Use of Inpatient Psychiatric Facilities by Medicare Beneficiaries with Dementia

Alzheimer's dementia, the most common type of dementia, afflicts an estimated 6 million Americans. More than 80 percent of people living with dementia (PLWD) live in the community either with caregivers or alone. As cognitive impairment becomes more severe, behavioral and psychological symptoms of dementia (BPSD) become more difficult to manage, and those with severe or dangerous BPSD may need to be treated in an inpatient psychiatric facility (IPF). The immediate goal of an IPF stay is to stabilize patients in a psychiatric crisis. IPFs primarily serve people with serious mental illness (SMI) and substance use disorders, whose care needs may overlap with but also differ from the needs of PLWD.

Little is known about PLWD who use IPFs. The goal of this research was to conduct exploratory analyses focused on PLWD who use IPFs to (1) characterize the population and compare them with IPF users without dementia, (2) examine characteristics and utilization patterns for different services and settings that may be associated with IPF stays, and (3) analyze outcomes following IPF stays.

We used Medicare fee-for-service (FFS) data to conduct descriptive analyses characterizing beneficiaries with dementia who experienced an IPF stay in 2018 and compare them with beneficiaries without dementia. We used regression analyses to explore predictors of IPF utilization and service use and outcomes after IPF discharge.

Key Findings

Sociodemographic characteristics. In 2018, approximately 205,000 Medicare FFS beneficiaries had at least one IPF stay. Their mean age was 60, and 56 percent were younger than 65 at the time of their first stay during the year. Slightly more than one-half of beneficiaries with an IPF stay were female. About 78 percent were White, non-Hispanic, 15 percent were Black, non-Hispanic, and a small percentage were Hispanic, Asian, or from another racial or ethnic group. Many IPF users were low-income, as measured by dual eligibility status for Medicare and Medicaid (57 percent) and eligibility for the Medicare Part D Low Income Subsidy (LIS) (63 percent). About 21 percent of IPF users resided in rural counties, and many lived in underserved regions, with 92 percent in counties with at least partial shortages in primary care and 94 percent in counties with at least partial shortages in mental health care.

More than 40 percent of the IPF users had dementia. Compared with IPF users without dementia, IPF users with dementia were older (76 percent versus 22 percent over age 65), more often female (56 percent versus 48 percent), and more often white (84 percent versus 74 percent) (Figure 1). They were less frequently dually eligible for Medicare and Medicaid (48 percent versus 64 percent) and the Medicare Part D LIS (51 percent versus 71 percent). PLWD were more frequently residents of rural counties (25 percent versus 18 percent) compared with those without dementia.

A bar chart shows the percentages of IPF users who are people living with dementia and people not living with dementia across six different characteristics.

PLWD	People not living with dementia
Age 65+	76 percent	22 percent
Female	56 percent	48 percent
White, non-Hispanic	84 percent	74 percent
Dual eligible	48 percent	64 percent
Part D LIS eligible	51 percent	71 percent
Rural county	25 percent	18 percent

NOTE: IPF = inpatient psychiatric facility; LIS = low-income subsidy; PLWD = people living with dementia.

Health and utilization characteristics. The typical beneficiary admitted to an IPF had a high level of illness burden. Of all IPF users, about 70 percent were mildly, moderately, or severely frail. In addition to more than 40 percent of IPF users having dementia, SMI was common: 43 percent had schizophrenia, 59 percent had bipolar disorder, and 83 percent had major depressive disorder. Multiple physical health comorbidities were also common, including rheumatoid arthritis (71 percent), ischemic heart disease (63 percent), and chronic kidney disease (53 percent).

About 72 percent of beneficiaries with any IPF stay in 2018 had only one IPF stay during the year; the rest had multiple IPF admissions, and 5 percent had four or more stays (Figure 2).

A bar chart shows the percentages of beneficiaries in 2018 with 1 through 4 or more stays.

Number of IPF Stays in 2018	Percentage of Beneficiaries with One or More IPF Stays
1	72 percent
2	17 percent
3	6 percent
4 or more	5 percent

NOTE: IPF = inpatient psychiatric facility.

Consistent with their older age, PLWD were frailer compared with those without dementia among IPF users (51 percent versus 11 percent moderately to severely frail) (Figure 3). IPF users with dementia were more often admitted to hospital psychiatric units than freestanding units (70 percent versus 52 percent) and had longer IPF stays, on average (14.4 versus 11.1 days), compared with people not living with dementia. Although comorbid SMI occurred for PLWD, they more frequently had a first diagnosis of schizophrenia and other psychotic disorders during IPF stays (16 percent versus 9 percent) compared with people not living with dementia, which may be related to the need to justify the use of antipsychotic medications.

A bar chart shows the percentages of IPF users who are people living with dementia and people not living with dementia across three health characteristics.

PLWD	People not living with dementia
Moderately to severely frail	51 percent	11 percent
IPF psychiatric unit	70 percent	52 percent
First diagnosis of schizophrenia and other psychotic disorders during IPF stay	16 percent	9 percent

NOTE: IPF = inpatient psychiatric facility; PLWD = people living with dementia.

Predictors of IPF utilization. Among PLWD, key predictors of IPF admission include long- term nursing home residence (odds ratio [OR] = 8.94; 95 percent confidence interval [CI], 8.12 to 9.86), frailty (OR = 5.12; 95 percent CI, 4.98 to 5.28 for moderately to severely frail compared with non-frail or pre-frail), SMI or SUD (e.g., OR = 2.78; 95 percent CI, 2.69 to 2.87 for schizophrenia), emergency department (ED) visit in the first quarter prior to IPF admission (OR = 3.59; 95 percent CI, 3.52 to 3.65), prior IPF stays in the past year (e.g., OR = 3.72; 95 percent CI, 3.21 to 4.33 for a prior IPF stay in the previous quarter), and use of antipsychotic medications in the first quarter prior to IPF admission (OR = 3.47; 95 percent CI, 3.36 to 3.59).

The associations between these types of utilization and IPF admission are stronger immediately before IPF admission. Figure 4 shows the associations over the four quarters prior to IPF admission for PLWD with ED visits and those with antipsychotic medication use. This pattern holds true for both PLWD and those without dementia; however, the relationship is stronger for PLWD. Furthermore, PLWD with antipsychotic medication use in two to four quarters prior to IPF admission were less likely to have an IPF admission, suggesting that new antipsychotic medication use in the one quarter prior may be a precursor to IPF admission.

This figure contains two separate bar charts.

Left bar chart:

ED visit	Days prior to IPF admission
271 to 360 days	Odds ratio: slightly above 1.0
181 to 270 days	Odds ratio: slightly above 1.0
91 to 180 days	Odds ratio: about 1.25
2 to 90 days	Odds ratio: 3.59

Right bar chart:

Antipsychotic medication use	Days prior to IPF admission
271 to 360 days	Odds ratio: about 0.8
181 to 270 days	Odds ratio: about 0.9
91 to 180 days	Odds ratio: about 0.7
2 to 90 days	Odds ratio: 3.47

NOTE: Error bars indicate 95 percent CIs. ED = emergency department; IPF = inpatient psychiatric facility; PLWD = people living with dementia.

Outcomes following IPF stays. PLWD have different patterns following IPF stays compared with people not living with dementia. Most notably, PLWD were less frequently discharged to home or self-care (48 percent versus 87 percent) and were more frequently discharged to skilled nursing facilities (SNFs) (25 percent to 2 percent). Compared with people not living with dementia, PLWD were more likely to use SNFs and less likely to use outpatient services in the 30 days following IPF discharge.

Risk of death after discharge from an IPF was substantial. Overall, about 2 percent of beneficiaries died within 30 days of discharge, and about 7 percent died within six months; these rates were much higher for PLWD (about 4 percent and 15 percent, respectively). Controlling for a variety of factors, including age, chronic conditions, and prior health care service and prior medication utilization, regression analysis of time to death after IPF discharge produced a hazard ratio (HR) of 1.76 (95 percent CI, 1.69 to 1.82) for PLWD compared with those without dementia.

Admission to long-term nursing homes was also more likely for PLWD compared with those without dementia, with an HR of 3.55 (95 percent CI, 3.40 to 3.71). Among beneficiaries who entered a long-term nursing home after IPF discharge, those with dementia were more likely to experience declines in cognition than those without dementia; however, no difference was found for functional status.

Policy Implications

Because IPFs primarily serve patients with SMI and SUD, these facilities may not have adequate resources and training to address BPSD. In addition to challenges in managing severe behavioral symptoms, PLWD using IPFs are often frail and have comorbid conditions, suggesting that they likely have complex health needs requiring care coordination.

Better understanding of the trajectory and care needs of PLWD could help inform the design of interventions and policies to support PLWD and caregivers coping with BPSD. Prior service and setting utilization associated with IPF admission could be points of intervention for evidence-based programs and policies that support PLWD and caregivers in managing BPSD. For example, we observed increased prevalence of ED visits and psychotropic medication use in the four quarters preceding IPF stays, suggesting possible worsening of BPSD over time. More work needs to be done to identify pathways of worsening BPSD that result in psychiatric crises that require stabilization in IPFs, as well as potential intervention points and strategies.

Similarly, improvements in care transitions and coordination following IPF discharges would support this population. Compared with people not living with dementia, PLWD are more likely to have SNF stays and less likely to have outpatient behavioral health visits for follow-up care after IPF discharges.

Even adjusted for age, comorbidities, utilization of health care service and psychotropic medications prior to IPF admission, and other factors, PLWD are more likely to enter long-term nursing homes and to die following IPF stays compared with people not living with dementia. Further work to better understand their care needs could apply qualitative methods to gather information from patients, caregivers, and providers across different settings regarding unmet needs and necessary resources.