UCLA study finds no clear survival benefit from hospital admission for borderline dementia cases, despite higher costs

A new study by UCLA Fielding researchers finds no clear evidence that hospital admission either increases or reduces mortality among persons with dementia, while finding that admission is associated with higher health care spending.

“The research did not find clear evidence that hospital admission increases or reduces mortality, but we did find that hospital admission was associated with higher health care spending,” said UCLA Fielding’s Dr. Elizabeth Rose Mayeda, associate professor in the Department of Epidemiology and a co-author of the study. “These findings suggest that for some persons with dementia presenting at the hospital, health care spending might be reduced without compromising health outcomes, which highlights opportunities to improve decision-making around admission and acute care for this vulnerable population.”

The peer-reviewed study - “Estimating the Effect of Hospital Admission on Health Care Outcomes and Spending Among Persons With Dementia: A Quasi-experimental Study” - was released June 9 in the journal Annals of Internal Medicine, published by the American College of Physicians. In the study, researchers examined more than 870,000 emergency department visits in the United States by persons 66 or older with dementia between 2017-19 and analyzed deaths and the amount spent per case within 30 and 90 days of the emergency department visits.

“Older adults living with dementia are frequently seen in emergency departments and are hospitalized at high rates, yet the effect of hospital admissions on their health outcomes and health care spending remains unknown,” said Dr. Ryo Ikesu, a physician and assistant professor at Kyoto University in Japan who recently completed his PhD at UCLA and is the study's lead author. "Transitions in care settings can be disorienting for persons with dementia and may trigger delirium or other complications, raising the concern that hospital admission may not always benefit this population."

In the U.S., persons with dementia visit the emergency department more frequently and are hospitalized at higher rates than older adults without dementia, and make nearly 1.4 million emergency department visits annually, of which approximately 40% to 55% result in hospital admission.

A key challenge in studying the effect of hospital admission is that sicker patients are more likely to be admitted, which can make admitted patients appear to have worse outcomes even if admission itself is not the cause, researchers said.

To address this bias, the researchers used an instrumental variable approach, a causal inference method that leverages the fact that which emergency physician treats a patient is essentially random, determined by who happens to be on shift. By comparing outcomes for patients seen by high-admission-tendency physicians versus low-admission-tendency physicians, the researchers were able to estimate the causal effect of admission while removing the influence of patient severity. 

The researchers found that among the 872,085 emergency department visits they reviewed, 482,208, or 55.3% resulted in hospital admission. The team found no clear evidence that hospital admission affected 30-day or 90-day mortality. The estimated effect on 30-day mortality was −2.6 percentage points (95% confidence interval, −5.2 to +0.1 percentage points), indicating that the data are consistent with both a modest mortality benefit and a small mortality increase.

“We found, for example, that 30 days after admission, costs were roughly $2500 more for a hospitalized dementia patient than for a non-hospitalized person with the same conditions, but we found no clear evidence that hospitalization was associated with patient outcomes, including mortality,” said study co-author Dr. Yusuke Tsugawa, a physician and associate professor in UCLA Fielding’s Department of Health Policy and Management. 

Health care spending was $2,547 higher for admitted cases than for non-admitted cases. Much of that increase came from post-discharge care, particularly home health care ($472) and nursing facility care ($2,080), suggesting that hospital admission may alter patients’ care trajectories in ways that generate substantial downstream spending. Similar patterns were observed for 90-day spending.

"Our findings do not suggest that hospital admission is inappropriate for all persons with dementia; for patients with severe acute illness, such as serious pneumonia, hip fracture, or acute heart failure, hospital admission is often essential," Ikesu said. "But for borderline cases where the admission decision is uncertain, these findings highlight an opportunity to more thoughtfully consider alternatives such as home-based acute care or outpatient follow-up, which could serve the interests of patients, families, and the health care system alike."

Background

The authors note several limitations to this study. The study focused on people with dementia who were 66 or older and covered by Medicare in the United States, so the results may not apply to younger patients, those with other types of insurance, or people in other countries. The researchers used a method that compares patients seen by doctors who tend to admit more often with those seen by doctors who admit less often. This helps reduce bias from how sick patients are, but it also centers on patients whose need for admission is not clear, not about every person with dementia who comes to the emergency department. The estimates for death within 30 and 90 days are rough, so the results are consistent with both a modest benefit and a small risk from hospital admission, meaning the true effect on survival is still uncertain. Finally, the study relied on billing and administrative records, which do not include many important clinical details, including how severe a patient’s symptoms were, how well they could function day to day, or what patients and families preferred for their care.

Funding

This study received no direct funding. Dr. Ikesu received support from the University of California Graduate Division Dissertation Year Award and the Mimi and Roger Detels Fellowship, as well as the universities and institutions where the authors are affiliated.