Background: The Health Extension Program (HEP) is Ethiopia’s flagship community health program, launched in 2003. Health Extension Workers (HEWs) are key vehicles for the delivery of the HEP. While it is believed that there is high attrition among HEWs, the magnitude of or reasons for attrition is unknown. Their intention to leave their jobs in the next 5 years has also never been investigated on a national scale. This study aimed to assess the magnitude of, and factors affecting HEWs’ attrition and intention to leave in Ethiopia. Methods: The study used mixed methods to address the research objectives. Using stratified random sampling and regions as strata, 85 districts from nine regions were randomly selected in Ethiopia. Within each study district, six kebeles (village clusters) were randomly selected, and all HEWs working in these kebeles were interviewed to capture their 5-year intention to leave. The study team developed a data-extraction tool for a rapid review of district-level documents covering the period June 30, 2004 through June 30, 2019 to gather their attrition figures. We used survival analysis to model attrition data and checked model goodness-of-fit using the Cox–Snell residual test. We additionally collected qualitative data from HEWs who had left their positions. Results: The attrition of HEWS over the lifespan of the HEP was 21.1% (95% CI 17.5–25.3%), and the median time to exit from HEWs workforce was 5.8 years. The incidence rate was 3.1% [95% CI 2.8–3.4]. The risk of attrition was lower amongst HEWs with level four certifications, with children, and among those working in urban settings. By contrast, HEWs who were not certified with a certificate of competency (COC), who were deployed after 2008, and those who were diploma/degree holders were more likely to exit the HEWs workforce. The magnitude of intention to leave was 39.5% (95% CI 32.5–47%) and the primary reasons to leave were low incentives, dearth of career development opportunities (50.8%), high workload (24.2%), and other psychosocial factors (25%). Conclusion: Although the magnitude of attrition is not worryingly high, we see high magnitude in HEWs’ intention to leave, indicating a dissatisfied workforce. Multiple factors have contributed to attrition and intention to leave, the prevalence of many of which can be reduced to fit the needs of this workforce and to retain them for the sustained delivery of primary healthcare in the country. Ensuring HEWs’ job satisfaction is important and linked with their career development and potentially higher rates of retention.

Background: Household contacts of tuberculosis (TB) patients are at a greater risk of infection and developing TB as well. Despite recommendations to actively screen such high-risk groups for TB, it is poorly implemented in Ethiopia. A community-based household contact screening was conducted to compare the yield of two different screening approaches and to identify factors associated with TB occurrence. Methods: Smear-positive pulmonary TB index cases from six health facilities in six districts of Silti Zone were identified and enrolled prospectively between September 2020 and December 2022. Trained healthcare workers conducted house visits to screen household contacts for TB. WHO (World Health Organization) recommended symptom-based screening algorithms were used. The yield of screening was compared between a two-time screening at study site I and a single baseline screening at study site II, which is the current programmatic approach. Generalized estimating equation was used to run multivariate logistic regression to identify factors associated with TB occurrence. Results: A total of 387 index TB cases (193 at site I and 194 at site II) with 1,276 eligible contacts were included for analysis. The TB yield of repeat screening approach did not show a significant difference compared to a single screening (2.3% at site I vs. 1.1% at site II, p < 0.072). The number needed to screen was 44 and 87 for the repeat and single screening, respectively, indicating a high TB burden in both settings. The screening algorithm for patients with comorbidities of asthma and heart failure had a 100% sensitivity, 19.1% specificity and a positive predictive value of 5.6%. Cough [AOR: 10.9, 95%CI: 2.55,46.37], fatigue [AOR: 6.1, 95%CI: 1.76,21.29], daily duration of contact with index case [AOR: 4.6, 95%CI; 1.57,13.43] and age of index cases [AOR: 0.9, 95%CI; 0.91–0.99] were associated with the occurrence of TB among household contacts. Conclusion: Our study showed that the yield of TB was not significantly different between one-time screening and repeat screening. Although repeat screening has made an addition to case notification, it should be practiced only if resources permit. Cough, fatigue, duration of contact and age of index cases were factors associated with TB. Further studies are needed to establish the association between older age and the risk of transmitting TB.

Background: Ensuring regular supervision, capacity building and motivation are crucial for the successful retention of health extension workers (HEWs). Failure in these aspects could increase the attrition rate of HEWs. To date, there has not been a comprehensive nationwide study on HEW attrition that could act as a source of evidence for policy makers. This study explored HEW attrition, including leaving the health sector entirely and its regional variation, trends and predictors of attrition out of the health system. Methods: This study explored the attrition of HEWs from the beginning of the program until the end of 2018. A district-based mixed method study was conducted to review the personnel files of HEWs. A multistage sampling technique was employed to select 3,476 HEWs, and a probability weight was assigned for each observation. Descriptive statistics were calculated for the outcome and predictor variables. A logistic regression model was used to model attrition out of the health system. A qualitative study was conducted to understand the reasons why HEWs leave their jobs. Thematic analysis was performed using Nvivo version 12. Results: The magnitude of attrition of HEWs was found to be 21.1% during the fifteen years of HEP implementation. Of the total 704 who left their job as an HEW, 530 (73%) left the health system altogether. Number of biological children [AOR = 0.61, 95% CI; 0.42–0.89], having an additional education [AOR = 8.34, 95% CI; 3.67–18.98], obtaining official recognition [AOR = 0.29, 95% CI; 0.10–0.83], administrative reprimand [AOR = 1.66, 95% CI; 1.07 -2. 56), distance between district health office and health post [AOR = 1.75, 95%CI; 1.18–2.59) and COC status [AOR = 2.06, 95%CI 1.39–3.06) were independent predictors of leaving the health sector. High regional variation in attrition was observed, ranging from 38.5% in Addis Ababa to just 6.1% in the Harari region. The trend of attrition has steadily increased over time, with a high of 1,999 attritions per 10,000 HEW in 2018. Psychosocial factors, administrative issues, career advancement incentives, and workplace-related problems were the themes that emerged from the qualitative study as reasons for attrition of HEWs. Conclusion: Even though the magnitude of attrition was relatively low, there was high regional variation and incremental trends. Moreover, the out-of-health sector attrition is also high. Critically examining the HEP policy environment to increase the number of HEWs deployed per health post to reduce workload and improving HEW incentives, including career development, may assist in increasing HEW job satisfaction, which in turn could help to reduce attrition, including leaving the health sector.

Smear-positive TB patients greatly contribute to community-level transmission of this disease. Locating hotspots would make it easier to prioritize and target control interventions. This study aimed to assess the spatial distribution of smear-positive index TB cases and their secondary cases and the predictors of clustering of smear-positive TB cases. This study was conducted in the Silti Zone of Central Ethiopia from 2020 to 2022. The data of smear-positive index TB patients were collected from the unit TB registries of healthcare facilities. Contacts of all index TB patients were screened in the community and tested to identify secondary TB patients. We performed spatial analysis, including Moran’s I statistic, the Getis-Ord Gi* statistic and geographically weighted regression (GWR), to assess the global distribution, local clustering and predictors of clustering of smear-positive TB patients, respectively. Additionally, we used inverse distance weighting (IDW) interpolation to predict the distribution of smear-positive TB cases and develop a continuous raster map for places with no data. Spatial autocorrelation analysis revealed that the distribution of smear-positive TB patients exhibited significant clustering (Moran’s I = 0.70029; p value < 0.000). The Getis-Ord Gi* output indicated the presence of statistically significant hotspots as well as cold spots in the study area. Significant hotspots were found in 11 Kebeles of the Silti, Dalocha and Misrak Silti districts. Significant coldspots were also found in five kebeles of the Silti and Misrak districts. GWR analysis revealed that no education, primary education, family size and thatched roof houses were significant predictors of the spatial clustering of smear-positive TB cases. We also found that the majority of the secondary TB cases were found in hotspots identified through spatial analysis. The study revealed a heterogenous distribution of smear positive TB in the study area and it could act as a model that can be replicated in other regions. The identified hotspots of TB could be targeted through location-based interventions such as systematic active screening in the form of outreach programs to improve the performance of TB prevention and control, including reducing the transmission of TB. Educational status, family size and housing type were some of the factors that significantly influenced the spatial distribution of smear-positive TB in the study area. The distribution of secondary TB cases found through household contact screening coincided with the identified hotspots, indicating greater transmission of the disease in these places.

For presumptive Tuberculosis (TB) case referral to be effective, most of the referred cases need to present themselves to health facilities for assessment and testing. Otherwise, cases of TB could be missed, and these cases are at an increased risk of delayed diagnosis, complications and death. Further, their care incurs significantly higher costs. This study assessed referral compliance as well as factors attributable to compliance/non-compliance to referral of presumptive TB cases in Silti district, Southern Ethiopia. We applied a mixed design involving both quantitative and qualitative methods. A randomly selected sample of 384 presumptive TB cases referred between January, 2014 and July 2021 were included in this study from the records of 12 health posts. Purposefully selected presumptive TB cases and Health Extension Workers were also interviewed to get in-depth information on the reasons for compliance and non-compliance to referral. STATA version 14 was employed to model the data using logistic regression. Qualitative data were analyzed using thematic content analysis. Of the 384 referred presumptive TB cases, close to 49% did not present themselves to the referral facilities. About 66% (n = 249) of the referred cases were women, and 62% (n = 119) of those who complied to referral were women. In multivariate analysis, cough [AOR = 3.4, 95%CI: 1.54–7.32], and chest pain [AOR = 2.7, 95%CI: 1.45–5.05] were independent predictors of compliance to referral. Nearly 5.5% (n = 21) of TB cases of all types were identified. The qualitative data analysis revealed that severe disease symptoms, HEW’s recommendations, and social issues as reasons improving compliance while personal and social factors, financial problems, lack of awareness about TB and transportation were reasons impeding compliance to referral. Our study showed a high level of non-compliance to referral among referred presumptive TB cases. We also found that more women were referred and also complied with the referral. Strengthening community awareness about the disease symptoms and the existence of free treatment, addressing misconceptions about TB, supporting the elderly and disabled, and checkup house visits after referral could improve compliance to referral.

Tuberculosis (TB) transmission within households remains a significant challenge in Ethiopia. Various factors including household, index TB case and contact characteristics may have a role in its transmission. Predicting which contacts of TB patients are at greatest risk of developing TB could help focus screening and intervention efforts on early detection and treatment initiation. In this study, we employed various machine learning models to predict TB risk among household contacts. We used data from a cross-sectional study of household contacts of index TB patients in Central Ethiopia. Data were collected through household visits conducted by trained health workers as part of a community-based TB contact investigation program in Central Ethiopia. A set of individual, household, and index case variables were used to train and evaluate multiple supervised machine learning models. Models included Random Forest (RF), Logistic Regression, Artificial Neural Networks (ANN) and XGBoost classifiers. Model performance was assessed using recall, precision, accuracy, ROC-AUC, and F1-score with recall (sensitivity) being prioritized because of the existence of class imbalance. Among 1,277 household contacts screened, 23 (1.8%) were diagnosed with TB. Random Forest and Balanced Random Forest classifiers outperformed logistic regression, achieving recall scores of 0.85. Logistic regression, a commonly used classifier, had an accuracy of 0.92 but a modest recall of 0.43 and an AUC of 0.76. Important TB predictors among household contacts identified using the best performing Balanced Random Forest model included being a presumptive TB case at baseline, providing sputum, having a cough and a productive cough, fatigue and loss of appetite. Machine learning models can identify household contacts at high risk for TB using routinely collected contact investigation data. Integration of predictive modeling into contact investigation could support more targeted and efficient TB screening efforts.