Growth of young HIV-infected and HIV-exposed children in western Kenya: A retrospective chart review

Introduction The objective of this study was to determine the growth patterns, rates of malnutrition, and factors associated with malnutrition in children born to HIV-infected mothers in western Kenya using data from an electronic medical record system. Methods This study was a retrospective chart review of HIV-infected (HIV+) and–exposed (HEU) children (<5 years) using data collected prospectively in the course of routine clinical care and stored in the electronic medical record system in western Kenya between January 2011 and August 2016. Demographics and anthropometrics were described, with Chi-square testing to compare proportions. Multiple variable logistic regression analysis was used to identify correlates of children being stunted, underweight, and wasted. We also examined growth curves, using a resampling method to compare the areas under the fitted growth curves to compare males/females and HIV+/HEU. Results Data from 15,428 children were analyzed. The children were 51.6% (n = 7,955) female, 5.2% (n = 809) orphans, 83.3% (n = 12,851) were HEU, and 16.7% (n = 2,577) were HIV+. For HIV+ children assessed at 24 months, 50.9% (n = 217) were stunted, 26.5% (n = 145) were underweight, and 13.6% (n = 58) were wasted, while 45.0% (n = 577) of HEU children were stunted, 14.8% (n = 255) were underweight, and 5.1% (n = 65) were wasted. When comparing mean z-scores, HIV+ children tended to have larger and earlier dips in z-scores compared to HIV-exposed children, with significant differences found between the two groups (p<0.001). Factors associated with an increased risk of malnutrition included being male, HIV+, and attending an urban clinic. Maternal antiretroviral treatment during pregnancy and mixed feeding at 3 months of age decreased the risk of malnutrition. Conclusions HIV+ and HEU children differ in their anthropometrics, with HIV+ children having overall lower z-scores. Continued efforts to develop and implement sustainable and effective interventions for malnutrition are needed for children born to HIV+ mothers. This is a retrospective study using data collected prospectively in the course of routine clinical care and stored in the electronic medical record (EMR) system. Data were pulled for all children who were <5 years of age between January 2011- August 2016 and enrolled in a large HIV clinical care system in Kenya, the Academic Model Providing Access to Healthcare (AMPATH). Born from a 20-year partnership between Indiana University School of Medicine (IUSM), Moi University School of Medicine (MUSM), and the Moi Teaching and Referral Hospital in Eldoret, Kenya, the AMPATH HIV care program has enrolled over 160,000 patients and currently provides care for approximately 15,000 HIV+ and HEU children in 65 clinics in western Kenya [19]. During this period, AMPATH clinical data were captured on standardized paper encounter forms and then entered into the AMPATH Medical Records System (AMRS), a resource for both patient care and research evaluations. AMRS was sub-Saharan Africa’s first comprehensive EMR for HIV care, pioneering the effective use of EMRs in such settings [20, 21]. Outcomes of HIV+ children in this large cohort have previously been reported for retention in care, therapy, and HIV transmission rates [22–24]. This study was approved by the IUSM Institutional Review Board and the United States’ Office of Human Research Protections-approved MUSM Institutional Research and Ethics Committee. The requirement for informed consent was waived by both ethical governing bodies. The data for this study were handled and stored within Health Insurance Portability and Accountability Act of 1996-compliant secured servers. Participants were eligible if they were seen in any of the AMPATH clinics between January 2011 and August 2016, were <5 years of age when enrolling in care during that time period, were born to HIV+ mothers, and had at least one anthropometric measurement recorded. In this setting, HIV-exposed children receive confirmatory HIV testing at 18 months of life. Because this dataset encompassed data from a specific range of time, there are individuals who were monitored by AMPATH but had not yet reached 18 months of age prior to completion of data collection. This group was termed “HIV-indeterminate” for this study. Also included in this group were the individuals who were lost-to-follow-up prior to 18 months of age. HIV-indeterminate children (n = 1,576) were removed from the dataset. All children who were exposed to HIV but had negative confirmatory testing during the data collection period are referred to as being HEU. HEU children were eligible to receive free follow-up care with well-child visits in the AMPATH system until 5 years of age. No HIV-unexposed children are included in this cohort. Per AMPATH and Kenya’s Ministry of Health protocols, during the period of data collection, pregnant mothers with HIV were expected to be on an antiretroviral treatment (ART) regimen of tenofovir, lamivudine, and efavirenz as the first-line combination (Option B+) which would continue for life. Exclusive breastfeeding was recommended for all HIV-exposed children for the first 6 months of life and to continue breastfeeding with appropriate complementary feeding introduced thereafter [25]. Nevirapine prophylaxis was given for 6 weeks to all infants attending the AMPATH clinics and co-trimoxazole was given from 6 weeks of age until the time of confirmatory testing at 18 months of age [25]. If a child was found to be HIV+ at any time, they would be switched to either a regimen of zidovudine/lamivudine/nevirapine or abacavir/lamivudine/nevirapine as first-line treatment. In late 2013, the first-line regimen for children was changed to abacavir/lamivudine/lopinavir/ritonavir, in accordance with changes in the World Health Organization (WHO) guidelines [26]. Per AMPATH standard operating procedures, from 2010- December 2015 all children included in this study were recommended to attend monthly follow-up visits until 5 years of age [27]. In 2016, this was changed to be in line with the national policy of monthly visits until 18 month and then every 6 months until 5 years of age [25]. The variables included within this study were collected from clinic visit forms completed by a clinical officer (a mid-level provider), medical officer, or pediatrician and entered into the AMRS system. These variables included age at enrollment, sex, clinic location/type, person accompanying child, orphan status, visit height/weights, caregiver-reported feeding method, clinician-reported maternal or child ART, HIV+ sibling, mean CD4 for HIV+ children, and final HIV testing result. There were also places where a clinician could indicate whether they believed a child was considered developmentally delayed or failure to thrive, although no standard operating procedures or guidelines were available outlining the definitions of those terms. Patient identifiers, including name, address, and contact information, were removed during data extraction. All analyzed data were handled and transferred using password protected, encrypted methodologies compliant with the United States’ Health Insurance Portability and Accountability Act standards. For this study, z-scores and standardized WHO definitions were used for presentation and analysis of anthropometric data, with z-scores calculated using the modeling defined by the WHO [28]. Z-scores are expressed anthropometric values as several standard deviations below or above the reference mean or median values, that are helpful for grouping growth data by age and sex [29]. To characterize malnutrition, the following categories were evaluated: stunting, underweight, and wasting. In these analyses, “stunting” refers to moderate-to-severe stunting, (height-for-age (HFA) z-scores of ≤-2). “Underweight” refers to moderate-to-severe underweight status (weight-for-age (WFA) z-scores of ≤-2). “Wasting” refers to moderate-to-severe wasting (weight-for-height (WFH) z-scores of ≤-2). The WHO defines “moderate-severe malnutrition” based on these three variables. To minimize the influences of recording errors and data irregularity, we conducted a due-diligence examination of the height and weight growth data: We restricted the weight change to no more than +/-3 kg per month, the height change to less than 10cm per month. We excluded height measures that were shorter than previously recorded heights, and z-scores (WFA, HFA, WFH) that changed more than +/-2 units per month. These thresholds were determined upon review of World Health Organization growth charts, review of the current dataset, and by using clinical judgement. Participant characteristics at study entry were summarized in a tabular form. Frequencies and percentages were calculated for categorical variables. Any visits occurring at a certain age point included a +/- 1 month window. For example, children who came to clinic between the ages of 11 and 13 months were marked as having a 12 month visit. In the event that multiple visits occurred within that window, only the first visit’s data was included. Mean and standard deviation were calculated for continuous variables. Descriptive statistics for participant characteristics were evaluated for the full sample and for the subsamples defined by the HIV status. Comparisons between HIV+ and HEU children of characteristics presented as proportions were analyzed using Chi-squared test, while comparisons of means between the two groups were analyzed using independent t-tests. We also compared the proportions of stunting, underweight, and wasting among HIV+ and HEU children, using chi-square tests. We then performed a visit-level analysis, assessing WFA, HFA, and WFH at multiple points of observations. A random effects logistic regression model was used to accommodate within-subject correlations over time to assess the effects of factors that were associated with the risks of stunting, underweight, and wasting. Two authors (M.S.M. and R.C.V) reviewed all available variables and selected those which may be potential confounders using clinical judgement and knowledge of the literature. Univariate analysis was then performed with each of these variables, and those found to reach statistical significance (α set at 0.05) were included in the multiple variable logistical regression models. Estimated adjusted odds ratios and related confidence intervals were reported. Finally, we comparatively examined the mean HFA, WFA, and WFH curves among the two HIV subgroups, over the entire age range. Comparisons of the curves were made using a resampling-based test [30]. All analyses were implemented in SAS 9.4 (SAS Institute, Cary NC) and R 3.4.0. P values less than 0.05 were considered statistically significant.