Search

Background. Machine learning (ML) refers to computational algorithms designed to learn from patterns in data to provide insights or predictions related to that data. Objective. Multiple studies report the development of predictive models for triage or identification of critically ill children. In this study, we validate machine learning models developed in South Africa for the identification of critically ill children presenting to a tertiary hospital. Results. The validation sample comprised 267 patients. The event rate for the study outcome was 0.12. All models demonstrated good discrimination but weak calibration. Artificial neural network 1 (ANN1) had the highest area under the receiver operating characteristic curve (AUROC) with a value of 0.84. ANN2 had the highest area under the precision-recall curve (AUPRC) with a value of 0.65. Decision curve analysis demonstrated that all models were superior to standard strategies of treating all patients or treating no patients at a proposed threshold probability of 10%. Confidence intervals for model performance overlapped considerably. Post hoc model explanations demonstrated that models were logically coherent with clinical knowledge. Conclusions. Internal validation of the predictive models correlated with model performance in the development study. The models were able to discriminate between critically ill children and non-critically ill children; however, the superiority of one model over the others could not be demonstrated in this study. Therefore, models such as these still require further refinement and external validation before implementation in clinical practice. Indeed, successful implementation of machine learning in practice within the South African setting will require the development of regulatory and infrastructural frameworks in conjunction with the adoption of alternative approaches to electronic data capture, such as the use of mobile devices. [ABSTRACT FROM AUTHOR]

Abstract Background Nucleic acid amplification tests (NAATs) have offered hope for rapid diagnosis of tuberculosis (TB). However, their efficiency with smear-negative samples has not been widely studied in low income settings. Here, we evaluated in-house PCR assay for diagnosis of smear-negative TB using Lowenstein-Jensen (LJ) culture as the baseline test. Two hundred and five pulmonary TB (PTB) suspects with smear-negative sputum samples, admitted on a short stay emergency ward at Mulago Hospital in Kampala, Uganda, were enrolled. Two smear-negative sputum samples were obtained from each PTB suspect and processed simultaneously for identification of MTBC using in-house PCR and LJ culture. Results Seventy two PTB suspects (35%, 72/205) were LJ culture positive while 128 (62.4%, 128/205) were PCR-positive. The sensitivity and specificity of in-house PCR for diagnosis of smear-negative PTB were 75% (95% CI 62.6-85.0) and 35.9% (95% CI 27.2-45.3), respectively. The positive and negative predictive values were 39% (95% CI 30.4-48.2) and 72.4% (95% CI 59.1-83.3), respectively, while the positive and negative likelihood ratios were 1.17 (95% CI 0.96-1.42) and 0.70 (95% CI 0.43-1.14), respectively. One hundred and seventeen LJ culture-negative suspects (75 PCR-positive and 42 PCR-negative) were enrolled for follow-up at 2 months. Of the PCR-positive suspects, 45 (60%, 45/75) were still alive, of whom 29 (64.4%, 29/45) returned for the follow-up visit; 15 (20%, 15/75) suspects died while another 15 (20%, 15/75) were lost to follow-up. Of the 42 PCR-negative suspects, 22 (52.4%, 22/42) were still alive, of whom 16 (72.7%, 16/22) returned for follow-up; 11 (26.2%, 11/42) died while nine (21.4%, 9/42) were lost to follow-up. Overall, more PCR-positive suspects were diagnosed with PTB during follow-up visits but the difference was not statistically significant (27.6%, 8/29 vs. 25%, 4/16, p = 0.9239). Furthermore, mortality was higher for the PCR-negative suspects but the difference was also not statistically significant (26.2% vs. 20% p = 0.7094). Conclusion In-house PCR correlates poorly with LJ culture for diagnosis of smear-negative PTB. Therefore, in-house PCR may not be adopted as an alternative to LJ culture.

Abstract Background Ugandan national guidelines recommend initiation of combination antiretroviral therapy (cART) at CD4+ T cell (CD4) count below 350 cell/μl, but the implementation of this is limited due to availability of medication. However, cART initiation at higher CD4 count increases survival, albeit at higher lifetime treatment cost. This analysis evaluates the cost-effectiveness of initiating cART at a CD4 count between 250–350 cell/μl (early) versus Methods Life expectancy of cART-treated patients, conditional on baseline CD4 count, was modeled based on published literature. First-line cART costs $192 annually, with an additional $113 for patient monitoring. Delaying initiation of cART until the CD4 count falls below 250 cells/μl would incur the cost of the bi-annual CD4 count tests and routine maintenance care at $85 annually. We compared lifetime treatment costs and disability adjusted life-expectancy between early vs. delayed cART for ten baseline CD4 count ranges from 250-350 cell/μl. All costs and benefits were discounted at 3% annually. Results Treatment delay varied from 6–18 months. Early cART initiation increased life expectancy from 1.5-3.5 years and averted 1.33–3.10 disability adjusted life years (DALY’s) per patient. Lifetime treatment costs were $4,300–$5,248 for early initiation and $3,940–$4,435 for delayed initiation. The cost/DALY averted of the early versus delayed start ranged from $260–$270. Conclusions In HIV-positive patients presenting with CD4 count between 250-350 cells/μl, immediate initiation of cART is a highly cost-effective strategy using the recommended one-time per capita GDP threshold of $490 reported for Uganda. This would constitute an efficient use of scarce health care funds.

7-11 November 2010, Tenth International Congress on Drug Therapy in HIV Infection, Glasgow, UK [ABSTRACT FROM AUTHOR]