The university sample included students interested in their nutritional intake and quality. Interest in nutritional intake was ensured by offering participants feedback on their diet quality by Accredited Practising Dietitians as non-monetary incentive. Participants were recruited between October 2021 and May 2022 using methods such as word-of-mouth, including physical flyers around university campus, advertisements and invitations on university websites, and social media posts. To be eligible, participants had to be aged between 19 and 30 years inclusive, own a smartphone or digital camera, and must not have ever had an eating disorder or concerns about disordered eating. Participants completed a basic demographic questionnaire, providing information such as gender (female, male, other), height (cm), weight (kg), residential postcode, and highest degree or level of school completed (year 11 or below, year 12, Certificate III or IV, diploma or advanced diploma, bachelor’s degree, graduate certificate or graduate diploma, master’s degree or doctoral degree). Participants’ self-reported weight (kg) and height (cm) were used to calculate body mass index (BMI) (underweight < 18.5 kg/m2, healthy weight 18.5–24.9 kg/m2, overweight 25.0–29.9 kg/m2, obese ≥ 30.0 kg/m2), which has been found to be sufficiently accurate [32]. Socio-economic status was determined using the Socio-Economic Indexes for Areas 2016 Index of Relative Socio-economic Advantage and Disadvantage [33] based on participants’ postcodes. The top five deciles were labelled as high, and the bottom five deciles were labelled as low.

Each participant captured images of all foods and drinks consumed on three days using their smartphone. Images were captured immediately prior to the meal, snack or beverage so that the time stamp on the image reflected the start time of consumption. On the same three days, participants also kept a record of their intake on a commercial smartphone application (app) [34] or a paper-based food diary. On the app, participants recorded all foods and drinks consumed and amounts consumed. If a food or beverage they consumed was not on the app’s database, the closest substitute was used at the participant’s discretion. After completion, data on the app were emailed directly to researchers in a format that could readily be imported into the Foodworks Professional nutrient analysis software [35] and analysed using the Australian Food and Nutrient Database (AUSNUT 2011-13) [36, 37]. For participants who used a paper-based food diary, data were manually entered into the Foodworks Professional nutrient analysis software by researchers. This component of the study was approved by the University of Sydney Human Research Ethics Committee on the 9th of March 2022 (2021/513).

The community sample included young adults from the sub-study (n = 133) [38] of a larger cross-sectional MYMeals project (n = 1001) [39]. Data from this group were collected in a similar manner to that of the university sample but used automated wearable cameras instead of user-initiated smartphone cameras and researcher-administered 24-hour recalls instead of food diaries recorded by participants.

Participants were aged 18–30 years, consumed foods or beverages prepared outside the home at least once a week, owned a smartphone, and were able to read and write English. Participants who were pregnant, lactating or had ever had an eating disorder were excluded. Recruitment and data collection methods have previously been described in the MYMeals study protocol [40]. In brief, participants in the sub-study wore an Autographer camera on a lanyard around the neck for all waking hours over three days. The camera automatically captured images from a first-person perspective every 30 s. On the same three days, they completed daily 24-hour dietary recall interviews with research dietitians via the Australian version of the Automated Self-Administered 24-hour recall Australia program [41]. Demographics were collected similarly to the university sample. Participants received an AUD $100 voucher as monetary incentive if they completed the study and returned the camera upon completion.

All images captured by the Autographer camera were coded by Accredited Practising Dietitians for the presence or absence of food and beverages, and matched with their 24-hour recall data. Two researchers independently matched the two sources (V.C. and A.D.) [42]. This component of the study was approved by the University of Sydney Human Research Ethics Committee on the 15th of July 2016 (2016/546).

After the food images were matched with the food diaries and 24-hour recalls, data from participants with no unmatched main meals over the three days of data collection and no more than one unmatched snack or beverage per day were included for analysis. The data of all other participants were excluded.

The times of eating occasions were collated using the time and date stamps available from the food images. An eating occasion was defined as the consumption of any food or beverage with ≥210 kJ of energy [7]. Images captured within ≤ 15 min of each other were combined to form one eating occasion as per previous studies [7, 8] and the time of the latest image taken prior to food consumption was used to label the eating occasion. Each day was defined as the 24-hour period from 00:00 h to 23:59 h the same calendar day.

Eleven metrics measuring chrononutritional patterns were extracted – five eating pattern metrics, five meal timing variability metrics, and evening eating.

Eating pattern metrics including the chronological clock time of the first and last eating occasions, caloric midpoint, number of eating occasions per day, and eating window were extracted for all days and averaged over the three days of data collection. The caloric midpoint was defined as the time at which 50% of daily energy was consumed. The eating window was defined as the duration from the start time of the first eating occasion to the start time of the last eating occasion of the same day.

The variability of the above eating pattern metrics across the three days of data collection were used to measure meal timing variability: standard deviation of the first eating occasion (SD First), standard deviation of the last eating occasion (SD Last), standard deviation of the caloric midpoint (SD Caloric Midpoint), coefficient of variation of the daily number of eating occasions (CV Eating Occasions), and coefficient of variation of the eating window (CV Eating Window). These metrics were selected based on the methods of previous studies that used standard deviation to measure the variability of the first and last eating occasions and caloric midpoint [12, 16, 43] and coefficient of variation to measure the variability of the daily number of eating occasions and eating window [15]. The cut-offs used to categorize standard deviation and coefficient of variation scores as low, moderate, high, and very high variability were based on the difference in hours or number of eating occasions (Table 1). For example, participants with low variability in SD First varied in meal timing by less than two hours over the three days of data collection. An example of this is the consumption of the first meal or snack at 09:00 h on the first day, 10:00 h on the second day, and 10:45 h on the third day. These cut-offs were adapted from the methods of a previous study measuring meal timing variability in young adults [6].

Evening eating was defined as continuing to eat at or after 20:00 h [44, 45], as determined using food image time stamps. Images with time stamps ≥20:00 h were further examined and labelled by food type as well as by food group, that is, predominately foods from the five food groups (grains, vegetables, fruits, dairy, lean meat and alternatives); or predominately discretionary, that is, foods and drinks that are high in saturated fat, added sugar, added salt, and/or alcohol that should only be consumed sometimes and in small amounts [46, 47]. If an eating occasion consisted of both five food group and discretionary foods, the one that provided more energy was used to label the eating occasion.

Diet quality was measured using the Healthy Eating Index for Australian Adults (HEIFA-2013), one of the best performing diet quality indices used in Australian adults [48] based on an inventory of diet quality indices construction criteria [49, 50]. It is a validated, gender-specific tool that assesses adherence to the Australian Dietary Guidelines [51]. The scoring system for this tool has been described elsewhere [52]. Briefly, the index consisted of 11 components: one for each of the five food groups; one for discretionary foods; four for specific nutrients (fatty acids, added sugar, sodium, alcohol), and one for water intake. Each component was scored a maximum of 10 points except for water intake and alcohol, which were scored a maximum of five. This totalled to an overall maximum score of 100. A higher score indicated a closer adherence to the dietary guidelines. For three of the five food group components (grains, vegetables, fruit), five of the 10 points were assigned to the adequate consumption of these food groups and the other five were assigned to the number of serves of wholegrains or how much variety was present in the types of fruit and vegetables consumed. For the fatty acid component, five of the 10 points were assigned to minimising saturated fat intake, and the other five were assigned to the adequate consumption of poly- and monounsaturated fats.

Within each of the 11 components, points were given incrementally for specified increases or decreases in number of serves consumed. Components where higher scores were given for a lower consumption were discretionary, saturated fat, sodium, added sugar, and alcohol. Increments and serves were different for each component and for different genders. For example, for the lean meat and alternatives component, males consuming ≥3.0 serves earned 10/10 points, 2.5 to < 3.0 serves earned 8/10, 2.0 to < 2.5 serves earned 6/10, 1.5 to < 2.0 serves earned 4/10, 1.0 to < 1.5 serves earned 2/10, and ≤0.5 earned 0/10. For females, ≥2.5 serves earned 10/10 points, 2.0 to < 2.5 serves earned 8/10, 1.5 to < 2.0 serves earned 6/10, 1.0 to < 1.5 serves earned 4/10, 0.5 to < 1.0 serve earned 2/10, and 0.0 serves earned 0/10. This scoring system was used on each day of the participants’ 24-hour recalls or food diaries via the Foodworks output and averaged across the three days of data collection to provide an overall diet quality score, as well as scores for individual diet quality components.

Statistical analyses were conducted using SPSS software, v27.0 for Windows (IBM, Armonk, NY, USA) [53]. Descriptive statistics (frequency, mean, standard deviation, and percentage (%)) were used to summarise sample characteristics, chrononutritional variables, and diet quality. Differences between university and community groups in chrononutritional variables and diet quality were determined using the t-test for normally distributed data and the Mann-Whitney U test for non-normal data. Normality was determined using the Shapiro-Wilk test. Linear regression was used to identify associations between chrononutritional variables and diet quality, including overall diet quality and individual diet quality components (discretionary, total vegetable, total fruit). Univariate general linear models tested for differences in diet quality between participants who continued to eat at or after 20:00 h and participants who concluded eating by 20:00h, as well as between university and community participants who continued to eat at or after 20:00h. In the linear regression and univariate general linear model, analyses were adjusted for gender, BMI, and socio-economic status. A p value of ≤ 0.05 was considered statistically significant.