Substance Use is Associated With Frequent Emergency Department Visits in Cardiac Patients
Tai Metzger, David A. Berger, Ramin Homayouni
Abstract
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsCardiac Health and Mental Health · Pharmaceutical Practices and Patient Outcomes
Background/Objectives
Social and behavioral determinants of health (SBDoH) influence healthcare utilization and a variety of health outcomes, including among cardiovascular disease (CVD) patients. Currently, the majority of outpatient and emergency department (ED) patients are not adequately screened although their needs may be documented in the clinical notes. The use of artificial intelligence (AI) text processing algorithms to analyze vast amounts of data in the electronic health record (EHR) may provide a more comprehensive view of SBDoH needs across a patient population. AI, which is already used in EDs to provide initial machine read of electrocardiograms, best practice alerts, and detection of hemorrhage on stroke head CT readings, may also facilitate assessment of patient risk of high ED use. Our objective was to apply a novel natural language processing (NLP) approach that we developed to determine which SBDoH are associated with frequent ED use for patients with CVD.
Methods
We included patients 18–65 years with a history of atrial fibrillation, acute myocardial infarction, ischemic heart disease, or non-ischemic heart disease during a one-year period (9/1/2022–8/31/23) at a large metropolitan hospital in southeast Michigan. Patients over 65 years old were excluded to focus on younger and healthier patients whose visits may be more affected by SBDoH. We used a custom algorithm to combine ICD-10 codes, SDoH screening responses, and SBDoH detected from the clinical notes with NLP. SBDoH factors were compared between high- (≥ 5 visits/year) and non-high utilizers (< 5 visits/year). Logistic regression with backward selection was used to find significant associations between high ED use and demographics, chronic conditions and 17 different SBDoH factors.
Results
A total of 4,844 patients met inclusion criteria, with 526 (10.9%) having high ED use. Univariate analysis comparing high and low ED use showed significant differences in sex, race, payer mix, average number of chronic conditions, and average number of SBDoH factors (Table 1). Multivariable regression revealed female sex, African American race, financial strain, unreliable transportation, inadequate support system, uninsured/underinsured, medication affordability concerns, depression, alcohol, and opioid abuse were significantly associated with high ED utilization (Table 2). In particular, patients with documented opioid abuse (adjOR 3.25, 95% CI 2.60–4.07, p< .0001) and alcohol abuse (adjOR 2.22, 95% CI 1.75–2.84, p< .0001) had significantly increased odds of frequent ED use. Payer mix was not included in the regression analysis because of the high degree of correlation between Medicaid status and SBDoH factors.
Conclusions
Using our unique NLP approach, patients with CVD and specific SBDoH factors were associated with high ED use. Consistent with other studies, we found that alcohol and opioid abuse were associated with two- and three-fold higher rates of ED use, respectively. Importantly, substance abuse is not screened in the standard SDoH tools, which emphasizes the importance of aggregating these data from multiple sources within the EHR, including the clinical notes. Future work could consider whether strategically addressing SBDoH, particularly substance use disorder among CVD patients, could reduce high ED use.