Introduction

Osteoarthritis (OA), a degenerative joint disease, frequently affects the knees. It is one of the most common causes of pain in older individuals [1]. Moderate-intensity physical activity, such as walking, is an effective intervention and has been linked to numerous health benefits [2]. However, people affected by this chronic health condition tend to avoid physical activity [3]. Therefore, it is essential to understand why this population commonly chooses to not adhere to a physical activity [4, 5] that could help circumvent the deleterious effects of OA on functional ability. As demonstrated in a recent systematic review of Loew et al. [2], studies involving physical activity interventions in individuals with OA reported low adherence rates ranging from 27 to 64 % [2]. Adherence is considered a key criterion to evaluate the therapeutic effectiveness of an exercise program. It refers to the extent to which a person follows and accepts a treatment recommended by health professionals [6, 7] and is able to successfully reach the therapeutic goals [7]. Poor adherence indicates that important internal as well as external barriers in the implementation or successful completion of a physical activity intervention exist. The success of implementing an evidence-based intervention may perhaps be affected by factors related to adherence [6]. Though previous researchers identified adherence factors impacting arthritis [8–10], a clear consensus of influencing factors to improve adherence has not yet been established. Thereby, there is a need to identify the factors related to adherence and to adopt new tailored strategies to promote the adherence of physical activity interventions or programs.

To date, there remains a significant gap in knowledge regarding the process through which individuals with OA fail to regulate physical activity over time. To address this knowledge gap, two main theoretical frameworks have been proposed to identify potential factors affecting adherence to physical activity in the management of chronic diseases [11–13]. The social cognitive theory of Bandura has shown that personal characteristics contribute to an individual’s level of motivation to adhere to physical activity [12–14]. Behavior is regarded as a reciprocal interaction between personal and environmental components [15]. The latter is defined as social or physical factors that facilitate or hinder behavior. Given the importance of these concepts, a framework pertaining to physical activity behavior change was developed by the World Health Organization (WHO) [13], based on the concepts of social cognitive theory [15]. The WHO published a conceptual framework to better illustrate how adherence is a multi-dimensional phenomenon not only defined exclusively by personal determinants but also by important environmental determinants [13]. In fact, adherence incorporates the broader notions of an external determinant called “concordance” (i.e., influence of the health professional on an individual’s treatment decision, while promoting harmony between the two parties) [16].

The five dimensions of adherence in this model illustrate that participant adherence is a multi-dimensional behavior construct, determined by five dimensions: (1) health system (e.g., community support, relationship with health professionals), (2) social/economic (e.g., social status, external environment), (3) therapy related (e.g., benefits, treatment effects), (4) condition related (e.g., illness related, level of physical/emotional disability), and (5) patient related (e.g., level of knowledge, beliefs) [13]. Therefore, this conceptual framework facilitates the theoretical understanding of adherence to physical activity behavior [13] and was used to guide the present study (see Fig. 1 for additional details).

Fig. 1
figure 1

Reproduced, with the permission of the publisher, from the World Health Organization ([13], Fig. 3, section II, chapter V, p. 41)

Full size image

Concerns regarding adherence still persist, even when the effectiveness of a physical activity intervention has been established [2]. When participants are not engaged in the clinical decision-making, they may feel less empowered, resulting in decreased adherence [12]. The treatment preferences of individuals thereby need to be considered, allowing the participants to express themselves. In fact, the term “participant exercise preference” (PEP) refers to the individual’s personal expression of a value following consideration of benefits and risks of the interventions proposed, based on his/her values, beliefs, and needs [17]. PEP can be used as a knowledge transfer strategy to implement an evidence-based intervention. This strategy has not yet been applied to an aerobic walking program trial [18] nor has it been studied among older people with OA and investigated with adherence as the primary outcome [19]. Given these gaps and the relevance of the PEP factor, adherence is considered an essential outcome to successfully transfer evidence-based guidelines into clinical practice and patient care but remains challenging for stakeholders [20].

Aim

The general objective of the present study was to identify potential factors that could affect adherence and consequently influence the implementation of an evidence-based structured walking program [2], among older individuals diagnosed with mild to moderate knee OA. Since the primary outcome measured in the main study—The PEP trial [21]—was walking adherence, we expected to encourage OA participants to successfully adhere to an evidence-based effective walking program, by implementing a PEP strategy based on the evidence-based clinical practice guidelines [2]. In the present study, we evaluated three main research questions:

  1. 1.

    Question 1: What are the most potential factors related to walking adherence?

  2. 2.

    Question 2a: What are the five most potential factors that best describe each of the five dimensions of adherence, based on the WHO conceptual framework?

    Question 2b: What is the potential influence of participants’ preference as it relates to walking adherence over and above the factors identified within each dimension of adherence, based on the WHO conceptual framework?

Materials and methods

Protocol

This study describes the findings from the first 3 months of a larger 9-month study using a preference study design [21]. The main objective of the larger study was to evaluate the effect of PEP, by examining the hypothesis that participants who followed their preferred intervention (supervised vs. unsupervised aerobic walking program) would be more likely to adhere throughout the 9-month study period. The inclusion criteria were participants (1) diagnosed with mild to moderate OA of the knee, (2) aged between 55 and 80 years old [1], (3) able to walk for a minimum of 20 min, (4) available to walk 3 times/week, and (5) give informed consent. A total of 69 adults were recruited (50 women (72.4 %) and 19 men (27.5 %)). Participants were stratified on whether they did or did not indicate a preference for supervision related to the walking program ((1) preference for supervised, (2) preference for unsupervised, or (3) no preference). The reader is referred to the larger study for a more detailed overview of the study methodology [21].

Primary outcomes

Walking adherence

Walking adherence was monitored as a percentage of the number of walking sessions attended and completed by each participant divided by the number of walking sessions recommended in the Ottawa Panel guidelines [2, 5]. Participants who completed at least on average two of three sessions (66 %) of the prescribed walking sessions per week were considered as adherent. This clinically relevant cut-off point was selected based on a previous physical activity study in which significant changes in health outcomes were demonstrated (e.g., improvement in pain, function, and overall health) among an OA population [22]. A review of aerobic activities for individuals with arthritis confirmed that walking demonstrated the highest clinical improvements with an adherence rate ranging between 68 and 88 %, compared with other modes of aerobic exercises (e.g., swimming, cycling, dance) [23]. To the best of our knowledge, only one study has used a participation threshold of 100 % [24].

Adherence factors

The purpose of this exploratory behavioral study was to identify factors that could potentially affect adherence and consequently influence the implementation of an evidence-based structured walking program [2]. To our knowledge, no detailed survey pertaining to physical activity adherence factors for individuals with knee OA has been published. To this end, a self-reported survey was developed based on two main sources.

To date, few studies have provide a thorough description of adherence factors, with the majority either only provided keywords for adherence concepts or described factors without developing a survey [6, 25–27]. Nonetheless, some adherence factors (e.g., self-efficacy, motivation) have been recognized in the literature [8–10, 12–14, 17, 18] and were considered to develop this self-reported survey.

Since the literature on determinants influencing adherence to walking programs for individuals with OA is very limited, participants were therefore asked to complete an initial evaluation. The participants had to identify (a) barriers that limit their participation in physical activity and (2) facilitators that help them engage in physical activity. These responses were integrated when developing the proposed survey on adherence factors.

Whether the adherence factors were identified from the literature or by the participants, they were categorized according to the WHO’s conceptual framework which integrates five key concepts (see Fig. 1) [13]: (1) social or economic, (2) health system, (3) therapy related, (4) condition related, and (5) patient related. See Table 1 for a list of all the potential adherence factors identified.

Table 1 A summary of all the potential adherence influencing factors selected are listed below, where they are separated into five categories, based on the conceptual framework presented by WHO [13]
Full size table

Data collection

Data were collected using the aforementioned survey. The compilation of data was verified independently by two members of the research staff. For the PEP, three types of preference were collected:

  1. 1.

    Initial preference: at baseline, preference was evaluated where the participants had to respond “yes” or “no”: Do you have a preference for participating in a supervised or unsupervised walking program?

  2. 2.

    Preferred choice: at baseline, the preferred choice was also examined, where the participants had to respond “yes” or “no”: Did you receive your preferred group?

  3. 3.

    Preference at 3 months: participants enrolled in the study completed the online survey at 3 months after enrolment to a walking program. They responded whether the preference factor (“I was enrolled in my preferred group”) influenced their adherence to the study negatively (−1), had no impact (0), or positively (+1).

Data analyses

The data analysis focused on the evaluation of factors identified as most important to potentially predict the adherence of older adults diagnosed with knee OA participating in an evidence-based walking program. The dependent variable (adherence) was considered as a dichotomous binary variable since it was represented by two categories: 1, participant successfully attended and completed the walking program; 0, participant did not successfully attended but completed the walking program.

Adherence was influenced by variations in the independent variables, following a 3-point Likert ordinal scale: negative (−1), no impact (0), or positive (+1). This common rating scale was concise and allowed the participants to better express their opinion [28]. Adherence with regard to the influencing factors was determined using a polychotomous binary logistic regression model [29]. It evaluated the probability to adhere to the walking program in accordance to the explanatory variables, in this case, all the identified influencing factors. A multivariate analysis was performed using the variables that indicated a p value of ≤0.2 in order to identify the predictor variables, i.e., only the most important potential factors. As frequently selected in behavioral and social sciences research, this p value was used since a large number of categorical variables were entered in the equation [29]. Odds ratio could not be interpreted as a relative risk because adherence is not a rare event [29]. To be significant, the odds ratio of a factor needed to be in the range of 95 % (confidence interval (CI)), without including 1 within this range.

To assess research question 1, the most potential factors related to walking adherence were identified, among all the factors included in the online survey, using a univariate quantitative analysis to look at each factor separately. A set of the five most important variables reaching statistical significance was conserved.

To assess research question 2a, the five most important factors that best describe adherence were identified using a univariate analysis according to each dimension of adherence (WHO’s conceptual framework [13]). A set of the five most potential variables was taken to do a stepwise procedure (one factor represented each dimension). To assess research question 2b, we determined the importance of PEP related to walking adherence, in conjunction with the five same potential variables, selected in question 2a, in order to perform a stepwise procedure with the three types of preference.

The most important potential factors related to walking adherence were assessed based on all of the participants who adhered (successfully or not), and those who dropped out (n = 69) using the intention-to-treat approach, since the factors affecting an individual’s decision to drop out [30] were significantly related to the factors identified as important for adherence, as demonstrated by a Chi-square test (Chi-square = 39.403, p < 0.001). In light of the fact that 20 participants withdrew from the study at 3 months, secondary analyses were performed with the 49 participants who completed the 3-month intervention.

Results

At baseline, 54 out of 69 (78 %) participants indicated a preference for participating in a supervised (20/54 (37 %)) or unsupervised (34/54 (63 %)) walking program. Based on their stated exercise preference, participants were then assigned to one of the two modes of supervision for the effective walking program. Twenty-nine participants who indicated a preference were able to participate in their preferred walking supervision mode (54 %). Out of the 69 participants, 43 participants (62.3 %) successfully adhered to the walking program.

Among the 20 participants (18 women) who dropped out within 3 months, 19 did not adhere successfully prior to dropping out (95 %). At baseline, 95 % (19/20) of the participants indicated a preference for participating in a supervised or unsupervised walking program, and 11 of them who stated a preference did not obtain their preferred group after being assigned (58 %).

Since the range of value (95 % CI) included 1 in the results, evidence of the odd ratios was not statistically significant and tendencies were therefore suggested for each factor that were potentially important. By using the intention-to-treat approach, the most potential factors related to adherence were identified, at 9 months (p value ≤0.2): (1) change in medication use, (2) work/volunteering schedule, and (3) fear (fear of falling, walking causing pain). Results indicated a tendency that participants were 20 % less likely to adhere if change in medication use (p value, 0.236) and work/volunteering schedule (p value, 0.158) both represented a negative influence. Participants were 31 % less likely to adhere if their fear of falling negatively impacted their walking adherence (p value, 0.117) (see Table 2).

Table 2 Most important potential factors related to adherence
Full size table

With regard to the 49 participants who completed the study at 3 months, 42 (85.7 %) successfully adhered to the walking program. Thirty-five (71 %) participants indicated a preference for participating in a supervised or unsupervised walking program, in which 21 of them were able to participate in their preferred group (60 %). At the end of the study program, the majority of participants (90 %) specified that participating in their preferred group had a positive influence on their adherence (44/49).

Table 3 shows the results according to research question 1. The five most important factors potentially influencing the adherence rate as determined from the survey response at 3 months were (p value ≤0.2) (1) level of satisfaction during walking and (2) emotional involvement. Specifically, participants were 3.4 times more likely to adhere if their level of satisfaction influenced their walking adherence in a positive way (p value, 0.293) Moreover, we observed a tendency that people were 53 % less likely to adhere to the walking program if they characterized the factor “emotional involvement” (i.e., attitude and unbearable emotions toward physical activity) as having a negative influence on their adherence.

Table 3 Most important potential factors related to adherence
Full size table

With regard to research question 2a, the five most important potential factors influencing walking adherence representing each adherence dimension [13] were (p value ≤0.2) (see Table 4):

Table 4 Most potential factors identified to better understand the five dimensions of adherence compared with the PEP factor
Full size table
  1. (A)

    Environmental factors:

  2. 1.

    Health system: participants were 2.9 times more likely to adhere if they felt being supervised by an exercise therapist had a positive impact on adherence (p value, 0.230).

  3. 2.

    Social/economic: participants who indicated that family or friend support had no impact on their walking adherence were 3.7 times more likely to adhere (p value, 0.246).

  4. 3.

    Therapy related: participants were 3.6 times less likely to adhere if they perceived a change in their medication use during the study. It seemed to have a negative influence on their adherence since it was a sign of more pain or additional health problems (p value, 0.073).

  5. (B)

    Personal factors:

  6. (i)

    4. Condition related: participants were 3.1 times less likely to adhere if they considered their physical fitness level as a negative influence on walking adherence (p value, 0.191).

  7. (ii)

    5. Patient related: participants were 11 times more likely to adhere if unbearable feelings, to the point of losing one’s ability, had no influence on their walking adherence (p value: 0.133).

With respect to research question 2b, Table 4 details the influence of preference compared with the five most potential adherence factors identified above (according to each dimension of adherence; p value ≤0.2). Preference was not considered the most important factor; it pertained to medication use and emotional involvement. The logistic regression revealed a tendency that participants who indicated a strong preference for being supervised or unsupervised were 46 % less likely to adhere if emotional involvement seemed to have a negative influence on walking adherence, in comparison with preference at baseline. Participants who had a preference and did not obtain their choice were 49 % less likely to adhere if again the factor “emotional involvement” had a negative influence on their adherence. At 3 months, results showed a tendency that participants were 77 % less likely to adhere if change in medication use had a negative influence on adherence.

Discussion

The purpose of this study was to identify factors that could potentially affect adherence and consequently influence the implementation of an evidence-based structured walking program [2], among older adults diagnosed with knee OA. We found that the five following most potential factors were deemed important to walking adherence and were therefore linked to the overall success of the walking program: participants’ (1) level of satisfaction during walking, (2) emotional involvement, ( 3) fear, (4) physical fitness level, and (5) change in medication use.

It has been established that adherence to physical activity is a multi-dimensional behavior construct, determined by the interaction between five dimensions of adherence [13]. The conceptual framework by the WHO [13] stipulates that physical activity behavioral change is influenced by both personal determinants and environmental factors, without identifying precise factors to better understand each dimension of adherence. Preliminary results delineate the key factors that had a greatest influence on adherence to physical activity in the context of the five dimensions. This was an innovative feature of the study, as it was never examined in studies with people diagnosed with OA. These are as follows:

  1. (A)

    Environmental factors:

  2. 1.

    Health system: supervision by exercise therapists’ seemed to have a positive impact on the participant’s adherence.

  3. 2.

    Social/economic: adherence seemed to be positively influenced if the participant received support from his family or friends.

  4. 3.

    Therapy related: change in the medication use seemed to negatively influence the participant’s adherence.

  1. (B)

    Personal factors:

  2. 4.

    Condition related: high physical fitness level seemed to positively affect adherence.

  3. 5.

    Patient related: absence of emotional involvement seemed to have a positive impact on the participant’s adherence.

The implementation of the innovative strategy using PEP was essential to address the recommendations of the Ottawa Panel guidelines, in order to successfully transfer evidence-based clinical practice guidelines into clinical practice [20]. Even though preference was not considered the most important factor, walking adherence was dependent on the PEP. Participants who stated that preference had a positive influence, were two times more likely to adhere (n = 69). Preference should be considered as an important influencing factor related to dropout as well as adherence.

This exploratory study supports the results of other studies examining adherence and drop out related to physical activity programs [17–19, 31]. Crandall et al. [32] suggested that considering preference when determining which type of intervention the participant should follow, can improve adherence. Our aforementioned findings are in accordance with those of Henry et al.’s [33] since, the authors confirmed that perception of individual physical capacity, beliefs about physical activity and motivational approaches should be considered and examined when working with a physical activity population as they can influence their behavior [33]. According to several other authors, self-efficacy [14, 25, 34] and motivation [25, 34–39] are factors that can significantly influence adherence, albeit these same factors did not reach statistical significance in our analyses even though they were evaluated.

Moreover, a new observation emerged from the analyses. Participants who perceived a change in their medication use during the study seemed to adhere poorly. It is plausible that a change in their medication prescription implies poorer health status, and as a consequence, they do not view physical activity as being able to provide any health benefits. Without perceived benefits on health or beliefs in physical activity, individuals tend to be less motivated, which influences their adherence. More research is warranted to examine this observation.

This study was novel in that it considered many factors to potentially impact adherence to physical activity, and a quantitative methodology was used to define the relative importance of each of these potential factors. However, since the majority of the factors influenced each other, a strong multicollinearity could be noticed, which may have decreased the significance of each factor. A strong multicollinearity increases the standard error and confidence intervals for the coefficients. Thus, multicollinearity may have to lead to some statistically non-significant analyses. To address this issue, a multivariate analysis commonly employed in behavioral and social sciences research was performed using only the variables that indicated a p value of ≤0.2, in order to identify the predictor variables, i.e., most potential factors. This p value was selected since a large number of categorical variables were entered in the equation. While the sample size was relatively small, a power calculation was performed prior to the study to justify the sample size. Our findings allowed for the development of a new relevant survey that can guide health professionals in their practice. The survey can be used by health professionals to assess their patients’ perceptions regarding adherence factors, and therefore adopt a more patient-centered approach. Ultimately, the goal is to help researchers and health care providers to better promote long-term adherence to a physical activity program and manage these potential factors in order to enable individuals with mild to moderate OA of the knee to participate in these exercise programs. Health professionals should also assess and respect their patients’ preferences for treatments [40, 41]. This will, in turn, allow the health care practitioners to adopt a patient-centered approach, to better accommodate the need of their patients, and increase their chance of success [41].

Conclusion

This exploratory study identified different potential factors, such as supervision, social support, medication use, fitness level, and emotional involvement that can influence walking adherence in clinical practice and research. Future studies should adopt an equally rigorous methodology, but factors which are inter-related and have strong associations should be removed (by performing a multicollinearity test) to avoid multi-collinearity and a larger sample size must be considered to confirm the factors that influence adherence.

In conclusion, there are many factors influencing adherence to walking program designed specifically for individuals with knee OA. These should be considered when developing physical activity interventions with this population. The most important ones included emotional involvement, mode of supervision, family/friends support, medication intake, and physical fitness level.