Relationship between family caregiver burden and physical frailty in older adults without dementia: a systematic review

We conducted a systematic search using Ovid MEDLINE(R) (1946 to present), Embase (1974 to 2015), PsycINFO (1987 to 2015), Cochrane Central Register of Controlled Trials, and CINAHL. The search was conducted by a trained research librarian from inception to April 2015 (Additional file 2). We also emailed three noted scholars in the fields of frailty and/or caregiver burden, who did not identify any additional studies.

Our inclusion criteria were defined a priori. We included English-language studies that included family caregivers of community-dwelling older adults and reported on caregiver burden. Experimental [randomized controlled trial (RCT), controlled trial, quasi-randomized trial, and quasi-experimental], observational (before/after study, prospective cohort study, retrospective cohort study, case-control study, analytic or descriptive cross-sectional study, case report, and case series), and qualitative studies were considered eligible. Systematic reviews addressing our research question were included to identify additional eligible studies. Book chapters, symposium and conference proceedings, essays, editorials, letters, commentaries, narrative reviews, and protocols were excluded.

We defined family caregivers as unpaid caregivers who are typically close relatives (e.g., spouses and children) and provide the bulk of the care and support received by a dependent older adult living in the community (not in a hospital, retirement community, nursing home, or any form of long-term care) [15]. Studies involving solely institutionalized older adults were excluded on the basis that caregiver burden has previously been explored extensively in this setting. Family caregiver burden was defined as any construct or variable representing the physical, psychological, and financial cost of providing care for a loved one [16]. Studies reporting caregiver-related outcomes related to subjective burden (e.g., effects on emotional, social, financial, physical, and spiritual functioning), objective burden (e.g., amount of time spent on caregiving, the caregiving tasks that are performed, and possible financial problems), or both were considered eligible. Studies that employed qualitative measures and quantitative measures (e.g., Caregiver Assessment Tool, Caregiver Burden Screen, and Caregiver Stress Scale) were both included.

Care recipient frailty was defined as any property or syndrome that may relate to physical frailty. Examples of such properties included, but were not limited to, unintentional weight loss, self-reported exhaustion, weakness, slow walking speed, low physical activity, and loss of energy, physical ability, or health. Articles defining frailty solely as a limitation in functional activity were considered ineligible. Finally, studies of older adults with dementia were excluded.

We note that our registered protocol did not include this last exclusion criterion. However, early abstract screening revealed that a large number of studies of caregivers of older adults included only patients with dementia. Based on the known substantial positive association between dementia and caregiver burden and our interest in identifying to the extent possible the impact of physical frailty in particular, we elected to exclude such studies on the assumption that they would be highly confounded [17, 18].

Two investigators (TR, AA) independently screened every original title and abstract identified. Consensus was reached on decisions to advance studies to full text screening, with discrepancies resolved by discussion between the two reviewers and adjudication by a third reviewer (AAH) as necessary. Full text versions of eligible studies were subsequently retrieved for detailed review and were independently screened in duplicate (TR, AM), with discrepancies resolved by discussion and adjudication by a third reviewer (AAH) as necessary. The reference lists of review articles encountered during the full text screening process were examined by both reviewers (TR, AM) to identify additional potentially eligible studies. Title and abstract as well as full text screening were conducted on the DistillerSR platform using pre-designed online forms piloted by a pair of reviewers (TR, AM) and refined to ensure usability.

After determining article inclusion, two reviewers (TR, AM) extracted study data into extraction tables in duplicate using piloted and standardized extraction forms and reconciled conflicts by discussion with adjudication by a third reviewer (AAH) as necessary. The following information was extracted from every included study: study design, country, study objectives, general description of care recipient population, care recipient age, care recipient gender, other descriptive information about care recipients (e.g., mini-mental status examination score), general description of caregiver population including age, gender, and additional details (e.g., whether caregivers resided with the recipient; relationship to care recipient). Outcome data regarding how frailty was defined and measured, how family caregiver burden was defined and measured, and reported results regarding the relationship between caregiver burden and physical frailty were extracted into summary tables.

All included papers were critically appraised in duplicate by two reviewers (TR, AM) using piloted and standardized assessment forms, with discrepancies resolved by discussion and adjudication by a third reviewer (AAH) as necessary. Included RCTs were rated on sequence generation, allocation concealment, blinding of participants, personnel and outcome assessors, incomplete outcome data, selective outcome reporting, and other sources of potential bias as per the Cochrane Collaboration’s tool for assessing risk of bias [19]. Included cross-sectional studies were appraised on sample selection, comparability, and outcome assessment based on a published adapted version of the Newcastle-Ottawa Scale [20]. As mentioned below, five papers reported on essentially the same study. However, each paper was appraised separately, since critical appraisal considers not just the analysis and design of a study but the manner in which its results are reported.

We note that our registered protocol had contemplated using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach to assess the level of evidence across studies, which posits a systematic and explicit approach to making judgments about the quality of evidence supporting clinical recommendations, such as a practice guideline statement about the effect of a particular intervention [21]. However, factors such as the heterogeneity of designs and outcomes among included studies, and the paucity of interventional studies, precluded the identification of any evidence-based recommendations (regardless of quality), thus rendering GRADE inapplicable.

Given significant heterogeneity in study populations, outcomes, and measures across eligible studies, it was not possible to perform a meta-analysis. Accordingly, results were synthesized narratively, with summary tables presenting the key study characteristics, findings, and limitations, and risk of bias assessments associated with each study.

The Geriatric Education and Research in Aging Sciences (GERAS) Centre employed and provided partial funding for the work of two co-authors (AP, AAH) but did not play a role in the selection of the topic, conduct of the study, or interpretation or presentation of the results as an institution acting in a corporate capacity. The GERAS Centre has no financial interest in the outcome of the study.

Characteristics of the five included studies are presented in Table 1 [25, 27‐30]. One study was an RCT [29]. The remainder were cross-sectional in design [25, 27‐30].

Although our review identified nine papers, we have decided that it is most accurate to characterize our review as including five studies. Five of the papers [22‐26] had the same principal investigator and involved the same population: caregivers of frail older adults who were enrolled in a RCT of a multifactorial interdisciplinary intervention aimed at reducing mobility-related disability. We note that the study reporting on this RCT was not captured by our database search and would not have met our review’s criteria because the population of the RCT itself did not include caregivers [32]. In a recruitment exercise distinct from the RCT itself, caregivers designated by RCT participants were sent a set of questionnaires.

We contacted the principal investigator on these five publications in December 2016 to clarify the relationships between them. Each paper consists of a different cross-sectional analysis of the data collected through the above-referenced questionnaires. Two of the papers describe cross-sectional analyses conducted at an earlier point in the underlying RCT and involved the caregivers of 93 RCT participants. The remaining three papers describe cross-sectional analyses which were conducted at a later point and involved caregivers of 119 participants in the same RCT. Because each paper reflects a different analysis of essentially the same questionnaire responses from the same population, but was published separately, we have designated one paper as the index study for purposes of extraction and appraisal and refer to it hereinafter as “the Aggar study” [25] and term the remaining publications “the related papers” [22‐24, 26].

The largest study had 606 caregiver-care recipient dyads in total, of which 135 included a frail care recipient [30]. The four remaining studies had fewer than 120 dyads [25, 27‐30]. The smallest had 33 dyads [29].

Table 2 summarizes key findings from the included studies.

The care recipient and caregiver populations included in each study were similar in terms of most descriptive characteristics. Of note, two studies only included care recipients who met certain minimums of cognitive function based on the Mini-Mental Status Examination (MMSE): >18 for the Aggar study [25] and >15 in the RCT [29]. One study only included patients with mild or worse cognitive impairment (MMSE <27) [28]. Among studies reporting on co-residence, approximately half of caregivers co-resided with their care recipient (57.1% [25] in the Aggar study; 47% (control) to 50% (intervention) in the RCT [29]).

No study included physical frailty as one of its areas of focus. Only two studies [25, 29] included participants deemed frail using a validated measure (Fried Frail Scale (FFS) [33].

Because they were recruited from an RCT involving a frailty intervention, all participants in the Aggar study were frail (FFS >2) as a criterion for inclusion [25]. These studies did not otherwise report on participants’ frailty. All participants in the included RCT were drawn from a study population previously determined to be frail (FFS ≥2), so recipients’ degree of frailty was not measured as part of the study [29].

Two studies neither explicitly defined or measured frailty qualitatively or quantitatively [27, 28]. Each study, however, included variables that reviewers agreed could be considered potential proxy measures for components of the FFS, such as gait ataxia or extrapyramidal gait disorder [28] which could be considered proxy measures for one of the FFS criteria (slow gait speed), a surrogate indicator for decreased physiologic reserve [33]. Another study analyzed care recipients' “timed up and go”, poor balance, and use of walking aid [27]. In one study, care recipients were deemed frail if their caregiver described them as “frail due to age” [30].

There was considerable heterogeneity in caregiver burden measures used across included studies. With one exception [30], all studies employed a standardized instrument used to assess caregiver burden. These included the Caregiver Reaction Assessment (CRA), Hospital Anxiety and Depression Scale (HADS), Caregiver Strain Index (CSI), and Relative Stress Scale. One study employed a total of four instruments [29]. Kim et al. asked participants to report their physical strain, emotional stress, and financial hardship on a 5-point scale (1 = not at all; 5 = great deal/very much) [30].

All included studies suggested that caregivers of physically frail older adults experienced negative reactions to caregiving, as well as anxiety and depression. No studies included normative data for purposes of comparison with non-frail populations.

The Aggar study found that caregivers of a population of physically frail older adult care recipients experienced negative reactions to caregiving as measured using the CRA [25]. However, 68% of caregivers found it rewarding to provide care [25].

One study, which directly compared caregivers of frail adults with those of other types of patients, found that caregivers of older adults whom the caregivers themselves described as “frail due to old age” experienced the least physical strain, emotional stress, and financial hardship when compared with caregivers of patients with cancer, diabetes, and dementia [30].

Finally, two studies found that indirect indicators of frailty such as timed up and go, ataxic gait, use of walking aid, poor balance, and weight loss were not associated such measures as caregiver strain, irritability, or tension [28, 29].

Risk of bias assessments for the four cross-sectional studies and one RCT are presented in Tables 3 and 4, respectively [19]. Summative risk of bias scores were not calculated for each study, as this practice is disfavoured [19]. The Aggar study drew on a narrow population of study participants from a small geographical region, raising concerns about sample representativeness [25]. Not all of the cross-sectional studies used validated instruments to measure caregiver distress. All included studies relied on self-reported measures of caregiver burden, although this is appropriate given the outcome of interest in these studies was subjective caregiver burden. In our estimation, the one included randomized controlled trial reported incomplete outcome data and did not adequately conceal allocation [29]. As previously discussed, an overall assessment of the quality of the evidence using a tool such as the GRADE approach was deemed inapplicable.

Our study used well-defined a priori criteria and a rigorous systematic methodology. We do note three potential limitations. First, non-English-language studies were excluded. Second, the relatively small number of included papers, and the fact that five publications consisted of different analyses of essentially the same data and were treated as papers related to a single index study, may impact generalizability of findings and speaks to a substantial gap in the evidence. Finally, the findings are limited by the paucity of normative comparisons in the included studies.