Causal Inference and Data Science Methods: fixing broken experiments!

My program of research is about developing causal inference methods for large healthcare data, with a particular emphasis on applications of data science tools. See publications page for more details about each of the following:

Methodological Research:

• Use of machine learning / AI approaches in the causal inference context

  • TMLE and variants
  • hdPS
  • Double machine learning
  • Autoencoders for dimensionality reduction

• Practical issues dealing with large health administrative databases

  • Longitudinal studies: dealing with data sparseness in Marginal structural models
  • Incorporating proxy information via high-dimensional propensity score modelling and it’s machine learning versions
  • Addressing unmeasured confounding via supplimentary information
  • Time-dependent treatment and time-dependent confounding
    • Addressing immortal time bias in longitudinal studies
  • Complex causal diagrams and respective analysis strategies
    • Mediation analysis: decomposing the effect of direct and indirect effects
  • Exposure misclassification in observational studies

• Pragmatic trials: Incomplete Medication Adherence in Pragmatic trials

  • Inverse probability of adherence weighting per-protocol effect in sustained treatment strategies
    • Finite sample properties
    • Differential adherence
    • Dealing with data sparseness for post-baseline prognostic factors
  • Extending the principal stratification framework for multi-category of partial adherence
  • Comparing the per-protocol effect estimation methods with instrumental methods in point-treatment scenario

• Survey data analysis

  • Incorporating causal inference and data science approaches within the survey data analysis framework
  • Motivating examples from DHS, NHANES, CCHS.

Real-World Applications

• Impact in healthcare
  • Multiple sclerosis
  • Tuberculosis
  • Arthritis
  • HIV
  • Opioid use disorder
  • Spinal cord
  • Cardiovascular risk
  • Mental health: depression and mood disorder
  • Pediatric intensive care
  • Nephrology and kidney disease
  • human papillomavirus vaccine
  • Child Hemoglobin
  • Antenatal care
  • Multimorbidity
  • Residential instability and hospitalizations among homeless and vulnerably housed individuals
• Impact in interdisciplinary research
  • Forestry
  • Nursing

Building Capacity in Data Scienece and Causal Inference

• Proposing guidelines for propensity score analyses in disease specific areas
• See teaching page for in-class teaching
• See workshops page for webinars, workshops
• Supervision of graduate and undergraduate students / trainees.

Grant support

Some notable funding sources listed below (excluding start-up fundings):

• MS Canada
• Michael Smith Foundation for Health Research,
• Natural Sciences and Engineering Research Council of Canada,
• BC SUPPORT Unit, and other start-up funds.
• Co-investigator and Collaborator in various other grants (NIDA/NIH, BC SUPPORT Unit, CIHR, and SSHRC).