Use this URL to cite or link to this record in EThOS:
Title: The use of propensity score methods to examine comparative effectiveness of treatments provided to children in Kenyan hospitals
Author: Malla, Lucas Owor
ISNI:       0000 0004 7430 7443
Awarding Body: University of Oxford
Current Institution: University of Oxford
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Background: Although randomised controlled trials are the best means of measuring efficacy of interventions and inferring causality challenges of limited generalisability to real life populations, cost and applicability for some types of intervention have prompted efforts to develop alternative approaches using observational data. Such approaches are very rarely used in low income settings. This research utilised a large dataset of well characterised admission events to paediatric wards in Kenya's county referral hospitals to develop approaches to, and execute, comparative effectiveness analyses for treatment of diarrhoea and pneumonia as the top causes of morbidity and mortality in children. Specifically, the two clinical questions addressed were: (i) comparative effectiveness of penicillin monotherapy versus penicillin plus gentamicin in children with (indrawing) pneumonia; (ii) effectiveness of Zinc in treatment of children admitted with diarrhoea. Methods: To answer these questions, a strategy for multiple imputation of missing data was developed. Propensity score based methods were then developed and evaluated to identify balanced groups receiving alternative treatment allocations (for example different antibiotic regimens for childhood pneumonia) using key covariates. In the next step the effectiveness of alternative treatment outcomes were compared. Finally, sensitivity analyses were then conducted to verify missing data assumptions and also to examine the influence of unmeasured confounders. Further, methods were tested on how to estimate propensity scores after multiple imputation. Results: The antibiotic analyses indicated that there was no statistical difference between penicillin monotherapy and penicillin plus gentamicin in the treatment of (indrawing) pneumonia. By extension, treatment using penicillin plus gentamicin would likely not offer any advantage over using amoxicillin, the recently recommended treatment of indrawing pneumonia, as equivalence had been demonstrated between penicillin monotherapy and amoxicillin in previous trials, although non-trivial mortality was observed in this group suggesting other measures need to be investigated to prevent deaths. In the second case, Zinc was demonstrated to be effective in shortening the length of stay for children admitted with diarrhoea including those who were well-nourished (WAZ >-2) and those aged 1-5 months. Conclusion: The analyses in this thesis provided potentially useful information to support development of clinical treatment policies. However, there remains some concern that comparative effectiveness evaluation using routine observational data may not fully mitigate risks of bias due to unobserved confounders. Rigorous approaches to such analyses using propensity scores, with detailed accounts of strategies to reduce bias and appropriate sensitivity analyses may be useful when randomised controlled trials are not feasible. However, it is probably better to pursue pragmatic trials (where randomisation is feasible) where possible to generate high quality, generalisable evidence of treatment effectiveness.
Supervisor: Perera, Rafael ; English, Mike ; McFadden, Emily Sponsor: Nuffield Department of Medicine
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available