Ehsan Karim

Background: Analyzing health administrative data is challenging both for causal inference and predictive analytics because data are not collected for research. Considering tuberculosis (TB) as a motivating example, we compared different epidemiological methods to analyze health administrative data for survival outcomes. We primarily focused on (i) evaluating the performance of high-dimensional disease risk score (hdDRS) to deal with residual confounding bias in causal inference, (ii) assessing the benefits of repeated nested case-control and matched-cohort analyses to estimate a stable hazard ratio (HR) with time-dependent exposure in causal inference, (iii) developing LASSO-based survival prediction models with multiple imputed datasets, and (iv) determining the performance of high-dimensional prediction model (hdPM) for enhancing risk prediction by supplementing unobserved clinical predictors.Methods: We developed a retrospective cohort of immigrants in British Columbia, Canada, 1985-2019 by utilizing linked health administrative databases. Notable methods include (i) designing plasmode simulations to compare eight hdDRS methods varied by different approaches to fit the risk score model; (ii) conducting extensive data analysis with Monte Carlo simulations by varying the number of controls with and without repeated sampling; (iii) conducting extensive data analysis and plasmode simulations to explore the performance of three distinct statistical methods: prediction average, performance average, and stacked; (iv) conducting comprehensive plasmode simulations to in scenarios where a strong or weak clinical predictor was unavailable to the analyst. Results: Key methodological findings included: (i) hdDRS effectively reduced residual confounding bias, (ii) repeating NCC and matched-cohort analyses multiple times and pooling the estimates improved the stability of the HR estimate, (iii) LASSO models applied to stacked datasets produced the most robust survival predictions, and (iv) hdPM with LASSO-shrinkage enhanced model accuracy by incorporating high-dimensional predictors. Conclusions: Evidence was generated to overcome some limitations of prior studies analyzing health administrative data with survival outcomes. Specifically, we minimized residual confounding bias, stabilized estimates for time-dependent exposures, developed LASSO-based survival prediction after addressing missing data challenges using multiple imputation technique, and improved risk prediction after supplementing unobserved important clinical predictors. These methodological advancements have implications for the analysis and interpretation of health administrative databases in TB literature and beyond.

View record

Background: Although some progress has been made in understanding the period preceding the classical recognition of MS, further research is necessary to fully define this period and assess the influence of bias. The thesis objectives in the period before a first demyelinating disease claim or MS symptom onset included: (1) to describe the diversity of clinical pathways; (2) to describe the joint trajectories of physician visits, hospitalizations, and prescription drug classes; (3) to re-examine the association between MS and depression/anxiety after correcting for three major sources of bias; (4) to investigate the mediating relationship between MS, physician visits, and hospitalizations; and (5) to investigate the association between MS and fractures, dislocations/sprains/strains, and burns.Methods: This was a cohort study using clinical and population-based health administrative data in British Columbia, Canada (1991–2020). A multichannel state-sequence analysis of clinical pathways and group-based multi-trajectory modelling of physician visits, hospitalizations, and prescription drug classes was conducted. Probabilistic quantitative bias analyses were used to correct for misclassification and differential surveillance, and time-distribution matching to correct for misclassified person-time. A causal mediation analysis was conducted of MS, physician visits, and hospitalizations. A modified Poisson regression analysis was used to investigate fractures, dislocations/sprains/strains, and burns, with hdPS adjustment and targeted maximum likelihood estimation.Results: In the state-sequence analysis, nine pathways were identified before a first demyelinating claim, and five before MS symptom onset. In the trajectory analysis, low, moderate, and high health care use trajectories were identified before a first demyelinating claim and MS symptom onset. Correction for misclassification, differential surveillance, and misclassified person-time yielded a stronger association between depression/anxiety and MS. Before both a first demyelinating disease claim and MS symptom onset, the largest proportion of the total effect of MS on hospitalizations was mediated through physician visits related to the nervous system and mental health, and the risk of a fracture, dislocation/sprain/strain, and burn was higher among MS cases.Conclusion: Findings suggest that there is potential to identify and manage MS earlier. The advanced methods used enabled a more comprehensive understanding of the period preceding the classical recognition of MS.

View record

BACKGROUND: Improvements in and expansion of tuberculosis (TB) diagnosis and treatment have yielded a growing population of TB survivors, with an estimated 155 million alive in 2020. While TB is preventable and curable, there is accumulating evidence of elevated chronic disease risk among survivors. Research objectives: (1) estimate the relative risk of non-TB mortality among TB survivors compared with controls, (2) systematically review the literature on cardiovascular disease (CVD) in TB and latent TB infection, (3) estimate the relative risk of airway disease among respiratory TB survivors compared with controls, and (4) estimate the relative risk of depression among TB survivors compared with controls, mediated by hospital length of stay (LOS).METHODS: A cohort of immigrants to British Columbia, Canada, 1985-2015, with linked health administrative and TB registry data was used for retrospective cohort studies of TB survivor health. Cox proportional hazards (PH) and time-varying models were used in statistical analyses. Causal mediation analysis of depression, mediated by hospital LOS, estimated depression risk. A prospectively registered systematic review and random-effects meta-analysis of TB and CVD was performed. RESULTS: In a time-varying Cox regression of non-TB mortality, an adjusted hazard ratio (aHR) of 1.69 (95% CI:1.50-1.91) was observed between TB exposed and non-TB exposed time. In the systematic review and meta-analysis, we found increased risk of major adverse cardiovascular events (MACE) among TB patients compared with non-TB controls (pooled RR = 1.51; 95% CI: 1.16-1.97). A higher risk of airway disease among respiratory TB survivors compared with non-TB controls was observed in our Cox PH regression (aHR=2.08; 95% CI: 1.91-2.28). In the causal mediation analysis of depression, TB survivors had aHR=1.24 (95% CI: 1.14-1.34) for depression by TB, decomposed into a natural direct effect of aHR=1.11 (95% CI: 1.02-1.21) and indirect effect of aHR=1.11 (95% CI: 1.10-1.12), indicating 50% (95% CI: 35-82%) mediation through hospital LOS.CONCLUSION: TB survivors face higher mortality from non-TB causes, and higher risk of airway disease, CVD, and depression, compared with non-TB controls. Chronic disease screening and models of care development are needed to support TB survivors’ health-related quality of life, during and after TB treatment.

View record