Statistical methodology for the analysis of Electronic Health Records

Electronic health records contain millions of patient histories, but the data is riddled with gaps and errors that can skew results if analysed with standard statistical tools. This matters because researchers and doctors increasingly rely on these records to answer basic questions—how common a disease is, who gets it, and how it progresses. The data was never designed for research; it was collected for clinical care. Simple methods that ignore missing or incorrect entries can produce misleading conclusions, potentially leading to wrong screening recommendations or treatment decisions. The researcher is developing new statistical methods that handle missing data, combine information from different sources, and account for uncertainty. They are also building software tools so other researchers can apply these methods easily. If successful, this work will improve the accuracy of any study that uses electronic health records. That includes helping doctors decide when to invite people for screening, guiding treatment choices, and giving a clearer picture of how diseases unfold over time. The impact is not dramatic or visible—it is a quiet improvement to the analytical backbone of modern epidemiology and clinical decision-making.