Study Setting፡ Addis Ababa is an administrative and economic capital of Ethiopia, and home for an estimated population of 5.2 million. It has 11 sub-cities and more than 115 functional public health facilities (13 public hospitals and 102 health centers) that provide healthcare services for the public. The aggregate data from district health information system (DHIS2) showed that there are 78 Health centers and 11 public hospitals which are providing ART services. These ART sites compile and sent monthly report to the Addis Ababa Regional Health bureau on a regular base. Including the public health facilities, 114 ART sites use an Enhanced Electronic Medical Record- ART database system (EMR-ART system) developed by ICAP (International Center for AIDS care and treatment Program)- Ethiopia, to collect and manage individual level ART data(29). The study collected the database backup data from 18 public health facilities in Addis Ababa. The list of public facility sites from where the database backup file collected were eight public hospitals: Zewditu Memorial, ALERT, Black Lion, St.Paul, St. Peter, Yekatit-12, Ras Desta Memorial and Gandhi Memorial hospital, and ten public health centers: Efoita, Feres, Hiwotamba, JagemaKelo, Keraneyo, Kirkose, kolfe, Lideta, Mikilayland, and Kolfe Woreda 9 health center. All clients receive ART services in these public facilities free of charge.
Study design
We conducted a multicenter retrospective cohort analysis of the effect of the same day ART initiation on virological outcome among adults ≥ 15yrs of age PLHIV who started ART from 2017 to 2021 in Addis Ababa. We took individuals who started ART within five years of duration as sample size wouldn’t be enough when the exclusion criteria applied. The study used secondary extracted from enhanced standalone EMR- ART database backup files from 18 public health facilities, and merged to a single database.
Study population
Adults with age ≥ 15 years at time of ART enrolment and started ART from January 2017 to February 2021 who were on treatment for a minimum of 12 months, and with at least one documented viral load status at 12th month of treatment were included in this study.
We excluded those transferred in, pregnant, TB and cryptoccocal meningitis cases, and those who were transferred out before 12 months of treatment from the study.
Sample size
To achieve representative sample size of the population who started ART on the same day and after seven days ART initiation for viral load suppression at 12 months, we considered the proportions of the viral suppression at 12 months in North West Ethiopia as estimates 73.4 percent for p1 (the proportion of viral suppression among clients who started ART on the same day diagnosed with HIV), and 83.7 percent p2 (for those who started ART after seven days diagnosed with HIV). With 95% confidence interval, and 80% study power, the sample size calculated using Epiinfo version 7.2.5.0 software. With additional 15 percent to compensate missing records, the total sample size was 616.
Sampling procedure
We collected database back-up files from 8 public hospitals and 10 health centers. The database backup files were collected from each health facility starting from July 2022 to January 2023 at various time during EMR-ART system upgrading task performed by ICAP. Using a database tool known as Microsoft SQL (Structured Query Language) 2019 Express Management Studio, the database backup file for each facility was restored to a separate instance of database, and merged to a single database by querying the different tables and views. During this procedure, we deliberately removed the individual’s full name and phone number from the merged database and inserted Health Facility name to make records confidential and allow analysis by facility type respectively. Then, we further extracted records of individuals who started ART from January 2017 to February 2021. As next step, we created different cascaded views by removing deprecated records, and records of individuals on ART and who were under 15 yrs. of age, transfer in, pregnant, TB and Cryptococci meningitis case at time of ART initiation.
Before the final extraction of the data from Microsoft SQL 2019 express, functional status code was translated, Months on ART, Months Viral Load performed after ART initiation, and number of days elapsed after HIV diagnosis were computed. Using Microsoft Excel 2013, the data set was extracted from SQL database to Microsoft excel. Using add-in tool of MS Excel, ‘Ablebits’, we trans-positioned the individual’s follow up records, and created a single row individual’s longitudinal records.
Data collection tools & procedures
The source of this research data was enhanced Electronic Medical Recording System–ART software. This software upgraded and deployed to a total of 99 public and non-governmental ART service providing sites in Addis Ababa by ICAP in 2018. It was initially developed and deployed by Tulane University in 2015 to ART sites in Ethiopia. Using free version of Microsoft SQL 2019 Express, we were able to restore each database file of 18 public health facilities and merged to a single database instance.
Data quality
Those deprecated (not usable any more) records were removed from the merged database. Unique ART number and Follow-up date were used to remove the duplicated initial and follow-up records. Since there was no time to correct records with invalid data with the paper register (taking it as a reference), we removed a total of 236 individual records before transferring the final dataset to SPSS version 20 statistical software. To understand the volume of data completeness or extent of the missing values to each independent variable, we applied multiple imputation pattern analysis technique using SPSS.
Study Variables
The same day ART initiation was the exposure variable and computed as the difference between the dates by which ART was initiated and the individual was diagnosed to have HIV (Date HIV diagnoses confirmed). For analysis purpose, we categorized ART initiation date of the participant into two groups, those participants who started ART on the same day, and those who started ART after seven days diagnosed to have HIV. The study outcome variable was viral load status at 12th month. The viral load status was recorded as ‘suppressed or ‘undetectable’ in the EMR database if viral load count was less or equal to 1000 copies per one milliliter (ml) of blood. Whereas, the viral load status was labeled as ‘unsuppressed’ or detectable’ when the viral load count was greater than 1000 copies per 1 ml of blood. The viral load count was not recorded for all individuals along with the viral load status in the database, and thus couldn’t further categorize the level of viral load status among the study participants. The independent variables included in the study were sex, age at enrollment, educational level, marital status, baseline CD4 count, Body Mass Index (BMI), WHO Clinical Stage, Functional status, ART regimen, Adherence, and Months on ART. For the dependent variable of the 12th month viral load status, we included those records with documented viral load status at 10th, 11th 12th, 13th, 14th and 15th months after starting ART. We categorized the values of the dependent and independent variables to the appropriate group for analysis and comparison with other studies.
Data Analysis
Data were extracted form enhanced EMR-ART and merged to a single database on Microsoft SQL 2019 Express. The data was imported to Microsoft 2013 excel for further data cleaning and transposed the follow-up records to create the longitudinal record for each individual case on ART treatment. After importing the data to SPSS version 20, we transformed the data to be ready for analysis. Descriptive statistics were used to characterize the dependent and independent variables of the two groups. We conducted bivariate analysis to examine whether there was an association between each independent variable of the study with viral load status. In addition, multi-variables logistic regressions analysis was conducted to adjust for potential confounders of the predicators of viral load suppression. Comparisons of the proportion of viral load suppression between the same day and after 7 days ART initiators were achieved using Pearson’s chi-square test.
To conduct the data analysis for the secondary objective, we examined the pattern of missing values among the independent variables using the multiple imputation analysis technique, and decided which variable to include or exclude from the multivariable analysis. As a result, we found that the baseline CD4 count variable had missing value of 88.7 percent (721/812), and occupation had 88.1%( 716/812) of missing values. The missing values for initial WHO clinical stage was 37.9 (308/812). Initial BMI with 33% (268/812), Initial functional status with 27.3% (222/812), Education with 13.1%( 111/812), Marital status with 11.3% (92/812) and religion with 7.6 Percent (62/812) of the records had missing values. The other independent variables Sex, Age, Regimen, and Facility type had no missing values. Therefore, we excluded those variables with missing values more than 30 percent from the multi-variable analysis.
Operational definitions
Based on the nature of the data and the World Health Organization (WHO) standards, we defined the exposure and the outcome variables as follows
-
Viral suppression at 12th month: In this research work, the viral load suppression is achieved if a viral load count is 1000 or lower copies per ml at 12th after initiation of ART. Because of the viral load test service availability and accessibility in Ethiopia, especially prior to COVID 19 era, it was difficult to get documented viral load test result exactly at 12 month. Thus, we considered test results document at 10th, 11th, 12th, 13th, 14th and 15th month of antiretroviral treatment as Viral load test at 12th month.
-
The same day ART initiation: When PLHIV starts ART on the day he or she is diagnosed to have HIV infection.
Ethical considerations
Ethical clearance was obtained from Addis Continental Institute of Public health and shared with Addis Ababa Regional Health Bureau. The individual identifiers were masked or de-identified before data extraction and manipulation started.