Cognitive Impairment among Medicare Home Health Patients: Comparing Available Measures

Abstract

Background:

There is growing interest in understanding home health utilization and outcomes for those with cognitive impairment (CI). Yet, approaches to measuring CI during home health vary widely across studies, with little known regarding potential implications for findings.

Methods:

Among a nationally representative sample of community-living Medicare beneficiaries receiving home health (2011–2016), we compare estimated CI prevalence using four different measures and evaluate measure-specific strengths and limitations.

Results:

CI prevalence estimates ranged from 18.4% of the sample with probable dementia from national survey data; to 27.8% with diagnosed dementia, from Medicare claims; to 26.7% with memory deficit and/or impaired decision-making and 43.9% with reduced cognitive function, from OASIS.

Conclusions:

Researchers must be deliberate in their choice of CI measure and transparent regarding its benefits and limitations. Regardless of the measure used, a sizable portion of the home health patients have CI, supporting the importance of ongoing research in this area.

INTRODUCTION:

Cognitive impairment has wide-reaching implications for older adults and their family caregivers, affecting health care utilization and spending, care needs, and caregiver tasks and burden.¹ Cognitive impairment is an encompassing term referring to reduced cognitive function (often relating to memory and decision-making) present in those with mild cognitive impairment, Alzheimer’s Disease (AD), and non-AD dementia.² There is strong interest in improving models of care for older adults with cognitive impairment and enacting policies to better support these individuals and their caregivers.^1–3 In particular, a growing field of research is aimed at understanding and improving the accessibility and quality of home health care for this population.^3–9 Home health care, a Medicare benefit which provides eligible beneficiaries with skilled nursing and therapy services delivered in their home, is one of few widely accessible and affordable options for home-based care for community-living older adults with cognitive impairment.^3,10–13

Health services research aimed at improving care for older adults with cognitive impairment is challenging in part due to the difficulties of measuring cognitive function. Small-scale studies can rely on individual cognitive testing—for example, performing a Mini-Mental State Examination for each study participant. However, for research that is highly generalizable and considers implications of broad-scale shifts in practice and policy, secondary analysis of existing, population-based data sources may be most appropriate and primary data collection may not be feasible. Yet, previous studies that compare measures of cognitive impairment from population-based data sources reveal significant variation by both data source and individual characteristics.^14–17

In the context of home health care, accurate measurement of patients’ cognitive function is hampered by underdiagnosis of dementia (particularly among minority racial and ethnic groups^16–19 who are more heavily represented in the home health patient population),²⁰ limited psychometric testing of cognition-related items included in standardized home health patient assessments,^21–23 and the dynamic and longitudinal nature of the underlying causes of cognitive impairment. Currently, the measures and data sources used to ascertain cognitive impairment among home health patients vary widely across studies, with little commentary on the potential implications for findings.

To our knowledge, there is no published work comparing and contrasting existing measures of cognitive function for home health patients. Such information could be valuable in informing researchers as they design and frame analyses and guiding readers as they engage with existing literature in this area. We compare four measures of cognitive function, from three distinct data sources, for a nationally representative cohort of community-living Medicare beneficiaries receiving home health between 2011–2016. We present the estimated prevalence of those with cognitive impairment across a range of definitions (from memory deficit to diagnosed dementia), and comment on measure-specific strengths and limitations. Findings provide foundational information regarding the tradeoffs between available measures of cognitive impairment for the Medicare home health patient population.

MATERIALS & METHODS

Data

We linked three datasets at the individual level. 1) The National Health and Aging Trends Study (NHATS), an annual, nationally-representative survey of Medicare beneficiaries aged 65 and older, using in-person interviews to gather information on participants’ health and functional status. 2) The Outcome and Assessment Information Set (OASIS), a standardized patient assessment completed during Medicare home health care. 3) The Medicare Master Beneficiary Summary File chronic conditions segment, which analyzes claims data to identify the presence of 27 chronic conditions for an individual beneficiary.

Sample

Our analytic sample included 1,214 (weighted n=5,856,333) community-dwelling older adults who participated in the NHATS and received Medicare-funded home health care within one year of NHATS survey, between 2011–2016. We exclusively examine the index, or first, home health episode in the study timeframe, thus each individual appears in the dataset only once. We exclude those enrolled in Medicare Advantage, due to lack of available claims data.

Measures

We compared four measures of cognitive impairment. 1) We measured probable dementia present in the year preceding the home health episode from NHATS, determined by a validated algorithm that uses self- or proxy-reported physician diagnosis of Alzheimer’s Disease or Related Dementias (ADRD), proxy respondent responses to a dementia screening tool, and older adult performance on brief cognitive tests of memory, executive function, and orientation.²⁴

2) We measured impaired cognitive function during the home health episode from OASIS item 1700, based on HHC clinician (registered nurse or physical therapist) ratings of patient cognitive function on a 5-item scale. Those rated a 1 or greater during the start of care assessment—ratings which range from “requires prompting” to “totally dependent”—were categorized as having impaired cognitive function on the date of assessment.²⁵ This construction has been used in previous literature.⁵ (See Appendix for additional detail on this OASIS item.)

3) We measured memory deficit and/or impaired decision-making during the home health episode from OASIS item 1740. This item asks the HHC clinician to document cognitive and behavioral symptoms “that are demonstrated at least once a week (reported or observed)”, including:

• Memory deficit, “failure to recognize familiar persons/places, inability to recall events of past 24 hours, significant memory loss so that supervision is required”

• Impaired decision-making, “failure to perform usual ADLs or IADLs, inability to appropriately stop activities, jeopardizes safety through actions”.²⁵

4) Finally, we measured diagnosed dementia, using an indicator for ADRD included in the Master Beneficiary Summary File (MBSF) chronic conditions warehouse which reflects the presence of at least one Medicare claim with an ICD-9 dementia code over a 3-year reference period.²⁶

We measured select descriptive characteristics of our sample (including sociodemographic information such as age, sex, race, and Medicaid enrollment, as well as receipt of paid caregiving) from participant self-reports in NHATS. We determine whether the individual was referred to home health from the community or following an acute care stay using OASIS.

Analysis

We calculated the prevalence of cognitive impairment in the analytic sample, as defined by each of the four measures detailed above, using NHATS weights and design variables to account for complex survey design.

RESULTS

Among our analytic sample, the average age was 79.8 years, over half (59.6%) were female, 1 in 5 (19.5%) were non-white, and 1 in 3 (29.0%) were referred to home health care from the community (Table 1).

Table 1.

Descriptive Characteristics of Community-Living Older Adults Receiving Medicare-Funded Home Health Care between 2011–2016 (unweighted n=1,214; weighted n=5,856,333)

Patient characteristics n (%) or mean ± SE
Age	79.8 ± 0.3
Sex:
Female	748 (59.6)
Male	466 (40.4)
Race:
Non-white	399 (19.5)
White	815 (80.5)
Medicaid-enrolled	249 (16.6)
Has a paid caregiver	249 (19.7)
Lives alone	438 (34.9)
Home health referral source:
Community	422 (29.0)
Post-acute	792 (71.0)

Percentages weighted to account for complex survey design. Data presented for a single cross-sectional sample of Medicare home health patients who participated in the National Health and Aging Trends Study (NHATS) and received home health care within one year of NHATS survey from 2011–2016. Data sources are linked at the individual level and include: Medicare Chronic Conditions Warehouse, National Health and Aging Trends Study, and Outcomes and Assessment Information Set.

The proportion of our analytic sample with cognitive impairment, by measure, is presented in Figure 1 along with major strengths and limitations of each measure. We observed notable variation by measure and data source. The prevalence of cognitive impairment ranged from 18.4% of the sample with probable dementia, as measured by NHATS, to 43.9% with impaired cognitive function, as measured by OASIS.

Figure 1.

Data presented for a single cross-sectional sample of Medicare home health patients who participated in the National Health and Aging Trends Study (NHATS) and received home health care within one year of NHATS survey from 2011–2016. Data sources are linked at the individual level and include: Medicare Chronic Conditions Warehouse, National Health and Aging Trends Study, and Outcomes and Assessment Information Set. Percentages weighted to account for NHATS complex survey design and provide nationally representative estimates. Prevalence of Cognitive Impairment among a Sample of Community-Living Medicare Beneficiaries Receiving Home Health Care 2011–2016: A Comparison of Available Data Sources*. (unweighted n = 1,214; weighted n = 5,856,333).

DISCUSSION

Among a nationally representative sample of Medicare beneficiaries receiving home health care from 2011–2016, estimates of the prevalence of cognitive impairment varied meaningfully depending on the data source. While 43.9% of these older adults had impaired cognitive function based on clinician evaluation in the home health patient assessment data, just 18.4% had probable dementia based on an algorithm to identify dementia in a national survey. Prior work estimates prevalence of diagnosed dementia among all Medicare beneficiaries at 14.4%²⁷ and among Medicare home health patients at 31.4%.²⁸ Researchers conducting analyses to identify or evaluate effects of cognitive impairment in this patient population must be cognizant of the relative strengths and weaknesses of available measures and the distinctions between the constructs they capture, and frame results accordingly. Regardless of the measure used, a sizable portion of the home health patient population has some form of cognitive impairment, supporting the importance of ongoing research in this area.

This brief report has notable strengths and limitations. We linked rich data sources at an individual level across a limited timeframe to directly compare measures of cognitive impairment. Using NHATS survey weights and design variables, we generated nationally representative estimates. However, the unweighted sample size is small compared to the overall Medicare home health patient population and due to lack of claims data availability, Medicare Advantage enrollees were excluded. There are other potentially valuable, nationally representative secondary data sources that could be used to further research in this area which we did not examine, including the Health and Retirement Study (HRS) and Medicare Expenditure Panel Survey (MEPS).

The measures compared in this study are designed to capture related, yet distinct underlying constructs and facets of cognitive impairment. Therefore, our results should not be taken as commentary on which measure is most accurate, but rather as additional information to contextualize findings and inform variable choice for future analyses of this patient population. The observed variation in estimates of cognitive impairment prevalence derived from these measures for the same group of home health patients speaks to the difficulty of ascertaining cognitive function in a population with significant underlying clinical complexity and social vulnerability.²⁰

In particular, it is likely that the OASIS measures we considered are more sensitive to mild cognitive impairment than the MBSF or NHATS measures. The OASIS measures are designed to capture broader symptoms of cognitive decline and may also reflect delirium among recently hospitalized or ill patients (although cognitive impairment prevalence is actually higher among community-referred patients⁶) and/or pre-clinical stages of dementia: previous research finds that HHC utilization significantly increases two years before an incident dementia diagnosis.²⁹ Regardless of the cause of impairment, information regarding symptoms of cognitive impairment is clinically relevant and highly likely to inform clinicians’ decisions relating to care planning and treatment.

When choosing an appropriate measure of cognitive impairment among home health patients, researchers must weigh a number of competing priorities. First, investigators must consider whether their interest lies in a particular underlying cause (e.g. vascular dementia), stage (e.g. mild cognitive impairment, early stage dementia), or symptom (e.g. memory loss) of cognitive impairment. Those interested solely in vascular dementia could isolate this cause from claims data, but not from surveys or OASIS. Those interested in early stage dementia may wish to use an instrument that captures symptoms of cognitive decline before an official diagnosis, such as OASIS, rather than claims data which often fail to capture mild/early-stage dementia.^30–32 Finally, those focused on memory loss may find the most value in OASIS which measures this symptom specifically.

Investigators must also consider their intended audience and the actions (e.g. federal policymaking, home health agency-level clinical practice) which they hope to inform. Researchers whose goal is to comment on existing Medicare payment policies, for example, may wish to rely on measures from claims or OASIS as these data are collected for all Medicare home health patients, and these instruments are already linked to the payment system. In contrast, those aiming to provide practice recommendations for home health agencies might gravitate towards OASIS measures only, as these are gathered during the routine course of care and available to the provider in real-time. Finally, investigators must be cognizant of the varying sensitivity of these measures for patient sub-groups. Notably, claims data routinely underestimate rates of dementia in racial and ethnic minority populations,^16–19 potentially limiting the utility of this data for studies focused on disparities.

CONCLUSION

Estimated rates of cognitive impairment among older adults receiving Medicare-funded home health care are sensitive to the particular data source and measure used. To produce reliable and replicable analyses, researchers should be thoughtful and transparent in their choice of measure and its benefits and limitations when disseminating results. Amid expanding knowledge regarding the trajectory and presentation of cognitive decline in older adults, there is a need for additional investigation of the psychometric properties of existing measures designed to capture cognitive impairment and a commitment to continued development of data infrastructure to support research in this critical field.

DATA AVAILABILITY STATEMENT:

Data used in this manuscript is drawn from a linked dataset including the National Health and Aging Trends Study (publicly available at www.nhats.org) and Medicare home health assessment and claims data (not publicly available, accessed via DUA RSCH-2020–55461 with the Centers for Medicare and Medicaid Services).