Variation in tuberculosis treatment outcomes and treatment supervision practices in Uganda

A machine learning approach to explore individual risk factors for tuberculosis treatment non-adherence in Mukono district

Despite the availability and implementation of well-known efficacious interventions for tuberculosis treatment by the Ministry of Health, Uganda (MoH), treatment non-adherence persists. Moreover, identifying a specific tuberculosis patient at risk of treatment non-adherence is still a challenge. Thus, this retrospective study, based on a record review of 838 tuberculosis patients enrolled in six health facilities, presents, and discusses a machine learning approach to explore the individual risk factors predictive of tuberculosis treatment non-adherence in the Mukono district, Uganda. Five classification machine learning algorithms, logistic regression (LR), artificial neural networks (ANN), support vector machines (SVM), random forest (RF), and AdaBoost were trained, and evaluated by computing their accuracy, F1 score, precision, recall, and the area under the receiver operating curve (AUC) through the aid of a confusion matrix. Of the five developed and evaluated algorithms, SVM (91.28%) had the highest accuracy (AdaBoost, 91.05% performed better than SVM when AUC is considered as evaluation parameter). Looking at all five evaluation parameters globally, AdaBoost is quite on par with SVM. Individual risk factors predictive of non-adherence included tuberculosis type, GeneXpert results, sub-country, antiretroviral status, contacts below 5 years, health facility ownership, sputum test results at 2 months, treatment supporter, cotrimoxazole preventive therapy (CPT) dapsone status, risk group, patient age, gender, middle and upper arm circumference, referral, positive sputum test at 5 and 6 months. Therefore, machine learning techniques, specifically classification types, can identify patient factors predictive of treatment non-adherence and accurately differentiate between adherent and non-adherent patients. Thus, tuberculosis program management should consider adopting the classification machine learning techniques evaluated in this study as a screening tool for identifying and targeting suited interventions to these patients.

Trends of notification rates and treatment outcomes of tuberculosis cases with and without HIV co-infection in eight rural districts of Uganda (2015 - 2019)

Background

The End TB Strategy aims to reduce new tuberculosis (TB) cases by 90% and TB-related deaths by 95% between 2015 – 2035. We determined the trend of case notification rates (CNRs) and treatment outcomes of TB cases with and without HIV co-infection in rural Uganda to provide an interim evaluation of progress towards this global target in rural settings.

Methods

We extracted retrospective programmatic data on notified TB cases and treatment outcomes from 2015 – 2019 for eight districts in rural Uganda from the District Health Information System 2. We estimated CNRs as the number of TB cases per 100,000 population. Treatment success rate (TSR) was calculated as the sum of TB cure and treatment completion for each year. Trends were estimated using the Mann–Kendall test.

Results

A total of 11,804 TB cases, of which 5,811 (49.2%) were HIV co-infected, were notified. The overall TB CNR increased by 3.7-fold from 37.7 to 141.3 cases per 100,000 population in 2015 and 2019 respectively. The increment was observed among people with HIV (from 204.7 to 730.2 per 100,000, p = 0.028) and HIV-uninfected individuals (from 19.9 to 78.7 per 100,000, p = 0.028).

There was a decline in the TSR among HIV-negative TB cases from 82.1% in 2015 to 63.9% in 2019 (p = 0.086). Conversely, there was an increase in the TSR among HIV co-infected TB cases (from 69.9% to 81.9%, p = 0.807).

Conclusion

The CNR increased among people with and without HIV while the TSR reduced among HIV-negative TB cases. There is need to refocus programs to address barriers to treatment success among HIV-negative TB cases.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-022-13111-1.

Factors associated with poor treatment outcomes among tuberculosis patients in Kyangwali Refugee Settlement, Uganda, 2016-2017

Communicable diseases, alone or in combination with malnutrition, account for most deaths in complex emergencies including refugee settings. Tuberculosis and HIV/AIDS are increasingly becoming an important cause of morbidity and mortality in refugee settings. We described the treatment outcomes of TB patients and explored factors associated with treatment outcomes among TB patients attending two facilities in Kyangwali Refugee Settlement in Kikuube District, 2016-2017. We abstracted data on laboratory-confirmed patient data from TB registers from 2016 to 2017, in Kikuube Health Centre IV and Rwenyawawa Health Centre II, both located in Kyangwali Refugee Settlement. We abstracted data on socio-demographic variables including age and sex. Other variables were height, weight, final treatment outcomes, demographics, HIV status, TB treatment category, and history of TB. Treatment outcomes were categorized into favorable (including patients who were cured or those who completed treatment) and unfavorable (those in whom treatment failed, those who died, those lost to follow-up, or those not evaluated). We used logistic regression to identify factors associated with unfavorable treatment outcomes. We identified a total of 254 TB patients with a median age of 36 (IQR 26-48) years; 69% (175) were male and 54% (137) were refugees. The median weight was 50.4 kg (range 4-198). Overall, 139 (55%) had favorable outcomes while 115 (45%) had unfavorable outcomes. Refugees formed 53% (71) of those with favorable outcomes and 47% (63) of those with unfavorable outcomes 63(47%). We found that increasing age was statistically associated with unfavorable outcomes, while diagnosis with MDR-TB was associated with decreased odds for unfavorable treatment outcomes. The treatment success rate was lower compared to 85% recommended by WHO. However, the rates are similar to that reported by other studies in Uganda. Innovative approaches to improve treatment success rates with particular focus on persons aged 41-80 years should be devised.

Digital adherence technology for tuberculosis treatment supervision: A stepped-wedge cluster-randomized trial in Uganda

BACKGROUND

Adherence to and completion of tuberculosis (TB) treatment remain problematic in many high-burden countries. 99DOTS is a low-cost digital adherence technology that could increase TB treatment completion.

METHODS AND FINDINGS

We conducted a pragmatic stepped-wedge cluster-randomized trial including all adults treated for drug-susceptible pulmonary TB at 18 health facilities across Uganda over 8 months (1 December 2018-31 July 2019). Facilities were randomized to switch from routine (control period) to 99DOTS-based (intervention period) TB treatment supervision in consecutive months. Patients were allocated to the control or intervention period based on which facility they attended and their treatment start date. Health facility staff and patients were not blinded to the intervention. The primary outcome was TB treatment completion. Due to the pragmatic nature of the trial, the primary analysis was done according to intention-to-treat (ITT) and per protocol (PP) principles. This trial is registered with the Pan African Clinical Trials Registry (PACTR201808609844917). Of 1,913 eligible patients at the 18 health facilities (1,022 and 891 during the control and intervention periods, respectively), 38.0% were women, mean (SD) age was 39.4 (14.4) years, 46.8% were HIV-infected, and most (91.4%) had newly diagnosed TB. In total, 463 (52.0%) patients were enrolled on 99DOTS during the intervention period. In the ITT analysis, the odds of treatment success were similar in the intervention and control periods (adjusted odds ratio [aOR] 1.04, 95% CI 0.68-1.58, p = 0.87). The odds of treatment success did not increase in the intervention period for either men (aOR 1.24, 95% CI 0.73-2.10) or women (aOR 0.67, 95% CI 0.35-1.29), or for either patients with HIV infection (aOR 1.51, 95% CI 0.81-2.85) or without HIV infection (aOR 0.78, 95% CI 0.46-1.32). In the PP analysis, the 99DOTS-based intervention increased the odds of treatment success (aOR 2.89, 95% CI 1.57-5.33, p = 0.001). The odds of completing the intensive phase of treatment and the odds of not being lost to follow-up were similarly improved in PP but not ITT analyses. Study limitations include the likelihood of selection bias in the PP analysis, inability to verify medication dosing in either arm, and incomplete implementation of some components of the intervention.

CONCLUSIONS

99DOTS-based treatment supervision did not improve treatment outcomes in the overall study population. However, similar treatment outcomes were achieved during the control and intervention periods, and those patients enrolled on 99DOTS achieved high treatment completion. 99DOTS-based treatment supervision could be a viable alternative to directly observed therapy for a substantial proportion of patients with TB.

TRIAL REGISTRATION

Pan-African Clinical Trials Registry (PACTR201808609844917).