Family Caregivers of Veterans Experience Clinically Significant Levels of Distress Pre- and During Pandemic: Implications for Caregiver Support Services

Abstract

Background.

Of the 26.4 million family caregivers in the U.S., nearly 40% report high levels of emotional strain and subjective burden. However, for the 5 million caregivers of Veterans, little is known about the experiences of caregivers of Veterans during the COVID-19 pandemic.

Objectives.

To examine pandemic-related changes of caregiver wellbeing outcomes.

Research Design, Subjects, and Measures.

Using a pre/post design and longitudinal data of individual caregivers captured pre- and during COVID-19, we use multilevel generalized linear mixed models to examine pandemic-related changes to caregiver well-being (n=903). The primary outcome measures include Zarit Subjective Burden, Center for Epidemiologic Studies Short Depression Scale, perceived financial strain, life chaos, and loneliness.

Results.

During the pandemic, we observe slight improvements for caregivers across well-being measures except for perceived financial strain. Prior to the pandemic, we observed that caregivers screened positive for clinically significant caregiver burden and probable depression. While we do not observe worsening indicators of caregiver well-being during the COVID-19 pandemic, the average predicted values of indicators of caregiver well-being remain clinically significant for caregiving subjective burden and depression.

Conclusions.

These findings illuminate pandemic-related impacts of caregivers receiving support through the VA pre- and during the COVID-19 pandemic while caring for a population of frail, older care-recipients with a high burden of mental illness and other chronic conditions. Considering the long-term impacts of the pandemic to increase morbidity and the expected increased demand for caregivers in an aging population, these consistently high levels of distress despite receiving support highlight the need for interventions and policy reform to systematically support caregivers more broadly.

1. Introduction

The Coronavirus 2019 (COVID-19) pandemic has been a collective global crisis with effects felt across health, social, and economic sectors in every country. The negative health impacts are well documented—as of December 2021, over 5.4 million people died and over 288 million contracted the virus with 30% of individuals diagnosed reporting persistent symptoms lasting at least 6 months.^1,2 In the United States (U.S.), the pandemic stifled economic growth due to stay-at-home orders for non-essential workers from March 2020 through May 2020 which negatively impacted mental health for almost 50% of Americans.³ While economic growth improved by July 2021 as vaccinations were made widely available to those 12 and older, Americans living alone (~28%) and households including people with pre-existing and complex conditions remained isolated for long periods of time.⁴

The COVID-19 pandemic illuminated how poor public policy supports for vulnerable groups, such as family caregivers, compound disparities in negative health and economic outcomes.^5,6 Family caregivers provide care for aging, ill, or disabled family members or friends without pay or training. In the U.S., family caregivers provide most long-term services and supports, e.g., 75% of Americans aged 65 or older with a disability receive help from family caregivers.⁷ Yet family caregivers operate without pay, benefits, or retirement assistance because they are not recognized as a formal profession. The COVID-19 pandemic exacerbated the reliance on family caregivers as “home” became the de facto setting for managing the long-term and medical needs of older adults and adults with chronic illness and functional impairments.⁸ During this time, caregiving responsibilities increased and caregivers often cared for their medically vulnerable relatives with reduced access to home and community-based services (HCBS) to minimize risk of exposure to COVID-19.⁹ Reduced access to HCBS is one example of how poor public policy supports may have inadvertently engendered financial hardship for these families as during the pandemic, some caregivers, of whom 61% in the U.S. are employed, reduced their work hours or left work completely to care for their loved one.^10–12

The cadre of 26.4 million family caregivers in the U.S. already experiences high rates of mental health burden—nearly 40% of U.S. family caregivers report high levels of emotional strain and subjective burden.^10,13 Evidence from quasi-experimental studies shows that caregiving increases depression and anxiety in caregivers compared to non-caregivers and that mental health effects can be long-lasting.^14–25 Recent studies conducted worldwide suggest that pandemic-induced lockdowns may have exacerbated caregiver mental health symptoms.^12,26–29 One US-based study (n=77) reported that caregiver subjective burden increased during the pandemic lockdown.²⁶ In a cross-sectional survey of 422 US-based caregivers conducted from June 2020 to August 2020, 83% reported increased caregiving-related stress. Another cross-sectional study from Italy found that 87% of caregivers reported at least one stress-related symptom and nearly 30% reported 4+ symptoms; symptoms were exacerbated by loss of services and supports.²⁷ Across studies, caregivers consistently reported that their primary worries were that they or their loved one, who may be at higher risk, would contract the virus.^{9,12,27,29,31}

The 5+ million caregivers of military Veterans may have had unique COVID-19 pandemic-related experiences.^32,33 Due to military service-related injuries, these caregivers are often providing care for loved ones with higher rates of mental illness and musculoskeletal conditions than civilian caregivers which may be associated with different pandemic impacts.³² Furthermore, the U.S. Department of Veterans Affairs (VA) offers one of the most comprehensive and widely available caregiver support programs and these programs remained operational during the pandemic as they are based at medical centers. The Program of General Caregiver Support Services (PGCSS) provides caregivers of Veterans enrolled in VA health care access to training, peer-support, respite care, training, and caregiver support coordinators at each VA medical center to help caregivers navigate the VA.³³ Despite these potential differences, pandemic-related changes in the wellbeing of military caregivers have not been documented. The needs of caregivers of Veterans could have important implications for VA and non-VA health systems. Pandemic-related exacerbations of poor outcomes could be a harbinger of future costs to public health systems through health care utilization and long-term mental health sequela. On the other hand, available caregiver supports and expansion of telehealth services^30,34,35 might have decreased caregiver burden and increased access to care for care-recipients thereby providing support for interventions that could alleviate pandemic-related effects.

Using telephone-based survey data collected annually from 2017 through 2020 from caregivers of Veterans, we examine pandemic-related changes for measures of caregiver wellbeing outcomes: subjective burden, depressive symptoms, financial strain, life chaos, and loneliness. First, we describe trends in caregiver outcomes from surveys completed March 1–August 31, 2018; March 1–August 31, 2019; and March 1–August 31, 2020. Second, using multilevel mixed-effects generalized linear models, we compare caregiver wellbeing outcomes during the pandemic to caregiver wellbeing outcomes one and two years prior to the pandemic. We hypothesized that indicators of caregiver wellbeing would be worse during the COVID-19 pandemic compared to the pre-COVID-19 periods. We also examined the association of caregiver/care-recipient characteristics and caregiver outcomes of well-being across years.

Current evidence provides an important foundation to understand the negative impacts of the COVID-19 on the emotional wellbeing of family caregivers. Yet, these studies often rely on self-reported perceived changes which may be subject to recall bias.^{12 36,37} Our pre/post view of 903 unique caregivers will strengthen the inferences of our results by allowing us to examine wellbeing indicators of caregivers using data collected before (historical control) and during the pandemic. Identifying factors associated with worse outcomes during the pandemic could inform outreach efforts to further support those caregivers at risk for worse outcomes.

2. Methods

2.1. Sample and Data

The analysis uses data collected through a longitudinal national survey of caregivers of Veterans. A sample of caregivers of Veterans enrolled in VA PGCSS from October 2016 through July 2018 was identified through the VA Caregiver Support Program’s Caregiver Application Tracker. To qualify, caregivers must care for a Veterans with demonstrated functional limitations with activities of daily living (ADLs) or instrumental activities of daily living (IADLs) ≥6 months.³³ While not a requirement, PGCSS predominately serves caregivers of Veterans living in the community who served pre-9/11.³³ Between October 2017 through September 2020, evaluators invite caregivers to participate annually in a survey conducted over the phone about their experiences as a caregiver for a Veteran. If enrolled, then caregivers are contacted and surveyed once a year for up to four years. To ensure geographic representation, caregiver recruitment was stratified by VA Medical Center. See Shepherd-Banigan et. al (2020) for detailed recruitment procedures.³³

Caregivers completed the initial baseline survey at different times, and therefore caregivers in our sample have varying numbers of follow-up periods. For example, caregivers completing the baseline survey in 2018 could contribute up to three periods of data (2018, 2019, and 2020). In contrast, caregivers recruited in June 2020 may only contribute one period of data for this analysis window. A total of 1,518 surveys were completed during March-August 2018, March-August 2019, and March-August 2020 with 976 unique caregivers completing one, two or three surveys during any of the three time periods. We excluded 73 respondents as they did not have complete covariate data and/or missing all outcome measures data. Our final analytical sample is 903 unique caregivers who contributed data at least one of the three time periods with complete data for covariates for adjustment and contributed at least one complete outcome measure. Exactly 206 caregivers completed at least one survey pre-pandemic and one survey during the pandemic. See Table 1 for the pattern of completed surveys by number of unique caregivers.

Table 1.

Pattern of Surveys completed by caregivers

Number of Caregivers	March-August 2018	March-August 2019	March-August 2020
379	X
173	X	X
141	X	X	X
103			X
56	X		X
42		X
9		X	X
Total N = 903
Number of Caregivers	Baseline	12-month Follow-up	24-month Follow-up
486	X
178	X	X
141	X	X	X
57	X		X
30		X
8			X
3		X	X
Total = 903

NOTES: 12-month follow-up surveys with respondents must have been completed within a 6–8 week period one year after the baseline survey. For example, if a respondent completed the baseline survey on February 12, 2018, and completed the 12-month follow-up survey on March 3, 2019. Thus, the outcomes from the 12-month follow-up survey, but not baseline, are retained in the analysis. Thus, our analytic sample may include respondents with a 12-month and/or 24-month follow-up survey, but not a baseline survey.

2.2. Measures

We conduct secondary analysis of existing survey data, thus we examine measures selected by the VA Caregiver Support Program Partnered Evaluation Center as informed by adopting the Organizing Framework for Caregiver Interventions.^33,38,39

2.2.a. Outcome Measures

Caregiver Subjective Burden.

The Zarit Caregiver Burden instrument measures caregiver subjective burden as a twelve-item scale indicating frequency of level of stress experienced.^40,41 Potential responses include “Never”, “Rarely”, “Sometimes”, “Quite Frequently”, or “Nearly Always”. Questions cover the elements often mentioned by caregivers as problems, including health, psychological well-being, finances, social life, and the relationship shared by the caregiver and care recipient. Scores range from 0 to 48, where higher scores indicate higher burden and “a score >16 suggests clinically significant caregiver burden” with a Cronbach’s alpha of 0.89 in our sample.⁴⁰

Caregiver Depressive Symptoms.

The Center for Epidemiologic Studies Depression 10-item Scale (CESD-10) captures caregiver depressive symptoms.⁴² Potential responses include “Never”, “Rarely”, “Sometimes”, or “Often” regarding statements of frequency of depressive symptoms experienced. Each response is scored for a range of 0–3 for a total score range of 0–30, where higher scores indicate greater depressive symptoms. Depending on the use of the CESD-10, a score of ≥8 or ≥10 is often used to indicate screening positive for depressive symptoms and probable depression, respectively.⁴² We calculated a Cronbach’s alpha of 0.85 in the analytic sample.

Caregiver Perceived Financial Strain

Perceived financial strain is assessed through the three item Impact on Finances subscale from the Caregiver Reaction Assessment in which respondents indicate how much they agree with degree of financial strain experienced.⁴³ Potential responses include “Strongly Disagree”, “Disagree”, “Neither Agree nor Disagree”, “Agree” or “Strongly Agree” resulting in scores ranging from 3 to 15, where a higher score indicates higher strain with a Cronbach’s alpha of 0.80 in the analytic sample.

Life Chaos.

Life chaos is ascertained with the six-item survey validated by Zullig et al. in which responses include “Definitely True”, “Somewhat True”, “Unsure”, “Somewhat False”, “Definitely False” or “Prefer Not to Answer”.⁴⁴ Higher scores indicate greater life chaos. We calculated a Cronbach’s alpha of 0.74 in the analytic sample.

Loneliness.

Loneliness is captured with the three-item survey validated for surveys administered by phone by Hughes et al. where a higher score indicates greater loneliness.⁴⁵ We calculated a Cronbach’s alpha of 0.85 in the analytic sample.

2.2.b. Key Explanatory Variables.

Indicator variables capture whether the survey was completed during the pandemic (March 1-August 30, 2020), the year prior to the pandemic (March 1-August 30, 2019), or 2 years prior to the pandemic (March 1-August 30, 2018). The referent category in the models is March 1-August 30, 2019.

2.2.c. Covariates for Adjustment.

We measure caregiver and care-recipient characteristics assessed at the first survey a caregiver completed. These factors included caregiver age, gender, relationship to Veteran (spouse, child, parent, other), marital status, living distance from Veteran, highest level of education, race, ethnicity, Veteran status of caregiver, self-reported health status, health insurance coverage, living in a rural area, and caregiver self-reported work status as employed (full- or part-time) or unemployed (not employed, retired, disabled or student). Veteran factors included the Activities of Daily Living rating (excellent/good, mild impairment, moderate impairment, severe impediment, or total impairment).

2.3. Statistical Analysis

As the sample includes caregivers with varying numbers of completed surveys, we have imbalanced panel data. Thus, we use multivariable, multilevel generalized linear mixed models to examine caregiver well-being before and during the pandemic as the model uses covariates and observed outcome measures to account for missingness of outcome measures data in survey time periods (March 1-August 2020; March 1-August 2019; March 1-August 2018). We include a caregiver random effect to account for correlation across observations that are repeated over survey time periods (see Appendix A for equation). We are primarily interested in changes in the outcome measures from March-Aug 2020 (hereafter referred to as the COVID-19 period) compared to prior years. We analyze trends over time using 2019 as the reference period. We identified model distributions and links as informed by modified Park’s test and Pregibon’s link test. We used a Gaussian distribution with an identity link to model caregiver subjective burden, depressive symptoms, and life chaos. We used a gamma distribution with log link to model caregiver perceived financial strain and loneliness. We calculated model estimated mean outcome values and marginal effects adjusted for covariates as defined in section 2.2.c. All covariates were examined for multicollinearity and no correlations >0.75 were observed. Analyses were conducted using the MEGLM command in StataMP 16.

To assess differential missingness patterns, we compared descriptive statistics and outcomes of two groups: (1) the most recent pre-COVID-19 period survey by caregivers who did complete a follow-up survey during COVID-19 period (i.e., contributed at least one pre-COVID survey and during-COVID survey) and (2) the most recent pre-COVID-19 period surveys by caregivers who did not complete follow-up survey during the COVID-19 period (i.e., only completed pre-COVID survey). Specifically, we examined balance of absolute standardized differences of the outcomes and baseline demographics between the two groups. A common threshold for balance of standardized differences is 20% with a preferred threshold of 10%.⁴⁶

As a sensitivity analysis, we also conducted an individual-level fixed effects regression analysis on responses of caregivers who responded to at least one pre-COVID-19 survey and one during the COVID-19 pandemic.

2.4. Ethical Considerations

This non-research evaluation was conducted under the authority of the CSP and Quality Enhancement Research Initiative (QUERI) and thus is classified as quality improvement. VHA Handbook 1058_05 (Veterans Health Administration 2011) provides guidance about authorization of manuscripts that have been developed through non-research activities (i.e., without institutional review board approval under the authority of VHA operations). All VHA authors of this article attest that the activities that resulted in producing this manuscript were conducted as part of the non-research evaluation conducted under the authority of the CSP and QUERI. Caregiver responses were kept confidential to the researchers and anonymous to the operational partners in the VA CSP.

3. Results

3.1. Descriptive Results

Caregivers are on average 62 years old and 74% white, 15% Black, 7% Hispanic/Latino(a), mostly female (96%), mostly the spouse of the care recipient (83%), 88% co-reside with the Veteran, and 32% have a college degree or higher (Table 2). Approximately 6% of caregivers are themselves Veterans and 6% of caregivers are uninsured. Thirty-four percent of caregivers self-reported a health status of Fair or Poor health. The Veteran’s health status as reported by the caregiver indicated substantial impairments, on average. Approximately 13%, 16% and 63% reported moderate, severe, and total impairment in ADLS and IADLS, respectively.

Table 2.

Descriptive Statistics of Caregivers at Baseline

	Percent (n=903)
Caregiver Demographics
Age, mean (SD, min-max)	61.66 (13.67, 20–91)
Race
Only Asian, American Indian, or Pacific Islander marked	2.44%
Only Black marked	14.84%
Only white marked	73.64%
Only other marked	4.87%
Multiple races marked	4.21%
Hispanic/Latino(a)	7.31%
Gender
Female	96.01%
Male	3.99%
Relationship of CG to Veteran
Wife/Husband or Significant Other	82.84%
Mother/Father	1.55%
Daughter/Son	10.74%
Other	4.87%
Marital Status
Married	86.60%
Living together	1.22%
Widowed	1.55%
Divorced/Separated	5.74%
Single, Never Married	4.87%
Distance from CG to Veteran
In the same house	87.60%
Within walking distance	1.44%
Less than 20 minutes	6.31%
More than 20 minutes	4.65%
Highest Level of Education
High school graduate/GED or less	23.59%
Trade/Technical/Vocational School, some college credit, Associates degree	44.41%
Bachelor’s degree	18.16%
Postgraduate work or graduate degree	13.84%
Insurance Status
No health insurance for myself	5.76%
Public insurance	47.18%
Private insurance	26.02%
Both public and private	21.04%
Caregiver is a Veteran	6.31%
Caregiver Self-Reported Employment
Working full-time (36+ hours per week)	12.29%
Working part-time	10.74%
Not working, searching for paid work	1.88%
Not working, not searching for paid work	22.81%
Retired	45.29%
Disabled	5.54%
Student	1.44%
Years spent as a caregiver at baseline, mean (SD, min-max)	7.02 (7.69, 0.125–50)
Caregiver Self-reported Health Status
Excellent	6.76%
Very good	20.82%
Good	38.65%
Fair	27.46%
Poor	6.31%
Veteran Health Status
Number of ADLs/IADLs requiring assistance	8.50 (3.50, 0–13)
ADL Rating
Excellent - Good	1.22%
Mild Impairment	6.31%
Moderate Impairment	13.07%
Severe Impairment	15.95%
Total impairment	63.47%

Table 3 presents the average, unadjusted outcome measures of respondents during each time period. We observe across all time periods high levels of caregiver subjective burden, depressive symptoms, financial strain, life chaos, and loneliness. The average Zarit and CESD-10 scores decreases slightly over time, but remain, on average, higher than the clinical threshold indicating clinically significant burden and probable depression, respectively. We observe similar patterns for perceived financial strain, life chaos, and loneliness scores – high rates of strain, life chaos and loneliness, with slightly lower scores during the pandemic. Among the subset of caregivers who contributed survey data to all three periods we observe similar patterns across all outcome measures (Appendix B, Table B1). Finally, when asked if the COVID-19 pandemic impacted their responses to the survey, 45% of caregivers self-reported no impact.

Table 3.

Unadjusted Descriptive Statistics of Measures of Caregiver Well-being

	March – August 2018	March – August 2019	March – August 2020	Total Observations
Zarit Burden Score, mean (SD, N)	21.85 (9.50, 723)	21.20 (9.26, 349)	17.92 (9.71, 300) *	1,372
CESD-10 Score, mean (SD, N)	11.70 (7.10, 741)	11.29 (6.87, 358)	10.12 (6.54, 306) *	1,405
Perceived Financial Strain Score, mean (SD, N)	9.19 (3.69, 735)	8.47 (3.95, 360) *	8.16 (3.53, 301) *	1,396
Life Chaos Score, mean (SD, N)	16.91 (4.86, 732)	16.76 (5.05, 362)	15.93 (4.54, 305) *	1,399
Loneliness Score, mean (SD, N)	6.12 (2.11, 727)	6.04 (2.12, 354)	5.68 (2.15, 305) *	1,386

Notes: A higher score indicates a worse outcome, e.g., greater depressive symptoms. As not all respondents had a score for each outcome (e.g., declining to answer enough questions of the instrument to generate a valid score), the number of observations per outcome varied. The total number of observations was 1,372 (n=882 unique caregivers) for the Zarit Subjective Burden; 1,405 (n=894 unique caregivers) for the CESD-10 Score; 1,396 (n=892 unique caregivers) for the Perceived Financial Strain; 1,399 (n=891 unique caregivers) for Life Chaos; and 1,386 (n=855 unique caregivers) for the Loneliness score. Clinically significant thresholds are greater than or equal to 16 for Zarit burden and greater than or equal to 8 or 10 for CESD-10.

^*Indicates statistically significant differences (p<0.05) as indicated via regressing indicators of time periods on each outcome clustered by individual where March-August 2018 is the referent period.

3.2. Examination of Systematic Differences between Respondents Based on Number of Responses

We find little evidence of systematic differences in outcome measures based on number of responses, see Table 4. All outcome measures and covariates have standardized mean differences <20%. All outcomes except Perceived Financial Strain have a standardized difference <10%.

Table 4.

Comparison of Outcomes in the most recent pre-COVID period by those who completed follow-up survey during COVID compared to those who did not complete follow-up survey during COVID

	Caregivers who did not complete survey during COVID	Caregivers who did complete survey during COVID	Absolute Value Standardized Mean Difference
Outcomes from most recent pre-COVID Survey
Zarit Burden Score, mean (SD, range)	21.91 (9.59, 0–47)	21.37 (9.60, 0–44)	5.57%
CESD-10 Score, mean (SD, range)	11.77 (7.06, 0–30)	11.64 (7.31, 0–30)	1.91%
Perceived Financial Strain Score, mean (SD, range)	9.05 (3.83, 3–15)	8.55 (3.80, 3–15)	13.29%
Life Chaos Score, mean (SD, range)	16.90 (4.68, 6–30)	16.75 (5.47, 6–30)	3.01%
Loneliness Score, mean (SD, range)	6.10 (2.10, 3–9)	6.09 (2.14, 3–9)	0.49%

Notes: Clinically significant thresholds are greater than or equal to 16 for Zarit burden and greater than or equal to 8 or 10 for CESD-10.

3.3. Association of COVID-19 with Caregiver Outcomes

After adjusting for caregiver characteristics and care-recipient baseline functioning, we observe consistently high levels of burden, depression, perceived financial strain, life chaos and loneliness across all time periods, see Figure 1 for model estimated mean outcome measure values. See Appendix C for full model output. Compared to individuals included in the March-August 2019 period, individuals during March-August 2020 had a 2.87 unit decrease in Zarit score (95% Confidence Interval (CI), [−3.83, −1.90]), 1.29 unit decrease in CESD-10 score (95% CI [−2.03, −0.56]), 0.65 unit decrease in Life Chaos score (95% CI [−1.18, −0.13]), and 0.33 unit decrease in Loneliness score (95% CI [−0.58, −0.09]). We observe no evidence of a statistically significant association with perceived financial strain (−0.15, 95%CI [−0.63, 0.33]). As higher scores for each measure suggest worse outcomes, a decrease in the score suggests improvements across measures. Importantly, during the March-August 2020 period, model estimated mean caregiver CESD-10 scores and Zarit Subjective Burden scores remain higher than clinical thresholds of significance (≥10 and ≥16, respectively).

Figure 1.

Model Estimated Mean Outcome Values with 95% Confidence Intervals

Notes: A higher score indicates a worse outcome, e.g., greater depressive symptoms. As not all respondents had a score for each outcome (e.g., declining to answer enough questions of the instrument to generate a valid score), the number of observations retained in each regression varied. The total number of observations was 1,372 (n=882 unique caregivers) for the Zarit Subjective Burden; 1,405 (n=894 unique caregivers) for the CESD-10 Score; 1,396 (n=892 unique caregivers) for the Perceived Financial Strain; 1,399 (n=891 unique caregivers) for Life Chaos; and 1,386 (n=855 unique caregivers) for the Loneliness score. Clinically significant thresholds are greater than or equal to 16 for Zarit burden and greater than or equal to 8 or 10 for CESD-10. Notably, the scale for the Zarit burden is already greater than the clinical thresholds of 16 and CESD-10 scores for all time-points are greater than the clinical threshold of 10.

The sensitivity analysis results were consistent with the primary results (Appendix D).

4. Discussion

Using longitudinal data of individual caregivers of Veterans captured pre- and during COVID-19, we observe slight improvements for caregivers across well-being measures except for perceived financial strain. Prior to the pandemic, we observed that caregivers on average screened positive for clinically significant caregiver burden and probable depression. While we do not observe worsening indicators of caregiver well-being during the COVID-19 pandemic, the average predicted values of indicators of caregiver well-being remain clinically significant for caregiving subjective burden and depression.

Our findings may be explained through several mechanisms, which may explain some of the discrepancies with existing evidence. First, as we observe high scores (indicating worse outcomes) across all measures over time, it is possible that we observe ceiling effects in this sample. Our sample consists of caregivers of Veterans with high functional impairments, as measured by ADLs and IADLs, with over 63% of the sample reporting providing care to an individual with total impairment at baseline. Second, few caregivers in this sample reported working prior to the pandemic and thus may have been less impacted by pandemic-related financial shocks. Third, lockdowns, may not have negatively impacted caregivers in our sample as they are primarily co-residing, spousal caregivers, unlike caregivers of care-recipients in congregate residential facilities. Finally, it is possible caregivers surveyed have greater supports than caregivers otherwise represented in the literature as they are enrolled in PGCSS. For example, evidence suggests approximately 50% of PGCSS caregivers had contacted a local VA Caregiver Support Coordinator.⁴⁷ While the evaluation is not designed to assess the effect of PGCSS itself, our results may differ from existing literature due to the support provided by PGCSS. Future work should evaluate whether such an association exists. Prior non-VA qualitative analyses have identified lack of services and supports for caregivers and the need to strengthen HCBS.⁴⁷ Moreover, our findings are consistent with caregivers in the analytic sample who were asked to report the impact of the COVID-19 pandemic on their outcomes, with 45% reporting no impact.

This analysis is subject to several limitations. First, while our analytic approach is relatively robust to missing data, if missingness is driven by unobserved characteristics, then results may be biased. For example, if caregivers experiencing the most pandemic-induced increase in strain dropped out of the survey, then survey results in 2020 may underestimate or show false improvements. However, sensitivity analyses did not indicate that there was a difference in the baseline characteristics of caregivers who did and did not participate in 2020. Second, results may have limited generalizability to caregivers of civilians given that surveyed caregivers were identified from those registered for VA caregiver services. However, the cohort’s demographics and functional limitations are similar to those of caregivers of Veterans who served pre-9/11 and civilian caregivers.³² Thus, results may be relevant for caregivers of civilians with high levels of functional impairments. Third, we do not capture long-term effects of the pandemic. Fourth, we only adjust for caregiver and Veteran characteristics collected at the time of the first survey. Finally, the survey sample included a high proportion of married caregivers who live together with low employment pre-pandemic; thus, results may have limited generalizability to the experiences of working and younger caregivers not co-residing with care-recipients.

While our analysis examines subjective caregiver burden, depressive symptoms, loneliness, financial strain, and life chaos, we are unable to speak to the positive aspects of caregiving. Existing research suggests caregiver resilience may moderate how caregivers experience the pandemic.⁴⁸ Specifically, some caregivers experienced positive aspects of caregiving during the COVID-19 pandemic, e.g., forging deeper connections between caregivers and care-recipients.^29,49

Our findings are a unique contribution to the literature, given the longitudinal data on caregivers, allowing us to observe caregiver outcome measures from validated instruments over time. Furthermore, we assess pandemic-related effects for a sample with higher-than-average access to supports and services through PGCSS, e.g., skills development/training resources. These findings further illuminate the experiences of caregivers for a population of frail, older care-recipients with a high burden of mental illness, such as posttraumatic stress disorder, and high rates of other chronic conditions who are receiving caregiver supports prior to and during the COVID-19 pandemic.³³ Despite these findings, we also document persistently low levels of well-being among caregivers of Veterans that underscores the need for interventions to ensure caregiver long-term well-being and high quality of care to military Veterans. More broadly, we observe high distress among caregivers receiving systematic support which raises concerns for caregivers of frail, older, civilian care-recipients who may not have access to supports. Considering the long-term impacts of the pandemic⁵⁰ and expected increased demand for caregivers in an aging population, these high levels of lasting distress highlight the need for interventions and policy reform to systematically support caregivers more broadly.