Return to search

Leveraging Natural Language Processing to Identify Risk for Hospitalizations Among Older Adult Home Healthcare Patients with Urinary Incontinence

Background: Persistently elevated hospitalization rates in the home healthcare setting indicate the need to prioritize patients with undertreated conditions that can lead to negative outcomes. Urinary incontinence affects approximately 40% of older adults in home healthcare, yet often remains unaddressed. This leaves older adults with urinary incontinence at risk for potentially serious complications that can lead to emergency department visits, hospitalizations, and mortality. Multiple comorbidities, co-occurring symptoms, and disparities in care fuel the complexity of older adults in the home healthcare setting. The overall purpose of this dissertation was to leverage natural language processing to understand symptom clusters and factors associated with acute care utilization among older adults with urinary incontinence in home healthcare to improve comprehensive assessment, treatment, and outcomes.

The aims of this dissertation were to: 1) identify relevant comorbidities among community-dwelling older adults with urinary incontinence; 2) develop and test a natural language processing algorithm to extract symptom information from home healthcare free-text clinical notes for older adults with urinary incontinence and analyze differences by race or ethnicity; 3) identify symptom clusters among older adults with urinary incontinence in home healthcare and examine differences by sociodemographic and clinical correlates; and 4) determine factors associated with the risk of emergency department visits or hospitalizations among older adults with urinary incontinence in home healthcare, including the impact of symptom clusters.

Methods: This dissertation comprised four studies: 1) a scoping review of the literature to identify comorbidities to broadly characterize community-dwelling older adults with urinary incontinence, 2) a secondary analysis of cross-sectional electronic health record data using natural language processing to extract symptoms from free-text clinical notes and analyze differences by race or ethnicity using Chi-square tests and logistic regression models, 3) a secondary analysis of cross-sectional electronic health record data using hierarchical clustering to analyze the natural language processing-extracted symptom variables and examine differences in sociodemographic and clinical correlates using Chi-square tests, and 4) a retrospective secondary analysis of electronic health record data to identify factors, including symptom clusters, associated with emergency department visits or hospitalizations using Chi-square tests and backward stepwise logistic regression.

Results: In the scoping review, we synthesized findings from 10 studies that identified comorbidities among community-dwelling older adults with urinary incontinence across neurologic, cardiovascular, respiratory, endocrine, genitourinary, musculoskeletal, and psychologic systems. In the natural language processing study, we identified eight symptoms of older adults with urinary incontinence (i.e., anxiety, constipation, dizziness, syncope, tachycardia, urinary frequency/urgency, urinary hesitancy/retention, and vision impairment/blurred vision) that were extracted from free-text clinical notes from approximately 29% of home healthcare episodes. Compared to White patients, home healthcare episodes for Asian/Pacific Islander, Hispanic, and Black patients were less likely to have any symptoms documented in clinical notes. In the clustering analysis, we identified five distinct symptom clusters: Cluster 1 (anxiety), Cluster 2 (broadly symptomatic), Cluster 3 (dizziness and anxiety), Cluster 4 (constipation, anxiety, and dizziness), and Cluster 5 (no symptoms) that correlate with sociodemographic and clinical characteristics. Finally, in the retrospective analysis, we found that Clusters 1-4 had higher odds of emergency department visits or hospitalizations, in addition to home healthcare episodes for Black and Hispanic patients, males, patients with an unhealed skin ulcer, and patients with a urinary tract infection 14 days prior to home healthcare admission.

Conclusion: Older adults with urinary incontinence in home healthcare have complex physical and psychosocial needs, increasing the risk of negative outcomes. Improving comprehensive assessment and treatment for older adults with urinary incontinence is an urgent priority, given high hospitalization rates in home healthcare. Leveraging natural language processing, this dissertation identified key symptom clusters and factors associated with emergency department visits or hospitalizations, providing valuable insight for multidimensional interventions. Findings provide preliminary evidence to inform improvements in clinical practice, healthcare policies, and future research to enhance the care of older adults with urinary incontinence and reduce negative outcomes in the home healthcare setting.

Identiferoai:union.ndltd.org:columbia.edu/oai:academiccommons.columbia.edu:10.7916/jx8h-6472
Date January 2024
CreatorsScharp, Danielle
Source SetsColumbia University
LanguageEnglish
Detected LanguageEnglish
TypeTheses

Page generated in 0.0023 seconds