Introduction

The term “caregiving” may not have much significance to the average person unless they have actually been or known a caregiver (CG) or care recipient (CR). However, there are over 44 million American CGs, and over 17 million of them are caregiving for someone over the age of 64.1,2 Caregiving generally refers to the informal and unpaid work performed by a CG to assist a CR. In the United States, in 2013, the economic value of this work was estimated to be $470 billion.3 Caregiving may include but is not limited to emotional support, physical assistance with personal care and household tasks, and more recently, financial support, home medical care and management, decision making, and the navigation of complex medical systems such as hospitals and insurance companies.1,2

It is the eldest in society, who are likely to have the most significant physical and cognitive limitations, and thus, who are the most in need of caregiving.2 As the U.S. population ages, it is likely that the informal and unpaid workload of caregiving will also increase. In 2010, there were 40.2 million people over the age of 65 living in the United States.4 The youngest of the baby boom generation (people born between 1946 and 1964) will be in their mid-60s by 2030—that is roughly 20% of the country’s population over the age of 64.4,5 By 2050, that number is projected to more than double to 88.5 million.4 The eldest of the elders, those over 85 years old, will represent 2.3% of the population in 2030, and are projected to nearly double by 2050.4 Although this generation will also be facing mortality declines by and after 2050, their presence will be an unprecedented increase in the total number of older adults living within the U.S. population.

It is not only the CRs who are aging but also the CGs. This study explored the specific relationship of elder CGs who are 64 years old and older that care for elder CRs who are also 64 years old and older. This is referred to herein as elder-elder (E-E) caregiving. The primary aim of this study was to evaluate trends in the E-E caregiving population from 1997 to 2014. Secondary aims included describing demographics, relationships and self-described burden on these older caregivers.

Methods

This paper used the National Alliance for Caregiving’s open data sets.6,7 The following two data sets were explored: (1) Caregiving in the U.S. 2015 and (2) Caregiving in the U.S. 2009. The 2015 files contained surveys completed in 2014, and from here on, are referred to by that year. The data set contained within Caregiving in the U.S. 2009 also included raw data from the years 1997 and 2004. Inclusion of these years was not indicated in the file title, and thus, not expected. The 2009 raw data was separated by year into separate files for statistical analysis.

Both data sets included codebooks explaining the data’s nomenclature. The 2009 files included the final survey questions in English and Spanish, and a report of their methodology. All four years of data were gathered from a random sample of CGs via telephone and online surveys. All datasets were determined to be deidentified and deemed nonhuman subjects research by the Geisinger Institutional Review Board.

Over 200 variables were collected through the survey questions. Variables analyzed in this report included age and sex for both CG and CR; the length of time the CG had been caregiving; the relationship of the CR to the CG; the CG’s health status; and the level of burden experienced by the CG due to caregiving.

Data exclusion criteria included any incomplete surveys. Additionally, survey responses were further whittled to the population of interest to include only CGs aged 64 and over who care for CRs aged 64 and over. Survey respondents who withheld or did not know their own age or the age of their CR were also excluded.

Data analysis

IBM’s SPSS Statistics Version 26 was used to organize the data, and Prism 8 and 9 were used to perform the statistical tests and create the figures. Summary statistics for age, sex, length of time caregiving, relationship of CR to CG, and health status of CG were calculated for each data set by year. The ROUT outlier test was used to look for outliers among CG and CR ages respectively. Inclusion and exclusion of outlier data is explicitly stated within the results of each test.

The one-way Brown-Forsythe and Welch ANOVA tests (alpha=0.05) compared the mean E-E CG and CR ages in 1997, 2004, 2009, and 2014 respectively. Parameters for these statistical tests assumed normal Gaussian distribution and unequal standard deviation for each group. The post-hoc test used was the Games-Hollow multiple comparison test (alpha=0.05), which was chosen because the sample sizes were over 50.

A two-tailed Pearson Correlation matrix was performed to see if there was an association between CG age and the level of burden experienced during any one year. A second Pearson Correlation matrix was performed to explore any association between CG age and self-reported health status during any given year.

Results

The resulting E-E caregiving sample was n = 147, 115, 282, and 445 for years 1997, 2004, 2009, and 2014 respectively. The mean age of E-E CGs was 71, 70, 71, and 74 years old for years 1997, 2004, 2009, and 2014 respectively. The largest increase in mean age of E-E CGs was from 70 years old in 2004 to 74 years old in 2014. The median, minimum, and maximum ages for both E-E CGs and E-E CRs from 1997 to 2014 (with outliers) is shown in Figure 1. However, the median age of CRs showed minimal variation during the 17-year time frame (Figure 1).

Figure 1.Median age of E-E CG and CR.

Summary information for E-E CGs (left) and E-E CRs (right) including median and min-max ages in 1997, 2004 (with outliers), 2009, and 2014.

In 2004, the maximum age for CGs (99 years) was older than the maximum age for CRs (97 years). Analysis found five outliers within the 2004 E-E CG data, including the two CGs aged 99 years old. No other year had outliers, nor were there outliers in the CR data. The maximum age of CRs (107 years old) occurred in 2014, the most recent year of data collection.

Both the Brown-Forsythe ANOVA (F(3,721.9) = 24.09, p <0.0001) and Welch ANOVA (W(3, 357.8) = 22.70, p <0.0001) tests found a significant statistical difference in the mean age of E-E CGs (outliers removed) between at least two groups. The post-hoc test found statistically significant differences for the mean age of E-E CGs for each year when compared to 2014 (Table 1). The ANOVA tests found no statistical difference between the mean age of E-E CRs; thus, no follow-up testing was done on that group.

Table 1.Results of the post-hoc Games-Howell multiple comparison test comparing mean ages (in years) of E-E CRs in 1997, 2004 (without outliers), 2009, and 2014.
Mean
Difference
Adjusted
P Value
1997 vs 2004 0.5416 0.8480
1997 vs 2009 -0.5576 0.7652
1997 vs 2014 -3.237**** <0.0001
2004 vs 2009 -1.099 0.3075
2004 vs 2014 -3.778**** <0.0001
2009 vs 2014 -2.679**** <0.0001

**** Indicates a statistically significant p-value.
 

The majority of E-E CGs each year were female, with an uptick in male CGs in 2014, as shown in Table 2. CRs were also predominantly female with the most equal sex division being in 2004. The relationship between the E-E CG and the CR was most often spouse, mother, or friend/non-relative/neighbor in any given year (Figure 2). The identifier shown (e.g., mother) refers to the relationship from the point of view of the CR (e.g., I care for my mother.)

Table 2.Breakdown of E-E CG and CR Sexes in 1997, 2004, 2009, and 2014.
CG Sex CR Sex
1997 M: 24%
F: 76%
M: 25%
F: 36%
UK: 38%
2004 M: 37%
F: 63%
M:40%
F: 55%
DK: 5%
2009 M: 32%
F: 68%
M: 33%
F: 67%
2014 M: 43%
F: 57%
M: 35%
F: 65%

M = Male, F = Female, UK = Unknown, DK = Don’t know.
 

Figure 2.Relationships of E-E CGs to their E-E CRs for 1997, 2004, 2009, and 2014.

The identifying relationship was made by the CR. In 2014, six relationships (and Other) are shown because “Brother” and “Sister” were each 3%. Note: Values may not add to 100% due to rounding.

The average length of time for a person to be an E-E CG was approximately five years with a standard deviation ranging from approximately five to ten years, which remained constant over time (Figure 3). The Pearson Correlation matrix found no correlation between E-E CG age and the level of burden experienced, with r = 0.04, 0.112, 0.014, and 0.022 for 1997, 2004, 2009, and 2014 respectively. Relatedly, when asked, "How has caregiving affected your health status?" the vast majority of E-E CGs responded that caregiving had not affected their health (Figure 4). However, from 2004 to 2014, there was a slight uptick in those reporting that caregiving made their health “worse” from 19% to 22%.

Figure 3.Mean length of time in years (with standard deviation) that E-E CGs report caregiving in 1997, 2004, 2009 (without outliers), and 2014.
Figure 4.E-E CG responses (without outliers) regarding the impact of caregiving on self-reported Health Status for 2004, 2009, and 2014.

This variable was not collected in the 1997 survey. Values may not add to 100% due to rounding.

Discussion

These results reflect that the E-E CG population is aging, mirroring U.S population trends, with the mean age of CGs increasing from 70 to 74 years old. While the traditional view of E-E caregiving might be of two similarly aged spouses caring for each other that was not the only relationship present in E-E caregiving. For each year in the study, one-quarter to one-third of E-E caregiving occurred between generations, as 64-year-old-and-older children took of care their more senior parents, parent-in-laws, aunts, and uncles.

Another interesting trend was the decrease in caregiving of non-relative friends (29.9% to 18.4%) with the increase in spousal caregiving (22.5% to 38.9%) from 1997 to 2014. It is possible that this shift in relationship between E-E CGs and CRs might be partly due to a more accepting cultural shift, such as legalization of same-sex marriage, so that non-heterosexual partners were more comfortable labelling their partner as a spouse instead of a non-relative friend when answering the survey questions.

Prior research found that 70% of CGs who cared for an older adult did so for two to seven years.2 The results of this study maintain the notion that CG is not a short-term responsibility, even when the CGs themselves are older adults. The larger standard deviations in 1997 (10.4 years) and 2014 (10.0 years) compared to the other two years may be due to the survey’s limitations. When CGs were asked, “how long they had been caregiving,” some CGs may have responded by stating, “the [CR’s] whole life.” In such cases, interviewers had been instructed to enter the CR’s age as the answer; thus, potentially skewing the data.

Although this study found no significant change in an E-E CG’s self-reported health status, it is important to recognize that the CG’s self-assessment may be subject to social desirability bias, particularly given the high percentage of familial and spousal E-E caregiving relationships. E-E CGs might be intentionally or unintentionally disinclined to express any consequences to their health due to being a caregiver for a variety of reasons, such as feelings of guilt or pride, cultural gender roles, or even concern that it may call attention to their own physical or mental deficits.

The primary limitation on this research stemmed from the method of data collection: surveys. Although the surveys were thorough (over 200 variables), allowed for some open-ended answers, and gave interviewers cues on how to “probe” a respondent, the fact remained that subjective survey data can often have less than optimal accuracy. The original study design did not include any additional follow-up or verification of the information. However, this limitation was not specific to this study, but rather characteristic of the field of caregiving.

From the literature research, findings between studies sometimes appeared contradictory. For example, there was no clear consensus on CG burden among those caring for older CRs in variables such as mortality, depression, and mental stress.2,8,9 The Level of Burden score used in this study’s correlation test was a consolidated score meaning the study makers asked multiple questions regarding hours of care and number of daily activities performed in the survey, assigned values to those answers, and then consolidated a respondent’s answers into the Level of Burden scale.7 More interesting results could come from performing correlation tests between the specific variables asked (regarding financial, emotional and physical stress) and an E-E CGs age.

Conclusion

The field of caregiving is a difficult one to study, largely because much of the research is opinion-based and completed through surveys. Although imperfect, the current literature does underscore the clinical need and real-world application for this type of research given that some people living with a chronic illness receive the majority of their care from lay caregivers at home.1

As CGs as a group get older, it is imperative that health care professionals ensure the needs of these CGs are being sufficiently addressed. The lack of thorough investigation on the E-E caregiving population may be concealing the needs of this specific population. Follow-up studies should explore caregiving subset populations such as generation spanning E-E caregiving, particularly when the CG is the older participant, or non-relative, non-spousal E-E caregiving, because such relationships likely have interpersonal dynamics and complexities that are not being addressed elsewhere.

Similarly, additional research should investigate how best to shorten the average length of time that E-E CGs undergo such responsibilities. Likewise, with no clear consensus in the field of caregiving on the positive and negative impacts of caregiving on CGs, particularly E-E CGs, new studies need to explore potential benefits and burdens. Such research could help society prioritize education and inform action plans to assist caregivers, ensuring that (a) they have a high quality of life near the end of life, and (b) that their informal, unpaid workload does not shift to institutional health care settings.


Acknowledgements

Thank you to Elizabeth Kuchinski, MPH, and Brian Piper, PhD, at Geisinger Commonwealth School of Medicine for their support on this research.

Disclosure Statement

The author declares that they have no competing interests.

Publisher’s Notes

This article was reformatted after publication as part of The Guthrie Journal’s move to a new platform so that all of our articles would have a consistent look. The article was published November 18, 2021, and reformatted in March 2022. In addition to small formatting changes, tables and figures were moved to appear near their mention in the text.

A preliminary version of this article appeared in Scholarly Research in Progress (SCRIP), which is “aimed at promoting and disseminating student scholarly activity at Geisinger Commonwealth School of Medicine.” See VanDeMark, SooYoung H. Investigating trends in elder-elder caregiving in the United States. Scholarly Research in Progress. 2020;4:85-89. This version underwent full peer review by The Guthrie Journal prior to acceptance. Published with permission.