COVID-19 Data Quality and Study Stabilisation

Led the recovery of a high-risk clinical study using a large, paper-based dataset with significant data quality issues.

• Problem: Missing and implausible data were compromising the reliability of the study and delaying progress
• Approach: Designed and implemented data quality control checks and clinician-facing reports to identify, track, and resolve data issues
• Outcome: Enabled successful publication in the New England Journal of Medicine and improved the reliability of the underlying clinical dataset

Behavioural Analysis of Gambling Decision-Making

Developed novel approaches to analyse real-world behavioural data from gambling environments, addressing challenges not captured in traditional experimental designs.

• Problem: Existing methods were not suited to complex, real-world behavioural data with high variability and noise
• Approach: Designed and applied advanced statistical models and developed tools to capture and analyse behavioural patterns in real-world settings
• Outcome: Generated new insights into decision-making behaviour and contributed to peer-reviewed publications in the field

Reconstruction of Individual Patient Data from Published Survival Curves

Developed a flexible pipeline to extract and reconstruct individual patient data from published Kaplan–Meier curves across varied data conditions.

• Problem: Published survival data are often only available as graphical outputs of varying quality, limiting reuse and robust comparative analysis
• Approach: Evaluated multiple extraction tools and custom methods, and designed a standardised pipeline with built-in quality control checks to reconstruct individual patient data
• Outcome: Enabled faster, more consistent, and reliable downstream analyses across projects, improving efficiency and analytical robustness