Study settings, design, and period
A health facility-based cross-sectional study was carried out from February to March 2020 among HIV-infected children on ART between September 2016 and December 2019. In Ethiopia, chronic HIV/AIDS care and treatment services are provided at hospital and health center settings throughout the country. This study was conducted in Bahir Dar town public health facilities, which is 565 km away from Addis Ababa. The study was conducted in four selected public health facilities (i.e. Bahir Dar Health Center, Abay Health Center, Han Health Center, and Felege Hiwot Comprehensive Specialized Hospital).
Study participants, sample size, and sampling technique
All HIV-infected children (aged <15years) taking ART in Bahir Dar Town public health facility were the target population. All HIV-infected children who took ART at least for nine months with documented viral load test results were included. Conversely, children with incomplete medical records (i.e. age of child at ART initiation, baseline ART regimen, current CD4 cell counts and unknown outcome status) were excluded.
The minimum required sample size was determined using a single population proportion formula. To compute the sample size, the following statistical assumptions were considered: - prevalence of VF =50% because there was no published study done on this topic in Ethiopia, margin of error =5% and the value of Zα/2 =1.96, which is the corresponding Z score of 95% confidence interval (CI).
Where, n= the required sample size, Zα/2= Standard normal variation for type 1 error, p=prevalence (0.5) & d= Margin of sampling error tolerated (0.05).
The calculated sample size was 385. By considering a 10 % contingency rate for incomplete charts, the final minimum required sample size of this study was 424 medical records.
This study was conducted in three randomly selected health centers, and one purposively (because it has high patient flow and case-team composite) selected comprehensive specialized hospital. First, a sampling frame was prepared based on the patient's medical registration number (MRN) from each health facility's recorded documents. The total sample sizes were then allocated for each healthy facility proportionally to the number of population size. Finally, medical charts of HIV-infected children taking ART at the selected public health facilities were sorted and selected using a simple random sampling technique (Figure.1).
Data extraction procedure
Available information from patient's chart was extracted using structured checklist, prepared in English. The data extraction checklist was adapted from the Ethiopian Federal Ministry of Health ART clinic intake and follow-up forms and included socio-demographic, clinical, laboratory, and treatment-related characteristics. Data were extracted by trained health professionals through document review. Five ART trained nurses were recruited as data collector. Orientation about the objectives of the study, contents of the tool, and data extraction procedures was given for data collectors and supervisors for one day. The assigned supervisors and principal investigator closely monitored the whole data collection process. Besides, the consistency between collected data and medical records was checked using randomly selected reviews of previously extracted charts.
Virologic failure: was diagnosed when the viral load above or equal to 1000 copies/ml under ART based on two consecutive VL results three months apart, with adherence support following the first viral load test after at least six months of ART (5).
Adherence: was defined as “good”, “fair” or “poor” if the patient took ≥95% (missing one from 30 doses or two out of 60 doses), 85-94% (missing 2-4 doses out of 30 doses or 4-9 from 60 doses) or ˂85% (missing ≥ 5 doses from 30 doses or >10 from 60 doses), respectively of monthly doses (4).
Treatment interruption: was defined as a treatment interruption for at least one week during the previous six months.
Data management and statistical analysis
Data were entered into Epi Data Version 3.1 and exported to Statistics Package for Social Science (SPSS) Version 25 for further analysis. Tables were used to present descriptive results. Additionally, frequencies, percentages, proportions, and summary statistics (mean, median) were used to summarize the study population characteristics. Variables with p-values < 0.25 in the bivariable analysis were entered into the multivariable analysis to control the effects of confounders. Goodness of fit of the model was checked using Hosemer-Lemeshow goodness of fit test. In the multivariable analysis, variables with p-values less than 0.05 were considered as statistically significant factors. Lastly, odds ratios with their correspondence 95% CIs were used to assess the strength and the direction of association.