Comparison of logistic regression with regularized machine learning methods for the prediction of tuberculosis disease in people living with HIV: cross-sectional hospital-based study in Kisumu County, Kenya.

Background: Tuberculosis (TB) is a major public health concern, particularly among people living with the Human immunodeficiency Virus (PLWH). Accurate prediction of TB disease in this population is crucial for early diagnosis and effective treatment. Logistic regression and regularized machine learning methods have been used to predict TB, but their comparative performance in HIV patients remains unclear. The study aims to compare the predictive performance of logistic regression with that of regularized machine learning methods for TB disease in HIV patients. Methods: Retrospective analysis of data from HIV patients diagnosed with TB in three hospitals in Kisumu County (JOOTRH, Kisumu sub-county hospital, Lumumba health center) between [dates]. Logistic regression, Lasso, Ridge, Elastic net regression were used to develop predictive models for TB disease. Model performance was evaluated using accuracy, and area under the receiver operating characteristic curve (AUC-ROC). Results: Of the 927 PLWH included in the study, 107 (12.6%) were diagnosed with TB. Being in WHO disease stage III/IV (aOR: 7.13; 95%CI: 3.86-13.33) and having a cough in the last 4 weeks (aOR: 2.34;95%CI: 1.43-3.89) were significant associated with the TB. Logistic regression achieved accuracy of 0.868, and AUC-ROC of 0.744. Elastic net regression also showed good predictive performance with accuracy, and AUC-ROC values of 0.874 and 0.762, respectively. Conclusions: Our results suggest that logistic regression, Lasso, Ridge regression, and Elastic net can all be effective methods for predicting TB disease in HIV patients. These findings may have important implications for the development of accurate and reliable models for TB prediction in HIV patients.

Implementation of evidence-based multiple focus integrated intensified TB screening to end TB (EXIT-TB) package in East Africa: a qualitative study.

INTRODUCTION: Tuberculosis (TB) remains a major cause of morbidity and mortality, especially in sub-Saharan Africa. We qualitatively evaluated the implementation of an Evidence-Based Multiple Focus Integrated Intensified TB Screening package (EXIT-TB) in the East African region, aimed at increasing TB case detection and number of patients receiving care. OBJECTIVE: We present the accounts of participants from Tanzania, Kenya, Uganda, and Ethiopia regarding the implementation of EXIT-TB, and suggestions for scaling up. METHODS: A qualitative descriptive design was used to gather insights from purposefully selected healthcare workers, community health workers, and other stakeholders. A total of 27, 13, 14, and 19 in-depth interviews were conducted in Tanzania, Kenya, Uganda, and Ethiopia respectively. Data were transcribed and translated simultaneously and then thematically analysed. RESULTS: The EXIT-TB project was described to contribute to increased TB case detection, improved detection of Multidrug-resistant TB patients, reduced delays and waiting time for diagnosis, raised the index of TB suspicion, and improved decision-making among HCWs. The attributes of TB case detection were: (i) free X-ray screening services; (ii) integrating TB case-finding activities in other clinics such as Reproductive and Child Health clinics (RCH), and diabetic clinics; (iii), engagement of CHWs, policymakers, and ministry level program managers; (iv) enhanced community awareness and linkage of clients; (v) cooperation between HCWs and CHWs, (vi) improved screening infrastructure, (vii) the adoption of the new simplified screening criteria and (viii) training of implementers. The supply-side challenges encountered ranged from disorganized care, limited space, the COVID-19 pandemic, inadequate human resources, inadequate knowledge and expertise, stock out of supplies, delayed maintenance of equipment, to absence of X-ray and GeneXpert machines in some facilities. The demand side challenges ranged from delayed care seeking, inadequate awareness, negative beliefs, fears towards screening, to financial challenges. Suggestions for scaling up ranged from improving service delivery, access to diagnostic equipment and supplies, and infrastructure, to addressing client fears and stigma. CONCLUSION: The EXIT-TB package appears to have contributed towards increasing TB case detection and reducing delays in TB treatment in the study settings. Addressing the challenges identified is needed to maximize the impact of the EXIT-TB intervention.

Diagnostic performance of oral swabs for non-sputum based TB diagnosis in a TB/HIV endemic setting.

OBJECTIVE: We evaluated diagnostic performance of oral swab analysis (OSA) for tuberculosis (TB) in a high HIV/TB burden setting in Kenya. METHODS: In this cross-sectional study, buccal swabs and sputum were collected from 100 participants with suspected TB in outpatient clinics in Kenya at enrollment and subsequent morning visits. Buccal swabs underwent IS6110-targeted qPCR analysis. Sputum was evaluated by Xpert MTB/RIF (Xpert) and culture. Diagnostic performance of OSA for TB diagnosis was evaluated relative to a combined reference of sputum Xpert and culture. RESULTS: Among 100 participants, 54% were living with HIV (PLHIV). Twenty percent (20/100) of participants had confirmed TB (19/20 [95%] culture-positive, 17/20 [85%] Xpert-positive). Overall buccal swab sensitivity was 65.0% (95% CI 40.8-84.6%) vs. sputum Xpert/culture and 76.5% (95% CI 50.1-93.2%) vs. sputum Xpert alone. Specificity was 81.3% (95% CI 71.0-89.1%) and 81.9% (95% CI 72.0-89.5%) compared to sputum Xpert/culture and Xpert alone, respectively. Sensitivity among PLHIV (n = 54) with suspected TB was 83.3% (95% CI 35.9-99.6%) vs. sputum Xpert/culture and 100% (95% CI 47.8-100.0%) vs. sputum Xpert alone. Among participants with TB, mean OSA threshold quantitation cycle (Cq) value was lower (stronger signal) at subsequent morning compared to enrolment visit (33.4 SD ± 3.7 vs. 35.2 SD ± 2.9, p = 0.009). CONCLUSIONS: In this pilot study, results confirm M. tuberculosis DNA is detectable in oral swabs including among PLHIV with fair diagnostic performance. Further work is needed to optimize OSA and evaluate its utility in diverse settings.

Integrating interferon-gamma release assay testing into provision of tuberculosis preventive therapy is feasible in a tuberculosis high burden resource-limited setting: A mixed methods study.

The World Health Organization recommends the scale-up of tuberculosis preventive therapy (TPT) for persons at risk of developing active tuberculosis (TB) as a key component to end the global TB epidemic. We sought to determine the feasibility of integrating testing for latent TB infection (LTBI) using interferon-gamma release assays (IGRAs) into the provision of TPT in a resource-limited high TB burden setting. We conducted a parallel convergent mixed methods study at four tertiary referral hospitals. We abstracted details of patients with bacteriologically confirmed pulmonary tuberculosis (PBC TB). We line-listed household contacts (HHCs) of these patients and carried out home visits where we collected demographic data from HHCs, and tested them for both HIV and LTBI. We performed multi-level Poisson regression with robust standard errors to determine the associations between the presence of LTBI and characteristics of HHCs. Qualitative data was collected from health workers and analyzed using inductive thematic analysis. From February to December 2020 we identified 355 HHCs of 86 index TB patients. Among these HHCs, uptake for the IGRA test was 352/355 (99%) while acceptability was 337/352 (95.7%). Of the 352 HHCs that were tested with IGRA, the median age was 18 years (IQR 10-32), 191 (54%) were female and 11 (3%) were HIV positive. A total of 115/352 (32.7%) had a positive IGRA result. Among HHCs who tested negative on IGRA at the initial visit, 146 were retested after 9 months and 5 (3.4%) of these tested positive for LTBI. At multivariable analysis, being aged ≥ 45 years [PR 2.28 (95% CI 1.02, 5.08)], being employed as a casual labourer [PR 1.38 (95% CI 1.19, 1.61)], spending time with the index TB patient every day [PR 2.14 (95% CI 1.51, 3.04)], being a parent/sibling to the index TB patients [PR 1.39 (95% CI 1.21, 1.60)] and sharing the same room with the index TB patients [PR 1.98 (95% CI 1.52, 2.58)] were associated with LTBI. Implementation challenges included high levels of TB stigma and difficulties in following strict protocols for blood sample storage and transportation. Integrating home-based IGRA testing for LTBI into provision of TB preventive therapy in routine care settings was feasible and resulted in high uptake and acceptability of IGRA tests.

Causes of death in HIV-infected and HIV-uninfected children aged under-five years in western Kenya.

OBJECTIVES: Describe the causes of death among infants and children less than 5 years stratified by HIV status. DESIGN: Cross-sectional analysis of causes of death ascertained through minimally invasive tissue sampling (MITS) in the Kenya Child Health and Mortality Prevention Surveillance site. METHODS: We included decedents aged 28 days to less than 5 years, whose death was reported within 36 h, underwent MITS, and had HIV test results and causes of death determined. MITS specimens were tested using Taqman Array Cards, culture, cytology, histopathology and immunohistochemistry and HIV PCR. A panel evaluated epidemiologic, clinical, verbal autopsy and laboratory data to assign causes of death using ICD-10 guidelines. Causes of death and etiological agents were stratified by HIV status. RESULTS: Of 176 included decedents, 14% (n = 25) were HIV-infected, median viral load was 112 205 copies/ml [interquartile range (IQR) = 9349-2 670 143). HIV-disease (96%; n = 24) and malnutrition (23%; n = 34) were the leading underlying causes of death in HIV-infected and HIV-uninfected decedents, respectively. Malnutrition was more frequent in the causal chain of HIV-infected (56%; n = 14) than HIV-uninfected decedents (31%; n = 49) (P value = 0.03). Viral pneumonia was twice as common in HIV-infected (50%; n = 9) than HIV-uninfected decedents (22%; n = 7) (P value = 0.04). CONCLUSION: Nearly all HIV-infected decedents' underlying cause of death was HIV disease, which was associated with malnutrition. Our findings underscore the need for strengthening early identification and management of HIV-infected children. Prevention, early diagnosis and treatment of malnutrition could be instrumental in improving the survival of HIV-infected and HIV-uninfected children.

Gaps related to screening and diagnosis of tuberculosis in care cascade in selected health facilities in East Africa countries: A retrospective study.

INTRODUCTION: East Africa countries (Tanzania, Kenya, and Uganda) are among tuberculosis high burdened countries globally. As we race to accelerate progress towards a world free of tuberculosis by 2035, gaps related to screening and diagnosis in the cascade care need to be addressed. METHODS: We conducted a three-year (2015-2017) retrospective study using routine program data in 21 health facilities from East Africa. Data abstraction were done at tuberculosis clinics, outpatient departments (OPD), human immunodeficiency virus (HIV) and diabetic clinics, and then complemented with structured interviews with healthcare providers to identify possible gaps related to integration, screening, and diagnosis of tuberculosis. Data were analyzed using STATA™ Version 14.1. RESULTS: We extracted information from 49,454 presumptive TB patients who were registered in the 21 facilities between January 2015 and December 2017. A total of 9,565 tuberculosis cases were notified; 46.5% (4,450) were bacteriologically confirmed and 31.5% (3,013) were HIV-infected. Prevalence of tuberculosis among presumptive pulmonary tuberculosis cases was 17.4%. The outcomes observed were as follows: 79.8% (7,646) cured or completed treatment, 6.6% (634) died, 13.3% (1,270) lost to follow-up or undocumented and 0.4% (34) treatment failure. In all countries, tuberculosis screening was largely integrated at OPD and HIV clinics. High patient load, weak laboratory specimen referral system, shortage of trained personnel, and frequent interruption of laboratory supplies were the major cited challenges in screening and diagnosis of tuberculosis. CONCLUSION: Screening and diagnostic activities were frequently affected by scarcity of human and financial resources. Tuberculosis screening was mainly integrated at OPD and HIV clinics, with less emphasis on the other health facility clinics. Closing gaps related to TB case finding and diagnosis in developing countries requires sustainable investment for both human and financial resources and strengthen the integration of TB activities within the health system.