Study design and population
This cross-sectional study was conducted at a single diabetes clinic at Meir Medical Center, which is part of Clalit Health Services. It is a tertiary medical center serving approximately 1 million residents of a mainly urban area.
The inclusion criteria were age >18 years and a diagnosis of T2D. People with diabetes who were diagnosed less than six months prior to the diabetes clinic visit, pregnant women and people who were unable to converse in Hebrew or provide informed consent were excluded. Participants were recruited from those attending the diabetes clinic on randomly selected days between July 2022 and May 2023. Based on a sample size calculation to detect a correlation coefficient of 0.25 representing a moderate-to-large effect in behavioral research [15, 16], with type-I error of 5% and power of 80%, we determined that 123 participants were required. To account for potential data incompleteness, we aimed to recruit 134 participants (~9% above target). Of those we approached, 134 participants (~80%) agreed to participate and completed a 30-min interview with a trained interviewer. The remaining individuals either declined participation or could not be interviewed due to clinic scheduling constraints and interviewer availability. The study protocol was approved by the Meir Medical Centre Ethics Committee (MMC-0076-22).
Study variables
The exposure variable was the participant’s dietary knowledge. Several instruments assess diabetes knowledge, incorporating dietary knowledge as one component of self-care practices [17,18,19]. Using these instruments often requires adaptations in language and cultural context especially for their dietary knowledge component. At the time of the study, there was no validated questionnaire in Hebrew specifically assessing diabetes-related dietary knowledge. Therefore, we used and adapted the dietary component of the Diabetes in the Arab Population in Israel (DAPI) questionnaire [20], which had been previously developed and validated in Israeli Arab populations with type 2 diabetes. Using closed-ended questions, the participants were asked to state whether several common food items “can be consumed freely”, “can be consumed in moderation” or “can be consumed only in rare situations (such as hypoglycemia)”. To make the questionnaire suitable for both Arab and Jewish Israeli populations, we made targeted cultural adaptations while maintaining the core construct—knowledge of appropriate food choices for diabetes management. Specific changes included: (1) replacing Arab-specific food items with culturally neutral equivalents from the same food category—for example, “dates” and “grapes” (both high in natural sugars), “bulgur or frike” with “whole grains”, and “biscuits and baked cookies” with “sweet pastries”; (2) expanding the item “rice” to “white rice or white bread or pita” to reflect staple carbohydrate sources consumed by both populations; and (3) adding questions about commercially packaged foods labeled for people with diabetes and preferred fat sources to align with current nutrition recommendations and the Israeli food environment. These modifications ensured the questionnaire was culturally relevant to both populations while preserving the assessment of knowledge across the same nutritional domains as the original DAPI.
The adapted questionnaire underwent two validation steps. First, a multidisciplinary team of endocrinologists, nutritionists, and public health specialists assessed face validity to ensure clarity and appropriateness. Second, we established content validity by individually presenting the questionnaire to 9 registered nutritionists specialized in diabetes care, working in hospital or community settings. As registered nutritionists are the main providers of diabetes nutrition education in Israel, they were well-positioned to determine clinically relevant content and correct responses. Items were retained if at least 75% of nutritionists agreed on the correct answer [21, 22]. For instance, if over 75% of the nutritionists indicated that Coca-Cola beverage should be consumed ‘only in rare cases, such as hypoglycemia,’ this was the designated correct answer (worth one point) and the item was retained. Because this was a cultural adaptation of an existing validated instrument rather than de novo scale development, full psychometric validation was not conducted. The final version included 17 items. Internal consistency was assessed using Cronbach’s alpha (α = 0.52).
The main outcome variable was MedD adherence, measured by the 17-item I-MEDAS questionnaire [23], an adaptation of the 14-item MEDAS questionnaire from the Spanish PREDIMED study [24, 25]. In the original MEDAS questionnaire, participants were asked to estimate their daily or weekly intake of various foods, earning 1 point per item if predefined criteria were met. For instance, two servings or more of non-starchy vegetables servings per day, with 1 serving defined as 200 g, earned 1 point). The I-MEDAS includes food items that are in accordance with the MedD principles and are widely consumed in Israel (e.g., tahini, hummus and low-fat dairy) instead of food items that appeared in the original MEDAS and are rarely consumed locally (e.g., shellfish and savory tomato sauce). The I-MEDAS has been previously validated and shown to correlate with all-cause mortality in a large Israeli cohort study [23].
The participants’ electronic medical records were thoroughly reviewed to collect socio-demographic, clinical, biochemical, and administrative data. Routine laboratory tests taken up to 6 months prior to the interview visit were recorded. Information on macrovascular and microvascular diabetes complications was based on documented diagnoses or laboratory tests. Nephropathy was considered present if noted by the clinician or evident based on lab tests of serum creatinine level ≥1.3 mg/dL or urine albumin/creatinine ratio >30. Area-based socioeconomic status was determined using the Israeli Socioeconomic Score (1–10), assigned based on participants’ residential address and periodically published by the Central Bureau of Statistics [26], where higher scores indicate greater affluence. This ecological index incorporates neighborhood-level indicators including average income, education level, employment rate, standard of living, and demographic characteristics.
Statistical analysis
In this cohort, both dietary knowledge and MedD adherence scores were relatively high with limited variability. Therefore, dietary knowledge was categorized into tertiles, while MedD adherence was dichotomized into the highest tertile versus the lower two tertiles, to better capture potential differences in the “high-score” subgroup while still retaining a sufficiently large comparison group.
Baseline characteristics across dietary knowledge tertiles are presented as means and standard deviations for continuous variables and as frequencies and percentages for categorical variables. For categorical variables, Chi-square tests were used. For continuous variables, normality was assessed; normally distributed variables were compared using one-way ANOVA, while non-normally distributed variables were analyzed using the Kruskal–Wallis test.
The primary analysis used multivariable logistic regression to assess the association between dietary knowledge tertiles, using the knowledge lowest tertile as a reference, and high MedD adherence (upper tertile of MedD adherence, dichotomic variable), with adjustment for age, sex, duration of diabetes and socio-economic status. Odds ratios (OR) were calculated with 95% confidence intervals (CIs). Logistic regression assumptions were verified, including absence of multicollinearity among predictors (variance inflation factor <2.0 for all variables). The goodness of fit of the model was evaluated with the Hosmer–Lemeshow statistic and Nagelkerke R².
For descriptive purposes, the unadjusted association between dietary knowledge and MedD adherence tertiles was assessed using chi-square test.
Two sensitivity analyses were conducted. First, the primary logistic regression model was further adjusted to additional clinical variables, including BMI ≥ 30 kg/m², HbA1c, number of hypoglycemic agents, and number of diabetic complications. Second, linear regression was used with the outcome variable of MedD adherence defined as a continuous score, adjusted for the same covariates as in the main model. Data were analyzed using IBM SPSS statistics software version 29.0 (SPSS Inc., Chicago, IL, U.S.A.). A two-sided P-value < 0.05 was considered statistically significant.
Dining and Cooking