The Palliative Care Paradox: Investigating Unexpected Outcomes in Dementia Care

Community-dwelling persons with dementia experience a high prevalence of neuropsychiatric symptoms (e.g., depression, apathy), clinical symptoms (e.g., pain, insomnia), and unmet care needs (e.g., assistance with activities of daily living) [1, 2]. They also frequently receive burdensome treatments near the end of life, including hospitalizations and intensive care unit (ICU) admissions [3]. Palliative care for individuals with dementia may therefore play a key role in ensuring that individuals receive care that optimizes quality of life, supports family and care partners, and facilitates goal-concordant care, which may entail avoiding certain aggressive medical interventions [4]. However, palliative care services are a limited resource in health systems worldwide. As a result, identifying the optimal timing for initiating palliative care among community-dwelling persons with dementia to meet clinical needs and reduce burdensome acute care utilization is critical [5]. In this issue of the Journal of the American Geriatrics Society, Quinn and colleagues used a unique cohort of 50,961 community-dwelling older adults with advanced dementia in Ontario, Canada to investigate this question of optimal timing of palliative care initiation; specifically, during a window of time when these individuals have both limited life expectancy and increased care needs as indicated by a referral for home care services [6]. They hypothesized that palliative care initiation within 30 days of this window would reduce acute care utilization (e.g., emergency department use, hospitalizations, and ICU admissions), which may not align with the values and goals of this population. Their analytic approach used propensity-score overlap weighting to minimize confounding by indication, given that sicker patients are more likely to receive palliative care. Overlap weighting is a relatively novel method that lends greater emphasis to individuals who could have easily ended up in either group (palliative care vs. no palliative care) by assigning larger weights to those with a likelihood of being in either group close to approximately 50% [7]. Overlap weighting produces an estimate of the average treatment effect in the overlap population, answering the question, “Among those who could realistically receive palliative care, does initiation of palliative care actually lead to reduced acute care utilization?” Unexpectedly, in the overlap-weighted analysis, the authors found that palliative care was associated with an increased hazard of hospitalization within the first 90 days (HR 1.43 at 30 days, 95% CI = 1.25–1.64). Importantly, this increased hazard of hospitalization was observed up to 180 days for hospitalizations with palliative intent (HR 1.45, 95% CI = 1.13–1.87), but not hospitalizations with non-palliative intent (HR 0.94, 95% CI = 0.79–1.13). The classification of “palliative intent” was based on diagnostic codes used during the hospital admission, including whether the hospital's primary diagnosis was coded as palliative care and whether the patient was seen by a palliative care physician during the admission [8]. The association between palliative care initiation and hospitalization decreased over time and was no longer statistically significant from 180 days up to 1 year. There was also an increased hazard of ICU admissions in the first 60 days, but not beyond 90 days. Palliative care was not associated with ED use. The study authors highlight several potential explanations for these surprising findings, including unmeasured confounders and mechanisms, which we summarize in Figure 1. First, the final cohort is a highly selected subset of individuals with dementia with likely unique care needs that prompted palliative care initiation. Among 871,986 community-dwelling adults with dementia, 50,961 individuals met eligibility criteria. Of these, 1115 (2%) received palliative care within 30 days of homecare initiation. The final cohort of individuals who ultimately received palliative care was quite sick, with 50.3% alive at 1 year compared to 88.2% of individuals who did not receive palliative care. While the authors minimized confounding through the overlap weighted analysis, many factors are not captured in administrative data. Second, many individuals referred to palliative care were likely in a state of relative crisis and rapidly evolving health needs, which increased their risk of hospitalization. For example, 23% (n = 258/1115) of the persons with dementia in the palliative care group had their first contact with the palliative care team during a hospitalization following the initiation of home care services. For those seeing palliative care for the first time in the community, palliative care initiation may reflect a “paradox of discovery,” in which the palliative care team uncovers an acute illness or uncontrolled symptom that prompts early hospitalizations because (1) there was insufficient time to engage in goals of care conversations, (2) the hospitalization aligned with the goals of patients and caregivers at the time, or (3) the individual's symptoms could not be safely managed within their current support system. Third, there are important differences between palliative care delivery in the United States and Canada. In general, physician-delivered palliative care in Canada is more likely to be provided by individuals without formal specialized palliative care training (e.g., family medicine) [9]. Additionally, in many cases, palliative care may not be provided by physicians (e.g., only 66% of individuals in this study's palliative care group had physician involvement). The authors also note that the healthcare system in Ontario, Canada lacks a rich infrastructure of hospice networks with a less well-defined hospice benefit compared to the United States, in which hospice is a highly regulated insurance benefit for individuals with a < 6-month prognosis. What are the study's implications? For now, we caution against using these results to dissuade clinicians from initiating palliative care services for persons with dementia whom they believe may benefit. While palliative care services may unmask high care needs that lead to acute hospitalizations, longitudinal follow-up can help address distressing symptoms and improve illness understanding to ensure that future treatment aligns with values and preferences. As the authors allude, the primary implication may be that merely enrolling high-risk individuals with advanced dementia in palliative care services may not prevent acute care utilization unless accompanied by more robust support. These individuals have needs beyond what palliative care services can provide, including unpaid and paid caregiver support and caregiver training. The fact that palliative care initiation was associated with increased hospitalizations with palliative intent but not hospitalizations without palliative intent suggests that some hospitalizations may have been avoided if better supports were in place. In randomized clinical trials (RCTs), palliative care interventions among persons with dementia have generally found an increase in the documentation of advance directives and goals of care discussions without a clear impact on hospitalizations [10]. Most recently, the Indiana Palliative Excellence in Alzheimer Care Efforts (IN-PEACE) RCT compared a predominantly telehealth palliative care intervention led by a trained nurse or social worker to usual care among patients with moderate to severe dementia [11]. While the trial did not report a statistically significant difference in its primary outcome of neuropsychiatric symptoms, it did show fewer combined ED and hospitalization events with palliative care (mean 1.06 vs. 2.37 over the 24-month study, between-group difference −1.31 [95% CI, −1.93 to −0.69]). Thus, palliative care under the right circumstances may reduce acute care use [12]. Caring for persons with dementia in the community, therefore, requires a multifaceted approach. Several models of care have been described, but their efficacy and sustainability remain unclear [13]. In the United States, the Centers for Medicare and Medicaid Services is evaluating the Guiding an Improved Dementia Experience (GUIDE) model to deliver comprehensive, coordinated dementia care to persons with dementia and their caregivers. Models such as the Care Ecosystem, which are centered on care team navigators, have shown benefits in reducing emergency department visits and improving quality of life [14]. Most recently, the Dementia Care Study (D-CARE) found no significant difference in its primary outcomes (behavioral symptoms and caregiver strain) between a health system-based dementia care program (led by dementia care specialists, such as a nurse practitioner) versus community-based program (led by care consultants, such as a social worker) versus usual care [15]. However, the results pave the way for future evaluations to identify whether benefits are seen when targeting specific populations. For example, models specifically adapted to homebound older adults with dementia, such as Dementia Care Quality at Home, may prove more effective [16]. Finally, future research is needed to guide referral criteria for palliative care among persons with dementia [5, 17]. There is currently no consensus on referral criteria for primary palliative care (i.e., palliative care delivered by healthcare professionals who are not specialists). While a 2025 Delphi study identified several criteria for specialist palliative care referral among persons with dementia, additional studies are needed to examine how these criteria can be integrated into care models to ensure sustainability and to identify community-dwelling individuals with limited life expectancy [18-20]. In summary, Quinn and colleagues examine the important question of optimal timing for palliative care for community-dwelling people with advanced dementia in a system with limited palliative care and hospice resources. What was uncovered is that palliative care initiation alone without a robust network of support may be inadequate to reduce costly acute care use in this high-need population. Instead, comprehensive dementia care must address the physical and psychosocial needs of persons with dementia and their caregivers, including the provision of additional in-home support. These results invite researchers and healthcare policymakers to rethink the current care model in Canada and other settings. All authors participated in the preparation of this editorial and have approved the final manuscript. W. James Deardorff was supported by the National Institute on Aging (NIA) under grants K76AG094730 and P30AG044281. Ashwin Kotwal was supported by the NIA under grant K23AG065438. W. James Deardorff was supported by the National Institute on Aging (NIA) under grants K76AG094730 and P30AG044281. Ashwin Kotwal was supported by the NIA under grant K23AG065438. The authors have nothing to report. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The authors declare no conflicts of interest. This publication is linked to a related research article by Quinn et al. To view this article, visit https://doi.org/10.1111/jgs.70282.