Combination Of Diagnostic Codes Research Articles

Development and Internal Validation of the Cleveland Clinic Bleeding Model to Predict Major Bleeding Risk at Admission in Medical Inpatients

BackgroundGuidelines recommend pharmacological VTE prophylaxis for acutely ill medical patients at acceptable bleeding risk, but only the International Medical Prevention Registry on Venous Thromboembolism (IMPROVE) model has been validated for bleeding risk assessment. ObjectiveWe developed and internally validated a risk assessment model (RAM) to predict major in-hospital bleeding using risk factors at admission and compared our model to IMPROVE. MethodsWe selected patients admitted to medical services at 10 hospitals in the Cleveland Clinic Health System from 2017 to 2020. We identified major bleeding according to the International Society on Thrombosis and Hemostasis criteria, using a combination of diagnostic codes and laboratory values, and confirmed events with chart review. We fit a LASSO logistic regression model in the training set and compared the discrimination and calibration of our model and IMPROVE in the validation set. ResultsAmong 46,314 admissions, 268 (0.58%) had a major bleed. The final RAM included 16 risk factors, of which prior bleeding (OR = 4.83), peptic ulcer (OR = 3.82), history of sepsis (OR = 3.26), and steroid use (OR = 2.59) were the strongest. The Cleveland Clinic Bleeding Model (CCBM) had better discrimination than IMPROVE (AUC = 0.85 vs. 0.70, p < .001) and, at equivalent sensitivity (52%), categorized fewer patients as high-risk (7.2% vs. 11.8%, p < .001). Calibration was adequate (Brier score = 0.0057). ConclusionUsing a large population of medical inpatients with verified major bleeding events, we developed and internally validated a RAM for major bleeding whose performance surpassed the IMPROVE model.

Subclinical hypothyroidism in Wales from 2000 to 2021: A descriptive cohort study based on electronic health records.

Subclinical hypothyroidism (SCH) is a biochemical thyroid disorder characterised by elevated levels of Thyroid Stimulating Hormone (TSH) together with normal levels of thyroid hormones. Evidence on the benefits of treatment is limited, resulting in persistent controversies relating to its clinical management. This study describes the demographic and clinical characteristics of patients identified as having subclinical hypothyroidism in Wales between 2000 and 2021, the annual cumulative incidence during this period and the testing and treatment patterns associated with this disorder. We used linked electronic health records from SAIL Databank. Eligible patients were identified using a combination of diagnostic codes and Thyroid Function Test results. Descriptive analyses were then performed. 199,520 individuals (63.8% female) were identified as having SCH, 23.6% (n = 47,104) of whom received levothyroxine for treatment over the study period. The median study follow-up time was 5.75 person-years (IQR 2.65-9.65). Annual cumulative incidence was highest in 2012 at 502 cases per 100,000 people. 92.5% (n = 184,484) of the study population had TSH levels between the upper limit of normal and 10mIU/L on their first test. 61.9% (n = 5,071) of patients identified using Read v2 codes were in the treated group. 41.9% (n = 19,716) of treated patients had a history of a single abnormal test result before their first prescription. In Wales, the number of incident cases of SCH has risen unevenly between 2000 and 2021. Most of the study population had mild SCH on their index test, but more than a third of the identified patients received levothyroxine after a single abnormal test result. Patients with clinically recorded diagnoses were more likely to be treated. Given the expectation of steadily increasing patient numbers, more evidence is required to support the clinical management of subclinical hypothyroidism.

One-year prevalence of cluster headache, hemicrania continua, paroxysmal hemicrania and SUNCT in Norway: a population-based nationwide registry study

BackgroundThere is lack of population-based studies evaluating the prevalence of paroxysmal hemicrania, hemicrania continua and short-lasting unilateral neuralgiform headache attacks.ObjectivesThe aim of this study was to investigate the gender-specific 1-year prevalence of cluster headache, paroxysmal hemicrania, hemicrania continua, and short-lasting unilateral neuralgiform headache attacks.MethodsA nationwide study was conducted from January 1 2022 and December 31 2022 by linking diagnostic codes from Norwegian Patient Registry and prescription of relevant drugs from Norwegian Prescription Database on an individual basis. The 1-year prevalence with 95% confidence intervals (CI) of cluster headache, paroxysmal hemicrania, hemicrania continua and short-lasting unilateral neuralgiform headache attacks are estimated based on the combination of diagnostic codes, prescription of drugs and corresponding reimbursement codes.ResultsAmong 4,316,747 individuals aged ≥ 18 years, the 1-year prevalence per 100,000 was 14.6 (95% CI 13.5–15.8) for cluster headache, 2.2 (95% CI 1.8–2.7) for hemicrania continua, 1.4 (95% CI 1.0–1.8) for paroxysmal hemicrania, and 1.2 (95% CI 0.8–1.4) for short-lasting unilateral neuralgiform headache attacks. For all the trigeminal autonomic cephalalgies, cluster headache included, the prevalence was higher for women than men.ConclusionsIn this nationwide register-based study, we found a 1-year prevalence per 100,100 of 14.6 for cluster headache, 2.2 for hemicranias continua, 1.4 for paroxysmal hemicranias, and 1.2 for short-lasting unilateral neuralgiform headache attacks. This is the first study reporting higher prevalence of cluster headache for women than men.

Association of Antepartum and Postpartum Air Pollution Exposure With Postpartum Depression in Southern California

Women are especially vulnerable to mental health matters post partum because of biological, emotional, and social changes during this period. However, epidemiologic evidence of an association between air pollution exposure and postpartum depression (PPD) is limited. To examine the associations between antepartum and postpartum maternal air pollution exposure and PPD. This retrospective cohort study used data from Kaiser Permanente Southern California (KPSC) electronic health records and included women who had singleton live births at KPSC facilities between January 1, 2008, and December 31, 2016. Data were analyzed between January 1 and May 10, 2023. Ambient air pollution exposures were assessed based on maternal residential addresses using monthly averages of particulate matter less than or equal to 2.5 μm (PM2.5), particulate matter less than or equal to 10 μm (PM10), nitrogen dioxide (NO2), and ozone (O3) from spatial interpolation of monitoring station measurements. Constituents of PM2.5 (sulfate, nitrate, ammonium, organic matter, and black carbon) were obtained from fine-resolution geoscience-derived models based on satellite, ground-based monitor, and chemical transport modeling data. Participants with an Edinburgh Postnatal Depression Scale score of 10 or higher during the 6 months after giving birth were referred to a clinical interview for further assessment and diagnosis. Ascertainment of PPD was defined using a combination of diagnostic codes and prescription medications. The study included 340 679 participants (mean [SD] age, 30.05 [5.81] years), with 25 674 having PPD (7.54%). Increased risks for PPD were observed to be associated with per-IQR increases in antepartum and postpartum exposures to O3 (adjusted odds ratio [AOR], 1.09; 95% CI, 1.06-1.12), PM10 (AOR, 1.02; 95% CI, 1.00-1.04), and PM2.5 (AOR, 1.02; 95% CI, 1. 00-1.03) but not with NO2; PPD risks were mainly associated with PM2.5 organic matter and black carbon. Overall, a higher risk of PPD was associated with O3 during the entire pregnancy and postpartum periods and with PM exposure during the late pregnancy and postpartum periods. The study findings suggest that long-term exposure to antepartum and postpartum air pollution was associated with higher PPD risks. Identifying the modifiable environmental risk factors and developing interventions are important public health issues to improve maternal mental health and alleviate the disease burden of PPD.

Assessment of a modified MINDS-based protocol for management of alcohol withdrawal syndrome on an inpatient medical service.

To determine if a novel symptom-based alcohol withdrawal syndrome (AWS) protocol in a US Veterans cohort leads to significant clinical improvements in patient outcomes and safety. Prior studies of AWS management, oftentimes using the revised version of the Clinical Institute Withdrawal Assessment for Alcohol (CIWA-Ar) index, have demonstrated the effectiveness of symptom-triggered therapy for AWS. The Minnesota Detoxification Scale (MINDS) is an alternative to the CIWA-Ar index but remains unevaluated outside of the intensive care unit (ICU) setting. This study assesses outcomes in AWS management prior to and after the implementation of a novel MINDS-based AWS protocol (SDAWP) utilizing a revised MINDS index (MINDS-rev) in an inpatient medical ward setting. Retrospective cohort study including encounters prior to (n = 342) and after (n = 338) the implementation of the protocol. Pre- and post-protocol encounters were selected by combinations of diagnostic codes and charting elements. Outcome measures of AWS management were obtained in both groups. The primary endpoint was median total benzodiazepine exposure. Secondary outcomes included median length of hospitalization, median duration of benzodiazepine administration, and the incidence of complications. The median total benzodiazepine exposure in the post-SDAWP group was significantly lower than the pre-SDAWP group (21.2 vs. 12.0 mg, p < 0.0001) and for a significantly shorter median duration of time (4.0 vs. 3.0 days, p < 0.0001). There was no significant difference in the median length of stay (4.0 vs. 4.0 days, p = 0.50). The incidence of delirium tremens (21 vs. 7, p = 0.01) and need for transfer to a higher level of care (33 vs. 12, p = 0.002) was significantly lower in the post-SDAWP group. The SDAWP has provided significant improvements in AWS management in our institution and may potentially serve as a template for wider use in other inpatient settings.

ODP008 Nonalcoholic Fatty Liver Disease (NAFLD) Impacts Weight Loss after Bariatric Surgery

Abstract Background Gastric bypass (GB) and sleeve gastrectomy (SG) remain the most effective intervention for weight loss in obesity. Our previous data showed a significant effect of preexisting diabetes on weight loss over 5 years. Since nonalcoholic fatty liver disease (NAFLD) is implicated as a precursor to clinical diabetes in obesity, we sought to determine if preexisting NAFLD modified weight loss outcomes. Methods 714 patients [79. 0% females, median age: 45 (37-53)] undergoing bariatric surgery were followed for 5 years. Clinically significant NAFLD at baseline was defined with a composite definition, combining the presence diagnosis on the medical record, definitive evidence for fat on imaging studies, presence of biopsy proving NAFLD, and/or elevation of liver function tests without another explanation. To screen for the presence of NAFLD, a natural language-processing tool EMERSE was utilized, and results were verified manually. Presence of diabetes, hypertension, and dyslipidemia were also defined using a combination of diagnostic codes, natural language processing and manual verification. Weight loss was tracked during annual visits and calculated using baseline weight and BMI. Total percent weight loss from baseline (%TWL) was also calculated. Results Of the 714 patients, 380 underwent GB [80.3% females, age 45 (35-53), median BMI 46.4 (41.6-51.9)] and 334 underwent SG [77.5% females, age 46 (39-54), BMI 49.7 (43.7-54.1)]. Overall, 31. 0% patients had clinically significant NAFLD at baseline, while the prevalence of diabetes, hypertension and dyslipidemia were 36. 0%, 54.9%, and 29.4% respectively. %TWL was impacted significantly by NAFLD in the GB group, as individuals without NAFLD lost medians of32.30 (26. 07-37.35), 32.33 (25. 04-39.21),31.12 (21.89-37.64), 26.40 (19.96-36.20), and 25.86 (18.16-34.79)%TWL versus29.63 (24.14-35.28),27.88 (21.44-35.11),24.15 (18.80-34.42),22. 09 (17. 03-30. 07), and21. 06 (14.92-30.25)%TWL compared to those with NAFLD by years1, 2, 3, 4, and 5 (p=0. 014,0. 002,0. 001,0. 002, and 0. 008). Among the SG group, differences in %TWL in NAFLD and non-NAFLD groupswere also significant, but only during year 1 and 2; non-NAFLD group had medians of25.73 (21.36-31.97) and24.55 (18.35-31.24) %TWL vs,22.48 (17.10-27.63) and22.28 (14.75-28.31)%TWL in NAFLD-group (p=&amp;lt;0. 001 and 0. 02). In a multivariable model for %TWL, preexisting NAFLD remained a highly significant covariate when adjusted for baseline BMI, age, gender, surgery type, time (year of follow up), time by surgery type, preexisting diabetes, baseline cholesterol and triglyceride levels (Beta; 95% CI: - 2.288; -3.938, -0.639 p=0. 0067). Of note, the only other significant covariate with regards to comorbidities was presence of baseline diabetes (Beta; 95% CI: -2.811; -4.602, -1. 021, p=0. 0022). Conclusions Preexisting NAFLD impacts weight loss outcomes in patients undergoing bariatric surgery, more so in GB than SG. Mechanisms for the role of clinically significant NAFLD or diabetes in reducing weight loss potential after bariatric surgery require further studies. Presentation: No date and time listed

Comparing Electronic Health Record Domains' Utility to Identify Transgender Patients.

Earlier literature has reported on the utility of diagnostic codes and demographic information for identifying transgender patients. We aim to assess which method identifies the most transgender patients utilizing readily available tools from within the electronic health record (EHR). A de-identified patient database from a single EHR that allows for searching any discrete data point in the EHR was used to query International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) diagnostic codes and demographic data specific to transgender patients from January 2011 to April 2019. Demographic data and ICD-10 codes yielded 1494 individual EHRs with transgender-specific data domains. ICD-10 diagnostic codes alone identified 942 (63.05%) unique EHRs. Demographics alone identified 218 (14.59%) unique EHRs. A total of 334 (22.36%) unique EHRs had both ICD-10 and demographic identifiers. Of those identified by transgender-specific demographic data (552), 294 (53.26%) were trans masculine, 215 (38.95%) were trans feminine, and 43 (7.79%) were nonbinary. Of the 552 demographic-identified transgender patients, 141 (25.86%) were identified by a two-part gender identity demographic question. ICD-10 diagnostic codes, not demographic data, identified the most transgender patient records, but neither diagnostic codes alone nor demographic data captured the full population. Only 26.36% of the charts identified as transgender patients had both ICD-10 codes and demographic data. We recommend that when identifying transgender populations through EHR domains, a combination of diagnostic codes and demographic data be used. Furthermore, research is needed to optimize disclosure and collection of demographic information for gender minority populations.

Temporal trends of catheter ablation for patients with atrial fibrillation: A Korean nationwide population-based study.

Catheter ablation (CA) for atrial fibrillation (AF) is used as a treatment to restore and maintain sinus rhythm in patients with AF. However, limited data exist regarding the temporal trends of AF ablation in Asia. This study aimed to describe the temporal trends of CA for AF in Korean over 11 years. The nationwide claims database in Korea was utilized. Patients underwent CA for AF were identified using combinations of diagnostic codes, claims history, and procedure codes. Comorbidities and complications were also identified, and their temporal trends were evaluated. The numbers of patients underwent CA for AF were observed to gradually increased over 11 years (452 patients in 2007 vs. 3035 patients in 2017). Mean age of the study population increased (55.4 in 2007-2010 vs. 58.9 in 2015-2017); and mean CHA2 DS2 -VASc score also increased (1.9 in 2007-2010 vs. 2.2 in 2015-2017). Risks of complications decreased during the study period but risks of all-cause deaths did not changed significantly. Older age, women, hypertension, cerebrovascular accident, chronic obstructive pulmonary disease, chronic kidney disease, general anesthesia, and small procedure volume were independent predictors of complications but, only diabetes and occurrence of any complication were associated with mortality after CA. CA for AF has become an increasingly important treatment option. Although the proportion of high-risk patients increased, risks of complications decreased over time. Performing procedure without complications and prompt managements are essential to improve the outcome of the patients with AF underwent CA.

PRESCRIPTIONS OF OPIOID-CONTAINING DRUGS IN PATIENTS WITH CHRONIC COUGH

SESSION TITLE: Wednesday Electronic Posters 5 SESSION TYPE: Original Inv Poster Discussion PRESENTED ON: 10/23/2019 09:45 AM - 10:45 AM PURPOSE: Chronic cough (CC) of eight or more weeks affects up to 10% of adults and may lead to numerous treatments, including opioid-containing cough suppressants (OCCS), but the use of OCCS to treat CC has not been characterized. We identified and characterized the utilization of such drugs among a cohort of patients. METHODS: Using a combination of diagnostic codes, information about prescriptions, and natural language processing to analyze text, we used electronic health records to identify patients aged 18 to 85 years who had cough and received care in a Midwestern academic medical institution between 2006 and 2015. The index date was the first date with cough during the study period. We identified patients with cough as CC (three or more encounters with cough within 56-120 days) or non-CC (one or two encounters). OCCS were opioids containing codeine, dihydrocodeine, or hydrocodone. Descriptive statistics identified the CC cohort’s age, gender, race, primary insurance, and prescriptions of opioids. Demographic comparisons were made to the non-CC cohort using Student’s t-test or Pearson’s chi-square test as appropriate. Multivariable modeling of number of prescriptions was conducted using zero-inflated Poisson models. RESULTS: Compared to non-CC (N=445,116), the CC cohort (N=25,593) had more women (65% vs. 58%) and Medicare (37% vs. 22%), and was older (mean age 54 vs. 47 years) and more white (72% vs. 64%). The CC cohort had 18% African-Americans, 30% commercial insurance, and 19% Medicaid. In the first year after the index date, OCCS prescriptions represented 71% of 21,627 antitussive orders, and 22% of patients (vs. 6.3% of those with non-CC). Among these patients, 35% had at least three OCCS prescriptions. In the CC cohort, OCCS drugs were ordered in 39% with Medicaid, 19% with Medicare, and 16% with commercial insurance. The number of OCCS prescriptions per 100 patients was 60 for CC (total of 15,274) vs. 12 for non-CC. More than ten OCCS prescriptions were ordered for 157 (0.6%) of the patients. In multivariable modeling adjusting for cohort, age, race, gender, urbanicity, insurance, and year, patients were more likely to have at least one OCCS prescription if they had CC (OR 2.9 compared to non-chronic cough), or were non-white (OR 1.2), urban (OR 1.2), or with Medicaid (OR 2.4) or self-pay insurance (OR 1.2, compared to commercial insurance). More OCCS orders occurred in patients with CC (1.8-fold increase compared to non CC) or self-pay insurance (1.4-fold). CONCLUSIONS: About one in five patients with CC received an OCCS prescription, which was more likely in this cohort than in patients with non-CC. Among patients with CC, OCCS prescriptions were more likely among patients with Medicaid, compared to those with other primary insurance. Most antitussive prescriptions for patients with CC contained opioids. CLINICAL IMPLICATIONS: Findings could foster improvements in effectiveness of treating CC. DISCLOSURES: No relevant relationships by Paul Richard Dexter, source=Admin input Employee relationship with Merck & Co., Inc. Please note: $20001 - $100000 Added 02/15/2019 by Ishita Doshi, source=Web Response, value=Salary No relevant relationships by Ashley Griffith, source=Web Response No relevant relationships by Ziyue Liu, source=Admin input No relevant relationships by Anna Roberts, source=Web Response Employee relationship with Merck Please note: $1-$1000 Added 03/15/2019 by Eric Sarpong, source=Admin input, value=Salary Employee relationship with Merck & Co., Inc. Please note: >$100000 Added 03/16/2019 by Jonathan Schelfhout, source=Web Response, value=Salary Employee relationship with Merck & Co. INc. Please note: >$100000 Added 03/12/2019 by Jessica Weaver, source=Web Response, value=Salary Grantee relationship with Merck & Co., Inc. Please note: >$100000 Added 03/04/2019 by Michael Weiner, source=Web Response, value=Grant/Research Support

Association of Clinician Behaviors and Weight Change in School-Aged Children

This study uses clinical practice data to determine whether recommended weight management clinician behaviors are associated with weight status improvement in children aged 6-12 years who are overweight or obese. Electronic health record data (2009-2014) from 52 clinics were used. Weight status was examined from 1 visit to the next as dichotomous improvement (versus worsening or no change) and change in percentage overweight (over sex/age-specific BMI95). The primary predictor was a clinician behavior variable denoting attention to high BMI alone or with assessment of medical risk/comorbidities and was defined using combinations of diagnostic codes and electronic health record orders. Covariates included time between visits and medications associated with weight gain or loss. Adjusted multilevel regression models examined the association of the clinician behavior variable with weight status improvement. Analyses were conducted from 2015 to 2018. Children (n=7,205) had a mean age of 8.9 years; 45.5% were overweight, 54.5% obese, and 81.1% publicly insured. For 62% of overweight children, and 38%, 21%, and 11% of those in obesity classes 1-3, respectively, no attention to high BMI/medical risk assessment at any visit was identified. Children with evidence of clinician attention to high BMI alone and who underwent a medical risk assessment had significantly greater AOR of improvement in percentage of BMI95 and percentage of BMI95 change: BMI alone, AOR=1.2 (p<0.001) and β= -0.3 (p>0.05); BMI/medical risk, AOR=1.2 and β= -0.5 (both p<0.001). Other factors associated with weight status improvement included prescription medications (1 or more prescriptions associated with either weight loss or none associated with weight gain) and fewer months between visits. This is the first study to use electronic health record data to demonstrate that widely recommended clinician behaviors are associated with weight status improvement in children aged 6-12 years who are overweight or obese.

Developing Phenotypes from Electronic Health Records for Chronic Disease Surveillance

ObjectiveTo utilize clinical data in Electronic Health Records (EHRs) to develop chronic disease phenotypes appropriate for conducting population health surveillance.IntroductionChronic diseases, including hypertension, type 2 diabetes mellitus (diabetes), obesity, and hyperlipidemia, are some of the leading causes of morbidity and mortality in the United States. Monitoring disease prevalence guides public health programs and policies that help prevent this burden. EHRs can supplement traditional sources of chronic disease surveillance, such as health surveys and administrative claims datasets, by offering near real-time data, large sample sizes, and a rich source of clinical data. However, few studies have provided clear, consistent EHR phenotypes that were developed to inform population health surveillance.MethodsRetrospective EHR data were obtained for patients seen at New York University Langone Health in 2017 (n=1,397,446). To better estimate chronic disease burden among New York City (NYC) adults, the patient population was limited to NYC residents aged 20 or older, who were seen in the ambulatory primary care setting (n=153,653). Rule-based algorithms for identifying patients with hypertension, statin-eligibility, diabetes, and obesity were developed based on a combination of diagnostic codes, lab results or vitals, and relevant prescriptions. We compared the performance of our metric definitions to selected phenotypes from the literature using percent agreement and Cohen’s kappa. Patients with discordant disease classifications between the two sets of definitions were analyzed through natural language processing (NLP) on the patients’ 2017 medical notes using a support vector machine model. Statin-eligibility is a novel phenotype and therefore did not have a comparable definition in the literature. Sensitivity analyses were conducted to determine how disease burden changed under alternative rules for each metric.ResultsOf 153,653 adult ambulatory care patients in 2017, an estimated 53.7% had hypertension, 12.4% had diabetes, 27.8% were obese, and 30.0% were statin-eligible under our proposed definitions. The estimated prevalence of hypertension increased from 28.1% to 53.7% when diagnostic codes were supplemented with blood pressure measurements and anti-hypertensive medications, while the estimated prevalence of diabetes increased less than one percentage point with inclusion of diabetes-related medications and elevated A1C measurements. There was high agreement between our obesity (94.5% agreement, k=0.86) and diabetes (96.2% agreement, k=0.81) definitions and selected definitions from the literature and moderate agreement between the hypertension definitions (74.8% agreement, k=0.41). NLP classification of discordant cases had greater alignment with the classification results of our definitions for both hypertension (78.0% agreement) and diabetes (71.2% agreement) but did not show strong agreement with either obesity algorithm. Sensitivity analyses did not have large impacts on prevalence estimates for any of the indicators, with all estimates within two percentage points of the final algorithms.ConclusionsOur proposed rule-based phenotypes using prescriptions, labs, and vitals improved ascertainment of conditions beyond diagnostic codes and were robust to modifications per sensitivity analyses. Results from our algorithms were highly consistent with standard phenotypes from the literature and may improve case capture for surveillance purposes. These algorithms can be replicated across diverse EHR networks and can be weighted to generate population prevalence estimates.

Efficient Genome-wide Association in Biobanks Using Topic Modeling Identifies Multiple Novel Disease Loci

Biobanks and national registries represent a powerful tool for genomic discovery, but rely on diagnostic codes that can be unreliable and fail to capture relationships between related diagnoses. We developed an efficient means of conducting genome-wide association studies using combinations of diagnostic codes from electronic health records for 10,845 participants in a biobanking program at two large academic medical centers. Specifically, we applied latent Dirichilet allocation to fit 50 disease topics based on diagnostic codes, then conducted a genome-wide common-variant association for each topic. In sensitivity analysis, these results were contrasted with those obtained from traditional single-diagnosis phenome-wide association analysis, as well as those in which only a subset of diagnostic codes were included per topic. In meta-analysis across three biobank cohorts, we identified 23 disease-associated loci with p < 1e-15, including previously associated autoimmune disease loci. In all cases, observed significant associations were of greater magnitude than single phenome-wide diagnostic codes, and incorporation of less strongly loading diagnostic codes enhanced association. This strategy provides a more efficient means of identifying phenome-wide associations in biobanks with coded clinical data.

An algorithm for identification and classification of individuals with type 1 and type 2 diabetes mellitus in a large primary care database.

BackgroundResearch into diabetes mellitus (DM) often requires a reproducible method for identifying and distinguishing individuals with type 1 DM (T1DM) and type 2 DM (T2DM).ObjectivesTo develop a method to identify individuals with T1DM and T2DM using UK primary care electronic health records.MethodsUsing data from The Health Improvement Network primary care database, we developed a two-step algorithm. The first algorithm step identified individuals with potential T1DM or T2DM based on diagnostic records, treatment, and clinical test results. We excluded individuals with records for rarer DM subtypes only. For individuals to be considered diabetic, they needed to have at least two records indicative of DM; one of which was required to be a diagnostic record. We then classified individuals with T1DM and T2DM using the second algorithm step. A combination of diagnostic codes, medication prescribed, age at diagnosis, and whether the case was incident or prevalent were used in this process. We internally validated this classification algorithm through comparison against an independent clinical examination of The Health Improvement Network electronic health records for a random sample of 500 DM individuals.ResultsOut of 9,161,866 individuals aged 0–99 years from 2000 to 2014, we classified 37,693 individuals with T1DM and 418,433 with T2DM, while 1,792 individuals remained unclassified. A small proportion were classified with some uncertainty (1,155 [3.1%] of all individuals with T1DM and 6,139 [1.5%] with T2DM) due to unclear health records. During validation, manual assignment of DM type based on clinical assessment of the entire electronic record and algorithmic assignment led to equivalent classification in all instances.ConclusionThe majority of individuals with T1DM and T2DM can be readily identified from UK primary care electronic health records. Our approach can be adapted for use in other health care settings.

Extracting Primary Open-Angle Glaucoma from Electronic Medical Records for Genetic Association Studies.

Electronic medical records (EMRs) are being widely implemented for use in genetic and genomic studies. As a phenotypic rich resource, EMRs provide researchers with the opportunity to identify disease cohorts and perform genotype-phenotype association studies. The Epidemiologic Architecture for Genes Linked to Environment (EAGLE) study, as part of the Population Architecture using Genomics and Epidemiology (PAGE) I study, has genotyped more than 15,000 individuals of diverse genetic ancestry in BioVU, the Vanderbilt University Medical Center’s biorepository linked to a de-identified version of the EMR (EAGLE BioVU). Here we develop and deploy an algorithm utilizing data mining techniques to identify primary open-angle glaucoma (POAG) in African Americans from EAGLE BioVU for genetic association studies. The algorithm described here was designed using a combination of diagnostic codes, current procedural terminology billing codes, and free text searches to identify POAG status in situations where gold-standard digital photography cannot be accessed. The case algorithm identified 267 potential POAG subjects but underperformed after manual review with a positive predictive value of 51.6% and an accuracy of 76.3%. The control algorithm identified controls with a negative predictive value of 98.3%. Although the case algorithm requires more downstream manual review for use in large-scale studies, it provides a basis by which to extract a specific clinical subtype of glaucoma from EMRs in the absence of digital photographs.

A novel method for studying the temporal relationship between type 2 diabetes mellitus and cancer using the electronic medical record.

BackgroundWe developed an algorithm for the identification of patients with type 2 diabetes and ascertainment of the date of diabetes onset for examination of the temporal relationship between diabetes and cancer using data in the electronic medical record (EMR).MethodsThe Marshfield Clinic EMR was searched for patients who developed type 2 diabetes between January 1, 1995 and December 31, 2009 using a combination of diagnostic codes and laboratory data. Subjects without diabetes were also identified and matched to subjects with diabetes by age, gender, smoking history, residence, and date of diabetes onset/reference date.ResultsThe final cohort consisted of 11,236 subjects with and 54,365 subjects without diabetes. Stringent requirements for laboratory values resulted in a decrease in the number of potential subjects by nearly 70%. Mean observation time in the EMR was similar for both groups with 13—14 years before and 5–7 years after the reference date. The two cohorts were largely similar except that BMI and frequency of healthcare encounters were greater in subjects with diabetes.ConclusionThe cohort described here will be useful for the examination of the temporal relationship between diabetes and cancer and is unique in that it allows for determination of the date of diabetes onset with reasonable accuracy.

Automated Identification of Patients With Pulmonary Nodules in an Integrated Health System Using Administrative Health Plan Data, Radiology Reports, and Natural Language Processing

Lung nodules are commonly encountered in clinical practice, yet little is known about their management in community settings. An automated method for identifying patients with lung nodules would greatly facilitate research in this area. Using members of a large, community-based health plan from 2006 to 2010, we developed a method to identify patients with lung nodules, by combining five diagnostic codes, four procedural codes, and a natural language processing algorithm that performed free text searches of radiology transcripts. An experienced pulmonologist reviewed a random sample of 116 radiology transcripts, providing a reference standard for the natural language processing algorithm. With the use of an automated method, we identified 7112 unique members as having one or more incident lung nodules. The mean age of the patients was 65 years (standard deviation 14 years). There were slightly more women (54%) than men, and Hispanics and non-whites comprised 45% of the lung nodule cohort. Thirty-six percent were never smokers whereas 11% were current smokers. Fourteen percent of the patients were subsequently diagnosed with lung cancer. The sensitivity and specificity of the natural language processing algorithm for identifying the presence of lung nodules were 96% and 86%, respectively, compared with clinician review. Among the true positive transcripts in the validation sample, only 35% were solitary and unaccompanied by one or more associated findings, and 56% measured 8 to 30 mm in diameter. A combination of diagnostic codes, procedural codes, and a natural language processing algorithm for free text searching of radiology reports can accurately and efficiently identify patients with incident lung nodules, many of whom are subsequently diagnosed with lung cancer.