Study design and setting
An institution-based interrupted time series design was conducted at public health facilities in Debre Markos town from October to November 30/2020. Debre Markos is the capital town of the East Gojjam zone, located 300 kilometers from Addis Ababa, the capital city of Ethiopia, and 265 kilometers from Bahir-Dar, the capital of the Amhara regional state. The town has four public health facilities sites, which provide ART services for the town and the catchment area population. Debre Markos health center and Debre Markos referral hospital started to implement ASM and viral load tests, whereas Wuseta and Hidase health centers do not provide ASM. According to health management information system (HMIS) reports, there were a total of 5060 clients taking ART in Debre Markos town. From these, 3,168 clients were enrolled in ASM until June 2020.
Population and sample
All adult HIV patients who were enrolled into ASM for HIV care and treatment were the source population. Whereas, the study population were those patients who had at least two viral load measurements before and after ASM in the selected public health facilities were included. Patients who were lost to follow-up, discontinued from ASM, died, and transferred out from the catchment area were excluded. The target sample size was calculated by considering the previous study report as 68% of patients had virological suppression in the standard of care (three months follow-up) and 79% had viral suppression after being enrolled in ASM[27]. Using G-power version 3.1.9.4 software by assuming 5% margin of error, 95% level of confidence, and 80% power, then the sample size to determine viral suppression was calculated by the following formula:

The final sample size was 276 people after a 10% attrition rate. A simple random sampling technique with proportional allocations was used to determine the participants from each facility.

Study Variables and Operational Definition
The dependent variable was virological suppression, whereas socio-demographic and clinical characteristics like age, sex, religion, marital status, educational level, occupation, residence disclosure status, treatment supporter, nutritional status, regimen type, adherence, Isoniazid, and cotrimoxazole preventive therapy (CPT), WHO clinical stages, TB-co infection, functional status, and opportunistic infections other than TB were collected.
Virological suppression: Viral load below the detected threshold using viral assay (<1000copies of viral RNA/ml of blood)after taking plasma and separated from whole blood[1, 28].
Appointment spacing model: Stable adult patients are offered the opportunity to provide six months' worth of ART and have a clinical follow-up at each visit to an appointment[29].
Stable patient: Defined as an individual’s>15 years old and on ART for at least one year, no adverse drug reactions requiring regular monitoring, good understanding of lifelong adherence, two consecutive viral load measurements <1000copies/ml, no acute illness, not pregnant or breastfeeding[29].
Adherence: Drug adherence was defined as the percentage of doses taken as prescribed; using a number of doses missed in the last one month.; good if equal to or greater than 95% adherence i.e., missing only 1 out of 30 doses or missing 2 from 60 doses. Fair if the clients taking 85-94% of the prescribed medications and poor less than 85% adherence[30]. However, drug non-adherence was defined if there was history of at least one poor or fair drug adherence throughout the study period.
Data collection and quality control
Data were collected using an extraction checklist prepared in English and extracted from HMIS reports, patient medical charts, and computer databases. Four data collectors (two data clerks and two monitoring and evaluation officers), and two supervisors who were working at ART treatment initiation centers were recruited. Two days’ intensive training regarding the objective of the study, and how to review the documents, as well as about confidentiality of information was given to data collectors and supervisors. Before data collection, records (both baseline and follow-up) were reviewed and identified by their medical registration.
Data management and analysis
Data were checked for completeness, edited, coded, and entered using Epi-data version 3.1 and exported to STATA version 14 software for analysis. Descriptive statistics, including frequencies, proportions, and scatter plots to show the patterns and trends, were computed to summarize the variables. For the goodness of fit of the model, R and adjusted R square were used. The presence of 1st order-autocorrelation was tested by using Durbin-Watson (DW) statistics and co-linearity as well as multi co-linearity between independent variables were assessed by a variance inflation factor (VIF) and tolerance test. A logistic segmented regression model was used to estimate trend and level changes from pre-intervention to post intervention.
We estimated the level and trend changes before and after ASM at public health facilities in Debre Markos town using a segmented regression model. The following multivariable regression model was specified to estimate the level and trend of virological suppression among clients enrolled in the ASM. Yt=βo+β1timet+β2 interventiont+β3time after interventiont+ et. Here, Yt is the mean virological suppression per patient in month t; time is a continuous variable indicated in months at time t from the start of the observation period; intervention is an indicator for time t occurring before (intervention=0) or after (intervention=1) ASM, which was implemented at month 30 in the series; and time after the intervention is a continuous variable containing the number of months after the intervention at time t coded as 0 before ASM and (time-24) after the ASM. In this model, βο estimates the baseline levels of the outcome; mean virological suppression per patient per month, at time zero; β1 estimates the change in the trend of virological suppression per patient for each observation; β2 estimates the level change just after the intervention, and βꝪ estimates the change in the trend of virological suppression per patient after the implementation of ASM, which compares the trend before ASM and after ASM. The sum of β1 and βꝪ is the post-intervention slope. To estimate the level and trend change associated with the intervention, controlling the baseline level and trends. The error term (et) at time t represents random variability not explained by the model. Parameter estimates from the segmented regression model effects of ASM on virological suppression; adherence was included in the model since it was the most influential variable in this study.
Regression coefficients with a 95% confidence interval (CI) were used to determine the strength of the association between the dependent and independent variables. Variables with a p-value < 0.05 in the segmented regression model were considered as statistically significant predictors of virological suppression.