Clinical and Demographic Profiles of Home Care Patients with Alzheimer’s Disease and Related Dementias: Implications for Information Transfer across Care Settings

Abstract

Home health care (HHC) clinicians serving individuals with Alzheimer’s Disease and Related Dementias (ADRD) do not always have information about the person’s ADRD diagnosis, which may be used to improve the HHC plan of care. This retrospective cohort study examined characteristics of 56,652 HHC patients with varied documentation of ADRD diagnoses. Data included clinical assessments and Medicare claims for a 6-month look-back period and 4-year follow-up. Nearly half the sample had an ADRD diagnosis observed in the claims either prior to or following the HHC admission. Among those with a prior diagnosis, 63% did not have it documented on the HHC assessment; the diagnosis may not have been known to the HHC team or incorporated into the care plan. Patients with ADRD had heightened risk for adverse outcomes (e.g. urinary tract infection, aspiration pneumonia). Interoperable data across healthcare settings should include ADRD-specific elements about diagnoses, symptoms, and risk factors.

Introduction

Health policy leaders have recognized the importance of standardized and interoperable data exchanges across healthcare settings, including inpatient, outpatient, and post-acute care such as home health care (HHC). The Improving Medicare Post-Acute Care Transformation (IMPACT) Act of 2014 specified that post-acute and other providers use a common set of standards and definitions to provide access to longitudinal information that would facilitate care coordination and improve health outcomes for Medicare beneficiaries (Hall, Connor, & O'Malley, 2019). In 2019, the Centers for Medicare & Medicaid Services (CMS) expanded the IMPACT Act with requirements to ensure that a patient’s healthcare information follows them after discharge from a hospital or post-acute provider (Hall et al., 2019; Stefanacci, 2019). Interoperability and data flow across settings are critical for improving care for older adults with multiple chronic conditions and those with Alzheimer's disease and related dementias (ADRD), who are particularly vulnerable to losses and inaccuracies in the transfer of information across healthcare settings. This study examined the demographic and clinical profiles of post-acute HHC patients with different statuses regarding the documentation of ADRD diagnosis, providing insights into care coordination and interoperability needs along the ADRD trajectory.

ADRD represents a looming public health crisis, affecting roughly 5 million people and 11% of older adults in the United States (US) (Alzheimer's Association, 2013). Pathways to diagnosis and treatment for ADRD vary, with important implications for quality of life. An earlier qualitative study provided unique insights into these pathways, suggesting that some individuals have a “crisis event” pathway where ADRD diagnosis is precipitated by an emergency, while others experience a “dead-end” pathway where the patient and family receive fragmented care and inadequate information (Hinton, Franz, & Friend, 2004). Research on care transitions suggests that older adults with chronic conditions are particularly vulnerable to inadequate information transfer across providers, increasing their risk for diminished quality of care, medication errors and complications of chronic conditions (Bayliss et al., 2015; Foust, Naylor, Bixby, & Ratcliffe, 2012; Mondor et al., 2017; O'Connor et al., 2016). Older adults with ADRD are often clinically complex; roughly half have three or more chronic conditions (Lin, Fillit, Cohen, & Neumann, 2013). They are at heightened risk for potentially preventable hospitalizations for chronic conditions (e.g. diabetes, hypertension), unplanned hospital readmission, urinary tract infection, sepsis, and other adverse outcomes (Daiello, Gardner, Epstein-Lubow, Butterfield, & Gravenstein, 2014; Lin, Zhong, Fillit, Cohen, & Neumann, 2017; Shen, Lu, & Li, 2012). Some of these outcomes may be avoidable with more effective care coordination and information transfer across settings (Amjad et al., 2018; Naylor et al., 2014; Samus et al., 2014).

Research on home-based care coordination for persons with ADRD suggests that the home is a viable setting for detecting and addressing symptoms, safety concerns, and caregiver stress (D'Souza et al., 2015; Gitlin et al., 2018; Lau, Chan, & Szeto, 2019; Samus et al., 2014). In the post-acute HHC setting, nurses may be uniquely positioned to observe symptoms and respond with tailored interventions. However, their ability to do so may be limited by inadequate transfer of information about the patient’s cognitive status prior to entering HHC. A study of home-based care for individuals with dementia demonstrated the potential for adverse consequences when the content of care does not match the person’s specific needs, potentially resulting from incomplete assessment information and a fragmentation of care among multiple home-based caregivers (Reckrey, Bollens-Lund, & Ornstein, 2020). Although care for ADRD patients is the focus of a growing body of research, more evidence is needed on the characteristics, needs, and risk factors of HHC patients with ADRD.

This retrospective cohort study examined the demographic and clinical profiles of HHC patients with different statuses regarding the documentation of ADRD diagnosis, including those who have a known diagnosis prior to entering HHC and those who have cognitive symptoms and may be on a path toward ADRD diagnosis. The study used a uniquely configured dataset comprised of HHC clinical assessments linked with multiple years of Medicare claims data. This allowed us to capture more information about the care patterns of HHC patients with ADRD than is typically available within the data systems maintained by HHC providers. Evidence in this area may inform how interoperability efforts can address the needs of individuals at different stages of ADRD and how HHC clinicians can best support these patients and their caregivers.

Methods

Study design.

The analysis presented here is part of a retrospective, longitudinal cohort study examining care trajectories and outcomes in a racially and socioeconomically diverse sample of community-dwelling individuals with ADRD and those who have cognitive symptoms and may be on a path toward ADRD diagnosis. The cohort consists of 56,652 HHC patients served by the Visiting Nurse Service of New York (VNSNY), a large non-profit HHC provider in the U.S. The data include the federally mandated Outcome and Assessment Information Set (OASIS) clinical assessment administered at the start of HHC, linked with Medicare claims data for a 6-month look-back period and 4 years of follow up. Patients were included in the sample if they were admitted to HHC during the period of July 1, 2010 through December 31, 2012 with any diagnosis and had complete Medicare Fee-for-Service data for the look-back and follow-up periods. Claims data spanned the calendar years of 2010-2016, allowing for uniform timeframes.

Data sources.

OASIS is a standard assessment tool mandated by CMS with roughly 100 items on clinical status and service needs during an HHC episode. OASIS includes data on patient demographics, living arrangements, informal supports, co-morbidities, symptom severity, risk factors, prognosis, therapies, medication/equipment management, pain, wounds, neurocognitive/behavioral status, and physical function (ADL/IADL) (Kinatukara, Rosati, & Huang, 2005; Tullai-McGuinness, Madigan, & Fortinsky, 2009). OASIS data were extracted from the VNSNY electronic health record (EHR) along with additional EHR items, including language and medication information.

OASIS and other EHR data were linked with Medicare enrollment and claims data acquired through a Data Use Agreement (DUA) from CMS. These data included demographics, insurance enrollment information, chronic condition diagnoses, and claims for all services provided under Fee-for-Service Medicare Parts A and B in inpatient (e.g. hospitals, skilled nursing facilities), outpatient, office, HHC and hospice settings. All claims files contained service dates, diagnoses, and procedure codes.

Study sample.

We identified all patients who were admitted to HHC between July 1, 2010 and December 31, 2012 with any diagnosis and a Medicare beneficiary number available in the EHR. Demographic data were linked to the CMS data via beneficiary number, successfully matching 106,429 persons out of 106,469. For matched patients, we extracted the first start-of-care (SOC) OASIS assessment on record during 7/1/2010-12/31/2012. This SOC assessment was the index time point for each patient. Patients were eligible for inclusion in the study sample if they had complete claims data for 6 months preceding the index SOC OASIS and 4 years after (or died within the 4-year follow-up). Patients were excluded if they: (a) did not have complete Medicare Parts A and B coverage during this period (10,146 patients excluded for this reason); or (b) had Medicare Advantage (managed care) at some point during this period (39,631 excluded), since managed care claims were not available. These criteria yielded a sample of 56,652 patients.

ADRD subgroup classifications.

The sample of N=56,652 was categorized into groups according to ADRD diagnostic status, anchored on the index SOC OASIS. We created a binary indicator of whether the patient had an ADRD diagnosis documented in the OASIS as an HHC payment diagnosis. We identified ADRD diagnoses on the OASIS using the list of ICD-9 codes included in the ADRD indicator provided by CMS in the Chronic Condition Warehouse (CCW) (Alzheimer’s Disease ICD-9: 331.0; Related Dementias ICD-9: 331.1, 331.11, 331.19, 331.2, 331.7, 331.82; 290.0, 290.1, 290.10, 290.11, 290.12, 290.13, 290.20, 290.21, 290.3, 290.40, 290.41, 290.42, 290.43, 291.2, 294.0, 294.1, 294.10, 294.11, 797) (Chronic Condition Data Warehouse, 2018). Additionally, we used the CCW indicator to determine whether the patient had an ADRD diagnosis on record in the claims, using the historical date of first occurrence of the diagnosis, recorded for any date prior to 12/31/2016 (the last date observed in our dataset). This allowed us to determine, for those with an ADRD diagnosis in the claims, whether it appeared prior to or after the index OASIS.

Patient characteristics.

In addition to the ADRD subgroups, we used OASIS item M1740 to identify patients indicated by the HHC nurse to have memory deficit and/or impaired decision-making. Memory deficit is defined as “failure to recognize familiar persons/places, inability to recall events of past 24 hours, significant memory loss so that supervision is required,” and impaired decision-making as “failure to perform usual ADLs [activities of daily living] or IADLs [instrumental activities of daily living], inability to appropriately stop activities, jeopardizes safety through actions.” We examined the extent to which documentation of these cognitive symptoms correlated with ADRD diagnosis.

Additional demographic and clinical characteristics were derived from the OASIS assessment, Medicare beneficiary data, and additional EHR items. These measures included but were not limited to: age; sex; race; living arrangements; dual eligibility; language spoken; ADL/IADL dependencies; risk factors for hospitalization (e.g. history of falls, frailty indicators); and medication count.

Adverse health outcomes (AHOs).

AHOs were defined as hospitalizations during the 4-year follow-up period for selected conditions for which ADRD patients are at heightened risk: urinary tract infection (UTI); falls; pneumonia and influenza; pneumonitis due to solid and liquids (i.e. aspiration pneumonia); and dehydration (Jorgensen et al., 2018; Lampela et al., 2017; Lin et al., 2017; Manabe et al., 2017; Marshall et al., 2016; Nair et al., 2018; Roitto et al., 2018; Shen et al., 2012). We constructed these measures using the Medicare inpatient claims data, including ICD-9/ICD-10 codes and service dates.

Analysis.

Descriptive analysis compared the demographic and clinical profiles of the ADRD subgroups, using chi-square tests and analysis of variance for categorical and continuous variables, respectively. To compare the incidence of AHOs by subgroup, we needed to account for subgroup differences in survival, which otherwise could bias the findings on AHOs observed during the 4-year follow-up. Accordingly, we conducted negative binomial regression on the count of each AHO, with the independent variable being subgroup assignment and adjusting for length of survival. Data preparation and descriptive analyses were conducted in SAS Version 9; negative binomial regression was conducted in Stata Version 14. All study activities were approved by the Institutional Review Board of VNSNY (Protocol #1124570-1).

Results

In the overall sample of 56,652 HHC patients, 49% (N=27,521) had an ADRD diagnosis observed in at least one of our data sources (the SOC OASIS and/or the claims data). Within the group of 27,521 patients with ADRD observed somewhere in the data, we identified four subgroups according to where and when in the data the diagnosis appeared (Figure 1). The first two groups had ADRD documented on the SOC OASIS. Group 1 (N=6,736; 12% of overall sample) had ADRD indicated in the claims prior to the SOC OASIS. Using the historical indicator of the date of first ADRD occurrence, we found that the first occurrence appeared in the claims on average 50 months prior to the SOC (SD=42, maximum=168). Group 2 (N=487; 1% of sample) had ADRD in the claims during the 4 years following the SOC, with a mean of 2 months until the diagnosis appeared in the claims (SD=6, maximum=44).

Figure 1.

Patient classification into ADRD subgroups

The other two subgroups did not have ADRD indicated on the SOC OASIS but did have it in the claims. Group 3 (N=11,397; 20% of sample) had ADRD indicated in the claims prior to the SOC, with the first occurrence appearing in the claims an average of 46 months prior (SD=42, maximum=170). Group 4 (N=8,901; 16% of sample) had the first occurrence sometime during the 4-year follow-up period, with an average of 19 months after the SOC (SD=14, maximum=49).

Of the total of 18,133 patients who had an ADRD diagnoses appearing in the claims prior to the SOC OASIS (Groups 1 and 3 combined), only 37% (Group 1) had ADRD indicated on the SOC OASIS. This suggests that, for the remaining 63% (Group 3), the previously established ADRD diagnosis may have been unknown to the HHC team and/or not factored into the plan of care.

Table 1 presents demographic and clinical profiles of the four ADRD subgroups and the comparison group without ADRD. All variables in Table 1 had statistically significant differences across subgroups (p<0.0001), though statistical significance is not surprising given the large sample size. The overall sample had a mean age of 79 and was 63% female. Patients with ADRD documented on the OASIS were older than the other groups. Group 1 had a mean age of 86 and a greater proportion of female patients than other groups. The overall sample was diverse in terms of race and language, with one third nonwhite and 21% non-English speakers. About a third was dually-eligible for Medicare and Medicaid, a proxy for low socioeconomic status (SES), with greater proportions among patients with an ADRD diagnosis prior to HHC admission.

Table 1.

Selected patient characteristics – overall sample and by ADRD subgroup

Subgroup	Overall Sample	ADRD Dx on Start-of-Care OASIS		ADRD Dx Not on Start-of-Care OASIS
		ADRD in claims prior to OASIS Group 1	ADRD in claims after OASIS Group 2	ADRD in claims prior to OASIS Group 3	ADRD in claims after OASIS Group 4	No ADRD in claims: Comparison group
Size of subgroup	56,652	6,736	487	11,397	8,901	29,131
Subgroup as percentage of overall sample	100%	12%	1%	20%	16%	51%
Age, mean (SD)	79.3 (11.3)	85.9 (7.3)	85.5 (8.3)	82.8 (9.7)	82.2 (9.3)	75.5 (11.8)
Female	63%	69%	63%	64%	66%	60%
Race
White	66%	66%	67%	67%	68%	66%
Black	15%	14%	17%	14%	15%	16%
Hispanic	12%	14%	10%	12%	11%	11%
Asian / Pacific Islander	5%	4%	*	5%	5%	5%
Other / American Indian or Alaska Native / Unknown	2%	2%	*	2%	2%	2%
Language spoken
English	73%	68%	71%	69%	75%	75%
Spanish	8%	12%	7%	10%	8%	7%
Russian	3%	3%	3%	6%	2%	2%
Other/not recorded	16%	17%	19%	15%	15%	16%
Living alone	43%	32%	44%	43%	48%	45%
Dually eligible (Medicare and Medicaid)	34%	43%	34%	42%	33%	29%
Months from ADRD to SOC (or viceversa), mean, SD, maximum	--	−50 (42) 168	2 (6) 44	−46 (42) 170	19 (14) 49	--
Impaired decision-making and/or memory deficit	15%	51%	34%	18%	11%	6%
Medication count at home care admission	9.2 (4.3)	8.4 (4.1)	7.8 (4.2)	9.6 (4.0)	9.0 (4.3)	9.3 (4.4)
Prescription med count at home care admission	3.7 (4.0)	3.5 (3.8)	3.8 (3.8)	4.0 (4.2)	3.7 (3.9)	3.6 (3.9)
Risk factors for hospitalization
Recent decline in mental, emotional, or behavioral status	15%	36%	37%	17%	13%	9%
2 or more hospitalizations in past 12 months	23%	19%	13%	25%	20%	23%
History of falls	17%	23%	26%	22%	21%	13%
Frailty indicators (e.g. weight loss, self-reported exhaustion)	30%	33%	37%	32%	30%	28%
Sum of ADL/IADL dependencies at home care admission	5.9 (3.2)	8.8 (3.1)	6.7 (3.3)	6.8 (3.3)	5.5 (2.9)	5.0 (2.8)
Risk of developing pressure ulcers	28%	50%	32%	36%	25%	20%
Urinary incontinence	38%	69%	50%	49%	39%	25%

Note: All differences in patient characteristics across subgroups in Table 1 were significant at the level of p-value <0.0001.

In addition to the ADRD subgroups, we identified patients marked as having a memory deficit and/or impaired decision-making on the OASIS. The proportions of patients with at least one of the two cognitive symptoms documented on the OASIS varied by ADRD subgroup. In Group 1, where patients had an ADRD diagnosis in the claims prior to HHC admission as well as on the OASIS, 51% had at least one of these cognitive symptoms documented. In Group 3, where patients had a previously established ADRD diagnosis that was not documented on the OASIS, 18% had at least one of these symptoms. Six percent of the comparison group (with no ADRD diagnosis indicators) had one or both of these cognitive symptoms.

The ADRD subgroups also varied in other clinical and functional characteristics, including risk factors for hospitalization identified on the OASIS. For example, the proportion with two or more hospitalizations in the past 12 months was highest among Group 3, at 25%, whereas 19% of Group 1 had this risk factor. This compared to 23% in the reference group with no ADRD indicators. The proportion of patients with a history of falls ranged from 21-23% among the ADRD subgroups, compared to 13% in the reference group. The average medication count in the overall sample was 9.2, ranging from 7.8 in Group 2 to 9.6 in Group 3. Functional dependencies in ADLs/IADLs varied by ADRD subgroup, with Group 1 having the greatest needs (mean count of 8.8 dependencies), and Group 4 having the lowest among ADRD patients (mean of 5.5).

Table 2 shows comparisons across the ADRD subgroups in key adverse health outcomes (AHOs), both unadjusted and adjusted for differences in length of follow-up due to death. AHOs were defined as selected diagnoses in an inpatient admission during 4-year follow-up. In unadjusted comparisons, UTI and dehydration are among the most common AHOs, with 29% and 39% of the overall sample having at least one occurrence, respectively. For all unadjusted comparisons of AHO occurrence by subgroup, the differences were statistically significant with a p-value of p<0.0001, except for pneumonitis due to solids and liquids (p=0.0003). However, there were also significant differences by subgroup (p<0.0001) in rates of death during 4-year follow-up. The percentage of patients who died ranged from 59% (Group 4) to 73% (Group 1), compared to 43% of the no-ADRD reference group. Accordingly, the subgroups differed in the mean length of the follow-up period in which patients could be observed for AHO events.

Table 2.

Adverse health outcomes: unadjusted and adjusted by months of follow-up – overall and by subgroup

Subgroup	Overall Sample (N=56,652)	ADRD Dx on Start-of-Care OASIS		ADRD Dx Not on Start-of-Care OASIS
		ADRD in claims prior to OASIS Group 1 (N=6,736)	ADRD in claims after OASIS Group 2 (N=487)	ADRD in claims prior to OASIS Group 3 (N=11,397)	ADRD in claims after OASIS Group 4 (N=8,901)	No ADRD in claims: Comparison group (N=29,131)
Adverse outcome of selected inpatient stay diagnoses during 4-year follow-up period, % with one or more occurrences unadjusted
Urinary tract infection (UTI)	29%	40%	34%	37%	45%	19%
Falls	13%	14%	15%	16%	23%	9%
Pneumonia and influenza	24%	26%	25%	30%	33%	19%
Pneumonitis due to solids and liquids	9%	15%	14%	12%	12%	4%
Dehydration	39%	37%	36%	43%	54%	33%
Any of the above adverse health outcomes	59%	66%	62%	67%	77%	49%
Died during 4-year follow-up period	56%	73%	68%	62%	59%	43%
Months of follow-up, mean (SD)	35 (19)	29 (20)	32 (19)	33 (19)	38 (17)	37 (19)
Count of adverse outcomes during 4-year follow-up period, negative binomial adjusted mean (SE) (95% CI)
Urinary tract infection (UTI)	0.71 (0.01) (0.69, 0.72)	1.22 (0.03) (1.15, 1.28)	0.76 (0.08) (0.60, 0.91)	0.98 (0.02) (0.94, 1.02)	1.04 (0.02) (1.00, 1.09)	0.36 (0.01) (0.35, 0.37)
Falls	0.17 (0.002) (0.16, 0.17)	0.21 (0.01) (0.19, 0.22)	0.21 (0.03) (0.16, 0.26)	0.22 (0.01) (0.21, 0.23)	0.29 (0.01) (0.27, 0.30)	0.10 (0.001) (0.09. 0.10)
Pneumonia and influenza	0.52 (0.01) (0.51, 0.53)	0.61 (0.02) (0.57, 0.65)	0.51 (0.07) (0.38, 0.64)	0.67 (0.02) (0.64, 0.70)	0.67 (0.02) (0.63,0.70)	0.38 (0.01) (0.37, 0.40)
Pneumonitis due to solids and liquids	0.16 (0.003) (0.15, 0.16)	0.36 (0.02) (0.33, 0.40)	0.23 (0.04) (0.15, 0.31)	0.25 (0.01) (0.23, 0.27)	0.18 (0.01) (0.17, 0.20)	0.06 (0.001) (0.06, 0.06)
Dehydration	0.98 (0.01) (0.96, 1.00)	0.96 (0.03) (0.91, 1.01)	0.85 (0.08) (0.68, 1.02)	1.16 (0.02) (1.12, 1.21)	1.35 (0.03) (1.29, 1.40)	0.77 (0.01) (0.75, 0.79)
Any of the above adverse health outcomes	2.00 (0.10) (1.97, 2.03)	2.49 (0.05) (2.39, 2.60)	2.03 (0.16) (1.73, 2.34)	2.52 (0.03) (2.45, 2.60)	2.71 (0.04) (2.62, 2.79)	1.40 (0.01) (1.37, 1.42)

Note: All differences across subgroups shown in Table 2 were significant at the level of p-value p<0.0001, except for the unadjusted adverse outcome of pneumonitis due to solids and liquids (p=0.0003).

To determine whether differences in outcomes were still significant after accounting for differences in length of follow-up, we performed negative binomial regressions on each adverse outcome as a count variable (e.g. the number of UTI events as an inpatient stay diagnosis), adjusting for the number of months of follow-up. Subgroup differences in the adjusted means generated from the negative binomial regressions were statistically significant for all AHOs (p<0.0001). The non-ADRD comparison group had the lowest incidence for all outcomes, as was the case with the unadjusted indicators.

To identify clinically meaningful findings from these regressions, we examined the confidence intervals corresponding to the adjusted means and identified ADRD subgroups with confidence intervals that did not overlap with other subgroups for each outcome. Group 1 had the highest incidence of UTI, with a mean of 1.22 events (SD=0.03); the confidence interval for this subgroup (1.15-1.28) had no overlap with any other subgroup, suggesting a distinct pattern of heightened risk of hospitalization for UTI within this group. Group 1 also had a higher mean number of pneumonitis events due to solids or liquids – often referred to as “aspiration pneumonia” – with a mean of 0.36 events (SD=0.02) and no overlapping confidence intervals with other subgroups. Group 4, which was comprised of patients who were diagnosed with ADRD during the 4-year follow-up period, had the greatest incidence of falls, with a mean of 0.29 events (SD=0.01) and a distinct confidence interval. This compared to a mean ranging from 0.21 (SD=0.01) to 0.22 (SD=0.01) in Groups 1, 2, and 3 and a mean of 0.10 (SD=0.001) in the no-ADRD comparison group. Group 4 also had the greatest incidence of dehydration, with a mean of 1.35 events (SD=0.03) and a distinct confidence interval, compared to a range of 0.85 (SD=0.08) to 1.16 (SD=0.02) in Groups 1, 2, and 3 and a mean of 0.77 (SD=0.01) in the no-ADRD group.

In sensitivity analyses, we reran the negative binomial regression removing the no-ADRD comparison group in order to discern whether this group was driving the statistical significance in the models. We also reran the models collapsing Groups 1 and 2, given that the latter is relatively small and possibly reflective of billing anomalies in the claims data. Differences across subgroups remained statistically significant in all sensitivity analyses (p<0.0001), with no notable differences in the findings.

Discussion

This study examined the demographic and clinical profiles of HHC patients with different ADRD statuses with reference to the HHC admission. Overall, nearly half of the 56,652 patients eligible for inclusion in the analytic sample had an ADRD diagnosis in at least one of the data sources, including the SOC OASIS assessment and/or the claims data, either in the historical record prior to HHC admission or during the 4-year follow-up period. Prior research estimated ADRD prevalence up to 29% among patients referred for home-based services (Kronish, Federman, Morrison, & Boal, 2006). In our sample, the comparable estimate is comprised of the subgroups of patients who had been diagnosed with ADRD prior to HHC admission, which is 32% (combining Groups 1 and 3). The fact that our sample has a slightly higher prevalence at the time of HHC admission compared to prior estimates may be partly due to the complexity of the population served by VNSNY. Moreover, the sample selection criteria filtered out patients who enrolled in Medicare Advantage during the study period, which may have resulted in a sample selection of older, more clinically complex patients with a greater prevalence of ADRD than the full Medicare population receiving HHC.

In addition to the 32% diagnosed prior to HHC, 16% were diagnosed during the 4-year follow-up period (Group 4), suggesting that a significant portion of the patients who were not previously diagnosed at the time of HHC admission may have been on a path toward ADRD diagnosis while receiving HHC. The small proportion that had ADRD documented on the OASIS but not in the claims data until sometime after the HHC admission (Group 2, representing 1% of the sample) might reflect situations where a physician made an ADRD diagnosis shortly after HHC admission, since HHC clinicians are not positioned to make new diagnoses or document diagnoses not indicated by the referring physician in the HHC plan of care.

The finding that roughly half of this Medicare Fee-for-Service sample receiving HHC is affected by ADRD suggests that the impact of ADRD on the HHC population may be broader than previously understood. Moreover, among those who had an ADRD diagnosis in the claims prior to HHC admission (18,133 patients, Groups 1 and 3 combined), 63% did not have the ADRD diagnosis documented on the OASIS (Group 3). Although the OASIS is not designed to comprehensively document all comorbidities, it is designed to capture the diagnoses that inform the plan of care in the HHC setting, as well as HHC payment. Thus, the 63% of previously diagnosed patients may not have had sufficient documentation of their ADRD diagnosis to allow for the HHC team to fully incorporate ADRD-related needs into the home-based plan of care. Although the physician referring the patient to HHC may have been aware of the diagnosis, the HHC nurse and other clinicians on the HHC team – such as physical therapists, occupational therapists, and social workers – may not have had this information while implementing the physician-ordered plan of care. This gap in information potentially introduces missed opportunities to address patient and caregiver needs related to cognitive decline – such as behavioral issues, wandering, and other safety concerns such as falls prevention.

The sizeable number of patients in Group 3 also suggests that the HHC clinician overseeing the patient’s care might not be aware that the patient had been previously diagnosed with ADRD. HHC clinicians typically do not have direct access to the medical records of other settings outside of the HHC organization, and it is usually not within the scope of practice for an HHC clinician – e.g. a nurse, physical or occupational therapist, or social worker – to make an ADRD diagnosis when it is not documented by the referring physician in the order for HHC. This raises questions about potential inadequacies in the transfer of critical information from prior healthcare settings to the HHC team which could be used to improve home-based supports for patients and their caregivers along the ADRD trajectory.

Moreover, the OASIS items for documenting cognitive symptoms (e.g. memory deficit and impaired decision-making) do not consistently agree with the presence of an ADRD diagnosis. This suggests that an HHC clinician cannot rely on the snapshot assessment of cognitive symptoms to detect potential needs for cognitive evaluation and/or supportive services related to cognitive decline. An individual with ADRD may not always appear to be cognitively impaired at a particular snapshot in time, especially during the early stages of ADRD (Pocnet et al., 2015). Therefore, an HHC clinician assessing a patient without historical information about their ADRD status will not necessarily detect patient needs related to cognitive status while relying on the OASIS assessment process alone.

By and large, the demographic profiles of the different ADRD subgroups were consistent with expectations. Group 1 was comprised of patients with a prior diagnosis, including patients with a longer time since diagnosis who are likely further along in the ADRD trajectory; thus, it is not surprising that this group is older, on average, than the other subgroups. The greater proportion of women in Group 1 is likely correlated with the older age distribution. The greater proportion of dual eligible patients in Group 1 is consistent with prior research suggesting that a transition from Medicare-only to dual eligible for Medicare and Medicaid is common among older adults with ADRD, driven by a complex set of factors including increased medical need over time and potential spenddown of financial resources (Feng, et al., 2019). That fewer patients in Group 1 lived alone is also intuitive; however, the fact that roughly one third of this group lives alone raises potential safety concerns. The subgroups’ clinical and functional profiles are also consistent with the longer history of ADRD in Group 1; this group had greater ADL/IADL needs and greater risk of pressure ulcers identified at HHC admission. That Group 1 had a slightly lower medication count than Groups 3 and 4 may be due to some patients in more advanced stages of ADRD being taken off certain medications near the end of life (Denholm, Morris, & Payne, 2019).

The finding that all ADRD subgroups had greater incidence of AHOs during the 4-year follow-up period compared to the non-ADRD comparison group – after adjusting for differences in survival – is consistent with prior research demonstrating heightened risks for adverse outcomes among ADRD patients (Daiello et al., 2014; Lin et al., 2017; Manabe et al., 2017; Nair et al., 2018; Shen et al., 2012). Among the subgroups, Group 4 had the greatest incidence of falls and dehydration, and it matched the high incidence in Group 3 of pneumonia/influenza. Given that patients in Group 4 were diagnosed with ADRD at some point during the 4-year follow-up period, it is possible that these AHO events were part of a series of crisis events that precipitated evaluation and ultimately ADRD diagnosis. This would be consistent with a prior qualitative study suggesting that some individuals have a pathway to ADRD diagnosis characterized by crisis events that trigger the recognition of cognitive decline and the need for cognitive evaluation (Hinton et al., 2004). Also consistent with prior evidence is that the group with the longest time since ADRD diagnosis (Group 1) had the highest incidence of aspiration pneumonia, which is a known risk in advanced stages of ADRD as swallowing difficulties increase (Manabe et al., 2017; Sato et al., 2014). The higher incidence of UTI in Group 1 may be related to this group’s higher prevalence of urinary incontinence, a potential risk factor for UTI especially among older women (Caljouw, den Elzen, Cools, & Gussekloo, 2011; Eriksson, Gustafson, Fagerstrom, & Olofsson, 2010).

Some differences in clinical characteristics upon HHC admission were challenging to interpret. In particular, Group 3 – which had a history of ADRD diagnosis that was not documented on the OASIS – had a higher proportion with a history of 2 or more hospitalizations within the past year, as well as the highest mean count of medications upon HHC admission. This potentially suggests a distinct pattern in clinical complexity within this group, but it is unclear whether this distinction is clinically meaningful and what unmeasured sources of variation might account for this difference.

Some study limitations are noted. First, patients were excluded if they enrolled in Medicare Advantage (managed care) during the study period, since managed care claims were not available in the DUA with CMS. Second, the available data did not include direct measures of SES, such as income and education; we relied on dual eligibility as a proxy for low SES. Third, the study was limited to a single large, urban HHC provider; this may limit generalizability to other providers and geographies. The focus on the one provider served the purpose of a larger study of disparities in healthcare use and outcomes in the ADRD population in HHC. In the larger study, the diverse VNSNY population was advantageous for examining subpopulations that have insufficient representation in national longitudinal datasets, thereby enhancing generalizability to understudied groups. A subsequent study can reproduce these methods with data from one or more additional agencies in different regions to further demonstrate reproducibility.

Conclusions and policy implications

The study findings suggest potential strategic priorities surrounding ADRD in efforts to improve care coordination and interoperability under the IMPACT Act. Providers referring patients to HHC need a standard mechanism for transferring ADRD diagnostic information to HHC providers. This information should not only be captured in the EHR of the receiving HHC organization, but also in the standardized OASIS assessment to incorporate ADRD-related information into the HHC plan of care. This information may be used to identify patients in need of interventions to manage risks that are heightened for those with ADRD, such as UTI and aspiration pneumonia. Additionally, for patients without an established ADRD diagnosis but who may be demonstrating signs and symptoms, OASIS indicators of cognitive symptoms could be used to trigger referral for a comprehensive cognitive evaluation to facilitate more timely ADRD diagnosis, as well as access to caregiver support services. Efforts to improve interoperability of data need to include the development and adoption of standard, ADRD-specific data elements to effectively transmit information about existing diagnoses, symptoms, and risk factors across acute, post-acute, and outpatient settings. These interoperable elements could be used to better understand the ADRD disease burden in the post-acute care setting and improve quality of care for community-dwelling individuals along the ADRD trajectory.