Electronic Health Record Algorithm Research Articles

Abstract 4137570: Atherosclerotic Cardiovascular Disease Risks in Children: An electronic health record data analytics approach

Background and Aims: There is little data on atherosclerotic cardiovascular disease (ASCVD) risk conditions, factors, and tiers in the pediatric population. This study aimed to develop and validate electronic health record (EHR) data analytic algorithms to attain the prevalence of 24 risk conditions, 7 factors, and 4 tiers in children in a rural community. Methods: This study included all the 495,902 subjects from 0 to 18 years at the Geisinger Medical Center from 2004 to 2022. The demographic and clinical data were collected using ICD-10 codes, clinical and laboratory data, and a dual-coded computational algorithm was developed based on the AHA guideline to derive the ASCVD risk conditions, factors, and tiers. To validate the EHR algorithms, the algorithm (rater 1) and a clinician (rater 2) extracted and analyzed the data from 40 randomly selected subjects from this cohort. Kendall’s coefficient of concordance (W) and Gwet’s agreement coefficient were used to evaluate algorithm and clinician agreement within the ASCVD tiers, conditions and factors, respectively. Results: High concordance was observed between algorithmic and clinician raters for risk tiers (W=80%), conditions (80-97%), and factors (49% for blood pressure, and 92% to 97% for the rest of the risk factors). In this cohort, the median age was 3.9 years, 51.0% were male, 80.6% were Non-Hispanic/Latino, and 88.1% were White. Overall, 15.5% (N = 76,945) were at-risk, 7.0% (N = 34,530) at moderate risk, and 4.2% (N = 20,761) at high risk for ASCVD, respectively. The most prevalent risk conditions were obesity (9.1%, N = 45,074), chronic inflammatory conditions (6.8%, N = 33,685), severe obesity (2.7%, N = 13,442), insulin resistance with comorbidities (2.7%, N = 13,397), and hypertension (2.1%, N = 10,436). The most prevalent risk factors were elevated BMI (27.4%, N = 135,904), blood pressure (12.5%, N = 62,015), and fasting lipid profile (8.7%, N = 43,100). Conclusion: Our EHR data analytics algorithm is in high agreement with the clinician rater. The high, moderate, and at-risk ASCVD risk tiers are prevalent in a large contemporary pediatric cohort at Geisinger. Both ASCVD risk conditions and factors are also prevalent, especially obesity, diabetes, hypertension, and hyperlipidemia. These findings suggest the further application and refinement of these algorithms to design innovative approaches for ASCVD risk reduction and prevention in children.

A SCOPING REVIEW OF SYSTEMS-LEVEL INTERVENTIONS FOR FALL PREVENTION IN LONG-TERM CARE

Abstract Older adults residing in long-term care facilities (LTCFs) are at high risk for falls. In addition to individual resident-level interventions to prevent falls and fall-related injury (FRI), system-focused interventions are also necessary to adequately address fall prevention. Relatively little attention has been given to research involving systems-focused interventions. We aimed to synthesize studies on the effects of system-focused interventions for fall and FRI prevention in LTCFs with the goal of identifying promising strategies and gaps. We searched Medline, CINAHL, and EMBASE databases from 2017 to 2022 following the Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews guideline. We conducted a narrative synthesis to summarize included studies. In the initial review, 301 studies were identified and underwent title and abstract screening resulting in 98 articles were retrieved for full-text review. Fifteen studies were included in data extraction. Fall prevention interventions evaluated in LTCF settings include: (1) multicomponent fall prevention programs, (2) staff education programs, (3) training in safe handling/transfers, (4) environmental adaptations, (5) electronic health record algorithms and prompts, (6) local culture change, (7) video analysis of falls to change local interventions, and (8) facility-level physical activity programs. About half of reported studies reported significant effects of system-focused interventions to reduce fall and FRI in LTCFs. Multicomponent fall prevention programs and physical activity interventions are the most effective systems-level interventions, while electronic health record interventions were the least effective. Overall, little attention has been given in the literature to evaluation of environmental adaptations at the systems level.

Identification of Persons Living with Dementia Using the Electronic Health Record

AbstractBackgroundAlzheimer’s disease and related dementias (ADRD) are underdiagnosed in the primary care setting. Proactive screening for cognitive impairment and subsequent evaluation and diagnosis of patients who screen positive are not routinely and reliably performed. As a result, identification of persons living with dementia (PLWD) for enrollment in clinical programs, clinical trials, and research studies can be challenging. The electronic health record (EHR) represents a potentially viable tool to efficiently and accurate identify PLWD.MethodsIn a large academic health care system in the U.S., we performed an EPIC EHR search for PLWD with the intent of enrolling them in an outpatient dementia care management program. The EHR algorithm used the following criteria for dementia: (1) an ICD‐10 dementia diagnosis in the problem list or past medical history or (2) an FDA‐approved drug on their active medication list. We performed chart review for patients who were identified to have potential dementia using this algorithm to confirm dementia diagnosis is documented in clinical notes.ResultsA total of 985 patients were identified by EHR criteria as potentially having dementia. After chart review, 747 (75.8%) were confirmed to have dementia documented in their clinical history and progress notes. One hundred thirty (13.2%) did not have a confirmed dementia diagnosis, 58 (5.9%) had an incomplete dementia evaluation, and 50 (5.0%) had only mild cognitive impairment. In addition, 55 (5.6%) were deceased.ConclusionWe found that an EHR algorithm can identify patients with dementia in a health plan population. This model may be a useful tool for a focused recruitment of potential participants in clinical trials, research studies, and clinical programs.

Alerting providers to hospitalized persons with dementia using the electronic health record.

While patients with dementia entering the hospital have worse outcomes than those without dementia, early detection of dementia in the inpatient setting is less than 50%. We developed and assessed the positive predictive value (PPV) and feasibility of a novel electronic health record (EHR) banner to identify patients with dementia who present to the inpatient setting using data from the medical record. We developed and implemented an EHR algorithm to flag hospitalized patients age ≥65 years with potential cognitive impairment in the Epic EHR system using dementia ICD-10 codes, FDA-approved medications, and the use of the term "dementia" in the emergency department physician note. Medical records were reviewed for all patients who were flagged with an EHR banner from October 2022 to May 2023. A total of 344 individuals were identified who had a banner on their chart of which 280 (81.4%) were either diagnosed with dementia or were on an FDA-approved dementia medication. Forty-three individuals who had confirmed dementia were identified by a medication only (15.4%). Of the patients without confirmed dementia, the majority (N = 33, 9.6%) had a diagnosis of altered mental status, cognitive dysfunction, or mild cognitive impairment. Only 31 individuals (9.0%) had no indication of dementia or cognitive decline in their problem list, past medical history, or medication list. We found that it was feasible to implement an EHR algorithm for prospective dementia identification with a high PPV. These types of algorithms provide an opportunity to accurately identify hospitalized older individuals for inclusion in quality improvement projects, clinical trials, pay-for-performance programs, and other initiatives.

Performance Characteristics of a Rule-Based Electronic Health Record Algorithm to Identify Patients with Gross and Microscopic Hematuria.

Two million patients per year are referred to urologists for hematuria, or blood in the urine. The American Urological Association recently adopted a risk-stratified hematuria evaluation guideline to limit multi-phase computed tomography to individuals at highest risk of occult malignancy. To understand population-level hematuria evaluations, we developed an algorithm to accurately identify hematuria cases from electronic health records (EHRs). We used International Classification of Diseases (ICD)-9/ICD-10 diagnosis codes, urine color, and urine microscopy values to identify hematuria cases and to differentiate between gross and microscopic hematuria. Using an iterative process, we refined the ICD-9 algorithm on a gold standard, chart-reviewed cohort of 3,094 hematuria cases, and the ICD-10 algorithm on a 300 patient cohort. We applied the algorithm to Geisinger patients ≥35 years (n = 539,516) and determined performance by conducting chart review (n = 500). After applying the hematuria algorithm, we identified 51,500 hematuria cases and 488,016 clean controls. Of the hematuria cases, 11,435 were categorized as gross, 26,658 as microscopic, 12,562 as indeterminate, and 845 were uncategorized. The positive predictive value (PPV) of identifying hematuria cases using the algorithm was 100% and the negative predictive value (NPV) was 99%. The gross hematuria algorithm had a PPV of 100% and NPV of 99%. The microscopic hematuria algorithm had lower PPV of 78% and NPV of 100%. We developed an algorithm utilizing diagnosis codes and urine laboratory values to accurately identify hematuria and categorize as gross or microscopic in EHRs. Applying the algorithm will help researchers to understand patterns of care for this common condition.

Identification of opioid use disorder using electronic health records: Beyond diagnostic codes

BackgroundWe used structured and unstructured electronic health record (EHR) data to develop and validate an approach to identify moderate/severe opioid use disorder (OUD) that includes individuals without prescription opioid use or chronic pain, an underrepresented population. MethodsUsing electronic diagnosis grouper text from EHRs of ~1 million patients (2012–2020), we created indicators of OUD—with “tiers” indicating OUD likelihood—combined with OUD medication (MOUD) orders. We developed six sub-algorithms with varying criteria (multiple vs single MOUD orders, multiple vs single tier 1 indicators, tier 2 indicators, tier 3 and 4 indicators). Positive predictive values (PPVs) were calculated based on chart review to determine OUD status and severity. We compared demographic and clinical characteristics of cases identified by the sub-algorithms. ResultsIn total, 14,852 patients met criteria for one of the sub-algorithms. Five sub-algorithms had PPVs ≥0.90 for any severity OUD; four had PPVs ≥0.90 for moderate/severe OUD. Demographic and clinical characteristics differed substantially between groups. Of identified OUD cases, 31.3% had no past opioid analgesic orders, 79.7% lacked evidence of chronic prescription opioid use, and 43.5% lacked a chronic pain diagnosis. DiscussionIncorporating unstructured data with MOUD orders yielded an approach that adequately identified moderate/severe OUD, identified unique demographic and clinical sub-groups, and included individuals without prescription opioid use or chronic pain, whose OUD may stem from illicit opioids. Findings show that incorporating unstructured data strengthens EHR algorithms for identifying OUD and suggests approaches limited to populations with prescription opioid use or chronic pain exclude many individuals with OUD.

Using machine learning to develop a clinical prediction model for SSRI-associated bleeding: a feasibility study

IntroductionAdverse drug events (ADEs) are associated with poor outcomes and increased costs but may be prevented with prediction tools. With the National Institute of Health All of Us (AoU) database, we employed machine learning (ML) to predict selective serotonin reuptake inhibitor (SSRI)-associated bleeding.MethodsThe AoU program, beginning in 05/2018, continues to recruit ≥ 18 years old individuals across the United States. Participants completed surveys and consented to contribute electronic health record (EHR) for research. Using the EHR, we determined participants who were exposed to SSRIs (citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, sertraline, vortioxetine). Features (n = 88) were selected with clinicians’ input and comprised sociodemographic, lifestyle, comorbidities, and medication use information. We identified bleeding events with validated EHR algorithms and applied logistic regression, decision tree, random forest, and extreme gradient boost to predict bleeding during SSRI exposure. We assessed model performance with area under the receiver operating characteristic curve statistic (AUC) and defined clinically significant features as resulting in > 0.01 decline in AUC after removal from the model, in three of four ML models.ResultsThere were 10,362 participants exposed to SSRIs, with 9.6% experiencing a bleeding event during SSRI exposure. For each SSRI, performance across all four ML models was relatively consistent. AUCs from the best models ranged 0.632–0.698. Clinically significant features included health literacy for escitalopram, and bleeding history and socioeconomic status for all SSRIs.ConclusionsWe demonstrated feasibility of predicting ADEs using ML. Incorporating genomic features and drug interactions with deep learning models may improve ADE prediction.

FRI-134 - Automation of hepatitis C screening through electronic health record algorithm

FRI-134 - Automation of hepatitis C screening through electronic health record algorithm

Healthcare personnel interactive pathogen exposure response system.

Exposure investigations are labor intensive and vulnerable to recall bias. We developed an algorithm to identify healthcare personnel (HCP) interactions from the electronic health record (EHR), and we evaluated its accuracy against conventional exposure investigations. The EHR algorithm identified every known transmission and used ranking to produce a manageable contact list.

479. Performance Differences in Electronic Health Record Algorithm for Bipolar Disorder by Race and Ethnicity: An Assessment From the Psychemerge Diversity Initiative

479. Performance Differences in Electronic Health Record Algorithm for Bipolar Disorder by Race and Ethnicity: An Assessment From the Psychemerge Diversity Initiative

Developing electronic health record algorithms that accurately identify patients with juvenile idiopathic arthritis

BackgroundThe objective of this study was to develop an algorithm that accurately identifies juvenile idiopathic arthritis (JIA) patients in the electronic health record (EHR). MethodsAlgorithms were developed in a de-identified EHR by searching for a priori JIA ICD-9 (International Classification of Diseases, Ninth Revision) and ICD-10-CM (International Classification of Diseases, Tenth Revision, Clinical Modification) codes and JIA-related keywords. Exclusion criteria were selected to remove other autoimmune diseases. A training set of 200 patients was randomly selected from patients containing ≥1 occurrence of a JIA ICD-9 or ICD-10-CM code. Case status was determined by a rheumatology clinic note documenting a JIA diagnosis before age 20. For each algorithm, positive predictive value (PPV), sensitivity, and F-measure were determined using the training set. ResultsWe developed 103 algorithms using combinations of ICD codes, keywords, and exclusion criteria. The algorithm requiring 4 or more counts of JIA ICD-9 or ICD-10-CM codes, keywords “enthesitis” and “uveitis”, and exclusion of ICD-9 or ICD-10-CM codes for systemic lupus erythematosus, dermatomyositis, polymyositis, and dermatopolymyositis had the highest PPV of 97% in the training set with an F-measure of 87%. There were 1,131 JIA cases returned by this algorithm. We validated the highest performing algorithm in a separate cohort from the training set with a PPV of 92% and an F-measure of 75%. ConclusionWe developed and validated JIA EHR algorithms with ICD-9 and ICD-10-CM codes to accurately identify a JIA cohort. Three algorithms achieved PPVs of 97%, each with different algorithm criteria, allowing for users to select an algorithm to best fit their research needs.

Abstract 11005: The Impact of Time Censoring on Machine Learning Models Which Identify Patients With Undiagnosed Cardiac Amyloidosis

Introduction: Cardiac amyloidosis (CA) is a common cause of progressive heart failure. New therapies can improve outcomes but most CA patients remain undiagnosed and untreated. Machine learning models deployed on electronic health record (EHR) data may be able to find patients with undiagnosed CA. To date, most models have focused on identification of undiagnosed amyloid from uncensored data modalities (Fig 1). Hypothesis: We hypothesized that lack of post-diagnosis censoring when training CA models leads to poor performance in predicting patients with undiagnosed CA whereas training with appropriate time censoring improves performance. Methods: We used 41 EHR features (demographics, labs, electrocardiogram/echocardiography measurements, vitals) to train a boosted decision tree model with and without time censoring. This was applied to 112 patients with confirmed CA and 22,400 controls matched on age, sex, encounter frequency and timespan of EHR. We also compared our findings to a web-based CA algorithm that was publicly available in 2020. Results: The EHR algorithm had modestly higher performance on at-risk, time-censored patients when trained with and without time censoring (area under the receiver operating characteristic curve (AUROC) 0.84±0.09 vs 0.79±0.07). Testing on temporally uncensored data showed higher performance (AUROC: 0.91±0.05) which may be unrepresentative of deployment scenarios where post-diagnostic features are unavailable for model use. The publicly available algorithm demonstrated a similar trend when tested on uncensored data (AUROC: 0.67±0.03) as compared to an appropriately censored feature set (AUROC: 0.54±0.04). Conclusions: EHR algorithms can be trained to find patients with high risk of undiagnosed cardiac amyloidosis. These models should be evaluated on temporally censored data so that post-diagnostic features do not artificially inflate performance estimates and negatively impact real-world deployment.

Development of algorithms for determining heart failure with reduced and preserved ejection fraction using nationwide electronic healthcare records in the UK

BackgroundDetermining heart failure (HF) phenotypes in routine electronic health records (EHR) is challenging. We aimed to develop and validate EHR algorithms for identification of specific HF phenotypes, using Read codes...

An electronic health record alert increases human immunodeficiency virus screening and case identification in a high-risk emergency department population

Increased screening for HIV is required to reduce mortality and transmission. Patients with risk factors for HIV may lack access to routine care and emergency departments are an important site for screening and linkage to care. We implemented an electronic health record algorithm to identify patients meeting criteria for HIV screening. Compared to unstructured clinical judgement, the EHR alert increased the number of patients screened and case identification.

Optimized Embedded Healthcare Industry Model with Lightweight Computing Using Wireless Body Area Network

Wireless technology is offering numerous growth to develop communication systems. The Internet of Things (IoT) is combined with the sensing ecosystem to transfer and process the physical environment. Recently, IoT devices have collaborated with wireless devices to improve embedded medical applications. Many solutions are proposed to decrease the power consumption of the sensing ecosystem and support the health industry. However, optimizing the transformation of collected data with lightweight power consumption is still a burning research issue. Moreover, uncontrolled network devices and healthcare professionals are remotely accessed by such embedded systems. Thus, securing sensitive information is also a significant factor for mobile communications. Therefore, this research presents an optimized embedded healthcare industry model with lightweight computing using a wireless body area network (WBAN), aiming to lessen the control overheads and improve the power consumption in mobile e-health services. To begin, it employs an optimal learning algorithm to lower the management costs of embedded systems in order to transform and administer the electronic health record (EHR) more efficiently. Second, with the help of trustworthy gateways, it delivers a safe EHR algorithm as well as lightweight computing resources for embedded systems. The proposed model is tested with a variety of experiments and demonstrates its significant improvement over state-of-the-art techniques.

Inviting patients to identify diagnostic concerns through structured evaluation of their online visit notes.

BackgroundThe 21st Century Cures Act mandates patients’ access to their electronic health record (EHR) notes. To our knowledge, no previous work has systematically invited patients to proactively report diagnostic concerns while documenting and tracking their diagnostic experiences through EHR-based clinician note review.ObjectiveTo test if patients can identify concerns about their diagnosis through structured evaluation of their online visit notes.MethodsIn a large integrated health system, patients aged 18–85 years actively using the patient portal and seen between October 2019 and February 2020 were invited to respond to an online questionnaire if an EHR algorithm detected any recent unexpected return visit following an initial primary care consultation (“at-risk” visit). We developed and tested an instrument (Safer Dx Patient Instrument) to help patients identify concerns related to several dimensions of the diagnostic process based on notes review and recall of recent “at-risk” visits. Additional questions assessed patients’ trust in their providers and their general feelings about the visit. The primary outcome was a self-reported diagnostic concern. Multivariate logistic regression tested whether the primary outcome was predicted by instrument variables.ResultsOf 293 566 visits, the algorithm identified 1282 eligible patients, of whom 486 responded. After applying exclusion criteria, 418 patients were included in the analysis. Fifty-one patients (12.2%) identified a diagnostic concern. Patients were more likely to report a concern if they disagreed with statements “the care plan the provider developed for me addressed all my medical concerns” [odds ratio (OR), 2.65; 95% confidence interval [CI], 1.45–4.87) and “I trust the provider that I saw during my visit” (OR, 2.10; 95% CI, 1.19–3.71) and agreed with the statement “I did not have a good feeling about my visit” (OR, 1.48; 95% CI, 1.09–2.01).ConclusionPatients can identify diagnostic concerns based on a proactive online structured evaluation of visit notes. This surveillance strategy could potentially improve transparency in the diagnostic process.

Developing and Validating Methods to Assemble Systemic Lupus Erythematosus Births in the Electronic Health Record.

Electronic health records (EHRs) represent powerful tools to study rare diseases. Our objective was to develop and validate EHR algorithms to identify systemic lupus erythematosus (SLE) births across centers. We developed algorithms in a training set using an EHR with over 3 million subjects and validated the algorithms at 2 other centers. Subjects at all 3 centers were selected using ≥1 code for SLE International Classification of Diseases, Ninth Revision (ICD-9) or SLE International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Clinical Modification (ICD-10-CM) and ≥1 ICD-9 or ICD-10-CM delivery code. A subject was a case if diagnosed with SLE by a rheumatologist and had a birth documented. We tested algorithms using SLE ICD-9 or ICD-10-CM codes, antimalarial use, a positive antinuclear antibody ≥1:160, and ever checked double-stranded DNA or complement, using both rule-based and machine learning methods. Positive predictive values (PPVs) and sensitivities were calculated. We assessed the impact of case definition, coding provider, and subject race on algorithm performance. Algorithms performed similarly across all 3 centers. Increasing the number of SLE codes, adding clinical data, and having a rheumatologist use the SLE code all increased the likelihood of identifying true SLE patients. All the algorithms had higher PPVs in African American versus White SLE births. Using machine learning methods, the total number of SLE codes and an SLE code from a rheumatologist were the most important variables in the model for SLE case status. We developed and validated algorithms that use multiple types of data to identify SLE births in the EHR. Algorithms performed better in African American mothers than in White mothers.

Application of 2021 American Diabetes Association Glycemic Treatment Clinical Practice Recommendations in Primary Care.

We aimed to identify the proportion of primary care patients meeting criteria for sodium-glucose cotransporter 2 inhibitors (SGLT2is) and glucagon-like peptide 1 receptor agonists (GLP-1 RAs) for cardiorenal comorbidities per 2021 American Diabetes Association (ADA) Standards of Care recommendations using readily available electronic health record (EHR) characteristics. We applied 2021 ADA recommendations to a primary care cohort of 13,350 adults with type 2 diabetes (T2D). We found that 33% of patients with diabetes would be eligible for an SGLT2i or GLP-1 RA based on cardiorenal comorbidities, 13% of patients met criteria for an SGLT2i based on heart failure or albuminuric chronic kidney disease (CKD), and 18% of patients met criteria for either agent based on atherosclerotic cardiovascular disease or CKD with an albumin-to-creatinine ratio of ≤300 mg/g. This EHR algorithm identified one-third of primary care patients with T2D as meeting criteria for SGLT2i and GLP-1 RA based on strict comorbidity definitions according to 2021 ADA recommendations.

439 Continuous Positive Airway Pressure and Cardiovascular Risk in a Large Clinical Sample of Obstructive Sleep Apnea Patients

Abstract Introduction While studies support the beneficial short-term role of continuous positive airway pressure (CPAP) therapy on cardiometabolic risk in adults with obstructive sleep apnea (OSA), its sustained effect on cardiovascular disease (CVD) risk is unknown. CPAP use data linked to electronic health records (EHR) offer opportunities to understand the role of OSA treatment for preventing CVD. We evaluated the association between CPAP use and CVD incidence in patients referred to a sleep study at a large health system. Methods We included adult patients with available apnea-hypopnea index (AHI) between 01/2018-02/2020 in Kaiser Permanente Southern California. At baseline, eligible participants had >1 year of continuous insurance coverage allowing gaps <90 days and were free of CVD 1 year prior to OSA diagnosis. Participants were distributed into three groups: no OSA (AHI<5), OSA (AHI≥5) with any CPAP use (median[IQR] 2.5[0.7–5.0] h/day) and OSA without evidence of CPAP use. CVD incidence was defined as first occurrence of myocardial infarction, stroke, unstable angina, heart failure or CVD death, based on validated EHR algorithms. We used Cox proportional hazards models to assess the association between OSA with or without CPAP use and CVD incidence, adjusted for baseline age, sex, body mass index, race/ethnicity, Charlson comorbidity index, and use of anti-hypertensives and lipid-lowering medications. Stratified analyses were conducted based on OSA severity. Results We included 11,145 patients without OSA, 13,898 with OSA and CPAP use, and 20,884 patients without CPAP use. Median follow-up was 262 days (IQR=129–409). CVD incidence rates were, respectively, 0.26%, 0.45% and 0.56%. In adjusted models, moderate-severe OSA (AHI≥15) without CPAP use was associated with increased CVD incidence when compared to no OSA (HR=1.71; 95%CI=1.11–2.64; p=0.016). OSA with any CPAP use was associated with lower CVD incidence (HR=0.68; 95%CI=0.50–0.93; p=0.016) when compared to OSA patients with no CPAP use. Stronger effects were observed when restricting the sample to moderate-severe OSA (HR=0.56; 95%CI=0.39–0.81; p=0.002). Conclusion Our analysis in a large observational clinical sample suggests that moderate-severe OSA with no CPAP use is associated with increased CVD incidence. Moreover, OSA with CPAP use was associated with decreased CVD incidence relative to no CPAP use. Support (if any) AASM Foundation (194-SR-18;205-SR-19); AHA (20CDA35310360).

What Is the Role of Sphingosine-1-Phosphate Receptors in Pain?

<h3>Background</h3> Systemic Lupus Erythematosus (SLE) is a systemic autoimmune disease with diverse manifestations that can occur over a long period of time. Electronic health record (EHR) data presents a rich source of information that can be used to understand the varied presentation of SLE. We examined three clinical classification criteria for SLE to assess whether they could be a foundation for phenotype-based detection of patients with SLE in EHR data. <h3>Methods</h3> We assessed algorithm performance over 600 medical records from the Northwestern Medicine Electronic Data Warehouse, 472 of which had definite SLE and 128 which did not, based on chart review. We developed algorithms, based on the American College of Rheumatology (ACR), Systemic Lupus International Collaborating Clinics (SLICC) and the proposed European League Against Rheumatism/American College of Rheumatology (EULAR/ACR) classification criteria using only structured data elements (diagnosis codes (ICD9/ICD10) and lab results) to determine whether patients met the classification criteria for definite SLE. <h3>Results</h3> As shown in table 1, the overall identification rate of SLE ranged from 58% to 78% across the three algorithms. All three criteria-based algorithms had greater than 95% specificity and greater than 98% positive predictive value (PPV). Sensitivity of the algorithms ranged from 52% to 69% and negative predictive value (NPV) from 35% to 55%. The SLICC-based algorithm had the overall highest performance, detecting 78% of the patients with definite SLE as determined by chart review, with 99% PPV, 69% sensitivity, 98% specificity and 55% NPV. <h3>Conclusions</h3> The ACR-, SLICC- and proposed EULAR/ACR- based EHR algorithms all detect a significant proportion of patients that were classified as having definite SLE by chart review, with high PPV and specificity. Low NPV of all three algorithms likely reflects undetected cases of SLE resulting from low detection of clinical and laboratory criteria (such as arthralgia and ANA tests) that are not consistently documented in structured data in the medical record. Use of structured data improves portability of the algorithms to other EHR datasets, but may have reduced the ability of the algorithms to detect important/highly weighted classification criteria that are documented primarily in free text notes. All three algorithms may improve through use of natural language processing (NLP) of physician notes for criteria that were difficult to detect using only diagnosis codes and labs, but may reduce portability as a result of the customization required for NLP to be effective. <h3>Funding Source(s):</h3> National Institute of Arthritis and Musculoskeletal and Skin Diseases (R21 AR072263) and NHGRI U01HG006388