Background:Adverse events (AEs) are often measured using administrative data, which has validity limitations. Electronic Medical Records (EMRs) provide rich, unstructured data that can improve AE detection through data science methods. Objective: To evaluate the accuracy of EMR-based algorithms for AE detection and estimate AE prevalence. Methods:We conducted a retrospective chart review validation study at a tertiary-care hospital (population ∼1.5M). Adult patients admitted between 2017-2022 with discharge summaries or narrative EMRs were included; only the most recent admission was analyzed. Seventeen algorithms were developed for specific AE categories and an overall "any AE" category using natural language processing (NLP) and machine learning. Prevalence, sensitivity, specificity, PPV, NPV, and model performance (AUC, accuracy, F1 score) were calculated using chart review as a reference. Results:Of 10,939 charts, 10,659 were linked to discharge data; 2,069 patients had ≥1 AE. AE prevalence ranged from 0.2%-6.0%. Sensitivity and specificity varied (57%-100%, 74%-96%). For "any AE," sensitivity was 63%, specificity 94%, PPV 72%, NPV 91%. Overall algorithm performance: AUC 78.4% (95% CI, 76.0-80.8), accuracy 87.6%, F1 score 67.2. Conclusion:NLP-based algorithms applied to EMRs show strong performance despite low AE prevalence, offering a scalable approach to enhance AE surveillance and patient safety.

Objective:To develop and evaluate a Transformer-based temporal representation learning framework for predicting incident diabetes using longitudinal laboratory data, and to compare its performance with state-of-the-art pretrained models and traditional machine-learning baselines. Approach:We deterministically linked a cardiac registry cohort (2015-2019) in Alberta, Canada, and retrieved three years of laboratory tests preceding each patient's diabetes diagnosis. Each laboratory event was modelled as a triplet consisting of test type, test value, and time gap with the previous test, to preserve temporal structure. Our Transformer architecture learned contextual embeddings that capture both intra-test dynamics and inter-test interactions, indicating the onset of diabetes. Performance was benchmarked against two pretrained Transformer models (Moment, PatchTST) and an XGBoost baseline trained on aggregated test summaries (mean, standard deviation, minimum, maximum). Results:The final cohort included 30,462 patients and 19 laboratory test types relevant to diabetes. The proposed method achieved the highest predictive accuracy (AUC = 0.92), outperforming Moment (0.80), PatchTST (0.74), and XGBoost (0.88). It also demonstrated superior sensitivity (0.70) and positive predictive value (0.777) while maintaining high specificity (0.938) and negative predictive value (0.911). Conclusion: Modelling raw laboratory trajectories with a dedicated temporal Transformer substantially improves diabetes prediction compared with both pretrained sequence models and aggregation-based baselines. Implications:This framework provides a generalizable pathway for leveraging routine laboratory data in early disease detection. It highlights the clinical value of sequence-level modelling and supports the integration of temporal representation learning into population-level surveillance and decision-support systems.

Background: Patient Safety Indicators (PSIs) derived from administrative data are widely used for monitoring and improving hospital care quality. However, the validity of ICD-10-based PSI algorithms remains uncertain, particularly in terms of their sensitivity and specificity. Objectives: To evaluate the diagnostic performance of ICD-10-CA-based algorithms for identifying fifteen PSIs using chart review as the reference standard. Research Design: Multicenter retrospective cohort validation study. Subjects: A random sample of 10,665 adult patients admitted to 4 acute care hospitals in Calgary, Alberta, between January 1, 2017, and March 31, 2022.Measures:Fifteen PSIs were identified using ICD-10-CA codes and validated against detailed chart reviews. Diagnostic performance was measured using sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall accuracy. Analyses were stratified by diagnosis code type and relevant patient characteristics. Results: Among 10,665 patients, 1688 had at least one PSI confirmed by chart review. ICD-10-CA coding detected any PSI with 67.0% sensitivity (95% CI, 64.7%-69.2%), 72.8% specificity (95% CI, 71.8%-73.7%), 31.6% PPV (95% CI, 30.1%-33.1%), 92.2% NPV (95% CI, 91.5%-92.8%), and 71.8% accuracy (95% CI, 71.0%-72.7%). Restricting PSIs to conditions that occurred after admission (limited diagnosis type II code) improved specificity (95.7%; 95% CI, 95.3%-96.1%) and PPV (56.5%; 95% CI, 53.2%-59.7%) but reduced sensitivity (29.6%; 95% CI, 27.4%-31.8%). Validity varied by PSI and patient characteristics, with higher sensitivity and PPV among older adults, males, and those with greater comorbidity, longer hospital and ICU stays, 30-day readmission, or in-hospital death. Conclusions :ICD-10 coded administrative data demonstrate high specificity and NPV but varied sensitivity and PPV in identifying PSIs. Restricting to type II codes improves PPV but reduces sensitivity. Tailoring coding strategies to specific surveillance or quality improvement goals is critical.

OBJECTIVE:To assess how family physicians use and perceive the International Classification of Diseases, 11th Revision (ICD-11) and the International Classification of Primary Care, Version 3 (ICPC-3) compared to the ICD-9. DESIGN:Cross-sectional, pre-implementation study. SETTING:Canada. PARTICIPANTS:Family physicians. MAIN OUTCOME MEASURES:This mixed-methods study used an online patient vignette coding exercise and postcoding survey. Time taken to code and coding consistency were assessed and determined using the coding exercise. Physician-reported satisfaction measures and perspectives related to the 3 systems were evaluated using the postcoding survey. RESULTS:A total of 161 family physicians participated in this study. Participants coded significantly faster with the ICPC-3 (median=48 seconds, interquartile range [IQR]=64 seconds), compared to the ICD-11 (median=59 seconds, IQR=71 seconds) and the ICD-9 (median=57 seconds, IQR=75 seconds). There were no differences in coding consistency based on the coding system used. More participants rated both the ICD-11 and ICPC-3 positively compared to the ICD-9 with respect to comprehensiveness, ability to find codes, precision, ease of use, ability to capture necessary information, and ability to capture medical and social complexity. CONCLUSION:This study demonstrated that the alternative coding systems, ICD-11 and ICPC-3, were associated with better utility and a more positive user experience among family physicians compared to the ICD-9. These study findings can be used as a foundation for future discussions toward an improved coding system in Canada.

Objective Large language models (LLMs) have the potential to revolutionize how population-level health research is conducted by automatically abstracting data that would otherwise be unavailable. However, few results are available on real clinical notes. We developed an LLM leaderboard showing how open-source LLMs perform at entity extraction on unseen clinical notes. Approach EMR data, including free-text notes, were linked to a chart-review cohort comprising 10,659 adults admitted to a hospital in Calgary, Canada, between 2017 and 2022, with data on comorbidities. We then attempted to replicate this chart review with multiple open-source LLMs in a secure computing environment. Chart review results served as the reference standard. Results There was a wide variation in performance among the examined LLMs: the smallest, Llama 3.2 3B, had a high mean sensitivity of 0.97 but low PPV at 0.4; the largest, Llama-3-70B, showed a high mean sensitivity of 0.96 and greatly improved PPV of 0.7; in-between these in size, phi 4 demonstrated a more balanced performance with a mean sensitivity of 0.81 and PPV of 0.77. However, the results varied considerably across conditions, with quirks specific to each model. Conclusions LLMs are already available that can perform entity extraction well enough to be considered in place of some administrative data. With rapid developments in the field, a leaderboard based on real clinical data is vital for informing researchers on best practices for integrating the latest AI techniques into their data practices.

Reducing prolonged sedentary behaviour is recommended to improve population health. The development of behaviour change interventions requires understanding context-specific determinants, which may differ across domains of transportation, occupational, and leisure time sedentary behaviour. This study aimed to use the Behaviour Change Wheel to inform the development of domain-specific intervention strategies for reducing prolonged sedentary behaviour among Canadian adults. One-on-one semi-structured interviews were conducted with a diverse sample of adult participants from across Canada. Determinants of reducing prolonged sedentary behaviour were identified through thematic analysis and characterized according to their relevance to domains of transportation, occupational, and leisure time sedentary behaviour. The Behaviour Change Wheel and Behaviour Change Techniques Taxonomy (BCTTv1) were used to identify intervention strategies, components, and modes of delivery specific to and across behavioural domains. Thirty participants were interviewed (63

Leveraging routinely collected health services data to conduct scientifically robust research requires a clear understanding and standardization of data from disparate clinical information systems which often change over time. Within Alberta Provincial Research Data Services (PRDS) a large team of analysts utilize a vast Enterprise Data Warehouse containing dozens of datasets with various update cycles and availability, generally spanning more than 20 years of data for a Canadian province of over 4 million people. We examine two recently developed standardized data marts for mortality and medical lab tests. These data marts are now updated daily using Snowflake Tasks to ensure timely data can be accessed as needed to support research studies. The mortality data are linked from three sources to ensure comprehensive coverage: Vital Statistics, Provincial Registry, and Connect Care (the provincial Epic based electronic medical record) and resolves inconsistencies in personal health numbers and discrepancies in dates of death. The Lab data mart uses test level data from several different historical lab systems and the current Connect Care system and consolidates coding and definitions over time to create a clean, easy to use data mart. Both use cases create a single, validated and linkable source of truth that is easy to use for research studies, negating the need for each analyst to re-engineer from several different source tables each time. The process used here will be applied to other subject matters internally and could be similarly applied data warehouse environments in other jurisdictions.

BACKGROUND:While the incidence of hospital adverse events appeared to be declining before 2019, the COVID-19 pandemic may have changed its course. This study aimed to evaluate adverse event incidence rates and trends during the pandemic and analyse differences in patient outcomes. METHODS:This retrospective electronic chart review included a random sample of adult patients admitted to four acute care hospitals in Calgary between 2017 and 2022. 18 adverse events and patient information were extracted. We calculated the observed and risk-standardised incidence rates of adverse events. Interrupted time series analysis was employed to determine the impact of COVID-19 on adverse events trends. Outcome differences were evaluated using mixed-effects logistic regression and negative binomial models. RESULTS:Among 10 673 patient admissions, 2310 adverse events were identified, resulting in an incidence rate of 21.64 (95% CI 20.77 to 22.54) per 100 patient admissions, or 26.85 (95% CI 25.77 to 27.97) per 1000 patient days. After adjusting for patient characteristics, seasonal variations and overall trends, the adverse event incidence rate increased by 14% (incidence rate ratio (IRR) 1.14, 95% CI 1.01 to 1.29) during the COVID-19 pandemic. In multivariable mixed-effects models, adverse events were associated with significantly longer hospital stays (IRR 3.13, 95% CI 2.97 to 3.30), increased odds of 30-day readmission (OR 1.4, 95% CI 1.17 to 1.68) and in-hospital death (OR 1.72, 95% CI 1.43 to 2.08). CONCLUSION:The incidence of adverse events was high but relatively stable in acute healthcare settings before the COVID-19 pandemic and increased during the pandemic. Strengthening healthcare resilience and prioritising patient safety initiatives are crucial as we transition into the post-pandemic era.

Abstract Importance Heart failure readmissions remain common following hospitalization, but accurately identifying which patients will be readmitted after discharge remains challenging. Improved prediction could support targeted transitional care interventions and more efficient allocation of clinical resources. Objective In this study we attempted to improve readmission prediction after heart failure hospitalization by using variables chosen through a modified Delphi process, and using inpatient Electronic Medical Record (EMR) data, focusing on clinical notes. Design This prognostic study developed competing risk survival models to predict readmission after heart failure hospitalization. Variables were chosen using a modified Delphi process, and extracted from EMR notes using various natural language processing techniques or from other EMR elements where appropriate. Patients were admitted between 2011 through 2019, and at least one year of follow-up was available for all patients. Models were evaluated using C-statistics, as well as sensitivity, specificity, positive and negative predictive values. Setting During the study period, all acute-care facilities in Calgary, Alberta used the same EMR system, from which patients were selected. Participants Patients were 18 years or older, resided in Alberta, and were admitted to a Calgary hospital. All corresponding admissions with a most responsible diagnosis of heart failure were included (n=15,160). Main Outcomes and Measures The main outcome of interest was readmission within 30 days, though 90– and 365-day time frames were also analyzed. Death was treated as a competing risk and analysed at those time frames as well. Results Many of the identified variables believed to be important for predicting readmission were not collected reliably enough to be used in readmission models. The final model had a C-statistic of 64.4 when predicting readmission within 30 days, in line with previous studies. Conclusions and Relevance Efforts to improve heart failure readmission prediction should work with clinical teams to ensure variables believed to be important are collected during hospitalization.

Background Accurate detection of inpatient depression is essential for surveillance and resource allocation. Traditional International Classification of Diseases (ICD) codes suffer from low sensitivity and delayed availability. Objective To develop and validate machine learning (ML) phenotyping algorithms for depression using unstructured inpatient clinical notes from electronic medical records (EMRs), compared with ICD-10-CA codes and manual chart review. Methods This retrospective cohort study included a derivation cohort (N=3032 adults, Jan–June 2015) and a temporally independent external validation cohort (N=10659 adults, 2017–2022) from Calgary acute care hospitals. Both cohorts were deterministically linked to DAD and EMR databases. Manual chart reviews defined the reference standard for depression status. We compared ICD-10-CA codes against three EMR-based algorithms: Keyword Search, Concept Model and Document-Concept Model. The latter two employed a supervised ML classifier (XGBoost) utilizing clinical concepts extracted via the cTAKES natural language processing (NLP) pipeline. Primary measures were prevalence, sensitivity, positive predictive value (PPV), specificity, negative predictive value (NPV), and F1 score. Results In derivation cohort (prevalence: 18.1%), the Concept Model achieved 76.4% sensitivity and 76.7% F1 score, significantly outperforming Keyword Search (F1: 66.7%), Document-Concept Model (F1: 73.7%), and ICD-10-CA codes (sensitivity: 20.0%; F1: 32.6%). In external validation (prevalence: 13.1%), the Concept Model retained robust performance (sensitivity: 92.3%; F1: 80.1%), compared with ICD-10-CA (sensitivity: 6.1%; F1: 11.4%). Conclusion An EMR-based Concept Model using NLP and ML markedly improves depression case detection over ICD coding in inpatient settings. The methods could potentially facilitate psychiatric surveillance and epidemiological research using real-world clinical data.

BACKGROUND AND OBJECTIVE:Coding by human coders is burdensome to healthcare systems requiring advanced computerized techniques. Epic has collaborated with Intelligent Medical Objects (IMO) solution to integrate a solution where the clinical interface terminology is mapped to the International Classification of Diseases (ICD). This study assesses the agreement between the solution mapped ICD-10-CA codes in Alberta, Canada's EHR system (based on Epic) to the human coded ICD-10-CA codes in hospital Discharge Abstract Database (DAD). DESIGN AND SETTING:In this retrospective analysis conducted in Alberta, Canada, we linked records in the acute care hospital DAD with the province wide EHR system for admissions between April 2021 and March 2024. MAIN OUTCOME(S) AND MEASURE(S):The primary outcome was the level of agreement between the solution mapped ICD-10-CA codes from the 'hospital problem list' in the EHR and hospital DAD data. The analysis was conducted at 3-digit and 4-digit ICD code level for main diagnosis and any diagnosis, and further stratified by physician specialty, hospital type and location, and length-of-stay. RESULTS:A total of 603,437 unique hospital records were linked between hospital DAD and EHR. The average level of agreement at 3-digit level of ICD-10-CA code for main diagnosis was 47.5% and any diagnosis was 37.0%. The average number of diagnoses coded by human coders in hospital DAD was higher than the solution mapped data in EHR. The agreement varied by specialty and length-of-stay with specialties with more complex patients and longer stays showing the lowest levels of agreement. CONCLUSION:Level of agreement between solution mapped EHR and hospital DAD for ICD-10-CA data was low, indicating significant differences between terminology mappings and the coding process.

Comorbidity measures, such as the Charlson Comorbidity Index, are commonly used in risk adjustment models to account for variability in disease burden. This narrative synthesis describes and critiques available comorbidity indices and offers implementation guidance to researchers based on a critical review of existing literature. First, common comorbidity measures are described. Instruments derived using case definitions, grouping of International Classification of Diseases (ICD) codes, and mapping of dispensed medications to chronic conditions are presented. Comorbidity indices that combine diagnostic and medication data are also introduced. No single option consistently outperforms the rest. Next, important considerations when applying a comorbidity index are described. It is crucial to respect temporality and exclude health events that arise after the study index date. Researchers must also weigh the interpretability of using a weighted sum against the flexibility of using a large set of binary variables. When modelling long-term outcomes, there are benefits to applying a one-year look-back window and augmenting data via linkage. For short-term outcomes, certain chronic conditions may exhibit a protective association; however, not all indices capture these relationships. Implementation of these findings will improve the interpretability of comorbidity measures and the quality of future studies.

Background Heart failure (HF) readmission rates have been a significant concern for healthcare systems globally. Accurate predictive models are essential to identify patients at high readmission risk and implement timely interventions. Current models often lack comprehensive variables that reflect both clinical and patient and/or caregiver perspectives. We aimed to develop a consensus-driven approach to identify essential variables for inclusion in HF hospital readmission risk prediction algorithms. Methods A Delphi panel comprised of clinicians and patient and/or caregiver partners was assembled. The Delphi panelists were recruited from the province of Alberta, Canada. The panel consisted of 13 individuals, including 9 healthcare providers and 4 patients and/or caregivers. The review panel was provided with a list of variables from a previously completed systematic literature review. Three rounds were conducted. The panel also determined the directionality of the association. Results A total of 99 variables were identified through literature and physician input. Panelists reached a consensus on 61 variables, which were deemed to be associated with the risk of readmission for any cause within 30 days of discharge after HF hospitalization. Clinician ratings on consensus were consistently higher than those of nonclinicians. Conclusions This study successfully identified 61 variables associated with HF readmission risk through a modified Delphi process, incorporating both clinician and patient and/or caregiver perspectives. These findings provide a foundation for future research and the development of more comprehensive and accurate risk prediction models. Including diverse stakeholder input highlights the importance of integrating medical expertise and patient experiences in improving HF management and reducing readmission rates.

Sepsis is a major global health problem. However, it lacks a true reference standard for case identification, complicating epidemiologic surveillance. Consensus definitions have changed multiple times, clinicians struggle to identify sepsis at the bedside, and differing identification algorithms generate wide variation in incidence rates. The two current identification approaches use codes from administrative data, or electronic health record (EHR)-based algorithms such as the Center for Disease Control Adult Sepsis Event (ASE); both have limitations. Here our primary purpose is to report initial steps in developing a novel approach to identifying sepsis using unsupervised clustering methods. Secondarily, we report preliminary analysis of resulting clusters, using identification by ASE criteria as a familiar comparator. This retrospective cohort study used hospital administrative and EHR data on adults admitted to intensive care units (ICUs) at five Canadian medical centres (2015–2017), with split development and validation cohorts. After preprocessing 592 variables (demographics, encounter characteristics, diagnoses, medications, laboratory tests, and clinical management) and applying data reduction, we presented 55 principal components to eight different clustering algorithms. An automated elbow method determined the optimal number of clusters, and the optimal algorithm was selected based on clustering metrics for consistency, separation, distribution and stability. Cluster membership in the validation cohort was assigned using an XGBoost model trained to predict cluster membership in the development cohort. For cluster analysis, we prospectively subdivided clusters by their fractions meeting ASE criteria (≥ 50

OBJECTIVE:Electronic health records (EHR) are widely available to complement administrative data-based disease surveillance and healthcare performance evaluation. Defining conditions from EHR is labour-intensive and requires extensive manual labelling of disease outcomes. This study developed an efficient strategy based on advanced large language models to identify multiple conditions from EHR clinical notes. METHODS:We linked a cardiac registry cohort in 2015 with an EHR system in Alberta, Canada. We developed a pipeline that leveraged a generative large language model (LLM) to analyze, understand, and interpret EHR notes by prompts based on specific diagnosis, treatment management, and clinical guidelines. The pipeline was applied to detect acute myocardial infarction (AMI), diabetes, and hypertension. The performance was compared against clinician-validated diagnoses as the reference standard and widely adopted International Classification of Diseases (ICD) codes-based methods. RESULTS:The study cohort accounted for 3088 patients and 551,095 clinical notes. The prevalence was 55.4 %, 27.7 %, 65.9 % and for AMI, diabetes, and hypertension, respectively. The performance of the LLM-based pipeline for detecting conditions varied: AMI had 88 % sensitivity, 63 % specificity, and 77 % positive predictive value (PPV); diabetes had 91 % sensitivity, 86 % specificity, and 71 % PPV; and hypertension had 94 % sensitivity, 32 % specificity, and 72 % PPV. Compared with ICD codes, the LLM-based method demonstrated improved sensitivity and negative predictive value across all conditions. The monthly percentage trends from the detected cases by LLM and reference standard showed consistent patterns. CONCLUSION:The proposed LLM-based pipeline demonstrated reasonable accuracy and high efficiency in disease detection for multiple conditions. Human expert knowledge can be integrated into the pipeline to guide EHR note analysis without manually curated labels. The method could enable comprehensive real-time disease surveillance using EHRs.

OBJECTIVES:Administrative data are commonly used to inform chronic disease prevalence and support health informatic research. This study assessed the validity of coding comorbidities in the International Classification of Diseases, 10th Revision (ICD-10) administrative data. METHODS:We analysed three chart review cohorts (4008 patients in 2003, 3045 in 2015 and 9024 in 2022) in Alberta, Canada. Nurse reviewers assessed the presence of 17 clinical conditions using a consistent protocol. The reviews were linked with administrative data using unique patient identifiers. We compared the accuracy in coding comorbidity by ICD-10, using chart review data as the reference standard. RESULTS:Our findings showed that the mean difference in prevalence between chart reviews and ICD-10 for these 17 conditions was 2.1% in 2003, 7.6% in 2015 and 6.3% in 2022. Some conditions were relatively stable, such as diabetes (1.9%, 2.1% and 1.1%) and metastatic cancer (0.3%, 1.1% and 0.4%). For these 17 conditions, the sensitivity ranged from 39.6-85.1% in 2003, 1.3%-85.2% in 2015 and 3.0-89.7% in 2022. The C-statistics for predicting in-hospital mortality using comorbidities by ICD-10 were 0.84 in 2003, 0.81 in 2015 and 0.78 in 2022. DISCUSSION:The undercoding could be primarily due to the increase in hospital patient volumes and the limited time allocated to coding specialists. There is the potential to develop artificial intelligence methods based on electronic health records to support coding practices and improve data quality. CONCLUSION:Comorbidities were increasingly undercoded over 20 years. The validity of ICD-10 decreased but remained relatively stable for certain conditions mandated for coding. The undercoding exerted minimal impact on in-hospital mortality prediction.

This environmental scan aims to identify and describe initiatives implemented in Alberta, one of the few Canadian provinces with a unified health delivery system, to reduce heart failure (HF)readmissions. It also acknowledges the challenges in attributing direct benefits to these interventions. Using snowball sampling, we identified and recruited 11 employees and clinicians from Alberta Health Services (AHS) who possessed significant historical institutional knowledge about HF. Academic and grey literature were reviewed related to Alberta's readmission reduction initiatives and reported outcomes. Unstructured, in-depth interviews were conducted to clarify timelines and provide detailed descriptions of these interventions. Our findings indicate substantial clinician efforts over 15 years to address all-cause readmissions post HF hospitalization in Alberta, encompassing a range of interventions from small-scale projects to large multi-city, multi-stakeholder initiatives. Assessing the impact of smaller interventions on provincial readmission rates proved challenging; however, five major initiatives collectively led to a 1.8% reduction in 30-day all-cause readmissions province-wide (from 22.2% to 20.4%, p=0.04). Key factors that appeared to support these efforts included utilization of the EMR system, stakeholder engagement in standardized care, effective communication practices, and appropriate resource allocation. Clinical teams are now integrating successful components from these initiatives into the province-wide clinical information system, Connect Care, to enhance care coordination and patient outcomes. This environmental scan highlights various comprehensive initiatives in Alberta aimed at improving patient care and reducing readmission after HF hospitalization. While an overall 1.8% reduction in readmission rates was observed over the 15 years, attributing this change directly to the interventions is challenging due to various implementation barriers and the complexity of healthcare delivery. Continued efforts towards personalized care and innovative EMR utilization hold promise for further improvement in HF readmission rates. Key Words: Heart failure, readmissions, environmental scan, clinical pathways, health outcomes, patient-centered care ### Competing Interest Statement The authors have declared no competing interest. ### Funding Statement This study did not receive any funding ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: Informed consent was obtained prior to each interview. Ethics was obtained from the Conjoint Health Research Ethics Board REB20-0684 I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes Data cannot be shared publicly because to respect the privacy of the participants. Data are available from the University of Calgary Institutional Data Access / Ethics Committee (contact via cfreb{at}ucalgary.ca) for researchers who meet the criteria for access to confidential data.

Countries have been routinely abstracting health data from hospital charts and coding conditions using ICD-10. A main condition must be assigned to each admission. However, the definition of main condition is inconsistent across countries, and may be based on (1) the initial reason for admission; (2) the reason for admission, as understood at the end of the hospital stay; and (3) the condition that consumed the most hospital resources or hospital days. Now, ICD-11 standardizes the coding schema for main condition. This paper describes the ICD-11 coding guidelines for main condition and discusses their implications for data comparability.

OBJECTIVE:We estimated incidence rates of avoidable hospitalizations by persons with rheumatoid arthritis (RA) relative to the general population. METHODS:We identified individuals meeting a validated case definition for RA based on International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) and ICD-10-CA codes in years 2002 to 2023. Four general population controls were matched to each RA case by age and sex. We identified hospitalizations for ambulatory care sensitive conditions (ACSCs), including grand mal seizures, chronic lower respiratory diseases, asthma, diabetes, heart failure and pulmonary edema, hypertension, and angina, from 2007 to 2023 by established diagnostic codes. Incidence rate ratios (IRRs) three and five years from the date of diagnosis were calculated using a multivariable regression model adjusting for age, sex, and location of residence. A Cox proportional hazards model was used to identify predictors of avoidable hospitalizations among patients with RA. RESULTS:Persons with RA (n = 83,811) had 1.12 times the risk of hospitalization for heart failure and pulmonary edema compared to those without RA (n = 190,304) (IRR 1.12, 95% confidence interval [CI] 1.01-1.25). Significant predictors of ACSC hospitalizations for RA cases were increasing age, prolonged exposure to glucocorticoids, and having comorbid conditions, especially if the comorbid condition was an ACSC (hazard ratio 10.1, 95% CI 7.8-13.0). CONCLUSION:Persons with RA are at a higher risk of potentially avoidable hospitalizations three and five years after diagnosis compared to those without RA. Improved ambulatory care access and quality, inclusive of primary care and subspecialty care, is proposed to prevent unnecessary hospitalizations and reduce burden on the acute care system.