Diabetes mellitus (DM) is an increasingly global public health problem. The number of people (aged 20-79 years) affected by diabetes in world will increase from 6.4% (285 million adults) in 2010 to 7.7%, (439 million adults) by 2030. This increase will be 69% in developing countries while developed countries will experience a 20% increase during the same period [1].

Physical activity reduces the risk of over 25 chronic conditions, in particular, coronary heart disease, stroke, hypertension, breast cancer, colon cancer, osteoporosis, and type 2 diabetes [2]. Regular physical activity is recommended for type 2 diabetic patients, as it is commonly known to correct metabolic disorders and prevent complications such as cardiovascular diseases [3]. The Canadian Diabetes Association's (CDA) Clinical Practice Guidelines for the prevention and management of diabetes recommended that type 2 diabetic patients should participate in moderate-intensity physical activity, such as brisk walking and biking, for at least 150 minutes/week, over at least 3 nonconsecutive days [4]. However recent studies have shown that diabetic patients may undertake less physical activity than non-diabetic people. Up to one third of diabetic patients are completely sedentary, and only a third exercise regularly [5]. To change this to a more healthy behavior the need for theory-driven interventions are recommended [6, 7]. Theory provides a road map for studying problems, developing appropriate interventions, and evaluating their successes. Theory can also help planners identify the most suitable target audiences, methods for fostering change, and outcomes for evaluation [6]. The Theory of Planned Behavior (TPB) [8] is a well-established theory that has been used to further understand predictive factors of health behaviors in general [9, 10] and physical activity in particular [11–14]. Numerous studies have provided support for the TPB to explain why some people practice physical activity and some do not [15–19].

This theory suggests that people's intention to perform a specific behavior is predicted or influenced by three determinants: attitudes, subjective norms and perceived behavioral control. Attitude refers to a personal factor of like or dislike, subjective norms refers to an individual's perception of social pressure, and perceived behavioral control refers to a person's perceived confidence in the ability to perform a behavior [20]. Hagger et al. reviewed 79 studies in a meta-analysis and reported that the TPB explained 44.5% of the variance in physical activity intentions and 27.4% of the variance in physical activity behavior [14].

The application of TPB in studies of physical activity in diabetic patients are well documented. For instance Plotnicoff et al. in a study on diabetic patients found the TPB constructs explained almost 40% of the variance in intentions for type 1 and type 2 diabetes. In cross-sectional studies, the TPB accounted for 23 and 19% of the variance in PA for type 1 and type 2 diabetes, respectively. In prospective studies, the TPB explained 13 and 8% of the variance in PA for type 1 and type 2 diabetes, respectively. The findings provide evidence for the utility of the TPB for the design of PA promotion interventions for adults with either type 1 or type 2 diabetes [21]. Also Davies et al. found that intention explained 28% of the variance in physical activity behavior. Attitude, subjective norm and perceived behavioral control (PBC) explained 73% of variance in physical activity intention. Attitude and PBC mediated the relationship between conscientiousness and physical activity intention. These results showed that targeting constructs proximal to the behavior (attitudes and PBC) may be effective in overcoming inherent qualities such as personality in order to produce physical activity behavior change within this sample population [22]. Data show links between sedentary behavior and all-cause mortality, cardiovascular disease, obesity, and adverse metabolic profiles [23]. Preliminary findings from STIL Project suggest that inactivity is more complex than we sometimes think [24]. Thus, in order to develop effective interventions to promote habitual physical activity, the predictive factors of this behavior is needed to be identified [25].

At present there are no comprehensive validated scales for measuring diabetic patients' beliefs regarding physical activity. Such instruments could help to understand patients' perspectives and in turn design comprehensive interventional programs for diabetic patients. Available instruments review particular aspects of the beliefs such as perceived benefits and barriers [26–31], outcome expectancy [29–32], self-efficacy [30, 33], social support [29, 30], enjoyment of physical activity [29, 30], social influences [33], that were developed in the framework of other health belief models. There was one more questionnaire, developed by Blue et al., assesses the indirect measures of the theory of planned behavior including behavioral beliefs, normative beliefs and control beliefs to predict physical activity intentions of persons at risk for diabetes [34]. No questionnaire was found in the context of direct measures of the Theory of Planned Behavior. The main objective of this study was to design and develop an instrument in such a framework. However, this study included an additional construct (self-identity) to the original theory. As Ajzen suggests the Theory of Planned Behavior is open for further extension with additional constructs [35]. Studies have shown that individuals who identify themselves as exercisers have more favorable intentions and engage significantly more in exercise than those who do not [36–38]. It is argued that self-identity may play an important role in predicting physical activity [36–38]. Furthermore, this study was limited to a sample of female diabetic patients. It is estimated that more than 1.5 million people with diabetes live in Iran. The preliminary results derived from a national study indicated that the prevalence of type 2 diabetes was 3.6% among adults aged over 30 (4.3% of women and 2.6% of men) [39]. The figures clearly indicate that, as many other countries, women in Iran are suffering more from diabetes. In addition, the existing data from Iran indicate that prevalence of sedentary lifestyle in females is higher than males [40–42]. Sarrafzadegan et al. found that abdominal obesity was nearly six times as prevalent in women as in men (71.7% vs. 12%, P < 0.05) [43].

The purpose of this study was to develop a scale for assessing women's perceptions about physical activity and to test its psychometric properties. This scale was developed based on Theory of Planned Behavior framework. In Theory of Planned Behavior, each predictor may be measured directly e.g. by asking respondents about their overall attitude, or indirectly e.g. by asking respondents about specific behavioral beliefs and its outcome. Usually for developing a TPB-based questionnaire, it has been suggested that for direct measures one should use the same direct measures developed by Ajzen [57] and Francis [51]. For indirect (belief-based) measures it has been recommended to carry out an elicitation study to develop all predictor constructs in the TPB that are attitude; subjective norm; and perceived behavioral control. Direct and indirect approaches make different assumptions about the underlying cognitive structures and neither approach is perfect. When different methods are tapping the same construct, scores are expected to be positively correlated, so it is recommended that both be included in a TPB-based questionnaire [51]. Unfortunately, in this study, only direct measures of TPB were used.

As Francis et al. suggested, each construct should be measured using a minimum of three items. This will result in a minimum of 12 items for intention and direct measures of the predicting variables [51]. However, in the primarily draft of the questionnaire presented in this study, 22 items included. In fact according to Ajzen and Francis [51, 57], attitude, subjective norm, and perceived behavioral control each contained two subscales (attitude: instrumental and affective; subjective norm: injunctive and descriptive; and perceived behavioral control: self-efficacy and controllability).

Exploratory factor analysis was led to remove three items from a total of original 22 items and the final 19-item scale categorized in six factors explaining 60.3% of the variance observed. EFA with varimax rotation indicated that six factors including instrumental attitude, subjective norm, perceived behavioral control, affective attitude, self-identity, and intention can be extracted. As expected attitude was extracted with two subscales (instrumental and affective) but subjective norm and perceived behavioral control did load as single factors and expected subscales were not emerged separately.

Overall the internal consistency of the PAQ-DP was acceptable, although Cronbach's alpha for some dimensions were lower than 0.70. However, there was no significant increase in the Cronbachs' alpha when any items were removed. This result may be due to the fact that fewer items were included in these dimensions. Also, small sample size, high homogeneity of the patients, and small variability of the scores might decreased alpha coefficients [71]. In addition, as Ajzen suggested, adapting items used in previous studies might produce measures with relatively low reliabilities and lead to an underestimate of the relations among the theory's constructs and of its predictive validity [57]. It seems that increasing the sample size, and increasing the number of items in some dimensions could further confirm the reliability of the scale. This result is very similar to those previously reported by Tan et al. [76] and Leung et al. [77].

There are some practical implications for developing a scale to measure diabetic patients' beliefs toward regular physical activity. Valid and reliable instruments are needed for designing and evaluating health programs based on behavior change theories of social and behavioral sciences [78]. Interventions to promote regular physical activity can target the beliefs that guide behavior. In other words, identifying beliefs could lead to development of effective programs and perhaps help to achieve healthy lifestyle for target populations. Health care providers who work with diabetic patients can use the measures to identify beliefs that promote regular physical activity and in turn intervene to delay or prevent diabetes complications. The PAQ-DP also could be a useful instrument for evaluating intervention effects by comparing the scores before and after patients' receiving various interventions; and for explaining the correlations of beliefs of diabetic patients with their exercise behavior.

The findings from this study should be interpreted with caution. This study had several inherent limitations: only female engaged in the study, sample size was relatively small, the reliability coefficients of some factors were not satisfactory. Also, only direct measures were used in the study. In addition 6 factors and 19 items means that most factors might not be quite deep. The future studies using this questionnaire could help to overcome these problems.

Overall, the study findings suggest that the PAQ-DP is a valid and reliable instrument for assessing beliefs of female diabetic patients regarding physical activity. The findings of the current study support the Ajzen's Theory of Planned Behavior. Further studies are recommended to confirm its application in clinical practice.