Measuring retention in care for HIV-positive pregnant women in Prevention of Mother-to-Child Transmission of HIV (PMTCT) option B+ programmes: the Mozambique experience

Background: Failure of retention of HIV-positive pregnant women on ART leads to increased mortality for the mother and her child. This study evaluated different retention measures intended to measure women’s engagement along the continuum of care for PMTCT option B+ services in Mozambique. Methods: We compared ‘point’ retention (patient’s presence in care at 12-months post-ART initiation or any time thereafter) to the following definitions: alive and in care at 12 months post-ART initiation (Ministry of Health); attendance at a health facility up to 15 months post-ART initiation (World Health Organisation); alive and in treatment at 1, 2, 3, 6, 9, and 12 months post-ART initiation (Inter-Agency Task Team); and alive and in care at 12 months post-ART initiation with ≥75% appointment adherence during follow-up (i.e. ‘appointment adherence’ retention) or with ≥75% appointment done on-time during follow-up (i.e. ‘on-time adherence’ retention). Kaplan-Meier survival curves were produced to assess variability in retention rates. We used ‘on-time adherence’ retention as our reference to estimate sensitivity, specificity, and proportion of misclassified patients. Results: Considering the ‘point’ retention definition, 16,840 HIV-positive pregnant women enrolled in option B+ PMTCT services were identified as ‘retained in care’ 12 months post-ART initiation. Of these, 60.3% (95% CI 59.6– 61.1), 84.8% (95% CI 84.2–85.3), and 16.4% (95% CI 15.8–17.0) were classified as ‘retained in care’ using MOH, WHO, and IATT definitions, respectively, and 1.2% (95% CI 1.0–1.4) were classified as ‘retained in care’ using the ‘ ≥75% on-time adherence’ definition. All definitions provided specificity rates of ≥98%. The sensitivity rates were 3.0% with 78% of patients misclassified according to the WHO definition and 4.3% with 54% of patients misclassified according to the MOH

4 definition. The 'point' retention definition misclassified 97.6% of patients. Using IATT and 'appointment adherence' retention definitions, sensitivity rates (9.0% and 11.7%, respectively) were also low; however, the proportion of misclassified patients was smaller (15.9% and 18.3%, respectively). Conclusion: More stringent definitions indicated lower retention rates for PMTCT programmes. Policy makers and programme managers should include attendance at follow-up visits when measuring retention in care to better guide planning, scaling up, and monitoring of interventions.

Background
In 2015, WHO recommended that all HIV-positive pregnant women be provided with lifelong antiretroviral therapy (ART) regardless of their CD4 count and WHO staging, also called "Option B+" (1). Compared to previous prevention of mother-to-child transmission (PMTCT) of HIV regimen such as Option A (short term anti-retroviral prophylaxis) or Option B (triple anti-retroviral until end of breastfeeding) (2), Option B+ by providing the same antiretroviral triple-combination to all HIV-positive pregnant women for life has the advantage of lowering the risk of HIV transmission to male partners since being on ART will make women less infectious, reducing the risk of MTCT of HIV in subsequent pregnancies while reducing the risk of child infection to less than 5% in resources limited settings (3).
Retention in PMTCT care has both individual and public health implications. For mother-infant pair, being in care contributes to higher maternal ART adherence, better viral suppression, and ensures better post-natal care, including full ART prophylaxis and complete infant testing for HIV-exposed infants (HEI) (4,5). In contrast, failure to be retained in care can lead to delayed or inconsistent use of antiretroviral medications, higher risk of maternal viral load failure, and increased morbidity and mortality for the mother and her child (6)(7)(8). In addition, recent publications have reported an upward trend in acquired drug resistance to first-line ART across low-and middle-income countries, principally due to sub-optimal viral suppression (9)(10)(11). Therefore, non-achievement of viral suppression can lead to a higher risk of HIV transmission and secondary infections with acquired drug resistance strains and is, therefore, considered a public health threat (12).
When referring to retention in care, publications often consider 'point' retention, which is defined in relation to a patient's presence in care at a certain time point (13). For women and children enrolled in PMTCT services, attendance at a clinic at a certain time point is often considered to be full retention over this time period.
While such a simple definition is useful, it has been demonstrated that between one-and two-thirds of HIV-infected adult patients are not in regular care (14).
Therefore, regularity of attendance is an important parameter to consider if the quality of engagement in care is to be evaluated (15).
To address shortcomings in this simple definition, attempts have been made to better measure retention in care in PMTCT services (16,17). In 2014, a World Health Organisation (WHO) monitoring and evaluation working group published consolidated strategic information guidelines and proposed that retention in care of HIV-infected pregnant and breastfeeding women was equivalent to attendance at a health facility at 12-months post-initiation of ART, or at any time up to 3 months later (16). The 12-month time point was agreed to align with the adult ART monitoring guidelines; however, it fails to address HIV transmission risk beyond 12 months due to prolonged breastfeeding. In 2015, the Inter-Agency Task Team (IATT) on Children and HIV and AIDS defined maternal retention as 'the proportion of HIV-positive pregnant and/or breastfeeding women on ART alive and in treatment at 1, 2, and 3 months post-ART initiation (early retention), and then at 6, 9, and 12 months post-ART initiation', where retention was considered as a continuous engagement (17). Despite such efforts, there is no consensus on what constitutes 'retention in care', and no gold standard has been determined to date. In 2010, Mugavero et al. provided a synopsis of five commonly used definitions of retention in HIV care and treatment services, ranging from a simple count of the number of missed visits to a more complex medical visit performance measure that incorporated elements of appointment consistency and gaps in care (14). While that study provided important insights on methodological and conceptual strengths and limitations of each definition, no comparative analysis of these definitions has been undertaken. We hypothesised that 'point' retention' might overestimate the level of engagement in care of women enrolled in PMTCT programmes giving a false image of clinical and programmatic success. In a high HIV-burden country such as Mozambique, women's engagement should be more precisely assessed in order to identify windows of opportunity for possible improvement throughout the PMTCT care continuum and to better account for the number of children in need of ART (18).
In this study, we aimed to answer the following questions: what is the effect of

Study design
This was a retrospective cohort study involving secondary analysis of routinely collected data.

Setting
Data were extracted from all sites with available electronic patient-level databases

Comparing definitions
The 'point' retention definition was compared against the following five definitions: (17), 'appointment adherence' retention (14), and 'ontime adherence' retention (22) ( Table 1). Similar to the 'point' retention definitions, the MOH and WHO retention definitions assess retention at a single time point. For women's attendance at a clinic, we considered different types of visits including medical consultations, pharmacy drug refills, counselling, or laboratory analysis. This approach allowed us to consider HIV care as a holistic strategy, given that HIV care is provided in an increasingly team-based environment with non-prescribing healthcare professionals taking on expanded roles in direct patient care. Table 1 summarises the definitions of retention according to different measures and shows how failure events were accounted for in statistical analyses.

Statistical analysis
A total of 31,186 HIV-positive women enrolled in the PMTCT option B+ programme in 86 MOH facilities in Nampula and Zambézia provinces and initiated ART during the study period. Women's appointment adherence, as well as their on-time adherence to scheduled visits, were expressed as percentages, median, and interquartile range [IQR]. Women's retention status was classified into a binary outcome (retained vs. not retained in care) using a 75% threshold for clinical attendance. Each woman contributed to the analyses from the date of ART initiation to the first occurrence of a failure event, as described in Table 1 Kaplan-Meier survival curves were produced to estimate retention in care at 3, 6, 9, and 12 months post-ART initiation using each of the definitions described above. We assessed variability in the different measures of retention between respective definitions used. We did not consider 'appointment adherence' retention for the survival analysis, as this definition does not contemplate a time event, but rather a total number of visits completed during the observation period.
We assessed sensitivity, specificity, and the proportion of misclassified women for each alternative definition of retention compared to the reference of an 'ideal case scenario' of full 'on-time adherence' retention, defined as a woman alive and in care 12 months post-ART initiation and ³75% of scheduled visits attended on time (+/-15 days). Sensitivity and specificity values were calculated with 95% confidence intervals.  We considered the reference category to be the 407 women alive and in care 12months post-ART initiation with at least 75% on-time adherence. We calculated the sensitivity and specificity of other definitions to more accurately classify the women into a more suitable category of retention and determine the proportion of misclassified patients. While all definitions provided specificity rates of ≥98%, the sensitivity to detect a patient actually retained in care with ≥75% of on-time adherence was <12% ( Table 2). The WHO and MOH retention definitions provided the lowest sensitivity figures and the highest proportion of women misclassified as retained in care as follows: 3.0% sensitivity (95% CI 2.7%-3.3%) with 78% of patients misclassified according to the WHO definition and 4.3% sensitivity (95% CI 3.9%-4.7%) with 54% of patients misclassified according to the MOH definition. In addition, the 'point' retention definition misclassified 97.6% of patients. Although sensitivity rates were also low when using the IATT and 'appointment adherence' retention definitions, the proportion of misclassified patients was relatively smaller (15.9% and 18.3%, respectively).

Discussion
We analysed six different definitions of 'retention in care' for HIV-positive pregnant women enrolled in PMTCT option B+ programmes ranging from the most lenient definition using a 'point' retention approach to the most stringent definition using an 'on-time adherence' retention approach. Of the six definitions, three definitions assess retention at a single time point ('point' retention, WHO, and MOH). The IATT, the 'appointment' and the 'on-time adherence' retentions approaches reflect a more continuous engagement in care, where follow-up visits between initiation of ART and the endpoint date are considered.
As expected, we found that the more stringent the definition, the lower the probability for women to be defined as retained in care; and when levels of attendance to visits during follow-up were considered, estimates were even lower (85% and 60% when using the WHO and MOH definitions compared to 16% and 1% when using the IATT and 'on-time adherence' retention definitions). At the end of the first year, most women starting ART had not adhered to their scheduled visits, with median appointment adherence at 66% and median on-time adherence at 40% among the 16,840 women considered as retained in care using the 'point' retention definition. Of these, only 16% and 2% were actually defined as retained in care 12 months post-ART initiation, if we considered a threshold of ≥75% for appointment and on-time attendance, respectively. Initial visits in the first year of ART initiation are critical to ensure successful engagement in care, involving intensive counselling, peer support, conveying important preventive healthcare and risk reduction messages, and monitoring patterns of healthier behaviour for the women and their exposed children. This interaction between the health system and the HIV-infected women determines the likelihood of survival of the pair and the mother-to-child transmission of HIV. Adding the frequency and regularity of visits into the concept of retention increases the quality of the PMTCT programme evaluations, enabling a distinction to be made between women who attend for all or some care and treatment, and those who completely fail to adhere to scheduled visits and are therefore considered not to be engaged in care. Table 3  When considering 'on-time adherence' retention as the reference, all other definitions provided very low sensitivity rates to accurately detect patients retained in care, with high rates of misclassified patients. The WHO and MOH retention definitions provided the lowest sensitivity rates, that is, 3% and 4%, respectively.
For the 'point' retention, WHO, and MOH definitions, which did not consider continuous follow-up in their calculations, the proportion of misclassified patients ranged from 54% to 97%. The 'appointment adherence' definition provided the highest sensitivity rate (11.7%), with a fairly low proportion of misclassified patients (18.3%) compared to all other definitions analysed. We found that the median time to first failure of correct follow-up was short (≤3 months), if visit attendance during follow-up was considered, which demonstrated the need to implement early preventive measures to avoid patients on the PMTCT programme disengaging from care. Such strategies should focus on already well-known barriers to regular attendance, such as a lack of disclosure, poor staff attitudes, competing personal priorities, medication side-effects, or transport difficulties (22).
While our definition of 'on-time adherence' retention represents the ideal situation in which an HIV-positive pregnant woman enrols in a PMTCT programme and starts ART under option B+ should be considered as fully engaged in care to ensure an optimal viral response, it is perhaps too stringent for PMTCT programme evaluations. It is also time consuming to use on a routine basis and may not be compatible with current PMTCT service settings. Therefore, we recommend either using the IATT or the 'appointment adherence' retention definitions to better measure levels of engagement in care for mother-infant pairs. Our analysis had several limitations. To accurately evaluate attendance at scheduled visits, it is desirable to distinguish visits cancelled in advance (either by the patient or the care provider) from 'no show' visits that are actually visits missed by the patient (14). Our data did not allow this distinction to be made as this information was not captured in the database. Therefore, our results may have underestimated the real values of women's engagement in care when analysing 'appointment' and 'on-time' adherence retentions.
Some women, although defined as having a full attendance at scheduled visits, did not follow per se the recommended schedule of visits as per national guidelines.
More frequent visits may indicate a more advanced HIV disease, problems in counselling that need to be addressed, or drug or laboratory reagent stock-outs (25). Less frequent clinical visits may reflect a stabilised disease stage or temporary silent transfers (i.e. women receiving ART services elsewhere for a short period of time). For this reason, we chose to use all scheduled visits, regardless of whether the visits fell into expected time intervals. We did not analyse treatment interruptions and gaps in care for similar reasons. Further research is warranted to investigate the effect of a lack of observation of the recommended schedule of visits as per national guidelines on PMTCT programme outcomes.
Finally, we analysed data from health facilities with ePLD in two provinces in Mozambique. Although these two provinces are among the most populated in the country, with large volumes of patients seen at health facilities, our results may not be generalisable to other regions in Mozambique or to other resource-limited settings.

Conclusion
Balancing the need to reflect correct attendance and capture the regularity of visits without involving complex computer processing of the available routine data can be a daunting task. In addition, applying more stringent definitions is likely to result in lower reported rates of retention in care for PMTCT programmes. However, policy makers and programme managers should consider systematically taking into account attendance at follow-up visits to better guide decision-making concerning planning or scaling up interventions for mother-infant pairs. A consensus on a practicable definition is urgently needed, based on data availability, clinic scheduling practices, and local analytical capacities. Without overall agreed guidelines, findings from different programmes and interventions cannot be satisfactorily compared. Finally, a retention indicator is used in modelling estimates for vertical transmission of HIV. Without accurate and correct estimates, it is not possible to evaluate programme achievements effectively in respect of elimination of mother-to-child transmission of HIV. This consideration may become particularly important as recent programmatic data seem to indicate a stagnation in the achievements of PMTCT programmes worldwide, which suggests the need for revised global, national, and local strategies to end the HIV epidemic by 2030. (26).       Estimated Kaplan-Meier survival curves for retention over time, by retention definition.