Medical foster home is less costly than traditional nursing home care

Abstract

Objective

To compare the costs of Community Nursing Homes (CNHs) to Medical Foster Homes (MFHs) at Veteran Health Administration (VHA) Medical Centers that established MFH programs.

Data Sources

Episode and costs data were derived from VA and Medicare files (inpatient, outpatient, emergency room, skilled nursing facility, dialysis, and hospice).

Study Design

Propensity scores matched 354 MFH to 1693 CNH Veterans on demographics, clinical characteristics, health care utilization, and costs.

Data Extraction Methods

Data were retrieved for years 2010‐2011 from the VA Corporate Data Warehouse, VA Health Data Repository, and the VA MFH Program through the VA Informatics and Computing Infrastructure (VINCI).

Principal Findings

After matching on unique characteristics of MFH Veterans, costs were $71.28 less per day alive compared to CNH care. Home‐based and mental health care costs increased with savings largely attributable to avoiding CNH residential care. When average out‐of‐pocket payments by Veterans of $74/day are considered, MFH is at least cost neutral. Mortality was 12 percent higher among matched Veterans in CNHs.

Conclusions

MFHs may serve as alternatives to traditional CNH care that do not increase total costs with mortality benefits. Future work should examine the differences for functional disability subgroups.

1. INTRODUCTION

In fiscal year 2012, 9 percent of VA health care spending, $4.9 billion of VA's $54 billion budget, was spent on long‐term care (LTC).1 While the general population is increasingly interested in containing costs for nursing home (NH) care, the VA must respond to demands for NH care with even more urgency because 40 percent of the Veteran population is over age 65 compared to only 13 percent of the general population.2

Expenditures for one segment of LTC, those receiving Community Nursing Home (CNH) care, in the Veterans Health Administration (VHA) in FY16 were estimated at $970 million, an increase from $385 million in FY06.3 These costs are expected to rise as the number of Veterans 65 years of age or older for whom the VHA is obligated to pay for NH care, increases from approximately 600 000 in 2016 to over 1 million by 20251 and VA nursing homes, which are known as Community Living Centers (CLCs), are increasingly utilized for short‐stay rehabilitation and hospice care rather than LTC for Veterans.

In the JAMA article from The National Academy of Medicine, “Preparing for Better Health and Health Care for an Aging Population,” the authors identify the need to study new models of care delivery and expand the eldercare workforce as vital components of a response to these projected deficiencies.4 The Home Based Primary Care (HBPC) interdisciplinary team model is not new, but when provided in adult foster care homes as part of the VA Medical Foster Home (MFH) program, it represents a new, rapidly expanding care model with a potential to add foster caregivers to the eldercare workforce.5

As an extension of the HBPC program, the MFH program is an adult foster care program established in 2008 and designed to provide an alternative community‐based living arrangement. Up to three Veterans who cannot live independently because of physical or mental disabilities (these eligibility criteria are identical for CNH Veterans) live in each MFH. These Veterans are unable to participate in the traditional HBPC program because they cannot care for themselves at home and do not have a familial caregiver to provide 24‐7 care.6

The VA MFH program provides primary care to the Veteran through HBPC interdisciplinary staff. The HBPC care team provides caregiver training tailored to the VA plan of care, home assessment, and patient care. Veterans pay out‐of‐pocket for MFH room and board, while the VA pays for and provides medical care coordinated by the home care team. Veterans are paired with a community caregiver who typically has prior experience working in a NH, caring for loved ones, or other caregiving experience.6 Potential MFH caregivers are initially interviewed by the MFH coordinator who is a senior‐level social worker. A HBPC occupational therapist also visits the home to assess suitability of the home for the MFH program in terms safety and any needed home modifications. Veterans’ daily care is provided in the caregiver's home 24 hours a day, 7 days a week. The MFH coordinator makes unannounced, monthly visits to asses MFH safety. Veterans only have a choice for MFH if: (a) the program is offered at their VA Medical Center (the program is not mandated—authorizing the program is at the discretion of each medical director), (b) they became aware of the program through advertisement, word of mouth, or a VA staff member, and (c) MFHs beds and caregivers are available.

H.R. 5693 The Long‐Term Care Veterans Choice Act authorizes VA to place up to 900 Veterans for whom the VA is required to provide NH care in a MFH at the expense of the United States.7 Understanding utilization and costs of care prior to this expansion is important to Veterans and families seeking these programs, referring clinicians, and policymakers trying to provide the highest value care at the lowest cost; however, much of the literature is dated or focused on persons with developmental disabilities.8, 9, 10, 11, 12, 13, 14, 15, 16, 17

The objective of this study was to compare costs of care for a cohort of Veterans in the MFH program to a similar sample of Veterans receiving LTC in a CNH. We hypothesized that MFH total costs of care would be lower for Veterans in the MFH program compared to those residing in CNHs with savings attributable to lower inpatient and residential care costs.

2. STUDY DESIGN AND METHODS

2.1. Study design

This was a retrospective study using observational data retrieved from the VA Corporate Data Warehouse, VA Office of Geriatrics and Extended Care Operations, VA Health Data Repository, and the VA MFH Program through the VA Informatics and Computing Infrastructure (VINCI).18 VINCI is a secure, virtual platform that provides access to research software, software development tools, and VA national health data. The Institutional Review Board approved the study.

2.2. Sources

Cost data were derived from the following two data types: Cost Accounting for Care Purchased in the Community (inpatient, outpatient, and ancillary) and Cost Accounting for Care Purchased by VA (inpatient and outpatient). Data were derived from a file containing patient‐level longitudinal records of enrollment across LTC settings of care.19 Both Medicare and VA data included the following costs for the following types of care: inpatient, outpatient, emergency room, skilled nursing facility, dialysis, and hospice.

2.3. Sample

Administrative data were used to identify Veterans with a CNH visit in fiscal years 2010 or 2011 (October 1, 2009‐September 30, 2011; Figure 1). Patients with prior HBPC, CLC, or CNH claims were removed because the intent was to study an incident cohort of new VA LTC users. No such exclusions were made for the MFH cohort to maximize sample size. An analysis that examined the differences between a CNH cohort that included patients with these claims resulted in an increased CNH sample size by 2505 Veterans, but these Vetrans had higher comorbidity scores and a lower propotion of Veterans in the CNH cohort for whom CNH care is paid for by VHA. Based on observed covariates, the CNH cohort with these restrictions is more similar to the MFH cohort. Thus, we opted to rely on the restricted CNH cohort. Both MFH and CNH were available at all sites but how each program was presented to Veterans was not standardized and Veterans were not randomized into one of the two programs. Veterans may have been referred by a VA staff member or self‐referred after learning about the program from a staff member, another Veteran or from the media following a personal interest story about the MFH program.

Figure 1.

Inclusion algorithm for Veterans in community nursing homes

Veterans in a CNH for fewer than 60 days were assumed to be admitted for short‐term rehabilitation and not with the same intent as those enrolled in MFH, in which the intention is to remain in the environment long term. MFH Veteran stays shorter than 30 days were excluded because these stays were believed to be too short to allow Veterans sufficient exposure to the MFH program during which changes in health care utilization would be expected (N = 38). Discharge status for these 38 exclusions were as follows: 31 percent died, 21 percent home with family or under self‐care, 18 percent nursing home, 13 percent hospitalized, 4 percent hospice, 2 percent assisted living facility, and 11 percent had missing discharge status data. We performed a sensitivity analysis that matched the minimum length of stay in the two cohorts at 15, 30, 45, and 60 days. Generally, we observed that this resulted in increased comorbidity scores in both cohorts, increased CNH predicted mortality and morbidity rates, and a reducuction in the number of Veterans with long‐stay CNH payment. This suggests that shifting these inclusion criteria changes some Veterans characteristics, but not in a manner that clearly provides a better representation of MFH Veteran characteristics. Moreover, the increased rates of sicker Veterans in the CNH cohort resulted in larger on average differences between the MFH and CNH cohorts, which justifies the use of unequal length of stays. CNH Veterans that were at VAMCs without MFH programs were also excluded from the analysis because stakeholder input indicated that medical centers without a MFH program in the choice set may be systematically different than those with a program. We speculated that these potential unmeasured organizational differences might introduce greater bias than the possibility of unobserved patient‐level covariates within systems that offer both. Thus, we opted to only consider systems that offer both. Veterans receiving hospice at enrollment were also excluded. The final cohorts before propensity score analysis comprised 354 MFH Veterans and 1693 CNH Veterans.

2.4. Covariates

Demographic variables included age, gender, ethnicity, and marital status. Income levels were derived from U.S. Census Bureau data.20 The number of emergency room visits in the year prior was used to provide an indicator of utilization. We also recorded indicators of severity of illness such as the Elixhauser Comorbidity Index, the JEN Frailty Index and the Care Assessment Need (CAN) prediction of one‐year mortality. We chose the Elixhauser Comorbidity Index (typically reported as 0, 1, 2‐3, and >3) over other indices (eg, Charlson) because it accounts for diagnoses such as hypertension, electrolyte imbalance, paralysis, depression, psychoses, alcohol use, and drug use which are more prevalent in Veteran populations.21, 22 The JEN Frailty Index creates a risk score using 13 impairment categories pulled from 1800 diagnosis related to risk of LTC admission and is significantly correlated with Medicare and Medicaid expenditures as well as mortality.23 A score of 0‐3 is considered low risk, 4‐6 moderate risk, and ≥7 high risk. Care Assessment Need (CAN) Scores are based on 36 input variables from a composite index derived from the US Census American Community Survey and the Rank and Service Branch from the VA and Department of Defense Identity Repository. Scores range between 0 and 1 with higher scores representing higher probability of mortality. These scores are derived weekly, and in our analysis, we relied on the 1‐year mortality predictions scores.24 Existing literature suggests that CAN scores are accurate predictors of 1‐year mortality for VHA patients receiving primary care. The CAN score sorted patients who died from patients who lived correctly 86 percent of the time (c‐statistic) at 90‐days and 85 percent of the time (c‐statistic) at 1 year.25

Veteran benefit groups were included because payment for services, co‐pays and other financial implications vary based on level of service connection.26 VHA is obligated to pay for nursing home care (referred to as Priority Group 1a) when Veterans have a VA service‐connected disability or when they are unemployable because of service‐connected disability. Veterans who have a catastrophic disability or who receive regular aid and attendance or housebound benefits qualify for Priority Group 4.

Costs were constructed as actual per Veteran costs with assignment of all costs to individual Veterans following enrollment in the respective programs. Cost categories included hospitalization, emergency room, HBPC visits, outpatient utilization (physician, chaplain, social work, screening, telephone, dental, home care, nursing home, day care), skilled nursing facility, hospice, dialysis, laboratory, mental health, procedures, radiology, specialty care, therapy, recreation therapy, home health, respite stays, and community nursing home encounters (outpatient, rehabilitation), CLC encounters, and CNH encounters. Indirect costs for CLC stays were not included in total costs.

The day of enrollment in each program was defined as the admission date to the program of interest. Any episode occurring on or 365 days before that event was defined as pre‐enrollment utilization and cost, and all episodes occurring after the day of enrollment were included in post‐enrollment utilization and costs. Length of stay was calculated using the last visit in the assigned setting of care as the Veteran's discharge date; for example, a Veteran may have been admitted to MFH on October 15, 2009, hospitalized December 15‐20, 2009, followed by a brief CLC stay from December 20‐30, and then remained in MFH until the end of the 365‐day follow‐up period on October 14, 2010. This Veteran would be considered in the MFH cohort, and utilization would be calculated for the entire 365 days despite being in the hospital and in a CLC for a period of time, because we intended to follow Veterans utilization and costs according to the setting of care to which they were initially assigned.

2.5. Analysis

To ensure objectivity of the analysis and to closely mimic a randomized clinical trial to estimate the effects of being placed in MFH versus CNH, we followed a two‐stage procedure.27, 28 The first stage comprised propensity score estimation and declaration of the procedures that would be used for data analysis. In the second stage, we treated the unobserved potential outcome for each unit as missing data29 and multiply imputed these missing potential outcomes to compare the mortality and costs for Veterans in MFH to Veterans in CNH.30, 31 Specifically, for all of the Veterans in MFH, we imputed the mortality and costs that would have been observed if they were in a CNH, and for all of the patients that were in a CNH, we imputed the mortality and outcomes if they were in a MFH. This imputation method that utilized the propensity score was shown to have good operating characteristics in comparison with matching and weighting.31 It is important to note that the first stage was implemented without viewing the outcome data.

2.5.1. Propensity score estimation

The propensity score was estimated using a logistic regression in a stepwise iterative fashion.32 Initial covariate selection for the propensity score model was based on prior research and clinical knowledge about factors that lead to selection of MFH and CNH.6, 33 Next, other covariates, as well as quadratic and interaction terms, were selected based on likelihood ratio testing. The final propensity score model included 35 variables consisting of socio‐demographics, baseline utilization, primary VA medical center where care was delivered, medical history, prognostic information, and some of their interactions. We excluded Veterans that were not within the range of the other groups’ propensity scores. This truncation excluded Veterans whose probability of choosing MFH or CNH approached 100 percent. Because this is a comparative effectiveness study, we were only interested in comparing participants who had some chance of utilizing either program. Sociodemographic characteristics were compared between those included and excluded from analyses. Generally, married Veterans in the VHA benefit group that covers CNH care, who were frailer and who experienced large numbers of prior hospitalizations, had very high propensity to stay at CNH. Similarly, non‐married Veterans in the benefit group that does not cover CNH, who were less frail but who had more ER visits, had higher propensity to stay at MFH.

In the MFH cohort, 84 subjects were dropped resulting in a final sample of 273. Dropped subjects were older (mean age 76.9 vs 70.4 years, P = 0.001), less often the Veteran benefit group for whom VHA covers CNH care (3.6 percent vs 34.8 percent, P < 0.001), and had lower mean annual income ($37 369 vs $47 950, P = 0.001). While a similar percentage were white (56.0 percent vs 58.1 percent), race was significantly different (P < 0.001) with fewer of the dropped subjects of black race (6.0 percent vs 19.6 percent), a higher percentage who were Hispanic (33.3 percent vs 3.7 percent) and fewer with an unknown race (3.6 percent vs 14.8 percent).

In the CNH cohort, 746 subjects were retained in the analysis and 947 dropped. Those who were dropped were older (mean age 76.03 vs 71.40 years, P < 0.001) with significant (P < 0.001) differences in marital status including a higher percentage married (70.0 percent vs 22.5 percent) and fewer divorced (12.0 vs 32.8), single (5.0 percent vs 16.2 percent), or widowed (9.8 percent vs 21.4 percent) among the dropped subjects. A higher percentage of dropped subjects were from the Veteran benefit group for whom VHA covers CNH care (79.8 vs 63.8, P < 0.001). Dropped subjects had lower Elixhauser comorbidity scores (2.38 vs 3.32, P < 0.001) and clinically similar but statistically significantly higher JEN Frailty Index scores (6.86 vs 6.29, P < 0.001). Race differed (P < 0.001) with more white subjects among the dropped Veterans (73.8 percent vs 61.8 percent) and fewer in the unknown race category (8.4 percent vs 16.6 percent) among dropped subjects.

The remaining patients who were not excluded were partitioned into 6 strata based on quantiles of the propensity scores. We then assessed covariates balanced between Veterans that utilized MFH to those who utilized CNH within and across strata. Balance of covariates across strata was assessed by testing the null hypothesis that the pseudo average difference in means between MFH and CNH Veterans for each covariate is equal to zero. Balance of the covariates within all strata was assessed using analysis of variance procedure to test the null hypothesis that the mean for the MFH Veterans is identical to the mean of CNH Veterans in each of the strata for each covariate.32 Imbalance of covariates was addressed by incorporating additional interactions, quadratic terms, and cubic terms to the initial propensity score model, and then, repeating the procedures above until the distribution of covariates was balanced across the CNH and MFH groups. An adequate balance was declared if the t‐statistics for all variables after the adjustment were <0.65. The standardized difference in means plot (Figure 2) displays absolute standardized differences for covariates means between MFH and CNH Veterans before (circles) and after propensity score subclassification (triangles).

Figure 2.

Standardized difference in means plot displays absolute standardized differences for covariates means between Medical Foster Home and Community Nursing Home Veterans before (circles) and after propensity score subclassification (triangles)

2.5.2. Imputation of missing potential outcomes

For each Veteran, we imputed their unobserved potential outcomes. Mortality at 6 and 12 months was modeled separately for Veterans in CNH and MFH using hierarchical logistic regression with cubic spline along the propensity score, linear adjustments of the covariates, and random intercepts for the different home facilities. Using these models, we multiply imputed the unobserved CNH mortality for Veterans residing in MFH and the unobserved MFH mortality for Veterans residing in CNH.

For costs expenditure categories, we estimated two separate models in the MFH and CNH groups, and using these models, we imputed the unobserved potential outcomes for each unit. Because some of the costs categories included many Veterans with $0 expenditures (eg, $0 for recreation therapy among CNH Veterans because this is included in CNH care and not billed to VA as a separate expense), we estimated costs using two‐stage models in each intervention group. In the first stage, we modeled whether a Veteran had any expenditure in that category using hierarchical logistic regression with cubic spline along the propensity score, linear adjustments for the other covariates, and random intercepts for the different home facilities. In the second stage, we relied on predictive mean matching (PMM) with a multilevel linear regression model of the logged outcome on a cubic spline along the propensity score, linear adjustments for the covariates, and random intercepts for the Veteran's home facilities.34 PMM imputes cost values using nearest‐neighbor donors with distance that is based on the predicted values from the aforementioned multilevel linear regression model of the logged outcome. This semi‐parametric method addresses the skewed distributions of the log‐transformed costs outcome.

The mortality and costs outcomes were imputed separately in the CNH and MFH groups, practically assuming that all the outcomes are independent given the covariates. The unobserved potential outcomes were imputed 600 times. We then calculated the mean difference between observed MFH outcome and the imputed CNH values as well as the standard deviation for mean difference for each of the 600 complete data sets and combined the results using Rubin's Rule for multiple imputation.35

3. RESULTS

Study population (Table 1): MFH Veterans were younger (71.93 vs 73.99, P < 0.005) with more Veterans younger than 65 years and fewer older than 85 years compared to the CNH cohort. Ethnicity differed (P < 0.0001) with more white CNH Veterans (68.5 percent vs 57.6 percent) and fewer Hispanic CNH Veterans (3.0 percent vs 10.7 percent). Marital status differed across all categories (P < 0.0001). The Veteran benefit group for whom VHA covers CNH care comprised 72.8 percent of the CNH Veterans compared to 27.4 percent of MFH Veterans. In contrast, 2.5 percent of CNH Veterans compared to 26.6 percent of MFH Veterans (P < 0.001) fell into the Veteran benefit group for whom VHA does not cover CNH care but who have a financial supplement as the result of catastrophic disability.

Table 1.

Characteristics of Veterans cared for in the Medical Foster Home Program compared to Veterans in community nursing homes during fiscal years 2010 through 2011 following matching procedures (N = 2047)

Characteristic	Community nursing home (CNH) (n = 1693)	Medical Foster Home (MFH) (n = 354)	P‐value (MFH/CNH)
Gender
Male	1645 (97.2)	341 (96.3)	0.503
Age
Mean	73.99 (12.42)	71.93 (13.06)	0.005
≤65	562 (33.2)	134 (37.9)	0.148
66‐70	149 (8.8)	40 (11.3)
71‐75	113 (6.7)	18 (5.1)
81‐85	238 (14.1)	48 (13.6)
>85	404 (23.9)	67 (18.9)
Ethnicity
White	1160 (68.5)	204 (57.6)	<0.0001
Black	247 (14.6)	58 (16.4)
Hispanic	50 (3.0)	38 (10.7)
Asian	16 (0.9)	7 (2.0)
Other	8 (0.5)	3 (0.8)
Native American	8 (0.5)	1 (0.3)
Unknown	204 (12.0)	43 (12.1)
Marital status
Married	830 (49.0)	46 (13.0)	<0.0001
Divorced	359 (21.2)	142 (40.1)
Widowed	253 (14.9)	79 (22.3)
Separated	37 (2.2)	13 (3.7)
Single	168 (9.9)	70 (19.8)
Mean length of stay (d)	207	282	<0.0001
Veterans Health Affairs (VHA) Benefit Group
VHA covers CNH care	1232 (72.8)	97 (27.4)	<0.001
VHA covers CNH care only if NH need is the result of a service‐connected disability	143 (8.4)	15 (4.2)
VHA does not cover CNH care but supplement added to pension as a result of disability	43 (2.5)	94 (26.6)
Missing	4.0 (0.2)	0 (0.0)
Other	271 (16.0)	148 (41.8)
Elixhauser comorbidity mean (standard deviation)	2.79 (2.59)	3.77 (2.67)	<0.001
JEN frailty index mean (standard deviation)	6.61 (2.29)	6.16 (2.33)	0.001
Predicted mortality (Care Assessment Need Score)
1 y	0.15 (0.16)	0.16 (0.17)	0.413
Annual income level
Mean (standard deviation)	51 429 (23 499)	45 439 (24 031)	<0.001
<10 000	0.08 (0.08)	0.12 (0.11)	<0.001
10 000‐14 999	0.06 (.0.06)	0.08 (0.07)	<0.001
15 000‐24 999	0.12 (0.08)	0.13 (0.08)	0.003
25 000‐34 999	0.11 (0.07)	0.11 (0.06)	0.997
35 000‐49 000	0.15 (0.07)	0.14 (0.07)	0.078
50 000‐74 999	0.19 (0.08)	0.17 (0.08)	<0.001
75 000‐99 999	0.11 (0.07)	0.10 (0.06)	<0.001
100 000‐149 999	0.11 (0.08)	0.09 (0.09)	0.001
150 000‐199 999	0.04 (0.04)	0.03 (0.04)	0.011
>199 999	0.03 (0.06)	0.03 (0.05)	0.312
Emergency room visits year prior to enrollment mean (standard deviation)	2.26 (2.73)	2.55 (3.26)	0.082

Comorbidity burden defined by the Elixhauser score was lower among CNH Veterans (2.79 vs 3.77, <0.001), and while the JEN Frailty Index was statistically different, the average differences would not be considered clinically meaningful (JEN Frailty Index 6.61 vs 6.16, P = 0.001). One‐year mortality prediction was similar in both groups (0.15 CNH vs 0.16 MFH, P = 0.413). Of note, the Pearson correlation between all three clinical measures across all Veterans was highest for CAN and JEN Frailty Index (0.28) and the lowest for Elixhauser and JEN Frailty Index (0.21), which indicates that these measures represent different clinical domains.

Significant differences were present between mean income at the 9‐digit zip code level ($51 428 CNH vs $45 439 MFH, P < 0.001) and the percentage of Veterans with income strata <$10 000 (8 percent vs 12 percent, P < 0.001), $10‐14 999 (6 percent vs 8 percent, P < 0.001), $50 000‐74 999 (19 vs 17, P < 0.001), and $75 000‐99 999 (11 percent vs 9 percent, P < 0.001). Both groups had a similar average number of ER visits (2.26 vs 2.55, P = 0.82) in the year prior to enrollment in the MFH and CNH programs. Mean length of stay was significantly lower among CNH Veterans at 207 days compared to 282 days for MFH Veterans (P < 0.0001).

Following propensity subclassification, cost comparisons (Table 2) demonstrated significantly higher costs among MFH Veterans for HBPC ($14 381, 95% CI [10 824, 17 938]) and mental health ($1902, 95% CI [223, 3581]) and lower costs for CNH care in the MFH Veterans (−$43 055, 95% CI [−48 340, −37 769]) resulting in lower total MFH costs of −$25 061 (95% CI [−33 635, −16 486]). The difference in costs per day alive was $71.28 less for MFH Veterans (95% CI [−107.24, −35.44]).

Table 2.

Adjusted cost differences of care in a medical foster home compared to a community nursing homes among propensity matched Veterans

Type of care	Adjusted cost medical foster home‐community nursing home		95% Confidence interval
	Mean difference	Standard error
Hospitalization	−$2449.43	$4759.12	−11 777.13	6878.27
Home based primary care	$14 381.58	$1814.98	10 824.27	17 938.89
Mental health	$1902.29	$856.47	223.64	3580.95
Recreation therapy	$1531.95	$1006.02	−439.83	3503.74
Home health	$1890.35	$1886.22	−1806.57	5587.28
Inpatient costs not including outpatient inpatient costs	−$530.05	$5709.91	−11 721.28	10 661.18
Therapy costs	$1159.66	$634.72	−84.37	2403.69
Recreation therapy costs	$1414.40	$905.15	−359.69	3188.49
Outpatient costs	$1405.32	$1149.96	−848.56	3659.20
Hospice costs	−$139.63	$1151.44	−2396.42	2117.17
VA (Community Living Center) Nursing Home	−$82.33	$2531.36	−5043.70	4879.05
Community Nursing Home	−$43 054.46	$2696.76	−48 340.02	−37 768.890
Skilled Nursing Home	−$727.83	$447.58	−1605.07	149.41
Total costsa	−$25 060.69	$4374.73	−33 635.00	−16 486.38
Costs per day alive	−$71.28	$18.29	−107.12	−35.44

^aTotal costs include Medicare (inpatient, outpatient, emergency room, skilled nursing facility, dialysis, and hospice), Veterans Health Administration, and costs paid by Veterans Health Administration to community providers. Costs not shown in table but included in total costs include outpatient social work encounters, outpatient chaplain encounters, dental services, outpatient day care, contract radiation therapy, contract Computerized Tomography examinations, outpatient research encounters, and telephone encounters. Nursing home costs paid by Medicaid are not included.

At 1 year, mortality was significantly higher than predicted for CNH Veterans compared to MFH Veterans (−0.12, 95% CI [−0.18, −0.05, ranging from −1 to 1]) corresponding to a 12 percent higher mortality rate among CNH Veterans. No statistically significant difference was observed between MFH and CNH Veterans at 6 months, but there was a trend to higher mortality for MFH Veterans (−.05, 95% CI [−0.12, 0.01]) (Table 3).

Table 3.

Six‐month and one‐year mortality differences in Medical Foster Home compared to Community Nursing Home Veterans

Mortality following program enrollment	Adjusted mortality difference Medical Foster Home‐Community Nursing Home	Standard error	95% Confidence interval
One yeara	−0.12	0.03	−0.18	−0.05
Six months	−0.05	0.03	−0.12	0.01

^aPrior to propensity score analysis, mortality difference was statistically significant −0.18 (SE 0.03) with 0.16 for Medical Foster Home and 0.34 for Community Nursing Home.

4. DISCUSSION

We examined costs in the year following enrollment into a CNH or a MFH among Veterans who required care for at least 60 days in a CNH or 30 days in a MFH. Costs were less for MFH Veterans per day alive than for similar CNH Veterans, and savings were derived primarily from lower residential care costs in the MFH despite significantly longer lengths of stay in MFH. While patient‐level out‐of‐pocket MFH and CNH costs were not available for inclusion in this analysis, MFH would be at least cost neutral if Veterans paid the monthly average of $2227 per month ($74/day), and had similar life expectancy given the finding of $71/day savings per day alive.

While the objective of this study was to compare costs of care for a cohort of Veterans in the MFH program to a similar group of Veterans receiving traditional LTC in a CNH, we found that the groups were quite different at baseline. MFHs are currently serving a unique cohort of Veterans who have lower incomes, higher comorbidity, more often lack a spouse but have similar levels of frailty and nearly identical predicted mortality compared to current CNH Veterans. It is also worth noting that while Veterans entitled to benefits covering the cost of CNH care would be expected to preferentially select CNH care, 27 percent who had this option chose instead to pay out‐of‐pocket for MFH. This suggests a demand for MFH as an alternative care option, and that shifts in selection of MFH instead of CNH may occur if Veterans with CNH coverage were offered MFH as an alternative.

In this analysis, we identified Veterans that used CNH and MFH and were similar in terms of observed clinical and demographic characteristics as a first step in understanding the implications of expanding VA payments to cover MFH for Veterans eligible for CNH payment. Using this cohort of similar Veterans, we estimated that costs for Veterans that used CNH are $71 higher per day alive compared to Veterans who resided in MFH. In addition, annual mortality is 12 percent higher in CNH compared to MFH.

Prior research that explored whether adult foster care is a suitable alternative for NH care concluded that frailty was similar among adult foster care and NH residents, but NH residents were more impaired physically and cognitively.12 In a subsequent study that examined adult foster care, investigators determined that for every additional foster care resident in a county, nursing homes in that county lost residents in an almost one‐to‐one substitution ratio.16 While these studies of substitutability were conducted in Oregon over 20 years ago, substitution does appear to be occurring. There was a decline in Oregon's nursing facility occupancy rate from 72 percent in 2000 to 64 percent in 2015, which represents the lowest nursing facility occupancy rate in the nation.36 These changes are not entirely accounted for by the adult foster care program as utilization of home and community‐based services and assisted living also explain declines in NH occupancy in Oregon. Based on the ability to propensity match a sufficient number of CNH and MFH Veterans yet also the need to drop many CNH Veterans because of many more CNH Veterans with benefits paying for CNH, our study suggests that substitutability exists at each site necessary to sustain a MFH program and incentives for MFH enrollment would likely exist if MFH payment were available to Veterans who already have CNH coverage. Additional research is needed to assess if these differences are the result of inherent Veteran characteristics that drive preference for MFH care, selective pressures of the health care system, and/or limitations placed on staff who are introducing the program to Veterans.

Mortality differences were an unexpected finding. Our objective in comparing mortality was to ensure that survival was not lower in the MFH where informal caregivers are caring for Veterans with complex care needs. The significantly higher mortality among CNH Veterans may be the result of unmeasured variables or true differences in outcomes resulting from care patterns. It is interesting to note that observed mortality for MFH was similar to the one predicted using the CAN score while observed CNH mortality exceeded these predicted values. Several studies that compared mortality rates among older persons with similar health status who remain at home to those who are relocated to a NH indicate that mortality is higher for those relocated to the NH.37, 38, 39, 40, 41 Some have proposed relocation as a stress characterized by disruption to routines and social connections while others have proposed new exposure to infectious agents and potential for injury in an unfamiliar environment as potential causes for higher mortality among those admitted to NHs. In this study, both groups of Veterans were relocated to new care settings. Prior work indicated that the odds of adverse care events and other unwanted outcomes among MFH Veterans were significantly lower while other outcomes were similar among Veterans in VA NHs.42 Importantly, none of the outcomes in prior work were worse among MFH Veterans, and favorable mortality rates for MFH Veterans in this study suggest that overall management patterns did not adversely impact survival.

Limitations of this analysis include omissions of costs that may impact our findings. First, the CNH sample includes Veterans who transitioned out of the CNH beyond the initial 60‐day inclusion requirement such that some CNH Veterans did not incur daily CNH residential care costs for every day they were studied. Another important limitation is that while these results were intended to represent costs to VA and Medicare, Veteran out‐of‐pocket payments for MFH and CNH care were not available for this analysis, so we do not know whether some of the cost burden may have shifted to Veterans in these programs. In addition, while the costs of monitoring caregivers were included, the costs associated with initial screening and recruitment of caregivers were not included.

A different limitation stems from the fact that this is an observational study in which the Veterans were not randomized to MFH or CNH. We have included many clinical and demographic characteristics and have only considered VHA systems that offer both MFH and CNH programs. These VAMCs were selected to reduce variation in practice and utilization between systems that offer both programs to those that offer only CNH. This variation across sites is hypothesized to contribute to large biases in the average differences between the outcomes. Additional bias may be introduced by unobserved covariates that influenced the Veteran's decision to reside in MFH or CNH. For example, functional status is highly correlated with survival and was not included in this analysis because of limited availability.43, 44, 45 However, we have included multiple measures of illness severity to capture clinical differences between the groups of Veterans utilizing MFH and CNH. Because these measures are highly correlated with functional status, we expect that functional status would also be similar across the two groups.

Another limitation is that this study was limited to Veterans who receive LTC which restrict our analysis only to this group of Veterans. Moreover, we restricted the analysis to Veterans who had positive propensity to enroll to either MFH or CNH in order to increase internal validity. This group does not represent the entire Veteran cohort that may use LTC, and thus, these findings may not be generalizable to populations not represented in this analysis.

Lastly, this study used a blunt measure of costs for all hospitalizations and future work should examine measures of hospitalization appropriateness46, 47 and long‐term outcomes. Medicaid payments for CNH care were not included in the analysis which may result in an underestimation of the total costs, particularly CNH costs for Medicaid eligible Veterans.

In conclusion, this is one of the first studies to provide cost comparison data for Veterans cared for in the VA MFH program compared to Veterans in traditional CNH care. Efforts to expand the MFH program appear warranted given that the program adds to the array of LTC options for persons in need of NH care while achieving cost savings and a mortality advantage when compared to traditional CNH care.

CONFLICTS OF INTERESTS

No conflicts of interests are identified.

Supporting information

 

Levy C, Whitfield EA, Gutman R. Medical foster home is less costly than traditional nursing home care. Health Serv Res. 2019;54:1346–1356. 10.1111/1475-6773.13195