an Option B+ Enhanced Adherence Package (BEAP) on initiation of ART and retention at 30 days [28]. The BEAP used community workers to provide follow-up for missed visits as well as other optional services, including individual counselling, home-based couples counselling and testing, male partner HIV testing and appointment reminders. The study enrolled 454 pregnant WLHIV aged 18 years or older (median age, 27 years) who had no previous ART use or had initiated ART during the past seven days. This study found no effect on retention after 30 days in an intention-to-treat analysis, although the intervention was found to significantly improve retention in a per-protocol analysis (92% vs. 80% in the control arm). The authors noted that age did not modify the intervention effect but did not state the age groups examined in these sub-group analyses [28].

| Learning from strategies to improve retention among adolescents and young WLHIV
Despite repeated calls for the development of evidence-based approaches to improve retention in antenatal and/or postpartum care among adolescents and young WLHIV, this review highlights the lack of evidence regarding effective strategies for these groups. We identified only two papers that reported results separately for adolescents and young WLHIV, and one that noted that age did not modify the intervention effect.
Two of the three studies were restricted to women aged 18 years or older at enrolment [26,28]. The MIR4Health combination intervention study improved mother-infant retention at six months' postpartum in the overall sample, but not among those aged 16 to 24 years at enrolment [27]. In contrast, age did not modify the effect of the BEAP combination intervention on retention after 30 days [28]. The MCH-ART study examined the effectiveness of integrated postpartum care, and showed that this intervention is effective in improving retention through 12 months' postpartum among women of all ages as well as women aged 18 to 24 years specifically [26]. This comparatively simple intervention may be effective through a range of mechanisms, including a lower burden of clinic visits as well as less HIV-related stigma through receiving ART in the Midwife Obstetric Unit. For young women who may receive limited financial support and have specific concerns related to stigma, these mechanisms may be particularly important. In addition, the MCH-ART intervention has been shown to be cost-effective [63], which is of particular importance in resource-limited settings, and we believe that this approach warrants further exploration.

WLHIV of all ages
The remaining 34 papers examined the effectiveness of interventions among WLHIV of all ages ( Table 2). Although this does not allow conclusions to be drawn about the effectiveness of these interventions among adolescents and young women, the effective interventions identified could be examined among younger women specifically or adapted where necessary. The most frequently examined approaches in these studies were the use of mobile health (mHealth) technologies, often in the form of SMS messaging; enhanced support for mothers, typically in the form of additional social support provided to individual mothers by peer or mentor supporters or during group counselling; active follow-up and tracing; and integrated PMTCT and ANC services (ART given in ANC clinics during pregnancy and sometimes postpartum) or integrated mother-infant care. As stated above, integrated care is defined as the co-delivery of services at the same time and location. These approaches were either examined as standalone interventions or as part of combination approaches, and we present results according to this distinction below.
Six studies investigated mHealth technologies as a standalone approach, with three investigating the use of text messages to improve retention. In South Africa, a study of oneway, twice-weekly maternal health information showed higher levels of attendance at ANC visits among women receiving text messages (82% of women in the intervention arm attended ≥4 ANC visits vs. 59% in the control arm) [48]. In Kenya, a randomized controlled trial of text messaging with the option of responding, calling or sending inquiry text messages showed higher levels of retention at 8 weeks' postpartum (20% vs. 12% in the control) [55], while a larger clusterrandomized controlled trial of the intervention showed no difference in retention at eight weeks' postpartum [36]. A study in South Africa investigating both text messaging and telephone calls providing visit reminders, motivational support and health information showed no differences in retention at either 10 weeks' or 12 months' postpartum [42]. In contrast, ART, antiretroviral therapy; cHTC, couple HIV testing and counselling; PMTCT, prevention of mother-to-child transmission; RCT, randomized controlled trial; SMS, short message service.  [59]. Finally, a randomized controlled trial in Uganda used text message reminders triggered by late dose-taking (monitored using Wisepill wireless pill monitors) but found no differences in retention through three months' postpartum [41]. Five papers examined enhanced support for mothers as a standalone intervention, typically provided by mentor mothers or within groups, with four demonstrating no significant effect on retention [29,30,39,40]. Two of these papers reported results from the same study of group-based peer support in South Africa and showed no improvement in attendance at either antenatal [39] or postpartum care [40]. In contrast, a Nigerian study reported improved retention at six months' postpartum related to an enhanced support intervention delivered by mentor mothers (62% in the intervention arm vs. 25% in the control arm) [60]. Two studies examined active follow-up and tracing as a standalone intervention, with a before-after intervention study of a defaulter tracing system showing no effect on retention at delivery in Zimbabwe [44], while a grouprandomized controlled trial of appointment-and communitybased tracking systems in Tanzania showed a decrease in missed clinic visits in the intervention arm (from 37% to 34%) and an increase in the control arm (from 39% to 46%) [58]. Finally, besides the MCH-ART study and follow-up study described above, a cluster-randomized controlled trial in Kenya examined the effect of integrated services as a standalone intervention, but showed no effect of the intervention on retention during the first six months after entering care [43].
The remaining 12 studies examined combination interventions but did not report the effect of each component of the intervention. For those demonstrating effectiveness, it is thus not possible to identify the specific component(s) that improved retention. In addition, four of these studies used a continuous quality improvement approach, where the intervention was adapted and refined throughout the study period, which also does not allow for conclusions regarding the effectiveness of specific components. These included one study demonstrating no effect on retention in care through six months in Nigeria [37], two studies demonstrating improved retention in care in each of Kenya [51] and Tanzania [53] and one study from Uganda where retention in care at 12 months' postpartum improved in demonstration but not at scale-up facilities [49]. Three combination intervention studies showed no significant effect on retention. These included two studies from Tanzania, one of which tested an intervention package which included integrated services [31], and the other which tested an intervention package including adherence counselling and defaulter tracing by community health workers [35]. Finally, a 3-arm cluster-randomized controlled trial of (1) integrated services and (2) integrated services as well as SMS reminders, compared to the control, demonstrated no effect of either intervention on retention in care through 12 months' postpartum in Malawi [34].
Alongside the MIR4Health, BEAP and continuous quality improvement studies described above, five other studies demonstrated the effectiveness of combination interventions, but it is not possible to identify the most effective intervention components. These components included integrated care [47,50], active follow-up and tracing [50,54,56,57] and enhanced support through counselling [54,56] or support groups [57]. The duration of follow-up in most of these studies was short, ranging from increased retention during the antenatal period [50,56] to increased retention at six weeks' [54] and 12 weeks' postpartum [47]; only one study reported outcomes through two years [57]. This 3-arm cluster-randomized controlled trial examined (1) facility-based peer support and (2) community-based peer support, with support including individual-and group-based support as well as follow-up for missed visits, versus the standard of care. No difference in retention was observed at 12 months after ART initiation, but retention after 24 months was significantly higher in both intervention arms compared to the control (80% and 83% vs. 66%) [57].

| Learning from strategies to improve retention among WLHIV of all ages
Taken together, studies examining standalone interventions among WLHIV of all ages demonstrated mixed evidence for the effectiveness of these approaches on retention, including mHealth technologies (with three of six studies showing no effect on retention) and enhanced support provided by mentor mothers or within groups (with four of five studies showing no effect). One of two studies demonstrated the effectiveness of active follow-up and tracing, and one study showed no effect of integrated services during the first six months after entering care, in contrast to the MCH-ART study described above. The 12 studies examining combination interventions similarly showed mixed effectiveness and, as noted above, it is not possible to identify the effectiveness of individual components of these interventions.
Previous reviews have similarly shown mixed evidence for approaches to improve retention among WLHIV [22], including the effectiveness of integrated services [64,65]. Although our review found mixed evidence for the effectiveness of active follow-up and tracing, a previous review found that outreach services improved retention in PMTCT care among women of all ages [66]. In addition, peer support services and support groups have been found to improve retention in PMTCT care among women of all ages [66] and among men and WLHIV [67]. However, although peer support for adolescents living with HIV has been described as a promising approach to improve retention, there are few rigorous evaluations of this approach [68]. A recent cluster-randomized controlled trial in Zimbabwe demonstrated higher levels of viral suppression among non-pregnant adolescents who received a peer-led community-based support intervention (the Zvandiri intervention) compared to the standard of care [69], but evaluations among pregnant adolescents specifically are needed. Finally, it should be noted that interventions that are universally applied to WLHIV of all ages may result in less stigma compared to interventions for adolescents and young women specifically, but interventions that are effective overall should be examined in this vulnerable population.

| Secondary search: strategies to improve retention among adolescents, regardless of HIV status
This secondary search yielded 8977 abstracts (Figure 2). After removing duplicates (n = 1399) and non-relevant titles and abstracts (n = 7560), 18 full-text articles were assessed for eligibility. Of these, eight were excluded for not evaluating attendance as a standalone outcome (n = 5), not targeting adolescents specifically (n = 2), and for reporting combined results from previously published papers (n = 1), resulting in 10 papers that examined approaches to increase attendance at ANC and/or facility delivery among adolescents (Table 3). Two major differences between the results of the two searches are evident. In contrast to the primary search, most of the 10 studies identified in the secondary search were conducted in high-income countries, including five in the United States [70][71][72][73][74]. In addition, cluster-randomized controlled trials were the most common study design identified in the primary search, but only three of the 10 studies identified in the secondary search randomized participants [72,75,76]. Seven of the studies reported that the intervention was associated with improved attendance at ANC appointments [70,71,74,75,[77][78][79]; only one study examined facility delivery as a standalone outcome but the intervention was not effective in improving this outcome [79]. Although these studies do not allow conclusions to be drawn about the effectiveness of these interventions among adolescents and young WLHIV, the effective interventions identified could be examined among WLHIV specifically or adapted where necessary.
Three studies examined adolescent-focused services as an approach to improve adolescent outcomes [70,77,78]. This typically included multidisciplinary care provided by specifically trained providers and delivered within a young women's clinic [78] or as part of an outreach programme [77]. In addition, a study based at a medical home that offered a range of services to pregnant adolescents within a single location in the United States examined the provision of trauma-informed pregnancy care, including psychological and psychiatric services [70]. These studies reported that the interventions resulted in higher proportions of adolescents accessing ANC during the first trimester of pregnancy [77], an increase in the median number of antenatal visits attended [70] and decreased likelihood of attending <5 antenatal visits [78].
Continuity of care with the same provider was another common strategy [70,74,78]. In one study conducted in Australia, care was delivered by the same midwife through pregnancy, birth and postpartum, and women had 24-hour telephone access to their midwife throughout; this approach decreased the likelihood that women would attend <5 ANC visits [78]. The other two studies conducted in the United States reported that continuity of care by the same provider increased the number of ANC visits attended [70,74]. In one study, adolescents received care from the same clinician throughout pregnancy within an integrated service [70].
Three studies examined home visits, including one in the United States in which pregnant adolescents received two home visits per month until delivery, with planned receipt of at least six visits [71]. In this study, home visits were conducted by public health nurses and social workers for the purpose of assessing adolescents' needs, supporting adolescents to attend antenatal visits, and providing referrals for additional services, and home visits were associated with an increased number of antenatal visits attended [71]. In the second study, conducted in England, the intervention consisted of up to 64 home visits from family nurses between early pregnancy and two years' postpartum. This intervention had no effect on attendance at ANC visits and was noted to be extremely costly [76]. The third study was conducted in India, and demonstrated that home visits by community health workers combined with referrals to care and behaviour change counselling increased the use of ANC services among married adolescent girls [79].
Finally, two studies examined group ANC for pregnant adolescents. A retrospective cohort study conducted in the United States reported that the proportion of women who attended all ANC visits was higher among those attending group ANC (62%) compared to those accessing single-provider (52%) or multi-provider care (41%) [74]. In contrast, a clusterrandomized controlled trial in the United States demonstrated Full-text articles excluded, with reasons: -attendance not a standalone outcome (n = 5) -not specific to adolescents (n = 2) -combined results from previous articles (n = 1) Studies included in qualitative synthesis (n = 10)  no association between group ANC and the number of ANC visits attended, although women attending group ANC visits had more favourable outcomes including fewer infants who were small for gestational age [72]. Of note, this study reported substantial challenges to attendance at group ANC, with adolescents attending half of the group visits on average and one in five attending no group visits [72].
3.7 | Learning from strategies to improve retention among adolescents, regardless of HIV status Taken together, we found mixed evidence for strategies to improve ANC attendance among adolescents. Interventions shown to be effective included providing multidisciplinary adolescent-focused services, often provided at the same location; continuity of care with the same provider; and home visits (with two of three studies demonstrating that the intervention was effective); two studies examined group ANC for pregnant adolescents, with mixed results and notable challenges in attendance. Alongside the results of the primary search, our review of interventions among adolescents regardless of HIV status also failed to yield clear insights into evidence-based strategies that could inform efforts to address the needs of pregnant adolescents and young WLHIV. As noted above, a major difference between the results of this search and the primary search is the setting in which studies were conducted. All studies identified as part of the primary search were conducted in sub-Saharan African countries, while most studies identified for the secondary search of general adolescent populations were conducted in high-income countries. The approaches found to be effective in these highincome settings may not be applicable to low-resource settings with a high HIV burden. Many of the approaches identified were resource-intensive, including one that provided up to 64 home visits to women during pregnancy and the postpartum period [76], potentially limiting their feasibility in resource-limited settings. Similarly, continuity of care with the same provider may not be feasible in overburdened PMTCT programmes in resource-limited settings.

| Limitations
Several limitations of this review should be noted. First, it was beyond the scope of this work to contact authors and request that they provide age-disaggregated data. In addition, we did not assess the risk of bias in the studies included. The potential for confounding in non-randomized studies is a major concern, and the potential for publication bias cannot be excluded. We were restricted in our ability to report details of the studies included, in particular the behavioural theories in which they were grounded, due to this information not being routinely presented in papers. Finally, critiquing the design and content of interventions was beyond the scope of this review.

| CONCLUSIONS
Pregnant adolescents and young WLHIV are a vulnerable population with multiple complex needs, and prioritizing this group for intensive retention interventions could improve outcomes for both mothers and infants [14]. However, our review highlights the lack of evidence-based approaches to improve retention in care in this population, with clear implications for future research. In particular, this review highlights the lack of reporting of age-disaggregated results in studies examining retention interventions for pregnant WLHIV. Previous reviews have similarly highlighted this issue in studies examining approaches to improve retention among non-pregnant adolescents and adults [25]. To address this issue, journal reviewers and editors could consider routinely requesting agedisaggregated analyses of intervention effects in published manuscripts. Furthermore, there is a critical need for studies to include adequate numbers of adolescents and young women in study samples to allow for age-disaggregated analyses and to support research in the area of intervention development and evaluation for adolescent and young pregnant WLHIV specifically. Given the continued high incidence of both HIV and pregnancy among adolescents and young women in sub-Saharan Africa, prioritizing the development and evaluation of approaches to improve retention in this population could help improve both maternal and child health.