Free-text Electronic Medical Records Research Articles

Development of algorithms to identify individuals with Neurofibromatosis type 1 within administrative data and electronic medical records in Ontario, Canada

BackgroundThere is limited population-based data on Neurofibromatosis type 1 (NF1) in North America. We aimed to develop and validate algorithms using administrative health data and electronic medical records (EMRs) to identify individuals with NF1 in Ontario, Canada.MethodsWe conducted an electronic free-text search of 15 commonly-used terms related to NF1 in the Electronic Medical Records Primary Care Database. Records were reviewed by two trained abstractors who classified them as confirmed, possible, and not NF1. An investigator with clinical expertise performed final NF1 classification. Patients were classified as confirmed if there was a documented diagnosis, meeting NIH criteria. Patients were classified as possible if (1) NF1 was recorded in the cumulative patient profile, but no clinical information to support the diagnosis; (2) only one criterion for diagnosis (e.g. child of confirmed case) but no further data to confirm or rule out. We tested different combinations of outpatient and inpatient billing codes, and applied a free-text search algorithm to identify NF1 cases in administrative data and EMRs, respectively.ResultsOf 273,440 eligible patients, 2,058 had one or more NF1 terms in their medical records. The terms “NF”, “café-au-lait”, or “sheath tumour” were constrained to appear in combination with another NF1 term. This resulted in 837 patients: 37 with possible and 71 with confirmed NF1. The population prevalence ranged from 1 in 3851 (confirmed NF1) to 1 in 2532 (possible and confirmed NF1). Billing code algorithms had poor performance, with overall low PPV (highest being 71%). The accuracy of the free-text EMR algorithm in identifying patients with NF1 was: sensitivity 85% (95% CI 74–92%), specificity 100% (95% CI 100–100%), positive predictive value 80% (95% CI 69–88%), negative predictive value 100% (95% CI 100–100%), and false positive rate 20% (95% CI 11–33%). Of false positives, 53% were possible NF1.ConclusionsA free-text search algorithm within the EMR had high sensitivity, specificity and predictive values. Algorithms using billing codes had poor performance, likely due to the lack of NF-specific codes for outpatient visits. While NF1 ICD-9 and 10 codes are used for hospital admissions, only ~ 30% of confirmed NF1 cases had a hospitalization associated with an NF1 code.

Natural language processing to measure the frequency and mode of communication between healthcare professionals and family members of critically ill patients.

To apply natural language processing (NLP) techniques to identify individual events and modes of communication between healthcare professionals and families of critically ill patients from electronic medical records (EMR). Retrospective cohort study of 280 randomly selected adult patients admitted to 1 of 15 intensive care units (ICU) in Alberta, Canada from June 19, 2012 to June 11, 2018. Individual events and modes of communication were independently abstracted using NLP and manual chart review (reference standard). Preprocessing techniques and 2 NLP approaches (rule-based and machine learning) were evaluated using sensitivity, specificity, and area under the receiver operating characteristic curves (AUROC). Over 2700 combinations of NLP methods and hyperparameters were evaluated for each mode of communication using a holdout subset. The rule-based approach had the highest AUROC in 65 datasets compared to the machine learning approach in 21 datasets. Both approaches had similar performance in 17 datasets. The rule-based AUROC for the grouped categories of patient documented to have family or friends (0.972, 95% CI 0.934-1.000), visit by family/friend (0.882 95% CI 0.820-0.943) and phone call with family/friend (0.975, 95% CI: 0.952-0.998) were high. We report an automated method to quantify communication between healthcare professionals and family members of adult patients from free-text EMRs. A rule-based NLP approach had better overall operating characteristics than a machine learning approach. NLP can automatically and accurately measure frequency and mode of documented family visitation and communication from unstructured free-text EMRs, to support patient- and family-centered care initiatives.

Prohibited abbreviations in electronic medical records.

To the Editor: The Joint Commission provides guidance to practitioners on safe medication practices through its “Do Not Use” list of abbreviations. In this list, clinicians are encouraged to avoid the use of prohibited abbreviations in patient history and physicals, progress notes, discharge summaries, and medication orders. Avoidance of prohibited abbreviations helps to reduce medication errors that can lead to patient harm. The use of electronic medical records (EMRs) has decreased medication errors caused by illegible handwriting by health care providers, but medication errors can be increased if prohibited abbreviations are used in these communications. Electronic communication must follow previous standards for safe medication practices. Pharmacists play an integral role on an institutional performance improvement committee through their knowledge of safe mediation practices. Our institution performed a rapid-cycle improvement initiative targeting health care provider’s use of institutionally recognized prohibited abbreviations in EMRs. In parallel, the impact of pharmacy on a quality improvement initiative was assessed. Pharmacy led education regarding the avoidance of prohibited abbreviations in free-text EMR documentation and evaluated compliance with our institution’s safe medication practice policy, which includes an expanded prohibited abbreviation list. Eighty patient charts were randomly selected for inclusion in this 10-week evaluation. The pre-education audit found 295 prohibited abbreviations in 40 EMRs, with medication abbreviations being the most common prohibited abbreviation (78%) (Table 1). Education interventions to reach practitioners were performed by pharmacy personnel, including publishing prohibited abbreviations in physician workrooms, nursing work areas, and morning resident and faculty meetings; distributing data collection results and educational material to physician department chairs; recruiting unit-based clinical pharmacists to educate medical teams; and participating in the nursing skills fair. The post education audit found that pharmacist-led education resulted in a 68% reduction of prohibited abbreviations in 40 EMRs. Table 1. Frequency of prohibited abbreviation use in electronic medical records (EMRs) before and after educational intervention (N = 40) As advances in technology are integrated into health care, communication is increased between health care providers. Providers must continue to avoid using prohibited abbreviations to ensure patient safety. Our performance improvement initiative highlights the impact pharmacists may have on an institution’s achievement of National Patient Safety Goals and encourages pharmacists to be at the forefront of safe medication practices at their institution.

Framework Design and Case Study for Privacy-Preserving Medical Data Publishing

With the pervasive using of Electronic Medical Records (EMR) and telemedicine technologies, more and more digital healthcare data are accumulated from multiple sources. As healthcare data is valuable for both commercial and scientific research, the demand of sharing healthcare data has been growing rapidly. Nevertheless, health care data normally contains a large amount of personal information, and sharing them directly would bring huge threaten to the patient privacy. This paper proposes a privacy preserving framework for medical data sharing with the view of practical application. The framework focuses on three key issues of privacy protection during the data sharing, which are privacy definition/detection, privacy policy management, and privacy preserving data publishing. A case study for Chinese Electronic Medical Record (ERM) publishing with privacy preserving is implemented based on the proposed framework. Specific Chinese free text EMR segmentation, Protected Health Information (PHI) extraction, and K-anonymity PHI anonymous algorithms are proposed in each component. The real-life data from hospitals are used to evaluate the performance of the proposed framework and system.

De-identification of primary care electronic medical records free-text data in Ontario, Canada

BackgroundElectronic medical records (EMRs) represent a potentially rich source of health information for research but the free-text in EMRs often contains identifying information. While de-identification tools have been developed for free-text, none have been developed or tested for the full range of primary care EMR dataMethodsWe used deid open source de-identification software and modified it for an Ontario context for use on primary care EMR data. We developed the modified program on a training set of 1000 free-text records from one group practice and then tested it on two validation sets from a random sample of 700 free-text EMR records from 17 different physicians from 7 different practices in 5 different cities and 500 free-text records from a group practice that was in a different city than the group practice that was used for the training set. We measured the sensitivity/recall, precision, specificity, accuracy and F-measure of the modified tool against manually tagged free-text records to remove patient and physician names, locations, addresses, medical record, health card and telephone numbers.ResultsWe found that the modified training program performed with a sensitivity of 88.3%, specificity of 91.4%, precision of 91.3%, accuracy of 89.9% and F-measure of 0.90. The validations sets had sensitivities of 86.7% and 80.2%, specificities of 91.4% and 87.7%, precisions of 91.1% and 87.4%, accuracies of 89.0% and 83.8% and F-measures of 0.89 and 0.84 for the first and second validation sets respectively.ConclusionThe deid program can be modified to reasonably accurately de-identify free-text primary care EMR records while preserving clinical content.