Title : Motivational , Volitional and Multiple Goal Predictors of Walking in People with Type 2 Diabetes

Background: Type 2 diabetes is a major public health problem. Effective diabetes self-management involves people engaging in multiple health behaviours, including physical activity. Walking is an effective, accessible and inexpensive form of physical activity, yet many people with Type 2 diabetes do not meet recommended levels. The present study aimed to: 1) identify demographic, motivational and volitional factors predictive of walking in people with Type 2 diabetes mellitus, and 2) test whether accounting for the perceived impact of other goal pursuits (goal facilitation and goal conflict) improved the prediction of walking. Methods: A theory-based cross-sectional study using the Health Action Process Approach was conducted in adults with Type 2 diabetes across Scotland. Assuming a 50% response rate 1000 questionnaires were mailed to achieve the target sample size (N = 500). Demographic information was collected, and intentional (outcome expectations, social support, risk perceptions), motivational (intention, self-efficacy), volitional (action planning, action control) and multiple goal (goal conflict, goal facilitation) factors were assessed as predictors of physical activity in general and walking specifically. Results: The final sample comprised 411 respondents. The majority (60%) were non-adherent to physical activity recommendations. Of 411 respondents, 356 provided walking data. Body Mass Index and age were the only demographic and anthropometric factors predictive of walking (overall R2 = 0.04). When motivational factors were added, intention and self-efficacy added to the prediction (overall R2 = 0.07). When volitional factors were added, only action control was predictive of walking (overall R2 = 0.08). Finally, goal facilitation explained an additional 7% variance in walking when added to the model (final overall R2 = 0.15). Conclusion: There was low adherence with physical activity recommendations in general and walking in particular. When testing predictors of motivational, volitional and competing goal constructs together, action control and goal facilitation emerged as predictors of walking. Future research should consider how walking can be embedded synergistically alongside other goal pursuits and how action control may help to ensure that they are pursued.

BACKGROUND 51 Diabetes is a common non-communicable chronic disease. The global prevalence of 8.3% is 52 expected to increase to 10.1% by 2030 (IDF, 2013). In Scotland, the prevalence of diabetes is 4.7%, 53 slightly above the UK average (S D. S. M. Group, 2012). Almost 90% of patients with diabetes 54 have Type 2 diabetes (WHO, 2006) and their life expectancy is up to 10 years less than people 55 without Type 2 diabetes . 56 Diabetes is a chronic, metabolic disease characterized by increased levels of blood sugar. Diabetes 57 occurs either when the pancreas produces no or insufficient insulin, or when the body cannot 58 effectively use the insulin it produces. Type 2 diabetes results from the body's insufficient 59 production and/or ineffective use of insulin. Hyperglycaemia (an increased concentration of glucose 60 in the blood) is a common effect of uncontrolled diabetes and over time leads to serious damage to 61 M A N U S C R I P T

Towards multiple behaviour approaches 159
Most popular social cognition models of health behaviour focus on understanding a single health 160 behaviour at a time. The ecological validity of such an approach has increasingly been questioned 161 (Presseau, Tait, Johnston, Francis, & Sniehotta, 2013). In everyday life, individuals pursue multiple 162 goals and perform multiple behaviours alongside the single health behaviour that is typically the 163 focus of tests of behavioural theory. These goal pursuits compete for time and energy such that 164 pursuit of some may help and/or hinder the pursuit of a particular health behaviour, such as physical 165 activity in general or walking specifically. 166

167
The extant literature has predominantly managed the concept of considering multiple goals by 168 focusing on the impact of goal conflict on health behaviour. Goal conflict can be described as M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
Running head: GOAL FACILITATION TYPE 2 DIABETES 8 occurring when the pursuit of multiple personal goals leads to situations where they interfere with 170 one another. For instance, working, childcare, relaxing and socialising may be common personal 171 goals that have the potential to conflict with walking by taking available leisure time, energy or 172 other resources that might otherwise be used go for a walk. The evidence on the link between goal 173 conflict on physical activity-related behaviour is mixed. There is a lack of support for this 174 relationship in between-subject predictive studies (Li & Chan, 2008;Presseau, Sniehotta, Francis, & 175 Gebhardt, 2010;Riediger & Freund, 2004). However, a study investigating actual time spent 176 pursuing goals that conflict with physical activity within-subjects was negatively predictive of 177 objectively assessed physical activity (Presseau et al., 2013), and a study investigating goal conflict 178 in more resource constrained contexts has also shown that goal conflict is negatively predictive of 179 behaviour (Presseau, Francis, Campbell, &Sniehotta, 2011). As people with Type 2 diabetes engage 180 in self-management regimens that inherently involve pursuing multiple behaviours and goals, it is 181 plausible that goal conflict may be a useful additional construct in this population. 182 183 By comparison, goal facilitation has received less research than goal conflict, yet is recurrently 184 shown to be predictive of physical activity-related behaviours. Goal facilitation involves instances 185 where the pursuit of other personal goals sets the stage or makes it more likely that physical activity 186 will take place (e.g. socialising with friends that involves walking in the park), or inherently 187 involves physical activity (e.g. commuting to work can be facilitative of physical activity when 188 involving active travel). The presumption is that the more one's other personal goals are aligned 189 with physical activity, the greater the physical activity. Goal facilitation has been demonstrated to 190 positively predict physical activity (Riediger & Freund, 2004), a relationship that is maintained 191 even when controlling for intention and self-efficacy (Presseau et al., 2010). However, it is not clear 192 whether these relationships persist when accounting for volitional (planning, action control) 193 processes, which could in themselves involve managing competing goals. For instance, action M A N U S C R I P T  (Presseau, Boyd, Francis, & Sniehotta, 2015 (Michie et al., 2005) to identify which theoretical domains and constructs were relevant to 216 understanding the adherence of people with Type 2 diabetes to physical activity recommendations 217 in general and walking in particular. The results were used to identify relevant items that were 218 included in a draft questionnaire. The questionnaire explored physical activity in general, and 219 walking in particular. Pre-piloting of the questionnaire was undertaken with five people using the 220 "think aloud" method (Jones, 1989;Lundgren-Laine & Salantera, 2010) where participants 221 verbalised their thoughts. Three participants with Type 2 diabetes were recruited from the Scottish 222 Diabetes Research Network (SDRN) (see later) and three were colleagues with Type 2 diabetes in 223 the Centre of Academic Primary. Minor revisions were made prior to the pilot study. The 224 questionnaire was piloted with 50 people with Type 2 diabetes, selected randomly from the SDRN 225 list, replicating the distribution process planned for the main survey (pre-notification letter, 226 questionnaire and covering letter, and a reminder letter and replacement questionnaire after two 227 weeks). To assess test-retest reliability, respondents were sent a second copy of the questionnaire 228 two weeks after returning their first questionnaire. 229 230

Sample and Recruitment 231
The sample size for this study was influenced by two factors: 1) having acceptable precision for the 232 M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
Running head: GOAL FACILITATION TYPE 2 DIABETES 11 estimation of adherence with physical activity (any precision within ± 5% that would be clinically 233 and statistically acceptable) and 2) the resources (time and money) available to undertake the 234 research. To achieve a balance between these two items, a sample size of 500 patients was required. 235 As previous research has shown compliance with physical activity to range from 19-30% (midpoint: 236 25%) (Kamiya et al., 1995;Kravitz et al., 1993), this allowed estimation of adherence with physical 237 activity of 25% with precision within ± 3.8% (95% CI 21.2% to 28.8%). Previous research in 238 community samples indicated a 50% response rate was likely, therefore 1000 questionnaires were 239 mailed to achieve the target of 500 evaluable responses. 240 Participants were recruited from the Scottish Diabetes Research Network (SDRN), a register of 241 patients with diabetes in Scotland who have consented to be contacted about potential participation 242 in research studies (SDRN, 2010). All SDRN registered patients in Grampian (n=388) were 243 identified and invited to participate, supplemented by a random sample of 612 of the 1279 patients 244 registered in Tayside exclusive of those who had taken part in the pilot study. A pre-notification 245 letter with a reply slip, that they could use if they did not want any further communication,was sent 246 to these 1000 patients two weeks before the questionnaire and accompanying invitation letter were 247 mailed. Two weeks after the first mailing, a reminder letter and another copy of the same 248 questionnaire were sent to non-respondents. The questionnaire was piloted with 50 people with 249 Type 2 diabetes, selected randomly from the SDRN list, replicating the distribution process planned 250 for the main survey (pre-notification letter, questionnaire and covering letter, and a reminder letter 251 and replacement questionnaire after two weeks). To assess test-retest reliability, respondents were 252 sent a second copy of the questionnaire two weeks after returning their first questionnaire. Questionnaire (IPAQ) (IPAQ, 2002). It measures physical activity over a short time frame. The 258 IPAQ was developed by consensus in 1998-1999 with support from the WHO to enable the cross-259 national assessment of physical activity in adults aged 18-65 years (Craig et al., 2003;Macfarlane, 260 Lee, Ho, Chan, & Chan, 2007;Papathanasiou et al., 2010). The short format of the IPAQ asks about 261 three types of activity in the four domains. Walking, moderate-intensity activities and vigorous-262 intensity activities are the specific types of activity which are assessed by the IPAQ short form 263 (IPAQ 2005). This version generates a total score by summation of the duration (in minutes per 264 day) and frequency (days) of walking, moderate-intensity activities and vigorous-intensity 265 activities. The IPAQ measures energy as Metabolic Equivalent of Task (MET). The IPAQ has been 266 used in a number of international studies (Craig et al., 2003;Guthold, Ono, Strong, Chatterji, & 267 Morabia, 2008) and acceptable reliability and validity has been reported (Craig et al., 2003;268 Hagstromer, Oja, & Sjostrom, 2006;Hallal et al., 2010;Macfarlane et al., 2007;Papathanasiou et 269 al., 2010). An international reliability and validity test of the IPAQ was conducted in 14 centres in 270 12 countries and reported that it has acceptable reliability and validity at least equal to other 271 established self-report tools for physical activity in diverse populations of 18-65 years (Craig et al., 272 2003). We focused specifically upon understanding predictors of walking as the primary outcome of 273 interest given the wording of our predictors focused upon walking. Walking was assessed using the 274 total time or energy (150 minutes or >600 MET minutes/week) spent on walking measured by the 275 IPAQ and served as the dependent variable in all predictive analyses. However we also aimed to 276 describe overall adherence to physical activity recommendations. 277

278
Adherence to physical activity was assessed by comparison with two different recommendations. 279 walking, moderate-intensity or vigorous-intensity physical activities (IPAQ, 2002). According to 284 IPAQ, <600 MET minutes/week, 600-2999 MET minutes/week, and >3000 MET minutes/week 285 are considered as low, moderate and vigorous physical activity, respectively (IPAQ, 2002). 286 287 Predictors of walking 288 The questionnaire assessed a number of potential demographic and theoretical predictors of 289 walking: demographic variables, self-efficacy, outcome expectations, risk perceptions, intention, 290 action planning and control, social support, goal facilitation and goal conflict. The demographic 291 variables age, gender, education, and employment items) were defined using the England household 292 version of the 2001 Census questionnaire (OFNS, 2002). All theoretical items were worded 293 according to the TACT principle (Target, Action, Context, and Time), specifying the behaviour of 294 interest as: "To increase (my) own walking level by 20% during the normal daily routine in the 295 forthcoming month" and described in detail below. 296 297 Self-efficacy. Self-efficacy was assessed using six items ranging from 1 (strongly disagree) to 5 298 (strongly agree) in relation to perceived capability to increase walking despite the presence of 299 barriers (Schwarzer et al., 2003). The stem "I am confident that I can increase my walking by 20% 300 in the next month even if...." had response options such as: "the weather is bad", "it is hard for me 301 physically", "I do not have much time". 302 303 Outcome Expectations. Two facets of outcome expectations were assessed (Schwarzer et al., 2003), 304 13 Metabolic equivalent of task (MET) is a concept frequently used to show the amount of energy or oxygen the body uses during physical activity. One MET is equivalent to the energy or oxygen that the body uses at rest, or consuming 3.5 millilitres of oxygen /kg of body weight/minute (1 MET= 50 kcal/hours/m2 body surface area) (Davis & Wilbrn, 2003).

M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT
Running head: GOAL FACILITATION TYPE 2 DIABETES 14 with scores for each item ranging from 1 (not at all) to 4 (exactly true): there were six items to 305 assess positive outcome expectations, and three items to assess negative outcome expectations. The 306 stem "if I increase my walking by 20% in the next month ...." had response options such as: "I 307 would feel better afterwards", "it would take up a lot of time ". 308 309 Risk Perception. Risk perception refers to the respondent's belief about their vulnerability to health 310 problems, or specifically in this patient group for their diabetes to worsen (Schwarzer et al., 2003). 311 Absolute and relative vulnerability were assessed using six items with response options ranging 312 from 1 (strongly disagree) to 7 (strongly agree). The items measuring absolute vulnerability had a 313 stem "If I am not physically active…" and response options such as: "... I am concerned that my 314 health in general will become worse", "... I am concerned that my diabetes in general will become 315 worse", "...I will worry about getting a serious medical condition". The items measuring relative 316 vulnerability had a stem "If I am not physically active…" comparing myself with an average person 317 of my age and sex, then I will be at higher risk of ..." and response options such as: " ... my diabetes 318 gets worse", "...having a serious medical condition". 319 320 Intention. Intention refers to a participant's intention to increase walking (Schwarzer et al., 2003) 321 and was assessed by four items with response options ranging from 1 (completely disagree) to 5 322 (totally agree). Intention was measured by items such as "I intend to walk more in the next month" 323 and " I am motivated to walk more to improve my health in general". 324 325 Action Planning. Action planning consisted of items assessing the extent to which participants had a 326 plan about when, where, and how to increase their walking (Schwarzer et al., 2003). Action 327 planning was assessed using four items . All items had response options 328 ranging from 1 (completely disagree) to 4 (totally agree). The stem "I have made a specific plan Action Control. Action control refers to perceived self-monitoring, awareness of standards and 334 effort  to increase walking of participants. Action control was assessed using 335 six items and all items had response options ranging from 1 (strongly disagree) to 4 (strongly 336 agree). The stem "During the last week I ...." had response options such as:" …regularly thought 337 about my intention to be regularly physically active", " …I have consistently checked to see 338 whether I am physically active enough". 339 340 Social Support. Social support items assessed support from colleagues, friends and household 341 members to increase walking using a modified version of the Molloy social support tool (Molloy, 342 Dixon, Hamer, & Sniehotta, 2010). All items (17 items) had response options ranging from 1 343 (strongly disagree) to 7 (strongly agree). Social support (friends/colleague) was measured by items 344 such as "I have a friend/colleague who thinks that I should increase my walking", and "I have a 345 friend/colleague who encourages me to increase my walking". Social support (household) was 346 measured by items such as "I have somebody to encourage me to increase my walking on the 347 regular basis", and "I have somebody to walk with me". 348 349 Goal Conflict and Goal Facilitation. Goal conflict (5 items) and goal facilitation (3 items) items 350 focus on the extent that a participant's personal goals conflicted with physical activity and were 351 adapted from general goal conflict and facilitation scales (Riediger & Freund, 2004). All the items 352 had response options ranging from 1 (never, not at all, or completely disagree) to 5 (very often, a 353 great deal, or completely agree). The items measuring goal conflict consisted of a stem "How often The primary outcome measure was the IPAQ walking criterion (MET minutes/week). A sensitivity 368 analysis was conducted using total MET minute/week. The extent of missing data varied across 369 variables. The variables with the greatest and smallest amount of missing data were walking level 370 (13.4%), and diabetes management method (2.1%). We used multiple imputation (Klebanoff & 371 Cole, 2008) to account for missing data which addresses missing data issues in the most robust 372 manner possible. All model testing was conducted on multiple imputed data and results presented as 373 pooled estimates. Hierarchical multiple regression analyses were conducted to test the sequential 374 contribution of demographic, motivational, volitional and multiple goal constructs as predictors of 375 walking. 376   (Table  406 2), but the median time (hours/week) spent for both moderate and vigorous physical activity was 407 zero ( Table 1). The median duration of walking was 5.25 hours per week. The proportion of total 408 physical activity reported as walking was 65.6%. 409 <Insert Table 1 here> 410 <Insert Table 2 here> 411 As shown in Table 3, which presents findings for the 356 respondents providing walking data, BMI, 412 action planning, action control and goal facilitation were significantly associated with walking 413 behaviour, and outcome expectations, social support, risk perceptions, self-efficacy, action 414 planning, action control, and goal conflict were significantly associated with walking intention. The 415 Cronbach's alpha of different subscales of HAPA questionnaire are presented in Table 3 indicating 416 that most subscales of the questionnaire had a good internal consistency. The negative outcome 417 expectations scale was omitted from any analyses due to low observed internal consistency. 418 419 <Insert Table 3 here> 420

Predicting walking 421
The hierarchical multiple regression was conducted in four steps. First, demographic factors and 422 predictors of intention from the HAPA were included. Next, motivational factors from HAPA were 423 added, then volitional, and finally multiple goal constructs. At each step, we tested whether the 424 added factors contributed to explaining additional variance in walking beyond factors in the model 425 from the previous steps, and which specific constructs explained this additional variance. In Step 1 426 of the hierarchical multiple regression, walking was regressed against demographic factors (BMI, 427 age, sex) and HAPA theory-based predictors of intention (outcome expectations, social support and 428 risk perception). As shown in Table 4, only BMI and age predicted walking, explaining 3.7% of the M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
Running head: GOAL FACILITATION TYPE 2 DIABETES 19 variance in walking. In Step 2, HAPA motivational constructs (intention and self-efficacy) were 430 added, with intention and self-efficacy adding to the prediction (∆R 2 = 0.03). In Step 3, the 431 volitional constructs of action planning and action control were added, with only the latter adding 432 significantly to the prediction (∆R 2 = 0.01) and intention and self-efficacy no longer significantly 433 contributing to predicting behaviour. In Step 4, the multiple goal constructs of goal conflict and 434 goal facilitation were added, with the latter significantly adding to the prediction of behaviour (∆R 2 435 = 0.07) whilst action control no longer significantly predicted behaviour. 436 437 <Insert Table 4 here> 438

439
The study showed that the majority (60%) of Type 2 diabetic patients were non-adherent to physical 440 which showed that 61% of the general population aged 16 and over did not meet physical activity 447 recommendations. Other evidence suggests that patients with Type 2 diabetes may be even less 448 physically active than the general population (Morrato et al., 2007). This was also the finding of a 449 study in USA of 23,283 adults, which showed that only 39% of individuals with Type 2 diabetes 450 were physically active compared with 58% of those without diabetes (Morrato et al., 2007). 451

452
The median duration of walking reported in the current study was 5.25 hours per week (IQR 1.5, 453 12). The proportion of walking as a percentage of total physical activity was 65.6% suggesting that 454 M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
Running head: GOAL FACILITATION TYPE 2 DIABETES 20 in some cases walking was the main type of physical activity undertaken by patients. This finding 455 reflects the behaviour of the general adult population (Monteiro et al., 2003;Morris & Hardman, 456 1997); therefore developing and evaluating interventions to increase and maintain this behaviour are 457 important. Walking is a common, accessible, inexpensive Type of physical activity. Walking 458 provides diverse health benefits of physical activity with few adverse effects. There is a large body 459 of evidence about the positive effect of walking to improve health in people with Type 2 diabetes. 460 This suggests that focusing on walking as a form of physical activity to improve peoples' adherence 461 with physical activity recommendations is important and could be an effective way to improve 462 physical activity. 463

464
In terms of the existing literature one study conducted with cardiac rehabilitation patients, was 465 found that measured action control as a predictor of physical activity . That 466 study reported that each of the three factors of planning, self-efficacy and action control made 467 unique contributions to translating intention into action . A study conducted 468 in students confirmed associations specified by the HAPA at the intrapersonal level: outcome 469 expectancies and self-efficacy, but not risk awareness, were positively associated with intentions for 470 physical exercise. Physical activity was positively associated with intentions, self-efficacy, action 471 control, but not with action planning (Scholz, Keller, & Perren, 2009). These findings are in 472 accordance with the results of this current study. Another study conducted in Type 2 diabetic 473 patients participating in a Diabetes Self-Management Education (DSME) (Bonner, 2010) showed 474 that self-efficacy was the strongest predictor of behavioural intention, followed by positive outcome 475 expectancy. The study (Bonner, 2010) revealed that behavioural intention, but not self-efficacy and 476 action planning could significantly increase initiation of a minimum level of physical activity. 477

478
The current study showed some degree of support for the tenets of the HAPA, whilst demonstrating 479 M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
Running head: GOAL FACILITATION TYPE 2 DIABETES 21 the importance of considering multiple goal pursuit in people with Type 2 diabetes. The majority of 480 respondents did not engage in physical activity at recommended levels. Action control and goal 481 facilitation were shown to be predictors of physical activity when considered alongside other HAPA 482 and demographic factors. Findings in relation to the HAPA with respect to intention (step 2 of the 483 regression) and action control (step 3) were consistent with previous research (Sniehotta. et al., 484 2005) and extend these findings by demonstrating the role for multiple goals constructs on physical 485 activity (in this case, goal facilitation). Conversely we did not show a predictive role for action 486 planning and in step two, there is an unexpected negative predictive relationship between self-487 efficacy and walking behaviour, although this becomes insignificant when the additional predictors 488 in steps three and four are added. Both findings are at odds with the HAPA model and most of the 489 literature investigating these relationships . Self-efficacy showed no 490 significant bivariate relationships with walking which may be due to the fact that the target 491 behaviour was 'increasing walking by 20%' which equates to large absolute changes for more 492 active respondents. Moreover, self-efficacy was significantly correlated with intention, so that the 493 negative beta-coefficient in the second step of the hierarchical regression analysis may be reflective 494 of an artefact, a statistical suppressor effect. Action planning showed a weak bivariate correlation 495 with walking and was significantly correlated with action control so that when action control was 496 simultaneously controlled for, there was not a unique predictive relationship between action 497 planning and walking. In the final model, neither of these variables was significant. 498 499 Findings regarding multiple goal constructs are also consistent with earlier research showing that 500 perceived goal facilitation but not perceived goal conflict were predictive of physical activity 501 (Presseau et al., 2010;Presseau et al., 2013;Riediger & Freund, 2004). 502

503
There is now growing evidence across a range of studies with diverse populations that particularly 504 M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
Running head: GOAL FACILITATION TYPE 2 DIABETES 22 support the role of goal facilitation as a key factor in physical activity and, with the present study's 505 findings, walking specifically. Goal facilitation is an indicator of the extent to which a target 506 behaviour (in this case, walking) "fits" synergistically alongside the other behaviours and goals that 507 individuals pursue in daily life. Findings from this study continue to support the role of goal 508 facilitation and also underscore its potential importance in understanding health behaviours; indeed, 509 even when controlling for predominant theoretical constructs reported in the literature, the 510 relationship between goal facilitation and walking robustly accounted for additional variability in 511 walking. With increasing recognition of the importance of considering the wider context of multiple 512 goal pursuit when understanding performance of a given health behaviour, the present study further 513 contributes evidence suggesting that goal facilitation may be a key indicator in the move towards 514 developing models that explicitly account for the impact of multiple goal pursuit. 515 516 There is also mounting lack of support for the role of goal conflict in understanding physical 517 activity. There may be a range of reasons for this. For instance, when considering the totality of an 518 individual's goal pursuits, individuals may be better able to perceive helpful goal relationships than 519 conflicting ones. Individuals may not be aware of the extent that their competing goals interfere 520 with their physical activity. When using diaries to assess actual time spent in pursuit of goals that 521 conflict with physical activity over time, goal conflict has been shown to be predictive of 522 objectively assessed physical activity (Presseau et al., 2013). This suggests that measures of 523 perceived goal conflict may need to be supplemented with behavioural assessments. This also 524 presents opportunities for feedback interventions by showing individuals which of their behaviours 525 is most interfering with their physical activity. In addition, when focusing the goal pursuit context 526 to a specific time and place rather than all of everyday life, both goal conflict and goal facilitation 527 have been shown to predict behaviour (Presseau et al., 2011). planning -facilitation planning -in their walking intervention, which was successful in increasing 543 and maintaining the increased walking behaviour. Planning when, where and how to perform 544 behaviours may facilitate action. To some extent, these may be preparatory behaviours, but goal 545 facilitation encompasses the broader spectrum of valued goals pursued in everyday life and may not 546 necessarily be preparatory in nature, whereas preparatory behaviours may not have any intrinsic 547 value to the actor. Nevertheless, the functional similarities between preparatory behaviours and goal 548 facilitation are noteworthy and future research should consider these two constructs in more detail. 549 550

Strengths and limitations 551
The present study is strengthened by its large sample size, robust development and inclusion of 552 theoretical factors as determinants of walking. Although the sample of 411 (356 for the main 553 analysis) was slightly short of the target of 500, this did not impact substantially upon the precision 554 M A N U S C R I P T The study also had limitations. Firstly, the cross-sectional study design only allows association, and 559 not causation, to be inferred. While there is no obvious suggestion of multicollinearity, the modest 560 bivariate correlations between predictors in the model should be considered in interpreting the 561 relative contribution of predictors in the model, particularly with respect to factors which were not 562 zero-order correlations, and were not bivariately associated with walking but which were associated 563 with walking when included in the multivariate analyses (i.e. age and self-efficacy). Future research 564 should aim to replicate findings using a prospective design or by embedding such questionnaires in 565 a theory-based process evaluation alongside a trial (Sedgwick, 2014). 566 567 A further limitation is that the study may have overestimated levels of physical activity in people 568

A C C E P T E D ACCEPTED MANUSCRIPT
with Type 2 diabetes. People living in Grampian and Tayside have slightly better self-reported 569 general health than the total population of Scotland (72% and 69.6% in Grampian and Tayside, 570 respectively versus 67.9% in Scotland) (The Scottish Census, 2011). Therefore, their self-reported 571 physical activity, used as the main outcome in this study, may also be higher than the general 572 national population. A further cause of over estimation could be that due to the patient population in 573 the current study i.e. patients with Type 2 diabetic registered with the SDRN may be more engaged 574 with their disease management compared with patients not registered with the SDRN. Social 575 desirability bias could also contribute to any over-estimation of self-reported physical activity. The 576 IPAQ has in fact been shown to overestimate self-reported time spent in physical activity compared 577 with accelerometer measured activity (Ekelund et al., 2006;Hallal et al., 2012).

M A N U S C R I P T
A C C E P T E D ACCEPTED MANUSCRIPT