De-identified Electronic Health Record Research Articles

Performance assessment of different machine learning approaches in predicting diabetic ketoacidosis in adults with type 1 diabetes using electronic health records data.

PurposeTo assess the performance of different machine learning (ML) approaches in identifying risk factors for diabetic ketoacidosis (DKA) and predicting DKA.MethodsThis study applied flexible ML (XGBoost, distributed random forest [DRF] and feedforward network) and conventional ML approaches (logistic regression and least absolute shrinkage and selection operator [LASSO]) to 3400 DKA cases and 11 780 controls nested in adults with type 1 diabetes identified from Optum® de‐identified Electronic Health Record dataset (2007–2018). Area under the curve (AUC), accuracy, sensitivity and specificity were computed using fivefold cross validation, and their 95% confidence intervals (CI) were established using 1000 bootstrap samples. The importance of predictors was compared across these models.ResultsIn the training set, XGBoost and feedforward network yielded higher AUC values (0.89 and 0.86, respectively) than logistic regression (0.83), LASSO (0.83) and DRF (0.81). However, the AUC values were similar (0.82) among these approaches in the test set (95% CI range, 0.80–0.84). While the accuracy values >0.8 and the specificity values >0.9 for all models, the sensitivity values were only 0.4. The differences in these metrics across these models were minimal in the test set. All approaches selected some known risk factors for DKA as the top 10 features. XGBoost and DRF included more laboratory measurements or vital signs compared with conventional ML approaches, while feedforward network included more social demographics.ConclusionsIn our empirical study, all ML approaches demonstrated similar performance, and identified overlapping, but different, top 10 predictors. The difference in selected top predictors needs further research.

COVID-19 in ocrelizumab-treated people with multiple sclerosis

BackgroundThere are limited data on the impact of coronavirus disease 2019 (COVID-19) caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) on people with multiple sclerosis (MS). ObjectiveTo better understand SARS-CoV-2 infection in ocrelizumab-treated people with MS. MethodsInternal Roche/Genentech data sources: Cases of COVID-19 from ongoing Roche/Genentech clinical trials and from post-marketing use of ocrelizumab until July 31, 2020 were identified and assessed using descriptive statistics.External real-world data (RWD) source: An MS COVID-19 cohort and an ocrelizumab-treated MS COVID-19 cohort were identified and assessed from the OPTUMⓇ de-identified COVID-19 electronic health record (EHR) database. ResultsRoche/Genentech clinical trial data: There were 51 (1.3%) suspected or confirmed cases of COVID-19 identified from 4,000 patients ongoing in 10 Roche/Genentech clinical trials. Of these, 26 (51%) were confirmed COVID-19 and 25 (49%) were suspected COVID-19. Sixteen (31.4%) patients were hospitalized. COVID-19 severity was mild to moderate in most patients (35, 68.6%). Ten (19.6%) patients had severe disease and there were three (5.9%) fatal cases. Most patients (43, 84.3%) recovered or were recovering. There was no association apparent between duration of exposure to ocrelizumab and COVID-19. Among COVID-19 patients with previous serum immunoglobulin status (27/51, 52.9%), all (27/27, 100%) had IgG levels within the normal range.Roche/Genentech post-marketing safety database data: There were 307 post-marketing cases of COVID-19 in the Roche/Genentech global safety database. Of these, 263 (85.7%) were confirmed and 44 (14.3%) were suspected COVID-19. 100 (32.6%) patients were hospitalized. COVID-19 was asymptomatic, mild or moderate in 143 (46.6%) patients, severe in 52 (16.9%) patients, and critical in 15 (4.9%) patients. There were 17 (5.5%) fatal cases. Information on severity was not reported in 80 (26.1%) cases. Most patients (211, 68.7%) recovered or were recovering at the time of the report.External RWD data source: As of July 13, 2020, the OPTUMⓇ database included EHRs for almost 1.2 million patients with suspected COVID-19, 130,500 of whom met the criteria for confirmed/clinically diagnosed COVID-19. A total of 357 patients with MS with confirmed COVID-19 were identified. Forty-eight (13.4%) were treated with ocrelizumab, of whom 12 (25.0%) were hospitalized and one died (2.1%). Similar rates of hospitalization, invasive ventilation, and death were observed in the ocrelizumab-treated and non-ocrelizumab-treated MS cohorts.Across the Roche/Genentech and RWD sources assessed, age, male sex, and the presence of comorbidities such as hypertension were associated with a more severe disease course of COVID-19. There was a higher number of comorbidities present in hospitalized versus non-hospitalized patients. ConclusionsThis assessment provides evidence that COVID-19 in ocrelizumab-treated people with MS is predominantly mild to moderate in severity with most patients not requiring hospitalization; in line with data reported from the general population and MS datasets. Risk factors known to be associated with severe COVID-19 outcomes in the general population also appear to influence COVID-19 severity in ocrelizumab-treated people with MS. Case fatality rates for ocrelizumab-treated people with MS were within published ranges for the general population and other MS cohorts.

The side effect profile of Clozapine in real world data of three large mental health hospitals.

Mining the data contained within Electronic Health Records (EHRs) can potentially generate a greater understanding of medication effects in the real world, complementing what we know from Randomised control trials (RCTs). We Propose a text mining approach to detect adverse events and medication episodes from the clinical text to enhance our understanding of adverse effects related to Clozapine, the most effective antipsychotic drug for the management of treatment-resistant schizophrenia, but underutilised due to concerns over its side effects. We used data from de-identified EHRs of three mental health trusts in the UK (>50 million documents, over 500,000 patients, 2835 of which were prescribed Clozapine). We explored the prevalence of 33 adverse effects by age, gender, ethnicity, smoking status and admission type three months before and after the patients started Clozapine treatment. Where possible, we compared the prevalence of adverse effects with those reported in the Side Effects Resource (SIDER). Sedation, fatigue, agitation, dizziness, hypersalivation, weight gain, tachycardia, headache, constipation and confusion were amongst the highest recorded Clozapine adverse effect in the three months following the start of treatment. Higher percentages of all adverse effects were found in the first month of Clozapine therapy. Using a significance level of (p< 0.05) our chi-square tests show a significant association between most of the ADRs and smoking status and hospital admission, and some in gender, ethnicity and age groups in all trusts hospitals. Later we combined the data from the three trusts hospitals to estimate the average effect of ADRs in each monthly interval. In gender and ethnicity, the results show significant association in 7 out of 33 ADRs, smoking status shows significant association in 21 out of 33 ADRs and hospital admission shows the significant association in 30 out of 33 ADRs. A better understanding of how drugs work in the real world can complement clinical trials.

Assessment of Readmission in a Rural Medical Center

Understanding and predicting hospital readmission has been of interest for more than three decades. To strategically place readmission reduction resources where most beneficial, organizations use readmission risk-stratification tools. However, common tools used to assess 30-day risk do not incorporate health disparity and it is unknown how modifying currently validated tools affects their predictive value. The aims of this retrospective study were to describe the population of people who are admitted and re-admitted for hospital care in a rural population and examine the effectiveness of a common risk satisfaction tool to predict 30-day readmission in a rural population experiencing health disparities. This retrospective cohort study examined data from de-identified Electronic Health Record and included adult patients admitted to one general medicine service. The factors identified in this study that influence readmission are also identified in the literature and include number of co-morbid conditions and insurance status.&#x0D; DOI: http://doi.org/10.14574/ojrnhc.v20i2.615

ML4 Application of Machine Learning Models to Evaluate COVID-19 Related ICU Utilization in a US Population

Prediction models that can accurately estimate the risk of poor prognosis could assist providers with proactively managing the at-risk population for the novel coronavirus disease 2019 (COVID-19). This study sought to identify predictors of intensive care unit (ICU) utilization among COVID-19 patients using Optum® de-identified COVID-19 Electronic Health Record (EHR) dataset (2007-2020) and machine learning (ML) techniques.

Reviewing a Decade of Research Into Suicide and Related Behaviour Using the South London and Maudsley NHS Foundation Trust Clinical Record Interactive Search (CRIS) System.

Suicide is a serious public health issue worldwide, yet current clinical methods for assessing a person's risk of taking their own life remain unreliable and new methods for assessing suicide risk are being explored. The widespread adoption of electronic health records (EHRs) has opened up new possibilities for epidemiological studies of suicide and related behaviour amongst those receiving healthcare. These types of records capture valuable information entered by healthcare practitioners at the point of care. However, much recent work has relied heavily on the structured data of EHRs, whilst much of the important information about a patient's care pathway is recorded in the unstructured text of clinical notes. Accessing and structuring text data for use in clinical research, and particularly for suicide and self-harm research, is a significant challenge that is increasingly being addressed using methods from the fields of natural language processing (NLP) and machine learning (ML). In this review, we provide an overview of the range of suicide-related studies that have been carried out using the Clinical Records Interactive Search (CRIS): a database for epidemiological and clinical research that contains de-identified EHRs from the South London and Maudsley NHS Foundation Trust. We highlight the variety of clinical research questions, cohorts and techniques that have been explored for suicide and related behaviour research using CRIS, including the development of NLP and ML approaches. We demonstrate how EHR data provides comprehensive material to study prevalence of suicide and self-harm in clinical populations. Structured data alone is insufficient and NLP methods are needed to more accurately identify relevant information from EHR data. We also show how the text in clinical notes provide signals for ML approaches to suicide risk assessment. We envision increased progress in the decades to come, particularly in externally validating findings across multiple sites and countries, both in terms of clinical evidence and in terms of NLP and machine learning method transferability.

Abstract 12993: Oral Anticoagulation Prescription Depends on Both Atrial Fibrillation Duration and CHA 2 DS 2 -VASc Score

Introduction: Current guideline recommendations for oral anticoagulation (OAC) in patients with atrial fibrillation (AF) are based on CHA 2 DS 2 -VASc score alone. In patients with cardiac implantable electronic devices (CIEDs), it is poorly understood how the interaction between AF duration and CHA 2 DS 2 -VASc score influence the decision to prescribe OAC. Methods: Data from the Optum® de-identified Electronic Health Record dataset were linked to the Medtronic CareLink database of CIEDs. An index date was assigned as the later of 6 months after device implant or 1 year after EHR data availability. CHA 2 DS 2 -VASc score was assessed via EHR data prior to the index date. Maximum daily AF burden (No AF, 6 minutes-23.5 hours, and &gt;23.5 hours) was assessed over the 6 months prior to index date. OAC prescription rates were computed prior to the index date as a function of both AF duration and CHA 2 DS 2 -VASc score. Results: A total of 35,779 patients with CHA 2 DS 2 -VASc scores ≥1 were identified, including 8,581 who were prescribed OAC and 27,198 not prescribed OAC during the observation period. The overall OAC prescription rate among the subset of 12,938 patients with device-detected AF &gt;6 minutes was 36.7% and the rate was nearly 60% higher for patients with a maximum daily AF burden &gt;23.5 hours (45.4%) compared to those with 6 minutes-23.5 hours (28.7%). OAC prescription rates increased monotonically with both increasing AF duration and CHA 2 DS 2 -VASc score, reaching a maximum of 67.2% for patients with AF &gt;23.5 hours and a CHA 2 DS 2 -VASc score ≥5 (Table). Conclusion: In contrast with guideline recommendations, real-world prescription of OAC increased with both increasing duration of AF and CHA 2 DS 2 -VASc score. This mirrors recent evidence demonstrating that stroke risk also increases with both of these features and highlights the need for further research into the role of AF duration, stroke risk, and the need for anticoagulation in patients with devices capable of long-term AF monitoring.

Characteristics of patients with at least one lipoprotein(a) [Lp(a)] assessment and those with Lp(a) levels equal to or greater than 70 mg/dL: a real-world study in the US

Abstract Background and purpose Elevated lipoprotein(a) [Lp(a)] has been associated with increased risk of cardiovascular disease (CVD). The objective of this real-world study was to characterize the patients in the United States (US) that have an Lp(a) assessment and those with Lp(a) levels ≥70 mg/dL. Methods This is a descriptive, non-interventional, retrospective cohort study of patients with at least one Lp(a) assessment and one-year continuous enrollment prior to index date (first Lp(a) assessment in the identification period i.e. 1/1/2008–30/6/2019) using the Optum® de-identified Electronic Health Record (EHR) dataset (2007–2019) in the US. Optum's longitudinal EHR repository include more than 700 Hospitals and 7000 Clinics; treating more than 95 million patients receiving care in the US. Results 26,997 patients (50.8% females) with ≥1 Lp(a) assessment having a mean (SD) age of 52.6 (14.1) years were evaluated. Patient distribution across Lp(a) levels ≥30, ≥50, ≥70, and ≥90 mg/dL were 32.4%, 20.2%, 12.3%, and 6.6%, respectively. Patients with Lp(a) ≥70 mg/dL (N=3,314) had a mean (SD) age of 54.7 (13.2) years and 55.2% were female. The overall population that gets tested for Lp(a) and the subgroup of patients with Lp(a) ≥70 mg/dL had prior myocardial infarction (4.8% and 6.2%), prior ischemic stroke (2.2% and 2.5%), peripheral artery disease (6.7% and 8.4%), dyslipidaemia (72.6% and 80.9%), hypertension (53.0% and 60.9%), diabetes mellitus (23.5% and 26.8%) and chronic ischemic heart disease (19.6% and 24.6%) at index. The use of baseline medications included statins (37.3% and 49.1%), ACEi/ARBs (28.2% and 34.0%) and beta-blockers (19.0% and 23.6%) for the overall population and those patients with Lp(a) ≥70 mg/dL, respectively. The mean (SD) low-density lipoprotein cholesterol, triglycerides and total cholesterol levels were 109.8 (39.8) mg/dL [n=7,371], 138.3 (119.3) mg/dL [n=8,179] and 190.5 (46.8) mg/dL [n=8,181], respectively for overall population and 114.8 (41.5) mg/dL [n=980], 129.6 (80.5) mg/dL [n=1,063] and 195.9 (47.6) mg/dL [n=1,056] for patients with Lp(a) ≥70 mg/dL. Conclusions Of patients with an Lp(a) assessment in the US, more than 10% had Lp(a) levels ≥70 mg/dL. These patients were frequently diagnosed with dyslipidaemia and when reported showed elevated LDL-C and total cholesterol values. Further analyses are required to better understand any differences in patient characteristics and treatment of patients across Lp(a) levels. Funding Acknowledgement Type of funding source: Private company. Main funding source(s): Novartis Pharma AG

A real-world assessment of the burden associated with hospitalizations in US patients with heart failure and left ventricular ejection fraction (LVEF) greater than 40%

Abstract Background and purpose Heart failure (HF) is one of the leading causes of hospitalization among older adults and is associated with a large burden of disease for the individual, the patient's family, healthcare systems, and society. This study assessed the burden of hospitalizations in patients with HF with LVEF &amp;gt;40% in the United States (US). Methods This retrospective, non-interventional study identified adult patients with incident or prevalent HF in Optum® de-identified Electronic Health Record (EHR) dataset (2007–2018) between 01/01/2013 and 31/12/2017. Optum's longitudinal EHR repository is derived from dozens of healthcare provider organizations in the US, that include more than 700 Hospitals and 7000 Clinics; treating &amp;gt;95 million patients receiving care in the US. Patients were followed for up to 1 year or until last active date whichever occurred first. Comorbidities, all-cause hospitalizations (AcH) and primary cause HF hospitalizations (HFH) were analysed. Results 120,606 patients with HF and LVEF &amp;gt;40% (54% female) with a mean (SD) age of 71 (13) yrs were included, representing 80,324.74 patient-yrs follow-up (days). Common comorbidities were hypertension (91.8%), ischemic heart disease (IHD, 71.4%), atrial fibrillation (AF, 54.8%), renal disease (54.1%), type 2 diabetes (T2D, 50.7%), obesity (44.6%) and anemia (39%). Comorbidities including IHD (72.9% vs. 68.4%), AF (56.4% vs. 51.6%) and T2D (51.1% vs. 49.9%) were more often recorded in patients with LVEF &amp;gt;40-≤60% than &amp;gt;60% cohort while hypertension (91.6% vs. 92.2%), renal disease (53.8% vs. 54.6%), obesity (43.9% vs. 46.1%) and anemia (38.1% vs. 40.9%) had significantly higher frequency in the LVEF &amp;gt;60% cohort. The annualized AcH rate for patients with LVEF&amp;gt;40% was 1.44 and annualized HFH rate was 0.24 with a median length of stay of 3 and 4 days, respectively. Annualized hospitalization rates were significantly higher for women than men (both AcH and HFH). AcH rates were significantly higher and HFH rates were significantly lower for patients with LVEF&amp;gt;60% compared with LVEF &amp;gt;40-≤60. Conclusions This study demonstrates that patients with HF and LVEF &amp;gt;40% experience significant burden from comorbidities and hospitalizations from any-cause and for HF. The hospitalization rates are higher in women (both AcH and HFH) or patients with LVEF &amp;gt;60% (AcH only). Further focus on reduction of hospitalizations and interdisciplinary management of patients with HF should be warranted. Funding Acknowledgement Type of funding source: Private company. Main funding source(s): Novartis Pharma AG

Patient and clinical characteristics of heart failure patients concomitantly prescribed SGLT2 inhibitors and sacubitril/valsartan, a database cohort study using the Optum electronic health record data

Abstract Background and purpose Sacubitril/valsartan (sac/val), an angiotensin receptor neprilysin inhibitor, reduces the risk for cardiovascular (CV) death or hospitalization for heart failure (HF) in HF with reduced ejection fraction (HFrEF). Sodium-glucose cotransporter-2 inhibitors (SGLT2i) are approved in patients with type 2 diabetes (T2D) and have shown to reduce the CV risk in T2D patients with established CV or at risk of CV disease. The SGLT2i dapagliflozin has shown to improve outcomes in patients with chronic HFrEF, with or without T2D, when used in addition to standard of care including sac/val. As the use of SGLT2i in HF evolves, and given the large overlap of HF and T2D populations, it is of interest to understand the population with concomitant use of sac/val and SGLT2i. This study describes the clinical characteristics of patients treated concomitantly with sac/val and SGLT2i or concomitantly with sac/val and dipeptidyl peptidase-4 inhibitors (DPP4i) or glucagon-like peptide-1 receptor agonists (GLP1), two comparable second line anti-diabetic drug classes. Methods This retrospective non-interventional study describes two mutually exclusive adult patient cohorts diagnosed with HF and T2D concomitantly prescribed sac/val and SGLT2i (cohort 1), or concomitantly prescribed sac/val and DDP4i/GLP1 (cohort 2). The index date was defined as the first date of concomitant use with prescriptions overlapping a minimum of 21 days. Patients were identified any time between 1/1/2015 and 30/6/2019 in the Optum® de-identified electronic health record (EHR) data from providers across the continuum of care. Results 2.3 million HF patients were identified, and 41.6% had a T2D diagnosis. 560 patients were concomitantly prescribed sac/val and SGLT2i (cohort 1) and 1,566 concomitantly sac/val and DDP4/GLP1 (cohort 2). There was a higher proportion of females in cohort 2 (35.0% vs 27.9%). Mean age was higher in cohort 2 (66.4 vs 61.4 years). The mean estimated glomerular filtration rate was 85.93 (SD 23.43) ml/min/1.73m2 (cohort 1) and 72.10 (Std. 27.11) ml/min/1.73m2 (cohort 2). The proportion of stage 3 CKD (&amp;lt;60 to &amp;gt;30 ml/min/1.73m2) was 11.8% (cohort 1) and 24.4% (cohort 2). Mean systolic blood pressure was similar, 120 mmHg (cohort 1) and 122 mmHg (cohort 2). Mean hemoglobin was 13.60 g/dl (cohort 1) and 12.43 g/dl (cohort 2). Median (IQR) NT-proBNP differed between the two cohorts, 914 (2154) pg/ml (cohort 1) and 2,290 (5,301) pg/ml (cohort 2) but with complete values available in only 17.7 and 19.0% of each cohort. Conclusions This descriptive analysis of concomitant prescription of sac/val and SGLT2i or DPP4/GLP1 highlights differences in the clinical characteristics between the two cohorts. The patients treated with sac/val and SGLT2i start with a more favorable clinical profile compared to the patients treated with sac/val and DPP4/GLP1. Further analyses are needed to determine if these differences are driven by age, gender or other factors. Funding Acknowledgement Type of funding source: Private company. Main funding source(s): Novartis Pharma AG

Temporal relationship between atrial fibrillation and ischaemic stroke in a large cohort with continuous rhythm monitoring

Abstract Background/Introduction Atrial fibrillation (AF) increases the risk of ischaemic stroke, but the temporal relationship is uncertain. Purpose To assess the temporal relationship between AF episodes and ischaemic stroke in patients with cardiovascular implantable electronic devices (CIED) utilizing a case-crossover analysis. Methods We linked a very large U.S. aggregated de-identified electronic health record database (2007–2017), containing claims for ischaemic stroke hospitalizations, to a manufacturer's CIED database with continuous AF monitoring. All stroke patients with ≥120 days of pre-stroke rhythm data were included. For each stroke patient, we compared the presence of a day with ≥5.5 hours of AF (the TRENDS study threshold) in the case period (days 1–30 pre-stroke) to that in the control period (days 91–120 pre-stroke). Matched analyses for dichotomous outcomes generated odds ratios with confidence intervals and p values. Results We identified 891 ischaemic stroke patients (71.3±10.5 years, 65% male, 27% pacemakers, 60% defibrillators, 13% insertable cardiac monitors) with continuous monitoring data in the 120 days pre-stroke. The vast majority had either no AF in both the case and control periods (n=682, 77%) or AF in both periods (n=143, 16%), i.e., non-informative records. However, among the 66 patients with informative, discordant arrhythmic states, 52 had AF in the case period versus 14 in the control period, for an odds ratio of 3.71 (95% C.I. 2.06–6.70, p&amp;lt;0.001). Analysed by 5-day periods, stroke risk was markedly increased within 30 days of the AF episode (figure). For days 1–5 following an AF episode, the odds ratio was 9.7 (95% C.I. 5.9–16.1). Risk diminished towards non-AF risk after 30 days. Conclusion Our analysis of this largest cohort with continuous rhythm monitoring prior to ischaemic stroke demonstrates that stroke risk is highest within a few days of an episode of AF and diminishes rapidly thereafter. Our findings support a strategy of time-limited anticoagulation for patients with infrequent episodes of AF and available continuous rhythm monitoring. Funding Acknowledgement Type of funding source: None

A real-world assessment of the distribution of lipid levels across increasing lipoprotein(a) levels in patients with established cardiovascular disease in the United States

Abstract Background and purpose Measurements of lipid levels are frequently used to assess the risk of developing cardiovascular disease (CVD), including future myocardium infarction (MI) and ischemic stroke. Lipoprotein(a) [Lp(a)] has been established as an independent risk factor for developing CVD, however it is not widely measured. This real-world study assessed the distribution of lipid (Total Cholesterol [TC], low-density cholesterol [LDL-C], high-density cholesterol [HDL-C] and Triglycerides [TGs]) levels across different Lp(a) strata of patients with established CVD and at least one Lp(a) assessment. Methods This was a descriptive, non-interventional, retrospective cohort study of patients with established CVD and at least one Lp(a) assessment in the Optum® de-identified Electronic Health Record dataset (2007–2019). The index date was defined as the date of the first established CVD (MI, ischemic stroke or peripheral artery disease diagnosis) in the identification period (1/1/2008–30/6/2018) following one-year continuous enrollment. Patients were followed for up to one year, until death or transfer out of the dataset, whichever occurred first. Results 5,434 patients (55.5% males) with established CVD and at least one Lp(a) assessment having a mean (SD) age of 62.5 (12.6) years were evaluated. During the 12 months pre-index period, 49.9%, 46.1%, 48.1% and 49.8% of these patients had a measurement of TC, LDL-C, HDL-C and TGs, respectively. Overall, the mean (SD) TC, LDL-C, HDL-C and TGs levels were 182.1 (47.0) mg/dL [n=2,712], 102.6 (40.1) mg/dL [n=2,503], 52.7 (16.7) mg/dL [n=2,614], and 140.3 (108.8) mg/dL [n=2,707], respectively. More than half of these patients were treated with at least one lipid lowering drug (includes statins, ezetimibe, fibrates, niacin or PCSK9is) during the pre-index period. The mean (SD) levels of TC, LDL-C, HDL-C and TG in patients with Lp(a) &amp;lt;30 mg/dL, 30–50] mg/dL, 50–70 mg/dL, 70–90 mg/dL and ≥90 mg/dL are presented in Figure 1. Conclusions Overall, no trends in the levels of total cholesterol, LDL-C, HDL-C or triglycerides were observed across increasing levels of lipoprotein(a). These findings suggest the measurement of lipoprotein(a) is required to ensure a complete characterization of a patient lipid profile. Funding Acknowledgement Type of funding source: Private company. Main funding source(s): Novartis Pharma AG

Age-appropriate cancer screening: A cohort study of adults with psoriasis prescribed biologics, adults in the general population, and adults with hypertension

Psoriasis is associated with increased risk of developing and dying from cancer. To evaluate whether psoriasis patients who are prescribed biologics receive the recommended screening for cervical, breast, and colon cancer. We conducted a retrospective cohort study using the Optum deidentified Electronic Health Record data set. Incidence rates for cervical, breast, and colon cancer screening were compared between psoriasis patients who were prescribed biologics and 2 matched comparator cohorts: general patient population and patients being managed for hypertension. Multivariable Cox proportional hazards regression was performed to assess for differences in the rates of cancer screening. Compared with those in the general population without psoriasis, psoriasis patients who were prescribed biologics had higher screening rates for cervical cancer (adjusted hazard ratio [aHR] 1.09; 95% confidence interval [CI] 1.02-1.16) and colon cancer (aHR 1.10; 95% CI 1.02-1.18). Compared with those with hypertension, patients with psoriasis who were prescribed biologics had lower screening rates for breast cancer (aHR 0.88; 95% CI 0.83-0.94) and colon cancer (aHR 0.89; 95% CI 0.83-0.95). Patients with psoriasis who are prescribed biologic therapies may not be receiving adequate age-appropriate cancer screening, especially for breast and colon cancer.

Association of Social Behaviors With Community Discharge in Patients with Total Hip and Knee Replacement

ObjectivesUnderstand the association between social determinants of health and community discharge after elective total joint arthroplasty. DesignRetrospective cohort design using Optum de-identified electronic health record dataset. Setting and ParticipantsA total of 38 hospital networks and 18 non-network hospitals in the United States; 79,725 patients with total hip arthroplasty and 136,070 patients with total knee arthroplasty between 2011 and 2018. MethodsLogistic regression models were used to examine the association among pain, weight status, smoking status, alcohol use, substance disorder, and postsurgical community discharge, adjusted for patient demographics. ResultsMean ages for patients with hip and knee arthroplasty were 64.5 (SD 11.3) and 65.9 (SD 9.6) years; most patients were women (53.6%, 60.2%), respectively. The unadjusted community discharge rate was 82.8% after hip and 81.1% after knee arthroplasty. After adjusting for demographics, clinical factors, and behavioral factors, we found obesity [hip: odds ratio (OR) 0.81, 95% confidence interval (CI) 0.76–0.85; knee: OR 0.73, 95% CI 0.69–0.77], current smoking (hip: OR 0.82, 95% CI 0.77–0.88; knee: OR 0.90, 95% CI 0.85–0.95), and history of substance use disorder (hip: OR 0.55, 95% CI 0.50–0.60; knee: OR 0.57, 95% CI 0.53–0.62) were associated with lower odds of community discharge after hip and knee arthroplasty, respectively. Conclusions and ImplicationsSocial determinants of health are associated with odds of community discharge after total hip and knee joint arthroplasty. Our findings demonstrate the value of using electronic health record data to analyze more granular patient factors associated with patient discharge location after total joint arthroplasty. Although bundled payment is increasing community discharge rates, post-acute care facilities must be prepared to manage more complex patients because odds of community discharge are diminished in those who are obese, smoking, or have a history of substance use disorder.

S0459 Irritable Bowel Syndrome Is Associated With the Primary and Secondary Mast Cell Disorders

INTRODUCTION: There is increasing evidence that some patients with irritable bowel syndrome (IBS) have increased number of mast cells (MCs) or increased MC degranulation. Yet, association of IBS with primary (mast cell malignancy (MCM), mastocytosis (MA), mast cell activation syndrome (MCAS)) and secondary (chronic spontaneous urticaria (CSU)) MC disorders (MCDs) has not been studied. We investigated this using two large databases. METHODS: The Explorys database (Cleveland, OH) is an aggregate of de-identified electronic health record (EHR) data from 26 US healthcare systems. Systemized Nomenclature of Medicine (SNOMED) classification was used to identify adult patients diagnosed with both IBS and each of MCM, MA (SNOMED code “urticaria pigmentosa”), and CSU (SNOMED code “urtiaria” with exclusion of “acute urticaria” and “urticaria pigmentosa”), from 1994-2020. As there is no SNOMED code for MCAS, we identified patients with IBS and MCAS using Epic's Slicer Dicer, a self-service tool for obtaining de-identified customized EHR data, from two local academic centers. We excluded patients with inflammatory bowel disease, celiac disease, thyroid disease, autoimmune disorders and vasculitis. We collected demographics, procedural history, medication exposures, relevant comorbidities and clinical features. In each cohort, whether IBS preceded MCD or vice versa was searched using the temporal attributes feature on Explorys. RESULTS: Of >53 million included patients from Explorys, 617,140 had IBS, of which nearly 4% (N = 23,200) had a MCD. We created three overlap cohorts: IBS + CSU (N = 23,050), IBS + MA (N = 100), IBS + MCM (N = 80). Of 8 million adults in slicer dicer 67,924 had IBS and we created the fourth overlap cohort, IBS + MCAS (N = 92). IBS was strongly associated with all four MCDs (Table 1). In all four overlap cohorts, IBS and MCD were equally likely to precede the other, approximately 50% had IBS-D and 30% IBS-C. Nearly half experienced nausea, but typical symptoms of MCDs (flushing, hypotension) were poorly reported. There was strong female predominance, most carried private insurance, and while few were underweight many were obese (Table 2). Comorbidities, procedures and medication exposures common to IBS or MCDs were found more frequently than IBS alone (Table 3). CONCLUSION: MCDs are at least 4 times more likely to be present in patients with IBS. This data supports the role of MC hyperactivity in the pathogenesis of some IBS patients. Further investigation of this mechanism is warranted.Table 1Table 2Table 3

Characteristics and Outcomes of a Real-world Population with Heart Failure with Reduced Ejection Fraction Representative of Clinical Trial Patients

Introduction Heart failure with reduced ejection fraction (HFrEF) incurs significant mortality, morbidity and cost. Despite recent advances, opportunities for therapeutic improvement remains. Omecamtiv mecarbil (OM) is a first-in-class cardiac myotrope currently in a phase-3 clinical trial (GALACTIC-HF) for HFrEF treatment. Given known differences between real world (RW) and clinical trial populations, we sought to characterize RW patients who might benefit from OM. Methods In the Vanderbilt University Medical Center de-identified electronic health record of in- and outpatients (Synthetic Derivative) from 2006-2019, we created two RW cohorts of HFrEF patients: 1) Based on the validated eMERGE Network phenotype of HFrEF (left ventricular ejection fraction (LVEF) ≤ 40%; ‘clinical cohort’) and 2) Mirroring eligibility criteria of the GALACTIC-HF trial (LVEF ≤ 35%; ‘GALACTIC-HF-like cohort’) including hospitalizations, medications and laboratory values. Diagnosis and procedure codes, structured and unstructured problem lists, laboratory values, and medications were used to characterize demographics, clinical characteristics and outcomes. HF hospitalization rates were calculated from 5 days post HF index date to last encounter or death. Results The clinical cohort included 3,955 HFrEF patients and the GALACTIC-HF-like cohort included 1,541 (median age at index date 65 and 61 years, respectively), both approximately 67% male and 80% white. Aside from higher median BNP level in the GALACTIC-HF-like cohort, clinical characteristics such as blood pressure and heart rate of the two cohorts were similar. (Table). Comorbidities and utilization of cardiac devices were somewhat higher in the GALACTIC-HF-like cohort including chronic kidney disease (39 v 30%), atrial fibrillation (52 v 48%), and cardiac resynchronization or implantable cardioverter defibrillator (52 v 40%). HF hospitalization rate (per 1000py) was 242 (95% CI 203, 280) in the clinical and 396 (95% CI 350, 442) in the GALACTIC-HF like cohorts during median follow up of 2.7 and 4.1 years, respectively. Conclusions Approximately 40% of RW HFrEF patients met GALACTIC-HF trial criteria, suggesting that RW HFrEF populations may benefit from new therapies such as OM. The higher prevalence of comorbidities and higher rate of HF hospitalization in the GALACTIC-HF-like cohort may indicate a high-need, high-risk population that may achieve better clinical outcomes with OM.

Depression and Buprenorphine Treatment in Patients with Non-cancer Pain and Prescription Opioid Dependence without Comorbid Substance Use Disorders

BackgroundDepression occurs in 40% of patients with prescription opioid dependence (POD). Existing studies of the association between depression and buprenorphine (BUP) treatment for POD are inconsistent and often include patients with comorbid substance use disorders (SUD). We estimated the association between depression and BUP use in patients with pain and POD and free of comorbid SUD. MethodsOptum® de-identified Electronic Health Record dataset from 2010 to 2018 was used to identify 5,529 patients with chronic pain, with and without depression, receiving prescription opioids and free of substance use disorder diagnoses for one year before POD diagnoses. Unadjusted and adjusted Cox proportional hazard models and negative binomial regression models were computed to estimate the association between depression and time to BUP start, number of BUP prescriptions in the year after BUP start and time to >30 day BUP gap. ResultsPatients’ mean age was 52.4 (SD±15.3) years, 62% were female and 84% were white and 4.9% (n=270) started BUP. Depression was not associated with BUP initiation.. Among BUP starters, depression vs. no depression, was significantly associated with receiving 29% fewer BUP prescriptions (RR=0.71; 95%CI: 0.51-0.98) and an increased risk for > 30 day gap (HR=1.76; 95%CI:1.01-3.09). LimitationsMissing data prevented measuring BUP dose. ConclusionsDepression is likely associated with earlier BUP treatment dropout. Depression related medication non-adherence or possible worsening of depression following BUP taper could explain results. Research is needed to determine if depression severity is associated with BUP dose trajectories and multi-year BUP retention..

Using de-identified electronic health records to research mental health supported housing services: A feasibility study.

Mental health supported housing services are a key component in the rehabilitation of people with severe and complex needs. They are implemented widely in the UK and other deinstitutionalised countries but there have been few empirical studies of their effectiveness due to the logistic challenges and costs of standard research methods. The Clinical Record Interactive Search (CRIS) tool, developed to de-identify and interrogate routinely recorded electronic health records, may provide an alternative to evaluate supported housing services. The feasibility of using the Camden and Islington NHS Foundation Trust CRIS database to identify a sample of users of mental health supported accommodation services. Two approaches to data interrogation and case identification were compared; using structured fields indicating individual's accommodation status, and iterative development of free text searches of clinical notes referencing supported housing. The data used were recorded over a 10-year-period (01-January-2008 to 31-December-2017). Both approaches were carried out by one full-time researcher over four weeks (150 hours). Two structured fields indicating accommodation status were found, 2,140 individuals had a value in at least one of the fields representative of supported accommodation. The free text search of clinical notes returned 21,103 records pertaining to 1,105 individuals. A manual review of 10% of the notes indicated an estimated 733 of these individuals had used a supported housing service, a positive predictive value of 66.4%. Over two-thirds of the individuals returned in the free text search (768/1,105, 69.5%) were identified via the structured fields approach. Although the estimated positive predictive value was relatively high, a substantial proportion of the individuals appearing only in the free text search (337/1,105, 30.5%) are likely to be false positives. It is feasible and requires minimal resources to use de-identified electronic health record search tools to identify large samples of users of mental health supported housing using structured and free text fields. Further work is needed to establish the availability and completion of variables relevant to specific clinical research questions in order to fully assess the utility of electronic health records in evaluating the effectiveness of these services.

In-Hospital Therapy for Heart Failure With Reduced Ejection Fraction in the United States

ObjectivesThis study sought to characterize in-hospital treatment patterns and associated patient outcomes among patients hospitalized for heart failure (HF) in U.S. clinical practice. BackgroundHospitalizations for HF are common and associated with poor patient outcomes. Real-world patterns of in-hospital treatment, including diuretic therapy, in contemporary U.S. practice are unknown. MethodsUsing Optum de-identified Electronic Health Record data from 2007 through 2018, patients hospitalized for a primary diagnosis of HF (ejection fraction ≤40%) and who were hemodynamically stable at admission, without concurrent acute coronary syndrome or end-stage renal disease, and treated with intravenous (IV) diuretic agents within 48 h of admission were identified. Patients were categorized into 1 of 4 mutually exclusive hierarchical treatment groups defined by complexity of treatment during hospitalization (intensified treatment with mechanical support or IV vasoactive therapy, IV diuretic therapy reinitiated after discontinuation for ≥1 day without intensified treatment, IV diuretic dose increase/combination diuretic treatment without intensified treatment or IV diuretic reinitiation, or uncomplicated). ResultsOf 22,677 patients hospitalized for HF with reduced ejection fraction (HFrEF), 66% had uncomplicated hospitalizations without escalation of treatment beyond initial IV diuretic therapy. Among 7,809 remaining patients, the highest level of therapy received was IV diuretic dose increase/combination diuretic treatment in 25%, IV diuretic reinitiation in 36%, and intensified therapy in 39%. Overall, 19% of all patients had reinitiation of IV diuretic agents (26% of such patients had multiple instances), 12% were simultaneously treated with multiple diuretics, and 61% were transitioned to oral diuretic agents before discharge. Compared with uncomplicated treatment, IV diuretic reinitiation and intensified treatment were associated with significantly longer median length of stay (uncomplicated: 4 days; IV diuretic reinitiation: 8 days; intensified: 10 days) and higher rates of in-hospital (uncomplicated: 1.6%; IV diuretic reinitiation: 4.2%; intensified: 13.2%) and 30-day post-discharge mortality (uncomplicated: 5.2%; IV diuretic reinitiation: 9.7%; intensified: 12.7%). ConclusionsIn this contemporary real-world population of U.S. patients hospitalized for HFrEF, one-third of patients had in-hospital treatment escalated beyond initial IV diuretic therapy. These more complex treatment patterns were associated with highly variable patterns of diuretic use, longer hospital lengths of stay, and higher mortality. Standardized and evidence-based approaches are needed to improve the efficiency and effectiveness of in-hospital HFrEF care.

An evaluation of recurrent hypoglycemia across Chicago, Illinois

AimsRecurrent hypoglycemia is understudied. This study evaluates recurrent hypoglycemia, fragmentation of care and mortality in a large urban center. MethodsThe Chicago HealthLNK Data Repository (CHDR), a de-identified electronic health record data set from institutions across Chicago, identified 9741 patients with diabetes (DM) who had hypoglycemia (emergency department (ED) or inpatient admission (IA)) from 2006 to 2012. Recurrence was defined as more than one hypoglycemia encounter, and fragmentation of health care was defined as an ED visit or IA for hypoglycemia at >1 site. Results187,644 patients were identified with DM; of 9741 patients with hypoglycemia, 2857 (29.3%) had recurrence. Patients with ≥4 hypoglycemic encounters (n = 1035) represented 10.6%, but accounted for 40.3% hypoglycemic encounters. Of 2857 patients with recurrence, 304 patients (10.6%) had fragmented care. In those with high hypoglycemic encounters (≥4), 22% (N = 226) had ≥10 encounters; race and insurance status differences were associated with number of hypoglycemic encounters. Having hypoglycemia was associated with increased mortality compared to no hypoglycemia (n = 2696, 27.7% vs n = 20,188, 11.4%; p < 0.00001 by chi-square). ConclusionA small subset of patients with hypoglycemia accounted for a large subset of hypoglycemia encounters. Targeted interventions in this high-risk, high mortality group are needed.