A cross-sectional representative face-to-face survey (Health Omnibus Survey) was conducted in South Australia (SA) by Harrison’s Research between September–December 2017. A multistage non-replacement sampling process was used to select a random sample of all individuals aged 15+ years in the state. Details of the methods have been reported previously [25, 26]. Individuals with a terminal illness or a mental incapacity (n = 60) or unable to speak English (n = 77) were ineligible for the study. Of the 4320 eligible participants, 1343 refused to answer the survey, providing a final sample of 2977 individuals (68.9%). Considering the low prevalence of CVD or their risk factors among younger participants [1], only those aged 35+ years were included in our analyses (n = 2384).
Outcome: engagement in lifestyle changes
Current engagement in lifestyle changes was assessed through four different questions [1) Are you currently increasing your consumption of fruit and vegetables?; 2) Are you currently increasing your level of exercise?; 3) Are you currently reducing your alcohol consumption?; 4) How many serious attempts have you made to quit smoking in the last year?]. Each of the first three questions allowed the options “No, but I have done so in the past”; “No, and I do not intend to in the next 6 months”; “No, but I am thinking about doing it in the next 6 months”; “No, but I intend to in the next 30 days”; “Yes, and I have been for less than 6 months”; “Yes, and I have been for more than 6 months”. Binary variables when then created based on these answers. Participants were considered as “positives” for increasing fruit and vegetable intake or increasing physical activity level when they answered “yes” to the correspondent questions, either it was happening for less or more than six months. This cut-off was defined considering the “action” and “maintenance” stages of change defined by the Transtheoretical Model [27], which is consistent with the follow-up time used in previous studies investigating lifestyle change [15,16,17]. Participants were classified as reducing their alcohol consumption when they answered “yes” to the correspondent question (either for less or more than six months) and individuals self-reported drinking alcohol in the last 12 months.
For smoking cessation, individuals who reported having smoked in the last 12 months were considered as “positives” for this lifestyle change when at least one attempt to quit smoking occurred in the last year.
Independent variables: cardiovascular conditions, risk factors and health assessments
The occurrence of CVD and its cardiometabolic risk factors was based on self-reported medical diagnosis (“Has a doctor ever told you that you have …” ), and included a list of CVD and their risk factors (i.e. myocardial infarction, angina, heart failure, stroke, high blood pressure, dyslipidaemia, diabetes mellitus), with the answering option “yes/no” for each of them. Body mass index (BMI), another risk factor for CVD, was defined based on self-reported weight and height and used to classify individuals as obese when the BMI was ≥30.0 kg/m2 [5]. All these binary variables were combined to create a new one categorized as 1) none of them (considered as ‘healthy’ in this paper); 2) at risk of CVD (with obesity, high blood pressure, dyslipidaemia, and/or diabetes, but not a diagnosed CVD), or; 3) with current or past CVD (regardless whether they had or did not have a cardiometabolic risk factor).
Participants were also asked if they had visited a GP in the last 12 months for any reason. They were then questioned about recommended preventive care/health assessments [1] performed by the GP during these visits, including 1) measurement of their weight and/or waist circumference, 2) a blood pressure check, 3) tested/requested a test to check their glucose levels, 4) tested/requested a test to check their lipid levels, 5) discussion or assessment of their diet or 6) physical activity levels, or 7) smoking status, or 8) alcohol intake, or 9) mental health status (diagnosed with anxiety, depression or other mental health problem), or 10) any sleeping problem or snoring. Each of these assessments was investigated separately as binary questions (yes/no). Moreover, they were all combined for analysis into a discrete variable (number of these health assessments performed by the GP, ranging from 0 to 10) and then transformed into an ordinal variable (0/did not visit the GP, 1–2, 3–5, 6–10 health assessments).
Confounding variables
Sociodemographic variables [1, 7, 8, 15, 28] included sex, age, marital status, residence area, attained educational level, working status, and the Socio-Economic Indexes for Areas Index of Relative Socio-economic Advantage and Disadvantage (SEIFA-IRSAD, an indicator of relative economic and social advantage/disadvantage of people and households within an area) [29].
Current lifestyle characteristics were investigated using separate questions for fruit intake, vegetable intake, physical activity, alcohol consumption and smoking [see Additional file 1]. Additionally, mental health status (“currently receiving treatment for anxiety, depression, or any other health problem”) was included as a possible confounder, considering it could affect answers related to lifestyle changes and assessments performed by the GP [4, 7].
Participants were also questioned about the number of times they visited a GP, any hospitalisations, or visits to an emergency department in the last 3 months [1, 4, 7].
Data analysis
All analyses were performed using STATA 15.1 (StataCorp, Texas, USA) and the results weighted to the inverse probability of the individual’s selection within the household and re-weighted to the estimated population in SA in 2016 [25, 26]. For reducing alcohol consumption or smoking cessation, analyses were restricted to those individuals that self-reported drinking alcohol (n = 1881) or smoking (n = 409) in the last 12 months.
Logistic regression was used in all analysis considering the binary nature of the outcomes (i.e. increasing fruit and vegetable intake, increasing physical activity level, reducing alcohol consumption, tried to quit smoking; all coded as yes/no), with adjustment for all sociodemographic variables and mental health status.
To test if the presence of CVD or cardiometabolic risk factors were associated with the engagement in lifestyle changes, results were additionally adjusted for current lifestyle (total portions of fruit/vegetable per day, days of physical activity, doses of alcohol/day, and cigarettes smoked/day), the number of visits to the GP, visits to the emergency room, and hospitalisations in the last 3 months [1, 7, 8, 15, 28]. To test the association between the number of health assessments performed by GPs and the engagement in lifestyle changes, results were further adjusted for the presence of CVD or their risk factors (no CVD, at risk of CVD, current or past CVD).
Maximum likelihood estimates (pseudolikelihood log values) for the full models were obtained, and Wald tests for heterogeneity or trend used to estimate the p-values due to the use of clustered weighted data. Results from all analyses were expressed as predicted adjusted prevalence instead of odds ratio to minimize confusion when interpreting study results, as many policy makers, clinicians, and researchers are not familiar with these measures of association [30].
The variance inflation factor (VIF) was investigated as an indicator of over-adjustment and collinearity between the explanatory variables. Furthermore, sex, age, educational level, the clinical health status (either with CVD or their risk factors), and the number of visits to the GP were investigated as possible effect modifiers of the relationship between the number of health assessments performed by GPs and the engagement in lifestyle changes. The interaction was tested by including in the final adjusted logistic regression models a multiplicative term between each of these variables and the number of assessments performed by the GP [31].
Ethics
Participants provided verbal rather than written informed consent, due to the practicalities of carrying out a large-scale study and the low-risk nature of the survey content. Written parental or guardian consent was obtained for participants aged 15–17 years. All procedures performed in this study were approved by the University of Adelaide Human Research Ethics Committee (project H-097-2010).