Introduction: The main objective of this thesis is to develop an algorithm to accurately identify cases of biopsy-proven Celiac Disease (CD) in children aged 6 months-14 years old from Ontario health administrative data. Method: CD cases diagnosed in 2005-2011 were identified from CHEO, and linked to the health administrative data to serve as reference for algorithms derivation. Algorithms based on outpatient physician visits for CD plus endoscopy billing code were constructed and tested. Results: The best algorithm selected based on performance from derivation study and clinical expertise consisted of an OHIP-based endoscopy billing claim followed by 1 or more adult or pediatric gastroenterologist encounters after the endoscopic procedure. The sensitivity, specificity, PPV, and NPV for the algorithm were 70.4%, >99.9%, 53.3% and >99.9% respectively. Conclusion: Study results suggest that the currently available Ontario health administrative data is not suitable for identifying incident pediatric CD cases.

Objectives: The objectives of this thesis were to: 1) develop and validate a coding algorithm to identify true cases of neonatal bacterial sepsis, and 2) apply the algorithm to calculate incidence rates and estimate temporal trends of neonatal bacterial sepsis. Methods: For Objective 1, the reference cohorts were assembled among neonates born in 2012-2017 using patient-level health care encounter data. Any neonates who met both the Diagnostic Criterion Ⅰ (microbiological confirmation) and Criterion Ⅱ (sepsis-related antibiotic administration) were included in the true-positive cohort. Potential coding algorithms were developed based on different combinations of ICD-10-CA codes on the hospitalization discharge abstract. For Objective 2, the coding algorithm with the most optimal characteristics was applied to provincial data to calculate incidence rates in Ontario during 2003-2017. Recent temporal trends were estimated by Poisson regression analysis. Results: In Objective 1, since all true-positive cases identified were born at preterm gestation, the study population in Objective 2 was limited to preterm infants. The final coding algorithm selected had sensitivity of 75.3% (95% CI, 66.8%-83.7%), specificity of 98.2% (95% CI, 97.8%-98.6%) and PPV of 50.0% (95% CI, 42.1%-58.0%). Using this algorithm, the annual incidence declined over time from 50.2 (95% CI, 45.4-55.4) per 1000 preterm infants in 2003 to 27.5 (95% CI, 20.4-36.9) per 1000 preterm infants in 2017. The trend over time was statistically significant with P-value <0.0001. Significant variation in bacterial sepsis incidence rates was noted across infant sex and gestational age. Conclusion: The coding algorithm developed in this study could not accurately identify neonates with bacterial sepsis from within health administrative database using the data available to us now. For the purpose of demonstrating the application of the algorithm, we carried out Objective 2; however, it is important to cautiously interpret the provincial rates given the the poor performance of the case-finding algorithm.

There is limited data on mortality, health service use and costs following treatment of malignant pleural effusions (MPE) in the real-world setting. We performed a retrospective population-based study using health administrative data of adults with indwelling pleural catheter (IPC) insertion (n=4,574) or pleurodesis (n=1,235) for MPE between 2015 to 2019. Inverse probability of treatment weighting using the propensity score was performed to adjust for baseline characteristic imbalances. After weighting to balance on baseline characteristics, there was no significant difference in post-procedure mortality between individuals receiving IPCs and pleurodesis, with IPCs inserted significantly later after an initial cancer diagnosis. IPCs with home nursing drainage were associated with reduced subsequent health resource use and healthcare costs compared to pleurodesis. A protocol was developed for a future economic evaluation to compare the cost-effectiveness of the procedures. This thesis provides the foundation for further research to help optimize the treatment of individuals with MPEs.

As the population ages, policymakers struggle to cope with the increasing demands for home care and institutional long-term care. This thesis project focuses on factors associated with the transition from home to institutional care for people with dementia. Using health administrative data at a population level, we construct a multivariable model that estimates the time between home care initiation after dementia diagnosis and placement in a long-term care home. From the model, we identify protective factors that allow people with dementia to remain at home for longer, with a particular emphasis on the health workforce and the contribution of formal and informal caregivers to delaying the transition from home to institutional care. Together, these results inform policymakers in capacity planning and in determining where investments should be targeted to maintain people with dementia at home, along with the associated health workforce implications.

Background: Investigators are often interested in using population-level health administrative data in inflammatory bowel disease (IBD) patients to study disease outcomes, risk factors and treatment effects to enhance knowledge, shape clinical practice and influence health care policy. A major limitation of using health administrative data for these purposes is the lack of detailed clinical data to adjust for the confounding effects of differential disease severity on observed associations. Methods to account for disease severity using administrative variables would offer a major advance to population-level studies in IBD patients. Thus, in this study we aimed to use a cohort of IBD patients from The Ottawa Hospital (TOH) to validate a model that was originally developed in Manitoba for estimating clinical disease course in IBD patients through healthcare utilization measures. Objectives: The objectives of this thesis are: 1) To identify and characterize a reference cohort of IBD patients in the ambulatory clinics of four gastroenterologists from TOH on clinical disease course in the preceding year (reference cohort), based on a Manitoba definition of clinical disease course; 2) To fit a partial proportional odds (PPO) model for predicting IBD course, derived using Manitoba health administrative data, to the reference cohort of IBD patients using Ontario health administrative data; 3) To derive new PPO models of IBD disease course for the reference cohort using Ontario administrative variables and compare model performance; and 4) To apply the models to the Ontario Crohn’s and Colitis cohort (OCCC) to estimate IBD course in Ontario, and compare the distribution to that of the Manitoba IBD population.Methods: We first identified a reference cohort of IBD patients in Ontario from the outpatient clinics at TOH during fiscal year 2015. Through chart review, we classified these patients into one of four clinical disease categories (remission, mild, moderate, or severe) using the Manitoba definition. We linked these patients to Ontario health administrative datasets. Given slight differences in data structure and coding between Manitoba and Ontario, we were unable to directly test the Manitoba model and instead fit a PPO model to the Ontario cohort using analogous administrative variables to those used in the final Manitoba model (“adapted model”). We subsequently derived new PPO models using unique Ontario administrative variables under three strategies: 1) Stepwise variable selection (“stepwise model”); 2) Forced fitting of all variables (“all-variables model”); and 3) Using a two-step modelling algorithm that considered IBD-related hospitalizations separate from other administrative variables (“two-step model”). We then compared model performance from the four strategies. Finally, we applied the models to the Ontario IBD population from 2004 to 2016 and compared model estimates to those from Manitoba. Results: We identified 963 patients with IBD from TOH outpatient clinics, of which 52.3% (n=504) were males, 64.6% (n=622) had Crohn's Disease, and 89.2% (n=859) resided in an urban setting. Based on the Manitoba definition, 64.9% of patients within our reference cohort were classified as remission, while 11.4%, 14.1%, and 9.6% were classified as mild, moderate, and severe disease course, respectively. The adapted model (c-statistic 0.77, goodness-fit p-value 0.28) performed comparably to the other models: the stepwise model (c-statistic 0.77, goodness-fit p-value 0.50), the all-variables model (c-statistic 0.77, goodness-fit p-value 0.53), and the two-step model (c-statistic 0.78, goodness-fit p-value 0.75). The adapted model also resulted in overall similar estimates with regards to the disease course distribution among the Ontario IBD population. However, on closer inspection, our two-step model, in which individuals who had been hospitalized for an IBD-related indication within the past year were assumed to have severe disease, performed better with respect to accurately classifying individuals with moderate or severe disease, without sacrificing discriminative ability. Based on the two-step model, from 2004 to 2016, 89.2-91.2% of the Ontario IBD population was in remission, 0% had mild disease, 2.4-3.2% had moderate disease, and 5.9-8.4% had severe disease. Distribution of disease course among IBD patients in Ontario differed considerably than that in Manitoba. Conclusion: In the absence of clinical information within health administrative data, we present and compare four different models that can be used to partially account for the confounding effect of disease course among IBD patients in future population-based studies using Ontario health administrative data. Given that our models did not perform as originally expected, especially with regards to accurately identifying individuals with more active disease states, we advise researchers to use these models at their own discretion.

Over the past century, the prevalence of immune-mediated inflammatory diseases (IMIDs) has increased worldwide. It has been identified that exposures to environmental factors early in life are associated with increased risk of these diseases. However, hypothesis-driven analyses do not always identify all risk or protective factors, nor do they adequately explain interactions between variables on the risk of disease. Data mining has the capability of exploring the data without considering specific a priori hypotheses, instead providing possible hypotheses for further analysis. Though, data mining techniques are still not popular among epidemiologists as a trustworthy analytical tool to analyze population-based diseases due to inexplicability of some of the methods (e.g., neural networks), unfamiliarity with, or uncommon use of machine learning and data mining methods in real-world health care applications. At the same time, large amounts of routinely collected health data are amassed as a matter of operating electronic health systems. Routinely collected health data are not collected for research purposes; however, they are great sources of information for research as a secondary use of the data. In this study, following the design science research methodology, we developed a methodology to reliably analyze health administrative data using data mining techniques to provide reproducible, reliable, and trustworthy findings. The reliable data mining methodology on health administrative data was designed in this study to address impartiality, validity, and sustainability concerns in five stages: Data Selection, Preprocessing, Modelling, Evaluation, and Feedback. As part of the main contributions, we developed two unique preprocessing guidelines as the key components of the designed methodology in order to standardize technical steps and address contextual sources of bias. While the proposed methodology is general in its design, to evaluate the designed methodology, we implemented it in several case studies on the real health administrative data housed at ICES, Ontario, first to analyze children suffering with an IMID in Ontario, predict new cases, and, most importantly, generate new hypotheses. The first case study was extended to a second one to narrow focus from all IMIDs to asthma which formed the majority of the IMID cases. Eventually, a third case study was implemented focusing on inflammatory bowel disease (IBD) and systemic autoimmune rheumatic diseases (SARDs) to better compare the findings. We applied both predictive and descriptive modelling techniques such as decision tree, neural network, logistic regression, and k-means clustering on the prepared datasets with more than 700K records and over 80 input variables. We built classification models with notable quality of performance (AUC of 68%), identified the significant factors associated to IMIDs, and extracted multifactorial rules causing protectiveness against or high risk of developing asthma, IBD, and SARDs. The factors that highly contributed to the extracted multifactorial rules were “general childhood infection”, “use of antibiotics”, “streptococcus pyogenes”, “respiratory infection”, “gastroenteritis”, “mother's prevalence of any IMID”, and “baby's sex”. The findings were evaluated and verified by health experts. Most data mining studies which are applied to health data do not handle bias and confounding in their work. However, the systematic errors were identified, and their risks were assessed in these case studies due to following the designed reliable methodology. The results with high risk of bias were reported to disregard. Therefore, this process allowed us to apply data mining techniques to discover new multifactorial rules and identify the factors with the highest impact among the 128 factors observed in the past epidemiological studies, while preserving the trust of domain experts in the results.

Inflammatory bowel disease (IBD), characterized by chronic gastrointestinal inflammation, represents a significant childhood chronic disease. In this thesis, a case ascertainment definition of paediatric-onset IBD was validated using administrative data and developed the Ontario Crohn’s and Colitis Cohort (OCCC). The epidemiology of paediatric IBD in Ontario was described, demonstrating that Ontario has one of the highest worldwide incidence rates. Statistically significant increases in incidence were noted in 0-4 year olds (5.0%/year, p=0.03) and 5-9 year olds (7.6%/year, p<0.0001), but not in other age groups. Lower income children were more likely to be hospitalized at least once (hazard ratio (HR) 1.17, 95% confidence intervals (CI) 1.05-1.30) or visit the ED (HR 1.21, 95% CI 1.09-1.35) and had more IBD-related physician visits (odds ratio (OR) 3.73, 95% CI 1.05-13.27). Lower income children with Crohn's disease (CD) (not ulcerative colitis [UC]) were more likely to undergo intra-abdominal surgery within 3 years of diagnosis (OR 1.22, 95% CI 1.01-1.49), especially if diagnosed after 2000 (OR 1.79, 95% CI 1.27-2.53). Finally, changes in health services utilization and surgical rates were described, as were changes in specialist care provision and immunomodulator use in children with IBD between 1994-2007. The changes to care included increased outpatient care provided by paediatric gastroenterologists, and increased immunomodulator use. Children diagnosed with CD, but not UC, in recent years had lower surgical rates. In CD patients, intra-abdominal surgical rates within three years of diagnosis decreased from 18.8% in children diagnosed in 1994-1997 to 13.6% in those diagnosed in 2001-2004 (P = 0.035). When stratified by age at diagnosis, this decrease was significant in children diagnosed ≥10 years old (OR 0.67, 95% CI 0.48-0.93). The OCCC will continue to be used to investigate the epidemiology and burden of paediatric IBD and to improve the care received by children with IBD in Ontario.

Inflammatory bowel disease (IBD), characterized by chronic gastrointestinal inflammation, represents a significant childhood chronic disease. In this thesis, a case ascertainment definition of paediatric-onset IBD was validated using administrative data and developed the Ontario Crohn’s and Colitis Cohort (OCCC). The epidemiology of paediatric IBD in Ontario was described, demonstrating that Ontario has one of the highest worldwide incidence rates. Statistically significant increases in incidence were noted in 0-4 year olds (5.0%/year, p=0.03) and 5-9 year olds (7.6%/year, p<0.0001), but not in other age groups. Lower income children were more likely to be hospitalized at least once (hazard ratio (HR) 1.17, 95% confidence intervals (CI) 1.05-1.30) or visit the ED (HR 1.21, 95% CI 1.09-1.35) and had more IBD-related physician visits (odds ratio (OR) 3.73, 95% CI 1.05-13.27). Lower income children with Crohn's disease (CD) (not ulcerative colitis [UC]) were more likely to undergo intra-abdominal surgery within 3 years of diagnosis (OR 1.22, 95% CI 1.01-1.49), especially if diagnosed after 2000 (OR 1.79, 95% CI 1.27-2.53). Finally, changes in health services utilization and surgical rates were described, as were changes in specialist care provision and immunomodulator use in children with IBD between 1994-2007. The changes to care included increased outpatient care provided by paediatric gastroenterologists, and increased immunomodulator use. Children diagnosed with CD, but not UC, in recent years had lower surgical rates. In CD patients, intra-abdominal surgical rates within three years of diagnosis decreased from 18.8% in children diagnosed in 1994-1997 to 13.6% in those diagnosed in 2001-2004 (P = 0.035). When stratified by age at diagnosis, this decrease was significant in children diagnosed ≥10 years old (OR 0.67, 95% CI 0.48-0.93). The OCCC will continue to be used to investigate the epidemiology and burden of paediatric IBD and to improve the care received by children with IBD in Ontario.

Background/Objectives: West Nile neuroinvasive disease (WNND) is a severe neurological illness that develops in approximately 1% of individuals infected with West Nile virus (WNV). Manifesting most frequently as encephalitis (WNE), meningitis (WNM), or acute flaccid paralysis (WNP), there is no cure for WNND beyond supportive care and rehabilitation, and death or permanent disability are common outcomes. As the virus arrived in North America less than 20 years ago, determinants of severe disease progression following infection are still being explored. This project is the first to examine comorbid conditions as risk factors of WNND in Ontario using a population-based study design. As prevention is the only avenue of defence against WNND, identifying comorbid risk factors of WNND would allow for public health prevention campaigns targeted to high-risk groups. The main objectives of this thesis were to explore whether pre-existing chronic diseases were associated with the development of WNND, or any of its three manifestations (i.e., encephalitis, meningitis, acute flaccid paralysis). Methods: This was a retrospective, population-based study including all Ontario residents with a confirmed diagnosis of WNV infection between January 1, 2002 and December 31, 2012. A cohort of individuals with WNV was identified from a provincial laboratory database and individually-linked to health administrative databases. In the WNV cohort, individuals with WNND and 13 comorbid conditions were identified using algorithms based on ICD-10-CA diagnostic codes. Incidence of WNND following WNV infection was then compared among individuals with and without comorbid conditions using relative risks estimated by log binomial regression. Additionally, risk ratios were calculated for associations between specific comorbid conditions and WNND neuroinvasive manifestation (i.e., encephalitis, meningitis, acute flaccid paralysis). Finally, associations between Charlson Comorbidity Index (CCI) scoring and development of WNND was examined through calculation of relative risk using log binomial regression. Results/Potential Impact: Risk factors for WNND included male sex (aRR: 1.21; 95% CI: 1.00-1.46) in addition to the combined effect of hypertension and increasing age (5-year intervals) (aRR: 1.16; 95% CI: 1.08-1.24); WNND was also associated with increasing CCI scores; individuals in low, medium, and high categories had increased risk compared to individuals with a score of zero, but the greatest risk was in the high CCI category (aRR: 3.45; 95% CI: 2.25-4.83) Male sex (aRR: 1.32; 95% CI: 1.00-1.76), increasing age (aRR: 1.02; 95% CI: 1.02-1.03), and being immunocompromised (aRR: 2.61; 95% CI: 1.23-4.53) were associated with development of WNE. No risk factors were identified for WNM and WNP. Identification of comorbid risk factors of WNND will allow public health officials to identify high-risk groups and to develop prevention strategies targeted for vulnerable individuals.