Applied health statistics
The core of the practice: choosing and applying statistical methods to answer real clinical, public health and health services questions, with the emphasis on interpretation that a non-statistician can act on.
Research interests and expertise
Methodological and applied research across health statistics, measurement and quantitative methods — the areas that shape how JH Stat approaches consultancy work.
Overview
Where methods meet health research
The research that informs JH Stat's consultancy sits between statistical methodology and applied health investigation: how to measure things well, how to model complex data structures honestly, and how to draw defensible conclusions from data that were rarely collected under ideal conditions.
Health data are messy in specific and predictable ways. Observations cluster within clinicians, practices and hospitals. People are followed over time and drop out non-randomly. Outcomes are censored, self-reported, or measured with instruments whose properties in the study population are unknown. Comparison groups are not randomised. The methods below are the ones that take those realities seriously rather than assuming them away.
Areas of expertise
Nine areas of concentration
Clinical and observational research
Design and analysis of trials, cohorts, case-control studies and registry analyses, including the practical problems of protocol deviation, loss to follow-up and non-random missingness.
Health services research
Evaluating how services are organised, delivered and used — including routinely collected data, service redesign evaluation, variation between providers and utilisation modelling.
Longitudinal and multilevel data
Modelling repeated measures and hierarchical structures: growth curves, mixed-effects models, GEE and trajectory modelling, with correct treatment of within-cluster correlation and attrition.
Psychometrics and outcome measurement
Development and validation of scales and patient-reported outcome measures: factor structure, reliability, validity evidence, measurement invariance and item response theory.
Survival and time-to-event analysis
Time-to-event modelling including Cox and flexible parametric models, competing risks, recurrent events, time-varying covariates and the assessment of proportional hazards.
Causal inference
Drawing defensible causal conclusions from non-randomised data using propensity score methods, instrumental variables, interrupted time series, difference-in-differences and formal confounding frameworks.
AI and machine learning in health research
Prediction modelling and applied machine learning where prediction is the genuine goal — with proper validation, calibration, fairness assessment and honest comparison against simpler models.
Research design and reproducibility
Study design, pre-specification, analysis planning and reproducible workflows — treating reproducibility as a working method rather than a compliance exercise.
Research interests
Current research interests
The themes below describe where the research and methodological interests of JH Stat are concentrated.
-
Quantitative research methods and health survey data analysis
Methods for designing and analysing quantitative health studies, with particular attention to the structure and complexity of survey data.
-
Multilevel modelling and longitudinal analysis of health outcomes
Modelling clustered and hierarchical data, and analysing health outcomes measured repeatedly over time.
-
Statistical and machine-learning methods in healthcare and social science research
Applying statistical and machine-learning methods to research questions across healthcare and the social sciences.
-
Geospatial data modelling and inference with geo-masked data
Modelling geospatial data, statistical inference where geolocations have been masked, and the preservation of respondent confidentiality.
-
Statistical consultation and collaboration in interdisciplinary research
Statistical consultation and sustained collaboration within interdisciplinary research teams.
Confidentiality. Much consultancy work is covered by non-disclosure or data-sharing agreements and cannot be described publicly, which is why individual projects are not listed here. Details of relevant experience can usually be shared confidentially on request. Get in touch if you would like to discuss comparable work, or see the publications for published research.
Collaboration
Open to research collaboration
JH Stat welcomes approaches from research teams looking for a statistical collaborator rather than a supplier — particularly where the methodological questions are interesting in their own right.
Particularly interested in
- Studies with complex data structures — clustered, longitudinal or multi-source
- Measurement development and validation in new populations or settings
- Evaluation of service change using routinely collected data
- Applications of causal methods to non-randomised health data
- Careful, well-validated prediction modelling in clinical settings
- Methodological work on reproducibility and reporting quality
Discuss a research question
If your study touches any of the areas above — or if you are not sure which methods apply — a short conversation is usually the fastest way to find out.