Study design, area, and period
A retrospective record review was conducted at Debre Markos Referral Hospital of applicable patient records from between January 1, 2012 and December 31, 2017. Debre Markos Referral Hospital is located in Debre Markos Town, which is located 299 km from Addis Ababa, the capital city of Ethiopia and 265 km from Bahir-Dar, the main city of Amhara Regional State. The hospital serves more than 3.5 million people in East Gojjam Zone and neighboring areas. Apart from other services, the hospital has been providing ART follow-up care services since 2005. In the hospital, the recorded number of HIV-positive people ever started ART was 3,716, of whom, 1, 569 HIV-positive adults started ART care between January 1, 2012 and December 31, 2017.
Population
All HIV-positive adults ever started ART at Debre Markos Referral Hospital and who had at least one month of ART follow-up from January 1, 2012 to December 31, 2017 were the target population for this study. All HIV-positive adults ever started ART from January 1, 2012 to December 31, 2017, however, HIV-positive adults on ART who had TB at the beginning of the follow-up and those who had incomplete baseline data for important variables (i.e., WHO stage, CD4 counts, hemoglobin, (Hgb), IPT, CPT and level of ART adherence) were excluded from the study.
Sample size determination and sampling procedures
The minimum required sample size for this study was calculated using a sample size determination formula for survival analysis using Stata™ Version 13 statistical software by considering CD4 count, functional status, and WHO clinical staging as major exposure variables. It was calculated by considering the following statistical assumptions: two-sided significant level (α) of 5 %, power 80 %, Za/2= value at 95 % CI =1.96, q1: proportion of subjects that are in group 1 (exposed), q0: proportion of subjects that are in group 2 (unexposed); 1-q1, HR: hazard ratio, and probability of event (E) for functional status were taken from a study conducted at the University of Gondar Teaching Hospital (0.33) [23]. Finally, the calculated sample size for this study was 544. To select the study participants, the records of all HIV-positive adults ever started ART and recorded from January 1, 2012 to December 31, 2017 were sorted. Then, the study participants were selected using a simple random sampling technique through computer-generated numbers. We selected this follow-up period for two reasons: : first, we intended to have a nearest six years of follow-up; and second, to take advantage of standardized ART documentation and reporting formats used during this period.
Data collection tool and procedures
The data extraction tool was prepared from the ART entry and follow-up forms. To ensure data quality, before data collection, the data extraction tool was prepared carefully from a standardized ART intake and follow-up forms. Furthermore, before the beginning of data collection, we verified consistency between data recording systems and the prepared checklist by randomly selecting and completing a few chart reviews which resulted in minor amendments of the data collection tool. Three BSc nurses who have been working in the ART clinic of Debre Markos Referral Hospital collected the data. Two days of training was given for both data collectors and supervisor concerning the data collection tool and collection process. The supervisor and principal investigators performed a strict follow-up and supervision throughout the entire data collection period. The most recent laboratory test results before starting ART were considered as a baseline value. If there were no pre-treatment laboratory test results, obtained at the time of ART initiation, test results done within one month of ART initiation were used as a baseline data. In case of two results obtained within a month, the mean value was computed and taken as a baseline.
Variables of the Study
The dependent variable was the time to develop TB.
The predictor variables were: Socio-demographic characteristics (age, sex, marital status, residence, family size, level of education, and occupation), Baseline clinical and laboratory characteristics (WHO clinical stage, CD4 cell count, hemoglobin level, history of TB, and history of opportunistic infections (OIs) and body mass index (BMI), and ART and other medication-related characteristics (type ART regimens, regimen change, ART side effects, ART adherence, IPT, and CPT).
In this study, the focus was on health care documentation of the event (the occurrence of TB) for HIV-positive adults in cases where TB developed after ART initiation until the end of the study. This was ascertained by review of patient records.
In this study, the study participants were classified as censored in either of the following conditions: if lost to follow-up or died before developing TB or if alive at the end of the study, but didn’t develop TB and took anti-TB medications. These elements were ascertained by reviewing patient records.
Lost to follow-up was defined as an HIV-positive patient missing an ART appointment for one to three months [26].
Adherence was classified as good, fair, or poor, according to the percentage of drug dosage calculated from a monthly total dose of ART drugs; hence, good was reported if equal to or greater than 95% or ≤ 3 dose missing per month, fair if 85-94% or 4-8 dose missing per month, or poor if less than 85% or ≥ 9 dose missing per month [27].
Low hemoglobin level was defined as Hgb level less than to 10 g/dl.
Opportunistic infections were diagnosed if HIV-positive adults developed any morbidities after starting ART, as documented by the health care professionals.
Data processing and analysis
Data were entered using EPI-data™ Version 4.2, and analyzed using STATA Version 13 statistical software. Patient’s follow-up characteristics for continuous data were described in terms of central tendency, dispersion, and frequency distribution for categorical data. At the end of follow-up, the outcome of each study participant was dichotomized into censored or event. The necessary assumption of Cox proportional hazard regression model was checked using Schoenfeld residual and Log-Log plot tests. In addition, the model goodness of fit was assessed using Cox-Snell residual test and model with the least value of Akaike’s information criteria selected as the best model. The Kaplan Meier survival curve was used to estimate the TB free survival time of HIV-positive adults on ART. Log rank test was used to compare the survival curves of different categorical explanatory variables. Bi-variable Cox-proportional hazard regression model was used to screen variables for the final model. Variables having p-value ≤ 0.25 in the bi-variable analysis were fitted into the multivariable Cox-proportional hazard regression model. Finally, adjusted hazard ratio with its corresponding 95% confidence interval was reported to declare the presence of significant association between the explanatory and outcome variables.