Mining the Health Disparities and Minority Health Bibliome: A Computational Scoping Review and Gap Analysis of 200,000+ Articles.

Without comprehensive examination of available literature on health disparities and minority health (HDMH), the field is left vulnerable to disproportionately focus on specific populations or conditions, curtailing our ability to fully advance health equity. Using scalable open-source methods, we conducted a computational scoping review of more than 200,000 articles to investigate major populations, conditions, and themes in the literature as well as notable gaps. We also compared trends in studied conditions to their relative prevalence in the general population using insurance claims (42MM Americans). HDMH publications represent 1% of articles in MEDLINE. Most studies are observational in nature, though randomized trial reporting has increased five-fold in the last twenty years. Half of all HDMH articles concentrate on only three disease groups (cancer, mental health, endocrine/metabolic disorders), while hearing, vision, and skin-related conditions are among the least well represented despite substantial prevalence. To support further investigation, we also present HDMH Monitor, an interactive dashboard and repository generated from the HDMH bibliome.

Large-scale characterization of gender differences in diagnosis prevalence and time to diagnosis.

We carry out an analysis of gender differences in patterns of disease diagnosis across four large observational health datasets and find that women are routinely older when first assigned most diagnoses. Among 112 acute and chronic diseases, women experience longer lengths of time between symptom onset and disease diagnosis than men for most diseases regardless of metric used, even when only symptoms common to both genders are considered. These findings are consistent for patients with private as well as government insurance. Our analysis highlights systematic gender differences in patterns of disease diagnosis and suggests that symptoms of disease are measured or weighed differently for women and men. Data and code leverage the open-source common data model and analytic code and results are publicly available.

Auditing Learned Associations in Deep Learning Approaches to Extract Race and Ethnicity from Clinical Text.

Complete and accurate race and ethnicity (RE) patient information is important for many areas of biomedical informatics research, such as defining and characterizing cohorts, performing quality assessments, and identifying health inequities. Patient-level RE data is often inaccurate or missing in structured sources, but can be supplemented through clinical notes and natural language processing (NLP). While NLP has made many improvements in recent years with large language models, bias remains an often-unaddressed concern, with research showing that harmful and negative language is more often used for certain racial/ethnic groups than others. We present an approach to audit the learned associations of models trained to identify RE information in clinical text by measuring the concordance between model-derived salient features and manually identified RE-related spans of text. We show that while models perform well on the surface, there exist concerning learned associations and potential for future harms from RE-identification models if left unaddressed.

An electronic pillbox intervention designed to improve medication safety during care transitions: challenges and lessons learned regarding implementation and evaluation.

BACKGROUND: Adverse drug events are common during transitions of care. As part of the Smart Pillbox study, a cluster-randomized controlled trial of an electronic pillbox designed to reduce medication discrepancies and improve medication adherence after hospital discharge, we explored barriers to successful implementation and evaluation of this intervention. METHODS: Eligible patients were those admitted to a medicine service of a large teaching hospital with a plan to be discharged home on five or more chronic medications. The intervention consisted of an electronic pillbox with pre-filled weekly blister pack medication trays given to patients prior to discharge. Pillbox features included alarms to take medications, detection of pill removal from each well, alerts to patients or caregivers by phone, email, or text if medications were not taken, and adherence reports accessible by providers. Greater than 20% missed doses for three days in a row triggered outreach from a pharmacist. To identify barriers to implementation and evaluation of the intervention, we reviewed patient exit surveys, including quantitative data on satisfaction and free-text responses regarding their experiences; technical issue logs; and team meeting minutes. Themes were derived by consensus among the study authors and organized using the Consolidated Framework for Implementation Research. RESULTS: Barriers to implementation included intervention characteristics such as perceived portability issues with the pillbox and time required by pharmacists to enter medication information into the software; external policies such as lack of insurance coverage for early refills and regulatory prohibitions on repackaging medications; implementation climate issues such as the incompatibility between the rushed nature of hospital discharge with the time required to deploy the intervention; and patient issues such as denial of previous problems with medication adherence. We founds several obstacles to conducting the study, including patients declining study enrollment and limited attempts by the hospital to streamline logistics by building the intervention into usual care. Several solutions to address many of these challenges were implemented or planned. Despite these challenges, many patients with the pillbox were pleased with the service and believed the intervention worked well for them. CONCLUSIONS: In this evaluation, several barriers to implementing and conducting a study of the effectiveness of the intervention were identified. Our findings provide lessons learned for others wishing to implement and evaluate HIT-related interventions designed to improve medication safety during care transitions. TRIAL REGISTRATION: Clinicaltrials.gov NCT03475030.

Health Care Use by Muslim Patients During Ramadan.

PURPOSE: Health care utilization during Ramadan has not been examined in the United States. METHODS: A retrospective review of billing and electronic health record data for Muslims (n = 2,919) and non-Muslims (n = 184,803) in primary care practices in Eastern Massachusetts. RESULTS: Muslim patients were younger, less educated, less often commercially insured, more likely to have Medicare, and less likely to be primary English speakers (p < .0001 for all comparisons). In multivariate models, during Ramadan, Muslims, compared with non-Muslims, had a higher rate of primary care visits (incidence rate ratio [IRR], 1.06; 95% confidence interval [CI], 1.01-1.11), emergency department visits (IRR, 1.60; 95% CI, 1.34-1.91), and hospitalizations (IRR, 1.18; 95% CI, 1.03-1.34). CONCLUSIONS: Important demographic differences exist between Muslim and non-Muslim patients. Muslims, compared with non-Muslims, had higher health care utilization during Ramadan.

Design of MARQUIS2: study protocol for a mentored implementation study of an evidence-based toolkit to improve patient safety through medication reconciliation.

BACKGROUND: The first Multi-center Medication Reconciliation Quality Improvement Study (MARQUIS1) demonstrated that implementation of a medication reconciliation best practices toolkit decreased total unintentional medication discrepancies in five hospitals. We sought to implement the MARQUIS toolkit in more diverse hospitals, incorporating lessons learned from MARQUIS1. METHODS: MARQUIS2 is a pragmatic, mentored implementation QI study which collected clinical and implementation outcomes. Sites implemented a revised toolkit, which included interventions from these domains: 1) best possible medication history (BPMH)-taking; 2) discharge medication reconciliation and patient/caregiver counseling; 3) identifying and defining clinician roles and responsibilities; 4) risk stratification; 5) health information technology improvements; 6) improved access to medication sources; 7) identification and correction of real-time discrepancies; and, 8) stakeholder engagement. Eight hospitalists mentored the sites via one site visit and monthly phone calls over the 18-month intervention period. Each site's local QI team assessed opportunities to improve, implemented at least one of the 17 toolkit components, and accessed a variety of resources (e.g. implementation manual, webinars, and workshops). Outcomes to be assessed will include unintentional medication discrepancies per patient. DISCUSSION: A mentored multi-center medication reconciliation QI initiative using a best practices toolkit was successfully implemented across 18 medical centers. The 18 participating sites varied in size, teaching status, location, and electronic health record (EHR) platform. We introduce barriers to implementation and lessons learned from MARQUIS1, such as the importance of utilizing dedicated, trained medication history takers, simple EHR solutions, clarifying roles and responsibilities, and the input of patients and families when improving medication reconciliation.

Benzodiazepines are Prescribed More Frequently to Patients Already at Risk for Benzodiazepine-Related Adverse Events in Primary Care.

BACKGROUND: Benzodiazepine use is associated with adverse drug events and higher mortality. Known risk factors for benzodiazepine-related adverse events include lung disease, substance use, and vulnerability to fracture. OBJECTIVE: To determine whether benzodiazepine prescribing is associated with risk factors for adverse outcomes. DESIGN: Longitudinal cohort study between July 1, 2011, and June 30, 2012. PARTICIPANTS: Patients who visited hospital- and community-based practices in a primary care practice-based research network. MAIN MEASURES: Odds ratio of having a target medical diagnosis for patients who received standard and high-dose benzodiazepine prescriptions; rates per 100 patients for outpatient and emergency department visits and hospitalizations. KEY RESULTS: Among 65,912 patients, clinicians prescribed at least one benzodiazepine to 15 % (9821). Of benzodiazepine recipients, 5 % received high doses. Compared to non-recipients, benzodiazepine recipients were more likely to have diagnoses of depression (OR, 2.7; 95 % CI, 2.6-2.9), substance abuse (OR, 2.2; 95 % CI, 1.9-2.5), tobacco use (OR, 1.7; 95 % CI, 1.5-1.8), osteoporosis (OR, 1.6; 95 % CI, 1.5-1.7), chronic obstructive pulmonary disease (OR, 1.6; 95 % CI, 1.5-1.7), alcohol abuse (OR, 1.5; 95 % CI, 1.3-1.7), sleep apnea (OR, 1.5; 95 % CI, 1.3-1.6), and asthma (OR, 1.5; 95 % CI, 1.4-1.5). Compared to low-dose benzodiazepine recipients, high-dose benzodiazepine recipients were even more likely to have certain medical diagnoses: substance abuse (OR, 7.5; 95 % CI, 5.5-10.1), alcohol abuse (OR, 3.2; 95 % CI, 2.2-4.5), tobacco use (OR, 2.7; 95 % CI, 2.1-3.5), and chronic obstructive pulmonary disease (OR, 1.5; 95 % CI, 1.2-1.9). Benzodiazepine recipients had more primary care visits per 100 patients (408 vs. 323), specialist outpatient visits (815 vs. 578), emergency department visits (47 vs. 29), and hospitalizations (26 vs. 15; p < .001 for all comparisons). CONCLUSIONS: Clinicians prescribed benzodiazepines and high-dose benzodiazepines more frequently to patients at higher risk for benzodiazepine-related adverse events. Benzodiazepine prescribing was associated with increased healthcare utilization.

Antiviral and antibiotic prescribing for influenza in primary care.

BACKGROUND: Anti-influenza antiviral medications reduce influenza-related morbidity, but may often be used inappropriately. OBJECTIVE: To measure the rate of antiviral and antibiotic prescribing, the appropriateness of antiviral prescribing, and evaluate independent predictors of antiviral and antibiotic prescribing for influenza in primary care. DESIGN AND PATIENTS: Retrospective analysis of 958 visits of clinician-diagnosed influenza in 21 primary care clinics in eastern Massachusetts from 1999 to 2007. We considered antiviral prescribing appropriate if patients had symptoms for 2 or fewer days, had fever, and any 2 of headache, sore throat, cough, or myalgias. MEASUREMENTS AND MAIN RESULTS: Clinicians prescribed antivirals in 557 (58%) visits and antibiotics in 104 visits (11%). Of antiviral prescriptions, 38% were not appropriate, most commonly because of symptoms for more than 2 days (24% of antiviral prescriptions). In multivariate modeling, selected independent predictors of antiviral prescribing were symptom duration of 2 or fewer days (odds ratio [OR], 12.4; 95% confidence interval [CI], 8.3 to 18.6), year (OR, 1.4 for each successive influenza season; 95% CI, 1.3 to 1.7), patient age (OR, 1.3 per decade; 95% CI, 1.2 to 1.5), and, compared to having no influenza testing, having a negative influenza test (OR, 5.5; 95% CI, 3.4 to 9.1) or a positive influenza test (OR, 11.4; 95% CI, 6.7 to 19.3). Independent predictors of antibiotic prescribing included otoscopic abnormalities (OR, 3.3; 95% CI, 1.8 to 6.0), abnormal lung examination (OR, 4.0; 95% CI, 2.1 to 6.2), and having a chest x-ray performed (OR, 2.2; 95% CI, 1.3 to 3.8). CONCLUSIONS: Primary care clinicians are much more likely to prescribe antivirals to patients with symptoms for 2 or fewer days, but also commonly prescribe antivirals inappropriately.