The process by which perceived autonomy support predicts motivation, intention, and behavior for seasonal influenza prevention in Hong Kong older adults

Ethical clearance was granted by the committee of Research Ethics and Safety (HASC) at Hong Kong Baptist University. We obtained approval and on-site assistance to recruit participants from one senior center in Hong Kong using convenience sampling. This center has over 1500 registered Chinese-speaking community-dwelling Hong Kong older adults aged 60 years and older. A total of 180 older adults expressed an interest in participating to senior center staff and members of the research team. Of these eligible participants, 141 agreed to participate and signed consent forms (response rate = 78.33%). The sample comprised of 14 males and 127 females (M age = 75.23, SD = 6.52, range 60 to 85) participated the study at Time 1. At Time 2, 137 participants (13 males, 124 females; M _age  = 75.10, SD = 6.49) remained in the study (retention rate = 97.16%).

Data were collected at two time points with a two-week interval in the November of 2014. At Time 1, during the peak season of winter influenza in Hong Kong, participants completed survey measures of study variables under the supervision of researchers following an introductory session in which the purpose and procedure of the study was explained and written informed consent obtained. The surveys took 10 to 15 min to complete. Senior center staff were not present during the completion and were not able to see participants’ responses. At Time 2, participants were invited to a face-to-face interview in a private room to ostensibly assess their knowledge of H1N1. The true purpose of the interview was to assess participants’ facemask wearing behavior for influenza prevention. During the course of the interview, the interviewer wore a facemask and feigned influenza-like symptoms (e.g., coughing, sneezing). Before the start of the interview, the interviewer told the participants in an offhand manner that he/she had caught “the ‘flu”, and that surgical facemasks were available at the desk that they could take and use for free. This allowed the interviewers to record participants’ facemask-wearing behavior. On completion of the interview, participants were debriefed regarding the cover story and informed that the interviewer had feigned influenza symptoms and worn the facemask as a prop to model the behavior.

Given that the Time 2 facemask wearing behavior was a categorical variable (yes/no), the proposed integrative model (see Fig. 1) was tested using path analysis (i.e., observed variables) with a variance-adjusted weighted-least squares (WLSMV) estimation method using the Mplus 7.3 [35]. Variables including facemask wearing habit, knowledge of facemask wearing benefits, frequency of influenza during the past 6 months, and perceived susceptibility were included as control variables. Adequacy of the fit of the proposed model with the data was based on multiple criteria for assessing goodness of fit including the comparative fit index (CFI), the root-mean-square error of approximation (RMSEA) and the weighted root mean square residual (WRMSR). Values exceeding .90 for the CFI and less than .08 and 1.00 for the RMSEA and WRMSR [36, 37], respectively, indicate good fit. For our tests of mediated effects, mediation was confirmed when the indirect effect of a predictor variable (e.g., autonomous motivation) on an outcome variable (e.g., intention) via a mediator (e.g., attitude) was statistically significant and the confidence interval of the effect size did not include zero.

In terms of the facemask wearing habit, most of the participants reported wearing facemasks previously (n = 124), while a small number of them reported that they did not been used to wearing facemasks previously (n = 17). With respect to knowledge of the benefits of facemask wearing, most of the participants reported knowing facemask wearing can prevent flu (n = 132), while a small number reported they did not know facemask wearing can prevent flu (n = 9). With regards to the frequency of influenza during the past 6 months, most of the participants reported that they had not caught influenza (n = 120). In comparison, only 11 participants reported that they caught influenza once, seven participants reported they had had the flu twice, and only three participants reported catching influenza three times. Missing data analysis revealed no significant pattern (missing data = 0.18%) and the small number if missing cases was replaced using mean substitution. Descriptive statistics and intercorrelations of the study variables are presented in Table 1.

The model exhibited adequate fit with the data, χ ² (3) = 1.58, p = .664, CFI = 1.00, WRMR = .178, RMSEA (90% CI) = .000 (.000, .112). Direct, indirect, and total effects of the path estimates of the integrated model are presented in Table 2 and Fig. 1. Effects of the control variables were not significant except the effect of frequency of influenza over the previous six months on autonomous motivation (β = −.18, 95%CI [−.344 to −.024], p = .025), and they are not, therefore, displayed in Fig. 1.

Focusing on the effects encompassed by hypothesis 1, we observed a statistically significant and positive effect of perceived autonomy support on autonomous motivation consistent with our hypothesis (hypothesis 1a). The indirect effects of perceived autonomy support on attitude and subjective norm were statistically significant as hypothesized (hypothesis 1b), although the direct effects were not. As predicted, the total, direct, and indirect effects of perceived autonomy support on perceived behavioral control were also statistically significant (hypothesis 1b).

Statistically significant and positive associations were observed for autonomous motivation on attitude, subjective norm, and perceived behavioral control, as hypothesized (hypothesis 2a). The indirect effects of autonomous motivation on intention mediated by attitude, subjective norm, and perceived behavior control were all statistically significant, consistent with predictions (hypothesis 2b).

We found statistically significant and positive effects of attitude, subjective norm, and perceived behavioral control on intention, as predicted (hypothesis 3a). However, contrary to predictions, the effect of intention on our measure of facemask wearing was small and not statistically significant (hypothesis 3b).

Current findings are consistent with the hypotheses of the integrated model and its component theories when it comes to predicting intentions. Perceived autonomy support was related to autonomous motivation for adopting health-related behaviors consistent with the predictions of SDT [16, 17, 38], a key component theory of the integrated model [22, 23]. The finding is also consistent with the findings of a previous study on facemask wearing [20]. The results suggest that establishing an autonomy supportive environment or ‘motivational climate’ toward influenza prevention behaviors such as facemask wearing will be effective in fostering autonomous motivation toward the behavior.

Consistent with previous studies of influenza prevention behaviors such as facemask wearing and vaccination using the TPB [14, 15, 20], attitudes, subjective norms, and perceived behavior control had positive, significant, and medium-to-large associations with the intentions of wearing facemasks in Hong Kong older adults. Among these three variables, perceived behavioral control had the strongest effects relative to attitudes and subjective norms. It was also the most prominent variable in the transmission of indirect effects of autonomous motivation on intention. Given that the TPB is, by and large, an integration of the personal (i.e., attitude) and social (i.e., subjective norm) antecedents of intention from the reasoned action approach [39, 40], and the self-efficacy related component from social cognitive theory [41], the increase of older adults’ personal resources (e.g., self-efficacy) and the decrease of the hindrance from barriers (e.g., easy access to facemasks when needed) seem to be most efficacious in accounting for variance in intentions to engage in the focal health behavior. Current findings provide formative evidence to support interventions to facilitate greater confidence in wearing facemasks and to overcome barriers to doing so.

The significant and positive direct effects of perceived autonomy support on attitude and perceived behavioral control indicated that a supportive motivational climate may affect the belief-based TPB variables. This is in line with previous studies [20, 24, 25, 27] and it further corroborates the integrated model [22, 23]. That is, autonomous motivation from SDT underpins the belief-based social cognitive variables of TPB, and the current study extends this to facemask wearing to prevent influenza transmission. The prominent role of control-related beliefs for this health behavior, and its mediating role in translating autonomous motives into intentions, is consistent with other theories which feature self-efficacy as a prominent component [42]. It is also consistent with previous research that has adopted the integrated model and demonstrated that perceived behavioral control is a key predictor and mediating factor of the influence of autonomous motivation on intentions [43].

The intentions of facemask wearing alone may be insufficient or unimportant when it comes to engaging in this particular behavior. In a post hoc analysis, we found that the majority of participants (89.05%) scored above the mid-point on the intention scale but only two fifths of them chose to wear the face masks (39.40%) in our scenario. Considering these data, it would be reasonable to assume that the majority of participants should be considered inclined abstainers or unsuccessful intenders [30, 44], or have weak or unstable [45‐47] intentions. Given that older adults’ facemask wearing habits, knowledge of whether wearing facemasks can prevent influenza, frequency of influenza over the previous 6 months, and perceived susceptibility to influenza were controlled for, it is reasonable to conclude that the current findings are consistent with the well-documented intention-behavior gap.

A number of factors may account for this ‘gap’. First, it emerged from the qualitative debriefing session conducted after the study had been completed that many of the participants chose not to wear a facemask because the interviewer was wearing a facemask, and they believed that they were, therefore, safe. Others reported that they were put off wearing facemasks themselves because it would make it difficult or uncomfortable to breathe, a notable barrier. As this is anecdotal evidence rather than a formal assessment, it may be important in future research to examine participants’ specific beliefs regarding the behavior, rather than global or direct measures of attitudes and perceived behavioral control. Such an approach may identify the influential behavioral (e.g., beliefs that people with illnesses will wear facemasks to reduce risk to others) and control (e.g., wearing masks makes it difficult or uncomfortable to breathe) beliefs that may influence the formation of intentions. Omission the beliefs that precede intentions aside, it may also be important to study the underlying volitional mechanisms from dual-phase models of action like the model of action phases [48] and the Health Action Process Approach [44] that determine the enactment of intentions. Incorporation of volitional components like implementation intentions, action planning, and coping planning into the study of facemask wearing may assist in identifying the possible moderators of the intention behavior relation.

Similarly, it may also be important to look at the implicit, non-conscious or ‘automatic’ influences on action. Research has indicated that self-reported habit predicts behavior as do implicitly measured attitudes and motives [49‐51]. Such influences likely account for action independent of the intentional route to behavior, and therefore determines action beyond an individual’s awareness. The likely process involved is that the implicit beliefs are associated with schema, or memory structures, which outline patterns of action which have been well learned and reinforced over time. Such patterns are activated when the implicit belief or attitude is cued, and, as a consequence, an individual’s behavior is initiated automatically [27, 52]. This does not mean that individuals act like ‘automatons’, blindly carrying out actions. The actual behavior may involve considerable planning and effort, it is just that the behavioral pattern is set in motion in a quick, efficient manner making the behavior subjectively very easy to enact. Implicit beliefs are also more likely to facilitate the enactment of the intended action, or failing to do the action, if it is either rewarding or relatively easy to do. Future research should examine individuals’ implicit beliefs toward illness and implicit beliefs toward the behavior, with respect to facemask wearing and assess their contribution to action. This would be consistent with further integrated models that have attempted to incorporate implicit components alongside motivational and volitional components of action in a unified framework [23].

The current study has numerous strengths including a focus on an important health-related behavior aimed at preventing influenza infection in a vulnerable population and has the potential to save lives, recruitment of a hard-to-reach population of older adults, adoption of an integrated theoretical model that has shown efficacy in predicting health behavior, and the use of valid and reliably instruments and a prospective design, and the use of an externally-verified observational measure of behavior instead of self-reports which have inherent biases.

Although our findings offer insight into the motivational processes that relate to facemask use in Hong Kong older adults to prevent influenza transmission during peak season, a number of limitations of the study should be identified and discussed. First, participants in the current study were ethnically and geographically homogeneous since they were recruited from the same senior center. All participants were Hong Kong Chinese, and most of them were female. In future, it would be important to replicate the findings in more diverse sample of older adults to provide further evidence on the validity and generalizability of study findings. Secondly, inference of causal relations among variables in the integrated model cannot be made even with the adoption of a prospective design. The causal direction of effects in the model can only be inferred from the theory rather than the data [53], it is, therefore, imperative that researchers also adopt experimental paradigms to manipulate key variables in the nomological network proposed by the theory and observe their effects on outcomes. Finally, the current study did not conduct an a priori statistical power analysis and adopted a manifest rather than latent variable approach given the relatively small sample. Future research should seek to conduct a power analysis beforehand, collect data from a larger sample and use a latent variable modeling approach which would control for measurement error [54].