Mortality and Discharge Location of Intensive Care Patients With Alzheimer Disease and Related Dementia

We used Optum’s deidentified Clinformatics Data Mart Database version 8.1, covering the years from 2016 to 2019. This insurance claims database includes 15 to 20 million annual lives from both commercial and MA enrollees dispersed across the 50 US states. The database provides demographic data and medical claims for both outpatient and inpatient services with regular follow-up until either the individual’s death or membership discontinuation. The database rigorously identifies death records from 6 sources: CMS, the US Social Security Administration’s death master file, facility claims where discharge status is “expired,” membership discontinuation due to death, Optum electronic health records, and externally sourced obituary data. The Indiana University Human Research Protection Program designated this study as not involving human subjects as defined in the Code of Federal Regulations¹⁴  and determined that it did not require institutional review board review.

We identified all MA beneficiaries who had an ICU admission in year 2018. The study included those with a first ICU admission in year 2018 (the index ICU stay) with no ICU admissions within 730 days of the index admission. Next, the study included those with continuous Medicare enrollment throughout the 2 years prior to the index admission, and throughout year 2018 or during year 2018 until the calendar month of death. Excluded were those whose age was 65 years or younger in year 2016 and who died before discharge.

We used diagnosis codes from inpatient and outpatient claims from the look-back period (the 730 days before the index ICU admission) to identify diagnostic ADRD among the study population.¹⁵  Patients with at least 1 inpatient claim or 2 outpatient claims at least 7 days apart with a qualifying diagnosis were identified as having ADRD.

The outcomes in this study included ICU discharge status (discharged to home vs some other facility) and mortality (death within the same calendar month as discharge and death within 12 months after discharge). The sample was evaluated for the presence of 10 common chronic conditions chosen on the basis of their high prevalence among older adults: hypertension, diabetes mellitus, coronary artery disease, osteoarthritis, renal disease, congestive heart failure, chronic obstructive pulmonary disease, cancer, stroke, and liver disease.^16-19  We used the same approach to identify these chronic comorbidities as we did for identifying ADRD.

In addition to chronic conditions, we included demographic data (ie, age, sex, and race), geographic and socioeconomic proxies (ie, dual eligibility for Medicare and Medicaid), and the number of hospitalizations before the ICU index admission.

We first used χ² tests to compare differences in binary measures, including age category, sex, race, eligibility, and each chronic comorbid condition. A t test was used to compare the difference in mean age between the 2 cohorts. Wilcoxon rank sum tests were used to test for differences in numbers of chronic comorbid conditions and hospitalizations. Patient outcomes (discharge status and mortality outcomes) were modeled as binary variables; multivariate logistic regression was used to compare outcomes between the ADRD and non-ADRD cohorts, adjusted for comorbidity, demographics, geographic and socioeconomic factors, and hospital clusters.

We identified a total of 473 371 patients with an ICU admission in 2018. Approximately 18% of these patients were excluded for having a previous ICU admission during the 2 years before the index ICU admission. After excluding others because of age or a lack of continuous MA enrollment, the final study population included 145 342 patients discharged from the ICU. Among this group, 10.5% had a diagnosis of ADRD (Figure 1).

Table 1 shows the differences in demographic characteristics (age, sex, race) and Medicare eligibility between the ADRD and non-ADRD cohorts. The mean age of the ADRD cohort (82.9 years) was significantly higher than that of the non-ADRD cohort (78.3 years, P < .001); almost half of the ADRD cohort was aged 85 years or older, versus less than a quarter of the non-ADRD cohort. Most ADRD patients were female (59.9% of ADRD patients vs 51.8% of non-ADRD patients, P < .001), but the race/ethnicity breakdown of the 2 cohorts was similar even though many differences were statistically significant. Notably, 14.6% of the ADRD cohort had dual Medicare/Medicaid eligibility versus only 5.2% of the non-ADRD cohort.

The comorbidity burden of the ADRD cohort was greater than that of the non-ADRD cohort. All the identified comorbidities except cancer and liver disease were more prevalent in patients with ADRD (Table 2). The diseases with the largest difference in prevalence between the ADRD and non-ADRD cohorts were osteoarthritis, renal disease, and congestive heart failure, each with a difference of more than 7 percentage points. Hospitalizations during the 2 years before the index ICU admission were more common in the ADRD cohort as well.

Although only a little more than one-third (37.6%) of the patients with ADRD went home after discharge, more than two-thirds (68.6%) of non-ADRD patients went home after discharge (odds ratio [OR] = 0.40; 95% CI = 0.38-0.41). Deaths in the ADRD cohort were almost twice as common within the same calendar month as discharge (19.9% vs 10.3%, OR = 1.54; 95% CI = 1.47-1.62) and within 12 months of discharge (50.8% vs 26.2%, OR = 1.95; 95% CI = 1.88-2.02, Figure 2) in comparison with deaths in the non-ADRD cohort. Results from adjusted logistic regression (Table 3) show that patients with ADRD were 60% less likely to be discharged to home than non-ADRD patients (odds ratio [OR], 0.40; 95% CI, 0.38-0.41). Dual-eligible (Medicare/Medicaid) beneficiaries had a higher risk of dying within the same calendar month as discharge (OR, 1.90; 95% CI, 1.76-2.05) than patients who were not dual-eligible and had an even greater risk of dying within 12 months of discharge (OR, 2.16; 95% CI, 2.01-2.32). Furthermore, ADRD patients faced a 54% higher risk of dying in the same month as discharge (OR, 1.54; 95% CI, 1.47-1.62) and a 95% higher risk of dying within 12 months (OR, 1.95; 95% CI, 1.88-2.02; Table 3).

In our study, we found that older adults with ADRD diagnoses who were admitted to the ICU were much less likely to be discharged home and faced almost twice the risk of death in the same calendar month as discharge and the 12 months after discharge in comparison with patients who did not have an ADRD diagnosis.

Previous research has demonstrated that patients with ADRD, especially older adults, experience an acute reduction in cognitive function and an increase in mortality after an acute illness.²⁰  As a result, the presence of ADRD should be carefully considered in ICU patients, especially when aggressive management is planned.^9,21 

It may be that some aggressive measures in older ICU patients with ADRD will have little benefit. Additionally, providers should consider the hospital care goals for patients with ADRD and introduce palliative care plans if appropriate. Family members of patients living with advanced ADRD almost universally agree that comfort should be the primary goal of care for their loved ones.²²  Incorporating palliative care in the emergency department, hospital, and posthospital discharge location may reduce system burden associated with patients with ADRD.²³ 

In the current study, patients with ADRD were older and had a larger disease burden and greater hospital utilization than their counterparts who did not have ADRD. This finding is consistent with previous studies associating ADRD prevalence with older age, more comorbidities, and a higher hospitalization rate.^22,24-26  Although these risk factors (age, comorbidity burden, hospitalization) could have also contributed to the increased mortality and reduced home discharge of patients with ADRD in our cohort, others have shown that ADRD is independently associated with a reduction in life expectancy and an increase in mortality risk.^25,27  Additionally, the difference in sex distribution we observed between the ADRD and non-ADRD cohort (59.9% female vs 51.8% female) is similar to that reported by Pisani et al⁸  in their study of ADRD in the ICU (65% vs 52%).

The strengths of this study include the large, geographically diverse MA sample and 12-month follow-up after ICU discharge. Our study adds to the scant literature about MA beneficiaries with ADRD. Existing work in MA populations with ADRD explores the prevalence of ADRD¹² ; the CMS hierarchical condition categories risk adjustment factor in ADRD^28,29 ; plan switching³⁰ ; and health services utilization, care satisfaction, and health status.³¹  Our study adds information about ICU outcomes and discharge status in patients with ADRD covered by MA plans.

In a recent study, Zhu et al³²  compared the prognoses of ICU patients with and without Alzheimer disease.However, the patient population differed from the population in our study. Our study focused on MA beneficiaries from across the United States, whereas the Zhu study’s population was derived from 1 urban hospital in the northeastern region of the United States.

Our results should serve as background information to inform future research aimed to optimize outcomes and payment models in ADRD. Our results add information pertinent to the policy dialogue surrounding comprehensive coverage to address the full range of biopsychosocial needs of caregivers, families, and patients dealing with ADRD. Our results are also timely, as CMS begins to assess innovations in the hospice component in value-based insurance designs in MA plans. Understanding death rates, death within the same calendar month as discharge, or 1 year after discharge, and the proportions of patients discharged to home after an ICU stay gives insight as to when and where to focus comprehensive care approaches and care coordination. It may help researchers understand areas to investigate to determine gaps in payment models at critical, possibly end-of-life, junctures in ADRD.^33,34 

The results of this study should, however, be viewed in light of some limitations. First, although this study included a large number of patients with ADRD, it did not explore mortality rates within different types of ADRD or between mild versus severe dementia. Understanding those differences would assist ICU physicians in determining the appropriate level of care for each patient. Second, our study did not include a risk adjustment for disease burden or severity. CMS removed the hierarchical condition categories risk adjustment factor for ADRD in MA plans in 2014 and did not reintroduce it again until 2020.²⁹  Third, we did not attempt to identify differences in the level of care or in postdischarge health encounters in ADRD versus non-ADRD patients, nor did we examine the cause of death. This, too, would further inform the clinical course of ADRD patients. Fourth, we did not attempt to determine whether patients were discharged to locations other than home for social reasons as opposed to care needs. Fifth, our study did not include information on the quality of inpatient acute care provided to patients living with or without ADRD during their hospital or ICU stay. This, or other unobserved underlying differences in health between the 2 groups, could contribute to the outcome differences we observed. Sixth, we do not know how many of our study population resided in a nursing home at the time of their ICU admission; estimates place MA enrollment in nursing homes at less than 20%,³⁵  but it is likely that nursing home residence was more common among the ADRD cohort than the non-ADRD cohort. Seventh, our study did not include data on the safety and quality of care management, which could have been a mediator for both outcomes of staying home and mortality; specifically, we lack data on the various chronic medical conditions affecting those with and without ADRD provided by the primary care and specialty providers before or after the ICU admission. Eighth, we could not determine from the data the time from initial ADRD diagnosis to time of hospital ICU admission. Finally, as an observational study using claims data, the results should not be interpreted as showing causal associations.

Patients with ADRD often have a limited life expectancy, which shortens sharply after acute events such as ICU admission.²  The lower home discharge and higher mortality rates after an ICU admission in patients with ADRD, as compared with the rates for the general ICU patient population, are cause for concern. Although beyond the scope of this study, these observations raise questions about whether treatment plans should be adjusted proactively to include strategies to diminish the likelihood of an ICU admission, such as vigilance with comorbid disease management, or patient, family, and caregiver involvement in early discussions about palliative care. Further research on discharge rates and mortality risks of patients with ADRD is needed, as well as research on how to best stratify the type and severity of ADRD for exploring post-ICU outcomes. Research comparing the level of care provided to ADRD patients relative to non-ADRD patients would also be valuable.

The authors thank Richard Meraz, manager, Advanced Biomedical IT Core for Indiana University’s University Information Technology Services Research Technologies, for his help in reviewing the analysis.