Physical activity in older people: a systematic review

The following methods were used to locate relevant published studies from January 2000 – 11 April 2011. Electronic searches of computerized databases were carried out on English language databases (The Cochrane Library, PubMed, MEDLINE, EMBASE, CINAHL, PsycINFO, British Nursing Index (BNI) and Scopus) and Chinese databases: Chinese Biomedical, VIP Chinese Science Journals and WANFANG DATA. Keyword combinations for electronic database searches are listed in Table 1. The search was limited to the English and Chinese languages.

Papers were reviewed if they met the following criteria: (1) original research; (2) independently reported the PA level of non-institutional older people (adults aged 60 years and over); (3) reported the proportion of any of PA recommendation or guidelines achieved by the sample.

No attempt was made to access unpublished studies or other ‘grey’ literature. The study selection process is set out in Figure 1.

All authors devised the research strategy. Initial screening was undertaken by one researcher (FS) and then checked by another (AEW). Disagreements about inclusion were discussed until agreement was reached. One researcher (FS) extracted the following data from the selected studies: country of study, study sample, measurements, recommendations and main results. Another researcher (AEW) verified the extracted data and made corrections where necessary. Disagreements were resolved by reference to the third researcher (IJN). All three researchers contributed to the synthesis of the data.

Definitions of “recommended” PA in older adults varied across the studies with some studies employing two or more guidelines. These guidelines are summarized in Table 2. The minimum recommended PA level in the guidelines in most studies was 150 minutes of moderate or vigorous PA per week and we adopted this standard as the desired PA level for the purpose of this review. We categorised the selected studies according to their underpinning PA guidelines, and identified studies as falling into one of two groups according to the detail provided by the study authors on the guidelines followed. Twenty six studies which stated only the total amount of PA per week were classified within the ‘less detailed guideline group’. Studies which stated the weekly frequency and daily dosage towards the total amount of weekly PA were classified as being in the ‘detailed guideline group’. According to the Physical Activity Guideline Advisory Committee’s report [5], detailed guidelines, such as 1995 CDC/ACSM guideline is too specific. In other words, the existing scientific evidence cannot distinguish the health benefits of 30 minutes of PA on 5 days a week from those gained through 50 minutes of PA on 3 days a week.

A total of 53 papers met the inclusion criteria. The main findings of the selected studies are reported in Tables 3 and 4.1

Most of the studies were conducted in the United States and Australia (USA n=19; Australia n=10; UK n=5; Brazil n=4; China n=4; Canada n=4; New Zealand n=1; Colombia n=1; South Africa n=1; Greece and Cyprus n=1; Cyprus n=1; Sweden n=1; Switzerland n=1). Forty nine papers reported cross-sectional studies and four reported longitudinal studies. The sample sizes ranged from 54 – 43,259. The number of studies conducted each year over the search period was not constant (1990–1994, n=2; 1995–1999, n=12; 2000–2004, n=32; 2005–2009, n=18; 2010–2011, n=2) with more than half being conducted between 2000 and 2004.

Forty seven studies measured PA intensity, duration, and frequency using subjective measures (interview or self-reported questionnaires) and six reported objective data gathered using an accelerometer. Two studies compared subjective and objective data measurements. While 39 studies recorded PA taken in periods of 10 minutes or more, other studies recorded all PA.

Physical activity comprises leisure-time PA, occupational PA, household PA and transportation PA. Leisure-time PA (LTPA) was most often measured and compared to the criterion for meeting PA recommendations. However, occupational, household and transportation PA were gathered in some studies. PA volume was calculated differently across the studies including: total metabolic equivalents (METs), minutes of weekly PA; minutes of participation in and frequency of PA during the week; kilocalories expended per kilogram of weight per day; and time in moderate to vigorous PA from accelerometers. To calculate minutes of weekly moderate to vigorous PA, some authors summed the duration of moderate and vigorous PA (MPA+VPA) while others doubled the time of vigorous PA because of its higher intensity (MPA+2VPA). The definition of moderate and vigorous PA also varied across the studies. For instance, the minimum cut off of moderate PA varied from 3 MET, 3.3 MET to 4 MET across the studies. Given the variety of methods of data collection and calculation of PA level, a meta-analysis was not attempted.

Thirty two studies adopted the more detailed PA guidelines while 26 studies were in the less detailed group with some studies employing two or more guidelines. Although most of the PA guidelines stated the total PA amount which should be accumulated in bouts of at least 10 minutes, 14 of the 53 studies recorded all moderate or vigorous PA (see Table 2).

Tables 3 and 4 present a synopsis of the findings relating to older people’s PA levels from each country. Across the 53 papers, the percentage of older adults meeting the guidelines ranged from 2.4% [55] to 83% [49] with most studies reporting that 20-60% of the samples met the guideline. When LTPA was measured, 20 studies excluded household work with reported PA prevalence ranging from 6.23% [58] to 67.51 % [27], while 14 included household work with reported PA prevalence of 10.86% [31] to 66.7% [16]. Seventeen studies measured all domains of PA (including occupational, household, transportation and recreational PA) with reported PA prevalence of 11.67% [31] to 77.22% [51], and two studies reported that 31.7% [13] and 62.4% [45] of older people achieved sufficient PA through walking. In six papers PA prevalence was reported by age group, gender group, or residential area group (rural versus urban), but not for the sample overall.

In 48 subjectively measured PA studies, 29 studies adopted the more detailed PA guidelines and 21 studies adopted the less detailed PA guidelines (5 studies adopted 2 criteria). Studies that employed the more detailed guidelines reported PA prevalence ranging from 6.2% in the Health Survey for England [38] to 82.6% in an urban Chinese sample [47]. Using the less detailed guidelines (criteria), sufficient PA in the other studies increased from 21.3 % [34] to 83.0% [49] in a small older female sample.

Those studies that accumulated PA data by 10 minutes bouts reported a relatively lower PA prevalence. The two studies which only included PA data of more than 30 minute sessions reported the lowest PA prevalence (10.9% [30]; 6.23% [58]) across all the subjectively measured PA studies.

Six studies used accelerometers and reported extractable data (i.e. proportion of the sample meeting the criterion of sufficient PA rather than measures of central tendency). The actigraph accelerometer was used in five studies and Actical in one study. When measured against the less detailed guidelines, the lowest prevalence (1.84%) was reported by Davis and Fox [3] based on an European sample and the highest was 17.2% from a US national survey, NHANES 2005–2006 [43]. Applying the more detailed guidelines, Davis and Fox [3] found nobody achieved sufficient PA and Colley [57] reported the highest proportion of 4.5% in this group with the exception of Hurtig-Wennlof et al.’s [51] exceptional finding of 87.04%.

However, there was a difference between the studies using self-report compared to the objective measurement of PA. In our review, two studies compared the subjective and objective data. Tucker et al.’s [43] analysis of the NHANES 2005–2006 data found that the sufficiently active group proportion defined by accelerometer measurement was 7.25% and 17.24% (using different guidelines); this increased to 54.2% when measured by questionnaire. Hurtig-Wennlof et al.’s Swedish study [51] reported a 87.04% objectively measured PA prevalence which was higher than the self-reported IPAQ data of 72.2%.

Twenty two studies reported the recommended PA prevalence in males and females separately. In general, men’s PA levels were higher than women’s. In the self-reported data, gender differences of PA ranged from 0.8% [30] to 21.4% [15], while the differences measured by accerlerometer were significantly lower (0.2% and 1.5%). However, we noticed that in the self-reported total PA, increased participation in PA by women exceeded that of men in three studies [40, 53].

Two Chinese papers reported the PA prevalence by place of residence and reported different results. Muntner et al. [22] reported sufficient all domain PA prevalence among 52.7% of their rural and 9.8% of their urban older people samples using the US CDC 1995 criterion from the InterASIAN Study. However, Xu et al. [47] reported that Guangdong province urban residents were more active than rural residents, with 82.6% of the urban and 60.9% of the rural samples attaining the HP 2010 goals measured by GPAQ.

Eighteen studies measured PA subjectively and two studies measured PA objectively across different age groups, however, the majority divided the samples into two age groups and reported that the older old were more sedentary than the younger old. Only five studies divided the group by relatively narrow age bands. Patterns of participation in PA decreased progressively with age for both men and women although there was variation across the studies. In McGuire et al.’s study [36] PA declined from 23.0% in the 65–69 year group to 13.6% in 85+ year group but in Mummery et al.’s study [35], the difference between the 60–64 year group and 80+ year group was 24.2%. There was an unexpected rise in the 70–74 year age group in the Sims et al.’s [42] study although the general trends decreased with age. Although Hamdorf et al. [12] did not report the detailed percentage for each age group, we can see the same gradually declining trend from 50.8% in the 60–64 year age group to 15.4% in 85+ year age group in Figure 2.

A crucial aspect of investigating trends over time is the comparability of methods. Differences in instruments, cut-off points, PA definitions and domains of activity investigated posed significant challenges to our review. Although nine studies reported the results of the BRFSS, not all the studies included national representative samples making it difficult to assess trends. Therefore, we only included the results of longitudinal studies or the findings from the same surveillance evaluated by the same criterion to establish the trend from seven studies. In Australia between 1990–2000 the proportion LTPA of older people aged 60 years and over meeting the PA guideline decreased from 28% to 25.8% in Merom et al.’s report [13] while the NSW Population Health Survey [11] detected a small rise in the 65 + year age group from 36.2% in 1998 to 39.8% in 2005 with a dip to 34.0% in 2003. The trend of regular walking [13] was similar to Chau et al.’s [11] report. The USCDC compared the BRFSS data in 2001 and 2005 and reported that the US’ older population was more active, with a rising prevalence of 43.1% to 44.5% in men and 32.2% to 36.3% in women [21]. Another paper using the BRFSS 2001 and BRFSS 2007 reported the same overall trend from 37.5% to 39.3% [48]. Conversely, Knuth et al. [26] reported a clear drop in PA prevalence between 2002 and 2007 in Brazil. In 2002, the proportions of older people reaching the PAGA 2008 guideline were 56.2% and 35.3% in the 60–69 year and 70+ year age groups respectively and decreased to 42.7% and 23.7% in 2007.

One of the challenges facing the development of disease prevention programmes is the lack of reliable data for PA levels and trends [59] and the data for PA levels in older people is no exception. Generally within the included studies, the older old age groups were less likely than those of younger age to be regularly active, and women were less likely than men to achieve regular PA, especially in LTPA across both the subjectively and objectively measured PA studies regardless of the PA guidelines adopted. The decline in PA with age may be the most consistent finding in PA epidemiology [60, 61] with the higher PA prevalence among males echoing the findings of previous studies [62] which may reflect increasing disability with age and cultural norms across the genders. However, one Australian study [42] found that the decline of LTPA across the age groups was not consistent with LTPA peaking in the 70–74 year group. Additionally three studies reported that women were more active than men in relation to total PA [25, 40, 53]. This pattern also appeared in other studies [63] which may reflect the inclusion of household activities and other non-leisure, non-sport activities [53] in the total PA, which, in many cases, are largely specific to women. However, this was not a consistent finding across the included studies regardless of the guidelines category, indicating the need to quantify household activities accurately to enable comparisons across genders.

In our review, there was a slightly increasing trend towards recommended PA levels in older people in Australia and the US over the last 10 years but a decrease in Brazil. Studies of time trends in PA have been conducted mostly in developed countries and their results indicate that LTPA levels appear to be increasing [31, 34, 64, 65]. Nevertheless, the data indicate that substantial numbers of older adults do not engage in sufficient PA to promote their health and there is considerable variation in the levels of PA reported across countries. Further, we found little data on time trends in PA in developing countries which was echoed in a recent systematic review [66]. However, the decline in PA in Brazil [26] is perhaps indicative of changes in occupational PA with the emergence of increasing sedentary work in developing countries which will present a future public health challenge especially when combined with dietary changes. Similarly Hallal et al. have also highlighted the lower proportions of adults who are physically active in south east Asia [59].

The only two studies comparing self-reported PA to accelerometer measured PA in our review yielded contradictory results [43, 51] with Tucker et al.’s [43] findings being consistent with the majority of other similar studies [67]. It is likely that the subjective PA studies reflect social desirability, recall bias and that there is underestimation of PA using objective measurement; PA monitors are worn typically on the hip which means that they do not accurately assess upper body activities or account for movements that require extra effort, such as walking uphill or carrying loads. Hurtig-Wennlof et al.’s [51] contradictory findings were probably caused by the lower cut-off point of MPA at 760 counts per minute (contrasting with about 2000 counts/min in the other studies) and the inclusion of PA under 10 minute bouts. This highlights the need for well designed studies using objectively measured PA to generate the evidence base for public health initiatives.

Limitations of this review arise from the discrepancies and inconsistencies in instrumentation, PA type measured, guidelines or recommendations adopted by different researchers, algorithms, sample frames and other confounding factors. These made it difficult to make full use of the extracted data and impossible to compare the PA prevalence between different regions/countries or assess PA trends with certainty.