BERT-Based Neural Network for Inpatient Fall Detection From Electronic Medical Records: Retrospective Cohort Study.

BACKGROUND: Inpatient falls are a substantial concern for health care providers and are associated with negative outcomes for patients. Automated detection of falls using machine learning (ML) algorithms may aid in improving patient safety and reducing the occurrence of falls. OBJECTIVE: This study aims to develop and evaluate an ML algorithm for inpatient fall detection using multidisciplinary progress record notes and a pretrained Bidirectional Encoder Representation from Transformers (BERT) language model. METHODS: A cohort of 4323 adult patients admitted to 3 acute care hospitals in Calgary, Alberta, Canada from 2016 to 2021 were randomly sampled. Trained reviewers determined falls from patient charts, which were linked to electronic medical records and administrative data. The BERT-based language model was pretrained on clinical notes, and a fall detection algorithm was developed based on a neural network binary classification architecture. RESULTS: To address various use scenarios, we developed 3 different Alberta hospital notes-specific BERT models: a high sensitivity model (sensitivity 97.7, IQR 87.7-99.9), a high positive predictive value model (positive predictive value 85.7, IQR 57.2-98.2), and the high F1-score model (F1=64.4). Our proposed method outperformed 3 classical ML algorithms and an International Classification of Diseases code-based algorithm for fall detection, showing its potential for improved performance in diverse clinical settings. CONCLUSIONS: The developed algorithm provides an automated and accurate method for inpatient fall detection using multidisciplinary progress record notes and a pretrained BERT language model. This method could be implemented in clinical practice to improve patient safety and reduce the occurrence of falls in hospitals.

Automated extraction of weight, height, and obesity in electronic medical records are highly valid.

Objective: Coding of obesity using the International Classification of Diseases (ICD) in healthcare administrative databases is under-reported and thus unreliable for measuring prevalence or incidence. This study aimed to develop and test a rule-based algorithm for automating the detection and severity of obesity using height and weight collected in several sections of the Electronic Medical Records (EMRs). Methods: In this cross-sectional study, 1904 inpatient charts randomly selected in three hospitals in Calgary, Canada between January and June 2015 were reviewed and linked with AllScripts Sunrise Clinical Manager EMRs. A rule-based algorithm was created which looks for patients' height and weight values recorded in EMRs. Clinical notes were split into sentences and searched for height and weight, and BMI was computed. Results: The study cohort consisted of 1904 patients with 50.8% females and 43.3% > 64 years of age. The final model to identify obesity within EMRs resulted in a sensitivity of 92.9%, specificity of 98.4%, positive predictive value of 96.7%, negative predictive value of 96.6%, and F1 score of 94.8%. Conclusions: This study developed a highly valid rule-based EMR algorithm that detects height and weight. This could allow large-scale analyses using obesity that were previously not possible.

Exploring the reliability of inpatient EMR algorithms for diabetes identification.

INTRODUCTION: Accurate identification of medical conditions within a real-time inpatient setting is crucial for health systems. Current inpatient comorbidity algorithms rely on integrating various sources of administrative data, but at times, there is a considerable lag in obtaining and linking these data. Our study objective was to develop electronic medical records (EMR) data-based inpatient diabetes phenotyping algorithms. MATERIALS AND METHODS: A chart review on 3040 individuals was completed, and 583 had diabetes. We linked EMR data on these individuals to the International Classification of Disease (ICD) administrative databases. The following EMR-data-based diabetes algorithms were developed: (1) laboratory data, (2) medication data, (3) laboratory and medications data, (4) diabetes concept keywords and (5) diabetes free-text algorithm. Combined algorithms used or statements between the above algorithms. Algorithm performances were measured using chart review as a gold standard. We determined the best-performing algorithm as the one that showed the high performance of sensitivity (SN), and positive predictive value (PPV). RESULTS: The algorithms tested generally performed well: ICD-coded data, SN 0.84, specificity (SP) 0.98, PPV 0.93 and negative predictive value (NPV) 0.96; medication and laboratory algorithm, SN 0.90, SP 0.95, PPV 0.80 and NPV 0.97; all document types algorithm, SN 0.95, SP 0.98, PPV 0.94 and NPV 0.99. DISCUSSION: Free-text data-based diabetes algorithm can yield comparable or superior performance to a commonly used ICD-coded algorithm and could supplement existing methods. These types of inpatient EMR-based algorithms for case identification may become a key method for timely resource planning and care delivery.

Use of Epic Electronic Health Record System for Health Care Research: Scoping Review.

BACKGROUND: Electronic health records (EHRs) enable health data exchange across interconnected systems from varied settings. Epic is among the 5 leading EHR providers and is the most adopted EHR system across the globe. Despite its global reach, there is a gap in the literature detailing how EHR systems such as Epic have been used for health care research. OBJECTIVE: The objective of this scoping review is to synthesize the available literature on use cases of the Epic EHR for research in various areas of clinical and health sciences. METHODS: We used established scoping review methods and searched 9 major information repositories, including databases and gray literature sources. To categorize the research data, we developed detailed criteria for 5 major research domains to present the results. RESULTS: We present a comprehensive picture of the method types in 5 research domains. A total of 4669 articles were screened by 2 independent reviewers at each stage, while 206 articles were abstracted. Most studies were from the United States, with a sharp increase in volume from the year 2015 onwards. Most articles focused on clinical care, health services research and clinical decision support. Among research designs, most studies used longitudinal designs, followed by interventional studies implemented at single sites in adult populations. Important facilitators and barriers to the use of Epic and EHRs in general were identified. Important lessons to the use of Epic and other EHRs for research purposes were also synthesized. CONCLUSIONS: The Epic EHR provides a wide variety of functions that are helpful toward research in several domains, including clinical and population health, quality improvement, and the development of clinical decision support tools. As Epic is reported to be the most globally adopted EHR, researchers can take advantage of its various system features, including pooled data, integration of modules and developing decision support tools. Such research opportunities afforded by the system can contribute to improving quality of care, building health system efficiencies, and conducting population-level studies. Although this review is limited to the Epic EHR system, the larger lessons are generalizable to other EHRs.

Development of machine learning models for the detection of surgical site infections following total hip and knee arthroplasty: a multicenter cohort study.

BACKGROUND: Population based surveillance of surgical site infections (SSIs) requires precise case-finding strategies. We sought to develop and validate machine learning models to automate the process of complex (deep incisional/organ space) SSIs case detection. METHODS: This retrospective cohort study included adult patients (age ≥ 18 years) admitted to Calgary, Canada acute care hospitals who underwent primary total elective hip (THA) or knee (TKA) arthroplasty between Jan 1st, 2013 and Aug 31st, 2020. True SSI conditions were judged by the Alberta Health Services Infection Prevention and Control (IPC) program staff. Using the IPC cases as labels, we developed and validated nine XGBoost models to identify deep incisional SSIs, organ space SSIs and complex SSIs using administrative data, electronic medical records (EMR) free text data, and both. The performance of machine learning models was assessed by sensitivity, specificity, positive predictive value, negative predictive value, F1 score, the area under the receiver operating characteristic curve (ROC AUC) and the area under the precision-recall curve (PR AUC). In addition, a bootstrap 95% confidence interval (95% CI) was calculated. RESULTS: There were 22,059 unique patients with 27,360 hospital admissions resulting in 88,351 days of hospital stay. This included 16,561 (60.5%) TKA and 10,799 (39.5%) THA procedures. There were 235 ascertained SSIs. Of them, 77 (32.8%) were superficial incisional SSIs, 57 (24.3%) were deep incisional SSIs, and 101 (42.9%) were organ space SSIs. The incidence rates were 0.37 for superficial incisional SSIs, 0.21 for deep incisional SSIs, 0.37 for organ space and 0.58 for complex SSIs per 100 surgical procedures, respectively. The optimal XGBoost models using administrative data and text data combined achieved a ROC AUC of 0.906 (95% CI 0.835-0.978), PR AUC of 0.637 (95% CI 0.528-0.746), and F1 score of 0.79 (0.67-0.90). CONCLUSIONS: Our findings suggest machine learning models derived from administrative data and EMR text data achieved high performance and can be used to automate the detection of complex SSIs.

The incidence rates of surgical site infections following total hip and knee arthroplasty were 0.5 and 0.52 per 100 surgical procedures. The incidence of SSIs varied significantly between care facilities (ranging from 0.53 to 1.71 per 100 procedures). The optimal machine learning model achieved a ROC AUC of 0.906 (95% CI 0.835­0.978), PR AUC of 0.637 (95% CI 0.528­0.746), and F1 score of 0.79 (0.67­0.90).

Hypertension identification using inpatient clinical notes from electronic medical records: an explainable, data-driven algorithm study.

BACKGROUND: Case identification is important for health services research, measuring health system performance and risk adjustment, but existing methods based on manual chart review or diagnosis codes can be expensive, time consuming or of limited validity. We aimed to develop a hypertension case definition in electronic medical records (EMRs) for inpatient clinical notes using machine learning. METHODS: A cohort of patients 18 years of age or older who were discharged from 1 of 3 Calgary acute care facilities (1 academic hospital and 2 community hospitals) between Jan. 1 and June 30, 2015, were randomly selected, and we compared the performance of EMR phenotype algorithms developed using machine learning with an algorithm based on the Canadian version of the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD), in identifying patients with hypertension. Hypertension status was determined by chart review, the machine-learning algorithms used EMR notes and the ICD algorithm used the Discharge Abstract Database (Canadian Institute for Health Information). RESULTS: Of our study sample (n = 3040), 1475 (48.5%) patients had hypertension. The group with hypertension was older (median age of 71.0 yr v. 52.5 yr for those patients without hypertension) and had fewer females (710 [48.2%] v. 764 [52.3%]). Our final EMR-based models had higher sensitivity than the ICD algorithm (> 90% v. 47%), while maintaining high positive predictive values (> 90% v. 97%). INTERPRETATION: We found that hypertension tends to have clear documentation in EMRs and is well classified by concept search on free text. Machine learning can provide insights into how and where conditions are documented in EMRs and suggest nonmachine-learning phenotypes to implement.

Training and experience of coding with the World Health Organization's International Classification of Diseases, Eleventh Revision.

BACKGROUND: The new International Classification of Diseases, Eleventh Revision for Mortality and Morbidity Statistics (ICD-11) was developed and released by the World Health Organization (WHO) in June 2018. Because ICD-11 incorporates new codes and features, training materials for coding with ICD-11 are urgently needed prior to its implementation. OBJECTIVE: This study outlines the development of ICD-11 training materials, training processes and experiences of clinical coders while learning to code using ICD-11. METHOD: Six certified clinical coders were recruited to code inpatient charts using ICD-11. Training materials were developed with input from experts from the Canadian Institute for Health Information and the WHO, and the clinical coders were trained to use the new classification. Monthly team meetings were conducted to enable discussions on coding issues and to select the correct ICD-11 codes. The training experience was evaluated using qualitative interviews, a questionnaire and a coding quiz. RESULTS: total of 3011 charts were coded using ICD-11. In general, clinical coders provided positive feedback regarding the training program. The average score for the coding quiz (multiple choice, True/False) was 84%, suggesting that the training program was effective. Feedback from the coders enabled the ICD-11 code content, electronic tooling and terminologies to be updated. CONCLUSION: This study provides a detailed account of the processes involved with training clinical coders to use ICD-11. Important findings from the interviews were reported at the annual WHO conferences, and these findings helped improve the ICD-11 browser and reference guide.

Health data governance for research use in Alberta.

Alberta has rich clinical and health services data held under the custodianship of Alberta Health and Alberta Health Services (AHS), which is not only used for clinical and administrative purposes but also disease surveillance and epidemiological research. Alberta is the largest province in Canada with a single payer centralised health system, AHS, and a consolidated data and analytics team supporting researchers across the province. This paper describes Alberta's data custodians, data governance mechanisms, and streamlined processes followed for research data access. AHS has created a centralised data repository from multiple sources, including practitioner claims data, hospital discharge data, and medications dispensed, available for research use through the provincial Data and Research Services (DRS) team. The DRS team is integrated within AHS to support researchers across the province with their data extraction and linkage requests. Furthermore, streamlined processes have been established, including: 1) ethics approval from a research ethics board, 2) any necessary operational approvals from AHS, and 3) a tripartite legal agreement dictating terms and conditions for data use, disclosure, and retention. This allows researchers to gain timely access to data. To meet the evolving and ever-expanding big-data needs, the University of Calgary, in partnership with AHS, has built high-performance computing (HPC) infrastructure to facilitate storage and processing of large datasets. When releasing data to researchers, the analytics team ensures that Alberta's Health Information Act's guiding principles are followed. The principal investigator also ensures data retention and disposition are according to the plan specified in ethics and per the terms set out by funding agencies. Even though there are disparities and variations in the data protection laws across the different provinces in Canada, the streamlined processes for research data access in Alberta are highly efficient.

Performance of machine learning algorithms for surgical site infection case detection and prediction: A systematic review and meta-analysis.

Background: Medical researchers and clinicians have shown much interest in developing machine learning (ML) algorithms to detect/predict surgical site infections (SSIs). However, little is known about the overall performance of ML algorithms in predicting SSIs and how to improve the algorithm's robustness. We conducted a systematic review and meta-analysis to summarize the performance of ML algorithms in SSIs case detection and prediction and to describe the impact of using unstructured and textual data in the development of ML algorithms. Methods: MEDLINE, EMBASE, CINAHL, CENTRAL and Web of Science were searched from inception to March 25, 2021. Study characteristics and algorithm development information were extracted. Performance statistics (e.g., sensitivity, area under the receiver operating characteristic curve [AUC]) were pooled using a random effect model. Stratified analysis was applied to different study characteristic levels. Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Diagnostic Test Accuracy Studies (PRISMA-DTA) was followed. Results: Of 945 articles identified, 108 algorithms from 32 articles were included in this review. The overall pooled estimate of the SSI incidence rate was 3.67%, 95% CI: 3.58-3.76. Mixed-use of structured and textual data-based algorithms (pooled estimates of sensitivity 0.83, 95% CI: 0.78-0.87, specificity 0.92, 95% CI: 0.86-0.95, AUC 0.92, 95% CI: 0.89-0.94) outperformed algorithms solely based on structured data (sensitivity 0.56, 95% CI:0.43-0.69, specificity 0.95, 95% CI:0.91-0.97, AUC = 0.90, 95% CI: 0.87-0.92). Conclusions: ML algorithms developed with structured and textual data provided optimal performance. External validation of ML algorithms is needed to translate current knowledge into clinical practice.

Field testing a new ICD coding system: methods and early experiences with ICD-11 Beta Version 2018.

OBJECTIVE: A beta version (2018) of International Classification of Diseases, 11th Revision for MMS (ICD-11), needed testing. Field-testing involves real-world application of the new codes to examine usability. We describe creating a dataset and characterizing the usability of ICD-11 code set by coders. We compare ICD-11 against ICD-10-CA (Canadian modification) and a reference standard dataset of diagnoses. Real-world usability encompasses code selection and time to code a complete inpatient chart using ICD-11 compared with ICD-10-CA. METHODS AND RESULTS: A random sample of inpatient records previously coded using ICD-10-CA was selected from hospitals in Calgary, Alberta (N = 2896). Nurses examined these charts for conditions and healthcare-related harms. Clinical coders re-coded the same charts using ICD-11 codes. Inter-rater reliability (IRR) and coding time improved with ICD-11 coding experience (23.6 to 9.9 min average per chart). Code structure comparisons and challenges encountered are described. Overall, 86.3% of main condition codes matched. Coder comments regarding duplicate codes, missing codes, code finding issues enabled improvements to the ICD-11 Browser, Coding Tool, and Reference Guide. Training is essential for solid IRR with 17,000 diagnostic categories in the new ICD-11. As countries transition to ICD-11, our coding experiences and methods can inform users for implementation or field testing.

International Classification of Diseases clinical coding training: An international survey.

BACKGROUND: The International Classification of Diseases (ICD) is widely used by clinical coders worldwide for clinical coding morbidity data into administrative health databases. Accordingly, hospital data quality largely depends on the coders' skills acquired during ICD training, which varies greatly across countries. OBJECTIVE: To characterise the current landscape of international ICD clinical coding training. METHOD: An online questionnaire was created to survey the 194 World Health Organization (WHO) member countries. Questions focused on the training provided to clinical coding professionals. The survey was distributed to potential participants who met specific criteria, and to organisations specialised in the topic, such as WHO Collaborating Centres, to be forwarded to their representatives. Responses were analysed using descriptive statistics. RESULTS: Data from 47 respondents from 26 countries revealed disparities in all inquired topics. However, most participants reported clinical coders as the primary person assigning ICD codes. Although training was available in all countries, some did not mandate training qualifications, and those that did differed in type and duration of training, with college or university degree being most common. Clinical coding certificates most frequently entailed passing a certification exam. Most countries offered continuing training opportunities, and provided a range of support resources for clinical coders. CONCLUSION: Variability in clinical coder training could affect data collection worldwide, thus potentially hindering international comparability of health data. IMPLICATIONS: These findings could encourage countries to improve their resources and training programs available for clinical coders and will ultimately be valuable to the WHO for the standardisation of ICD training.

Machine Learning Patient-Specific Prediction of Heart Failure Hospitalization Using Cardiac MRI-Based Phenotype and Electronic Health Information.

Background: Heart failure (HF) hospitalization is a dominant contributor of morbidity and healthcare expenditures in patients with systolic HF. Cardiovascular magnetic resonance (CMR) imaging is increasingly employed for the evaluation of HF given capacity to provide highly reproducible phenotypic markers of disease. The combined value of CMR phenotypic markers and patient health information to deliver predictions of future HF events has not been explored. We sought to develop and validate a novel risk model for the patient-specific prediction of time to HF hospitalization using routinely reported CMR variables, patient-reported health status, and electronic health information. Methods: Standardized data capture was performed for 1,775 consecutive patients with chronic systolic HF referred for CMR imaging. Patient demographics, symptoms, Health-related Quality of Life, pharmacy, and routinely reported CMR features were provided to both machine learning (ML) and competing risk Fine-Gray-based models (FGM) for the prediction of time to HF hospitalization. Results: The mean age was 59 years with a mean LVEF of 36 ± 11%. The population was evenly distributed between ischemic (52%) and idiopathic non-ischemic cardiomyopathy (48%). Over a median follow-up of 2.79 years (IQR: 1.59-4.04) 333 patients (19%) experienced HF related hospitalization. Both ML and competing risk FGM based models achieved robust performance for the prediction of time to HF hospitalization. Respective 90-day, 1 and 2-year AUC values were 0.87, 0.83, and 0.80 for the ML model, and 0.89, 0.84, and 0.80 for the competing risk FGM-based model in a holdout validation cohort. Patients classified as high-risk by the ML model experienced a 34-fold higher occurrence of HF hospitalization at 90 days vs. the low-risk group. Conclusion: In this study we demonstrated capacity for routinely reported CMR phenotypic markers and patient health information to be combined for the delivery of patient-specific predictions of time to HF hospitalization. This work supports an evolving migration toward multi-domain data collection for the delivery of personalized risk prediction at time of diagnostic imaging.

Decision algorithm for when to use the ICD-11 3-part model for healthcare harms.

Accurate data collection of healthcare-related adverse events provides a foundation for quality and health system improvement. The International Classification of Diseases for Mortality and Morbidity Statistics, 11th revision (ICD-11 MMS) includes new codes to identify harm or injury and the events or actions leading to the adverse events. However, it is difficult to choose the correct codes without in-depth understanding of which event may be classified as an injury or harm. A 3-part model will be available in the ICD-11 MMS to cluster the codes for the harm or injury that occurred, the causal factors, and the mode (mechanism) involved. While field testing coding of adverse events, our team developed a decision tree (algorithm), which guides when to use the 3-part model. The decision tree now resides in the ICD-11 Reference Guide. This paper is part of a special ICD-11 paper series and outlines the steps used in the decision-tree (algorithm) and provides examples to help understand the process.While it may take coders some time to gain experience to use the 3-part model and decision-tree, the ICD-11 Reference Guide and this paper can be helpful resources to help clarify the process.

Supervised consumption site enables cost savings by avoiding emergency services: a cost analysis study.

BACKGROUND AND AIMS: We report on a cost analysis study, using population level data to determine the emergency service costs avoided from emergency overdose management at supervised consumption services (SCS). DESIGN: We completed a cost analysis from a payer's perspective. In this setting, there is a single-payer model of service delivery. SETTING: In Calgary, Canada, 'Safeworks Harm Reduction Program' was established in late 2017 and offers 24/7 access to SCS. The facility is a nurse-led service, available for client drop-in. We conducted a cost analysis for the entire duration of the program from November 2017 to January 2020, a period of 2 years and 3 months. METHODS: We assessed costs using the following factors from government health databases: monthly operational costs of providing services for drug consumption, cost of providing ambulance pre-hospital care for clients with overdoses who could not be revived at the facility, cost of initial treatment in an emergency department, and benefit of costs averted from overdoses that were successfully managed at the SCS. RESULTS: The proportion of clients who have overdosed at the SCS has decreased steadily for the duration of the program. The number of overdoses that can be managed on site at the SCS has trended upward, currently 98%. Each overdose that is managed at the SCS produces approximately $1600 CAD in cost savings, with a savings of over $2.3 million for the lifetime of the program. CONCLUSION: Overdose management at an SCS creates cost savings by offsetting costs required for managing overdoses using emergency department and pre-hospital ambulance services.

THE PREVALENCE OF INSOMNIA AND SLEEP APNEA IN DISCHARGE ABSTRACT DATA: A CALL TO IMPROVE DATA QUALITY.

Insomnia and sleep apnea are associated with a variety of comorbid conditions and carry a symptom burden to patients. As the prevalence of insomnia and sleep apnea continue to rise, it is imperative that appropriate tools are implemented to accurately capture their prevalence in acute care settings. A retrospective chart review was conducted on 3,074 inpatient charts in Calgary, Alberta. The estimated prevalence of insomnia was 10.36 percent, and sleep apnea was 6.56 percent in inpatient visits between January 1, 2015, and June 30, 2015. The sensitivity of insomnia and sleep apnea were low, and the specificity was high when comparing the chart review to the ICD-10. As both insomnia and sleep apnea were associated with various comorbid conditions, it would be imperative that alternate methods are identified to capture and code them. This would enable clinicians to better identify and treat them, and ultimately improve patient care.

CREATE: A New Data Resource to Support Cardiac Precision Health.

BACKGROUND: The initiatives of precision medicine and learning health systems require databases with rich and accurately captured data on patient characteristics. We introduce the Clinical Registry, AdminisTrative Data and Electronic Medical Records (CREATE) database, which includes linked data from 4 population databases: Alberta Provincial Project for Outcome Assessment in Coronary Heart Disease (APPROACH; a national clinical registry), Sunrise Clinical Manager (SCM) electronic medical record (city-wide), the Discharge Abstract Database (DAD), and the National Ambulatory Care Reporting System (NACRS). The intent of this work is to introduce a cardiovascular-specific database for pursuing precision health activities using big data analytics. METHODS: We used deterministic data linkage to link SCM electronic medical record data to APPROACH clinical registry data using patient identifier variables. The APPROACH-SCM data set was subsequently linked to DAD and NACRS to obtain inpatient and outpatient cohort data. We further validated the quality of the linkage, where applicable, in these databases by comparing against the Alberta Health Insurance Care Plan registry database. RESULTS: We achieved 99.96% linkage across these 4 databases. Currently, there are 30,984 patients with 35,753 catheterizations in the CREATE database. The inpatient cohort contained 65.75% (20,373/30,984) of the patient sample, whereas the outpatient cohort contained 29.78% (9226/30,984). The infrastructure and the process to update and expand the database has been established. CONCLUSIONS: CREATE is intended to serve as a database for supporting big data analytics activities surrounding cardiac precision health. The CREATE database will be managed by the Centre for Health Informatics at the University of Calgary, and housed in a secure high-performance computing environment.

CONTEXTE: Les initiatives en matière de médecine de précision et les systèmes de santé apprenants ont besoin de bases de données riches et exactes sur les caractéristiques des patients. Nous présentons ici la base de données CREATE ( C linical Re gistry, A dminis t rative Data and E lectronic Medical Records), qui regroupe les données couplées de quatre bases de données populationnelles : le registre clinique national APPROACH ( A lberta P rovincial Pr oject for O utcome A ssessment in C oronary H eart Disease), le système de gestion des dossiers médicaux électroniques SCM (Sunrise Clinical Manager, utilisé à l'échelle municipale), la Base de données sur les congés des patients (BDCP), et le Système national d'information sur les soins ambulatoires (SNISA). Notre objectif est d'offrir une base de données portant précisément sur les maladies cardiovasculaires, afin de soutenir les activités en santé de précision nécessitant l'analyse de mégadonnées. MÉTHODOLOGIE: Nous avons utilisé une méthode de couplage déterministe pour apparier les données du système SCM à celles du registre APPROACH à l'aide de variables d'identification des patients. L'ensemble de données SCM-APPROACH a ensuite été couplé aux données de la BDCP et du SNISA, afin d'obtenir les données des cohortes des patients hospitalisés et des patients ambulatoires. Lorsque c'était possible, nous avons en outre validé la qualité du couplage en comparant les données à celles de la base de données du Régime d'assurance maladie de l'Alberta. RÉSULTATS: Nous avons obtenu un taux de couplage de 99,96 % pour les quatre bases de données. À l'heure actuelle, la base de données CREATE compte 30 984 patients ayant subi 35 753 cathétérismes. La cohorte des patients hospitalisés représente 65,75 % (20 373/30 984) de l'échantillon, tandis que la cohorte des patients ambulatoires représente 29,78 % (9226/30 984). L'infrastructure et le processus de mise à jour et d'expansion de la base de données ont été définis. CONCLUSIONS: La base de données CREATE est destinée à soutenir les activités d'analyse de mégadonnées nécessaires à la santé cardiaque de précision. Elle sera gérée par le Centre for Health Informatics de l'Université de Calgary et hébergée dans un environnement informatique à haut rendement sécurisé.

Enhancing ICD-Code-Based Case Definition for Heart Failure Using Electronic Medical Record Data.

BACKGROUND: Surveillance and outcome studies for heart failure (HF) require accurate identification of patients with HF. Algorithms based on International Classification of Diseases (ICD) codes to identify HF from administrative data are inadequate owing to their relatively low sensitivity. Detailed clinical information from electronic medical records (EMRs) is potentially useful for improving ICD algorithms. This study aimed to enhance the ICD algorithm for HF definition by incorporating comprehensive information from EMRs. METHODS: The study included 2106 inpatients in Calgary, Alberta, Canada. Medical chart review was used as the reference gold standard for evaluating developed algorithms. The commonly used ICD codes for defining HF were used (namely, the ICD algorithm). The performance of different algorithms using the free text discharge summaries from a population-based EMR were compared with the ICD algorithm. These algorithms included a keyword search algorithm looking for HF-specific terms, a machine learning-based HF concept (HFC) algorithm, an EMR structured data based algorithm, and combined algorithms (the ICD and HFC combined algorithm). RESULTS: Of 2106 patients, 296 (14.1%) were patients with HF as determined by chart review. The ICD algorithm had 92.4% positive predictive value (PPV) but low sensitivity (57.4%). The EMR keyword search algorithm achieved a higher sensitivity (65.5%) than the ICD algorithm, but with a lower PPV (77.6%). The HFC algorithm achieved a better sensitivity (80.0%) and maintained a reasonable PPV (88.9%) compared with the ICD algorithm and the keyword algorithm. An even higher sensitivity (83.3%) was reached by combining the HFC and ICD algorithms, with a lower PPV (83.3%). The structured EMR data algorithm reached a sensitivity of 78% and a PPV of 54.2%. The combined EMR structured data and ICD algorithm had a higher sensitivity (82.4%), but the PPV remained low at 54.8%. All algorithms had a specificity ranging from 87.5% to 99.2%. CONCLUSIONS: Applying natural language processing and machine learning on the discharge summaries of inpatient EMR data can improve the capture of cases of HF compared with the widely used ICD algorithm. The utility of the HFC algorithm is straightforward, making it easily applied for HF case identification.

Identifying Cases of Sleep Disorders through International Classification of Diseases (ICD) Codes in Administrative Data.

OBJECTIVES: Prevalence, and associated morbidity and mortality of chronic sleep disorders have been limited to small cohort studies, however, administrative data may be used to provide representation of larger population estimates of disease. With no guidelines to inform the identification of cases of sleep disorders in administrative data, the objective of this study was to develop and validate a set of ICD-codes used to define sleep disorders including narcolepsy, insomnia, and obstructive sleep apnea (OSA) in administrative data. METHODS: A cohort of adult patients, with medical records reviewed by two independent board-certified sleep physicians from a sleep clinic in Calgary, Alberta between January 1, 2009 and December 31, 2011, was used as the reference standard. We developed a general ICD-coded case definition for sleep disorders which included conditions of narcolepsy, insomnia, and OSA using: 1) physician claims data, 2) inpatient visit data, 3) emergency department (ED) and ambulatory care data. We linked the reference standard data and administrative data to examine the validity of different case definitions, calculating estimates of sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). RESULTS: From a total of 1186 patients from the sleep clinic, 1045 (88.1%) were classified as sleep disorder positive, with 606 (51.1%) diagnosed with OSA, 407 (34.4%) with insomnia, and 59 (5.0%) with narcolepsy. The most frequently used ICD-9 codes were general codes of 307.4 (Nonorganic sleep disorder, unspecified), 780.5 (unspecified sleep disturbance) and ICD-10 codes of G47.8 (other sleep disorders), G47.9 (sleep disorder, unspecified). The best definition for identifying a sleep disorder was an ICD code (from physician claims) 2 years prior and 1 year post sleep clinic visit: sensitivity 79.2%, specificity 28.4%, PPV 89.1%, and NPV 15.6%. ICD codes from ED/ambulatory care data provided similar diagnostic performance when at least 2 codes appeared in a time period of 2 years prior and 1 year post sleep clinic visit: sensitivity 71.9%, specificity 54.6%, PPV 92.1%, and NPV 20.8%. The inpatient data yielded poor results in all tested ICD code combinations. CONCLUSION: Sleep disorders in administrative data can be identified mainly through physician claims data and with some being determined through outpatient/ambulatory care data ICD codes, however these are poorly coded within inpatient data sources. This may be a function of how sleep disorders are diagnosed and/or reported by physicians in inpatient and outpatient settings within medical records. Future work to optimize administrative data case definitions through data linkage are needed.

Validation of two case definitions to identify pressure ulcers using hospital administrative data.

OBJECTIVE: Pressure ulcer development is a quality of care indicator, as pressure ulcers are potentially preventable. Yet pressure ulcer is a leading cause of morbidity, discomfort and additional healthcare costs for inpatients. Methods are lacking for accurate surveillance of pressure ulcer in hospitals to track occurrences and evaluate care improvement strategies. The main study aim was to validate hospital discharge abstract database (DAD) in recording pressure ulcers against nursing consult reports, and to calculate prevalence of pressure ulcers in Alberta, Canada in DAD. We hypothesised that a more inclusive case definition for pressure ulcers would enhance validity of cases identified in administrative data for research and quality improvement purposes. SETTING: A cohort of patients with pressure ulcers were identified from enterostomal (ET) nursing consult documents at a large university hospital in 2011. PARTICIPANTS: There were 1217 patients with pressure ulcers in ET nursing documentation that were linked to a corresponding record in DAD to validate DAD for correct and accurate identification of pressure ulcer occurrence, using two case definitions for pressure ulcer. RESULTS: Using pressure ulcer definition 1 (7 codes), prevalence was 1.4%, and using definition 2 (29 codes), prevalence was 4.2% after adjusting for misclassifications. The results were lower than expected. Definition 1 sensitivity was 27.7% and specificity was 98.8%, while definition 2 sensitivity was 32.8% and specificity was 95.9%. Pressure ulcer in both DAD and ET consultation increased with age, number of comorbidities and length of stay. CONCLUSION: DAD underestimate pressure ulcer prevalence. Since various codes are used to record pressure ulcers in DAD, the case definition with more codes captures more pressure ulcer cases, and may be useful for monitoring facility trends. However, low sensitivity suggests that this data source may not be accurate for determining overall prevalence, and should be cautiously compared with other prevalence studies.