Predicting hospital admission for older emergency department patients: Insights from machine learning

Highlights

• Little is known about the performance and utility of ML methods in predicting hospital admission among older ED patients.

• This is the first study to use ML for predicting hospital admission in older ED patients using geriatric syndromes and functional assessments.

• ML can accurately predict hospital admission in older ED patients using geriatric syndromes, functional assessments, and baseline care needs.

• Our predictions can inform decision-making about ED disposition and may expedite admission processes and proactive discharge planning.

Abstract

Background

Emergency departments (ED) are a portal of entry into the hospital and are uniquely positioned to influence the health care trajectories of older adults seeking medical attention. Older adults present to the ED with distinct needs and complex medical histories, which can make disposition planning more challenging. Machine learning (ML) approaches have been previously used to inform decision-making surrounding ED disposition in the general population. However, little is known about the performance and utility of ML methods in predicting hospital admission among older ED patients. We applied a series of ML algorithms to predict ED admission in older adults and discuss their clinical and policy implications.

Materials and methods

We analyzed the Canadian data from the interRAI multinational ED study, the largest prospective cohort study of older ED patients to date. The data included 2274 ED patients 75 years of age and older from eight ED sites across Canada between November 2009 and April 2012. Data were extracted from the interRAI ED Contact Assessment, with predictors including a series of geriatric syndromes, functional assessments, and baseline care needs. We applied a total of five ML algorithms. Models were trained, assessed, and analyzed using 10-fold cross-validation. The performance of predictive models was measured using the area under the receiver operating characteristic curve (AUC). We also report the accuracy, sensitivity, and specificity of each model to supplement performance interpretation.

Results

Gradient boosted trees was the most accurate model to predict older ED patients who would require hospitalization (AUC = 0.80). The five most informative features include home intravenous therapy, time of ED presentation, a requirement for formal support services, independence in walking, and the presence of an unstable medical condition.

Conclusion

To the best of our knowledge, this is the first study to predict hospital admission in older ED patients using a series of geriatric syndromes and functional assessments. We were able to predict hospital admission in older ED patients with good accuracy using the items available in the interRAI ED Contact Assessment. This information can be used to inform decision-making about ED disposition and may expedite admission processes and proactive discharge planning.

Introduction

Health systems around the world are challenged to adapt traditional care pathways to accommodate the complex physical and psychosocial needs of the growing geriatric population. The number of adults aged 65 years and older is expected to double by 2050, reaching 1.6 billion around the world [1]. The demand for health services will continue to increase alongside this demographic shift. Emergency departments (ED) are a common access point for older adults seeking medical attention, with the number of older ED patients increasing annually [2,3]. Older adults constitute a higher percentage of ED visits than younger persons, and they are more likely to visit for an urgent reason, resulting in hospital admission [[4], [5], [6]]. As the portal of entry into the hospital and the first source of contact for many seniors with their health care system, EDs are uniquely positioned to influence the health care trajectories of older adults presenting to the hospital for care [7,8].

The task-oriented and disease-centric focus of traditional emergency care pathways limit the department’s ability to attend to the multifaceted needs of older adults [9,10]. Geriatric complexity and multimorbidity often surpass the conventional ‘one-patient, one-problem’ approach to emergency care [10]. Assessing and treating older adults in the ED is often challenging for emergency clinicians [11,12], as these patients commonly present with heterogeneous symptoms, cognitive impairment, and a complex medical and social history [13]. Furthermore, the time pressures and high patient volumes seen in the ED can impede health care providers from providing comprehensive geriatric assessments and tailored treatment plans [14]. As a result, geriatric syndromes are commonly overlooked by ED clinicians [15,16], and older adults experience high rates of adverse outcomes both during their visit and following discharge [17]. Cognizant of the fact that older adults have distinct health care needs and a greater risk for adverse health events, clinical researchers and machine learning experts commonly aim to study the efficacy of clinical therapies and health services exposures in this vulnerable patient population [[18], [19], [20]]. Following suit, governing bodies responsible for ED accreditation and clinical practice guidelines collaborated in 2014 to create the first geriatric ED guidelines [8].

Determining patient disposition (admission vs. discharge) is an important and sometimes difficult decision for health care providers in the ED [21]. This is especially true for older adults, as they are at greater risk for functional decline, poor outcomes, and delayed discharge during in-patient stays [22,23]. The majority of hospital admissions come through the ED and constitute a significant source of health care spending. Furthermore, many hospital admissions are deemed to be potentially avoidable. As a result, health care providers, policymakers, and key stakeholders have all taken an interest in strategies to decrease inappropriate hospital admissions and extended stays [[24], [25], [26]]. Early identification of ED patients who require hospitalization can expedite admission processes and in-patient treatment [27], allowing for proactive discharge planning from the first patient contact. Both departmental barriers and the hazards associated with an inappropriate ED disposition underscore the utility of clinical decision support for older ED patients.

Machine learning (ML) has previously been used to predict disease, stratify patient risk, and inform clinical decision-making around patient disposition in emergency settings [28,29]. ML approaches are flexible and better able to identify hidden patterns and interactions among predictor variables [30,31]. The heterogeneous and complex presentation of older adults seeking emergency care makes ML an ideal candidate to inform clinical decision-making surrounding hospital admission in this high-risk patient population.

To date, the majority of studies aiming to predict hospital admission in ED patients focus on the general ED population [27,[32], [33], [34], [35], [36], [37], [38], [39]]. Two studies used text-mining and natural language processing to predict hospital admission [32,33], while the rest used non-linguistic methods to produce predictions. A study by Leegon et al. [40] attempted to predict hospital admission in the pediatric population. Only one study to date has attempted to predict hospital admission in older adults [41]. LaMantia and colleagues used demographic data, insurance status, chief complaint, vital signs, and triage acuity from a single hospital to fit a logistic regression model, which was moderately predictive of hospital admission in older ED patients (AUC = 0.73).

With little known about the performance and applications of ML approaches in older ED populations, we set out to identify whether ML could accurately predict hospital admission from the ED. More specifically, the objectives of this study are two-fold. First, we aimed to predict hospital admission using several ML approaches in older patients presenting to the ED for care. Second, we aimed to identify and describe the most important clinical and patient characteristics that are associated with hospital admission in older ED patients. Our study contributes to research in medical informatics, geriatrics, and emergency medicine by reporting on the utilities of ML as well as the importance of a comprehensive set of geriatric syndromes and functional assessments. These clinical assessments are not commonly available in an emergency setting.