Search

Introduction The Canadian Institute of Health Information's (CIHI) Discharge Abstract Database (DAD) contains standardised administrative data on all hospitalisations in Canada, excluding Quebec. Objectives We aimed to validate preterm birth related perinatal and neonatal data in DAD by assessing its accuracy against the reference standard of the Canadian Neonatal Network (CNN) database. Methods We linked birth hospitalization data between the DAD and CNN databases for all neonates born $

Objective To evaluate the performance of a natural language processing (NLP) method to develop an automated crosswalk between the 9th and 10th revisions of the International Classification of Diseases (ICD) for diagnosis codes in the Charlson comorbidity index (CCI). Approach SBERT, an advanced NLP transformer-based model, was used to produce sentence embeddings, numeric vectors that represent the semantic meaning of text, for the labels (i.e., descriptors) of 932 ICD-10-CA (Canadian Adaptation) codes in the CCI (up to six digits). Sentence embeddings were also produced for all ICD-9-CM (Clinical Modification) code labels (15,145). Cosine similarity scores (CSS) were calculated for all possible pairs of ICD-10-CA and ICD-9-CM code labels. CSSs were classified as equivalent (CSS = 1), high (0.8 ≤ CSS < 1), and low (CSS < 0.8). CSS categories for CCI diagnosis codes were compared to an ICD-9-CM to ICD-10-CA crosswalk file manually created by the Canadian Institute of Health Information. Results Of the 932 CSSs for ICD-10-CA codes in CCI, 84 (9%) were classified as equivalent, 284 (30.5%) were high, and 564 (60.5%) were low. For ICD-10-CA codes with low CSSs, the median was 0.67 (interquartile range 0.14). Conclusions and Implications An ICD-10-CA to ICD-9-CM crosswalk based on NLP had low accuracy for identifying semantically similar diagnosis code labels. The accuracy of this method might be improved by fine-tuning and training on task-specific data. Evaluation of different text analysis-based models would provide guidance for research involving ICD code labels.

Background The Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) is a standardized structure for observational data that enables collaborative health research. Health Data Research Network (HDRN Canada) is conducting a pilot study to implement this standard at partner data linkage centers. Objective Our objective is to describe the process used to implement the OMOP CDM, methodological and technical challenges, and the approach to evaluate CDM implementation at multiple sites. Approach CDM working group members include technical staff from data linkage centers in four provinces and a national data coordinating center, HDRN Canada operational and executive leads, and OMOP CDM consultants. The evaluation aims to collect quantitative and qualitative information on Extract, Transform, and Load (ETL) processes, performance of the CDM in a multi-site observational case study, and data quality. Results Initial steps have included collaborative learning about the CDM structure, importing open-source tools and licensed coding standards (i.e., Systematized Nomenclature of Medicine-Clinical Terms [SNOMED-CT], RxNorm) at data linkage centres, and exploring the technical requirements for developing ETL processes and hosting OMOP data. Creation of select OMOP tables has been accompanied by discussions about models for mapping clinical and prescription drug coding standards. A scoping review of evaluation strategies is underway. Conclusions and Implications The benefits of standardizing health data for multi-site research and supporting the production of findable, accessible, interoperable and reusable (FAIR) national data underscores the importance of HDRN Canada embarking on this OMOP CDM implementation project. The lessons learned will benefit other multi-site data standardization initiatives.

Objective Administrative data often lacks complete family linkages, particularly to fathers, hindering familial health research. We compared methods to address missing paternal linkages in research investigating the familial transfer of mental disorders. Approach A population-based cohort study of Manitoba (Canada) born adults, +18 years old between 1977 and 2017 with maternal linkages. Three methods were used to address missing paternal linkages: indicator category, complete case, and multiple imputation, which identified three candidate fathers for each individual within -4 to +7 years of mother’s age with same postal code. For each method, the association of maternal and paternal history with mental disorder risk during follow-up was tested using multivariable logistic regression models adjusted for demographics and comorbidities. Results The cohort included 142,549 individuals; 22.6% lacked paternal linkages. Using indicator category, maternal and paternal histories were associated with mental disorder risk, with odds ratio (OR) 1.49, 95% confidence interval (CI): 1.45-1.52 and OR 1.34, 95% CI: 1.30-1.37, respectively. Similar results were obtained in the complete case analysis (OR 1.49, 95% CI: 1.45-1.53 and OR 1.33, 95% CI: 1.30-1.37, for maternal and paternal history, respectively). With multiple imputation, maternal history’s association with mental disorder risk remained consistent with the other methods (pooled OR 1.52, 95% CI: 1.49-1.56); paternal history was associated with a smaller risk (pooled OR 1.25 95% CI: 1.22-1.29). Conclusions and Implications The three methods for addressing missing paternal linkages produced similar findings about the familial transfer of mental disorders. Future familial studies could incorporate multiple methods to demonstrate robustness of findings.

Background The International Classification of Diseases (ICD) is revised over time and there are region-specific versions, including ICD-10-CA (Canada) and ICD-9-CM (USA). Studies spanning multiple ICD versions require crosswalks to translate diagnosis codes across versions, but manual crosswalk development is costly and requires clinical expertise. Objective To evaluate the accuracy of a pre-trained large language model (LLM) to automatically translate chronic disease diagnosis codes from ICD-10-CA to ICD-9-CM. Approach Eight prompts were developed to instruct the OpenAI Generative Pre-trained Transformer 4 (GPT-4) LLM to translate 1,272 ICD-10-CA codes for the Elixhauser Comorbidity Index to ICD-9-CM. Prompt accuracy (%) was measured against a crosswalk developed by the Canadian Institute of Health Information. Variability was assessed by replicating each prompt three times. Mean accuracy ± standard deviation was reported for each prompt across replications, for both five-digit and truncated three-digit codes. Results The highest prompt performance was observed when assigning a persona of a medical coding specialist (40.8% ± 0.9%), requesting justification for the selected code (41.4% ± 1.1%), and providing diagnosis code labels (47.5% ± 0.7%). For truncated three-digit codes, these prompts achieved accuracy of 82.0% ± 0.5%, 80.8% ± 0.9%, and 82.7% ± 0.1%, respectively. Combining these three prompting techniques marginally improved accuracy to 48.6% ± 0.7% for five-digit codes and 84.3% ± 0.2% for truncated three-digit codes. Conclusion General-purpose LLMs are currently not sufficiently accurate at automating ICD code translation for chronic diseases. Implications Additional experiments with fine-tuning, task-specific training, and prompt engineering are needed to improve accuracy and reduce variability.

Background Research and surveillance using routinely collected health data rely on algorithms to ascertain disease cases or health measures. Where algorithm validation studies are not possible due to lack of a reference standard, algorithm feasibility studies can be used to create and assess algorithms for use in more than one population or jurisdiction. Publication of the methods used to conduct feasibility studies is critical for transparency and reproducibility. Existing guidelines applicable to feasibility studies, including the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) and REporting of studies Conducted using Observational Routinely collected health Data (RECORD) statements, may benefit from additional elements to capture aspects particular to multi-jurisdiction algorithm feasibility studies and ensure full reproducibility. Methods A subcommittee of members from Health Data Research Network (HDRN) Canada’s Algorithms and Harmonized Data Working Group (AHD-WG) reviewed items within the STROBE and RECORD guidelines and compared these to published feasibility studies. Items not contained within STROBE or RECORD but recommended to ensure transparent reporting of feasibility studies were identified. The AHD-WG reviewed and approved these additional recommended elements. Results Eleven additional recommended elements were identified: one element for the title and abstract, one in the introduction, five in the methods, and four in the results sections. Elements primarily addressed reporting jurisdictional data variabilities, data harmonization methods, and algorithm implementation. Significance Implementation of these recommended elements, alongside the RECORD guidelines, is intended to encourage consistent publication of methods that support reproducibility, as well as increase comparability and international collaborations.

Introduction The validity of chronic disease case definitions for administrative health data may change over time due to changes in data quality. Trend control charts to identify out-of-control (OOC; i.e., unexpected) observations in a time series may indicate where disease estimates are influenced by changes in data quality. Objective Apply and compare trend control charts methods for multiple sclerosis (MS) incidence and prevalence estimates using previously-validated case definitions for Manitoba, Canada. Methods Eight case definitions were identified from published literature and applied to Manitoba administrative health data from January 1, 1972 to December 31, 2018. Incidence and prevalence trends were modeled using negative binomial and generalized estimating equation models, respectively. Trend control charts were used to plot predicted case counts against observed case counts. Control limits to identify OOC observations were calculated using two methods: predicted case count ±0.8*standard deviation (0.8*SD) and predicted case count ±2*standard deviation (2*SD). Differences in proportion of OOC observations across case definitions was assessed using McNemar's test. Results The proportion of OOC observations ranged from 0.71 to 0.90 for incidence and 0.72 to 0.98 for prevalence when using the 0.8*SD control limits. A lower proportion of OOC observations (0.46 to 0.74 for incidence; 0.30 to 0.74 for prevalence) was observed for the 2*SD control limits. Neither method resulted in significant differences in OOC observations across case definitions. Conclusions The proportion of OOC observations in trend control charts varied with the control limit method adopted, but statistical significance did not. Trend control charts are a potentially useful tool for developing surveillance methods, but may benefit from disease-specific calibrated control limits.

Introduction Unstructured text data (UTD) are increasingly found in many databases that were never intended to be used for research, including electronic medical record (EMR) databases. Data quality can impact the usefulness of UTD for research. UTD are typically prepared for analysis (i.e., preprocessed) and analyzed using natural language processing (NLP) techniques. Different NLP methods are used to preprocess UTD and may affect data quality. Objective Our objective was to systematically document current research and practices about NLP preprocessing methods to describe or improve the quality of UTD, including UTD found in EMR databases. Methods A scoping review was undertaken of peer-reviewed studies published between December 2002 and January 2021. Scopus, Web of Science, ProQuest, and EBSCOhost were searched for literature relevant to the study objective. Information extracted from the studies included article characteristics (i.e., year of publication, journal discipline), data characteristics, types of preprocessing methods, and data quality topics. Study data were presented using a narrative synthesis. Results A total of 41 articles were included in the scoping review; over 50% were published between 2016 and 2021. Almost 20% of the articles were published in health science journals. Common preprocessing methods included removal of extraneous text elements such as stop words, punctuation, and numbers, word tokenization, and parts of speech tagging. Data quality topics for articles about EMR data included misspelled words, security (i.e., de-identification), word variability, sources of noise, quality of annotations, and ambiguity of abbreviations. Conclusions Multiple NLP techniques have been proposed to preprocess UTD, with some differences in techniques applied to EMR data. There are similarities in the data quality dimensions used to characterize structured data and UTD. While a few general-purpose measures of data quality that do not require external data; most of these focus on the measurement of noise.

Objectives Prediction of asthma risk can potentially be improved by including family history of asthma and related diagnoses, which reflect both genetics and shared environments. We tested the improvements in offspring asthma risk prediction using objectively-measured maternal, paternal, and offspring histories of comorbid conditions from administrative healthcare databases. Approach A population-based cohort study was conducted using data from Manitoba, Canada. Children born from 1974 to 2000 with linkages to at least one parent using family identification numbers were included. Asthma diagnosis and comorbidities were identified from hospital and outpatient physician visit records. Lasso regression models were used to assess performance and identify important predictors. The base model included offspring demographics, diagnosed allergic conditions and respiratory infections, and diagnosed parental asthma. Subsequent models included multiple comorbid chronic health conditions for offspring and parents. Results The cohort included 195,666 offspring; 51% were males, 13.6% had a parental asthma diagnosis, and 17.7% had an asthma diagnosis (median age at diagnosis: 6.0 years; interquartile range 3.0-11.0 years). The base model achieved a modest prediction performance with an area under the receiver operating characteristic curve of 0.60, sensitivity of 0.46 and a specificity of 0.67 using a threshold of 0.20. Sensitivity significantly improved when we included offspring chronic health conditions (sensitivity= 0.69; specificity = 0.66); both measures further improved when we additionally included parents’ chronic health conditions (sensitivity= 0.72; specificity = 0.70). Chronic obstructive pulmonary disease, noninfectious gastroenteritis and otitis media were among the variables that added incremental predictive value of asthma risk with odd ratios of 1.36, 1.25 and 1.18, respectively. Conclusions Including offspring and parents’ chronic health conditions, identified objectively from administrative healthcare databases, improved the performance of asthma risk prediction models in children. Health histories of comorbid conditions provide important factors to improve risk prediction models of chronic health conditions, which will facilitate disease prevention and treatment strategies.

Objective A family history of a chronic disease often predicts disease risk, with predictive value determined by heritability, the proportion of variation in risk explained by inherited genetic factors. Our objective was to assess the validity of disease heritability estimates from electronic healthcare records (EHRs) that capture family relationships and disease diagnoses. Approach A population-based investigation was conducted using healthcare records from Manitoba, Canada for 1970 to 2021. We constructed family relationships for up to four generations using health insurance registration information containing unique family and individual identifiers. Health histories for family members were created using diagnosis codes in hospital and physician visit records. Linear mixed-effects models were used to estimate heritability (h) for 130 chronic health conditions using open-source Clinical Classifications Software that defines clinically-meaningful disease categories. Comparisons between EHR-derived estimates and genetically-derived estimates from published studies were used to assess validity of the methodology. Results Health insurance registration data were used to construct 10,000 families that included 116,879 individuals. Median family size was 9 (interquartile range: 8). Median observation time was 39.6 years (interquartile range: 25.7). Males comprised half (51.0%) of family members. A total of 272,114 familial relationships were identified; slightly more than half (53%) were first degree (i.e., child and parent) relationships. One-third (33.2%) of families were comprised of four generations; only 15.3% were comprised of two generations. Heritability estimates were consistent with published genetically-derived estimates for several conditions, including diabetes (EHR h = 0.29 vs. 0.22), anemia (EHR h = 0.21 vs. 0.20), and asthma (EHR h = 0.34 vs. 0.33). However, inconsistencies were identified for pancreatic disorders, gastrointestinal conditions, some mental health conditions, and heart disease. Conclusion EHRs provide a promising approach to explore heritability of selected health conditions in large, diverse populations. Inconsistencies between EHR-derived and genetically-derived estimates are indicative of the limitations of diagnoses recorded for administrative purposes. Future research will explore sex-specific heritability estimates and effects of change in disease diagnosis coding over time.

Objectives First Nations (FN) organizations worked with public health and governments to improve FN access to COVID-19 vaccines by prioritizing FN communities in vaccination initiatives. FN researchers and data scientists partnered to test whether these efforts were associated with increased access to COVID-19 vaccines among FN compared with all other Manitobans. Approach This retrospective cohort study linked whole-population administrative data from (i) the First Nations research file, (ii) COVID testing and vaccination data, and (iii) health and social services for sociodemographic data and information on potential confounders. Several public health policies were created to improve access to COVID vaccines among FN; we tested whether FN received their 1st and 2nd vaccines sooner than all other Manitobans (AOM) using restricted mean survival time models. We adjusted for sociodemographic characteristics, comorbidities, and whether FN lived on- or off-reserve. We conducted sex-specific and effect modification analyses to test whether associations differed by sex. Results Prioritizing FN to receive vaccines was associated with increased vaccine uptake compared with AOM. After adjusting for various confounders, FN received their first dose 15.5 (95% CI 14.9 – 16.0) days sooner than AOM and their second dose 13.9 (13.3 – 14.5) days sooner than AOM. Sex-stratified and subsequent effect modification analyses using interaction terms, found that differences were greater for males than for females: FN males received their first dose 18.1 (17.3 – 18.8) days sooner than AOM males and FN females received their first dose 12.9 (12.2 – 13.7) days sooner than AOM females. This pattern held for second doses as well. FN with comorbidities also received vaccines sooner than AOM with similar comorbidity levels 20.9 days (23.1 – 18.8) among those with 3+ comorbidities. Conclusion Partnerships between public health entities and FN organizations that respect FN community sovereignty were instrumental in supporting FN health and well-being during COVID-19. Policies and programs that prioritized FN people for vaccines improved uptake saving lives. This partnership-based COVID-19 response can provide a framework for future public health efforts.

Objectives Red River Métis are Indigenous people hailing from the Canadian Prairies who have historically experienced poor health outcomes due to colonial practices. Researchers from the Manitoba Métis Federation (MMF) partnered with health services researchers to test whether MMF-led COVID initiatives were associated with access to COVID-19 testing and vaccines. Approach We linked the Métis Population Data-Base from the MMF (to identify Red River Métis) with whole-population COVID testing and vaccination data and health and social services administrative data (for information on sociodemographics and confounders) to complete this retrospective cohort study. We used restricted mean survival time models to test whether COVID-19 vaccination differed between Métis and all other Manitobans (AOM); models adjusted for demographics, comorbidities, and other characteristics (age, socioeconomic status, urbanicity, and mental health status). Data were stratified by sex and subsequent effect modification analyses tested whether associations differed by sex and physical health comorbidities. Results COVID testing rates were lower during the first year of the pandemic among Métis than among AOM. During the second year of the pandemic, this finding was reversed - Métis accessed tests at higher rates. There was no difference between Métis and AOM in accessing first vaccine doses before implementation of MMF-led initiatives. After initiatives were put in place, Métis received their second COVID vaccine, on average, 1.3 (95% CI 1.9-0.6) days sooner than AOM, after adjusting for confounders. Effect modification analyses showed this relationship was concentrated among females – female Métis received their second vaccine 1.7 (2.6-0.8) days sooner than female AOM; differences were non-significant for males. Métis with 2+ comorbidities received their vaccine second 2.9 (5.3-0.5) days sooner than AOM with 2+ comorbidities. Conclusion Public health initiatives prioritizing Métis for vaccines improved uptake. Initiatives led by Métis to improve COVID outcomes were critical to supporting Métis during the course of the pandemic. Public health response efforts need to operate from a standpoint that honours Indigenous sovereignty in their design and implementation.

Objective Post-acute COVID-19 (or ‘long COVID’) manifests as a wide range of long-lasting symptoms affecting multiple organ systems. We are developing criteria for identifying long COVID cases using administrative, clinical, survey and other data from Manitoba, Canada, with the ultimate goal of examining long COVID prevalence, risk factors, prognosis and recovery. Approach Given the lack of an accepted clinical definition and resulting lack of diagnostic codes, we are adopting several different creative and complementary strategies to identify long COVID cases. We are examining administrative and clinical data sources (laboratory data, physician claims, drug prescriptions, and electronic medical records) for information on positive COVID tests, common symptoms and complaints, and treatment provided. To identify people with long COVID who may not have sought healthcare, we are collecting survey data from a convenience community sample (members of a medical health fitness facility) and mining data on long COVID symptoms from Twitter. Results The combination of approaches we have adopted and the expanding scientific literature on long COVID are contributing to a more comprehensive understanding of the impacts of long COVID in Manitoba. Through preliminary work on the laboratory data (positive COVID tests March 2020-June 2021), we have developed and characterized a COVID-positive cohort (n=47,515). Work is now underway to develop an algorithm for long COVID using symptoms from free text in electronic medical records, ICD-9 codes, and changes in health-seeking behaviour (compared to the pre-positive COVID test period and a matched sample). This population data-driven approach will then allow us to examine how multiple underlying health conditions, COVID illness severity, COVID vaccination status, and various socio-demographic factors are related to risk of long COVID. Conclusion This research is generating actionable information by identifying risk factors to support clinical diagnosis of long COVID, making it easier for clinicians to recognize this new illness and develop plans to manage it, and will inform healthcare system planning by quantifying the burden of long COVID at the population level.

Objectives Validation of algorithms to identify health conditions (e.g., diabetes) or service use (e.g., high-cost users) in administrative data is time-consuming and expensive. Many algorithms are only assessed in a single jurisdiction, which may limit generalizability. Our study described the characteristics of multi-jurisdiction algorithms from a Canadian algorithm repository. Approach We summarized algorithms captured in the open-access Algorithms Inventory developed by Health Data Research Network (HDRN) Canada. This inventory contains published algorithms identified through a series of systematic reviews of peer-reviewed research. Algorithms included in the inventory were validated or assessed for feasibility of implementation in two or more provinces/territories; they encompass measures of population health, health service use, and determinants of health. Descriptive statistics were used to characterize the study data on such features as year and discipline of the study journal, algorithm topic area, jurisdictions included in the study, validation source data, and algorithm elements (i.e., diagnosis codes). Results The HDRN Canada Algorithms Inventory currently contains 166 algorithms from 63 published articles. The majority of articles were published in 2010 or later (89%) and more than half (56%) of the articles were found in journals with a clinical focus. Feasibility studies (79%) were conducted more often than validation studies (21%). Most algorithms used data from the provinces of British Columbia, Manitoba, Ontario, and Nova Scotia. The majority of algorithms (72%) measured population health concepts, such as chronic physical health conditions (63%; e.g., hypertension) and mental health conditions (14%; e.g., depression). Algorithms about the determinants of health (17%) mostly focused on measures of socioeconomic status (37%) derived from census data. Multi-jurisdiction algorithms about health service use were least common (11%). Conclusion This synthesis revealed few Canadian multi-jurisdiction validation studies have been conducted and not all provinces/territories are equally represented. New validation studies, particularly about health service use and determinants of health, will increase the consistency and accuracy of Canadian research. Reusing published algorithms from this inventory will facilitate research reproducibility.

Objectives Mental disorders are a major public health concern. Genetic and environmental factors, both reflected in family health histories, jointly contribute to the onset of mental disorders. We examined the intergenerational transmission of mental disorders using objectively-measured family health histories from three generations. Approach A population-based cohort study was conducted using administrative healthcare databases from Manitoba, Canada. The cohort included offspring who were 18 years or older between 1977 and 2020 with linkage to 1+ parent and 1+ grandparent. Mental disorders were identified using diagnosis codes from hospitalization and outpatient physician visit records and included mood and anxiety, psychotic, and substance use disorders. Logistic regression models were mutually adjusted for mental disorder history in grandparents, parents and/or siblings in addition to offspring demographics: sex, region, decade of birth and income quintile, and comorbidity. Odds ratios (ORs) and 95% confidence intervals (95% CIs) were estimated. Results Out of 125,070 individuals, 59.1% were females and 57.8% were urban residents. 41,552 (33.2%) had a mental disorder during study period and 108,682 (86.9%) had a family member with a mental disorder history. Individuals were more likely to have a mental disorder if they had a family history: mother (OR 1.52, 95% CI 1.48-1.56), father (OR 1.21, 95% CI 1.17-1.25), sibling (OR 1.33, 95% CI 1.28-1.39), grandparent (OR 1.06, 95% CI 1.03-1.09). Compared with other mental disorders, psychotic disorders had the strongest association with family history: mother (OR 2.37, 95% CI 2.00-2.82), father (OR 3.00, 95% CI 2.40-3.76), sibling (OR 2.34, 95% CI 1.79-3.05). However, there was no association between psychotic disorders and grandparent history (OR 1.00, 95% CI 0.90-1.11). Conclusions We observed a strong association between mental disorders family history across three generations and the risk of the mental disorders in offspring. This association was observed for all the investigated mental disorders. This work highlights the value of multigenerational data linkage in understanding the intergenerational transfer of mental disorders.

Introduction Notwithstanding Canada’s exceptional longitudinal health data and research centres with extensive experience transforming data into knowledge, many Canadian studies based on linked administrative data have focused on a single province or territory. Health Data Research Network Canada (HDRN Canada), a new not-for-profit corporation, will bring together major national, provincial and territorial health data stewards from across Canada. HDRN Canada’s first initiative is the $81 million SPOR Canadian Data Platform funded under the Canadian Institutes of Health Research Strategy for Patient-Oriented Research (SPOR). Objectives and Approach HDRN Canada is a distributed network through which individual data-holding centres work together to (i) create a single portal and support system for researchers requesting multi-jurisdictional data, (ii) harmonize and validate case definitions and key analytic variables across jurisdictions, (iii) expand the sources and types of data linkages, (iv) develop technological infrastructure to improve data access and collection, (v) create supports for advanced analytics and (vi) establish strong partnerships with patients, the public and with Indigenous communities. We will share our experiences and gather international feedback on our network and its goals from symposium participants. Results In January 2020, HDRN Canada launched its Data Access Support Hub (DASH) which includes an inventory listing over 380 datasets, information about more than 120 algorithms and a repository of requirements and processes for accessing data. HDRN Canada is receiving requests for multi-province research studies that would be challenging to conduct without HDRN Canada. Conclusion / Implications Thus far, HDRN Canada services and tools have been developed primarily for Canadian researchers but HDRN Canada can also serve as a prompt for an international discussion about what has/has not worked in terms of multi-jurisdictional research data infrastructure. It can also present an opportunity for the development of metadata, standards and common approaches that support more multi-country research.

Introduction: When designing prospective and retrospective longitudinal cohort studies, investigators must make decisions about study duration (i.e. length of follow-up) and frequency of outcome measurement. The impact of these decisions have been previously investigated in the prospective setting, but have not been described for retrospective cohort studies. Objectives: To examine the impact and potential challenges of longitudinal design decisions in retrospective cohort studies and illustrate the effects of varying study duration and frequency of outcome measurement in the retrospective setting using a numeric example. Methods: Linked administrative data from Manitoba was used. The cohort included all mothers who experienced the death of an infant between April 1, 1999 and March 31, 2012 and a matched (3:1) group of mothers who did not experience death. A generalized linear mixed model was used to model differences in the trend in the number of healthcare contacts for the two groups. Holding sample size constant, the model was fit to the data for combinations of duration and frequency. Estimated standard errors and regression coefficients were compared. Results: A total of 2576 mothers were included; 644 experienced death of an infant and 1932 were matched to this group. Thirteen combinations of frequency (1, 2, 3, 4 periods/year) and duration (1, 2, 3, 4 years) were compared. As frequency increased from 1 to 4 periods/year, the standard error of the group-time-time interaction decreased up to 98.9%. As duration increased from 1 to 4 years, the standard error of the interaction decreased up to 96.9%. As frequency and duration increased, the coefficients trended toward zero. Conclusions: Retrospective designs using administrative data offer greater flexibility to select time-related design elements than prospective designs, but present potential new challenges. Recommendations about how to select and report time-related design decisions in retrospective cohort studies should be included in reporting guidelines.

Chronic diseases have a major impact on populations and healthcare systems worldwide. Administrative health data are an ideal resource for chronic disease surveillance because they are population-based and routinely collected. For multi-jurisdictional surveillance, a distributed model is advantageous because it does not require individual-level data to be shared across jurisdictional boundaries. Our objective is to describe the process, structure, benefits, and challenges of a distributed model for chronic disease surveillance across all Canadian provinces and territories (P/Ts) using linked administrative data. The Public Health Agency of Canada (PHAC) established the Canadian Chronic Disease Surveillance System (CCDSS) in 2009 to facilitate standardized, national estimates of chronic disease prevalence, incidence, and outcomes. The CCDSS primarily relies on linked health insurance registration files, physician billing claims, and hospital discharge abstracts. Standardized case definitions and common analytic protocols are applied to the data for each P/T; aggregate data are shared with PHAC and summarized for reports and open access data initiatives. Advantages of this distributed model include: it uses the rich data resources available in all P/Ts; it supports chronic disease surveillance capacity building in all P/Ts; and changes in surveillance methodology can be easily developed by PHAC and implemented by the P/Ts. However, there are challenges: heterogeneity in administrative databases across jurisdictions and changes in data quality over time threaten the production of standardized disease estimates; a limited set of databases are common to all P/Ts, which hinders potential CCDSS expansion; and there is a need to balance comprehensive reporting with P/T disclosure requirements to protect privacy. The CCDSS distributed model for chronic disease surveillance has been successfully implemented and sustained by PHAC and its P/T partners. Many lessons have been learned about national surveillance involving jurisdictions that are heterogeneous with respect to healthcare databases, expertise and analytical capacity, population characteristics, and priorities.

Introduction The Canadian Chronic Disease Surveillance System (CCDSS) is a collaboration of provincial and territorial surveillance systems which generates estimates of chronic diseases using linked population-level administrative health databases and standard case definitions. We conducted an environmental scan of administrative data validation studies and identified opportunities for CCDSS case definition enhancement. Objectives and Approach The purpose of this project is to develop a methodology for and conduct an environmental scan, identifying opportunities for enhancing the CCDSS. This multifaceted approach consists of the following elements: 1) key informant interviews and stakeholder consultations to identify new and existing priority conditions for updating/validating within the CCDSS, and new areas of conceptual and methodological relevance for administrative data disease surveillance, 2) a systematic literature review of PubMed, Ovid and Embase from 2013-2017 using MeSH terms and a librarian peer-reviewed search strategy, and 3) a review of the grey literature. Results Key stakeholders identified the following priorities for validation work and/or case definition enhancement: diabetes, mood and anxiety disorders, schizophrenia, obesity, hypertension, chronic obstructive pulmonary disease, osteoarthritis, stroke, early-onset dementia, rheumatoid arthritis and gout. Scientific and grey literature reviews of validation work for these conditions examined the following concepts/methods: 1) evaluating validity of disease-specific case definitions over time, and in different ages, sub-populations and settings, 2) defining incidence versus prevalence using linked administrative data, 3) determining opportunities and constraints of using linked administrative data to conduct surveillance on diseases that are chronic versus episodic in nature and defining active versus lifetime prevalence, and 4) assessing the feasibility of using new sources of data for linkage to enhance case definition validity. Conclusion/Implications Utilization of linked administrative databases for chronic disease surveillance has expanded across many jurisdictions since the inception of the CCDSS. As disease estimates generated in this manner are increasingly being relied upon by policy makers working to enhance public health, the methodological opportunities and constraints identified here require consideration.

Introduction Researchers and decision makers from across Canada use linked provincial administrative data for analysis and to address research and policy questions. Currently there are several impediments to working harmoniously across provincial boundaries. A group of academic and policy researchers are working to address these multi-jurisdictional obstacles. Objectives and Approach Researchers and data organizations from across Canada are working together as the Pan-Canadian Real-World Health Data Network PRHDN). PRHDN aims to: (1) create harmonized data, algorithms and analytic protocols, and (2) link administrative databases to other types of data, including electronic medical records, clinical trials records, “omics data” and records from pan-Canadian cohort studies. PRHDN’s vision is to construct a unified, documented infrastructure to advance pan-Canadian population-based research and analysis. This presentation incorporates material that is part of PRHDN’s response to a funding call to create national, collaborative infrastructure. Results Scientists and staff at PRHDN organizations will create three main categories of infrastructure: 1) Algorithms: Reusable processes, ideally in the form of documented code, which implement a common approach or definition, e.g. to define cases or to create derived variables; 2) Harmonized Common Data: Based on the Sentinel model, we will establish a standardized subset of harmonized common data that are analysis-ready; 3) Common Analytic Protocols: Complementing work of the Canadian Network for Observational Drug Effect Studies (CNODES), we will establish processes for distributed analysis with common analytic protocols and meta-analysis of results to provide pan-Canadian estimates. Source data would remain within jurisdictional boundaries and only aggregate results would be pooled across jurisdictions. Details of these approaches will be presented. Conclusion/Implications This initiative will improve coordinated access to distributed data from across Canada that is built once then used by many stakeholders for a variety of purposes including: research, benchmarking, performance monitoring to identify gaps and opportunities for improvement, multi-jurisdictional evaluations of novel interventions and inter-jurisdictional comparisons.

Introduction Clinical registries are a potentially valuable resource to study the effects of medical interventions on outcomes, particularly patient-reported outcomes like health-related quality of life, which are not included in administrative data. However, because clinical registries are primarily intended for patient management and not for research, their validity must be established. Objectives and Approach Our objective was to validate patient self-reported health conditions in a clinical registry. Study data were from a population-based regional joint replacement registry in the Canadian province of Manitoba. The clinical registry data were linked to administrative health data. Validated administrative data algorithms for 12 conditions were defined using diagnosis codes in hospital and physician records and medication codes in prescription drug records for the period up to three years prior to the joint replacement surgery. Accuracy of the clinical registry data was estimated using Cohen’s kappa coefficient, sensitivity, specificity, and positive and negative predictive values (PPV; NPV); 95% confidence intervals were also estimated. Analyses were stratified by joint type, age group, and sex. Results The study cohort included 20,592 individuals (average age 66.3 years; 58.4% female); 8,424 (40.9%) had a total hip replacement. Sensitivity of the clinical registry data ranged from 16% (anemia) to more than 70% (diabetes, high blood pressure, rheumatoid arthritis); specificity was greater than 92% for all conditions, except back pain and high blood pressure. PPV ranged from 19% (back pain) to 83% (diabetes). Chance-adjusted agreement was very good for diabetes (kappa: 0.74), moderate for heart disease and high blood pressure (kappa range: 0.41-0.53) and poor or fair for back pain, anemia, cancer, depression, kidney disease, liver disease, rheumatoid arthritis and stomach ulcers (kappa range: 0.14-0.37). Estimates varied by sex (i.e., generally higher agreement for females) and age (i.e., generally lower agreement for older age groups), but not joint type. Conclusion/Implications Self-reported health conditions in registry data had good validity for conditions with clear diagnostic criteria, but low validity for conditions that are difficult to diagnose or rare (e.g., cancer). Linked registry and administrative data is strongly recommended to ensure valid and accurate comorbidity measures when developiong risk prediction models and conducting inter-jurisdictional comparisons of patient-reported outcome measures.

Introduction Public housing is a form of subsidized housing that is owned and/or managed by government. Previous research suggests that public housing has a positive impact on personal finances and education outcomes, but less is known about if/how it impacts health and healthcare use. Objectives and Approach Using linked administrative health and social data, we tested for changes in healthcare use among a cohort who moved into public housing in 2012 and 2013 in Manitoba, Canada, and compared utilization to a matched general population cohort who did not move into public housing. Generalized linear models with generalized estimating equations tested for differences in numbers of healthcare contacts in the years before and after the move-in date, adjusted for economic, residential mobility, and health characteristics. The data were modeled using a Poisson (rate ratio, RR), negative binomial (incident rate ratio, IRR), or a binomial (odds ratio, OR) distribution. Results There were 2619 residents in the public housing cohort; 99.7% were matched to the general population. The cohort by time interaction was statistically significant for inpatient days (p Conclusion/Implications Public housing residents were more likely to use healthcare services than the matched population, but changes in use were similar in the two cohorts. There is little evidence that public housing impacts healthcare use, but it serves an important function of meeting basic needs for a vulnerable population group.

Introduction Although RCTs remain the gold standard for generating clinical evidence, follow up of participants to study long-term effects is limited by cost and other logistical considerations. Linkage of participant information to routinely collected data potentially offers a cost-effective solution to achieving long-term follow-up of treatment effects Objectives and Approach This scoping review aimed to identify RCTs that had been extended by record linkage, and characterize these in terms of nationality, numbers of trials, disease areas and outcomes, types of data, linkage modes and duration of follow-up. We followed published guidelines for the conduct of scoping reviews, with a registered protocol and comprehensive literature search. Criterion-based selection of studies and extraction of date were performed in duplicate. Descriptive statistics were used to summarise the characteristics of eligible studies. Results One hundred thirteen RCTs had been extended by record linkage. Fifty-six were conducted in Nordic countries, 26 in the USA and 24 in the UK. Types of linkage data used were: vital statistics 36, adimistrative data 31, cancer registry 28, special registries 13 and others 11. The literature spanned 45 years, but 66 (58%) were published between 2010 and 2016. Linkage methods were reported as: deterministic 33, probabilistic 16 and unspecified 64. In 44 studies researchers reported ethics approval for linkage; this was not obtained in 39 cases and was absent in 30 reports. The overall follow up times achieved by record linkage were: 1-4 y (6 studies), 5-9y (34), 10-19y (48), 20-29y (21), 30-39y(4) and over 50y (1). Conclusion/Implications Although we uncovered over 100 RCTs that were extended by record linkage this is tiny compared with the number of trials that have been undertaken. Linkage to routinely collected data seems to be a feasible but under-used approach to extending the follow-up of clinical trial participants for very long periods.

Background In December 2017 the Canadian Institutes of Health Research (CIHR) issued a request for proposals to develop a pan-Canadian health data platform. This platform will enable cross-jurisdictional research by facilitating the use of rich provincial and national data and ensure engagement with patients and specific populations including Indigenous partners. Academics and policy makers from across Canada operating under the banner of the Pan-Canadian Real-World Health Data Network (PRHDN) have joined forces to address this call. Objectives Create national infrastructure that is built once then made available for research, benchmarking, performance monitoring, multi-jurisdictional evaluations and inter-jurisdictional comparisons to address pressing health and social policy problems in Canada. Methods Our approach will address several issues including creating significant efficiencies in data access, streamlining cross provincial/ territorial ethics and access approvals, establishing standards for data and methods harmonization and providing innovative and privacy-conscious solutions to data access and use. The presentation will focus on the plan to create harmonized common data, algorithms and analytic protocols, and link administrative data to electronic medical records and clinical trials to create an integrated and documented infrastructure for pan-Canadian studies. Comparisons to PopMedNet and the Sentinel Initiative in the US will be made. Conclusion Provincial centres across Canada hold rich sources of health and social data that are linkable at the person-level. With the exception of standardized data managed by the Canadian Institute for Health Information (CIHI), these data are often not comparable from one province to another, thereby limiting use to single-province studies. There is growing interest in Canada in creating an environment that would enable cross-jurisdictional data sharing and analysis’ and in sharing experiences to make effective use of linkable administrative data.

ABSTRACT Objectives Canadians are investing heavily in electronic medical records (EMRs) to inform primary care practice improvements. The Canadian Primary Care Sentinel Surveillance Network (CPCSSN) is a national practice-based network that has enrolled more than one million patients to date. Accurate CPCSSN EMR data are essential for unbiased research about chronic disease prevention and management. The study purpose was to test the accuracy of chronic disease case definitions in EMR data from one CPCSSN site. Approach This study linked CPCSSN EMR data, hospital records, physician billing claims, prescription drug records, and population registration files for the province of Manitoba. Individuals who had at least one encounter with a CPCSSN practice between 1998 and 2012, were at least 18 years of age, and had a minimum of two years of healthcare coverage before and after the study index date were included. Separate cohorts were defined for the following chronic diseases: chronic obstructive pulmonary disease (COPD), depression, diabetes, hypertension, and osteoarthritis. Validated case definitions based on diagnoses in physician and hospital records and prescription drug data were used estimate sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and kappa of each EMR chronic disease case definition. Results More than 74,000 individuals were included in each cohort, except for COPD which had 51,000. Approximately half of each cohort was comprised of urban residents. The average age ranged from 45.9 years for individuals with depression to 65.3 years for individuals with COPD. Hypertension had the highest prevalence (22.0%) in EMR data followed by depression (14.6%). Estimates of agreement (i.e., kappa) for EMR and administrative data ranged from 0.47 for COPD to 0.58 for diabetes. Sensitivity of the EMR data was lowest for COPD (37.4%; 95% CI 36.0-38.8) and highest for diabetes (57.6%; 95% confidence interval [CI] 56.6-58.6). PPV estimates were lowest for osteoarthritis (66.9%; 95% CI 66.0-67.8) and highest for hypertension (78.3%; 95% CI 77.7-78.9). Specificity estimates were consistently above 90% and NPV estimates were always greater than 80%. Validity estimates for the EMR case definitions were associated with demographic and comorbidity characteristics of the study cohorts. Conclusions Validity of EMR data, when compared to administrative health data, for ascertaining five different chronic diseases was fair to good; it varied with the disease under investigation. Further research is needed to identify methods for improving the accuracy of chronic disease case definitions in EMR data.

ABSTRACT Objective Public housing residents, who live in low income government rental housing, are often in poorer health than the rest of the population. However, few studies have been able to untangle the relationships between health and public housing residency, and to assess whether health contributes to the decision to apply. We used linked population-based administrative data from one Canadian province to compare the health and health service use of people who applied to public housing to that of people who did not apply. Approach Administrative data housed in the Manitoba Centre for Health Policy’s Population Health Research Data Repository were used to identify a cohort of individuals who applied to public housing in 2005 and 2006. They were matched one-to-one to a cohort from the general population using socio-demographic variables. A population registry provided demographic and geographic characteristics. Economic measures included receipt of income assistance and an area-level measure from the Statistics Canada Census. Measures of health and health service use were derived from hospital, physician, emergency department, and prescription drug databases. Conditional logistic regression was used to test the association between a public housing application and health status and health service use, after controlling for income. Results There were 10,324 individuals in each of the public housing applicant and matched cohorts; the majority were female (72.4%), young (62% less than 40 years), urban residents (61.2%), and received income assistance (52.8%). A higher percent of the public housing applicant cohort had physician-diagnosed physical and mental health conditions and used more health services compared to the matched cohort. Having a physician-diagnosed respiratory illness (odds ratio [OR] = 1.14, 95% confidence interval [CI] 1.05,1.25), diabetes (OR = 1.24, 95% CI 1.09,1.40), schizophrenia (OR = 1.58, 95% CI 1.30,1.92), affective disorders (OR = 1.37, 95% CI 1.27,1.48), and substance abuse disorders (OR = 1.46, 95% CI 1.25,1.71) were associated with an increased likelihood of applying for public housing, while being diagnosed with cancer (OR = 0.76, 95% CI 0.61,0.96) was associated with a decreased likelihood of applying, after controlling for income differences. High health service users were also more likely to apply for public housing, after controlling for income differences. Conclusion Individuals who move into public housing are in poor health before they apply. Health and social service supports that are co-located with public housing facilities may help to ensure that residents have successful tenancies.

ABSTRACT Objectives An increasing number of physicians are remunerated by alternative forms of payment, instead of conventional fee-for-service (FFS) payments. Changes in physician remuneration methods can to influence the completeness of physician billing claims databases, because physicians on alternative payments may not consistently complete billing records. However, there is no established technique to estimate the magnitude of data loss. This proof-of-concept study estimated completeness of physician claims by comparing them with prescription drug records. We applied the method to estimate completeness of non-fee-for-service (NFFS) and FFS physician claims data over time in Manitoba, Canada. Approach Our method uses information on the date of patient initiation of a new prescription medication, payment method of the prescribing physician, and presence/absence of a physician billing claim prior to the medication initiation date. A billing claim within 7 days of the medication initiation date was defined as a captured claim; if there was no claim in this observation window, it was classified as missed. Our method was applied to annual patient cohorts who initiated a common prescription medication (i.e., anti-hypertensives) between fiscal years 1998/99 and 2012/13. A sensitivity analysis used a 21-day observation window to identify captured/missing claims. Multivariable hierarchical logistic regression models tested patient and prescriber characteristics associated with missing claims. Results The cohort consisted of 274, 462 individuals with a new anti-hypertensive prescription medication. A total of 9.2% of the cohort had a NFFS prescribing physician in 1998/99; this increased to 20.2% in 2012/13 (linear trend p-value < .0001). The percentage of NFFS prescribers almost doubled, from 10.0% to 17.8%. The percentage of the annual cohorts with a FFS prescribing physician and a missing claim remained close to 13.0%. However, the percentage of the annual cohorts with a NFFS prescribing physician and a missing claim increased from 15.6% to 23.3% (linear trend p-value < .0001), and was always higher than the FFS percentage. Patient age, sex, and comorbidity and physician specialty and practice location were associated with the odds of a missing claim. Conclusion The percentage of missing claims was higher for patients with NFFS than FFS prescribing physicians, demonstrating the impact of physician remuneration on database completeness. The trend of greater data loss in later than earlier years suggests that completeness of physician billing claims data may be decreasing. Our method can be applied across jurisdictions to compare the impact of physician payment methods on data quality.

ABSTRACT Objectives The Public Health Agency of Canada (PHAC) established the Canadian Chronic Disease Surveillance System (CCDSS) in 2009 to facilitate national estimates of chronic disease prevalence, incidence, and health outcomes. The CCDSS uses population-based linked health administrative databases from all provinces/territories (P/Ts) and a distributed analytic protocol to produce standardized disease estimates. Our purpose is to describe the process, benefits, and challenges of implementing a distributed model for disease surveillance across thirteen jurisdictions with unique healthcare databases. Approach The CCDSS is founded on deterministic linkage of three administrative health databases in each Canadian P/T: health insurance registration files, physician billing claims, and hospital discharge abstracts. Disease case definitions are developed by expert Working Groups after literature reviews are completed and validation studies are undertaken. Feasibility studies are initiated in selected P/Ts to identify challenges when implementing the disease case definitions. Analytic code developed by PHAC is then distributed to all P/Ts. Data quality surveys are routinely conducted to identify database characteristics that may bias disease estimates over time or across P/Ts or affect implementation of the analytic code. The summary data produced in each P/T are approved by Scientific Committee and Technical Committee members and then submitted to PHAC for further analysis and reporting. Results National surveillance or feasibility studies are currently ongoing for diabetes, hypertension, selected mental illnesses, chronic respiratory diseases, heart disease, neurological conditions, musculoskeletal conditions, and stroke. The advantages of the distributed analytic protocol are: (a) changes in methodology can be easily made, and (b) technical expertise to implement the methodology is not required in each P/T. Challenges in the use of the distributed analytic protocol are: (a) heterogeneity in healthcare databases across P/Ts and over time, (b) the requirement that each P/T use the minimum set of data elements common to all jurisdictions when producing disease estimates, and (c) balancing disclosure guidelines to ensure data confidentiality with comprehensive reporting. Additional challenges, which include incomplete data capture for some databases and poor measurement validity of disease diagnosis codes for some chronic conditions, must be continually addressed to ensure the scientific rigor of the CCDSS methodology. Conclusions The CCDSS distributed analytic protocol offers one model for national chronic disease surveillance that has been successfully implemented and sustained by PHAC and its P/T partners. Many lessons have been learned about national chronic disease surveillance involving jurisdictions that are heterogeneous with respect to healthcare databases, expertise, and population characteristics.

ABSTRACT Objectives Electronic medical records (EMR) can be a cost-effective source for hypertension surveillance. However, diagnosis of hypertension in EMR is commonly under-coded and warrants the needs to review blood pressure and antihypertensive drugs for hypertension case identification. To advocate for the use of EMR data for research, we developed methods for defining hypertension using diagnosis codes, blood pressure measurements and antihypertensive drug prescription Approach We included all the patients actively registered in The Health Improvement Network (THIN) database, UK, on 31 December 2011. Three case definitions using diagnosis code, antihypertensive drug prescriptions and abnormal blood pressure, respectively, were used to identify hypertension patients. We compared the prevalence and treatment rate of hypertension in THIN with results from Health Survey for England (HSE) in 2011. Results Compared with prevalence reported by HSE (29.7%), the use of diagnosis code alone (14.0%) underestimated hypertension prevalence. The use of any of the definitions (38.4%) or the combination of antihypertensive drug prescriptions and abnormal blood pressure (38.4%) had the higher prevalence than HSE. The use of diagnosis code or two abnormal blood pressure records within a 2-year period (31.1%) had similar prevalence and treatment rate of hypertension with HSE. Conclusions Different definitions should be used for different study purposes. The definition of ‘diagnosis code or two abnormal blood pressure records with a 2-year period’ could be used for hypertension surveillance in THIN.