We collaborate with all OUCRU research groups as well as with the clinical trials unit in all stages of clinical and epidemiological studies. We advise on study conception, study design and data collection; we advise on analysis of the data or perform the analysis ourselves; and we give critical appraisal of how to interpret and communicate results. We also perform methodological research on statistical methods.

Besides, we enhance the statistical skills of our fellow researchers via teaching. We teach the courses Introduction to Medical Statistics and Principles of Data Visualization to all OUCRU staff, and together with the Mahidol Oxford Tropical Medicine Research Unit (MORU) we teach a yearly course on more advanced statistical methods. The topic of this course differs by year and has been causal inference, diagnostic and prognostic models, survival analysis, longitudinal data analysis, missing data and Bayesian statistics.

Some of our research projects

• Construction and validation of models for diagnosis and prognosis. We combine patient characteristics with clinical and laboratory examinations to improve the diagnosis of the infectious agent as well as the prognosis of disease outcomes. Our focus is on statistical regression models rather than machine learning algorithms. Such models have been developed for tuberculous meningitis (TBM), dengue, malaria and COVID-19 and some have been implemented in web calculators (Decision helper for TBM Diagnosis; Risk calculator for the absolute risk of death within 9 months in HIV-uninfected and HIV-infected patients with tuberculous meningitis and Dynamic prediction for death in HIV-infected and HIV-uninfected patients with tuberculous meningitis) These can be used to adapt clinical management to the individual patient. In a more theoretical study, we evaluated how best to deal with missing data in prediction models.
• Causal inference. The gold standard to infer causality is the randomized controlled trial (RCT). We design and perform the statistical analyses of RCTs that are performed at OUCRU and affiliated institutes, most importantly the hospital for tropical diseases. Causal inference based on observational data is more complex and requires the use of sophisticated statistical models. One example is the quantification of the effect of M. tuberculosis lineage on clinical phenotype.
• Trajectory of markers and their role in disease progression. Often the primary interest is in a clinical outcome like severe disease, death or cure. Markers of disease progression provide important intermediate information that reflects progression to the outcome. An example is the role of changes in RNA and platelet count in the disease course of dengue patients. We use statistical models for a proper description of their trajectories in combination with the clinical outcome.
• SARS-CoV-2 pandemic. The Vietnamese government implemented a policy of intensive contact tracing and testing, which resulted in unique and detailed data on community transmissions. These data can be used to estimate important parameters from the early stage of the infection: the latency time (time from infection to becoming infectious) and the incubation time (time from infection to development of symptoms).