Older adults with chronic illnesses form a large part of society [1
]. In general, their physical activity (PA) levels are low [2
]. This is associated with increased health problems and decreased cognitive functioning [3
]. Therefore, it is important to increase PA levels of older adults with chronic illnesses. PA programs for this population exist, but attendance and results vary [4
]. In the present manuscript, the effects of an already proven-effective computer-tailored PA stimulating intervention on PA behaviour in older adults with chronic illnesses (OACI) is examined. This is a secondary analysis of a larger study on the effects of this proven-effective PA intervention on cognitive functioning.
Almost 85% of the older adults in developed countries have at least one chronic illness, and around 60% suffer from multiple chronic illnesses [1
]. However, these chronic illnesses themselves might be not the only issue; the functional limitations and mobility restrictions older adults experience as a result of these chronic illnesses provide additional problems [8
]. Examples of these functional limitations and mobility restrictions are: (1) having difficulties with walking; (2) getting into a bed or a chair; or (3) stair climbing. These problems are highly prevalent and have a negative influence on the ability to maintain their activity levels [9
]. As a result, these functional limitations and mobility restrictions impede the ability to have an independent life and can lead to poorer health-related quality of life (HRQoL) [9
Increasing PA is a proven-effective strategy for both prevention as well as treatment of chronic illness [13
]. PA has other health benefits as well, such as weight control, strengthening of muscles and bones, and improvements of physical and cognitive functioning, mental health and HRQoL—all factors negatively affected by chronic illnesses [16
]. Despite these beneficial effects of PA, older adults are the least physically active age group, especially when they suffer from chronic illnesses [2
]. Most of them do not reach the recommendation of 150 min per week of moderate-to-vigorous physical activity (MVPA) supplemented with muscle, bone, and balance improving activities at least twice a week [18
]. This may be due to the many PA-related barriers (i.e., fatigue, pain) experienced by older adults with chronic illnesses [2
]. However, meta-analyses by Bullard et al. [17
] and de Vries et al. [10
] suggest that adherence to the guidelines for PA (≥150 min of MVPA/week) is highly feasible and effective among chronic disease populations.
Besides MVPA, the health benefits of light-intensity physical activity (LPA) have been established as well. Objectively measured light physical activity (LPA) is associated with lower all-cause mortality risk and improved cardiometabolic risk factors and cognitive functioning [20
]. LPA is perhaps more feasible for OACI than MVPA [23
]. Consequently, it is suggested that, next to MVPA, LPA should be taken into account when evaluating PA [24
Although a few PA programs exist for OACI (e.g., Coach2Move [25
], Strong-for-Life [26
], Life-P [27
]), most programs are not easily accessible, and often only reach already active older adults [5
]. They usually take place at a research site, gym, or physical therapist, which OACI need to visit 1–3 times per week. These programs are generally face-to-face and offer detailed and intensive supervision [17
]. Yet, these programs are also highly demanding, making it more difficult for OACI to adhere to these programs, especially on long term [28
]. However, not only clinic-based programs exist for OACI, but there are also proven effective home-based programs as described by Duijts et al. [29
], Gary et al. [30
] and Lee et al. [31
]. These programs generally provide participants with a personalised PA program to be executed at home. First, participants are taught by a research nurse how to correctly perform the aerobic exercise program and how to adjust the exercise prescription. Home-based programs typically provide more autonomy (e.g., more choices regarding training schedule and fewer transportation-related barriers) [17
]. Nevertheless, these existing programs for OACI are commonly focussed on physical functioning instead of physical activity. In other words, these programmes are more focussed on improving the capacity to execute habitual daily activities such as stair climbing than on having an active lifestyle with higher amounts of low, moderate and vigorous PA. Besides, a meta-analysis showed that offering exercise or PA in a program is not enough to stimulate OACI to become more physically active on their own in their daily lives outside of the gym [10
Based on this knowledge, the computer-tailored PA stimulating intervention Active Plus was developed for people aged over 50 years [32
]. At a later stage, the eHealth program was adapted to a more elderly (≥65 years) population who often suffered from chronic illnesses [34
]. Active Plus participants receive 3 personalised PA advice letters (online or print delivered) in 4 months. Previous research in people aged over 50 years showed that the Active Plus group was 1.5 h per week more active at moderate to vigorous intensity after 1 year compared to controls [35
], even in older adults with impaired mobility [36
]. Although a recent study showed positive effects of the adapted Active Plus intervention on PA in single older adults (≥65 years) with physical impairments 3 months after baseline, no effects were found after 6 months [37
]. However, this concerned an implementation study without a control group, making it impossible to draw definite conclusions.
In this paper, the effects of the Active Plus intervention on PA in OACI are examined. Although previous studies already showed the effectiveness of the Active Plus intervention on PA in the general older adults (≥50 years) population [35
], solely self-reported PA measures were used to assess intervention effects. However, self-reported PA questionnaires are known for their over-reporting of PA due to social desirability and recall accuracy [38
]. Furthermore, not all intensities of PA are validly assessed by questionnaires, in particular, light to moderate PA, which is the intensity of PA older adults are most likely to engage in [39
]. Therefore, in the present study, we also included objectively measured PA. The use of accelerometers in the assessment of PA became more user friendly [40
], and past research suggests that it captures the quantity and intensity of PA behaviour more accurately [41
Nevertheless, accelerometers too have disadvantages. Insight in which specific PA activities a person performs cannot be derived from accelerometer data [42
]. For example, solely on accelerometer data from a single device, it is still impossible to discriminate between sitting and standing. Furthermore, depending on the attachment site (hip versus upper leg or wrist), some accelerometers are not able to detect all kind of movements such as upper/lower body movement or stationary movement, while these behaviours are common in older adults during gardening, household chores, and cycling. Cycling in particular is a frequent PA activity in The Netherlands [43
]. In addition, most accelerometers are also not able to assess water-based activities, as not all accelerometers are waterproof [41
]. Therefore, it is recommended to both use objectively measured PA with an accelerometer and subjectively measured PA with a self-report PA questionnaire in assessing PA intervention effects [44
The present study will be one of the first studies to assess computer-tailored PA intervention effects on PA behaviour in a broad sense. We evaluate the effects of the Active Plus intervention in OACI on objectively measured LPA and MVPA, and on the likelihood to perform common PA activities and MVPA minutes per week during these activities assessed with a self-report questionnaire. As the intervention was aimed at increasing PA, it is hypothesized that the intervention group increased both their objectively measured and self-reported PA more than the waiting list control group. Although the intervention is individually tailored, it might be that not all subgroups of participants respond similarly to the intervention. Therefore, we examine in an exploratory way whether the effects differ for the degree of impairment, age, gender, body mass index, educational level, and marital status.
The current study assessed the effects of the computer-tailored Active Plus intervention on objectively measured and self-reported PA in older adults with chronic illness(es). Additionally, it explored the effectiveness of the intervention in subgroups.
Overall, the effects of the Active Plus intervention on PA in OACI were limited. The hypothesis that the intervention group would increase their objectively measured PA was not confirmed. Although at 6 months follow-up there seemed to be a positive intervention effect tendency for MVPA, there was no statistical evidence. The hypothesis that Active Plus would increase self-reported PA behaviour during specific activities was only partly confirmed. At 6 months after baseline, the intervention was effective in increasing the likelihood to perform self-reported cycling and gardening. For those individuals engaging in cycling at 6 months, the intervention was effective in increasing the amount of MVPA during cycling as well. At 12 months after baseline, the likelihood to perform self-reported walking improved significantly more in the intervention group. Besides, the intervention appeared effective in increasing the amount of MVPA during cycling at 12 months after baseline. In sum, we did not see any effects on objectively measured PA and only limited effects on self-reported PA.
However, these results are not in line with previous research on the Active Plus intervention. Peels et al. [35
] demonstrated that Active Plus was effective in increasing PA in adults of 50 years or older both in short- and long term. Participants in this previous RCT had a mean age of ±63 years and only ±40% of them had a chronic limitation. Hence, both populations differ substantially and this possibly explains the limited results we found in the present study. It is possible that the Active Plus intervention is too voluntary for OACI. In addition, a more recent study on the adapted version of the Active Plus intervention had a more comparable population consisting of single older adults over 65 years with a chronic impairment in PA. This study showed limited effects of Active Plus on PA too [37
]. However, this concerned an implementation study without a control group, making it impossible to draw definite conclusions. Both studies only measured self-reported PA, so our findings concerning the objectively measured PA could not be compared.
As there were differences in the way of measuring the degree of impairment in all three Active Plus studies, it is hard to conclude that an eHealth intervention like Active Plus is less effective in increasing PA in a more ill population. To our knowledge, there are no meta-analyses that studied the effect of PA promoting eHealth interventions in OACI. However, a meta-analysis by Chase [19
] showed that PA interventions, including home-based interventions, tested among healthier participants had larger effects than those tested among chronically ill populations, although effectiveness varies between different groups of chronic illnesses [66
]. In addition, starting levels of PA in the present study were already relatively high with a raw mean of objectively measured MVPA of 193–210 min per week. Hence, there may be less room for improvement in this population (ceiling effect). Therefore, very large effects should not be expected when studying already physically impaired older adults [10
Furthermore, next to the different target populations/ samples in the three studies and the already active population in the present study, the intervention itself and design of the study could explain the limited effects found in the present manuscript. The Active Plus intervention was specifically adapted to an older and a chronically ill population with the intervention mapping protocol based on a literature study, focus groups with the target population and expert interviews. However, constructs and affected determinants of PA of the original proven-effective intervention did not change. The degree of importance of certain determinants and consequently the tailored messages did change to some extent [34
]. Presumably, the tailored messages need to be fine-tuned even more to the target population or are not convincing enough to uptake PA behaviour in this specific target group.
In addition, in the present study, we saw a larger dropout of relatively older participants due to a loss of interest during the intervention period and especially at the time the follow-up questionnaire (needed to compose the third advice) was sent out. One might consider that the intervention up to that point was not what the older participants expected. Conceivably, the relatively older participants expected the intervention to involve contact moments. Furthermore, it is possible participants thought the questionnaire was too extensive. An eHealth intervention study of Van der Mispel et al. [67
] concluded that in eHealth interventions extensive questionnaires should be avoided as dropout rates were higher in interventions with rather lengthy questionnaires than in interventions with an interactive character. Although an extensive questionnaire allows for more adequate tailoring, it may be possible to shorten it. However, at this point, when the follow-up questionnaire was sent out, participants already received two out of three times advice which contained most of the information available in the intervention. It is possible that participants were already satisfied at this point, and got out of the intervention what they needed and expected. This may especially be the case for online interventions [68
]. A study into the appreciation of the intervention might provide more answers for why the dropout rate was higher for relatively older participants during the intervention period.
In contrast to the above, in the current manuscript, exploratory subgroup analyses suggest that more vulnerable OACI participants benefitted more from the Active Plus intervention on several PA outcome measures, especially on the lower intensity PA outcomes. Firstly, intervention participants who were more severely impaired increased the likelihood of performing walking behaviour in contrast to control group participants. In addition, participants with a BMI of ≥30 kg/m2 had borderline significant higher increases in LPA. Additional analyses suggest that participants with a BMI of ≥32 kg/m2 did increase LPA significantly. Besides, the likelihood of doing odd-jobs increased for participants of 80 years or over. Furthermore, participants with a medium level of education had improved MVPA minutes of household activities. Thus, it appears that different subgroups respond in diverse ways to specific parts of the intervention and more vulnerable participants improve mostly on the lighter intensity activities. Possibly these are better achievable in the more vulnerable population. However, none of de covariates was a moderator on more than one PA outcome measure. Therefore, these results should be taken into account with precaution. Nonetheless, the intervention could be tailored more to the specific needs of the non-responsive subgroups.
While only the computer-tailored PA stimulating Active Plus intervention is not sufficient to increase PA in the general OACI population, a possible solution to increase the effect could be a blended approach in which this eHealth intervention and face-to-face contact are combined [69
]. A blended approach could be a cost-effective solution, as it implies less costly face-to-face contact and improved feeling of self-regulation. For example, the Active Plus intervention, which contains solely personalised advice, could be combined with face-to-face contact with a physiotherapist or weekly meetings with a PA group for older adults. A blended approach is increasingly being applied in both healthcare and mental healthcare [69
]. There are already some examples of blended approach interventions aimed at promoting PA in older adults [70
]. However, only a few studies exist and results are varying. Accordingly, additional research (i.e., meta-analysis) is necessary to identify what type of intervention (i.e., web-based, supervised, blended, etc.) works best for whom and which healthcare professionals are most suitable to refer and guide OACI to the intervention and eventually guide them.
Although we did not find any significant improvements on objectively measured PA, some improvements on subjectively measured PA were found. Most benefits were seen in self-reported cycling behaviour, as participants in the Active Plus group were more likely to perform cycling at 6 months after baseline. In addition, participants who cycled performed this behaviour for a longer time at both 6 and 12 months after baseline. While using a validated questionnaire [56
], self-reported questionnaires assessing PA behaviour are known for their over-reporting [38
]. Nonetheless, waist-worn accelerometers do not detect all movements such as gardening, upper body strength exercises and cycling [42
]. As we found most improvements on self-reported cycling at 6 months after baseline, and this activity is difficult to measure with an accelerometer, this problem might explain not finding effects with the objective PA measurement. Certainly, because it was fall/winter at the time of the second measurement people were more likely to stay indoors at that time because of the increased rainfall and the decreased daylight [72
]. The Active Plus intervention pays attention to the weather and seasonal effects, as the intervention aims to increase the self-efficacy of participants to keep a higher PA level during bad weather/colder seasons and provides options to exercise at home. It is possible that this advice has helped to increase cycling behaviour despite the colder weather, but due to the limitations of the accelerometer (which cannot measure cycling properly), we only found limited effects. Therefore, our results can be considered of value and clinical relevance.
Some strengths and limitations should be noted. Firstly, the current study has a strong research design (RCT) in which both objective (accelerometer) and self-reported PA (questionnaire) information were assessed [44
] to give more insight in the complexity of PA behaviour. Although, both measures have their strengths and weaknesses. Self-report questionnaires are known for over-reporting, whereas accelerometers do not measure certain activities properly (e.g., cycling, swimming) [42
]. By assessing objectively measured LPA and MVPA and self-reported MVPA behaviour during common PA activities we tried to gain the best insight into PA behaviour. Secondly, our research population was fairly varied and therefore the generalizability with a general OACI population seems reasonable. For instance, our research sample consisted of almost equal groups of male and female participants, and a considerable part of the participants was low educated (e.g., 51%). The mean number of comorbidities (3.5) is also in accordance with numbers in the general older adult population [73
] of The Netherlands, as well as BMI levels [74
]. Thirdly, the statistical method we used (two-part generalized linear mixed-effects model) to analyse the self-reported PA outcomes, is not used often. Most studies apply a linear mixed-effects models, but with self-reported PA data being highly skewed and zero-inflated this is not the optimal method [61
]. Finally, by conducting multilevel analyses in this study, the most accurate way of handling missing data was applied [64
Limitations were selective dropout, statistical power for multilevel analyses, no correction for multiple testing, and no information on adherence to the intervention. Although the selective dropout (i.e., older participants and during the intervention period) may have affected our findings, this is expected to be less detrimental because of the relatively low dropout. A dropout rate of 25.1% per cent is considered low in (partly) digital health interventions [75
]. Next, as the power calculation was based only on subject level analyses and not municipality level, primary and moderator analyses may have been underpowered, as the ICC was 0.07 instead of the expected <0.01. Large inclusion numbers and a relatively low dropout rate may have limited this potential problem. Besides, we performed multiple tests to show the results of this study in a broader perspective and to give a more nuanced picture of PA behaviour. However, the more tests that are done, the more likely erroneous conclusions are drawn, because the probability of a Type 1 error is increased [76
]. A Bonferroni correction, however, assumes that all of the hypothesis tests are statistically independent, which is not the case in the current study and is, therefore, overly conservative. The probability of making a Type 1 error would be less than Bonferroni assumes, and the Bonferroni correction would be an overcorrection. Therefore, we did not apply a Bonferroni correction, but results should be taken into account with precaution. Finally, adherence to the intervention was not administered during this study, therefore it is not known to what extent participants read or used the Active Plus intervention advice. However, previous research on Active Plus showed that printed materials were read by more than 93% of the participants [77
]. As participants were provided with both printed and online materials, reading level/intervention exposure in the current study is expected to be similar.