The timeliness of early infant diagnosis (EID)—HIV testing of exposed infants 6–8 weeks after birth—and proportion of infants testing positive are cardinal indicators of prevention of mother-to-child transmission (PMTCT) program success.1 However, in routine program settings, completion of EID is assessed at individual facilities, from information across several paper-based registers, which may be incomplete and/or inaccurate.2,3
New approaches are needed to improve confidence in estimates of EID completion and survival of HIV-exposed infants to guide on-going quality improvement and inform national modeling estimates. These become increasingly important as countries such as Zimbabwe, with an HIV prevalence of 16.0% among women,4 seek to validate elimination of mother-to-child transmission.5
In this study, we randomly sampled lost-to-follow-up (LTFU) HIV-positive mother, HIV-exposed baby pairs (MB), pairs from routine health facility-based data. We actively traced MB pairs in the community to assess infant and maternal survival and EID uptake. Finally, we incorporated these findings to correct estimates of HIV-exposed infant mortality and timely EID in the entire clinic population of MB pairs using a probability weight. Although this sampling-based approach has been used in clinic-based cohorts of adult patients in HIV treatment programs and found to alter estimates of retention and mortality,6,7 our study represents the first time the approach has been used in PMTCT programs. We reflect on key lessons from our research on program progress and persisting gaps for strengthening EID coverage on the path to elimination of MTCT in Zimbabwe.
We defined LTFU as no documented sample taken for EID polymerase chain reaction testing among infants of HIV-positive antenatal care (ANC) patients within 90 days of delivery. At the time of study, Zimbabwe HIV care and treatment guidelines recommended HIV testing of all exposed children 6 weeks postnatal.8 Our sample included all women with HIV who accessed ANC at public sector health facilities in Mashonaland East Province (referral sites excluded), using a multistage clustered survey sampling approach.
Phase I: Register Tracing
We selected 5 clinics from each of the 9 districts within Mashonaland East Province based on probability proportional to size, using program data on HIV-positive women accessing ANC over the previous year. In the absence of interfacility-linked electronic patient monitoring systems, we used a combination of individual identifiers including clinic-allocated PMTCT number, patient name, and date of birth to trace each identified HIV-positive pregnant woman in ANC and her HIV-exposed infant through 5 paper-based registers (ANC, delivery, HIV-exposed infants, clinic registers, HIV infant diagnosis, and EID lab request form booklets) to determine LTFU for EID status. We targeted a random sample of at least 10% of lost MB pairs based on practical considerations regarding limitations of completeness and accuracy of patient information in multiple paper-based registers9 and available resources to intensively trace those identified as LTFU for EID over the study timeframe. Based on known challenges in tracing defaulters from HIV care and treatment,10–12 we over-sampled by 100% to ensure we met our target sample.
Phase II: Tracing of LTFU MB Pairs
In the second stage of sampling, a random sample of patients identified as lacking documented EID in Stage 1 was selected for active tracing to determine outcomes through direct patient interviews. We trained MOHCC village health workers as ascertainers due to their familiarity with the surrounding community and existing role to support health facilities with defaulter tracing.13,14
Village health workers were assigned selected LTFU patients residing within their geographical catchment area for tracing. If the individual was unable to be located, mother and infant vital status was obtained from informants (friends, neighbors, or relatives); however, to maintain confidentiality, informants were not asked any specific questions related to HIV services. Successfully traced and consenting mothers were interviewed using a standardized questionnaire to ascertain survival outcomes, EID status and timing, and reasons for failure to obtain EID or for self-transfer of care to another facility. Analyses were conducted using Stata v.13.1. This study was approved by the Medical Research Council of Zimbabwe (MRCZ/A/1844) and MOHCC authorities.
First, we estimated EID incidence based on clinic registry data only, using facility weights inverse to the probability of clinic selection to yield an estimate of EID completion based purely on paper registers kept at the facilities. Facility and individual factors on documented completion of EID were explored using Poisson regression with robust standard errors adjusted for confounders to estimate risk ratios (RRs).15,16
Second, we generated corrected estimates of EID completion using data from clinic registries as well as data ascertained through interviews with a random sample of MB pairs missing EID in facility-based registers.7 Deaths before EID (including fetal deaths) were considered as competing risks. To obtain an estimate corrected for outcomes not captured in the facility registers, we used additional sampling weights inverse to the probability of being successfully sought to represent all LTFU HIV-infected pregnant women at each clinic.
Among women with no documented EID in facility registers who were traced and interviewed, reasons for silent transfer to a different clinic from ANC for EID and for not bringing HIV-exposed child in for HIV testing (no EID) were grouped into 4 categories informed by a socioecological framework17,18: structural; clinic-based; psychosocial or patient-related; or medical factors.
For the weighted mortality estimate, the traced patients contributed time from the delivery date to the date of the patient interview or the date of the death of the infant at any time prior (ie, all deaths including those later than 90 days were included in the mortality estimate); patients with documented EID contributed time from the date of delivery to the date of the 6-week postnatal visit, which is the latest date at which we have confirmed vital status for those infants. Kaplan–Meier methods were used to estimate mortality.
Phase I: Facility Register Data
Among 18,065 women registered for ANC between April 2012 and May 2013, 2651 (14.7%) were HIV positive and 31.2% [95% confidence interval (CI): 29.5% to 33.0%] had documented uptake of EID for their infant within 3 months of delivery in clinic registers. After adjustment for register information and site characteristics, factors associated with documented EID completion included earlier gestational age at presentation (RR: 0.97 per 2 weeks; 95% CI: 0.95 to 0.99; P = 0.013), later calendar time of ANC presentation (RR: 1.04 per 30 days; 95% CI: 1.02 to 1.06, P = 0.011), and smaller site volume (RR: 1.85 1-200 ANC patient volume, 1001-1500 volume referent; 95% CI: 1.44 to 2.38, P < 0.001).
Phase II: Community Tracing LTFU
Among 1652 mother–baby pairs identified as LTFU with any documented locator information, a random sample of 643 (38.9%) was selected for community tracing between March 2015 and May 2015. In 371/643 (57.7%), updated vital or EID status information was obtained (22.5%; 371/1652 of the total LTFU sample). The primary reason for failure to locate clients was insufficient location information. Among 371 successfully traced patients, 256 (69.0% of located) mothers were interviewed directly on infant vital status and EID uptake, and for the remaining 115 (31.0%), informants were interviewed regarding MB pair survival outcomes (not HIV-related) (Fig. 1). Among the 371 mother–baby pairs for whom vital status outcomes were determined, 66 infants (17.8%; 95% CI: 14.0% to 22.1%) and 18 mothers (4.9%; 95% CI: 2.9% to 7.6%) were found to be deceased.
Most mothers interviewed (190/256; 74.2%; 95% CI: 68.4% to 79.5%) reported their infant had received HIV testing; although fewer than half received EID testing before 3 months of age (92/190; 48.4%; 95% CI: 41.1% to 55.8%). Our corrected estimate following tracing resulted in a cumulative incidence of EID with death as a competing risk of 60.0% (95% CI: 58.7% to 61.3%). We estimated a cumulative incidence of mortality among HIV-exposed infants at 3 months of 3.9% (95% CI: 3.4% to 4.4%) and at 1 year of 7.7% (95% CI: 4.7% to 13.5%). Among the 66 infants with no EID at any time, the most frequently cited reason for failure to have EID was “my child died” (36.6%; 95% CI: 25.8% to 49.0%), Figure 2A. Among infants with timing of death ascertained, most did not survive to the age of recommended EID testing (6 weeks) (26/42; 61.9%). Among mothers of living infants, “I didn't know I should have my child tested” was the most frequently cited reason for no EID (16/45; 35.6%). Relocation to a different area (21.8%; 95% CI: 13.2% to 32.6%) and transport being easier/cheaper at new clinic (20.5%; 95% CI: 12.2% to 31.1%) were the most commonly reported reasons for silent transfer (Fig. 2B).
In a representative sample of Mashonaland East Province of Zimbabwe, after tracing a sample of MB pairs identified as LTFU for EID, our corrected estimate of EID almost doubled (from 31.2% to 60.0%). These findings underscore the risk of equating LTFU of MB pairs in health information systems with disengagement from care where such systems do not facilitate electronic, longitudinal tracing of service uptake within and between health facilities.19 MOHCC is currently expanding efforts to strengthen facility-based documentation, retention monitoring, and longitudinal outcome reporting through use of the Mother-Baby Pair Register to track longitudinal outcomes of MB Pairs, together with appointment diary systems.20 As Zimbabwe transitions toward electronic health record systems, we indicate the value of sampling-based methods for improving accuracy of EID coverage estimates and informing current progress toward EMTCT validation.21
We found high mortality among HIV-exposed infants, with “my child died” being the most frequently cited reason for no EID testing. Passive facility-based monitoring may underestimate true mortality by up to 80%,22 and under-reporting of infant mortality in high HIV burden settings is acknowledged to bias child mortality estimates downward.23,24 Our findings reinforce the need to act early among HIV-positive mother–exposed infant pairs at high risk of defaulting or adverse clinical outcomes with enhanced PMTCT interventions.25,26 Subsequent MOHCC guidelines recommending prioritization of viral load monitoring for pregnant and lactating mothers, birth testing for high-risk infants,20,27 and case-based surveillance of infants testing HIV positive20,28 are intended strengthen program evidence and action among high-risk MB pairs. Our findings emphasize that monitoring of facility-level implementation fidelity, data quality, and robust analysis of resulting data will be central to realizing the benefit of such efforts for improved PMTCT program strategies and impact.
Finally, we not only improved our understanding of “true EID” rates but also developed evidence on factors influencing timely infant HIV testing for informing quality improvement. At patient-level, the predominance of psychosocial reasons (“I didn't know”) for failure to uptake EID among mothers of living infants emphasizes need to provide information about the importance of timely infant HIV testing during the initial engagement in ANC and continued emphasis at every subsequent visit. Reasons for silent transfer are consistent with studies of adult ART retention and highlight the role of structural29 and clinic-based factors30 for optimizing retention in care along the PMTCT cascade. Findings have guided PMTCT program planning and actions including strengthening of patient education and problem solving, appointment monitoring, active follow-up, and outcome documentation in Zimbabwe and other settings.20,31–33
Although oversampling enabled achievement of our targeted sample of at least 10% of MB pairs LTFU for EID (we traced 22%), our findings highlight need to strengthen routine documentation and other forms of observational data to ascertain uptake of services and mortality.34 We acknowledge nonresponse due to incomplete tracing details may introduce some uncertainty in the resulting estimates. However, our findings of lower LTFU and higher mortality after tracing are concordant with larger, more robust LTFU studies.35 In the absence of effective interfacility electronic patient-monitoring systems, we demonstrate the value of leveraging routine observational data in “real-life” program settings for improving accuracy of estimates, identifying bottlenecks and guiding programmatic actions at the local level.34,36,37
Our findings suggest that routinely reported EID completion in public health information systems may be substantially underestimated. Accurate determination of mother–infant pair outcomes in PMTCT programs is complicated by resource-constrained health information systems that involve multiple paper-based registers, lack of unique identifiers, and challenges with completeness and accuracy of information recorded. We demonstrate the value of sampling-based approaches in pediatric HIV research for providing important, context-based evidence for policy and programs.
The authors are grateful to the Zimbabwe Ministry of Health and Child Care, participating health facilities, village health workers, and study participants. The opinions expressed in this article do not necessarily reflect the views or policies of PEPFAR, USAID, NIH, CFAR, or Wellcome Trust.
1. Ghadrshenas A, Ben Amor Y, Chang J, et al. Improved access to early infant diagnosis
is a critical part of a child-centric prevention of mother-to-child transmission agenda. AIDS. 2013;27(suppl 2):S197–S205.
2. Mate KS, Bennett B, Mphatswe W, et al. Challenges for routine health system data management in a large public programme to prevent mother-to-child HIV transmission in South Africa. PLoS One. 2009;4:e5483.
3. Gourlay A, Wringe A, Todd J, et al. Challenges with routine data sources for PMTCT programme monitoring in East Africa: insights from Tanzania. Glob Health Action. 2015;8:29987.
4. Ministry of Health and Child Care (MOHCC). Zimbabwe Population-Based HIV Impact Assessment (ZIMPHIA) 2015-16: First Report. Harare, Zimbabwe: MOHCC; 2017.
5. WHO. Global Guidance on Criteria and Processes for Validation: Elimination of Mother-To-Child Transmission of HIV and Syphilis. 2nd ed. Geneva, Switzerland: WHO; 2017.
6. Geng EH, Bangsberg DR, Musinguzi N, et al. Understanding reasons for and outcomes of patients lost to follow-up in antiretroviral therapy programs in Africa through a sampling-based approach. J Acquir Immune Defic Syndr. 2010;53:405–411.
7. Geng EH, Odeny TA, Lyamuya RE, et al. Estimation of mortality among HIV-infected people on antiretroviral treatment in East Africa: a sampling based approach in an observational, multisite, cohort study. Lancet HIV. 2015;2:e107–116.
8. MOHCC. Guidelines for Antiretroviral Therapy for the Prevention and Treatment of HIV in Zimbabwe. Harare, Zimbabwe: NMTPAC and MOHCC; 2013.
9. Matyanga P, Webb K, Patel D, et al. Drowning in registers: task-oriented job analysis of rural health nurses in Zimbabwe in the era of Option B+ [WEPE451]. Presented at: 20th International AIDS Conference (AIDS 2014); July 20–25, 2014; Melbourne, Australia.
10. Vogt F, Ferreyra C, Bernasconi A, et al. Tracing defaulters in HIV prevention of mother-to-child transmission programmes through community health workers: results from a rural setting in Zimbabwe. J Int AIDS Soc. 2015;18:20022.
11. Tweya H, Gugsa S, Hosseinipour M, et al. Understanding factors, outcomes and reasons for loss to follow-up
among women in Option B+ PMTCT programme in Lilongwe, Malawi. Trop Med Int Health. 2014;19:1360–1366.
12. Tweya H, Feldacker C, Estill J, et al. Are they really lost? “true” status and reasons for treatment discontinuation among HIV infected patients on antiretroviral therapy considered lost to follow up in Urban Malawi. PLoS One. 2013;8:e75761.
13. Schmitz K, Basera TJ, Egbujie B, et al. Impact of lay health worker programmes on the health outcomes of mother-child pairs of HIV exposed children in Africa: a scoping review. PLoS One. 2019;14:e0211439.
14. MOHCC. The Village Health Worker Handbook. Harare, Zimbabwe: MOHCC; 2010.
15. Geng EH, Hare CB, Kahn JO, et al. The effect of a “universal antiretroviral therapy” recommendation on HIV RNA levels among HIV-infected patients entering care with a CD4 count greater than 500/muL in a public health setting. Clin Infect Dis. 2012;55:1690–1697.
16. Zou G. A modified poisson regression approach to prospective studies with binary data. Am J Epidemiol. 2004;159:702–706.
17. Geng EH, Odeny TA, Lyamuya R, et al. Retention
in care and patient-reported reasons for undocumented transfer or stopping care among HIV-infected patients on antiretroviral therapy in Eastern Africa: application of a sampling-based approach. Clin Infect Dis. 2016;62:935–944.
18. Rachlis B, Naanyu V, Wachira J, et al. Identifying common barriers and facilitators to linkage and retention
in chronic disease care in western Kenya. BMC Public Health. 2016;16:741.
19. Myer L, Phillips TK. Beyond “option B+”: understanding antiretroviral therapy (ART) adherence, retention
in care and engagement in ART services among pregnant and postpartum women initiating therapy in sub-saharan Africa. J Acquir Immune Defic Syndr. 2017;75(suppl 2):S115–S122.
20. MOHCC. National Strategic Plan for Eliminating New HIV Infections in Children and Keeping Mothers and Families Alive (2014–2018). Harare, Zimbabwe: Ministry of Health and Child Care; 2014.
21. WHO. Validation of EMTCT of HIV and/or Syphilis: Tools and Checklists for In-Country Evaluation of Four Required Components. Geneva, Switzerland: WHO; 2017.
22. Yiannoutsos CT, An MW, Frangakis CE, et al. Sampling-based approaches to improve estimation of mortality among patient dropouts: experience from a large PEPFAR-funded program in Western Kenya. PLoS One. 2008;3:e3843.
23. Hug L, Alexander M, You D, et al. National, regional, and global levels and trends in neonatal mortality between 1990 and 2017, with scenario-based projections to 2030: a systematic analysis. Lancet Glob Health. 2019;7:e710–e720.
24. Hallett TB, Gregson S, Kurwa F, et al. Measuring and correcting biased child mortality statistics in countries with generalized epidemics of HIV infection. Bull World Health Organ. 2010;88:761–768.
25. Bhardwaj S, Carter B, Aarons GA, et al. Implementation research for the prevention of mother-to-child HIV transmission
in sub-Saharan Africa: existing evidence, current gaps, and new opportunities. Current HIV/AIDS Rep. 2015;12:246–255.
26. Myer L, Essajee S, Broyles LN, et al. Pregnant and breastfeeding women: a priority population for HIV viral load monitoring. PLoS Med. 2017;14:e1002375.
27. MOHCC. Guidelines for Antiretroviral Therapy for the Prevention and Treatment of HIV in Zimbabwe, 2016. Harare, Zimbabwe: NMTPAC and MOHCC; 2016.
28. Webb K, Chitiyo V, Page-Mtongwiza S, et al. Too Little, Too Late: Early Lessons on MTCT Risk from HIV Positive Infant Case Investigation in Zimbabwe . Paper presented at: International Workshop on HIV Pediatrics; July 19–20, 2019; Mexico City, Mexico.
29. McCoy SI, Buzdugan R, Mushavi A, et al. Food insecurity is a barrier to prevention of mother-to-child HIV transmission
services in Zimbabwe: a cross-sectional study. BMC Public Health. 2015;15:420.
30. Zanolini A, Sikombe K, Sikazwe I, et al. Understanding preferences for HIV care and treatment in Zambia: evidence from a discrete choice experiment among patients who have been lost to follow-up. PLoS Med. 2018;15:e1002636.
31. Nance N, Pendo P, Masanja J, et al. Short-term effectiveness of a community health worker intervention for HIV-infected pregnant women in Tanzania to improve treatment adherence and retention
in care: a cluster-randomized trial. PLoS One. 2017;12:e0181919.
32. Webb K, Chitiyo V, Patel D, et al. The Action Birth Card: Evaluation of an Innovative Goal-Setting Tool to Increase Demand and Uptake of Underutilized Services along the PMTCT Cascade [MOPEC466]. 8th IAS Conference on HIV Pathogenesis, Treatment and Prevention (IAS 2015); 2015; Vancouver, Canada.
33. MOHCC. The Plan for Elimination of Mother to Child Transmission of HIV & Syphilis in Zimbabwe 2018–2022. Harare, Zimbabwe: Ministry of Health and Child Care; 2018.
34. Sohn AH, Judd A, Mofenson L, et al. Using observational data to inform HIV policy change for children and youth. J Acquir Immune Defic Syndr. 2018;15:S22–S26.
35. McMahon JH, Elliott JH, Hong SY, et al. Effects of physical tracing on estimates of loss to follow-up
, mortality and retention
in low and middle income country antiretroviral therapy programs: a systematic review. PLoS One. 2013;8:e56047.
36. Rice B, Boulle A, Baral S, et al. Strengthening routine data systems to track the HIV epidemic and guide the response in sub-Saharan Africa. JMIR Public Health Surveill. 2018;4:e36.
37. Mark D, Geng E, Vorkoper S, et al. Making implementation science work for children and adolescents living with HIV. J Acquir Immune Defic Syndr. 2018;78(suppl 1):S58–S62.