Estimating the burden of Paediatric HIV in an ‘A’ category district in India: An epidemiological study

Anju Sinha (  apradhandr@gmail.com ) Indian Council of Medical Research, New Delhi Reynold Washington St John’s Research Institute, Bangalore Rajeev Sethumadhvan Karnataka Health Promotion Trust, Belgaum Rajaram Potty St John’s Research Institute, Bangalore Shajy Isac Karnataka Health Promotion Trust, Belgaum Vasantha Thavraj Indian Council of Medical Research, New Delhi Ravindra Pandey All India Institute of Medical Sciences, New Delhi


Introduction
Globally the population of People living with HIV is on the rise and projected as a threat to public health.
Countries are not on track for the UNAIDS 2020 and SDG 2030 targets 1 . It is estimated that 1.94 million children will be living with HIV in 2020 2 . There are few estimates of the magnitude of Paediatric HIV. Mathematical modeling based estimates using the Spectrum and Estimation and Projection Package (EPP) model cover only  year old population 3,4 . In India, a country with 1.3 billion people, children up to 18 years constitute 41% of the total population 5 , but lacks accurate estimates of the pediatric HIV burden. This is needed to save and improve lives of HIV exposed children. The National AIDS Control Program provides estimates of Paediatric HIV that are based on its proportion to adult infections, but uses available data from other countries, in order to arrive at this estimate 6 . The WHO recommended various methods to estimate Paediatric HIV including case reporting, household surveys, immunization clinic surveys, in-school and out-of-school surveys, mortality data and vital registration. These have been used in countries with a high HIV prevalence: South Africa, Nigeria, Kenya, Thailand, Argentina, Mozambique, Malawi, Indonesia and others 7,8 . Similarly, the WHO and UNAIDS recommended measurement of HIV prevalence among children (0-14 years) in settings where the HIV prevalence among women in the reproductive age is 5% or greater, with high fertility rates, low coverage of prevention of mother to child transmission (PMTCT) services and where resources to conduct a large sample survey among children is feasible. India's low overall adult HIV prevalence and limited resources for large scale surveys do not meet these criteria and hence Paediatric HIV measurement is excluded in many large scale surveys partly due to limitation of sample sizes for a robust pediatric burden measure 3 . India therefore lacks direct Paediatric population level data that can help the country arrive at more accurate estimates of the magnitude of Paediatric HIV.
One of the goals of the National Strategic Plan to end AIDS (2017-2024) is the elimination of mother to child transmission (EMTCT) of HIV. Mother to child transmission (MTCT) remains the main reason for HIV infection among children in India. Despite the availability of tools and methods to prevent, identify and treat HIV in children, India's performance on this front has lacked lustre. By 2020, 95% of pregnant women should have received testing for HIV and syphilis and 95% of estimated positive pregnant women should be on antiretroviral treatment (ART) in order to achieve an MTCT of less than 5%. However, by 2015, India had registered only 74.2% of 29.7 million pregnancies during the antenatal period and only 59% of these were tested for HIV. Only 45% of an estimated 35,255 HIV positive pregnant women received ARV. Out of 10,677 HIV positive live births, 85% of babies were tested at least once in 2016-17, but only 59% were tested within two months. The lack of technology enabled platforms and inadequate utilization of front-line workers for this purpose, results in a linkage loss at every level 9 . The 'start free, stay free and AIDS free' platform offers India an opportunity for focused, coordinated action and renewed commitments to end Paediatric AIDS and to eliminate MTCT 10 . In the absence of ART, approximately 30% of untreated HIV-infected children die before their rst birthday and more than 50% die before they reach 2 years of age 11 . With early diagnosis and early initiation of treatment, survival improves substantially with only 20% of babies dying by their rst birthday 12 . Untreated HIV infection in children results in growth delays that may not be reversed by ART 13 . It is therefore crucial to have reliable estimates of Paediatric HIV, in order to plan, implement and monitor the coverage of prevention and control efforts in HIV infected children.
The Indian Council of Medical Research commissioned a task force study to estimate the burden of Paediatric HIV in a category 'A' district of a high prevalence state. The study protocol has already been published previously 8 .

Methods
Aim and study design: The aim of the study was to estimate the disease burden of pediatric HIV among children in 'A' category district of a high HIV prevalence state, using a multipronged approach Study setting: The study comprised of three distinct strategies including active surveillance, inclusion of public and private healthcare facilities, data collection from blood banks and NGOs and children attending general and specialized clinics for a better estimation of Pediatric burden. Information regarding study setting and design and sample size estimation have been previously described 14 . Brie y, District Belgaum was selected after a baseline assessment of three high prevalence districts in South India. Belgaum was chosen for the study as it had a high adult HIV prevalence of 1.43% in 2011 15 . Besides, the coverage of antenatal care and prevention of mother to child transmission (PMTCT) services was high and the district administration was supportive of the study. A mapping of all 971existing (Govt. & private) health care facilities (HCFs) in the district was conducted (ref published report). A total of 285 HCFs from ten talukas with stand-alone and reporting HIV testing facilities (149 Govt. and 136 private) were included in the study.
In 2011, Belgaum district had a total population of 4779661 (Males: 50.7%, Females: 49.3%), 75% of whom were rural and 73.5% (Males: 82.2%; Females: 64.6%) were literate. Strategy 1 used a prospective cohort design to measure the incidence rate of HIV by early case detection in infants and young children (0-22 months) born to a HIV positive pregnant woman registered at one of the public or private health care facilities of the district. The study team visited the identi ed HCF twice in a week. A crude line list was prepared from the secondary data collected from the HCFs, the list was re ned by applying eligibility and removing duplicates. All HIV infected pregnant women residing in any of the 10 talukas, who consented for age -speci c HIV blood test for their infants were eligible for enrolment. The women were contacted over phone and visited physically at home or elsewhere (if home visit was not permitted). Pregnant women were followed through their pregnancy & delivery, visited once in a month until delivery and afterwards till their infant was 22 months old. A mother and infant form was lled for each mother-infant dyad. Ageappropriate early testing in infants using DNA PCR dry blood spot (DBS) was conducted at 6-10 weeks, 6-9 months and antibody based ELISA tests at 18-22 months. Demographic information, mother -infant form and a pregnant Positive Women line-list was maintained. Each mother and infant Dyad was assigned a unique identi cation number.
Strategies II and III used a cross-sectional design. Strategy II aimed to detect HIV infection among children (0-14 years) by family screening of HIV positive parent(s) (PLHAs) referred from ICTC centers, blood banks and community based NGOs in all 10 talukas. If a positive male was detected his wife and children were tested, if a positive female was detected her husband and children were tested. Any HIV infected man/woman, of age 18-49 years, having a biological child of 0-14 years residing in any taluka of Belgaum, who consented for testing of their spouse and children for HIV were eligible and included in strategy II of the study. Public and private HCFs were visited twice a week by the designated teams to check information about HIV infected individuals 18-49 years, they were contacted to nd if they had any biological children 0-14 years. The spouse and all children of the positive person (male or female) were subjected to age appropriate HIV testing. Demographic details, data on testing were recorded. A unique identi cation number for the positive persons identi ed through family screening was created. The strategy III used screening of sick children visiting health care facilities, in four talukas of Belgaum that included 10 Health care facilities selected using strati ed random sampling on the basis of MTCT prevalence, government and nongovernment, and by levels of health care offered. IMCI -HIV criteria (applicable to 0-5 years age) was adapted by Indian experts to include children >5 to 14 years into the algorithm. Sick children (0-14 years) presenting with suspected signs and symptoms satisfying the 'Modi ed Integrated Algorithm' (including sign symptoms from the IAMI and 'special clues') 16 were tested at health care facilities by age appropriate HIV tests. Health care providers from the four participating talukas were trained in the use of the algorithm, operational de nitions were developed for each criteria included in the algorithm. The facilities were visited weekly by the research team to nd information about children screened positive, referred & tested. Tracking for test results was done through the unique identi cation number maintaining con dentiality. The strategies (I, II, & III) were expected to comb the district with pediatric cases hidden within families and presenting as masked sicknesses yet unidenti ed as HIV) Estimates derived from each of the strategies (1, II & III) were multiplied by an in ation factor derived in a workshop of investigators and the Project Advisory Group/subject experts. The results from the study were extrapolated to the population characteristics within the district, including total population (adult and child), estimated overall adult HIV prevalence, estimated prevalence among pregnant women and reported coverage of HIV testing among the antenatal sub-population. The outline of the study design is depicted in Figure 1. Study implementation: The ten talukas in the district were divided into three clusters, each of which had ve Field Investigators (FIs) and a supervisor (Senior Research Fellow, SRF). In each cluster, a team of one male and one female FI was given charge of two talukas each (strategy 1 & 2), and one additional FI was allotted for the strategy 3 related work. 14 FIs were recruited and trained over two periods, for a total duration of 10 days in technical skills as well as soft skills of maintaining con dentiality& gender sensitive approach. The overall eld study was supervised by a medically quali ed professional. SRFs planned daily visits of FIs, validated 5% of data and veri ed forms for accuracy and completion.
Data Management & statistical analysis: Data was double entered using Microsoft Access, cleaned and veri ed for consistency and analyzed using SPSS version 22.0 and STATA version 13.0 A Project Advisory Group (PAG) guided the study team to develop a Statistical Analysis Plan (SAP). The primary outcome in strategy 1 was cumulative incidence, calculated as the number of new infections per total number of children at risk. A child was at risk till rst positive result by any test at any age, by age appropriate testing. For censored observations, time was the duration of follow up. The SAP considered the limitations in coverage of services and response rates, and guided to determine the 'Net in ation factor' for each of the 3 strategies. Under strategy I the net in ation factor was derived using the estimated number of pregnancies in the district, proportion of un-tested pregnancies, pregnant mothers not enrolled and untested children.. In strategy II and III prevalence of HIV infection among children 0-14 years of age was calculated. Data was analyzed as per the SAP. It was based on the actual/projected 18-49 year population in the study period, estimated number of 0-14 year children, proportion of eligible index persons not recruited and not proportion of eligible children not tested. In strategy III the factors considered for Net In ation factor were: actual/projected 0-14 year population during the study period, estimated number of 0-14 year children experiencing any morbidity, estimated morbid children reaching a HCF for care, estimated children satisfying screening algorithm, and suitable in ation factors for geographical and institutional factors, morbid children not reaching selected HCF but reaching other facilities in the district, un-screened, non-enrolled and untested children. The estimates derived under the strategies were then multiplied with the in ation factor to come up with the overall estimate. The steps followed are described along with the results section.

Results
In Strategy 1, we line listed 750 HIV infected pregnant women from 285 selected health care facilities. After exclusion of duplicates, those resident/had moved outside the district, died or who were no longer pregnant, we recruited 469 HIV infected mothers who had delivered between 2011 and 2013, and who consented to participate in the cohort study. Twenty-seven mothers had a repeat pregnancy during the study period. Thus the total number of pregnancies was 496. Among the 496 pregnancies, 477 (96%) resulted in live births, 10 (2%) in abortions and 9 (2%) in still births. Among the live births, 10 were twin pregnancies. Thus, the total number of live born babies was 487. Of 487 HIV exposed live born children, 454 (92.3%) children were tested at least once, and 39 were found to be HIV positive during follow up by 22 months of age. The net cumulative incidence rates of vertical transmission of HIV per 100 pregnancies were 2.1, 5.3 and 7.8 at 0-10 weeks, 0-9 months and 0-22 months of age, respectively. The annual cumulative incidence rate (%) was calculated to be 5.32%.
In order to calculate the in ation factor at a population level, the following indicators as per the ow diagram ( Fig. 2) were conceptualized. The basic parameters were taken from the Census data (Table 1). Based on this, the indicators in the ow diagram were deduced as in the Table 2. Based on this in ation factor, the occurrence of new Paediatric HIV infections from MTCT is 41.2 (rounded off to 41) children by the age of 22 months.

Figure 2
Conceptual diagram for calculating in ation factors and burden of Paediatric HIV from Strategy 1.    12.8%; Females: 8.4%). The HIV prevalence was 18.6% (M: 23%, F:13%) in the < 5 year age group and 8% (M:9.2, F:6.9%) in the 5-14 year group. In order to project these ndings for the district population, we used the following criteria as shown in Fig. 3.

Figure 3
Conceptual diagram for calculating in ation factors and burden of Paediatric HIV from Strategy 2.
For this, the basic parameters required are taken from the Census data available (Table 3). For estimation of PLHIV in the district of Belgaum, the following calculation was adopted (Table 4). Based on this, the indicators in the ow diagram (in ation factors) were deduced (Table 5).      In strategy 3, of the total 33342 children who visited the 10 health care facilities during the study period, 24342 (73%) were screened by the trained eld investigators. 527 (2.2%) sick children were identi ed, 509 completed HIV testing requirements. Of these, 97 children turned out to be positive (HIV prevalence 19.1%), but 86 of them had prior knowledge of their HIV positive status. The study was therefore able to identify 11 (2.16%) new HIV infections from among the total 509 sick children. For strategy 3, it was assumed that HIV infection would be nil among the 23815 screened children who did not have any indicative symptoms or social risk criteria for HIV and AIDS. The parameters considered are given in Table 6.     The results although not generalizable across all districts in India, do provide useful information on how to estimate the burden of Paediatric HIV from existing data by using an excel based software (being published elsewhere).
The GBD framework 1 based on Spectrum and EPP use multiple methodological improvements, yet face limitations and biases due to use of variable data sources. A non-parametric back calculation method 20 used in Thailand studied HIV/AIDS trends for future predictions reported data adjustments to overcome surveillance reporting issues. Moreover, these methods did not include pediatric age groups. Present study is a sincere effort and a step further in deriving the pediatric burden estimate from real time data of a district.
The uniqueness of this study is its innovative epidemiological design, using robust combination of community and facility level data, inclusion of private and public sector health facility data, multi-pronged strategy of using cross-sectional and longitudinal data collection techniques, and extrapolation to correct for gaps in coverage using in ation factors. In most countries, including India, linkage loss tends to occur at various levels within the PMTCT programs 21,22,23,24 . The study staff supported the program to reduce the gaps in HIV testing and treatment linkages at various levels. They thus achieved high levels of coverage of testing of children identi ed within the families of index person living with HIV, as well as timely testing of the HIV exposed newly delivered infants. With well-de ned line listing and recruitment processes in place and individualized follow ups the study was able to quantify and reduce duplication and to better understand mobility of PLHIV within and outside the district. A number of doctors were also trained in the use of the Modi ed Integrated Algorithm based on IMCI-HIV and adapted IMAI guidelines, during the study. The response rates of eligible subjects for HIV testing were high with about 85% of eligible mothers completing the protocol for follow up and a similar proportion of eligible families with children completing HIV testing for all children. The study methodology is replicable for other settings and other diseases.
However, we also recognize a number of limitations in the study. During the study period, a number of changes in policy of HIV treatment and prophylaxis for pregnant women occurred including prophylaxis using single dose Nevirapine, to use of expanded regimens to option B, option B + and a 'treat all HIV infected pregnant women', irrespective of CD4 count, at the time of HIV diagnosis. HIV related prevention and treatment services were already district-wide and to scale. As a result of prevention and treatment initiatives, there were steady declines in HIV prevalence among the general population, and particularly among pregnant women that led to a delay in completion of the required sample size and follow up in strategy I. This delayed the timely achievement of required sample size and could have implications in our calculation of prevalence using the base population size for the district. However, because of the large overall population size within Belgaum district, the effect of these annual changes in estimated population size may not be substantial.
A second limitation is the assumption of a similar prevalence among the non-included subjects within the study, during the extrapolation exercise. The number of variables that we had collected was insu cient to completely match the characteristics of responders with non-responders. The non-response bias could make our estimate an under or an overestimate. However, with high levels of coverage as in the study, this too may not be much different.
There was a delay in initiation of strategy III, as development of the Modi ed Integrated Algorithm by the national experts/ICMR sub-committee took time, as did the training of health care providers in selected talukas. 86 of the 97 HIV positive children knew their HIV positive status before falling sick. This could be attributed to the late initiation. However, despite this, strategy III did yield new HIV positives among children attending a health facility for reasons other than HIV treatment.
Our study only considered children within a family unit. We did not include children who lived without a parent (child-headed homes) or who lived within an institution. Previous studies and strategy 3 results indicate that children who were orphans were much more likely to be HIV infected. We did pick up some of the HIV infected orphans in strategy 3. However, we were not able to estimate the HIV prevalence among orphans. A cross-sectional survey of orphanages for HIV prevalence could have added value to this study 25 .
Another primary assumption used in strategy I was that the new cases were contributed only by mother to child transmission. Recent studies from other countries have indicated that adolescents orphaned as a result of HIV are at greater risk and vulnerability for physical and sexual abuse, including HIV 26 . However, it is expected that these other modes of HIV transmission are rare and the numbers that add to the burden would be minimal. A last limitation is that we could not integrate mortality and migration estimates into this estimation. The study was not designed to systematically measure mortality among children living with HIV and the information on age-speci c mortality rates for children 0-14 years are not available. We could not calculate age-speci c death dates from the current study, as reporting of data for child deaths in the family were not forthcoming and the records were not available to verify the actual date of death, even when they were reported. We observed that HIV prevalence among the under 5 year age group was more than twice the HIV prevalence in the children 5-14 years. Interventions for PMTCT in Belgaum were almost non-existent ten years prior to the study period. Therefore, the only plausible reason for this reduction in prevalence in a cross-sectional strategy II could be that most children living with HIV had died. The nonintegration of mortality information into the estimate, could result in a higher than the real value.
Despite these limitations, the study is the rst of its kind in India and offers new information on methods to estimate Paediatric HIV. We put forward a number of recommendations for further studies. During the phase II ongoing study, it would also be useful to explore the feasibility of testing the baby for HIV at birth 27 , as many maternally exposed new-born died before they were due for the rst HIV test at six weeks.
Testing for HIV at birth is a current recommended CDC guideline that has not yet been adopted in India 28 . acknowledged that there could be other reasons for HIV transmission to children, especially amongst adolescents 31 . A cohort study among adolescents within these families could indicate the extent to which this occurs.

Conclusion
The study has used a unique innovative methodology for disease burden estimation of pediatric HIV in a high prevalence district in India, where such data do not exist. There is an increase in the burden estimate from the earlier projected gure of 6-10.4% in the study; it could be used by program planners for improvement in disease control efforts. The study methodology can be replicated in similar settings for HIV as well as for other infectious diseases.

Declarations
Ethics approval and consent to participate: Regulatory approvals for the study were obtained from the National AIDS Control Organisation (NACO) and the Karnataka State AIDS Prevention Society (KSAPS).
Ethical approval was obtained from the Institutional Ethics Committee of St John's Medical College and Research Institute, Bengaluru, India. Informed written consent was taken from all the study participants. Informed written consent was taken from the entire parent on behalf of the children under the age of 16 for participating in the study. All methods were performed in accordance with the relevant guidelines and regulations.

Consent for publication: Not applicable
Availability of data and materials: The datasets generated and/or analysed during the current study are not publicly available. The study data is available only to the collaborating scientists. The data may be available on request to the corresponding author Dr. Anju Sinha (apradhandr@gmail.com), Indian Council of Medical Research (ICMR), New Delhi. The data also may be available upon request for some of the collaborating institutions. Data will be sanitised to remove individual identi ers in order to comply with the local data protection laws. All data sharing is also subject to National AIDS Control Organisation (NACO) and ICMR approval.
Competing interests: Authors declare that they do not have any competing interests..
Funding: Indian Council of Medical Research, New Delhi. The funding agency had no role in study design, collection, analysis and interpretation of data and in writing the manuscript.
Authors' contribution: AS conceptualized the study, wrote the protocol, coordinated the study, contributed to analysis, and revised the draft manuscript. RW was responsible for study implementation, contributed to the study design, data analysis & interpretation, wrote the rst draft of the manuscript. RS was responsible for the day to day management of eld activities, supervised and monitored implementation, contributed to data analysis, RSP helped with data management and analysis, SI contributed to sample size calculation and reviewed the data analysis, RMP guided the statistical plan of analysis, contributed to data analysis and interpretation. VT was involved in initial phase of study conceptualization and supervision. All authors agreed and approved the nal draft for submission.

Figure 1
Outline of the research study Figure 3 Conceptual diagram for calculating in ation factors and burden of Paediatric HIV from Strategy 2.