Reasons for Missed Visits and Inuence of Phone Call Reminders on Retention of HIV-Infected and Uninfected Children Followed in Cameroon: The ANRS-PEDIACAM Cohort

As many longitudinal studies, follow-up in the ANRS-PEDIACAM study is disrupted by repeated absences of participants to scheduled visits. This lead to missing data which inuence the quality of results. We describe reasons for participants absence or non-compliance (NC) and assess the inuence of phone call reminders (CR) on retention in care.


Introduction
The Human Immunode ciency Virus (HIV) infection remains a major problem in sub-Saharan Africa countries with bear the heaviest global burden with nearly 67% of persons living with HIV (PLWH) in 2019 (1). The scale-up of HIV testing and antiretroviral therapy (ART) has helped to control the disease and improve the survival and quality of life of PLWH (2). These global efforts had also an impact on the prevention of new HIV infection among children born to HIV infected mother, leading to a 60% reduction of the number of HIV pediatric infection in sub-Saharan Africa between 2002 and 2013 (3) and increasing of the number of HIV-exposed uninfected (HEU) children.
Coverage of ART among HIV-infected children has also improved during these last years. In 2017, 52% of HIV-infected children received ART. However, ART programmatic success requires both a sustainable supply of ART at the facility level and lifelong adherence to treatment by PLWH (4). Poor ART retention in care and adherence may lead to virological failure, reducing future therapeutic options (5,6), in the context of very limited access to the second and third lines regimens, and a hindrance in achieving UNAIDS 90-90-90 targets set for 2020 (7,8). Retention in care also concerned HEU-children as it has been demonstrated that they are at high risk of mortality and morbidity compared to children born by HIVuninfected mother (HUU) during the rst year of life (9)(10)(11). The mechanisms of occurrence are not fully understood (12)(13)(14).
The ANRS-PEDIACAM cohort is constituted of HIV-infected, HEU and HUU children groups followed in Cameroon. As most prospective studies, it is threatened in its management by episodic and repeated absences to scheduled visits (8,15). In that cohort, the proportion of "Non-compliant (NC)" children increased gradually during follow-up, especially among HIV-uninfected children. This could have consequences not only on their health, but also on the analysis and interpretation of the results. The reminders methods based on mobile telephony tools might help in improving patient adherence in health facilities (16)(17)(18). Based on that hypothesis, we reorganized and intensi ed the use of phone calls within the ANRS-PEDIACAM study from April 2014.
The aims of this study were to describe the reasons for missing scheduled visits and to assess the in uence of phone call reminders (CR) on retention.

Methods
Study setting: the ANRS-PEDIACAM cohort study This study was developed within the framework of the ANRS-Pediacam cohort which is an ongoing longitudinal study coordinated by the Centre Pasteur of Cameroon, designed to assess the feasibility of early HIV-diagnosis and early treatment of HIV-infected children in resources limited settings (19,20). From November 2007 to October 2011, a total of 611 children were included in three groups and followed in three referral hospital in Cameroon (the Mother and Child Center of the Chantal Biya Foundation (MCC-CBF) and the Essos Hospital Center (EHC) in Yaounde, and the Laquintinie Hospital in Douala (LHD)).
These groups consist of: HIV-infected (n=210), HEU-children (n=205) and HUU-children (n=196). Followup was scheduled three-monthly until the age of two years, then six-monthly for HIV-infected children; while HIV-uninfected children were followed six-monthly.

Study population
Our study population consisted of children followed in the ANRS-PEDIACAM cohort. By April 2014, 65 (10.6%) died ( gure 3) leading to 546 children including 153 HIV-infected, 200 HEU, and 193 HUU considered in this study. Of these children, 160 (29.3%) were considered as non-compliant.
Intervention phone reminders strategy and data collection From April 2014 to April 2017, we issued every week a list of expected children for their visit. Phone calls were conducted by a Clinical Research Assistant as follow: At week i, all the NC children of the week.
If not seen, wait two weeks before calling back If still not seen restart the procedure immediately for HIV-infected children or wait for two months before restarting for uninfected children.
The topics discussed on phone were: children vital status, reminder of the missed visit, reasons for missing, and taking an appointment to return to care. The collected data were then entered into an Access database.

Data used
We used socio-demographic data collected at the study inclusion, the subject history of missing visit, the reasons for non-return to the visit for NC children and phone reminder from April 2014 to April 2017.

Outcome de nition
A child was considered to have missed a visit 'V i ' if he did not come to the expected date of the visit 'V i ', and never returned since that date until his next scheduled visit 'V i+1 '.
The outcome was "Non-compliant (NC)" de ned as children alive or unknown status who missed at least two consecutive visits.

Statistical analysis
Socio-demographic parameters and reasons for missed visits were described and compared between groups using respectively frequency, chi-square or Fisher test for categorical variables; means, median (IQR) and non-parametric Kruskal Wallis test for continuous variables. A multistate transition modelling approach was used to analyse the retention care cascade. Follow-up time was discretized into intervals centred on the scheduled visit date for each child. At each time interval, four discrete states were de ned using children-level data: seen, alive and not seen only at the current visit (missed visit), alive and not seen since the last scheduled visit (Non-Compliant), and deceased. Figure 1 is a graphical representation of the states and possible transitions. States transition matrix was calculated under the rst-order Markov dependence assumption. To evaluate the in uence of the call reminder enhancement performed on retention, we use a multinomial regression model (characterized by: retention to the care; duration of consecutive missing scheduled visit and mortality) to estimated the Relative Rate Ratio (RRR) (21) associated with each. A p-value of 0.05 or less was considered to be statistically signi cant. The R software version 3.3.3 was used to perform all statistical analysis.

Population characteristics
Of the 546 children considered at the start of this sub-study, 193 (35.3%) were HUU, 200 (36.6%) HEU and 153 (28.0%) HIV-infected. Among them, 48.5% (n = 265) were included at the MCC-CBF, 27.8% (n = 152) at EHC in Yaounde, and 23.6% (n = 129) at the LHD in Douala. The children characteristics at inclusion differed signi cantly among groups (Table 1). Compared to the HIV-infected group, mothers of uninfected groups (HUU and HEU) were more likely to be educated, employed, and involved in a stable relationship.

Non-compliance to visit
At the beginning of this study, 29.3% (160/546) were NC (April 2014), mostly HIV-uninfected children (36.9% (145/393) vs 9.8% (15/153), p < 0.001) (Fig. 3). Non-compliance status was also associated with the site of inclusion and follow-up group (Table 1) Tables 3 and 4 provide a summary of state transition rates before and after call reminder enhancement (CRE), using all data aggregated over time. Entries in the upper left indicate the cyclic nature of retention in care: among those seen at their current scheduled visit, the rate of retention in care is 84% and 89% respectively before and after CRE. Among those not seen only at the last visit, about 47% and 54% respectively before and after CRE, become NC at the current visit. The estimated probability transition matrix shows that children who are NC have a higher risk to not return for follow-up.

Phone calls results
About 1069 phone calls were made for the 868 visits scheduled to NC children when they were NC during April 2014 to April 2017, among these calls 693(64.8%) were unreachable. Of the 376(35.2%) reachable phone calls which resulted in an interview, 168(44.7%) ended with an appointment and 26.2%(44/168) were subsequently realized (Fig. 4). The proportion of unreachable phone calls increased signi cantly with a long delay since the last visit (35% for ≤ 7 months; 45% for 7-12 months, and 65% for > 12 months, p < 0.001). About 80.3%(171/213) of called children were at least once reachable while 19.7% (42/213) were always unreachable during the evaluation of which 71.4%(30/42) were already NC before the evaluation period.
Reasons why children and parent's pairs weren't returning for follow-up Of the 376 reachable phone calls realized, about 54%(115/213) of contacts provided reasons (n = 276) why they did not come back to visits ( Fig. 4 and Table 2). These reasons were mainly: changing of residence (30.4%), lack of time (23.6%), wish to stop follow-up (11.6%), parent and/or child trip (9.8%), school attendance by children (9.4%), forgetfulness (7.2%), illness or death of a relative (4.7%) and 3.3% other reasons including lack of motivation. This distribution did not change with those of children who were always NC. No signi cant difference was found between the reasons given in each follow-up group.
Two-third of parents whose reason of non-returning for follow-up was changing of residence have always been non-compliant ( Table 2).

Impact of phone reminders on retention to care
Of the 160 NC children at the beginning of this evaluation, 31.8% (n = 51) returned to follow-up (Fig. 3). Among children, whose parents were called and reached at least once, 21.2% (38/171) returned to followup (Fig. 4). We also noted that among the 2046 visits completed, 32.2% (659/2046) were after the scheduled date of appointment; and among them, 91% (600/659) were achieved through phone calls.
Factors associated with state transition probability The RRR in Table 7 indicates that children from a low-income family (< 50000) are most likely to remain non-compliant compared to children from high income family (> 150000).

Discussion
This paper describes the reasons for missing scheduled visits and the in uence of phone reminders on missing visit in the ANRS-PEDIACAM cohort from April 2014 to April 2017; which is an ongoing longitudinal study set up for more than ten years. Also, the in uence of phone reminders on missing visit state, from inclusion to April 2017 was modelled using discrete time multistate framework characterized by: retention to the care; duration of consecutive missing scheduled visit and mortality.
Overall, about 45% of children followed had been considered NC at least once during the evaluation period. The proportion of NC was signi cantly higher (56%) among HIV-uninfected children particularly among HEU-children (58%). This observation reinforces those reported by Sidze et al (22) in a previous study conducted in the same cohort which showed that the proportion of lost to follow up children was signi cantly lower in HIV-infected and HEU-children compare to HUU-children, and only 35% of the latter could be joined by phone calls. The high proportion of NC among HIV-uninfected children might be related to parent decision. As a longitudinal study, PEDIACAM has consisted of regular clinical visits. The parents whose child is not sick, or shows no signs of illness, may not perceive the importance of regular clinical visits and will even try to stop follow-up as was observed in 90.6%(29/32) of those who show interest to stop. On the other hand, a high proportion of mothers (42%) of HEU-children were unaware of their HIV status before prenatal HIV screening. The Woman who had not disclosed her own HIV infection status to the family (especially her partner) had di culties justifying frequent visits with the child to the health facility. Fear of violence and stigma from their families prevented them from utilizing HIV care services (23).
In this study, phone calls reminder had considered as a search tool on returning to care. The in uence of this intervention in improving patient adherence in health facilities had been shown in a similar context (16,17,24). Only 35.2% of the total phone calls made resulted in a phone interview. The reasons for NC to missed visits were mainly related to geographic mobility of population (relocation, travels), lack of time, and forgetfulness. These reasons are similar to those presented in other study (25). However, in our study, others important reasons like stigma, lack of money, and the cost of transport were not mentioned as indicated in some studies conducted in the resource's limited countries (22,(25)(26)(27)(28). In fact, in PEDIACAM study, to avoid missing scheduled visits, free medical support and reimbursement of transport fares were brought to the children. This is curious when we nd that it's the uninfected children who miss the most visits, and that is why we can think that this stigma reason could have been masked among the answer we collected.
The predominance of relocation (30.4%) among reasons given in our study could be explained by people dynamic related to study duration. Some of the parents were transferred for professional purpose, others relocated because of economic reasons or modi cation of the composition of the family with the death of one of the parents. Most of the parents who change the place of residence had children always NC indicating that speci c measures should be implemented to tackle such a problem.
Phone call reminders (CR) allowed to reduce signi cantly adjusted relative risk to miss one scheduled visit to two third. This result shows that CR are an effective intervention to improve the retention in care. Our results are in line with those recently published in a systematic review of current clinical evidence of CR on retention to care among HIV patients (29). The later used the results from nine studies carried out in four sub-Saharan African countries (Cameroon, Kenya, South Africa, Eswatini), Switzerland, United Kingdom and United States of America. However, this intervention shows weaknesses for children noncompliant. Indeed, a large number of calls performed among non-compliant children remained unreachable throughout the study period.
Multivariate analysis identi ed others economically and individual's factors associated with a missing visit or non-compliant: HIV infection, low family incomes, mother's occupation. In addition, low-income family live in a precarious situation, which leads to frequent relocation to nd a better situation or for seasonal work. This may result in an increase in the distance from the site of care, an increase in the cost of transport and therefore pre-nancing di culties. On the other hand, most of the mothers (40%) of HEUchildren were housewives or unemployed at the inclusion of their child in the study. However, it can be noted that mothers who were housewives or unemployed were not nancially independent and therefore needed nancial support from a partner or their families to come frequently to clinical visits.
This study has some limitations: the phone calls were the only research tool used in this study for reminders and collection of informations about the reasons for missing visits. More than half of calls performed were unreachable. As a result, reasons for missing visits were not always available, and could only be collected from those who were reachable. The time intervals between clinical visits were irregular throughout the follow-up. In multivariate analysis, we only considered covariables measured at inclusion in our study. Indeed, the model was the concern with missing visits, and at these visits, the dynamic data were missing. In this context, the modeling requires a step of the imputation of the missing data beforehand. In a recent paper, Lee and al (21) used a most recent observed value for imputing missing data, however this strategy to deal with missing data is not always appropriate. In future work, comparing this method with other imputation strategies may determine the best way to consider dynamic data in this framework.

Conclusion
Our ndings suggest that maternal HIV infection and socioeconomic status could in uence the attendance of children in HIV care. Strengthening counselling of mothers of HIV-uninfected children to better explain the objectives of the study and the reasons of their participation can help to improve the attendance of HEU-children as well as HUU-children. On the other hand, the phone call reminder is an effective intervention to improve attendance to care. However, this intervention is limited among noncompliant children, due to the high proportion of unreachable calls, what indicates that this strategy cannot be used alone and should always be associated to be more effective like community monitoring.

Declarations Ethics approval and consent to participate
The ANRS-Pediacam study was granted ethical approval in Cameroon by the National Ethics Committee and in France by the Biomedical Research Committee of the Pasteur Institute of Paris. The Cameroon Ministry of Public Health gave administrative authorization to start the study. We obtained free and informed consent from all study participants.

Consent for publication
We give our consent to the publication of this work.

Availability of data and material
The data and tools of this study remain available and are maintained by our teams.

Competing interests
We declare that this study is not subject to any con ict of interest.