Pain Impacting Quality of Life in Persons with Dementia Dying in the Nursing Home by Alternative Medicare Payment Model

Abstract

Background:

Medicare alternative payment models were created to improve health care value by controlling costs and improving care quality.

Objective:

To determine if prevalence of pain affecting quality of life (QoL) differs by Medicare payment model among nursing home (NH) decedents with dementia at the end of life.

Setting/Subjects:

NH decedents in 2017/2018 in the United States with dementia who self-reported pain on a Minimum Data Set assessment in the last 30 days of life.

Measurements:

Main outcome was pain impacting QoL (i.e., affecting day-to-day activities or sleep). Multivariable logistic analysis examined the association between payment model (traditional Medicare [TM], Medicare Advantage [MA], or accountable care organizations [ACOs]) and pain impacting QoL after controlling for potential confounders.

Results:

There were 115,757 NH residents with dementia who self-reported pain in the last 30 days of life. Of those, 17.8% (n = 20,585) reported having pain the last five days from assessment, which varied by Medicare payment model (17.7% in TM, 17.5% in MA, and 19.1% in ACOs; p < 0.001). Among decedents reporting pain, 23.6% of ACO decedents reported pain affecting QoL compared to 22.1% in MA and 21.6% in TM (p = 0.09). After adjustment, decedents in ACOs compared to TM had greater predicted probability of pain affecting QoL (absolute marginal difference 0.017, 95% CI 0.00–0.035, p = 0.05), and persons in MA did not differ from persons in TM (absolute marginal difference 0.005, 95% CI −0.008 to 0.019, p = 0.41).

Conclusions and Implications:

Among dementia decedents dying with pain, pain impacted QoL in more than one in five persons. All payment models can improve pain management.

Introduction

Alzheimer's disease and related dementias (ADRD) are an increasing cause of death in the aging US population. In the United States, over five million persons are diagnosed with ADRD, a set of incurable, progressive, degenerative neurological diseases, which by 2050 will cause an estimated 1.6 million deaths per year.^1,2 The ADRD disease course presents a number of challenges, including increased likelihood that symptoms in ADRD persons may go unrecognized, including pain.^3,4 This is due to cognitive impairment, which may impact individuals' ability to report pain.^3–5 This results in inadequate pain management.^4,6–8 Pain is a sentinel symptom that increases in the last days of life.^9–11 Outcomes measuring quality of life (QoL) or quality of dying can be representative of the quality of care provided.¹²

In a study of 623 bereaved family members, 16.9% reported that their loved one had “too little care” at the end of life; lack of pain relief was often the reason behind this perception.¹³ Pain is an especially important outcome to recognize in persons with dementia due to its impact on QoL, such as poor sleep, restriction of day-to-day activities, and mood, including a higher risk of developing depression.³

The Center for Medicare and Medicaid Services (CMS) is currently implementing programs to incentivize value-based care that aims to control cost and maintain or improve the quality of health care. Value-based care is an alternative to traditional fee-for-service Medicare (TM), which incentivizes volume-based care. In 2010, the Affordable Care Act introduced a new CMS-led value-based care initiative via accountable care organizations (ACOs),¹⁴ which commenced in 2012.¹⁵ ACOs are composed of groups of health care providers who seek to provide high-quality care to Medicare beneficiaries based on a risk-reward system.¹⁶ Medicare incentives that emphasize value may impact care practices, which affect not only quality of care, but QoL of those receiving care.^17–20

Previous research has found that among persons with dementia with a nursing home (NH) stay and who were enrolled in a Medicare Advantage (MA) program were less likely to be hospitalized in the last 30 days of life and less likely to die in the hospital compared to beneficiaries in TM and ACOs.²¹ A similar study found that persons with Alzheimer's disease who were hospitalized were less likely to have experienced a preventable hospitalization based on Medicare payment model type.²⁰ Cost savings from a reduction in hospitalizations are ethically defensible only if the resident is receiving high-quality palliative care in the NH setting. The goals of this report are to better understand if pain affecting QoL near the end of life differs by Medicare payment model type among NH residents with dementia.

Methods

Data and study population

Medicare beneficiaries with a Minimum Data Set (MDS) assessment that was completed in the last 30 days of life in 2017 or 2018 were identified. The MDS is a federally mandated assessment for persons residing in Medicare-certified NHs that is conducted on admission, yearly, quarterly, and at other prescribed intervals during skilled nursing stays. Using the MDS, staff in the NH report sociodemographic characteristics, diagnoses that impact health, Activities of Daily Living (ADL) impairments, cognitive functioning, and other key clinical characteristics. The cohort was defined based on decedents having a dementia diagnosis noted on the MDS. In addition, decedents must have had a Cognitive Function Score (CFS) of two or greater, indicating at least mild cognitive impairment, and two or more ADL impairments as published in previous literature.^21,22

Age and sex were drawn from the Medicare Master Beneficiary Summary File as collected by the Social Security Administration. Basic race and ethnicity variables were improved upon by using the Research Triangle Institute race and ethnicity code. Decedents were also required to be assigned to one of the following: TM, MA, or an ACO. Persons were considered in MA if they had MA coverage during the last month before death. CMS defines ACO attribution based on a beneficiary receiving a majority of evaluation and management visits between January 1 and December 31 from ACO-affiliated health care providers for primary care services. Finally, decedents must have been able to complete a self-reported pain assessment in section J of the MDS, and must have died in NHs with five or more decedents in 2017 or 2018 to conduct a multivariate logistic model that adjusted for clustering of persons in NHs.

Self-reported pain assessment

“Resident voice” is an important stated objective of the MDS 3.0.²³ One of the ways resident voice is incorporated is via section J “Health Conditions” in the MDS. Staff are prompted to attempt to allow residents to self-report their pain in an interview-type assessment unless the resident is rarely or never understood by staff; persons with dementia can often still report their pain experience despite cognitive decline.²³ Self-reported pain assessment items include (1) pain presence in the last five days, (2) pain frequency, (3) pain effect on function (sleep and day to day activities), and (4) pain intensity, which include a choice of either a numeric rating scale or verbal descriptor scale.²⁴

For each question in the self-reported pain assessment, in addition to the provided answers, there is an option for “unable to answer.” In this study, persons were only considered to have completed a self-reported pain assessment if they were able to answer all questions on the assessment.

Measures

The main exposure was the decedent's Medicare payment model at the time of death. A measure of “any pain” was used for both staff-reported persons and self-reporting persons as a method of describing differences in Medicare payment models. In self-reporting persons, persons were considered as having pain if they answered “yes” to the question “Have you had pain or hurting at any time in the last five days?.” In persons whose pain was assessed by staff, pain was defined as observing a person having one or more of the four signs of pain in the last five days. The four signs of pain for persons unable to self-report were nonverbal sounds, vocal complaints of pain, facial expressions, and protective body movements or postures.

The primary study outcome and dependent variable was whether decedents based on self-report experienced pain affecting QoL in the last five days and whether this differed by Medicare payment model type. We restricted this analysis to persons who reported “yes” to having experienced pain in the last five days and who were able to complete a self-reported pain assessment, including answering questions documenting whether pain impacting day to day activities or sleep. This measure of quality was created by combining persons who answered “yes” to MDS Section J, question J0500a “over the past five days, has pain made it hard for you to sleep at night?” or question J0500b “over the past five days, have you limited your day-to-day activities because of pain?.”

Analytic strategy

As part of this observational study, frequencies were used to describe the sample flow, including the completion of pain assessments and pain presence by dementia versus nondementia diagnosis, pain assessment type (staff or self-report), and by Medicare payment model. Persons who only partially completed a self-reported assessment or who had a completed staff assessment for pain were excluded from the primary analysis.

Next, frequencies and means with 95% confidence intervals were used to describe the cohort of persons with dementia with pain in the last five days from assessment by Medicare payment model type using sociodemographic variables, cognitive functioning, ADL impairments, whether the person was reported as having a life expectancy of less than six months, hospice use, comorbidities, the social deprivation index (SDI) based on the beneficiary zip code,²⁵ whether the resident was bedfast, and whether the decedent's MDS assessment type indicated a skilled NH stay.

Univariate association of possible covariates with the dependent variable and whether pain impacted QoL were examined using chi-square test for nominal variables and one-way analysis of variance for interval level variables. Next, a multivariate logistic regression with NH random effects was conducted to adjust for potential confounders. This analysis was completed for those with and without a dementia diagnosis adjusting for clustering at the NH level. Covariates included age, dual Medicaid status, sex, race/ethnicity, CFS level of cognitive impairment, number of ADL impairments, whether the person was bedfast, and whether the person had hospice in the last 14 days, whether the person had a life expectancy of less than six months, indicators of painful comorbidities when available (heart failure, diabetes, chronic obstructive pulmonary disease [COPD]/asthma/chronic lung disease, cerebrovascular accident [CVA]/traumatic brain injury [TBI]/stroke), decile of decedent's level of SDI, and whether or not the decedent had a hospitalization before their NH stay.

Two multivariate logistic models with random effects at the NH level were conducted, including the primary analysis and a sensitivity analysis excluding persons with a CFS of 4 or persons who had hospice to test whether our results were affected by severity of impairment, quality of NH, or increased perceptions of pain due to hospice care. Results are presented as adjusted average marginal effect in Stata version 17.0, which reports differences in predicted probability of the outcome measure. The use of these data was approved by a data use agreement. Based on secondary data analysis and a decedent population, Brown University Institutional Review Board waived informed consent for this population.

Results

Sample flow and description

During 2017 and 2018, 416,262 persons died with an MDS assessment in the last 30 days of life in a NH with 5 or more decedents. Among those persons, 42.8% (n = 178,129) had dementia. Figure 1 characterizes the sample flow. Among persons with dementia and a completed pain assessment (either self-report or staff-report), 176,043 were identified and 65.8% (n = 115,757) self-reported their pain experience.

FIG. 1.

Sample flowchart of path to final study sample of persons with dementia who reported pain in the last five days from assessment.

Table 1 describes the final sample of 20,585 persons with dementia who reported having pain on that MDS assessment. Persons in ACOs were more often white (91.3% [95% CI 90.1–92.4] vs. 86.0% [95%CI 85.8–87.0] in traditional Medicare [TM] and 85.2% [95% CI 84.2–86.2] in MA, p < 0.05). Persons in ACOs were less likely to be black (6.7% [95% CI 5.7–7.7] vs. 9.8% [95% CI 9.3–10.3] in TM and 11.4% [95% CI 10.5–12.3] in MA, p < 0.05), less likely to have a CFS score of 4 (3.8% [95% CI 3.0–4.6] vs. 5.1% [95% CI 4.7–5.4] in TM and 5.6% [95% CI 5.0–6.3] in MA, p < 0.05), or less likely to be bedfast (12.9% [95% CI 11.6–14.3] vs. 15.8% [95% CI 15.2–16.4] in TM and 14.8% [95% CI 13.8–15.8] in MA, p < 0.05).

Table 1.

Characteristics of 2017 and 2018 Medicare Decedents with Dementia Who Reported Pain in Traditional Medicare, Medicare Advantage, and Accountable Care Organizations

Decedent characteristics	TM (N = 13,256)	MA (N = 4909)	ACO (N = 2420)
Age, mean (95% CI)	85.9 (85.8–86.1)	86.0 (85.8–86.2)	86.4 (86.1–86.7)
Female, % (95% CI)	67.5 (66.7–68.3)	66.6 (65.2–67.9)	68.0 (66.1–69.8)
Dual Medicaid status, % (95% CI)	61.6 (60.8–63.5)	61.4 (60.0–62.7)	58.9 (56.9–60.9)
Race: white, % (95% CI)	86.4 (85.8–87.0)	85.2 (84.2–86.2)	91.3 (90.1–92.4)
Race: black, % (95% CI)	9.8 (9.3–10.3)	11.4 (10.5–12.3)	6.7 (5.7–7.7)
Race: Asian, % (95% CI)	0.9 (0.7–1.1)	0.9 (0.6–1.2)	Supp.a
Race: Hispanic, % (95% CI)	1.5 (1.3–1.7)	1.6 (1.3–2.0)	0.7 (0.4–1.2)
CFS Score: 2, % (95% CI)	41.5 (40.6–42.3)	41.0 (39.6–42.4)	43.4 (41.4–45.4)
CFS Score: 3, % (95% CI)	53.5 (52.6–54.3)	53.4 (52.0–54.8)	52.9 (50.8–54.9)
CFS Score: 4, % (95% CI)	5.1 (4.7–5.4)	5.6 (5.0–6.3)	3.8 (3.0–4.6)
No. of ADLs, mean (95% CI)	5.7 (5.7–5.2)	5.7 (5.6–5.7)	5.6 (5.6–5.7)
Life expectancy >6 months, % (95% CI)	9.1 (8.6–9.6)	9.0 (8.3–9.9)	9.1 (7.9–10.3)
Hospice in the last 14 days, % (95% CI)	11.2 (10.7–11.8)	11.6 (10.8–12.6)	10.7 (9.4–12.0)
Bedfast, % (95% CI)	15.8 (15.2–16.4)	14.8 (13.8–15.8)	12.9 (11.6–14.3)
Heart failure, % (95% CI)	31.7 (30.9–32.5)	31.2 (29.9–32.5)	32.4 (30.5–32.3)
Diabetes, % (95% CI)	34.5 (33.7–35.3)	35.6 (34.2–36.9)	34.1 (32.2–36.0)
COPD/asthma/chronic lung disease	28.4 (27.7–29.2)	27.7 (26.4–29.0)	25.5 (25.5–29.1)
CVA/TBI/stroke, % (95% CI)	10.6 (10.1–11.2)	9.5 (8.7–10.3)	11.6 (10.4–13.0)
Skilled NH stay	19.8 (19.1–20.5)	20.1(19.0–21.3)	20.9 (19.3–22.6)
SDI score, mean (95% CI)	48.5 (48.1–48.8)	46.8 (46.1–47.6)	44.4 (43.4–45.5)

^a Value suppressed due to small cells.

ACO, accountable care organization; CFS, Cognitive Function Score; COPD, chronic obstructive pulmonary disease; CVA, cerebrovascular accident; MA, Medicare Advantage; NH, nursing home; SDI, social deprivation index; TBI, traumatic brain injury; TM, traditional Medicare.

Pain outcomes

Among the 20,585 persons who reported pain, 22.0% (n = 4528) reported pain affecting QoL. This varied somewhat by Medicare payment model, with ACOs having slightly more decedents (23.6%; 95% CI 22.0–25.4) with pain impacting sleep and/or daily activities compared to TM (21.6%; 95% CI 20.9–22.4) and MA (22.1%; 95% CI 21.0–23.3), but this did not reach conventional statistical significance (p = 0.09).

In univariate analysis, race, CFS score, COPD/asthma/chronic lung disease, SDI score, and CVA/TBI/stroke were all significantly associated with pain affecting QoL based on p < 0.05 and difference of >2% between payment models. In multivariate analysis, we controlled for age, race, sex, number of ADL impairments, CFS score, bedfast, hospice, life expectancy of less than six months, comorbidities (heart failure, diabetes, COPD/asthma/chronic lung disease, CVA/TBI, stroke), SDI, whether the decedent had a skilled nursing stay while assessed, and Medicare payment model type.

Compared to persons with TM, persons in an ACO had a significantly greater predicted probability of experiencing pain affecting QoL, and persons in MA did not differ from persons in TM (Table 2). Absolute differences in predicted probability of experiencing pain affecting QoL from TM were 0.005 for MA (95% CI −0.008 to 0.019, p = 0.41) and 0.017 for ACOs (95% CI −0.000 to 0.035, p = 0.05).

Table 2.

Marginal Differences of Medicare Advantage, Accountable Care Organizations, and Traditional Medicare with Pain Affecting Quality of Life for Persons with Dementia

	TM		MA		ACOs
	Univariate, % (95% CI)	Marginal difference (95% CI)	Univariate, % (95% CI)	Marginal difference (95% CI)	Univariate, % (95% CI)	Marginal difference (95% CI)
Pain affecting QoL	21.6 (20.9 to 22.4)	Ref.	22.1 (21.0 to 23.3)	0.005 (−0.008 to 0.019)	23.6 (22.0 to 25.4)	0.017a (0.000 to 0.035)

^a Significant at p = 0.05.

QoL, quality of life.

Discussion

Alternative payment models, like MA and ACOs, were created by Medicare policy aimed at incentivizing health care organizations to achieve value through reducing health costs and maintaining or improving the quality of care. MA plans are profitable when they provide care within specific rules that is less costly than the per-member per-month premium paid by Medicare.²⁶ ACOs receive bonus payments when total annual beneficiary costs are lower than CMS's annual target for spending while also meeting minimum quality measure requirements.²⁷

In previous research comparing MA to TM, MA reduced hospitalizations in the last month of life in the care of persons with dementia diagnosis,²¹ and ACOs were reported as possibly improving quality of care for persons with dementia or Alzheimer's disease.²⁰ An important question is whether this savings in health care costs are associated with improved quality of care for persons dying in a NH. For dying persons and their families, a key part of the quality care at the end of life is symptom management.^10,11

We hypothesized that MA and ACOs would better control pain affecting QoL based on savings that would allow participating organizations to have more flexibility and creativity over how they care for and improve care for their beneficiaries residing in NHs. Contrary to the expected hypothesis that cost savings would be accompanied by improved quality of care, this study found that pain impacted QoL (pain affecting sleep and/or daily activities) in more than one in five decedents with dementia across all payment models. Neither MA nor ACOs improved the quality of pain management compared to TM. Persons in ACOs were slightly more likely to report pain impacting their sleep or day-to-day activities compared to those in TM.

Currently, CMS does not endorse any MDS care quality indicators involving pain unless a person is concurrently receiving hospice. There remains an important opportunity to improve pain management in the NH setting for dying persons.

Pain is an underreported symptom at the end of life.²⁸ There is evidence that documentation and treatment of pain decreases with increased cognitive impairment.^6,28–30 There is no evidence that persons with dementia experience less pain, and the literature suggests that they may actually experience increased sensitivity to pain stimuli⁵; ability for staff to recognize pain or for these persons to adequately report their pain may be impacted by their cognitive impairments.^6,28–30 Our finding that only 17.8% of self-reporting ADRD persons had pain was about half the rate of their counterparts without dementia, confirming previous research that found that pain is underestimated for this population.^28–31

It is important to note that patient preferences for pain relief at the end of life differ in that persons may prefer to experience pain over taking medications to control pain that may have side effects³²; reporting the prevalence of pain alone may be problematic in identifying poor care quality. Therefore, we studied pain that impacted QoL, which we defined as pain impacting sleep or pain interfering with day-to-day activities in a subpopulation where a majority of persons would want treatment of that pain. Even with concerns of opiate abuse, this pain in dying persons should not preclude dementia patients from treatment of pain affecting sleep or interfering with day-to-day activities, especially near the end of life.

Limitations

This study has limitations. First, only persons with dementia in a NH who were able to self-report pain were included in this study. Results are not generalizable to all persons dying with dementia. Previous research has demonstrated that majority of persons with dementia or other cognitive impairment can report their pain, but this requires health care providers with adequate training in pain assessment of persons with dementia.^33–35 Second, the frequency of MDS assessment depends on whether a person is receiving skilled nursing services. Thus, our sample may have more persons receiving skilled services. This again limits the generalizability of our results in that many dying persons did not have an MDS assessment in that time period or that persons with a skilled nursing visit may be overrepresented.

Third, the lookback period for pain on the MDS is only five days. Therefore, the pain experience in this article of persons with intermittent pain or chronic conditions with pain flare-ups may not be representative of the majority of their experience.²⁹ It is also possible that the report of pain in the last five days does not imply uncontrolled or unmitigated pain and that pain was addressed over that time period. Fourth, we were only able to adjust for differences in painful conditions that were available to us in the MDS.

Finally, patient self-report of pain is considered the gold standard for identifying pain.²³ Despite self-reporting pain, there are concerns that persons with dementia may find the questions too abstract or complex to complete accurately; however, the MDs' self-reported pain assessment items have been previously tested among cognitively impaired persons and were determined to be reliable in a validation study.³³ Despite these limitations, our results provide national data comparing the predominant Medicare payment models paying for medical care of dying persons with dementia. All Medicare payment models should strive for excellence in the reduction of pain affecting QoL, especially among its most vulnerable beneficiaries dying in NHs.

Conclusions

Persons in the NH with a dementia diagnosis comprise a vulnerable population, with current policy focused on constraining or reducing costs while maintaining or improving care quality. Prior research has found that among persons with dementia, MA reduced hospitalizations in the last 30 days of life. With alternative Medicare payment models incentivizing value and a key goal of reducing costs, our findings highlight opportunities to improve the quality of pain management under all payment models.

Across all payment models, one in five persons in pain experienced a level of pain that affected QoL, with MA and ACOs not improving this outcome compared to TM. While there are important concerns regarding the risks and benefits of opiates for chronic pain, renewed efforts are needed to ensure that dying, frail older persons in NHs have their pain ameliorated and their QoL improved in the last months of life. With pain as a sentinel event at the end of life, the cost-savings equation of value should not outweigh the importance providing high-quality care.

Funding Information

Research in this article was supported by the National Institute on Aging of the National Institutes of Health (NIH) (2P01AG027296-11). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Author Disclosure Statement

No competing financial interests exist.