Abstract
Background:
Serious illness is characterized by high symptom burden that negatively affects quality of life (QOL). While palliative care research has highlighted symptom burden in seriously ill adults with cancer, symptom burden among those with non-cancer serious illness and multiple chronic conditions has been understudied. Latent class analysis is a statistical method that can be used to better understand the relationship between severity of symptom burden and covariates, such as the presence of multiple chronic conditions. While latent class analysis has been used to highlight subgroups of seriously ill adults with cancer based on symptom clusters, none have incorporated multiple chronic conditions.
Objectives:
The objectives of this study were to (a) describe the demographic and baseline characteristics of seriously ill adults at the end of life in a palliative care cohort, (b) identify latent subgroups of seriously ill individuals based on severity of symptom burden, and (c) examine variables associated with latent subgroup membership, such as QOL, functional status, and the presence of multiple chronic conditions.
Methods:
A secondary data analysis of a palliative care clinical trial was conducted. The latent class analysis was based on the Edmonton Symptom Assessment System, which measures 9 symptoms on a scale of 0 to 10 (e.g., pain, fatigue, nausea, depression, anxiousness, drowsiness, appetite, well-being, shortness of breath). Clinically significant cut-points for symptom severity were used to categorize each symptom item in addition to a categorized total score.
Results:
Three latent subgroups were identified (e.g., low, moderate, and high symptom burden). Lower overall QOL was associated with membership in the moderate and high symptom burden subgroups. Multiple chronic conditions were associated with statistically significant membership in the high symptom burden latent subgroup. Older adults between 65 and 74 had a lower likelihood of moderate or high symptom burden subgroup membership compared to the low symptom burden class.
Discussion:
Lower QOL was associated with high symptom burden. Multiple chronic conditions were associated with high symptom burden, which underlines the clinical complexity of serious illness. Palliative care at the end of life for seriously ill adults with high symptom burden must account for the presence of multiple chronic conditions.
Keywords: multiple chronic conditions, palliative care, quality of life, serious illness, symptom burden
For seriously ill adults, the illness experience is accompanied by the presence of high symptom burden, functional impairment, high mortality risk, frequent health care utilization, and high caregiver stress, which can result in poor quality of life (QOL; Kelley, 2014). Among the defining characteristics of serious illness, poor QOL is often linked to symptom burden among those with serious illness, which can be intensified in the presence of multiple chronic conditions (MCC; Bernacki & Block, 2014; Grembowski et al., 2014). MCC is defined as an individual having two or more chronic conditions resulting in daily living limitations, requiring regular care, and is associated with clinical complexity (Grembowski et al., 2014; Parekh et al., 2011). There is limited research that examines how the presence of MCC is associated with symptom burden or QOL in those with serious illness at the end of life in the palliative care context. Palliative care is an approach to care that seeks to improve the serious illness experience through symptom management, advanced communication, care coordination, caregiver support, addressing spiritual and psychological needs with a goal of improving QOL (Kelly & Morrison, 2015). Seriously ill adults with MCC who report frequent burdensome symptoms can have increased health care utilization, complications, and adverse events (Kamal et al., 2015; Kelley & Morrison, 2015). Symptom burden can vary depending on the clinical presentation of different diagnoses, and as most symptom studies have been conducted in those with cancer. While those with cancer may also have MCC, there is a need to better understand symptom burden and its severity in seriously ill adults with MCC to provide goal-concordant and effective care (Kamal et al., 2015; Sanders et al., 2018).
Latent class analysis—a statistical method used to identify unmeasured latent subgroup membership using observed variables (categorical or continuous)—can be operationalized to increase the understanding of the severity of symptom burden among seriously ill adults (Miaskowski et al., 2015; UCLA Institute for Digital Research & Education, 2020). Prior studies have used latent class analyses (LCA) to identify subgroups based on symptom burden or have identified symptom clusters to examine relationships with QOL in patients with cancer (Astrup et al., 2017; Ganesh et al., 2017; Johnstone et al., 2017; Miaskowski et al., 2015; Papachristou et al., 2018). Miaskowski et al. (2015) identified latent subgroups based on symptom profiles and examined the association with demographic and clinical characteristics among those with cancer; the subgroup associated with “all high” symptoms had poorer QOL. Papachristou et al. (2018) found similar results when using LCA to identify subgroups of patients with cancer based on symptom burden, revealing that those in the highest symptom burden category had poorer QOL outcomes and higher comorbidity scores. Astrup et al. (2017) found that individuals in the highest symptom burden group had poorer functional status, multiple comorbidities, and poorer QOL outcomes among those with cancer. However, the evidence-base related to MCC and symptom burden in serious illness remains limited.
To date, few studies have examined symptom burden subgroups in non-cancer serious illnesses (Kamal et al., 2017; Yu et al.; 2016). Kamal et al. (2017) examined the symptom profiles and associated diagnoses for a variety of serious illnesses. They found that individuals with pulmonary disease or cancer had higher symptom burden than other diagnoses. In individuals with advanced heart failure, Yu et al. (2016) studied symptom clusters and QOL in those with heart failure using exploratory factor analysis and identified three factor types, including a distress cluster (i.e., shortness of breath, anxiety, depressive symptoms), a decondition cluster (i.e., fatigue, drowsiness, nausea, low appetite), and a discomfort cluster (i.e., pain, general discomfort), each of which was predictive of QOL. Greater focus on the management of symptoms at the end of life can improve QOL and reduce unnecessary hospitalization for seriously ill individuals (Kelley & Morrison, 2015). However, there is a significant degree of variation in the symptom burden experience of seriously ill individuals, which can make it challenging to develop specifically tailored palliative care approaches. In addition, the lack of inclusion of MCC further complicates efforts to provide effective, goal-concordant care. Latent class analysis can be operationalized to create latent subgroups of seriously ill adults at the end of life based on severity of symptom burden, from which a latent class variable then becomes the outcome variable representing symptom burden. Identifying latent subgroups of seriously ill adults can be useful for examining demographic and clinical variables associated with membership in the latent subgroups. Therefore, the objectives of this study were to (a) describe the demographic and baseline characteristics of seriously ill adults at the end of life in this palliative care cohort; (b) identify latent subgroups of seriously ill individuals based on severity of symptom burden; and (c) examine variables associated with latent subgroup membership, such as QOL, functional status, and the presence of multiple chronic conditions.
Methods
Design and Sample
This study is a cross-sectional secondary data analysis of randomized clinical trial data that examined the effect of continuing versus discontinuing statins at the end of life in seriously ill adults receiving palliative care. The initial trial results, the Multisite Randomized Trial of Continuing Versus Discontinuing Statins, are published elsewhere (Kutner et al., 2015). While the focus of the original trial was on statin discontinuation, the study included seriously ill adults at the end of life with a broad array of MCC, including cancer. It used validated measurement tools in palliative care, which provided rich data to meet the objectives of the secondary analysis.
Multisite Randomized Trial of Continuing Versus Discontinuing Statins
In the original trial, 381 participants met eligibility criteria for inclusion such as statin use for 3 months or longer for primary or secondary prevention of heart disease, deemed as having a life-limiting illness by a treating physician and having a declining functional status as measured by a reduction in the Australia-Modified Karnofsky Performance Status (AKPS) score to less than 80% in 3 months (Abernethy et al., 2005). Patient-centered secondary outcomes included performance status, QOL (McGill Quality of Life Questionnaire [MQOLQ]), symptom burden (Edmonton Symptom Assessment Scale [ESAS]), number of non-statin medications, statin-related effects, and patient satisfaction (Bruera et al., 1991; Cohen et al., 1997). The trial was registered with clinicaltrials.gov (NCT01415934). The study was conducted from 2011 to 2013 across 15 primary study sites spanning rural, suburban, and urban areas across the United States and Australia. Of note, participants in Australia receive universal health insurance called Medicare. For this secondary data analysis, participants who had baseline ESAS data were included. This study was exempt by the institutional review board at the respective institution as it contained solely deidentified clinical trial data.
Conceptual Model
To inform the conceptual basis for this study, the objectives and covariates identified for inclusion in our analyses were identified by a conceptual model that effectively guides how to integrate palliative care in serious illness and multiple chronic conditions research (Murali et al., 2020). In this model, relevant contextual factors that influence the needs and services received by seriously ill adults with MCC are conceptualized. Ultimately, the conceptual model highlights the primary outcomes of interest associated with integrating palliative care for seriously ill adults with MCC, such as symptom burden and QOL (Murali et al., 2020).
Demographic and Clinical Covariates
The demographic and clinical variables included in this analysis were: age, gender, race/ethnicity, education, insurance status, polypharmacy, hospice at baseline, presence of MCC, functional status, and QOL. Age was categorized into three groups: individuals younger than 65, between 65–74, and 75 and older based on categorizations in the literature to discern age-related differences among predictors of latent subgroup membership (Lee et al., 2018). Hospice enrollment was also considered a clinical covariate associated with serious illness.
Comorbidities in the sample were measured at baseline using the Charlson Comorbidity Index (CCI; Charlson et al., 1987). Participants were considered to have MCC if they had two or more diagnoses requiring regular medical care and limitations in the activities of daily living. Of note, all participants in the sample had functional status limitations and were receiving standard medical care, therefore, meeting the accepted definition for MCC in the literature (Grembowski et al., 2014; Parekh et al., 2011). The inter-rater reliability of the CCI is 0.74 to 0.945 in older adult cancer cohorts (Hall et al., 2004). Test–retest reliability is 0.92 (intraclass correlation coefficient; Katz et al., 1996). The CCI has been shown to have content validity across a variety of settings and moderate-to-good correlation with other measures of comorbidity (Hall et al., 2004, 2006).
The 9-item ESAS (e.g., pain, fatigue, nausea, depression, anxiousness, drowsiness, appetite, well-being, shortness of breath) was used to measure symptom burden. Participants self-reported symptoms on a 10-point numerical scale. Because the individual ESAS items were non-normally distributed, a categorical variable was created for each item based on symptom severity using qualitative labels (e.g., none, mild, moderate, severe), which are clinical cut-points reported previously by Hui and Bruera (2017) as follows: 0 (none), 1–3 (mild), 4–6 (moderate) and 7–10 (severe). The ESAS has been tested numerous times for reliability and validity; internal reliability shows a Cronbach’s alpha of 0.79. Test–retest reliability of the ESAS is high with a Cronbach’s alpha of 0.8. Types of validity testing have included convergent validity, predictive validity, and construct validity (Hui & Bruera, 2017). The total ESAS score—for which clinical cut-points are not reported, as they are not routinely used in clinical practice—was categorized using quartiles of the total ESAS score within the sample to capture the severity of symptom burden for analysis. The individual items and the quartile categories of the total ESAS score together were representatives of symptom burden.
QOL was measured using the MQOLQ (Cohen et al., 1997). This questionnaire has been shown to be reliable and valid across a variety of settings. The original complete MQOLQ had an internal consistency of 0.8 (Cronbach’s alpha). Reliability testing in the palliative care setting resulted in a Cronbach’s alpha of 0.70 (Cohen et al., 1997). The overall MQOLQ score is calculated as a mean of the values for each subscale, and the final score ranges from 0 to 10, in which higher scores indicate higher QOL. There are no clinically significant cut-points for the overall QOL score reported in the literature; therefore, quartile categories were created for analysis.
The AKPS is a modified version of the original Karnofsky Performance Status (KPS), which was validated for use in palliative care (Abernethy et al., 2005). The Kappa coefficient for agreement between the original scale and the AKPS was 0.84 (Abernethy et al., 2005). The AKPS was used to examine the functional status of participants. All study participants had functional status limitations or an AKPS score of 70 or lower. Those with an AKPS of 50 or below require “considerable assistance and frequent medical care.” Those with an AKPS score higher than 50 are considered less functionally impaired.
The number of medications that define high levels of polypharmacy varies considerably across studies. Because many of the participants in this sample had serious illnesses with MCC, polypharmacy was substantial. We created a categorical variable using the qualitative labels of minor (less than five medications), major (5 to 10 medications), excessive (11 to 15 medications), and very excessive (greater than 15 medications) polypharmacy based on definitions extrapolated from a systematic review by Masnoon et al. (2017).
Statistical Analysis
Descriptive statistics were performed to analyze continuous and categorical variables using Stata Statistical Software (Release 17; College Station, TX: StataCorp LLC.). Central tendency measures were conducted for continuous variables. For continuous variables, assumptions of normality were tested, and categorical variables were summarized with frequencies and percentages. LCA was used to identify subgroups within the sample based on symptom burden (UCLA Institute for Digital Research & Education, 2020), using Mplus (Version 8.3; Los Angeles, CA: Muthén & Muthén). The indicator variables included in the LCA were representative of severity of symptom burden, including each of the 9 ESAS symptom items categorized based on severity using clinical cut-points and quartile categories of the total ESAS score in a reversed Poisson distribution. We reversed the total ESAS score quartile categories using a Poisson distribution for the LCA given the substantial degree of symptom burden in the seriously ill sample equating to high total ESAS scores. Therefore, the reference category of “0” included individuals with the highest total ESAS scores, indicating high symptom burden. Maximum-likelihood estimation procedures and model fit statistics were used to identify optimal representation of latent subgroups based on the observed data. The fit statistics for choosing the appropriate latent class model were lowest Bayesian Information Criterion (BIC), significant parametric bootstrapped-likelihood ratio test, Vuong-Lo-Mendell-Rubin likelihood ratio test, adjusted Lo-Mendell-Rubin likelihood ratio test, and an entropy greater than 0.8 (Bakk et al., 2013). The percentage of individuals in each class reporting none, mild, moderate, or severe symptom scores for each ESAS item was compared to that of the overall sample. A difference of 5% or greater indicated posterior probability of membership in a latent subgroup based on symptom severity. The final latent subgroup outcome variable in the analysis was representative of distinct subgroups for which qualitative labels were assigned according to the degree of symptom burden (e.g., low, moderate, high).
Bivariate and multivariate analyses were conducted in Stata/MP 15.1. The means and standard deviations of the main independent variables (e.g., functional status, QOL) were summarized by latent subgroups. Bivariate analyses comparing the latent subgroups across demographic and clinical covariates were conducted using Pearson’s chi-square test and Fisher’s exact test, where appropriate. The final multivariate models were performed using multinomial logistic regression. The outcome was the three-category latent subgroup variable, and latent subgroup membership was predicted in unadjusted and adjusted models by overall MQOLQ and AKPS scores. Covariates were chosen for the final model based on the extant literature and the aforementioned conceptual model by Murali et al. (2020). The final multinomial logistic regression model examined membership in a latent subgroup after controlling for functional status, QOL, age, sex, race/ethnicity, education, insurance status, polypharmacy, hospice at baseline, and the presence of MCC.
Results
The final sample consisted of 287 seriously ill individuals who were enrolled in the original trial with available symptom burden data. Baseline characteristics are provided in Table 1. Seventy-six percent of participants in the sample were 65 and older. There were more men than women in the sample (56.5% vs. 43.5%, respectively). The racial and ethnic makeup of the sample mainly included non-Hispanic White individuals (80%). Those who went to college or completed graduate school made up 36% of the sample, and 54% completed high school. The majority of the sample received Medicare or Medicaid as a primary form of insurance (78.1%). At baseline, 30% of the sample was enrolled in hospice. 41% had an AKPS score of less than 50, indicating higher functional impairment. Two hundred thirteen patients (74.2%) in the sample met criteria for MCC, and 180 (62.7%) patients had a cancer diagnosis. One hundred forty (65.7%) participants had a diagnosis of cancer and met criteria for MCC. The univariate results of baseline symptom burden using symptom severity cutoffs and total ESAS score quartile categories are reported in Table 1. The average overall MQOLQ score was 6.15 ± 2.6. Based on quartile categories of the overall MQOLQ, 21.9% of participants reported an overall MQOLQ score between 0 and 5, indicating lower QOL.
Table 1.
Baseline Sociodemographic, Symptoms, and Clinical Characteristics
| Total | 287 |
|---|---|
| Age | M = 70.2 (SD = 9.9) |
| <65 | M = 57.7 (SD = 4.9) |
| 65–74 | M = 69.3 (SD = 2.8) |
| ≥75 | M = 81.4 (SD = 4.6) |
| Age Distribution | N (%) |
| <65 | 69 (24.0) |
| 65–75 | 112 (39.0) |
| ≥75 | 106 (36.9) |
| Sex | |
| Male | 162 (56.5) |
| Female | 125 (43.5) |
| Race/Ethnicity | |
| Non-Hispanic White | 229 (79.8) |
| Non-Hispanic Black | 44 (15.3) |
| Hispanic | 10 (3.5) |
| Other – Asian, Pacific Islander, Multiple | 4 (1.4) |
| Education | |
| Grad School | 40 (14.0) |
| College | 63 (22.0) |
| HS Grad | 155 (54.2) |
| Less than HS | 28 (9.8) |
| Insurance | |
| Medicare/Medicaid | 224 (78.1) |
| Private/Other | 61 (21.3) |
| Uninsured | 2 (0.7) |
| Polypharmacy (# meds) | |
| Minor (<5) | 14 (4.9) |
| Major (5–10) | 95 (33.1) |
| Excessive (11–15) | 104 (36.2) |
| Very Excessive (>15) | 74 (25.8) |
| Hospice at Baseline | |
| No | 199 (69.3) |
| Yes | 86 (30.0) |
| Functional Statusa | |
| Higher | 170 (59.2) |
| Lower | 117 (40.8) |
| MCCb | |
| No | 74 (25.8) |
| Yes | 213 (74.2) |
| Cancer Diagnosis | |
| No | 107 (37.3) |
| Yes | 180 (62.7) |
| MCC with Cancer | |
| No | 73 (34.3) |
| Yes | 140 (65.7) |
| Symptoms | |
| Pain | |
| None | 120 (41.8) |
| Mild | 62 (21.6) |
| Moderate | 67 (23.3) |
| Severe | 38 (13.2) |
| Fatigue | |
| None | 46 (16.0) |
| Mild | 60 (20.9) |
| Moderate | 94 (32.8) |
| Severe | 87 (30.3) |
| Nausea | |
| None | 222 (77.4) |
| Mild | 30 (10.5) |
| Moderate | 19 (6.6) |
| Severe | 16 (5.6) |
| Depression | |
| None | 169 (58.9) |
| Mild | 46 (16.0) |
| Moderate | 38 (13.2) |
| Severe | 34 (11.9) |
| Anxiousness | |
| None | 150 (52.3) |
| Mild | 43 (15.0) |
| Moderate | 59 (20.6) |
| Severe | 34 (11.9) |
| Drowsiness | |
| None | 94 (32.8) |
| Mild | 45 (15.7) |
| Moderate | 81 (28.2) |
| Severe | 67 (23.3) |
| Appetite | |
| Best | 80 (27.9) |
| Good | 58 (20.2) |
| Average | 76 (26.5) |
| Worst | 72 (25.1) |
| Well-Being | |
| Best | 60 (20.9) |
| Good | 58 (20.2) |
| Average | 100 (34.8) |
| Worst | 62 (21.6) |
| Shortness of Breath | |
| None | 134 (46.7) |
| Mild | 47 (16.4) |
| Moderate | 50 (17.4) |
| Severe | 55 (19.2) |
| Total ESAS Scorec | M = 27.2 (SD = 16.2) |
| Total ESAS Score Quartiles | N (%) |
| >38 (Quartile 4) | 65 (22.7) |
| >26 – ≤38 (Quartile 3) | 71 (24.7) |
| >16 – ≤26 (Quartile 2) | 74 (25.8) |
| ≥1 – ≤16 (Quartile 1) | 66 (23) |
| 0 (None) | 11 (3.8) |
| Overall MQOLQ Scored | M = 6.15 (SD = 2.6) |
| Overall MQOLQ Quartiles | N (%) |
| MQOLQ ≥8–10 (Quartile 4) | 104 (36.2) |
| MQOLQ 6–8 (Quartile 3) | 56 (19.5) |
| MQOLQ 5–6 (Quartile 2) | 58 (20.2) |
| MQOLQ 0–5 (Quartile 1) | 63 (21.9) |
Note: M = Mean; SD = Standard Deviation
Binary variable of the Australia-modified Karnofsky Performance Status (AKPS < 50 = Lower Functional Status; AKPS > 60 = Higher Functional Status)
MCC (no/yes) – Represents individuals with two or more conditions, are receiving regular care, and have limitations in the activities of daily living (all participants in this sample have functional status limitations).
Individual ESAS item clinical cut points (Hui & Bruera, 2017)
McGill Quality of Life Questionnaire (MQOLQ): For the overall score, responses are based on a 0 to 10 scale (0 equating ‘very bad’ QOL and 10 equating ‘excellent’ QOL). There are no clinical cut points described in the literature; therefore, quartile categories were created for analysis. Question is as follows: “Considering all parts of my life – physical, emotional, social, spiritual, and financial – over the past 2 days the quality of my life has been.”
Fit statistics for the six latent class models are provided in Table 2. Fit statistics indicated a three-class was optimal for the LCA. The descriptive labels assigned to each of the latent subgroups were based on the severity of symptom burden associated with each subgroup as interpreted by posterior probability of an individual belonging to a specific subgroup based on symptom burden. The low symptom burden class (n = 99) had participants primarily reporting symptom scores consistent with none or mild severity. The moderate symptom burden class (n = 127) was made up of a greater proportion of individuals who reported mild pain, mild–moderate fatigue, mild depression, mild–moderate anxiousness, above average or average appetite, average well-being, and mild shortness of breath. Participants in the low or moderate symptom burden classes did not have nausea. The high symptom burden class (n = 61) consisted of those with moderate–severe pain; severe fatigue; mild, moderate, and severe nausea; moderate–severe depression; moderate–severe anxiousness; severe drowsiness; worst appetite; average or worst well-being; and moderate–severe shortness of breath.
Table 2.
Fit Statistics for the Latent Class Analysis Models
| Number of Classes | Akaike Information Criterion | Bayesian Information Criterion | Parametric Bootstrapped Likelihood Ratio Test p-value | Vuong-Lo-Mendell-Rubin Likelihood Ratio Test p-value | Lo-Mendell-Rubin Likelihood Ratio Test p-value | Entropy |
|---|---|---|---|---|---|---|
| 1 | 7305.1 | 7407.6 | - | - | - | - |
| 2 | 6826.4 | 7035.0 | < .0001 | .04 | .04 | .83 |
| 3 | 6687.3 | 7002.0 | < .0001 | .14 | .14 | .83 |
| 4 | 6636.6 | 7057.4 | < .0001 | .76 | .76 | .88 |
| 5 | 6604.1 | 7131.1 | < .0001 | .79 | .79 | .86 |
| 6 | 6613.3 | 7246.4 | 1 | .84 | .84 | .89 |
Bivariate results of demographic and clinical characteristics by latent subgroups are reported in Table 3. There were significant associations between latent subgroup membership and age, polypharmacy, functional status, MCC, and overall QOL. The high symptom burden subgroup had a greater number of individuals reporting overall quality of scores in Quartile 1, or scores between 0 and 5, indicating low QOL. In contrast to the low and moderate symptom burden subgroups, the high symptom burden subgroup had a greater percentage of individuals with lower functional status.
Table 3.
Sociodemographic, Symptom, and Clinical Characteristics by Latent Subgroups
| Subgroup 1: Low Symptom Burden | Subgroup 2: Moderate Symptom Burden | Subgroup 3: High Symptom Burden | ||
|---|---|---|---|---|
| N (%) | N (%) | N (%) | p-value† | |
| Total | 99 (34.5) | 127 (44.25) | 61 (21.25) | - |
| Age | .03 | |||
| <65 | 14 (14.1) | 36 (28.4) | 19 (31.2) | |
| 65–74 | 38 (38.4) | 50 (39.4) | 24 (39.3) | |
| >=75 | 47 (47.5) | 41 (32.3) | 18 (29.5) | |
| Sex | .92 | |||
| Male | 57 (57.6) | 70 (55.1) | 35 (57.4) | |
| Female | 42 (42.4) | 57 (44.9) | 26 (42.6) | |
| Race/Ethnicity | .37 | |||
| Non-Hispanic White | 74 (74.8) | 107 (84.3) | 48 (78.7) | |
| Non-Hispanic Black | 20 (20.2) | 13 (10.2) | 11 (18.0) | |
| Hispanic | 3 (3.0) | 6 (4.7) | 1 (1.6) | |
| Other – Asian, Pacific Islander, Multiple | 2 (2.0) | 1 (0.8) | 1 (1.6) | |
| Education | .12 | |||
| Grad School | 10 (10.1) | 24 (19.1) | 6 (9.8) | |
| College | 21 (21.2) | 29 (23.0) | 13 (21.3) | |
| HS Grad | 55 (55.6) | 67 (53.2) | 33 (54.1) | |
| Less than HS | 13 (13.1) | 6 (4.8) | 9 (14.8) | |
| Insurance | .85 | |||
| Medicare/Medicaid | 79 (79.8) | 96 (75.5) | 49 (80.3) | |
| Private/Other | 19 (19.2) | 30 (23.6) | 12 (19.7) | |
| Uninsured | 1 (1.0) | 1 (0.8) | 0 | |
| Polypharmacy (# meds) | .03 | |||
| Minor (<5) | 8 (8.1) | 5 (3.9) | 1 (1.6) | |
| Major (5 to 10) | 35 (35.4) | 44 (34.7) | 16 (26.2) | |
| Excessive (11 to 15) | 36 (36.4) | 50 (39.4) | 18 (29.5) | |
| Very Excessive (>15) | 20 (20.2) | 28 (22.1) | 26 (42.6) | |
| MCCa | .03 | |||
| No | 33 (33.3) | 32 (25.2) | 9 (14.8) | |
| Yes | 66 (66.7) | 95 (74.8) | 52 (85.3) | |
| Cancer Diagnosis | .055 | |||
| No | 46 (46.5) | 43 (33.9) | 18 (29.5) | |
| Yes | 53 (53.5) | 84 (66.1) | 43 (70.5) | |
| MCC with Cancer | .56 | |||
| No | 26 (39.4) | 31 (32.6) | 16 (30.8) | |
| Yes | 40 (60.6) | 64 (67.4) | 36 (69.2) | |
| Hospice at Baseline | .38 | |||
| No | 66 (68.0) | 86 (67.7) | 47 (77.1) | |
| Yes | 31 (31.96) | 31 (32.3) | 14 (22.95) | |
| AKPSa | M = 58.4 (SD = 11.7) | M = 57.3 (SD = 12.4) | M = 52.95 (SD = 12.3) | |
| Functional Status | .02 | |||
| Higher | 66 (66.7) | 77 (60.6) | 27 (44.3) | |
| Lower | 33 (33.3) | 50 (39.4) | 34 (55.7) | |
| Total ESAS Score | 11.6 ± 7.4 | 28.3 ± 6.5 | 50.2 ± 11.1 | |
| Pain | <.0001 | |||
| No | 65 (65.7) | 43 (33.9) | 12 (19.7) | |
| Yes | 34 (34.3) | 84 (66.1) | 49 (80.3) | |
| Fatigue | <.0001† | |||
| No | 42 (42.4) | 4 (3.2) | 0 (0) | |
| Yes | 57 (57.6) | 123 (96.9) | 61 (100) | |
| Nausea | <.0001 | |||
| No | 98 (98.9) | 99 (77.9) | 25 (40.9) | |
| Yes | 1 (1.01) | 28 (22.1) | 36 (59.0) | |
| Depression | <.0001 | |||
| No | 90 (90.9) | 67 (52.8) | 12 (19.7) | |
| Yes | 9 (9.1) | 60 (47.2) | 49 (80.3) | |
| Anxiousness | <.0001 | |||
| No | 89 (89.9) | 56 (44.4) | 5 (8.2) | |
| Yes | 10 (10.1) | 70 (55.6) | 56 (91.8) | |
| Drowsiness | <.0001 | |||
| No | 69 (69.7) | 19 (14.9) | 6 (9.8) | |
| Yes | 30 (30.3) | 108 (85.0) | 55 (90.2) | |
| Decreased Appetite | <.0001 | |||
| No | 56 (56.6) | 19 (15.1) | 5 (8.2) | |
| Yes | 43 (43.3) | 107 (84.9) | 56 (91.8) | |
| Compromised Well-Being | <.0001 | |||
| No | 52 (54.2) | 3 (2.4) | 5 (8.3) | |
| Yes | 44 (45.8) | 121 (97.6) | 55 (91.7) | |
| Shortness of Breath | <.0001 | |||
| No | 73 (73.7) | 46 (36.5) | 15 (24.6) | |
| Yes | 26 (26.3) | 80 (63.5) | 46 (75.4) | |
| Overall MQOLQc | M 7.37 (SD = 2.35) | M = 5.89 (SD = 4.68) | M = 4.75 (SD = 8.990 | |
| Quartile Categories of Overall MQOLQ | <.001 | |||
| ≥8–10 Quartile 4 (ref) | 57 (58.8) | 33 (26.8) | 14 (22.95) | |
| 6–8 Quartile 3 | 15 (15.5) | 34 (27.6) | 7 (11.5) | |
| 5–6 Quartile 2 | 17 (17.5) | 28 (22.8) | 27 (44.3) | |
| 0–5 Quartile 1 | 8 (8.3) | 28 (22.8) | 27 (44.3) |
Note. M = mean, SD = standard deviation, CCI = Charlson Comorbidity Index, ESAS = Edmonton, MCC = multiple chronic conditions
Pearson’s chi-squared, Fisher’s Exact Test.
MCC (no/yes) –Represents individuals with two or more conditions, are receiving regular care, and have limitations in the activities of daily living (all participants in sample have functional status limitations).
Australia-modified Karnofsky Performance Status (AKPS <50 = Lower Functional Status; AKPS >60 = Higher Functional Status)
McGill Quality of Life Questionnaire (MQOLQ): For the overall score, responses are based on a 0 to 10 scale (0 equating ‘very bad’ QOL and 10 equating ‘excellent’ QOL). There are no clinical cutoffs described in the literature; therefore, quartile categories were created for analysis. Question is as follows: “Considering all parts of my life – physical, emotional, social, spiritual, and financial – over the past 2 days the quality of my life has been.”
The results of the unadjusted and adjusted multinomial logistic regression models are provided in Table 4. In the unadjusted model, individuals with lower functional status had 2.52 greater odds of high symptom burden latent subgroup membership compared to the low symptom burden subgroup (p = .006). However, the relationship between functional status and membership in the high symptom burden subgroup was no longer significant in the adjusted model after controlling for demographic and clinical variables. In the unadjusted model, individuals with an overall MQOLQ score of 8 or lower were significantly more likely to be in the moderate symptom burden subgroup. Those with an overall MQOLQ score of 6 or lower, indicating lower QOL, were substantially more likely to have high symptom burden latent subgroup membership compared to the low symptom burden subgroup. After adjusting for functional status, age, sex, race/ethnicity, education, insurance, polypharmacy, hospice at baseline, and MCC, the relationships between quartile categories of overall QOL and membership in the moderate symptom burden subgroup remained significant.
Table 4.
Multinomial Logistic Regression Models of Predictors of Membership in a Latent Subgroup Based on overall Quality of Life and Baseline Functional Status
| Model 1 – Unadjusted | Model 2 – Adjusted with Covariates | |||
|---|---|---|---|---|
| Moderate vs Low Symptom Burden | High vs Low Symptom Burden | Moderate vs Low Symptom Burden | High vs Low Symptom Burden | |
| OR [95% CI] | OR [95% CI] | aOR [95% CI] | aOR [95% CI] | |
| Functional Statusa | ||||
| Higher | 1.00 | 1.00 | 1.00 | 1.00 |
| Lower | 1.30 [0.75,2.25] | 2.52 [1.31,4.86]** | 1.07 [0.52,2.20] | 1.96 [0.83,4.60] |
| Quartile Categories of Overall MQOLQb | ||||
| ≥8–10 Quartile 4 (ref) | 1.00 | 1.00 | 1.00 | 1.00 |
| 6–8 Quartile 3 | 3.91 [1.86,8.25]*** | 1.90 [0.65,5.55] | 4.33 [1.85–10.1]** | 1.65 [0.53,5.15] |
| 5–6 Quartile 2 | 2.84 [1.36,5.97]** | 3.11 [1.23,7.90]* | 3.22 [1.38–7.51]** | 3.46 [1.16,10.3]* |
| 0–5 Quartile 1 | 6.04 [2.47,14.8]*** | 13.7 [5.14,36.7]*** | 8.07 [2.98–21.9]*** | 13.5 [4.45,41.1]*** |
| Age | ||||
| <65 | 1.00 | 1.00 | ||
| 65–74 | 0.36 [0.16,0.84]* | 0.34 [0.12,0.98]* | ||
| >=75 | 0.21 [0.08,0.52]** | 0.18 [0.06,0.57]** | ||
| Sex | ||||
| Male | 1.00 | 1.00 | ||
| Female | 1.68 [0.87,.25] | 1.24 [0.55,2.82] | ||
| Race/Ethnicity | ||||
| Non-Hispanic White | 1.00 | 1.00 | ||
| Non-Hispanic Black | 0.67 [0.29,1.56] | 1.12 [0.38,3.32] | ||
| Hispanic | 2.53 [0.45,14.2] | 0.63 [0.05,7.99] | ||
| Other – Asian, Pacific Islander, Multiple | 0.19 [0.01,3.76] | 0.48 [0.02,0.39] | ||
| Education | ||||
| Grad School | 1.00 | 1.00 | ||
| College | 0.39 [0.12,1.26] | 0.96 [0.23,3.91] | ||
| HS Grad | 0.31 [0.11,0.88]* | 0.88 [0.24,3.21] | ||
| Less than HS | 0.09 [0.02,0.36]** | 0.55 [0.11,2.75] | ||
| Insurance | ||||
| Medicare/Medicaid | 1.00 | 1.00 | ||
| Private/Other | 0.81 [0.37,1.78] | 0.58 [0.20,1.69] | ||
| Uninsured | 0.35 [0.04,3.41] | - | ||
| Polypharmacy | ||||
| Minor (<5) | 1.00 | 1.00 | ||
| Major (5–10) | 1.40 [0.43,4.56] | 2.63 [0.30,23.3] | ||
| Excessive (11–15) | 0.97 [0.30,3.11] | 2.15 [0.24,19.5] | ||
| Very Excessive (>15) | 1.04 [0.28,3.91] | 5.01 [0.53,47.8] | ||
| Hospice at Baseline | ||||
| No | 1.00 | 1.00 | ||
| Yes | 1.50 [0.71,3.15] | 0.63 [0.24,1.71] | ||
| MCCc | ||||
| No | 1.00 | 1.00 | ||
| Yes | 1.67 [0.80,3.46] | 3.26 [1.27,8.33]* | ||
Note. Latent Subgroup 1 (Low Symptom Burden) is the reference category
p-value < .05
p-value < .01
p-value < .001
Binary variable representing the Australia-modified Karnofsky Performance Status (AKPS <50 = Lower Functional Status; AKPS >60 = Higher Functional Status)
McGill Quality of Life Questionnaire (MQOLQ) Part A: For the overall score, responses are based on a 0 to 10 scale (0 equating ‘very bad’ QOL and 10 equating ‘excellent’ QOL). There are no clinical cutoffs described in the literature; therefore, quartile categories based on sample average were created for analysis. Question is as follows: “Considering all parts of my life – physical, emotional, social, spiritual, and financial – over the past 2 days the quality of my life has been”
MCC – Individuals with two or more conditions, are receiving regular care, and have limitations in the activities of daily living (all participants in this sample have functional status limitations).
In the adjusted model, those with an overall MQOLQ score between 0 and 5, representing lower overall QOL, were 13.5 times more likely to be in the high symptom burden subgroup (p < .001) and 8.1 times more likely to be in the moderate symptom burden subgroup compared to the low symptom burden subgroup. There was a significant relationship in the adjusted model based on age. Individuals who were 65 to 75 years of age had 0.36 and 0.34 lower odds of being in the moderate or high symptom burden subgroups compared to the low symptom burden subgroup, respectively. Those over the age of 75 were less likely to have moderate or high symptom burden subgroup membership compared to the low symptom burden subgroup (p = .001, p = .004, respectively). The adjusted model also revealed that individuals who completed high school (aOR = 0.31) or less than high school education (aOR = 0.09) had lower likelihood of being in the moderate symptom burden group compared to the low symptom burden group. Finally, MCC’s presence increased the odds of being in the high symptom burden subgroup by 3.26 (aOR) compared to the low symptom burden subgroup.
Discussion
Among the seriously ill adults receiving palliative care at the end of life, lower overall QOL and MCC were associated with membership in the moderate and high symptom burden subgroups. While this finding was consistent with prior literature, our study highlighted the significant relationship between the presence of MCC and membership in the high symptom burden latent subgroup. Among individuals with MCC, cancer prevalence was also high, which can further affect symptom burden and must be accounted for in palliative care strategies. To our knowledge, this study uniquely operationalizes clinically significant cut-points reported by Hui and Bruera (2017) to categorize severity of symptoms measured in the ESAS as latent predictor variables in the LCA modeling. Using symptom severity to conduct the LCA ensured that the symptom burden subgroups that emerged were reliable representations of the full spectrum of severity associated with symptom burden in this seriously ill cohort.
Interestingly, older age was less likely associated with moderate or high symptom burden subgroup membership. In prior studies, older adults have been shown to have higher levels of symptom burden (Amjad et al., 2019; Pandya et al., 2019; Patel et al., 2019; Portz et al., 2017). Research suggests that older adults may be less likely to perceive symptoms as burdensome or may have difficulty reporting symptoms to caregivers and health care providers (Amjad et al., 2019). Furthermore, older adults may experience difficulty expressing their symptoms due to stigma or fear of burdening their caregivers (Amjad et al., 2019). Studies have shown that older adults may perceive their symptoms and functional status limitations as negative signs of aging, resulting in underreporting or minimization symptom burden (Han, 2018; Tkatch et al., 2017). Nevertheless, developing strategies to identify older adults who are at risk of high symptom burden are necessary to tailor care strategies to meet the needs of symptomatic individuals as individuals age with serious illness. More qualitative or mixed studies are needed to better understand the lived experiences of symptom burden among older adults at the end of life.
Individuals with high school education or less had lower odds of being in the moderate or high symptom burden latent subgroups. Prior studies have not examined the effect of education on the reporting or perceptions of symptom burden in seriously ill adults. It is possible that these findings were obtained because we did not account for all the confounding in the study or by chance. These results were seen after adjusting for other demographic and clinical variables in the final model. While previous studies have not explored the relationship between educational status and symptom burden among seriously ill adults, there is a possibility that the link between educational level and health literacy explains this finding (Chesser et al., 2016).
A prior study showed that MCC affects functional status and is associated with high symptom burden but did not include clinically significant cut-points for symptom severity (Portz et al., 2017). In our study, polypharmacy and hospice status were other clinical variables not associated with latent subgroup membership after adjusting for covariates. In a prior analysis of the same data, polypharmacy was associated with higher symptom burden and lower QOL; however, there was no specific focus on the inclusion of MCC (Schenker et al., 2019). This study highlighted greater odds of high symptom burden subgroup membership among those with major, excessive, or very excessive polypharmacy. While the results are consistent with prior studies, they were not statistically significant. It is possible that the inclusion of MCC was moderating the relationships as having MCC was consistently associated with membership in the high symptom burden subgroup. This finding is important to note because prior studies had not evaluated the effect of MCC on the relationships between symptom burden and QOL. The presence of MCC being associated with membership in the high symptom burden subgroup may be attributable to the level of clinical complexity associated with having MCC. Still, this idea is not well understood, and more empirical data is needed to examine this phenomenon (Grembowski et al., 2014). The inclusion of MCC as a covariate is unique in that having MCC was considered as potentially contributing to the severity of symptom burden, which has not been studied before. We believe that the use of latent class analysis to identify subgroups of seriously ill adults with MCC based on their symptom burden was an innovative method to better understand this complex population. The inclusion of MCC as a variable associated with latent subgroup membership based on symptom burden contributes a new concept to the serious illness and palliative care evidence base.
This study sheds light on an essential facet of the serious illness experience at the end of life in those receiving palliative care. A better understanding of symptom burden among seriously ill adults can help clinicians provide patient-centered, goal-concordant palliative care that adequately addresses symptoms to improve QOL (Sanders et al., 2018). For example, individuals who have high symptom burden would benefit from tailored palliative care interventions. Such interventions should be mindful of the clinical complexity associated with concurrent serious illness and presence of MCC (Murali et al., 2020). Furthermore, nurses and other clinicians must be adequately trained to identify the needs of seriously ill adults at the end of life, especially in those with moderate to high symptom burden. Training should include evidence-based practice guidelines to manage individuals with MCC, ensuring that its cumulative effect on the serious illness experience is adequately managed through effective communication and care coordination (Murali et al., 2020). In addition, we must train nurses and other clinicians in primary palliative care skills to identify and manage symptom burden among seriously ill adults with MCC.
Limitations
There are limitations of this study. Because the study was a cross-sectional analysis of a specific seriously ill population receiving palliative care at the end of life with a prognosis of 6 months or less, findings must be interpreted cautiously as they may not necessarily be generalizable to all seriously ill adults. First, original limitations of the trial must be accounted for, including the fact that individuals with active cardiovascular disease requiring a statin or recent cardiovascular event were excluded from the study, which is a potentially symptomatic population. Although our analysis controlled for covariates such as age, sex, race/ethnicity, education, insurance, polypharmacy, hospice at baseline, and presence of MCC, it is possible that the analysis did not account for all possible confounding variables that would explain membership in a latent subgroup based on symptom burden, such as health care utilization or type, duration, and frequency of palliative care, as the available data limited us. It is also possible that the ESAS did not fully encapsulate the symptom profile of the participants in the sample.
Conclusions
This study quantitatively explored factors associated with latent subgroup membership based on severity of symptom burden in a seriously ill population receiving palliative care at the end of life. Our findings add to prior results that lower QOL is associated with moderate or high symptom burden. Care strategies should employ a comprehensive approach to identify those at risk of high symptom burden at the end of life to provide optimal care, especially among older adults. We found that individuals with MCC were more likely to be in the high symptom burden subgroup, which sets the groundwork for further studies exploring the effect of MCC in serious illness and high symptom burden. Ultimately, the findings of this study provide foundational evidence for studies that aim to describe seriously ill adults with MCC at the end of life to create high-quality, patient-centered interdisciplinary care strategies to improve outcomes and QOL in this complex patient population.
Acknowledgements:
At the time this research was completed, Komal P. Murali, PhD, RN, ACNP-BC was a doctoral student supported by the NYU Clinical and Translational Science Institute (NCATS/NIH: TL1 TR001447; UL1 TR001445) and is now receiving postdoctoral fellowship support from the Columbia University School of Nursing Comparative and Cost-Effectiveness Research Training for Nurse Scientists (CER2) (T32NR014205). Amy S. Kelley, MD, MSHS, FAAHPM, received funding support from the National Institute on Aging (K24-AG062785). The authors would like to extend thanks to the Palliative Care Research Cooperative for access to the original trial data. The authors would like to acknowledge the NYU Clinical and Translational Science Institute, NYU Rory Meyers College of Nursing P20 Center for Precision Health in Diverse Populations (NINR/NIH: 1P20NR018075-01), Columbia University School of Nursing P20 Center for Improving Palliative Care for Vulnerable Adults with Multiple Chronic Conditions (NINR/NIH: P20NR018072-01) and the Jonas Nurse Leader Scholar Program.
Footnotes
Conflict of Interest: The authors have no conflicts of interest to report.
Ethical Conduct of Research: The authors of this study maintained highest standards of ethical conduct of research. This study was exempt by the Institutional Review Board at the respective institution as it contained solely deidentified clinical trial data.
Clinical Trial Registration: This was a secondary data analysis of a clinical trial. The trial was registered on clinicaltrials.gov at https://clinicaltrials.gov/ct2/show/NCT01415934. The start date of the study was June 3, 2011 (NCT01415934) (NINR/NIH: UC4-NR012584).
Contributor Information
Komal P. Murali, Columbia University School of Nursing, New York, NY, USA.
Allison Vorderstrasse, University of Massachusetts Amherst, Amherst, Massachusetts, USA.
Amy S. Kelley, Department of Geriatrics and Palliative Medicine, Icahn School of Medicine at Mount Sinai, New York, NY, USA.
Abraham A. Brody, Hartford Institute for Geriatric Nursing, NYU Rory Meyers College of Nursing, New York, NY, USA.
References
- Abernethy AP, Shelby-James T, Fazekas BS, Woods D, & Currow DC (2005). The Australia-modified Karnofsky Performance Status (AKPS) scale: A revised scale for contemporary palliative care clinical practice [ISRCTN81117481]. BMC Palliative Care, 4, 7. 10.1186/1472-684X-4-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Amjad H, Snyder SH, Wolff JL, Oh E, & Samus QM (2019). Before hospice: Symptom burden, dementia, and social participation in the last year of life. Journal of Palliative Medicine, 22, 1106–1114. 10.1089/jpm.2018.0479 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Astrup GL, Hofsø K, Bjordal K, Guren MG, Vistad I, Cooper B, Miaskowski C, & Rustøen T (2017). Patient factors and quality of life outcomes differ among four subgroups of oncology patients based on symptom occurrence. Acta Oncologica, 56, 462–470. 10.1080/0284186X.2016.1273546 [DOI] [PubMed] [Google Scholar]
- Bakk Z, Tekle FB, & Vermunt JK (2013). Estimating the association between latent subgroup membership and external variables using bias-adjusted three-step approaches. Sociological Methodology, 43, 272–311. 10.1177/0081175012470644 [DOI] [Google Scholar]
- Bernacki RE, & Block SD (2014). Communication about serious illness care goals: A review and synthesis of best practices. JAMA Internal Medicine, 174, 1994–2003. 10.1001/jamainternmed.2014.5271 [DOI] [PubMed] [Google Scholar]
- Bruera E, Kuehn N, Miller MJ, Selmser P, & Macmillan K (1991). The Edmonton Symptom Assessment System (ESAS): A simple method for the assessment of palliative care patients. Journal of Palliative Care, 7, 6–9. 10.1177/082585979100700202 [DOI] [PubMed] [Google Scholar]
- Charlson ME, Pompei P, Ales KL, & MacKenzie CR (1987). A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation. Journal of Chronic Diseases, 40, 373–383. 10.1016/0021-9681(87)90171-8 [DOI] [PubMed] [Google Scholar]
- Chesser AK, Keene Woods N, Smothers K, & Rogers N (2016). Health literacy and older adults: A systematic review. Gerontology and Geriatric Medicine, 2, 2333721416630492. 10.1177/2333721416630492 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Cohen SR, Mount BM, Bruera E, Provost M, Rowe J, & Tong K (1997). Validity of the McGill Quality of Life Questionnaire in the palliative care setting: A multi-centre Canadian study demonstrating the importance of the existential domain. Palliative Medicine, 11, 3–20. 10.1177/026921639701100102 [DOI] [PubMed] [Google Scholar]
- Ganesh V, Zhang L, Chan S, Wan BA, Drost L, Tsao M, Danjoux C, Barnes E, McDonald R, Rowbottom L, Zaki P, Chow R, Hwang MK, DeAngelis C, Lao N, & Chow E (2017). An update in symptom clusters using the Edmonton Symptom Assessment System in a palliative radiotherapy clinic. Supportive Care in Cancer, 25, 3321–3327. 10.1007/s00520-017-3749-x [DOI] [PubMed] [Google Scholar]
- Grembowski D, Schaefer J, Johnson KE, Fischer H, Moore SL, Tai-Seale M, Ricciardi R, Fraser JR, Miller D, & LeRoy L (2014). A conceptual model of the role of complexity in the care of patients with multiple chronic conditions. Medical Care, 52, S7–S14. 10.1097/MLR.0000000000000045 [DOI] [PubMed] [Google Scholar]
- Hall SF (2006). A user’s guide to selecting a comorbidity index for clinical research. Journal of Clinical Epidemiology, 59, 849–855. 10.1016/j.jclinepi.2005.11.013 [DOI] [PubMed] [Google Scholar]
- Hall WH, Ramachandran R, Narayan S, Jani AB, & Vijayakumar S (2004). An electronic application for rapidly calculating Charlson comorbidity score. BMC Cancer, 4, 94. 10.1186/1471-2407-4-94 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Han J (2018). Chronic illnesses and depressive symptoms among older people: Functional limitations as a mediator and self-perceptions of aging as a moderator. Journal of Aging and Health, 30, 1188–1204. 10.1177/0898264317711609 [DOI] [PubMed] [Google Scholar]
- Hui D, & Bruera E (2017). The Edmonton Symptom Assessment System 25 years later: Past, present, and future developments. Journal of Pain and Symptom Management, 53, 630–643. 10.1016/j.jpainsymman.2016.10.370 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Johnstone PAS, Lee J, Zhou J-M, Ma Z, Portman D, Jim H, & Yu H-HM (2017). A modified Edmonton Symptom Assessment Scale for symptom clusters in radiation oncology patients. Cancer Medicine, 6, 2034–2041. 10.1002/cam4.1125 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kamal AH, Nipp RD, Bull J, Stinson CS, & Abernethy AP (2015). Symptom burden and performance status among community-dwelling patients with serious illness. Journal of Palliative Medicine, 18, 542–544. 10.1089/jpm.2014.0381 [DOI] [PubMed] [Google Scholar]
- Kamal AH, Taylor DH Jr., Neely B, Harker M, Bhullar P, Morris J, Bonsignore L, & Bull J (2017). One size does not fit all: Disease profiles of serious illness patients receiving specialty palliative care. Journal of Pain and Symptom Management, 54, 476–483. 10.1016/j.jpainsymman.2017.07.035 [DOI] [PubMed] [Google Scholar]
- Katz JN, Chang LC, Sangha O, Fossel AH, & Bates DW (1996). Can comorbidity be measured by questionnaire rather than medical record review? Medical Care, 34, 73–84. 10.1097/00005650-199601000-00006 [DOI] [PubMed] [Google Scholar]
- Kelley AS (2014). Defining “serious illness.” Journal of Palliative Medicine, 17, 985. 10.1089/jpm.2014.0164 [DOI] [PubMed] [Google Scholar]
- Kelley AS, & Morrison RS (2015). Palliative care for the seriously ill. New England Journal of Medicine, 373, 747–755. 10.1056/NEJMra1404684 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kutner JS, Blatchford PJ, Taylor DH Jr., Ritchie CS, Bull JH, Fairclough DL, Hanson LC, LeBlanc TW, Samsa GP, Wolf S, Aziz NM, Currow DC, Ferrell B, Wagner-Johnston N, Yousuf Zafar S, Cleary JF, Dev S, Goode PS, Kamal AH, … Abernethy AP (2015). Safety and benefit of discontinuing statin therapy in the setting of advanced, life-limiting illness: A randomized clinical trial. JAMA Internal Medicine, 175, 691–700. 10.1001/jamainternmed.2015.0289 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lee SB, Oh JH, Park JH, Choi SP, & Wee JH (2018). Differences in youngest-old, middle-old, and oldest-old patients who visit the emergency department. Clinical and Experimental Emergency Medicine, 5, 249–255. 10.15441/ceem.17.261 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Masnoon N, Shakib S, Kalisch-Ellett L, & Caughey GE (2017). What is polypharmacy? A systematic review of definitions. BMC Geriatrics, 17, 230. 10.1186/s12877-017-0621-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Miaskowski C, Dunn L, Ritchie C, Paul SM, Cooper B, Aouizerat BE, Alexander K, Skerman H, & Yates P (2015). Latent class analysis reveals distinct subgroups of patients based on symptom occurrence and demographic and clinical characteristics. Journal of Pain and Symptom Management, 50, 28–37. 10.1016/j.jpainsymman.2014.12.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Murali KP, Merriman JD, Yu G, Vorderstrasse A, Kelley A, & Brody AA (2020). An adapted conceptual model integrating palliative care in serious illness and multiple chronic conditions. American Journal of Hospice and Palliative Medicine, 37, 1086–1095. 10.1177/1049909120928353 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Pandya C, Magnuson A, Flannery M, Zittel J, Duberstein P, Loh KP, Ramsdale E, Gilmore N, Dale W, & Mohile SG (2019). Association between symptom burden and physical function in older patients with cancer. Journal of the American Geriatrics Society, 67, 998–1004. 10.1111/jgs.15864 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Papachristou N, Barnaghi P, Cooper BA, Hu X, Maguire R, Apostolidis K, Armes J, Conley YP, Hammer M, Katsaragakis S, Kober KM, Levine JD, McCann L, Patiraki E, Paul SM, Ream E, Wright F, & Miaskowski C (2018). Congruence between latent class and K-modes analyses in the identification of oncology patients with distinct symptom experiences. Journal of Pain and Symptom Management, 55, 318–333.E4. 10.1016/j.jpainsymman.2017.08.020 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Parekh AK, Goodman RA, Gordon C, Koh HK (2011). Managing multiple chronic conditions: A strategic framework for improving health outcomes and quality of life. Public Health Reports, 126, 460–471. 10.1177/003335491112600403 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Patel KV, Guralnik JM, Phelan EA, Gell NM, Wallace RB, Sullivan MD, & Turk DC (2019). Symptom burden among community-dwelling older adults in the United States. Journal of the American Geriatrics Society, 67, 223–231. 10.1111/jgs.15673 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Portz JD, Kutner JS, Blatchford PJ, & Ritchie CS (2017). High symptom burden and low functional status in the setting of multimorbidity. Journal of the American Geriatrics Society, 65, 2285–2289. 10.1111/jgs.15045 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sanders JJ, Curtis JR, & Tulsky JA (2018). Achieving goal-concordant care: A conceptual model and approach to measuring serious illness communication and its impact. Journal of Palliative Medicine, 21, S17–S27. 10.1089/jpm.2017.0459 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Schenker Y, Park SY, Jeong K, Pruskowski J, Kavalieratos D, Resick J, Abernethy A, & Kutner JS (2019). Associations between polypharmacy, symptom burden, and quality of life in patients with advanced, life-limiting illness. Journal of General Internal Medicine, 34, 559–566. 10.1007/s11606-019-04837-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Tkatch R, Musich S, MacLeod S, Kraemer S, Hawkins K, Wicker ER, & Armstrong DG (2017). A qualitative study to examine older adults’ perceptions of health: Keys to aging successfully. Geriatric Nursing, 38, 485–490. 10.1016/j.gerinurse.2017.02.009 [DOI] [PubMed] [Google Scholar]
- Yu DS, Chan HY, Leung DY, Hui E, & Sit JW (2016). Symptom clusters and quality of life among patients with advanced heart failure. Journal of Geriatric Cardiology: JGC, 13(5), 408–414. 10.11909/j.issn.1671-5411.2016.05.014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- UCLA Institute for Digital Research & Education. (2020). Latent class analysis in Mplus. https://stats.idre.ucla.edu/mplus/seminars/lca/