SNOMED CT Concept Hierarchies for Sharing Definitions of Clinical Conditions Using Electronic Health Record Data

Checklist and guidance on creating codelists for routinely collected health data research

Background

Codelists are required to extract meaningful information on characteristics and events from routinely collected health data such as electronic health records. Research using routinely collected health data relies on codelists to define study populations and variables, thus, trustworthy codelists are important. Here, we provide a checklist, in the style of commonly used reporting guidelines, to help researchers adhere to best practice in codelist development and sharing.

Methods

Based on a literature search and a workshop with researchers experienced in the use of routinely collected health data, we created a set of recommendations that are 1. broadly applicable to different datasets, research questions, and methods of codelist creation; 2. easy to follow, implement and document by an individual researcher, and 3. fit within a step-by-step process. We then formatted these recommendations into a checklist.

Results

We have created a 10-step checklist, comprising 28 items, with accompanying guidance on each step. The checklist advises on which metadata to provide, how to define a clinical concept, how to identify and evaluate existing codelists, how to create new codelists, and how to review, check, finalise, and publish a created codelist.

Conclusions

Use of the checklist can reassure researchers that best practice was followed during the development of their codelists, increasing trust in research that relies on these codelists and facilitating wider re-use and adaptation by other researchers.

Validation of an acute respiratory infection phenotyping algorithm to support robust computerised medical record-based respiratory sentinel surveillance, England, 2023

IntroductionRespiratory sentinel surveillance systems leveraging computerised medical records (CMR) use phenotyping algorithms to identify cases of interest, such as acute respiratory infection (ARI). The Oxford-Royal College of General Practitioners Research and Surveillance Centre (RSC) is the English primary care-based sentinel surveillance network.AimThis study describes and validates the RSC's new ARI phenotyping algorithm.MethodsWe developed the phenotyping algorithm using a framework aligned with international interoperability standards. We validated our algorithm by comparing ARI events identified during the 2022/23 influenza season in England through use of both old and new algorithms. We compared clinical codes commonly used for recording ARI.ResultsThe new algorithm identified an additional 860,039 cases and excluded 52,258, resulting in a net increase of 807,781 cases (33.84%) of ARI compared to the old algorithm, with totals of 3,194,224 cases versus 2,386,443 cases. Of the 860,039 newly identified cases, the majority (63.7%) were due to identification of symptom codes suggestive of an ARI diagnosis not detected by the old algorithm. The 52,258 cases incorrectly identified by the old algorithm were due to inadvertent identification of chronic, recurrent, non-infectious and other non-ARI disease.ConclusionWe developed a new ARI phenotyping algorithm that more accurately identifies cases of ARI from the CMR. This will benefit public health by providing more accurate surveillance reports to public health authorities. This new algorithm can serve as a blueprint for other CMR-based surveillance systems wishing to develop similar phenotyping algorithms.

Primary Infection Site as a Predictor of Sepsis Development in Emergency Department Patients

BACKGROUND

Sepsis is a life-threatening condition but predicting its development and progression remains a challenge.

OBJECTIVE

This study aimed to assess the impact of infection site on sepsis development among emergency department (ED) patients.

METHODS

Data were collected from a single-center ED between January 2016 and December 2019. Patient encounters with documented infections, as defined by the Systematized Nomenclature of Medicine-Clinical Terms for upper respiratory tract (URI), lower respiratory tract (LRI), urinary tract (UTI), or skin or soft-tissue infections were included. Primary outcome was the development of sepsis or septic shock, as defined by Sepsis-1/2 criteria. Secondary outcomes included hospital disposition and length of stay, blood and urine culture positivity, antibiotic administration, vasopressor use, in-hospital mortality, and 30-day mortality. Analysis of variance and various different logistic regression approaches were used for analysis with URI used as the reference variable.

RESULTS

LRI was most associated with sepsis (relative risk ratio [RRR] 5.63; 95% CI 5.07-6.24) and septic shock (RRR 21.2; 95% CI 17.99-24.98) development, as well as hospital admission rates (odds ratio [OR] 8.23; 95% CI 7.41-9.14), intensive care unit admission (OR 4.27; 95% CI 3.84-4.74), in-hospital mortality (OR 6.93; 95% CI 5.60-8.57), and 30-day mortality (OR 7.34; 95% CI 5.86-9.19). UTIs were also associated with sepsis and septic shock development, but to a lesser degree than LRI.

CONCLUSIONS

Primary infection sites including LRI and UTI were significantly associated with sepsis development, hospitalization, length of stay, and mortality among patients presenting with infections in the ED.

Phenotype execution and modeling architecture to support disease surveillance and real-world evidence studies: English sentinel network evaluation

Objective

To evaluate Phenotype Execution and Modelling Architecture (PhEMA), to express sharable phenotypes using Clinical Quality Language (CQL) and intensional Systematised Nomenclature of Medicine (SNOMED) Clinical Terms (CT) Fast Healthcare Interoperability Resources (FHIR) valuesets, for exemplar chronic disease, sociodemographic risk factor, and surveillance phenotypes.

Method

We curated 3 phenotypes: Type 2 diabetes mellitus (T2DM), excessive alcohol use, and incident influenza-like illness (ILI) using CQL to define clinical and administrative logic. We defined our phenotypes with valuesets, using SNOMED's hierarchy and expression constraint language, and CQL, combining valuesets and adding temporal elements where needed. We compared the count of cases found using PhEMA with our existing approach using convenience datasets. We assessed our new approach against published desiderata for phenotypes.

Results

The T2DM phenotype could be defined as 2 intensionally defined SNOMED valuesets and a CQL script. It increased the prevalence from 7.2% to 7.3%. Excess alcohol phenotype was defined by valuesets that added qualitative clinical terms to the quantitative conceptual definitions we currently use; this change increased prevalence by 58%, from 1.2% to 1.9%. We created an ILI valueset with SNOMED concepts, adding a temporal element using CQL to differentiate new episodes. This increased the weekly incidence in our convenience sample (weeks 26-38) from 0.95 cases to 1.11 cases per 100 000 people.

Conclusions

Phenotypes for surveillance and research can be described fully and comprehensibly using CQL and intensional FHIR valuesets. Our use case phenotypes identified a greater number of cases, whilst anticipated from excessive alcohol this was not for our other variable. This may have been due to our use of SNOMED CT hierarchy. Our new process fulfilled a greater number of phenotype desiderata than the one that we had used previously, mostly in the modeling domain. More work is needed to implement that sharing and warehousing domains.

Evaluation of SNOMED CT Grouper Accuracy and Coverage in Organizing the Electronic Health Record Problem List by Clinical System: Observational Study

Background

The problem list (PL) is a repository of diagnoses for patients' medical conditions and health-related issues. Unfortunately, over time, our PLs have become overloaded with duplications, conflicting entries, and no-longer-valid diagnoses. The lack of a standardized structure for review adds to the challenges of clinical use. Previously, our default electronic health record (EHR) organized the PL primarily via alphabetization, with other options available, for example, organization by clinical systems or priority settings. The system's PL was built with limited groupers, resulting in many diagnoses that were inconsistent with the expected clinical systems or not associated with any clinical systems at all. As a consequence of these limited EHR configuration options, our PL organization has poorly supported clinical use over time, particularly as the number of diagnoses on the PL has increased.

Objective

We aimed to measure the accuracy of sorting PL diagnoses into PL system groupers based on Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT) concept groupers implemented in our EHR.

Methods

We transformed and developed 21 system- or condition-based groupers, using 1211 SNOMED CT hierarchal concepts refined with Boolean logic, to reorganize the PL in our EHR. To evaluate the clinical utility of our new groupers, we extracted all diagnoses on the PLs from a convenience sample of 50 patients with 3 or more encounters in the previous year. To provide a spectrum of clinical diagnoses, we included patients from all ages and divided them by sex in a deidentified format. Two physicians independently determined whether each diagnosis was correctly attributed to the expected clinical system grouper. Discrepancies were discussed, and if no consensus was reached, they were adjudicated by a third physician. Descriptive statistics and Cohen κ statistics for interrater reliability were calculated.

Results

Our 50-patient sample had a total of 869 diagnoses (range 4-59; median 12, IQR 9-24). The reviewers initially agreed on 821 system attributions. Of the remaining 48 items, 16 required adjudication with the tie-breaking third physician. The calculated κ statistic was 0.7. The PL groupers appropriately associated diagnoses to the expected clinical system with a sensitivity of 97.6%, a specificity of 58.7%, a positive predictive value of 96.8%, and an F1-score of 0.972.

Conclusions

We found that PL organization by clinical specialty or condition using SNOMED CT concept groupers accurately reflects clinical systems. Our system groupers were subsequently adopted by our vendor EHR in their foundation system for PL organization.

Prevalence and Predictors of Subclinical Cardiomyopathy in Patients With Type 2 Diabetes in a Health System

INTRODUCTION

Diabetic cardiomyopathy (DbCM) is characterized by subclinical abnormalities in cardiac structure/function and is associated with a higher risk of overt heart failure (HF). However, there are limited data on optimal strategies to identify individuals with DbCM in contemporary health systems. The aim of this study was to evaluate the prevalence of DbCM in a health system using existing data from the electronic health record (EHR).

METHODS

Adult patients with type 2 diabetes mellitus free of cardiovascular disease (CVD) with available data on HF risk in a single-center EHR were included. The presence of DbCM was defined using different definitions: (1) least restrictive: ≥1 echocardiographic abnormality (left atrial enlargement, left ventricle hypertrophy, diastolic dysfunction); (2) intermediate restrictive: ≥2 echocardiographic abnormalities; (3) most restrictive: 3 echocardiographic abnormalities. DbCM prevalence was compared across age, sex, race, and ethnicity-based subgroups, with differences assessed using the chi-squared test. Adjusted logistic regression models were constructed to evaluate significant predictors of DbCM.

RESULTS

Among 1921 individuals with type 2 diabetes mellitus, the prevalence of DbCM in the overall cohort was 8.7% and 64.4% in the most and least restrictive definitions, respectively. Across all definitions, older age and Hispanic ethnicity were associated with a higher proportion of DbCM. Females had a higher prevalence than males only in the most restrictive definition. In multivariable-adjusted logistic regression, higher systolic blood pressure, higher creatinine, and longer QRS duration were associated with a higher risk of DbCM across all definitions.

CONCLUSIONS

In this single-center, EHR cohort, the prevalence of DbCM varies from 9% to 64%, with a higher prevalence with older age and Hispanic ethnicity.

Multiple emergency department encounters for acute musculoskeletal presentation with an existing mental health diagnosis

Reconceptualising acute Musculoskeletal (MSK) injuries with both stress- and tissue- based factors is required to consider prior influences of mental health disorders on acute persistent MSK pain presentations. This report describes repeated emergency presentations of an individual with acute persistent MSK pain in their twenties living with mental health. Their mental health diagnoses included depression, mood disorders, and anorexia nervosa. This person also had mental health related inpatient admissions that were not captured under the retrospective record review for a large district hospital emergency department using the Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT) classification system. This case report attempts to demonstrate that improving the understanding of preexisting vulnerabilities and mental health diagnoses may assist with informing healthcare design to develop specialised care pathways for acute injury presentations within triage settings.

Evaluating the representation of disaster hazards in SNOMED CT: gaps and opportunities

OBJECTIVE

Climate change, an underlying risk driver of natural disasters, threatens the environmental sustainability, planetary health, and sustainable development goals. Incorporating disaster-related health impacts into electronic health records helps to comprehend their impact on populations, clinicians, and healthcare systems. This study aims to: (1) map the United Nations Office for Disaster Risk Reduction and International Science Council (UNDRR-ISC) Hazard Information Profiles to SNOMED CT International, a clinical terminology used by clinicians, to manage patients and provide healthcare services; and (2) to determine the extent of clinical terminologies available to capture disaster-related events.

MATERIALS AND METHODS

Concepts related to disasters were extracted from the UNDRR-ISC's Hazard Information Profiles and mapped to a health terminology using a procedural framework for standardized clinical terminology mapping. The mapping process involved evaluating candidate matches and creating a final list of matches to determine concept coverage.

RESULTS

A total of 226 disaster hazard concepts were identified to adversely impact human health. Chemical and biological disaster hazard concepts had better representation than meteorological, hydrological, extraterrestrial, geohazards, environmental, technical, and societal hazard concepts in SNOMED CT. Heatwave, drought, and geographically unique disaster hazards were not found in SNOMED CT.

CONCLUSION

To enhance clinical reporting of disaster hazards and climate-sensitive health outcomes, the poorly represented and missing concepts in SNOMED CT must be included. Documenting the impacts of climate change on public health using standardized clinical terminology provides the necessary real time data to capture climate-sensitive outcomes. These data are crucial for building climate-resilient healthcare systems, enhanced public health disaster responses and workflows, tracking individual health outcomes, supporting disaster risk reduction modeling, and aiding in disaster preparedness, response, and recovery efforts.

Creation of a structured molecular genomics report for Germany as a local adaption of HL7's Genomic Reporting Implementation Guide

OBJECTIVE

The objective was to develop a dataset definition, information model, and FHIR® specification for key data elements contained in a German molecular genomics (MolGen) report to facilitate genomic and phenotype integration in electronic health records.

MATERIALS AND METHODS

A dedicated expert group participating in the German Medical Informatics Initiative reviewed information contained in MolGen reports, determined the key elements, and formulated a dataset definition. HL7's Genomics Reporting Implementation Guide (IG) was adopted as a basis for the FHIR® specification which was subjected to a public ballot. In addition, elements in the MolGen dataset were mapped to the fields defined in ISO/TS 20428:2017 standard to evaluate compliance.

RESULTS

A core dataset of 76 data elements, clustered into 6 categories was created to represent all key information of German MolGen reports. Based on this, a FHIR specification with 16 profiles, 14 derived from HL7®'s Genomics Reporting IG and 2 additional profiles (of the FamilyMemberHistory and RiskAssessment resources), was developed. Five example resource bundles show how our adaptation of an international standard can be used to model MolGen report data that was requested following oncological or rare disease indications. Furthermore, the map of the MolGen report data elements to the fields defined by the ISO/TC 20428:2017 standard, confirmed the presence of the majority of required fields.

CONCLUSIONS

Our report serves as a template for other research initiatives attempting to create a standard format for unstructured genomic report data. Use of standard formats facilitates integration of genomic data into electronic health records for clinical decision support.

Systematized Nomenclature of Medicine-Clinical Terminology (SNOMED CT) Clinical Use Cases in the Context of Electronic Health Record Systems: Systematic Literature Review

BACKGROUND

The Systematized Medical Nomenclature for Medicine-Clinical Terminology (SNOMED CT) is a clinical terminology system that provides a standardized and scientifically validated way of representing clinical information captured by clinicians. It can be integrated into electronic health records (EHRs) to increase the possibilities for effective data use and ensure a better quality of documentation that supports continuity of care, thus enabling better quality in the care process. Even though SNOMED CT consists of extensively studied clinical terminology, previous research has repeatedly documented a lack of scientific evidence for SNOMED CT in the form of reported clinical use cases in electronic health record systems.

OBJECTIVE

The aim of this study was to explore evidence in previous literature reviews of clinical use cases of SNOMED CT integrated into EHR systems or other clinical applications during the last 5 years of continued development. The study sought to identify the main clinical use purposes, use phases, and key clinical benefits documented in SNOMED CT use cases.

METHODS

The Cochrane review protocol was applied for the study design. The application of the protocol was modified step-by-step to fit the research problem by first defining the search strategy, identifying the articles for the review by isolating the exclusion and inclusion criteria for assessing the search results, and lastly, evaluating and summarizing the review results.

RESULTS

In total, 17 research articles illustrating SNOMED CT clinical use cases were reviewed. The use purpose of SNOMED CT was documented in all the articles, with the terminology as a standard in EHR being the most common (8/17). The clinical use phase was documented in all the articles. The most common category of use phases was SNOMED CT in development (6/17). Core benefits achieved by applying SNOMED CT in a clinical context were identified by the researchers. These were related to terminology use outcomes, that is, to data quality in general or to enabling a consistent way of indexing, storing, retrieving, and aggregating clinical data (8/17). Additional benefits were linked to the productivity of coding or to advances in the quality and continuity of care.

CONCLUSIONS

While the SNOMED CT use categories were well supported by previous research, this review demonstrates that further systematic research on clinical use cases is needed to promote the scalability of the review results. To achieve the best out-of-use case reports, more emphasis is suggested on describing the contextual factors, such as the electronic health care system and the use of previous frameworks to enable comparability of results. A lesson to be drawn from our study is that SNOMED CT is essential for structuring clinical data; however, research is needed to gather more evidence of how SNOMED CT benefits clinical care and patient safety.

Translating and evaluating historic phenotyping algorithms using SNOMED CT

OBJECTIVE

Patient phenotype definitions based on terminologies are required for the computational use of electronic health records. Within UK primary care research databases, such definitions have typically been represented as flat lists of Read terms, but Systematized Nomenclature of Medicine-Clinical Terms (SNOMED CT) (a widely employed international reference terminology) enables the use of relationships between concepts, which could facilitate the phenotyping process. We implemented SNOMED CT-based phenotyping approaches and investigated their performance in the CPRD Aurum primary care database.

MATERIALS AND METHODS

We developed SNOMED CT phenotype definitions for 3 exemplar diseases: diabetes mellitus, asthma, and heart failure, using 3 methods: "primary" (primary concept and its descendants), "extended" (primary concept, descendants, and additional relations), and "value set" (based on text searches of term descriptions). We also derived SNOMED CT codelists in a semiautomated manner for 276 disease phenotypes used in a study of health across the lifecourse. Cohorts selected using each codelist were compared to "gold standard" manually curated Read codelists in a sample of 500 000 patients from CPRD Aurum.

RESULTS

SNOMED CT codelists selected a similar set of patients to Read, with F1 scores exceeding 0.93, and age and sex distributions were similar. The "value set" and "extended" codelists had slightly greater recall but lower precision than "primary" codelists. We were able to represent 257 of the 276 phenotypes by a single concept hierarchy, and for 135 phenotypes, the F1 score was greater than 0.9.

CONCLUSIONS

SNOMED CT provides an efficient way to define disease phenotypes, resulting in similar patient populations to manually curated codelists.

The association of unconventional natural gas development with diagnosis and treatment of internalizing disorders among adolescents in Pennsylvania using electronic health records

BACKGROUND

Unconventional natural gas development (UNGD) introduces physical and psychosocial hazards into communities, which could contribute to psychosocial stress in adolescents and an increased risk of internalizing disorders, common and impactful health outcomes.

OBJECTIVES

To evaluate associations between a 180-day composite UNGD activity metric and new onset of internalizing disorders, overall and separately for anxiety and depressive disorders, and effect modification by sex.

METHODS

We used a nested case-control design from 2008 to 2016 in 38 Pennsylvania counties using electronic health records from adolescent Geisinger subjects. Cases were defined by at least two diagnoses or medication orders indicating new onset of an internalizing disorder, and controls frequency-matched 4:1 on age, sex, and year. To evaluate associations, we used generalized estimating equations, with logit link, robust standard errors, and an exchangeable correlation structure within community.

RESULTS

We identified 7,974 adolescents (65.9% female, mean age 15.0 years) with new onset internalizing disorders. There were no associations when we used data from the entire study period. When restricted to years with higher UNGD activity (2010-2016), comparing the highest to lowest quartile, UNGD activity was associated (odds ratio [95% confidence level]) with new onset internalizing disorders (1.15 [1.06, 1.25]). Associations were slightly stronger for depressive disorders. Associations were only present in females (p = 0.009).

DISCUSSION

This is the first epidemiologic study of UNGD in relation to adolescent mental health, an important health outcome in a potentially susceptible group to the environmental and community impacts of UNGD. UNGD activity was associated with new onset internalizing disorders in females in this large sample in an area of active UNGD.

Validation of the WATCH-DM and TRS-HFDM Risk Scores to Predict the Risk of Incident Hospitalization for Heart Failure Among Adults With Type 2 Diabetes: A Multicohort Analysis

Background The WATCH-DM (weight [body mass index], age, hypertension, creatinine, high-density lipoprotein cholesterol, diabetes control [fasting plasma glucose], ECG QRS duration, myocardial infarction, and coronary artery bypass grafting) and TRS-HF_DM (Thrombolysis in Myocardial Infarction [TIMI] risk score for heart failure in diabetes) risk scores were developed to predict risk of heart failure (HF) among individuals with type 2 diabetes. WATCH-DM was developed to predict incident HF, whereas TRS-HF_DM predicts HF hospitalization among patients with and without a prior HF history. We evaluated the model performance of both scores to predict incident HF events among patients with type 2 diabetes and no history of HF hospitalization across different cohorts and clinical settings with varying baseline risk. Methods and Results Incident HF risk was estimated by the integer-based WATCH-DM and TRS-HF_DM scores in participants with type 2 diabetes free of baseline HF from 2 randomized clinical trials (TECOS [Trial Evaluating Cardiovascular Outcomes With Sitagliptin], N=12 028; and Look AHEAD [Look Action for Health in Diabetes] trial, N=4867). The integer-based WATCH-DM score was also validated in electronic health record data from a single large health care system (N=7475). Model discrimination was assessed by the Harrell concordance index and calibration by the Greenwood-Nam-D'Agostino statistic. HF incidence rate was 7.5, 3.9, and 4.1 per 1000 person-years in the TECOS, Look AHEAD trial, and electronic health record cohorts, respectively. Integer-based WATCH-DM and TRS-HF_DM scores had similar discrimination and calibration for predicting 5-year HF risk in the Look AHEAD trial cohort (concordance indexes=0.70; Greenwood-Nam-D'Agostino P>0.30 for both). Both scores had lower discrimination and underpredicted HF risk in the TECOS cohort (concordance indexes=0.65 and 0.66, respectively; Greenwood-Nam-D'Agostino P<0.001 for both). In the electronic health record cohort, the integer-based WATCH-DM score demonstrated a concordance index of 0.73 with adequate calibration (Greenwood-Nam-D'Agostino P=0.96). TRS-HF_DM score could not be validated in the electronic health record because of unavailability of data on urine albumin/creatinine ratio in most patients in the contemporary clinical practice. Conclusions The WATCH-DM and TRS-HF_DM risk scores can discriminate risk of HF among intermediate-risk populations with type 2 diabetes.

Risk Prediction for Acute Kidney Injury in Patients Hospitalized With COVID-19

Rationale & Objective

Acute kidney injury (AKI) is common in patients hospitalized with COVID-19, but validated, predictive models for AKI are lacking. We aimed to develop the best predictive model for AKI in hospitalized patients with coronavirus disease 2019 and assess its performance over time with the emergence of vaccines and the Delta variant.

Study Design

Longitudinal cohort study.

Setting & Participants

Hospitalized patients with a positive severe acute respiratory syndrome coronavirus 2 polymerase chain reaction result between March 1, 2020, and August 20, 2021 at 19 hospitals in Texas.

Exposures

Comorbid conditions, baseline laboratory data, inflammatory biomarkers.

Outcomes

AKI defined by KDIGO (Kidney Disease: Improving Global Outcomes) creatinine criteria.

Analytical Approach

Three nested models for AKI were built in a development cohort and validated in 2 out-of-time cohorts. Model discrimination and calibration measures were compared among cohorts to assess performance over time.

Results

Of 10,034 patients, 5,676, 2,917, and 1,441 were in the development, validation 1, and validation 2 cohorts, respectively, of whom 776 (13.7%), 368 (12.6%), and 179 (12.4%) developed AKI, respectively (P = 0.26). Patients in the validation cohort 2 had fewer comorbid conditions and were younger than those in the development cohort or validation cohort 1 (mean age, 54 ± 16.8 years vs 61.4 ± 17.5 and 61.7 ± 17.3 years, respectively, P < 0.001). The validation cohort 2 had higher median high-sensitivity C-reactive protein level (81.7 mg/L) versus the development cohort (74.5 mg/L; P < 0.01) and higher median ferritin level (696 ng/mL) versus both the development cohort (444 ng/mL) and validation cohort 1 (496 ng/mL; P < 0.001). The final model, which added high-sensitivity C-reactive protein, ferritin, and D-dimer levels, had an area under the curve of 0.781 (95% CI, 0.763-0.799). Compared with the development cohort, discrimination by area under the curve (validation 1: 0.785 [0.760-0.810], P = 0.79, and validation 2: 0.754 [0.716-0.795], P = 0.53) and calibration by estimated calibration index (validation 1: 0.116 [0.041-0.281], P = 0.11, and validation 2: 0.081 [0.045-0.295], P = 0.11) showed stable performance over time.

Limitations

Potential billing and coding bias.

Conclusions

We developed and externally validated a model to accurately predict AKI in patients with coronavirus disease 2019. The performance of the model withstood changes in practice patterns and virus variants.

Supporting Structured Data Capture for Patients With Cancer: An Initiative of the University of Wisconsin Carbone Cancer Center Survivorship Program to Improve Capture of Malignant Diagnosis and Cancer Staging Data

PURPOSE

Structured data elements within electronic health records are health-related information that can be entered, stored, and extracted in an organized manner at later time points. Tracking outcomes for cancer survivors is also enabled by structured data. We sought to increase structured data capture within oncology practices at multiple sites sharing the same electronic health records.

METHODS

Applying engineering approaches and the Plan-Do-Study-Act cycle, we launched dual quality improvement initiatives to ensure that a malignant diagnosis and stage were captured as structured data. Intervention: Close Visit Validation (CVV) requires providers to satisfy certain criteria before closing ambulatory encounters. CVV may be used to track open clinical encounters and chart delinquencies to encourage optimal clinical workflows. We added two cancer-specific required criteria at the time of closing encounters in oncology clinics: (1) the presence of at least one malignant diagnosis on the Problem List and (2) staging all the malignant diagnoses on the Problem List when appropriate.

RESULTS

Six months before the CVV implementation, the percentage of encounters with a malignant diagnosis on the Problem List at the time of the encounter was 65%, whereas the percentage of encounters with a staged diagnosis was 32%. Three months after cancer-specific CVV implementation, the percentages were 85% and 75%, respectively. Rates had increased to 90% and 88% more than 2 years after implementation.

CONCLUSION

Oncologist performance improved after the implementation of cancer-specific CVV criteria, with persistently high percentages of relevant malignant diagnoses and cancer stage structured data capture 2 years after the intervention.

Prevalence and incidence of neuromuscular conditions in the UK between 2000 and 2019: A retrospective study using primary care data

BACKGROUND

In the UK, large-scale electronic primary care datasets can provide up-to-date, accurate epidemiological information on rarer diseases, where specialist diagnoses from hospital discharges and clinic letters are generally well recorded and electronically searchable. Current estimates of the number of people living with neuromuscular disease (NMD) have largely been based on secondary care data sources and lacked direct denominators.

OBJECTIVE

To estimate trends in the recording of neuromuscular disease in UK primary care between 2000-2019.

METHODS

The Clinical Practice Research Datalink (CPRD) database was searched electronically to estimate incidence and prevalence rates (per 100,000) for a range of NMDs in each year. To compare trends over time, rates were age standardised to the most recent CPRD population (2019).

RESULTS

Approximately 13 million patients were actively registered in each year. By 2019, 28,230 active patients had ever received a NMD diagnosis (223.6), which was higher among males (239.0) than females (208.3). The most common classifications were Guillain-Barre syndrome (40.1), myasthenia gravis (33.7), muscular dystrophy (29.5), Charcot-Marie-Tooth (29.5) and inflammatory myopathies (25.0). Since 2000, overall prevalence grew by 63%, with the largest increases seen at older ages (≥65-years). However, overall incidence remained constant, though myasthenia gravis incidence has risen steadily since 2008, while new cases of muscular dystrophy fell over the same period.

CONCLUSIONS

Lifetime recording of many NMDs on primary care records exceed current estimates of people living with these conditions; these are important data for health service and care planning. Temporal trends suggest this number is steadily increasing, and while this may partially be due to better recording, it cannot be simply explained by new cases, as incidence remained constant. The increase in prevalence among older ages suggests increases in life expectancy among those living with NMDs may have occurred.

Clinical Decision Support Systems

BACKGROUND

By combining up-to-date medical knowledge and steadily increasing patient data, a new level of medical care can emerge.

SUMMARY AND KEY MESSAGES

Clinical decision support systems (CDSSs) are an arising solution to handling rich data and providing them to health care providers in order to improve diagnosis and treatment. However, despite promising examples in many areas, substantial evidence for a thorough benefit of these support solutions is lacking. This may be due to a lack of general frameworks and diverse health systems around the globe. We therefore summarize the current status of CDSSs in medicine but also discuss potential limitations that need to be overcome in order to further foster future development and acceptance.

Definition and validation of SNOMED CT subsets using the expression constraint language

BACKGROUND

SNOMED CT Expression Constraint Language (ECL) is a declarative language developed by SNOMED International for the definition of SNOMED CT Expression Constraints (ECs). ECs are executable expressions that define intensional subsets of clinical meanings by stating constraints over the logic definition of concepts. The execution of an EC on some SNOMED CT substrate yields the intended subset, and it requires an execution engine able to receive an EC as input, execute it, and return the matching concepts. An important issue regarding subsets of clinical concepts is their use in terminology binding between clinical information models and terminologies for defining the set of valid values of codified data.

OBJECTIVE

To define and implement methods for the simplification, semantic validation and execution of ECs over a graph-oriented SNOMED CT database, and to provide a method for the visual representation of subsets in order to explore, understand and validate its content, as well as to develop an EC execution platform, called SNQuery, which makes use of these methods.

METHODS

Since SNOMED CT is a directed and acyclic graph, we have used a graph-oriented database to represent the content of SNOMED CT, where the schema and instances are represented as graphs and the data manipulation is expressed by graph-oriented operations. For the execution of ECs over the graph database, it is performed a translation process in which ECs are translated into a set of Cypher Query Language queries. We have defined some EC simplification methods that leverage the logic structure underlying SNOMED CT. The purpose of these methods is to reduce the complexity of ECs and, in turn, its execution time, as well as to validate them from a SNOMED CT Concept Model and logical definition points of view. We also have developed a graphic representation based on the circle packing geometrical concept, which allows validating subsets, as well as pre-defined refsets and the terminology itself.

RESULTS

We have developed SNQuery, a platform for the definition of intensional subsets of SNOMED CT concepts by means of the execution of ECs over a graph-oriented SNOMED CT database. Additionally, we have incorporated methods for the simplification and semantic validation of ECs, as well as for the visualization of subsets as a mechanism to understand and validate them. SNQuery has been evaluated in terms of EC execution times.

CONCLUSION

In this paper, we provide methods to simplify, semantically validate and execute ECs over a graph-oriented database. We also offer a method to visualize the intensional subsets obtained by executing ECs to explore, understand and validate them, as well as refsets and the terminology itself. The definition of intensional subsets is useful to bind content between clinical information models and clinical terminologies, which is a necessary step to achieve semantic interoperability between EHR systems.

[Extraction of features from clinical routine data using text mining]

BACKGROUND

Anti-VEGF drugs are currently used to treat macular diseases. This has led to a wealth of additional data, which could help understand and predict treatment courses; however, this information is usually only available in free text form.

OBJECTIVE

A retrospective study was designed to analyze how far interpretable information can be obtained from clinical texts by automated extraction. The aim was to assess the suitability of a text mining method that was customized for this purpose.

MATERIAL AND METHODS

Data on 3683 patients were available, including 40,485 discharge letters. Some of the data of interest, e.g. visual acuity (VA), intraocular pressure (IOP) and accompanying diagnoses, were not only recorded textually but also entered in a database and could thus serve as a gold standard for text analysis. The text was analyzed using the Averbis Health Discovery text mining platform. To optimize the extraction task, rule knowledge and a German language technical vocabulary linked to the international medical terminology standard systematized nomenclature of medicine (SNOMED CT) was manually added.

RESULTS

The correspondence between extracted data and the structured database entries is described by the F1 value. There was agreement of 94.7% for VA, 98.3% for IOP and 94.7% for the accompanying diagnoses. Manual analysis of noncorresponding cases showed that in 50% text content did not match the database content for various reasons. After an adjustment, F1 values 1-3% above the previously determined values were obtained.

CONCLUSION

Text mining procedures are very well suited for the considered discharge letter corpus and the problem described in order to extract contents from clinical texts in a structured manner for further evaluation.

Validating the Matching of Patients in the Linkage of a Large Hospital System's EHR with State and National Death Databases

Background  Though electronic health record (EHR) data have been linked to national and state death registries, such linkages have rarely been validated for an entire hospital system's EHR.

Objectives  The aim of the study is to validate West Virginia University Medicine's (WVU Medicine) linkage of its EHR to three external death registries: the Social Security Death Masterfile (SSDMF), the national death index (NDI), the West Virginia Department of Health and Human Resources (DHHR).

Methods  Probabilistic matching was used to link patients to NDI and deterministic matching for the SSDMF and DHHR vital statistics records (WVDMF). In subanalysis, we used deaths recorded in Epic ( n  = 30,217) to further validate a subset of deaths captured by the SSDMF, NDI, and WVDMF.

Results  Of the deaths captured by the SSDMF, 59.8 and 68.5% were captured by NDI and WVDMF, respectively; for deaths captured by NDI this co-capture rate was 80 and 78%, respectively, for the SSDMF and WVDMF. Kappa statistics were strongest for NDI and WVDMF (61.2%) and NDI and SSDMF (60.6%) and weakest for SSDMF and WVDMF (27.9%). Of deaths recorded in Epic, 84.3, 85.5, and 84.4% were captured by SSDMF, NDI, and WVDMF, respectively. Less than 2% of patients' deaths recorded in Epic were not found in any of the death registries. Finally, approximately 0.2% of “decedents” in any death registry re-emerged in Epic at least 6 months after their death date, a very small percentage and thus further validating the linkages.

Conclusion  NDI had greatest validity in capturing deaths in our EHR. As a similar, though slightly less capture and agreement rate in identifying deaths is observed for SSDMF and state vital statistics records, these registries may be reasonable alternatives to NDI for research and quality assurance studies utilizing entire EHRs from large hospital systems. Investigators should also be aware that there will be a very tiny fraction of “dead” patients re-emerging in the EHR.

Derivation With Internal Validation of a Multivariable Predictive Model to Predict COVID-19 Test Results in Emergency Department Patients

OBJECTIVES

The COVID-19 pandemic has placed acute care providers in demanding situations in predicting disease given the clinical variability, desire to cohort patients, and high variance in testing availability. An approach to stratifying patients by likelihood of disease based on rapidly available emergency department (ED) clinical data would offer significant operational and clinical value. The purpose of this study was to develop and internally validate a predictive model to aid in the discrimination of patients undergoing investigation for COVID-19.

METHODS

All patients greater than 18 years presenting to a single academic ED who were tested for COVID-19 during this index ED evaluation were included. Outcome was defined as the result of COVID-19 polymerase chain reaction (PCR) testing during the index visit or any positive result within the following 7 days. Variables included chest radiograph interpretation, disease-specific screening questions, and laboratory data. Three models were developed with a split-sample approach to predict outcome of the PCR test utilizing logistic regression, random forest, and gradient-boosted decision tree methods. Model discrimination was evaluated comparing area under the receiver operator curve (AUC) and point statistics at a predefined threshold.

RESULTS

A total of 1,026 patients were included in the study collected between March and April 2020. Overall, there was disease prevalence of 9.6% in the population under study during this time frame. The logistic regression model was found to have an AUC of 0.89 (95% confidence interval [CI] = 0.84 to 0.94) when including four features: exposure history, temperature, white blood cell count (WBC), and chest radiograph result. Random forest method resulted in AUC of 0.86 (95% CI = 0.79 to 0.92) and gradient boosting had an AUC of 0.85 (95% CI = 0.79 to 0.91). With a consistently held negative predictive value, the logistic regression model had a positive predictive value of 0.29 (0.2-0.39) compared to 0.2 (0.14-0.28) for random forest and 0.22 (0.15-0.3) for the gradient-boosted method.

CONCLUSION

The derived predictive models offer good discriminating capacity for COVID-19 disease and provide interpretable and usable methods for those providers caring for these patients at the important crossroads of the community and the health system. We found utilization of the logistic regression model utilizing exposure history, temperature, WBC, and chest X-ray result had the greatest discriminatory capacity with the most interpretable model. Integrating a predictive model-based approach to COVID-19 testing decisions and patient care pathways and locations could add efficiency and accuracy to decrease uncertainty.

Count me in: using a patient portal to minimize implicit bias in clinical research recruitment

OBJECTIVE

Determine whether women and men differ in volunteering to join a Research Recruitment Registry when invited to participate via an electronic patient portal without human bias.

MATERIALS AND METHODS

Under-representation of women and other demographic groups in clinical research studies could be due either to invitation bias (explicit or implicit) during screening and recruitment or by lower rates of deciding to participate when offered. By making an invitation to participate in a Research Recruitment Registry available to all patients accessing our patient portal, regardless of demographics, we sought to remove implicit bias in offering participation and thus independently assess agreement rates.

RESULTS

Women were represented in the Research Recruitment Registry slightly more than their proportion of all portal users (n = 194 775). Controlling for age, race, ethnicity, portal use, chronic disease burden, and other questionnaire use, women were statistically more likely to agree to join the Registry than men (odds ratio 1.17, 95% CI, 1.12-1.21). In contrast, Black males, Hispanics (of both sexes), and particularly Asians (both sexes) had low participation-to-population ratios; this under-representation persisted in the multivariable regression model.

DISCUSSION

This supports the view that historical under-representation of women in clinical studies is likely due, at least in part, to implicit bias in offering participation. Distinguishing the mechanism for under-representation could help in designing strategies to improve study representation, leading to more effective evidence-based recommendations.

CONCLUSION

Patient portals offer an attractive option for minimizing bias and encouraging broader, more representative participation in clinical research.

Pilot evaluation of sensitive data segmentation technology for privacy

BACKGROUND

Consent2Share (C2S) is an open source software created by the Office of the National Coordinator Data Segmentation for Privacy initiative to support electronic health record (EHR) granular segmentation. To date, there are no published formal evaluations of Consent2Share.

METHOD

Structured data (e.g. medications) codified using standard clinical terminologies (e.g. RxNorm) was extracted from the EHR of 36 patients with behavioral health conditions from study sites. EHRs were available through a health information exchange and two sites. The EHR data was already classified into data types (e.g. procedures and services). Both Consent2Share and health providers classified EHR data based on value sets (e.g. mental health) and sensitivity (e.g. not sensitive. Descriptive statistics and Chi-square analysis were used to compare differences between data categorizations.

RESULTS

From the resulting 1,080 medical records items, 584 were distinct. Significant differences were found between sensitivity classifications by Consent2Share and providers (χ2 (2, N = 584) = 114.74, p = <0.0001). Sensitivity comparisons led to 56.0 % of agreements, 31.2 % disagreements, and 12.8 % partial agreements. Most (97.8 %) disagreements resulted from information classified as not sensitive by Consent2Share, but sensitive by provider (e.g. behavioral health prevention education service). In terms of data types, most disagreements (57.1 %) focused on procedures and services information (e.g. ligation of fallopian tube). When considering value sets, most disagreements focused on genetic data (100.0 %), followed by sexual and reproductive health (88.9 %).

CONCLUSIONS

There is a need to further validate Consent2Share before broad use in health care settings. The outcomes from this pilot study will help guide improvements in segmentation logic of tools like Consent2Share and may set the stage for a new generation of personalized consent engines.

SNOMED CT Concept Hierarchies for Computable Clinical Phenotypes From Electronic Health Record Data: Comparison of Intensional Versus Extensional Value Sets

Background

Defining clinical phenotypes from electronic health record (EHR)–derived data proves crucial for clinical decision support, population health endeavors, and translational research. EHR diagnoses now commonly draw from a finely grained clinical terminology—either native SNOMED CT or a vendor-supplied terminology mapped to SNOMED CT concepts as the standard for EHR interoperability. Accordingly, electronic clinical quality measures (eCQMs) increasingly define clinical phenotypes with SNOMED CT value sets. The work of creating and maintaining list-based value sets proves daunting, as does insuring that their contents accurately represent the clinically intended condition.

Objective

The goal of the research was to compare an intensional (concept hierarchy-based) versus extensional (list-based) value set approach to defining clinical phenotypes using SNOMED CT–encoded data from EHRs by evaluating value set conciseness, time to create, and completeness.

Methods

Starting from published Centers for Medicare and Medicaid Services (CMS) high-priority eCQMs, we selected 10 clinical conditions referenced by those eCQMs. For each, the published SNOMED CT list-based (extensional) value set was downloaded from the Value Set Authority Center (VSAC). Ten corresponding SNOMED CT hierarchy-based intensional value sets for the same conditions were identified within our EHR. From each hierarchy-based intensional value set, an exactly equivalent full extensional value set was derived enumerating all included descendant SNOMED CT concepts. Comparisons were then made between (1) VSAC-downloaded list-based (extensional) value sets, (2) corresponding hierarchy-based intensional value sets for the same conditions, and (3) derived list-based (extensional) value sets exactly equivalent to the hierarchy-based intensional value sets. Value set conciseness was assessed by the number of SNOMED CT concepts needed for definition. Time to construct the value sets for local use was measured. Value set completeness was assessed by comparing contents of the downloaded extensional versus intensional value sets. Two measures of content completeness were made: for individual SNOMED CT concepts and for the mapped diagnosis clinical terms available for selection within the EHR by clinicians.

Results

The 10 hierarchy-based intensional value sets proved far simpler and faster to construct than exactly equivalent derived extensional value set lists, requiring a median 3 versus 78 concepts to define and 5 versus 37 minutes to build. The hierarchy-based intensional value sets also proved more complete: in comparison, the 10 downloaded 2018 extensional value sets contained a median of just 35% of the intensional value sets’ SNOMED CT concepts and 65% of mapped EHR clinical terms.

Conclusions

In the EHR era, defining conditions preferentially should employ SNOMED CT concept hierarchy-based (intensional) value sets rather than extensional lists. By doing so, clinical guideline and eCQM authors can more readily engage specialists in vetting condition subtypes to include and exclude, and streamline broad EHR implementation of condition-specific decision support promoting guideline adherence for patient benefit.

On beyond Gruber: "Ontologies" in today's biomedical information systems and the limits of OWL

The word "ontology" was introduced to information systems when only closed-world reasoning systems were available. It was "borrowed" from philosophy, but literal links to its philosophical meaning were explicitly disavowed. Since then, open-world reasoning systems based on description logics have been developed, OWL has become a standard, and philosophical issues have been raised. The result has too often been confusion. The question "What statements are ontological" receives a variety of answers. A clearer vocabulary that is better suited to today's information systems is needed. The project to base ICD-11 on a "Common Ontology" required addressing this confusion. This paper sets out to systematise the lessons of that experience and subsequent discussions. We explore the semantics of open-world and closed-world systems. For specifying knowledge bases and software, we propose "invariants" or, more fully, "the first order invariant part of the background domain knowledge base" as an alternative to the words "ontology" and "ontological." We discuss the role and limitations of OWL and description logics and how they are complementary to closed world systems such as frames and to less formal "knowledge organisation systems". We illustrate why the conventions of classifications such as ICD cannot be formulated directly in OWL, but can be linked to OWL knowledge bases by queries. We contend that while OWL and description logics are major advances for representing invariants and terminologies, they must be combined with other technologies to represent broader background knowledge faithfully. The ICD-11 architecture is one approach. We argue that such hybrid architectures can and should be developed further.