Predicting Risk of Potentially Preventable Hospitalization in Older Adults With Dementia

Donovan T. Maust, MD, MS; H. Myra Kim, ScD; Claire Chiang, PhD; Kenneth M. Langa, MD, PhD; Helen C. Kales, MD

Disclosures

J Am Geriatr Soc. 2019;67(10):2077-2084. 

In This Article

Abstract and Introduction

Abstract

Objectives: Reducing potentially preventable hospitalization (PPH) among older adults with dementia is a goal of Healthy People 2020, yet no tools specifically identify patients with dementia at highest risk. The objective was to develop a risk prediction model to identify older adults with dementia at high imminent risk of PPH.

Design: A 30-day risk prediction model was developed using multivariable logistic regression. Patients from fiscal years (FY) 2009 to 2011 were split into development and validation cohorts; FY2012 was used for prediction.

Setting: Community-dwelling older adults (≥65 years of age) with dementia who received care through the Veterans Health Administration.

Participants: There were 1 793 783 participants.

Measurements: Characteristics associated with hospitalization risk were (1) age and other demographic factors; (2) outpatient, emergency department, and inpatient utilization; (3) medical and psychiatric diagnoses; and (4) prescribed medication use including changes to psychotropic medications (eg, initiation or dosage increase). Model discrimination was determined by the C statistic for each of the three cohorts. Finally, to determine whether predicted 30-day risk strata were stable over time, the observed PPH rate was calculated out to 1 year.

Results: In the development cohort, .6% of patients experienced PPH within 30 days. The C statistic for the development cohort was .83 (95% confidence interval [CI] = .83-.84) and .83 in the prediction cohort (95% CI = .82-.84). Patients in the top 10% of predicted 30-day PPH risk accounted for more than 50% of 30-day PPH admissions in all three cohorts. In addition, those predicted to be at elevated 30-day risk remained at higher risk throughout a year of follow-up.

Conclusion: It is possible to identify older adults with dementia at high risk of imminent PPH, and their risk remains elevated for an entire year. Given the negative outcomes associated with acute hospitalization for those with dementia, healthcare systems and providers may be able to engage these high-risk patients proactively to avoid unnecessary hospitalization.

Introduction

Patients with dementia have an all-cause hospitalization rate approximately 1.4 times higher than other older adults, and potentially preventable hospitalization (PPH) is nearly 1.8-times higher.[1] PPH captures admission for ambulatory care-sensitive conditions such as congestive heart failure (CHF) or pneumonia, that, with optimal outpatient access and management, are potentially unnecessary. Reduction of PPH specifically in older adults with dementia is a goal of Healthy People 2020.[2,3] Their elevated hospitalization risk is worrying because, although all older adults are at increased risk of hospitalization-associated delirium, iatrogenic complications, and cognitive and functional decline,[4–6] the consequences are greater for patients with dementia,[7,8] for whom cognitive or functional decline are risk factors for institutionalization.[9,10]

As the population with dementia nearly triples by 2050,[11] even small reductions in the rate of PPH could have a large impact. Unfortunately, no controlled dementia care intervention trials have demonstrated a reduction in hospitalization.[12,13] One possible reason is the trials were not specifically designed to target patients at the highest risk of hospitalization. Given the potential adverse consequences of hospitalization for older adults with dementia, identifying those at high risk before they need to be hospitalized, especially admissions for conditions that could potentially be treated in an outpatient setting, is critical.

Approaches to risk-stratify patients with dementia that do not rely on overburdened primary care clinicians[14] is key to appropriately targeting supports that may benefit these older adults and their caregivers.[15] For this analysis, we used national data from the US Veterans Health Administration (VHA) electronic health record (EHR) to develop a multivariable logistic regression model to predict PPH admission within 30 days (developmental cohort). We then used the model to identify risk tiers among a different cohort of older adults with dementia (validation cohort) and determine whether the model could accurately predict risk among the new set of patients (prediction cohort).

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....