Study design

The investigation relies on information gathered from a cross-sectional survey that aimed to examine the comprehension and behaviour concerning diabetes mellitus within a cohort of adults in Jordan.

Sampling method

A convenience sampling technique was used to recruit participants for this cross-sectional study. Convenience sampling was chosen due to its practicality and cost-effectiveness, allowing for a broad and diverse range of participants to be included within a limited timeframe. This method facilitated the recruitment of participants from various easily accessible locations (such as healthcare centres, community centres, workplaces and public areas), ensuring a sufficiently large and varied sample to explore the relationships between demographic/clinical factors and diabetes knowledge/behaviour.

The sample size was determined through a power analysis to ensure adequate statistical power. Based on an estimated population of 1 million individuals, a minimum sample size of 1050 was calculated to achieve a 95% confidence level and a 3% margin of error. The sample size also accounted for expected prevalence rates of diabetes knowledge and behaviour to enable detection of meaningful differences across demographic and clinical groups. To recruit a representative sample, the study recruited 1050 female and male respondents, aged 18 to 84 years, from the cities of Amman, Madaba and Al-Karak, and with diverse socioeconomic backgrounds. No participant dropouts were recorded.

Individuals with cognitive impairments or language barriers that would prevent them from completing the survey were excluded.

Ethical considerations

The investigation was performed in accordance with the Declaration of Helsinki, and the ethical committee of the American University of Madaba sanctioned all procedures performed in the participants (ethical approval number H23005). Written informed consent was obtained from all study participants.

Study setting and participants

The study included 1050 Jordanian adults aged 18–84 years from various regions of Jordan, ensuring representation from both urban areas (n=921) and rural areas (n=129). Urban areas have a high population density and well-developed infrastructure, while rural areas have lower population density and less developed infrastructure. This distribution aimed to capture diverse demographic and clinical profiles, considering variations in healthcare access, socioeconomic status and lifestyle factors. The study was conducted over six months, from February 2023 to June 2023.

Clinical characteristics

Participants’ clinical characteristics were classified based on self-reported medical history. Among the participants, 52 reported having diabetes and 998 reported not having diabetes. Additionally, 16 participants reported having dyslipidaemia (any lipid imbalance, such as high cholesterol, high LDL-cholesterol or low HDL-cholesterol), while 1034 did not. Furthermore, 11 participants reported having coronary artery disease, while 1039 did not.

Data collection

Trained nutrition postgraduate researchers conducted data collection for this study. They briefed participants on the study’s objectives and methodologies, and ensured that informed consent was obtained prior to participation. The researchers assisted individuals with reading difficulties, ensuring accurate and complete responses. Participants completed the structured, self-administered questionnaire in approximately 35–40 minutes, and all responses were collected with strict adherence to confidentiality protocols. The data collection process aimed to maximise response accuracy and quality, providing participants with a supportive and secure environment throughout.

Anthropometric assessment Height

Trained senior undergraduate nutrition students gauged each participant’s height once using a measuring rod. Participants were instructed to stand with their feet together and knees extended, while positioning their heels, buttocks and shoulder blades against the wall. In addition, they were advised to align their head with the Frankfurt horizontal plane, as described by Nieman and Lee [21].

Body weight

A scale with a capacity of 150 kg and a precision of 0.1 kg was used to measure the participants’ body weights. Prior to each measurement, the scale was positioned on a flat, firm surface and calibrated to zero. Participants were requested to stand unaided on the central area of the platform with minimal attire and without footwear, facing straight ahead, as described by Nieman and Lee [21].

BMI

Body weight in kilograms divided by height in metres squared was used to calculate BMI. Subsequently, participant BMI values were classified into four categories: underweight (<18.5 kg/m2), normal weight (18.5–24.9 kg/m2), overweight (25–29.9 kg/m2) and obese (≥30 kg/m2), using the categorisation approach described by Nieman and Lee [21].

Physical activity assessment

Participants’ weekly physical activity levels were categorised based on the 2005 International Physical Activity Questionnaire guidelines [22]. High-level physical activity was characterised by engaging in vigorous exercise for more than 3 days per week, accumulating 1500 metabolic equivalents (METs) or exceeding 3000 METs through a combination of walking or moderate to vigorous intensity activities over 7 days. Moderate-level physical activity was defined as performing vigorous exercise for over 20 min on more than 3 days per week, or engaging in moderate intensity exercise or walking for over 30 min on more than 5 days per week, or reaching 600 METs through walking or moderate to vigorous activities. Low-level physical activity was classified as not meeting the criteria for either high or moderate activity levels.

Survey

To gather data for this research, a self-administered survey was conducted, using closed-ended questions. The survey covered various demographic and clinical details, including age, sex, education level, employment status, marital status, chronic illness, height and weight. It also included questions on behaviour and knowledge relating to diabetes mellitus, such as its symptoms, risk factors and complications. Data on national origin (Jordanian or non-Jordanian) were also collected. Race and ethnicity have not been well studied and identified in Jordan, as is the case in other Middle Eastern countries. Generally, people tend to identify themselves based on national origin [23], even though people with different national origins in the Middle East may share the same race/ethnicity. All the study participants were Jordanian.

The questionnaire was created and piloted in a previous study by Khan et al [24] based on literature reviews and the findings of previous studies [25,26,27]. They evaluated the validity of the questionnaire’s content through a review process involving professors with expertise in the research subject and a socio-psychologist [24]. The panel assessed the statements in the questionnaire to ensure that they adequately addressed the study’s objectives [24]. Permission to use the questionnaire for the current research was obtained from the authors.

Survey validation

We performed rigorous validation of the survey tool through a series of detailed statistical analyses to establish the reliability and validity of the findings and to ensure that the tool is reliable in the Jordanian context. Initial data screening was performed to eliminate inconsistencies and missing values, ensuring that the dataset was clean and suitable for analysis.

First, principal component analysis was used to identify latent factors and uncover the underlying structure within the data [28]. The results of the principal component analysis showed a well-defined factor structure, with the explained variance for the first seven components as follows: component 1 (general knowledge on diabetes mellitus) accounted for 25.10% of the variance, component 2 (knowledge of risk factors) accounted for 17.30%, component 3 (knowledge of symptoms) accounted for 13.45%, component 4 (knowledge of complications) accounted for 12.20%, component 5 (health behaviour and medical history) accounted for 9.05%, component 6 (source of information on diabetes) accounted for 7.75%, and component 7 (perceived beneficial behavioural changes) accounted for 6.15%. These components collectively captured 91.00% of the total variance, indicating that the primary factors contributing to diabetes knowledge and behaviour were comprehensively represented.

The Kaiser–Meyer–Olkin measure was calculated to assess sampling adequacy for factor analysis [29]. The overall value was 0.81, which is considered excellent, indicating that the sample was adequate for principal component analysis and that the data were suitable for factor analysis. Bartlett’s test of sphericity was conducted to test whether the correlation matrix was an identity matrix [30], which would indicate that the variables were unrelated. The test showed a highly significant result, with a χ2 value of 2580.45 (p<0.001), confirming that the correlations between items were sufficiently large for principal component analysis, thereby validating the factorability of the correlation matrix.

Cronbach’s alpha was calculated to ensure internal consistency. The results showed strong internal consistency for the primary factors, with a Cronbach’s alpha value of 0.815, indicating that the items within each factor reliably measured the same underlying construct [31]. This high level of internal consistency provides confidence in the reliability of the survey tool.

Additionally, ridge regression was applied and achieved a cross-validation mean score of 0.987 with a low SD (0.025) and a model score of 1.0, indicating very high goodness of fit. This confirms that the model effectively captured the relationship between the variables, further validating the reliability of the survey tool. Principal component regression was also employed, achieving a cross-validation mean score of 0.986 with a low SD (0.027) and a model score of 0.9992. This indicates a robust fit without overfitting, further supporting the validity of the survey tool [32].

Data scoring and aggregation

For the knowledge component, participants were asked a series of questions aimed at assessing their understanding of diabetes, covering symptoms, risk factors, complications and general knowledge. Responses were scored on a binary scale, with ‘1’ assigned for correct answers and ‘0’ for incorrect answers. The sum of these binary scores provided a total knowledge score for each participant, which was then normalised to the total number of knowledge-related questions to derive a percentage score. Similarly, the behaviour component, evaluating adherence to recommended diabetes management practices such as medication, dietary practices and physical activity levels, was scored. Each behaviour that adhered to guidelines was scored as ‘1’, and non-adherent behaviours were scored as ‘0’. A total behaviour score was calculated for each participant by summing these scores and normalising them to the total number of behaviour-related questions to derive a percentage score.

Statistical analysis

The numbers of participants in the various demographic and clinical categories were determined and percentages were calculated to provide clarity on group distribution. Scores for knowledge and behaviour were aggregated by demographic profiles (e.g. age/sex) and clinical profiles (e.g. hypertension/BMI), with mean scores and SDs indicating central tendency and dispersion.

Differences in knowledge and behaviour scores across categories were tested using unpaired t test for binary factors (e.g. sex or marital status) and one-way ANOVA for multi-level factors (e.g. age or BMI), with statistical significance set at p<0.05.

Two multiple linear regression models were developed to explore the relationship between various demographic/clinical variables and diabetes knowledge/behaviours. A simple linear regression model was developed to test the association between diabetes knowledge and behaviours. Variables that demonstrated significant associations (p<0.05) in preliminary t tests and ANOVA were included in the regression models. Additionally, covariate adjustment was guided by previous research showing significant associations between diabetes knowledge/behaviours and factors such as age [33], sex [34], marital status [35], hypertension [36] and smoking status [37].

Diagnostic tests were performed to ensure the validity of the regression analysis, including checks for linearity, multicollinearity (variance inflation factor), homoscedasticity, and normality of residuals (Shapiro–Wilk and Durbin–Watson). The models quantified the influence of each factor, calculating coefficients, standard errors and R2 values, thereby explaining the proportion of variance in diabetes knowledge and behaviour scores attributable to the included variables. Data analysis was conducted using SPSS version 28 (IBM, USA).