Abstract
Purpose:
Multimorbidity is associated with increased intensity of end-of-life healthcare. This association has been examined by number but not type of conditions. Our purpose was to understand how intensity of care is influenced by multimorbidity within specific chronic conditions to provide guidance for interventions to improve end-of-life care for these patients.
Methods:
We identified adults cared for in a multihospital healthcare system who died between 2010–2017. We categorized patients by 4 primary chronic conditions: heart failure, pulmonary disease, renal disease, or dementia. Within each condition, we examined the effect of multimorbidity (presence of 4 or more chronic conditions) on hospital and ICU admission in the last 30 days of life, in-hospital death, and advance care planning (ACP) documentation >30 days before death. We performed logistic regression to estimate associations between multimorbidity and end-of-life care utilization, stratified by the presence or absence of ACP documentation.
Results:
ACP documentation >30 days before death was associated with lower odds of in-hospital death for all 4 conditions both in patients with and without multimorbidity. With the exception of patients with renal disease without multimorbidity, we observed lower odds of hospitalization and ICU admission for all patients with ACP >30 days before death.
Conclusions:
Patients with dementia and multimorbidity had the highest odds of high-intensity end-of-life care. For patients with dementia, heart failure, or pulmonary disease, ACP documentation >30 days before death was associated with lower likelihood of in-hospital death, hospitalization, and ICU use at end-of-life, regardless of multimorbidity.
Introduction
Multimorbidity is experienced by at least a quarter of American adults and is occurring at younger ages. 1 -4 Among US Medicare enrollees, 35% of adults have 4 or more chronic conditions 5 and incur two-thirds of Medicare costs. 6 Patients with multimorbidity are more likely to experience emergency department visits and hospitalizations at end of life, 7 -10 which may not align with patient and family goals of care. Rather, such healthcare use may indicate unmet symptom needs, 11 patient inability to provide self-care, 12 patient distress, 13 or high caregiver burden. 14,15 Although previous studies have noted that a higher number of chronic conditions is associated with higher intensity healthcare use, 16,17 the specific types of conditions may also contribute to this association. 18
Advance care planning (ACP) and its documentation by providers offers an opportunity for patients to clarify their goals for treatment with their physicians, and if so desired, opt out of higher intensity care. However, ACP may be difficult for patients with multiple conditions who rely on a number of providers who may not communicate with each other, and none of whom may take responsibility for overseeing global aspects of care such as ACP. Without the kind of coordination that supports a shared knowledge of a patient’s care plans, patients may experience increased stress, medical errors, or adverse events. 19 -21
For palliative and end-of-life care to align with care that is concordant with patient wishes, additional data are needed on the relationship between end-of-life care and multimorbidity within specific types of conditions. Such data can inform which specialties may need additional support in providing primary palliative care to patients, an important consideration given the limitation of available clinicians trained in palliative care. 22 In this observational study, we examined the predictors of high-intensity end-of-life care in the last 30 days of life and in-hospital death among patients with dementia, heart failure, pulmonary disease, or advanced renal disease, with or without multimorbidity. We also examined the association of ACP documentation prior to the last month of life with intensity of end-of-life healthcare among these groups of patients, and the influence of multimorbidity on this association. These data on how intensity of care is associated with multimorbidity within specific chronic conditions and how ACP may affect care received may help inform the adaptation and design of interventions to improve end-of-life care.
Methods
Study Population
We used the Cambia Metrics database, which incorporates a linkage of Washington State death certificates to the EHR of the UW Medicine system. The largest public health system in the Puget Sound region, UW Medicine comprises an academic medical center, a county safety-net hospital that is the only Level 1 trauma center for a 5 state region, 2 community hospitals, outpatient clinics, and a National Cancer Institute-designated comprehensive cancer center.
In our analysis, we included patients ages 18 or older who died between 2010–2017 with one or more of the following conditions identified in the 24 months preceding death: poor prognosis cancer (metastatic solid tumor or aggressive hematologic malignancy, e.g. acute myeloid leukemia), chronic liver disease, chronic pulmonary disease, coronary artery disease, dementia, diabetes with end-organ damage, end-stage renal disease, heart failure, and peripheral vascular disease. 23 These conditions, along with accidents, suicide, and infectious disease, comprise the most common causes of death in America. 24 We chose 4 specific conditions to focus on based on their prevalence and contributions to death in the United States: chronic pulmonary disease, dementia, end-stage renal disease, and heart failure. 25,26 We refer to each of these 4 conditions as a primary condition. We have previously evaluated ACP and high-intensity end-of-life healthcare use for patients with cancer with or without multimorbidity in this database 27 and therefore did not include cancer as a primary condition group in this analysis. We included patients under the age of 65 because, although the average age of comorbidity onset is dropping in the United States, few data are available in the literature describing multimorbidity in adults under age 65. 1,2 We identified diagnoses in the EHR based on International Classification of Diseases, version 9 (ICD-9) and version 10 (ICD-10) codes using adapted Dartmouth Atlas criteria. 23
Outcomes
Our outcomes were high-intensity health care use and ACP documentation. Using previously published criteria, we defined high-intensity health care use in the last month of life as inpatient hospitalizations, ICU admissions, and in-hospital death. 28 -30 We used the EHR to identify healthcare utilization at an academic medical center and a public safety net hospital in the UW Medicine healthcare system and Washington State death certificates to identify place of death as in- or out-of-hospital. We recorded the presence/absence of ACP documentation more than 30 days prior to death, defining ACP documentation as a healthcare directive, a durable power of attorney for healthcare, or a Physician Order for Life-Sustaining Treatments present in the patient’s EHR. We focused on presence/absence of ACP documentation more than 30 days before death to ensure that the ACP documentation occurred before the outcomes of hospitalization or ICU admission in the last 30 days of life.
Predictors
We examined 2 predictors: 1) multimorbidity versus no multimorbidity within each of the 4 primary conditions described above as identified from the EHR according to ICD-9/10 codes during the 24 months prior to death; and 2) documentation of ACP more than 30 days before death as a predictor of high-intensity care, stratified by presence of multimorbidity within each primary condition. For these analyses, we defined multimorbidity as 4 or more chronic conditions, including the primary condition within each analysis. We chose this threshold based on the high prevalence of 4 or more chronic conditions among Medicare enrollees. 5,31
Covariates
We identified confounders a priori. Social determinants of health, including race, level of education (completed some college or higher), and marital status (married or not married at the time of death) were identified from Washington State death certificates. Age, sex, and insurance status (private, Medicare, Medicaid, other) were identified from the EHR.
Statistical Analysis
For all analyses, we examined subjects by primary condition groups (e.g., chronic pulmonary disease, dementia, heart failure, renal disease). This stratified approach accounted for lack of independence across comorbidity groups, that is, patients could appear in multiple analyses. Within each primary condition group, we compared patients with the primary condition and multimorbidity to those with the primary condition without multimorbidity. First, we described the characteristics of patients in each primary condition group, stratified by multimorbidity, using frequencies for categorical variables and means with standard deviation for continuous variables. Patients within each primary condition group could have varying comorbidities. We dichotomized responses for the study outcomes (i.e., ACP documentation more than 30 days before death, inpatient hospitalization in the last month of life, ICU admission in the last month of life, in-hospital death). We estimated multivariable logistic regression models to determine the odds of having each outcome if a patient had multimorbidity, controlling for a priori confounders of age, race, sex, marital status, insurance status, and highest education level attained. The reference group within each primary condition group was patients without multimorbidity. We report odds ratios (OR) with 95% confidence intervals (CI). We set statistical significance at 0.05 for all analyses.
To explore the association between ACP and healthcare use in the last month of life, we stratified our analysis by multimorbidity within each condition. We evaluated patients within each primary condition group with or without multimorbidity, with our reference group being patients without ACP. The UW Institutional Review Board approved this study and issued a waiver of consent to access decedent records in accordance with Washington State law. We performed all analyses using Stata version 16.0.
Results
Demographics
This study included 27,711 unique patients, of whom 6,746 had heart failure, 6,738 had pulmonary disease, 5,803 had renal disease, and 2,520 had dementia. Overall, multimorbidity was more common among men and among people of color (Table 1). Across most conditions, patients with multimorbidity were older, and education level varied among groups. Medicare was the most frequent insurer across all primary conditions. Patients with multimorbidity and dementia had a higher frequency of Medicaid insurance compared to their counterparts without multimorbidity and other conditions.
Demographics of Patients by Primary Condition Groups With or Without Multimorbidity (MM).*
* Patients are not in mutually exclusive clusters and can appear in multiple primary condition groups.
Inpatient Hospitalization
Across all primary conditions, patients with multimorbidity were more frequently hospitalized compared to patients with the same primary condition without multimorbidity (Table 2). The highest observed odds of hospitalization in the context of multimorbidity was among patients with dementia (OR 2.26, 95% CI 1.84–2.77) (Table 3). Patients with multimorbidity and pulmonary disease (OR 1.94, 95% CI 1.74–2.17), renal disease (OR 1.52, 95% CI 1.36–1.70) or heart failure (OR 1.45, 95% CI 1.31–1.61) also experienced increased odds of hospitalization compared to patients with the same primary condition without multimorbidity.
Prevalence of Healthcare Utilization and ACP by Primary Condition Group.
Association Between Multimorbidity, Healthcare Use, and ACP at the End of life by Primary Condition Group.
p < 0.01, bolded; p < 0.05, italics.
Multiple logistic regression, adjusted a priori for age, sex, marital status, insurance, race, education.
a Reference group, HF without MM, n = 3800.
b Reference group, pulmonary disease without MM, n = 4380.
c Reference group, renal disease without MM, n = 3122.
d Reference group, dementia without MM, n = 1869.
ICU Admission
Across all primary condition groups, patients with multimorbidity were more frequently admitted to the ICU at end of life compared to those without multimorbidity. The percentage difference ranged from 3% among patients with heart failure (30% versus 27%) to 11% for patients with dementia (21% versus 10%) (Table 2). In adjusted models, patients with multimorbidity had significantly higher odds of ICU admission compared to those with the same primary condition but without multimorbidity for all 4 primary conditions (Table 3).
In-Hospital Death
In-hospital deaths were significantly higher for patients with multimorbidity in 3 of the 4 primary condition groups: dementia (40% vs. 26%, OR 1.69, 95% CI 1.39–2.05), pulmonary disease (48% vs. 46%, OR 1.14, 95% CI 1.03–1.27) and renal disease (50% vs. 49%, OR 1.12, 95% CI 1.01–1.25). Although patients with multimorbidity and heart failure were slightly more likely to die in the hospital (52% vs. 51%), this difference was not associated with increased odds of in-hospital death (OR 1.01, 95% CI 0.91–1.12).
ACP Prior to Last Month of Life and End-of-Life Healthcare Use
For patients with heart failure, pulmonary, or renal disease, less than half of patients had ACP documentation >30 days before death, regardless of multimorbidity. However, 51% of patients with dementia and multimorbidity had ACP (Table 2). We found significantly higher odds of ACP documentation >30 days before death across all 4 primary conditions for people with multimorbidity versus those without multimorbidity: heart failure (OR 2.14, 95% CI 1.93–2.37), pulmonary disease (OR 1.97, 95% CI 1.77–2.20), dementia (OR 1.97, 95% CI 1.63–2.37) and renal disease (OR 1.93, 95% CI 1.73–2.15) (Table 3).
After adjusting for confounders, the relationship between ACP documented prior to the last month of life and healthcare utilization varied by primary condition as well as by presence of multimorbidity (Table 4). Patients with heart failure, pulmonary disease, or dementia who had ACP documentation had significantly lower odds of all healthcare utilization than those without ACP documentation, regardless of whether or not they had multimorbidity. Among patients with heart failure and multimorbidity, odds of inpatient hospitalization (OR 0.72, 95% CI 0.62–0.84), ICU admission (OR 0.72, 95% CI 0.61–0.85) and in-hospital death (OR 0.69, 95% CI 0.59–0.81) were significantly lower if ACP was present. These results are similar to those with heart failure without multimorbidity for hospitalization (OR 0.63, 95% CI 0.53–0.75), ICU admission (OR 0.63, 95% CI 0.52–0.75) and in-hospital death (OR 0.61, 95% CI 0.52–0.70).
Association Between ACP 30+ Days Prior to Death and Healthcare Use at End-of-Life, Stratified by MM Status Within Each Primary Condition Group.
p < 0.01, bolded; p < 0.05, italics.
Multiple logistic regression, adjusted a priori for age, sex, marital status, insurance, race, education.
a Reference group, HF + MM without ACP, n = 1701.
b Reference group, HF without MM, without ACP, n = 2806.
c Reference group, pulmonary disease + MM without ACP, n = 1372.
d Reference group, pulmonary disease without MM, without ACP, n = 3192.
e Reference group, renal disease + MM without ACP, n = 1468.
f Reference group, renal disease without MM, without ACP, n = 2173.
g Reference group, dementia + MM without ACP, n = 317.
h Reference group, dementia without MM, without ACP, n = 1228.
Patients with pulmonary disease as their primary condition had a significant association between ACP >30 days and reduced likelihood of ICU admission whether they had multimorbidity (OR 0.80, 95% CI 0.66–0.96) or not (OR 0.82, 95% CI 0.68–0.98). Similarly, all patients with pulmonary disease and ACP >30 days compared to those without ACP had lower odds of inpatient admission, regardless of multimorbidity (multimorbidity: OR 0.80, 95% CI 0.68–0.96; no multimorbidity: OR 0.68, 95% CI 0.58–0.81) and in-hospital death (multimorbidity: OR 0.76, 95% CI 0.65–0.90; no multimorbidity: OR 0.70, 95% CI 0.61–0.80).
For patients with dementia and multimorbidity, ACP was significantly associated with lower odds of in-hospital death (OR 0.63, 95% CI 0.45–0.87), inpatient hospitalization (OR 0.69, 95% CI 0.50–0.96), or ICU admission (OR 0.49, 95% CI 0.33–0.74). ACP was also significantly associated with a reduction in intensity of care for all outcomes among patients with dementia without multimorbidity. These patients had the lowest odds of in-hospital death (OR 0.60, 95% CI 0.48–0.77), hospitalization (OR 0.48, 95% CI 0.36–0.63), or ICU admission (OR 0.31, 95% CI 0.20–0.47).
For patients with renal disease and multimorbidity, ACP was associated with lower odds of inpatient admission (OR 0.77, 95% CI 0.66–0.90), in-hospital death (OR 0.72, 95% CI 0.61–0.84) and ICU admission (OR 0.83, 95%CI 0.70–0.99). Among patients with renal disease who did not have multimorbidity, those with ACP >30 days before death had significantly lower odds of in-hospital death (OR 0.76, 95% CI 0.65–0.89). We did not observe a significant association between ACP and ICU admission (OR 0.91, 95% CI 0.76–1.10) or inpatient admission (OR 0.87, 95% CI 0.73–1.03).
Discussion
In this analysis, we found that within each type of common, serious illness, patients with multimorbidity were more likely to experience a high number of hospitalizations, ICU admissions and in-hospital death, and that ACP was associated with lower odds of such outcomes. To ensure that such high-intensity care is aligned with patient goals for care, our findings suggest that ACP may play an important role for patients with multimorbidity. 32 -35 These patients may receive more high-intensity care than they would choose otherwise if they had had a chance to participate in ACP. 36 Patients with multimorbidity would benefit from having a provider be a “point person” for ACP to moderate the effects of the fragmented care and presence of multiple providers frequently experienced by patients with multimorbidity. 37
Within primary condition groups, we found that the odds of high-intensity healthcare at end of life were highest among patients with multimorbidity and dementia compared to patients with dementia without multimorbidity. This finding may be linked to a poor understanding by healthcare proxies, family caregivers and clinicians of patients’ preferences; previous work has reported that less than half of patients with dementia and their proxies had discussed health care wishes with health care providers. 38 Without this knowledge, surrogate decision makers may default to high-intensity care as the result of care facility or institutional policies, 39 even though this may be discordant with patient care preferences.
Overall, the high amount of hospitalization observed herein may be in part because ACP has been more slowly adopted for patients with other serious illnesses compared to patients with cancer. 40 Given the multiple decisions that patients with multimorbidity and serious illness face with respect to treatment options, 41 -43 innovative programs such as serious illness care models that integrate care coordination and palliative support have been found to increase ACP among adults with serious illness 44 while potentially lowering healthcare utilization for these patients. 45
Our finding that patients with multimorbidity more frequently had ACP documentation yet were more likely to experience an ICU admission compared to patients with the same primary condition without multimorbidity may be attributed to a number of circumstances. Although data suggest ACP documentation is often associated with preferences for lower intensity of care, 46 -48 patients’ preferences may vary depending on their prior experiences; patients who had previously undergone successful interventions for any of their multiple conditions, even at the level of the ICU, may consider more intensive care as acceptable. Alternatively, healthcare proxies or family caregivers may be unaware of patient preferences due to lack of access to ACP documentation and inability of patients to articulate preferences. It is also possible that an unexpected and abrupt change in clinical course for patients may result in the delivery of high-intensity care despite previously documented wishes. 49 For example, if these patients experienced an unexpected deterioration when they came to the hospital, or an infection that may be amenable to treatment, then they may have been admitted to the ICU for stabilization before clarifying their goals of care or before the admitting physician was aware of ACP documentation. 50
There are a number of limitations to this study. First, we used Washington State death certificates; patients dying outside of Washington State would be missed in this analysis. The geography of Seattle and Washington state makes this less problematic than some areas, but this remains a potential limitation. Second, for hospital and ICU admissions, patients may have received care in other healthcare systems, but these data were not available to us as we relied on one EHR, a known limitation of data linkages. 51 However, this was not a limitation for the outcome of in-hospital death, as this information is available on the death certificates for all patients. Since we saw similar trends across these 3 outcomes, this suggests that care at other hospitals may not be an important source of bias. Third, these analyses relied on evidence of ACP documents in the EHR and did not use provider notes or other evidence about ACP discussions. We may therefore have missed important information relevant to ACP activities. Fourth, we relied on ICD-9 and ICD-10 coding to determine the presence of various conditions in the EHR. While this indicates the presence of a condition, it does not give us information about condition severity. Finally, this represents a decedent analysis which can introduce potential bias. 52 However, since our goal was to examine end-of-life care, and since the conditions of interest were all highly likely to precede the observation period, this is less of a concern.
As chronic diseases and multimorbidity become more prevalent, it is important that we evaluate and improve end-of-life care for this population. These patients likely have unaddressed symptoms and unmet palliative care needs 53 that may contribute to our finding that multimorbidity is associated with more high-intensity end-of-life care. While additional palliative care assessments and interventions can benefit all patients, our findings indicate that patients with dementia and multimorbidity may need more targeted efforts to avoid high-intensity end-of-life care. Additional interventions to assess symptom burden, caregiver support, care coordination, and patient needs may help decrease the frequency of high-intensity care experienced by adults with multimorbidity, and ensure patients receive care aligned with their preferences.
Footnotes
Authors’ Note
Presentation: 2019 Annual Meeting of the American Geriatrics Society; accepted at the 2020 Annual Assembly of the American Academy of Hospice and Palliative Medicine (meeting canceled due to COVID-19).
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Pilot grant from the Palliative Care Research Cooperative Group to CLM; parent grant funded by the National Institute of Nursing Research U2CNR014637. This work was also supported by National Institutes of Health’s National Heart, Lung, and Blood Institute Grant K12HL137940 and the Cambia Health Foundation (JRC).
