**2. Materials and Methods**

The study was performed using an analytical cross-sectional design. Data were collected from the patients visiting the Primary Healthcare Centers (PHCs) in Majmaah City, KSA from February to April 2017. A systematic random sampling technique was used for the selection of patients based on the inclusion criteria, which were: clinically diagnosed cases of type 2 diabetes mellitus of either gender and in the age range of 35–55 years. The DM prevalence value of 23.7% [11] was used for sample size calculation, and the values were placed in the level of precision formula that yielded a sample size of 278. To compensate for potential missing observations/patients withdrawing from the study, the sample size was increased to 350. Each patient's consent was obtained prior to data collection. This research was approved by the ethical review committee of Majmaah University, KSA vides reference number: MURECApril.02/COM-2016.

The dietary attitude questionnaire (DAQ) was prepared following a thorough review of the literature and based on meetings with local experts to determine the pattern of questions suitable for assessing and evaluating the DA of patients with type 2 diabetes. The self-administered valid and reliable questionnaire was divided into three sections (Section A, B, and C). We have discussed the psychometric properties (face validity, content validity, exploratory factor analysis (EFA), and reliability) of the DAQ in a separate article [12]. The internal consistency reliability of the DAQ was excellent (Cronbach Alpha = 0.841). Based on the pilot study results of the EFA, the five factors were labelled as "food selection", "health impact", "healthy choices", "food restraint", and "food categorization" [12].

Section A contained questions related to demographic characteristics. Section B was comprised of 16 questions that assessed patients' general DA towards food. All of the questions were measured on a seven-point Likert scale (strongly agree, agree, somewhat agree, neutral, disagree, somewhat disagree, and strongly disagree). The DA was further classified as positive and negative based on mean values. Values at or above the mean were classified as having a positive DA, and values below the mean were referred to as having a negative DA [13,14]. Section C was also comprised of 16 questions: The first

15 questions assessed patients' DA towards specific food items with categories ("not" eating this food is healthy and necessary, eating this food "occasionally" is healthy and necessary, and eating this food "often" is healthy and necessary), and the last question was about "opinion regarding healthy diet" with the options "yes" and "no".

The data were entered and analyzed using IBM SPSS version 25 (IBM Corp., Armonk, N.Y., USA). Normality of the quantitative variables was assessed through a One-Sample Kolmogorov–Smirnov (KS) test. A univariate method (z-score) was used for the detection of outliers. Qualitative variables are expressed as frequencies and percentages, while a median and quartiles (25th–75th) are given for non-normally distributed variables. A one-sample non-parametric chi-squared test was used to assess the significance of overall and subgroup positive and negative DA. Pearson's chi-squared test was applied to compare the overall positive and negative DA between gender, body mass index (BMI), education status, and marital status. Binary logistic regression with the backward conditional approach was used to predict the set of variables assessing the DA of patients towards specific food items. The odds ratios were further converted into probabilities by using the equation (yˆ = odds/1 + odds). The statistical significance value was set at *p* < 0.05.
