The Association of Objectively Measured Physical Activity and Sedentary Behavior with (Instrumental) Activities of Daily Living in Community-Dwelling Older Adults: A Systematic Review

Introduction

Physical activity (PA), defined as bodily movement produced by the contraction of skeletal muscle that requires energy,¹ has been linked to various health benefits with increasing age.² Up to 60% of older adults worldwide do not meet PA guidelines³ due to physical impairments that arise with aging^4,5 or sedentary behavior (SB), which refers to waking activity (mainly performed while in a sitting, reclining, or lying posture) with little to no energy expenditure beyond the resting metabolic rate.⁶ Low PA (volume, duration, or intensity) and high SB (duration) can be distinct behaviors⁷ that independently amplify age-related decline in many physiological systems⁸ and may therefore affect endurance, muscle strength, and flexibility⁹ as well as cognition.¹⁰ However, these capacities are necessary to autonomously function in daily life, including engaging in activities of daily living (ADL), referring to self-care tasks, such as transferring in and out of bed, feeding, and dressing, as well as instrumental activities of daily living (IADL), which involve more complex and cognitively demanding tasks, such as housekeeping, shopping, and medication use.¹¹

Previous systematic reviews of longitudinal and cross-sectional studies have demonstrated that PA classified as of at least moderate intensity is positively associated with the ability to complete ADL and IADL,^12,13 whereas negative associations were found between SB and the ability to perform these activities.¹⁴ An important limitation of these findings is that conclusions are predominantly based on self-reported measures of PA and SB (ie, questionnaires), which are especially susceptible in older adult populations to overestimation of PA and underestimation of SB¹⁵ as a result of recall bias. Furthermore, self-reported measures of PA and SB often fail to capture activity at the lower end of the PA continuum, which comprises most of the PA in older adults (eg, light-intensity, short-duration tasks).¹⁶ PA and SB can be most accurately quantified with wearable technology (accelerometers, pedometers), which allows for the objective assessment of PA as well as continuous monitoring of activity in daily life¹⁷ (ie, frequency, intensity, duration). Objective measurements of PA and SB are therefore essential to advance knowledge by accurately quantifying the association of PA and SB with ADL and IADL, which can ultimately be targeted through public health clinical intervention.

This systematic review aimed to describe the association of objectively measured PA and SB with ADL and IADL in community-dwelling older adults.

Materials and Methods

The protocol of this review was registered in the PROSPERO International prospective register of systematic reviews with registration number CRD42018103910.

Information Sources and Search Strategy

Two assessors (the Vrije Universiteit librarian (RO) and AR) conducted a systematic literature search based on the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement,¹⁸ consulting the following electronic databases from inception to June 21, 2020: PubMed, Embase, the Cochrane Library (via Wiley), CINAHL, PsychINFO, and SPORTDiscuss (via EBSCO). The search terms “active or inactive lifestyle”, “motor activity”, and “people over 60 years of age” were used to ascertain articles that studied PA and SB in relation to any health outcome in older adults; the full search strategy is presented in Appendix A. Articles that reported associations of PA and SB with ADL and IADL were organized and managed in the software Endnote (Version X8.2 Clarivate Analytics, Philadelphia, USA) and Rayyan QRCI.¹⁹

Inclusion Criteria

Full-text articles published in English or Dutch were considered eligible for this systematic review based on the following criteria: 1) observational or experimental study, 2) participants’ mean or median age ≥60 years old, 3) study population consisted of community-dwelling older adults, 4) PA and SB were measured with an accelerometer or pedometer, 5) ADL was defined as any tool or questionnaire explicitly described as measuring ADL and/or IADL, and 6) PA and SB were studied in relation to ADL and/or IADL. For intervention studies, associations at baseline or control group data were included.

Article Selection

Search results were assessed for possible eligibility based on title and abstract screening by two independent assessors (KR and EvdR) using the Rayyan screening software.¹⁹ Full-text screening was performed in duplicate by two independent assessors (KR and LD, or AR) and differences in opinion with regard to inclusion and exclusion decisions were resolved by another assessor (AM). The references of all included articles were screened for additional eligible articles.

Data Extraction

Data extraction was performed by two independent assessors (EG and WZ) and disagreement was settled by a third assessor (KR). The following data were extracted: first author; year of publication; country; cohort; study design with, if applicable, follow-up period; characteristics of study population (population selection), sample size, age (in years), sex (number and percentage of females), device used for objective assessment of PA/SB (accelerometer, pedometer), device name, wearing location of device, number of monitor days, mean device wear time, minimum duration of device wear to define a valid day, number of valid days required for analysis, reported measures of PA/SB and their definitions, PA/SB scores, tools and definitions used for ADL and IADL assessment, activities included in an ADL or IADL tool/questionnaire, ADL/IADL scores, adjustment model(s), statistical analysis to study association(s), effect size(s) with 95% confidence interval (95% CI) or standard error (SE), and significance level (p-value).

Assessment of Study Quality

Study quality and risk of bias were assessed by two independent assessors (EG and WZ) using modified versions of the Newcastle–Ottawa scale (NOS) for cross-sectional and longitudinal studies,²⁰ customized for this systematic review. Three domains, selection (representativeness of study cohort and ascertainment of exposure), comparability (adjustment model(s) and statistical analysis), and outcome (assessment of outcome and, if applicable, adequacy to follow-up), were assessed and the median of total possible stars (points) was set as the cut-off to determine high or low quality, defined as ≥ or < 4 out of 7 and ≥ or < 5 out of 9 for cross-sectional and longitudinal studies, respectively (Appendix B).

Data Analysis and Visualization

Extracted information and associations between PA/SB and ADL or IADL were reported in tables, visualized in effect direction heat maps,²¹ and synthesized in albatross plots²² according to the PRISMA¹⁸ and Synthesis Without Meta-analysis (SWiM)²³ guidelines. Data were reported based on the following hierarchy of adjustment: 1) age and sex, 2) age and sex, and other factors (eg, cognitive function, number of chronic diseases, body mass index), 3) age or sex, and other factors, 4) other factors only, and 5) unadjusted (crude) model. When articles reported more than one type of statistical analysis for an association, the following hierarchy for reporting was considered: 1) adjusted linear regression, 2) adjusted logistic regression, 3) partial correlation, 4) unadjusted linear regression (including Pearson’s and Spearman correlation), 5) analysis of variance (ANOVA), and 6) Mann–Whitney test, Student’s t-test, or chi-squared test. Continuous measures of PA/SB were used if reported and categorical variables were used otherwise. P-values were calculated when these were not reported: for linear regression: the upper and lower limit of the 95% CI were used to acquire the SE, SE=((upper limit of 95% CI – lower limit of 95% CI)/(2*1.96)), which was then used to obtain the absolute value (abs) of the z-statistic (z), z=abs (regression coefficient/SE), and eventually calculate the p-value, p(calc)=exp((−0.717*z) - (0.416*(z²))). The aforementioned formulae were also used for ratio measures (odds ratio (OR), hazard ratio (HR), and risk ratio (RR)), except for that the upper and lower limits of the 95% and effect sizes were transformed into logarithms using natural log(ln) first.²⁴ For correlations, sample size (n) and coefficients (including Pearson’s R and Spearman’s Rho) were used to calculate the t-statistic (t), t=R*√((n-2)/(1/R)), of which the absolute value (abs) was compared to the two-sided student’s t-distribution using T.VERD.2T in Microsoft Excel to obtain p(calc). For comparison between groups, mean values and standard deviations (sd) were used to calculate t, t=((mean₁-mean₂)/√(((n₁-1)*((sd₁)²))+((n₂-1)*((sd₂)²)))/((n₁+n₂-2)*((1/n₁)+(1/n₂)))), which was then compared to the two-sided t-distribution using the earlier-mentioned function in Microsoft Excel. Associations with p-values that could not be calculated were conservatively estimated for effect direction heat maps as being ≥0.25 or the largest p-value derived from the reported information and excluded from albatross plots.²⁵

Effect Direction Heat Maps

Effect direction heat maps²¹ were created to provide a qualitative overview of all associations between PA/SB measures and ADL or IADL and were stratified by study design (longitudinal versus cross-sectional) and ordered by sample size. Articles that included combined measures of ADL and IADL were categorized as IADL because inability to carry out more complex and cognitively demanding activities precedes difficulty in ADL.²⁶ The observed direction of effect was determined based on whether higher PA and lower SB were associated with better (positive effect) or worse (negative effect) ADL and IADL, indicated by an upwards or downwards triangle, respectively. The following color scheme was used to present significance: p<0.001 (dark blue filled triangle), 0.001≤p<0.01 (blue filled triangle), 0.01≤p<0.05 (light blue filled triangle), 0.05≤p<0.1 (light grey empty triangle), 0.01≤p<0.25 (grey empty triangle), and p≥0.25 (dark grey empty triangle).

Albatross Plots

Albatross plots are scatter plots of sample size plotted against two-sided p-values, stratified by the observed effect direction to graphically present the estimated magnitude of associations²² (expressed as median with corresponding interquartile range, [IQR]). Each data point represents an association and based on whether higher PA and lower SB were associated with better (positive effect) or worse (negative effect) ADL and IADL, data points fall on the right or left side of albatross plots, respectively. Contour lines were superimposed on the plot to examine hypothetical effect sizes, here selected as standardized regression coefficients (βs), and derived from the following equation: N=(((1-β²)/β²)*(Z_p)²) in which Z_p denotes the z-value associated with given two-sided p-values. Separate albatross plots were made for ADL and IADL using the Stata Statistical Software, Release 16.0 (StataCorp LLC, College Station, Texas, United States), each stratified by measures of PA and SB. Sensitivity analyses were performed by stratifying albatross plots using population selection (disease versus general), study design (cross-sectional versus longitudinal), adjustment (adjusted versus unadjusted associations), device type (accelerometer versus pedometer), and device wearing location. For the latter sensitivity analysis, device wearing locations were entered into the albatross plots if reported for ≥5 associations to obtain an IQR.

Results

The literature search identified 18,806 articles of which 9660 articles were left after duplicate removal. Of the 1017 full texts assessed for eligibility, 30 articles^27–56 were included in this systematic review (Figure 1).

Figure 1 Flowchart of article selection process.

Characteristics of Studies

A total of 24,959 (range: 23 to 3749) community-dwelling older adults were included with mean or median age ranging from 60.0 to 92.3 years and, on average, populations were 54.6% female. In 11 articles, specific disease groups were studied: osteoarthritis (OA),^34,36,44,54 chronic obstructive pulmonary disease (COPD),^28,39,45,56 cirrhosis,³⁷ Parkinson’s disease,³⁸ and stroke survivors.⁴⁰ Longitudinal associations were reported in six articles^{27,31,34,36,53,54} (mean follow-up period of 3.1 years) and represented 7554 older adults with mean or median age ranging from 62.4 to 80.6 years (56.8% female); remaining articles reported cross-sectional associations (Table 1). The NOS categorized 26 out of 30 articles as high quality (Table 2).

Table 1 Characteristics of Included Studies

Table 2 Scoring of Study Quality Based on Modified Versions of the Newcastle–Ottawa Scale for Cross-Sectional and Longitudinal Studies

Measures of Physical Activity and Sedentary Behavior

Accelerometers were used in 28 studies, while two studies^27,28 used pedometers to objectively measure PA/SB (Table 3). The following measures of PA/SB were included: number of steps (or walking duration),^{27,28,37,38,41,44,45,50,55} activity counts (or accelerations, movement intensity),^{29,33,42,43,45,49,53,55,56} energy expenditure (EE),^31,37,45,50 duration (in different units of time) of total PA (TPA) (or mobile duration),^45,47,51,56 moderate to vigorous PA (MVPA) (or moderate PA (MPA) or vigorous PA (VPA) individual),^{30–32,34,36–40,46–49,51,52,54,55} light PA (LPA),^{34,40,47,49,52,55} and SB (or lying duration, immobile time),^{30–32,35,37,38,40,43,45,47,49,52,55} breaks per sedentary hour (SB break rate),⁵² and breaks in sedentary time (BST).^32,52,55

Table 3 Measurement Methods and Scores of Physical Activity and Sedentary Behavior

Assessment of Activities of Daily Living and Instrumental Activities of Daily Living

The association of PA/SB measures and ADL was studied in 20 articles using the following tools: London Chest Activities of Daily Living (LCADL) scale,^28,39 Katz Index of Independence in Activities of Daily Living (Katz),^29,53 Glittre-ADL test,⁴⁵ Western Ontario and McMaster Universities osteoarthritis index (WOMAC) functional limitation sub-scale,²⁷ Health Assessment Questionnaire Disability Index (HAQ-DI),⁴⁹ Barthel Index,⁴⁰ Composite Physical Function (CPF) scale,⁵² Knee injury and Osteoarthritis Outcome Score (KOOS) questionnaire function in daily life sub-scale,⁴⁴ Parkinson’s Disease Questionnaire-39 (PDQ-39) activities of daily living dimension,³⁸ Nottingham Extended Activities of Daily Living (NEADI),⁵⁶ and custom questionnaires^{30,31,35,36,43,48,50,51,55} (Table 4). In 13 articles, the association between measures of PA/SB and IADL was studied with the use of the following tools: Tokyo Metropolitan Institute of Gerontology Index of Competence (TMIG),³² Rosow-Breslau scale,³⁷ CPF scale,^47,52 Late-Life Disability Index (LLDI),⁵⁴ Late-Life Function and Disability Index (LLFDI),^40–42,46 and custom questionnaires^31,33,34,43 (Table 5).

Table 4 Assessment, Scores, and Breakdown of Activities in Tool Used for the Assessment of Activities of Daily Living

Table 5 Assessment, Scores, and Breakdown of Activities in Tool Used for the Assessment of Instrumental Activities of Daily Living

Associations of Physical Activity and Sedentary Behavior with Activities of Daily Living and Instrumental Activities of Daily Living

All associations are visualized by effect direction heat maps (Figure 2), standardized regression coefficients (βs) for each association are presented by albatross plo ts (Figure 3), and the sensitivity analyses (population selection, study design, adjustment, device type, and device wearing location) are demonstrated in Figure 4.

Figure 2 Effect direction heat map visualizing associations of objectively measured physical activity and sedentary behavior with (A) activities of daily living and (B) instrumental activities of daily living based on p-values, ordered by sample size, and stratified by study design (cross-sectional and longitudinal). ± indicate positive/negative effect direction (higher PA and lower SB are associated with better (+) or worse (-) activities of daily living (ADL) or instrumental activities of daily living (IADL). PA/SB measures: Counts=activity counts, EE=energy expenditure, TPA=total physical activity, MVPA=moderate to vigorous physical activity, LPA=light physical activity, SB=sedentary behavior, break rate=number of breaks per sedentary hour, BST=breaks in sedentary time. ▲/▼ (dark blue): p<0.001, ▲/▼ (blue): 0.001≤p<0.01, ▲/▼ (light blue): 0.01≤p<0.05, Δ/▽ (light grey): 0.05≤p<0.1, Δ/▽ (grey): 0.1≤p<0.25, Δ/▽ (dark grey): p≥0.25. *activities of daily living or instrumental activities of daily living as independent variables and PA/SB as dependent variable. dDisease population. Abbreviations: M, Males; F, Females.

Figure 3 Albatross plots depicting the magnitude of associations, provided as standardized regression coefficients (βs), of higher physical activity (PA) and lower sedentary behavior (SB) with (A) activities of daily living and (B) instrumental activities of daily living. ● (green) steps, ● (pink) activity counts, ● (yellow) energy expenditure, ■ (red) total physical activity, ■ (blue) moderate to vigorous physical activity, ■ (light green) light physical activity, ▲ (purple) inverse sedentary behavior, ▲ (orange) break rate (number of breaks per sedentary hour), ▲ (cyan) breaks in sedentary time. β = ±0.10, β = ±0.20, β = ±0.30.

Figure 4 Continued.

Figure 4 Albatross plots depicting the magnitude of associations, provided as standardized regression coefficients (βs), of higher physical activity (PA) and lower sedentary behavior (SB) with activities of daily living (ADL) and instrumental activities of daily living (IADL), stratified by (A) population (general versus disease), (B) study design (cross-sectional versus longitudinal), (C) adjustment (adjusted versus unadjusted associations), (D) device location (accelerometer versus pedometer), and (E) device wearing location. (A) population selection: ● general, ○ disease, (B) study design: ● cross-sectional, ○ longitudinal, (C) adjustment: ● adjusted, ○ unadjusted, (D) device type ● accelerometer, ○ pedometer, (E) device wearing location: ● (green) wrist, ● (pink) triceps, ● (yellow) hip. β = ±0.10, β = ±0.20, β = ±0.30.

Associations of PA and SB with ADL

Longitudinal associations between PA/SB measures and ADL were studied in four articles;^27,31,36,53 all associations were significant and effect directions showed that higher PA and lower SB were consistently associated with better ADL: lower MVPA and EE, and higher SB at baseline, were associated with an increased likelihood to become dependent in ADL after two years in community-dwelling older males,³¹ higher baseline activity counts was associated with a lower hazard of ADL dependence after 3.4 years in a general community-dwelling older adult population,⁵³ engaging in approximately one-hour MVPA was associated with a lower risk of becoming dependent in ADL after four years in an osteoarthritis population,³⁶ and a bidirectional association was identified between number of steps and ADL (a higher average number of steps was associated with better ADL from baseline and, additionally, worsened ADL from baseline was associated with a lower average number of steps) over five years in an osteoarthritis population.²⁷ These findings were supported by cross-sectional associations, which demonstrated that higher PA and lower SB were associated with better ADL; furthermore, three articles^28,51,55 studied ADL as independent and PA/SB as dependent variable, showing that limited ability to complete ADL was associated with lower PA and higher SB (Table 6; Figure 2A). The median [interquartile range] standardized regression coefficient (β) for all articles reporting associations between PA/SB measures and ADL was 0.145 [0.072, 0.280] (Figure 3A).

Table 6 Associations of Objectively Measured Physical Activity and Sedentary Behavior with Activities of Daily Living and Instrumental Activities of Daily Living in Community-Dwelling Older Adults, Stratified by Domain

Associations of PA and SB with IADL

Three articles studied longitudinal associations between PA/SB measures and IADL,^31,34,54 which were all significant and had a positive effect direction: community-dwelling older male adults with lower MVPA and EE, and higher SB at baseline were more likely to become dependent in IADL after two years³¹ and in two articles including older adults from the Osteoarthritis Initiative (OAI), after two years follow-up, higher MVPA and LPA at baseline³⁴ and increasing MVPA from baseline⁵⁴ were associated with a lower hazard for the development and progression of IADL dependence³⁴ and improved IADL,⁵⁴ respectively. Cross-sectional associations were in line with these results, showing that PA/SB measures were positively associated with IADL. Three studies investigated the cross-sectional association between measures of PA/SB and IADL with IADL as independent variable and PA/SB as dependent variable, showing that experiencing difficulty in IADL was associated with lower levels of PA (Table 6; Figure 2B). The median [interquartile range] standardized regression coefficient (β) for all articles reporting associations between PA/SB measures and IADL was 0.135 [0.093, 0.211] (Figure 3B).

Sensitivity Analyses

Sensitivity analyses demonstrated that population selection (general and disease populations) had an influence on the effect sizes of associations between PA/SB and, in particular, ADL with larger standardized regression coefficients found for disease populations (median [IQR]: β=0.314 [0.159, 0.460]) than general populations (median [IQR]: β=0.111 [0.067, 0.178]) (Figure 4A). Longitudinal associations presented smaller standardized regression coefficients (median [IQR] for ADL: β=0.078 [0.065, 0.120] and IADL: β=0.084 [0.069, 0.094]) when compared to cross-sectional associations (median [IQR] for ADL: β=0.157 [0.098, 0.301] and IADL: β=0.162 [0.113, 0.224]) (Figure 4B). For unadjusted associations larger standardized regression coefficients were found (median [IQR] for ADL: β=0.316 [0.304, 0.462] and IADL: β=0.170 [0.144, 0.176]) in comparison to adjusted associations, especially for the relationship between PA/SB and ADL (median [IQR] β=0.112 [0.072, 0.178]) (Figure 4C). In all studies, except for two that used a pedometer, accelerometers were used to monitor PA and SB (median β [IQR] for ADL: 0.145 [0.076, 0.266] and for IADL: 0.135 [0.093, 0.211]) (Figure 4D). For ADL, largest median standardized coefficient was observed when the device was located on the wrist (median [IQR] β=0.187 [0.082, 0.232], followed by a positioning on the hip (median [IQR] β=0.114 [0.064, 0.157]) and triceps (median [IQR] β=0.078 [0.059, 0.277]); whereas for IADL, device wearing location had no influence on the effect size (median β [IQR] for hip: 0.162 [0.090, 0.204] and for triceps: 0.158 [0.106, 0.213]) (Figure 4E).

Discussion

Higher PA and lower SB at baseline and increased PA from baseline were consistently associated with maintaining or improving the ability to complete ADL and IADL from baseline in community-dwelling older adults. These longitudinal associations were supported by the more frequently reported cross-sectional studies. Effect sizes were similar for associations between PA/SB and ADL or IADL; cross-sectional results yielded larger effect sizes for both ADL and IADL, and larger effect sizes were additionally found for ADL in disease populations and unadjusted analyses.

Objective measures of higher PA and lower SB showed associations with better ADL and IADL, which was in line with previous literature that purports health benefits from PA of any intensity and limited sedentary time.⁵⁷ This is also in accordance with intervention studies that provide evidence of improved functional capacities in response to PA, such as coordination, muscle strength, and balance, which are essential for ADL and IADL.⁵⁸

This systematic review identified similar standardized effect sizes for the association of PA/SB measures with ADL and IADL, which was unexpected considering differences in capacities required to complete ADL and IADL. ADL primarily depends on motor functions, such as upper limb control and postural stability, that are necessary to complete the most basic forms of self-care;²⁶ whereas, IADL additionally places a demand on cognition, particularly executive function during activities, such as grocery shopping.⁵⁹ Furthermore, IADL dependence precedes ADL with the latter hence indicating greater system-level impairment and severe loss of autonomy.⁶⁰ This is because ADL dependence is typically caused by musculoskeletal failure to where minimally demanding activities can no longer be performed.⁶¹ However, inclusion of exclusively community-dwelling older adults may have masked differences between ADL and IADL as to remain non-institutionalized requires a certain minimum ADL ability.⁶² While it is likely that the ability to complete ADL and IADL plays a role in determining to what extent someone can engage in PA, it is important to acknowledge that having the capacity to perform these activities does not ensure that the capacity is actually used to partake in PA.⁶³

Population selection revealed dissimilarity in the effect sizes for disease versus general populations, showing that associations were dependent on the population studied, which can be explained by the pathophysiological backing regarding the effect of disease on the engagement in PA. Chronic diseases, such as COPD and osteoarthritis (commonly studied populations within this systematic review), may modify the effect that PA has on ADL because engaging in PA may be more critical for physical functioning in the presence of disease-induced impairments, such as breathlessness and stiffness, and inversely, SB may be more detrimental in the presence of disease. Stratification by study design showed that there were smaller effect sizes for longitudinal studies when compared to cross-sectional studies, which may suggest that while baseline PA and SB are determinants of baseline and future ability to perform ADL and IADL, changes in ADL and IADL in short periods of time may be more affected by other factors involved in health status. Larger effect sizes found for unadjusted associations in comparison to adjusted associations strengthen the importance of our adjustment hierarchy, which was applied to prevent inflation by confounders, such as age and sex, that may mask the actual relationship of PA/SB with ADL and IADL.

PA and SB are, as highlighted in this systematic review, associated with the ability to independently accomplish ADL and IADL. Enhancing PA and reducing the time spent sedentarily are therefore promising strategies to maintain functional independence. With increasing age, however, multimorbidity and cognitive impairment are more prominent and threaten healthy aging.⁶⁴ It may therefore be that the inability to perform ADL and IADL influences PA engagement and involves higher levels of SB, resulting in an overall more inactive lifestyle. Conversely, an active lifestyle could protect older adults from a loss of functional independence, which is implicated by our longitudinal findings. To disentangle this reverse causation, future randomized controlled trials are advised to inform public health strategies about an attainable active lifestyle for older adults based on their functional capability.

Population aging is accompanied by an increase in disease burden among older adults, which threatens functioning in daily life and, therefore, underpins the clinical relevance of our findings that objectively measured PA and SB are modifiable lifestyle factors of the ability to carry out ADL and IADL. This can be used to determine the dose–response relationship of PA and SB with ADL and IADL to guide public health and clinical interventions for preventing and delaying loss of independence. Considering the importance of an active lifestyle for maintaining independence, as shown in this systematic review, PA may act as a target for future intervention studies. Future studies should aim to improve standardization in the assessment of PA and SB (eg, device-wearing location, cut-off points, and assessment of ADL and IADL) to unravel the dose–response relationships of PA and SB with ADL and IADL and, ultimately, establish thresholds to prevent deterioration in the ability to complete ADL and IADL.

The inclusion of solely articles that objectively measured PA and SB is a strength of this systematic review as it eliminates bias that is involved in self-reported assessment and thus provides the most accurate insight into PA and SB and the subsequent association with ADL and IADL. As older adults regularly spend most of their time in low-intensity activities, a broad range of PA measures, including LPA, is an additional strength because this metric is often neglected due to the difficulty of measuring LPA via self-report.⁶⁵ Furthermore, diverse community-dwelling older adults were included, without exclusion of specific disease groups, which allows for generalizability of our findings. Another strength is that the literature search focused on articles that were explicitly described as measuring ADL and/or IADL, in contrast to the liberal use of keywords related to these daily-life activities throughout the literature. Despite the important advantages of measuring PA and SB objectively, accelerometers and pedometers are limited in their ability to capture loading or resistance during PA, which represents a limitation to fully characterizing PA. Our strategy in making a hierarchy of adjusted covariates to address confounding by age and sex may have suppressed the true relationship between PA and SB with ADL or IADL due to over-adjustment. While we aim to include associations only adjusted for age and sex, in some studies the closest available model includes adjustments for a range of variables beyond age and sex that may have interfered in the causal pathway, which would therefore represent over-adjustment and lower effect sizes. In all studies, except for one study that included performance-based measures of ADL,⁴⁵ the ability to perform ADL and IADL was assessed by self-report of the participants themselves. Such a subjective approach in assessing ADL and IADL may lead to biases, including individual differences in self-perceived difficulty or ability to perform ADL or IADL and therefore presents a limitation. However, the ability to accurately self-assess ADL and IADL is likely easier than PA or SB given that the activities assessed are familiar and finite. Methodological challenges were also encountered in PA/SB measures due to large variability in units, definitions, and statistical analyses used to examine the association of interest. This limitation has precluded us from performing a meta-analysis and led to alternative methods to synthesize our results.

Conclusion

Higher PA and lower SB are significantly associated with better ADL and IADL in community-dwelling older adults. Future research should, based on older adults’ ability to function in daily life, aim to establish the optimal dose of PA to prevent development and progression of dependence in ADL and IADL, as well as investigating if higher PA and lower SB can recover loss of independence in one or more activities to, ultimately, design attainable lifestyle guidelines for older adults.