Individual characteristics associated with physical activity intervention delivery mode preferences among adults

Participants and procedure

The cross-sectional data analysed in this study was collected during July and August 2010 as part of the Queensland Social Survey, conducted by the Population Research Laboratory at the Central Queensland University in Australia. The survey was conducted by trained interviewers using computer assisted telephone interview (CATI) software. The target population included persons 18 years of age or older who, at the time of the survey, were living in a dwelling unit in Queensland that could be contacted by direct-dialled, land-based telephone service. A single respondent within each household was selected using pre-established guidelines to ensure an equal yet random selection of male and female participants. The selection and interview protocol has been described in detail elsewhere [17].

All participants provided informed consent at the start of the CATI survey. Approval from the Human Research Ethics Committee of CQUniversity Australia was obtained prior to data collection.

Measures

Analysis

Descriptive statistics were calculated for all variables and compared across intervention preference groups using chi-square tests (for categorical variables) and analyses of variance (for continuous data). Multinomial logistic regression models were conducted to model the likelihood of intervention preference choice relative to the reference category (‘a face-to-face program with an instructor’ served as the reference category in all analyses) as a function of demographic, behavioral, health and internet variables. ‘A face-to-face program with an instructor’ was chosen as the reference category because it is the most preferred intervention type among adults [11–15] and because it may be the least promising in terms of public health impact, due to the high cost associated with delivery and issues with reach and maintenance [24–26].

Univariate analyses were run with the regression of the dependent variable (preference for intervention delivery-mode) on each of the independent variables (potential correlate) using the whole sample. Only independent variables which had a p-value of 0.25 or less were selected for inclusion in the multinomial models [27]. Based on these analyses, income, internet access at home, self-rated health, years of schooling, work status, work hours and sitting at work were dropped from subsequent multivariate models. To screen for potential multicollinearity problems, a correlation matrix was produced examining the correlation between all remaining independent variables. If two variables were found to be strongly correlated with each other (i.e., >50%), [27] only one variable (i.e., the strongest predictor) was retained. Children under 18 years old living at home and use of social networking sites were found to be highly correlated with age (negative association), and were dropped from the analysis.

Separate models were conducted for demographics (model 1), health variables (model 2), and behavioural variables (including internet behaviors; model 3). Only variables demonstrating a significant relationship (p ≤ 0.05) with preference for delivery mode in these models were then entered into a combined model (model 4). Relative risks are reported to show the magnitude of association [27]. Stata version 11 (Stata-Corp) was used for all analyses.

Participants

Factors associated with preferred delivery mode

Model 1 found that being separated (RR = 2.06, P < 0.05), being within the 35–44 aged group (RR = 2.71, P < 0.05), and living in a rural area (RR = 1.77, P < 0.05) was associated with a greater likelihood of preferring a web-based intervention over a face-to-face intervention. Being older aged (55 + yrs old, RR = 3.15, P < 0.05) was associated with an increased likelihood of preferring a print-based intervention over a face-to-face intervention. Being female was found to decrease the likelihood of preferring a web-based intervention (RR = 0.52, P < 0.01) and a print-based intervention (RR = 0.52, P < 0.05). There were no significant demographic predictors for preferring a group-based intervention compared to a face-to-face intervention. The variance accounted for by the demographic model was 3% (Nagelkerke R² = 0.03, P = < 0.01).

Model 2 found that obesity (RR = 0.62, P < 0.01) was associated with a decreased likelihood of preferring a group-based program over a face-to-face program, and that having no chronic health conditions (RR = 0.53, P < 0.01) was associated with a decreased likelihood of preferring a print-based intervention over a face-to-face program. There were no health predictors significantly related to preferring a web-based intervention over a face-to-face intervention. The variance accounted for by the health model was 1% (Nagelkerke R² = 0.01, P = < 0.01).

Model 3 found that higher levels of total physical activity (RR = 0.99, P < 0.05) were associated with a decreased likelihood of preferring a web-based program over a face-to-face program. Occasional use of mediated sources (e.g., email, telephone) to contact family and friends (compared to very often, RR = 2.26, P < 0.01) was associated with increased likelihood of preferring a web-based program over a face-to-face program. Higher levels of internet usage in the last week was associated with a decreased preference for group-based programs (RR = 0.98, P < 0.05) compared to face-to-face programs. There were no significant behavioural predictors related to preferring a print-based intervention over a face-to-face intervention. The variance accounted for by the behavioural model was 2% (Nagelkerke R² = 0.02, P = < 0.01).

Comparison to previous research

Our results support the previous research findings that men are more likely to prefer mediated (print and online) interventions than women [14] and that people with lower physical activity levels are more likely to prefer mediated programs than people with higher levels of physical activity [15]. The preference for mediated intervention among men may be due to men’s desire to self-monitor their health, maintain their regular activities, and to obtain health information independently and judge illness severity before seeking help [32]. There may also be social factors contributing to this preference, such as men’s traditional social roles (e.g., difficulty relinquishing control; immunity and immortality; perception that men are not interested in prevention; and lack of male care providers) [33]. Among those with lower physical activity levels, mediated interventions may be perceived as less confronting, and hence preferred to face-to-face interventions. This seems plausible, since low activity levels are associated with lower self-efficacy, knowledge and skills [34].

The finding that having a chronic disease is associated with an increased preference for print-based interventions (in the health model) in the current study is in contrast to studies exploring intervention preferences among individuals with diabetes [12] and cancer [13]. In both of these studies, a strong preference for face-to-face interventions over mediated interventions (telephone, video-tape, pamphlet, and internet) was reported. One explanation for the different findings observed in the current study is that age may have had a confounding effect, since older age is associated with both the preference for print-based interventions and the likelihood of having a chronic disease [35, 36]. Alternatively, it may be that the distinct etymology associated with different chronic diseases impacts on an individual’s physical activity support needs. In the case of elderly people, a wide range of barriers associated with older age, such as lack of transportation facilities, financial considerations, lack of affiliation to the fitness center culture, or social embarrassment, may contribute to a preference for mediated interventions that are home-based [37].

To our knowledge, previous research has not examined the influence of geographical location, weight status or internet use on preference for intervention delivery mode. In doing so, we found some novel findings. Namely, living in a regional town was associated with preferring group-based interventions and living rurally was associated with preferring online interventions. This may be due to the unique social structures and availability of health services present in each of these settings [38, 39]. For example, individuals that live in rural Australia face transport and accessibility problems that are not experienced by those living in regional and urban areas [39], potentially making the practicality of online interventions more attractive. The interactivity of these interventions may also enhance their appeal, since social isolation is also often experienced in rural settings [38, 40]. Another novel finding was that obesity was strongly associated with preferring face-to-face interventions with an instructor over any other delivery mode. Possible reasons for this include being too shy or embarrassed to exercise in front of a group and/or a perceived need for greater support than what they believe could be provided via a mediated intervention [41]. This could have public health implications, given the health and economic burden associated with obesity [42, 43] and the need for sustainable approaches. Finally, this study also found that being in the 35–44 age group, compared to being in the 18–34 age group, was associated with a greater preference for web-based interventions over face-to-face interventions. This may reflect the increasing use of the internet among this age group at work (51%), and at home (52% access internet at home daily) [44], and/or the difficulty or perceived difficulty of attending face-to-face sessions at this stage of life due to a busy social or professional life [45].

Strengths and limitations

The major strength of this study is that it is the first to examine the relationship between physical activity intervention delivery mode preference and individual characteristics among adults. A second strength is that the research was conducted in a large sample recruited from the general population. Research in this field is scare and is predominantly limited to research describing intervention preferences in special sub-population groups. Hence, this study has contributed to the literature by expanding current knowledge about adults’ physical activity intervention delivery mode preferences and by highlighting what factors may influence these choices.

This study also has some limitations that should be considered when interpreting results. First, there was over-sampling of older adults and under sampling of younger adults. This is potentially due to the use of landline only sampling. Whilst recent studies have shown that the exclusion of mobile only households does not (yet) significantly influence survey results [46], it is possible that this sampling method led to response bias, since younger people increasingly live in a mobile only household [47]. Further, due to the under-sampling of adults aged 18–24 and 25–34 years it was necessary to combine these age groups in order to perform the analyses. Given age-related differences in several factors that may influence intervention preferences (e.g., stage of life and lifestyle) it is possible that examining these two age categories separately would yield different results. This will require further analysis. Second, the data for this study was collected in 2010. More recent data may yield different outcomes, given the rapidly changing telecommunications environment. However, our review of previous research suggests that our data reflects current intervention preference trends [10, 11]. Third, the individual characteristics explored in this study were not exhaustive. Although our study investigated a range of potential demographic, health and behavioural correlates of preference for intervention delivery mode, other potential correlates such as psychosocial factors were not explored. This may account for the small proportion (6%) of variance explained by the factors included in the combined model, since psychosocial factors typically explain 30-40% of the variance in physical activity behaviour [48–50]. Psychological factors that may be relevant to assess in future studies could include perceived physical activity barriers, preference for physical activity in groups or alone, social support and the importance placed on social interaction, as well as how each of these interacts with key socio-demographic, environmental and health factors [51]. Finally, due to the low proportion (<1%) of people who reported a preference for telephone based counselling these were excluded from the analysis and thus we were unable to explore preference characteristics related to this delivery mode which is still currently used in research settings and by Government agencies.

Implications for practice

Despite accounting for only a small proportion of variance, the results may have implications for practice as they suggest that certain population groups, such as males and females, the young and the old, those living in regional and rural areas and those with a high body mass index may be responsive to different interventions types. However, the results also demonstrate that the preference for mediated interventions is low at a population level, especially among women, the obese, and those living in urban areas. If we are to successfully reduce the public health burden associated with physical inactivity, an increased effort to promote the desirability and accessibility of mediated approaches among these groups may be needed. Furthermore, attempts to adapt intervention strategies to the local context and target them to well-defined groups or individual characteristics may also further enhance efficacy [52]. This study helps delineate these groups and provides some insight into what approach may be most appealing to whom.