A multicentre retrospective follow-up study was conducted from January 1, 2015, to December 31, 2024. The study was conducted in Amhara region’s comprehensive specialized hospitals. These include Woldia, Dessie, Debrebirhan, Tibebe Gion, Debretabor, the University of Gondar, Felege Hiwot, and Debre Markos comprehensive specialized hospital. The hospitals serve more than 2.7 million, 2.5 million, 2.8 million, 4.5 million, 3.1 million, 7 million, 5 million, and 3.5 million peoples who came from the catchment area in their respective orders. Apart from other services, pediatric antiretroviral therapy services have been delivered in all public hospitals since 2005. From January 1, 2015, to December 31, 2024, a total of 1004 HIV-infected children aged younger than 14 years have been initiated ART services across these hospitals and all children were followed till the event of interest developed or become censored. Moreover, the follow-up was stopped when the child become adult (age >=15) during the course of treatment by considering them as right censored. From the eight hospitals, four comprehensive specialized hospitals were selected: University of Gondar hospital, Felege Hiwot, Woldia, and Debre Birhan hospital.

All children under 15 years of age started ART in Amhara region comprehensive specialized hospital.

All children under 15 years of age who started anti-retroviral therapy in four randomly selected Comprehensive hospitals in Amhara region after the implementation of the new guideline, from January 01, 2015, to December 31, 2024.

Children aged less than 15 who started ART from January 01, 2015, to December 31, 2024, after the implementation of the new guideline.

Incomplete patient charts on the day of ART initiation and children transferred out to other facilities during the study period were excluded from the study.

Sample size determination

The sample size was calculated using a double population proportion formula by Epi-Info 7. To determine the sample size, a cohort study conducted at Felege Hiwot Referral Hospital in Bahirdar, Ethiopia was considered. ¹⁷ In that study, the statistically significant predictors of mortality were taking cotrimoxazole prophylaxis, low hemoglobin level, delayed developmental milestone, and absolute CD4 count below the threshold. Therefore, the sample size was calculated using the mortality rate of 10% in children with low hemoglobin levels (exposed) and the mortality rate of 4% in children with normal hemoglobin levels (unexposed), by using 95% CI, with a 1:1 ratio, 80% power and 5% type I error. The resulting sample size was 204 in unexposed and 204 in exposed with a total sample size of 408. After multiplying by a design effect of 1.5 to adjust for the sampling error due to the two-stage sampling method we used and to increase the precision of the study estimates, the final sample size became 612.

Sampling technique

Among eight hospitals that had pediatric ART clinics in the region with strong pediatric HIV/AIDS care and treatment centers, four comprehensive hospitals (University of Gondar Hospital, Felege Hiwot, Woldia, and Debrebirhan) were randomly selected. The sample size was proportionally allocated to the four selected hospitals based on the number of target children. The records of children aged less than 15 who started anti-retroviral therapy from January 01, 2015, to December 31, 2024, were filtered from the database. Then, the required samples were selected from each of the four selected hospitals using simple random sampling.

Independent variable

The independent variables were socio-demographic and socio-economic characteristics and family care (child age, child sex, caregiver educational status, caregiver marital status, caregiver and caregiver employment status). Clinical and immunologic information like the presence of Tuberculosis (TB), TB treatment, opportunistic infections (OI), developmental milestones, WHO clinical stages, prophylaxis, anthropometric measurement, CD4 count, ART adherence, hemoglobin level, and ART treatment regimen.

Dependent variable

The dependent variable was the time to death of HIV-infected children after initiation of HAART.

A data extraction tool was used for recording the data from electronic databases and patient cards. This form was developed using a standardized ART entry and follow-up form employed by the ART clinic. The laboratory results recorded before starting ART were used as baseline values. If there was no pretreatment laboratory test, the results obtained within one month of ART initiation were considered baseline values. In each facility, two ART nurses collected the data, and the data collection was supervised by a trained supervisor.

The data review tool was carefully designed and prepared in the English language using a standardized ART entry and follow-up form employed by the ART clinic. Before actual data collection, a pre-test was done to check data review tool validity and reliability using 5% of the sample size in one of the facilities not selected for the actual study. The training was given to both data collectors and a supervisor about how to use data review tools, how to select study participants according to eligibility criteria, and how to collect the data. The filled record review tools were gathered and checked for completeness by the supervisor daily. The principal investigator also checked the data collection process and supervised overall research project activities.

After data entry was completed, the data was exported and analyzed using Stata software. Data cleaning followed by exploration was undertaken to see if there were items that were not logical and then subsequent edits were made. The patients’ cohort characteristics were described in terms of mean with standard deviation (SD) and median value with interquartile range (IQR) for continuous data and in terms of frequency and percentage for categorical data. The study participants were followed from the date of ART initiation until the occurrence of death confirmed by reviewing medical registration in the hospital or registration by ART adherence supporters (from patient card) or, date lost to follow up (last visit) or the end date of the follow-up period (December 31, 2024). During analysis, the status of each participant was dichotomized into censored or death.

Life Table analysis was carried out to estimate the mortality rate of children and Kaplan Meier survival curve with a log-rank test was fitted to test for the presence of a difference in survival time among different predictor variables. Cox-proportional hazard regression analysis was used to identify predictors of time to death. Multicollinearity between predictor variables was checked using variance inflation factors (VIF) and there was no multicollinearity between predictor variables. The proportionality of the hazard assumption was checked using the Log (-Log) S (t) plots. The crude and adjusted hazard ratios with their 95% confidence intervals (CI) were estimated and a p-value less than 0.05 was used to declare the presence of a significant association between time to death and covariates.

From the 612 reviews done, 10 patient cards were excluded with many missing values, and 602 cards were reviewed which gave an enrollment rate of 98.3%. The mean age of children under follow-up was 7 years with a standard deviation (SD) of 4 years. Regarding the sex of the children, 318 (52.8%) of them were males. The majority of the caregivers, 356 (64.6%), were married, 390 (72.5%) had primary level and above educational status and 222 (41.4%) were government-employed. Regarding the place of residence, 394 (78.2%) live in the catchment area of the hospitals ( Table 1).

Among under five years children, 172 (91.9%) had an appropriate developmental status at admission and among five years and older children about 379 (93.5%) of them had working functional status. Regarding the WHO clinical staging of the children at the baseline, 157 (26.1%) of them were at WHO clinical stage two, and 222 (36.9%) of children were at WHO clinical stage three. The CD4 count at the baseline was below the threshold among 227 (38.9%) of the children and 105 (19.2%) were found to be anemic during the baseline hemoglobin measurement ( Table 2).

From the total of 602 children under follow-up, 269 (44.7%) developed OI, and diarrhea and TB were the common OIs; 94 (34.9%) and 75 (27.9%) respectively. CPT is given to children under ART either as a prophylaxis or as a treatment for different OIs and 436 (72.4%) of children took CPT. Similarly, anti-TB drugs were given to the children under ART either as a prophylaxis or as a treatment for those who had TB, and 92 (15.3%) of the total children under follow-up took these anti-TB drugs.

Regarding the children in the follow-up ART drug regimen, the majority of them took first-line ART regimen particularly; 287 (47.7%) of the children took 4g and 4f ART drug regimen and 184 (30.6%) them took 4j and 1j ART drug regimen. Only 14 (2.3%) took the second-line ART regimen. Among the total of 602 children under follow-up, 468 (77.7%) had a good or fair ARV treatment adherence while the rest 62 (10.3%) had poor adherence to ARV treatment ( Table 3).

Of the total children who were initiated on ART under the follow-up period, 45 (7.5%) of them died, 468 (77.7%) were alive and 89 (14.8%) were lost from follow-up. The overall median (IQR) follow-up time was 47 (26-60) months with a total follow-up time of 2140 years. The overall death rate was 2.1 (95% CI: 1.6-2.8) per 100 person-years of follow-up. The cumulative survival probabilities of HIV-infected children after 6, 12, 36, and 60 months of ART initiation were 98%, 97%, 93% and 89% respectively ( Table 4).

Using the Kaplan-Meir survival function, the survival experience of children was assessed among different categories of predictors. Among all independent variables, four children’s status and two caregiver characteristics showed significant differences within the different categories. Under five children have a shorter survival experience than children older than five years (Log-rank test X ²-value=12.9, P-value=0.002). Children with OIs have shorter survival experiences than those without OIs (Log-rank test X ²-value =24.4, P-value=0.0001). The survival experience of children who took TB prophylaxis/treatment during the follow-up was significantly longer than those who didn’t take TB prophylaxis/treatment (Log-rank test X ²-value =8.3, P-value=0.004). Children with good and fair ARV adherence have longer survival experiences than those with poor ARV treatment adherence (Log-rank test X ²-value=20.9, P-value=0.0001). Children with normal CD4 count at the baseline assessment have longer survival experiences than those with below the threshold CD4 count (Log-rank test X ²-value=25.1, P-value=0.0001). Similarly, children with baseline WHO classification of stage I & II have longer survival experiences than those with baseline WHO classification of stage III & IV (Log-rank test X ²-value =7.2, P-value=0.007) ( Figures 1- 4).

From the caregiver characteristics; children with a caregiver of no educational status had shorter survival experiences than those who attended a formal educational status (Log-rank test X ²-value=13.3, P-value=0.004). Finally, children from a married caregiver had a longer survival experience than children from a widowed or divorced caregiver (Log-rank test X ²-value=11.1, P-value=0.01).

The Cox proportional regression model was used to determine the predictors of survival of HIV-infected children. Before fitting the covariate into the model, the proportional hazard assumption was checked by examining Log (-Log S (t)) plots. Overall, five variables; child age, CD4 counts, WHO Clinical Stage, OI, ART adherence, and Caregiver educational status were found to be significantly associated with time to death.

According to the final Cox proportional regression model, children with a baseline CD4 count below the threshold were 2.6 times more likely to die than those children with normal baseline CD4 count [AHR: 2.6; 95% CI: 1.2-5.8], children with OIs during the follow up were 3.7 times more likely to die than children without OIs during the follow-up time [AHR: 3.7; 95% CI: 1.7-7.9]. Likewise, those who had poor ARV adherence are nearly three times more likely to die than their counterparts [AHR: 2.9; 95% CI: 1.4-5.7]. In addition, children whose caregivers with no formal educational status were 3.4 times more likely to die than those children with caregiver educational status of tertiary or above education [AHR: 3.4; 95% CI: 1.1-11.6] ( Table 5).

I want to inform our readers that the incidence of death may be undermined by the absence of survival data on the patient card, hence we, considered them as lost follow-up. Also, incomplete information concerning important variables like nutritional status and other clinical variables was not found, which might confound the results. In addition, the exact day of lost follow-up is unknown which may reduce the survival time which could affect the estimations and the need to get adequate information about the cause of death to identify the actual cause of death for those patients reported as dead.