Study area, design and population
We conducted a retrospective cohort study on 429 children on ART on treatment and care from January 1/2009 to December 31 /2018 at Assosa & Pawe general hospitals in Benishangule Gumuz regions. Both hospitals are located in this regional state in North West Ethiopia. This region is one of the nine regions in Ethiopia. Assosa is the capital city of this region and it is located at a distance of 659 km in west of Addis Ababa and Pawe hospital is also located a distance from 565 km from Addis Ababa in North West direction. This region has currently 2 general and 3 primary hospitals with one regional laboratory. This two selected Hospitals are routinely diagnose and treat tuberculosis based on the clinical findings, chest x-ray, AFB and XpertTB for suspected TB patients  . In both general hospitals there has been given ART care service 2007 pediatric HIV/AIDS guideline . Following the time of enrollment to ART care continuum, all children have started ARV at both hospitals. Among these, 238 and 191 children were on follow up and care at Assosa general hospitals and Pawe general hospitals, respectively. From the registration logbook, eight children with incomplete outcome data were excluded from the study.
Sample size determination and sampling procedure
Sample size for this study was calculated by using EPI INFO software using the following parameters. A) (α) of 5%, power 80%, Z = within 95% CI = 1.96 and AHR=2.39  (P1) =6.6% and (P2) = 15.8% obtained 408 by adding 5% incomplete data final sample size will be 421.Computer generated random number used for final study subject of study subject from two hospitals. Totally there existed 723 children started HAART and registered on computer SMART DATA sets of ART registration office since January 1st /2009 – 31st December 2018. Assosa hospitals 407 and Pawe general hospital 316 children were treated. There for proportionally allocated to selected 421 samples from two hospitals.
Assosa Hospital ni1= (N2)n = (421)(407)=236
Pawe Hospitals ni2= (N2)n = (316)(421)= 185
First by using Unique ART number of each individual card retrieved from SMART DATA set of ART children ART office. Then by using computer generated random number 236 study participants from Assosa hospitals and 185 study participants from pawe general hospitals were recruited,
In this study, the outcome variable was time to develop TB. Incident TB cases were only those who developed new TB (EPTB & PTB) during the follow up period. The outcome variables ascertained if TB occurred only after started ART during ART follow up times.
Independent variables included: Age of children, sex, residence, family size, WHO clinical stage TB contact history, CD4 counts, Hgb, functional status, Isoniazid preventive therapy, Cotrimoxazoles preventive therapy, vaccination status, weight for age (under nutrition), weight -for -height (wasting) and height –for –age (stunting).
Case ascertainment: The outcome variables (TB) was diagnosed based on bacteriological, molecular, histopathology and clinical methods by using ( microscope, sputum culture, chest x-ray, and Xpert or combinations) during patient presentation for TB symptoms .
Pulmonary tuberculosis: Pulmonary tuberculosis (PTB) refers to a case of TB involving the lung parenchyma. Military tuberculosis is also classified as pulmonary TB because there are lesions in the lungs. Extra pulmonary tuberculosis (EPTB): refers to a case of TB involving organs other than the lungs .Event: New occurrence of tuberculosis during ART care follows up times with study in periods Censored: HIV positive children who did not developed TB during ART follow up.
TB history of contact: Children during ART follow up before TB incidence developed, having history of survives or contact at any time with who has active PTB patient.
Opportunistic infection: for HIV infected children during the following if any one of diseased developed registered on ART follow up form by their code (BP=Bacterial pneumonia ,UL= oral ulcer , Z=Herpes zoster, PCP = pneumocystis carnie pneumonia ,DC/DA –chronic / acute diarrhea, CT= central nervous toxoplasmosis CM streptococcal meningitis .
CD4 :was classified as below the threshold according to the following age-specific thresholds: less than 15% for children aged 12–35 months, less than 10% for children aged 36–59 months or less than 100 cells/mm3 for children aged 5–15 years .
Seropositive: children<15 years were confirmed diagnosed of HIV /AIDS and under follow up.
Stunting, underweight, and wasting: The child being 2 standard deviations (SDs) below the normal for height for age, weight for age, or weight for height, according to the WHO 2006 curve. For children under or equal age 2, wasting was measured by weight for length Z-score; for children above age 2, wasting was defined by Z-score. Z-score ≥ −2 was defined as non-wasting; −3 ≤ Z-score ≤ −2 was defined as moderate wasting; Z-score ≤ −3 was defined as severe wasting. Stunting was measured by height/length for age Z-score. Z-score ≥ −2 was defined as non- stunting; −3 ≤ Z-score ≤ −2 was defined as moderate stunting; Z-score ≤ −3 was defined as severe stunting [10, 17, 22].
Data collection tools, procedures, and quality control
Four bachelor nurses and two supervisors were selected for data collection processes and all had took ART training. For quality of data collection process, one-day traing was given in two hospitals with two supervisors for data collectors. The principal investigator and two supervisor followed data. Data were collected using the data abstraction tool and medical history sheet prepared from Ethiopian Federal ministry of health HIV/AIDS follow up forms .
Data processing and analysis
Data entered into the computer using EPI-DATA version 3.1 & exported to STATA 14.1 for cleaning and analysis. Descriptive analysis, such as tables, graphs, Kaplan Meier survival curve and log rank test was done. Hazard ratio with 95%CI & P≤0.05 was used to measure association with independent variable. The overall survival graph and hazard failure estimated curve was used to show survival and hazed probability of risk group. Cox-regression model was fitted to identify predictors for incidence of pulmonary tuberculosis. All predictors that was associated with the outcome variables in the bivariable analysis at a hazard ratio of P-value 0.25 or lower was included in multi variable Cox-regression model. Variables with adjusted hazard ratio in multivariable Cox-regression with their corresponding 95% confidence interval with P-value <0.05 was considered as significant predictors .Cox–proportional hazard assumption was checked by (log-log plot) & expected versus observed Kaplan Meier graph test for each variable with schoenfield residuals test for each variable. No variables less than <0.05.After multivariable cox regression was built by transforming from bivariable P<0.25, for finally model selection was selected by AIC & BIC criteria .Finally, model adequacy was checked by Nelson Alana and Cox Snell residual combination was used for checked model adequacy & it became on straight line with Zero origin in X and Y axis.