Evaluating nutritional and food cost assessments: cash-register receipts may be an alternative for FFQs – accuracy and feasibility in a dietary study

Abstract

Objective

Comparing two dietary cost measurements at the individual level: ‘Food Frequency Questionnaire (FFQ) and supermarkets prices’ with ‘cash-register-receipts-items’.

Design

Method comparison study. Reference method: participants collected receipts of food purchases for 28 days; conventional method: participants completed a diet-specific online FFQ.

Setting

A Vegan Israeli Study substudy.

Participants

30 participants were recruited using advertisements on social media.

Main outcome measure

Average diet cost, energy and nutrients consumption, generated by: (1) items on receipts; (2) online FFQ with supermarkets prices.

Analysis

Examining correlations between methods and generating Bland-Altman graphs.

Results

Agreement between measurement tools increased when ‘eating-away-from-home’ dietary costs were omitted from the analysis, from differences of 1453 New Israeli Shekel (NIS)/28 days (414 US$/28 days) higher to 1010 NIS/28-days (288 US$/28days) lower compared with differences of 756 NIS/28 days (215 US$/28 days) higher to 1159 NIS/28 days (330 US$/28 days) lower. Moreover, the Pearson correlation between methods, which was r=0.29 (p=0.13), increased to r=0.52 (p<0.0042). Finally, Pearson correlations between questionnaire-based and receipt-based nutrients were: energy=0.58 (p=0.001); protein=0.46 (p=0.012); fat=0.50 (p=0.005); carbohydrates=0.76 (p<0.001); calcium=0.46 (p=0.012); and iron=0.37 (p=0.049).

Conclusions and implications

The dietary cost of the ‘FFQ-and-supermarket-prices’ method is more strongly correlated and agreeable with the ‘cash-register-receipts-items’ method when ‘eating-away-from-home’ items are omitted, indicating that ‘eating-away-from-home’ costs are poorly estimated when using the standard ‘FFQ-and-supermarket-prices’ method. Finally, estimating energy, carbohydrates, protein, fat, calcium and iron using ‘cash-register-receipts-items’ is feasible.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Estimating dietary costs at the individual level is challenging. Common methods use price databases linked to dietary assessment tools like Food Frequency Questionnaires but may not reflect actual food expenditures or account for eating outside the home.

WHAT THIS STUDY ADDS

• This study demonstrates that cash-register receipts can serve as a feasible tool to estimate dietary costs, as well as nutrient intake at the individual level, particularly when excluding ‘eating-away-from-home’ costs.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Findings support the inclusion of receipts-based methods in future nutrition research and highlight the need to better capture eating-away-from-home costs in standard dietary assessments.

Introduction

Improving public health is a cornerstone of public health research, with dietary costs potentially serving as an important modifiable factor for population diets. Nevertheless, it undesirably appears that people who eat healthier spend more money on their purchasing of food than do those who make less healthy choices.¹ It is uncommon for researchers to conclude otherwise,² as on average, healthier food is more expensive than its unhealthy alternative.³ Moreover, in England, an analysis revealed that these price gaps are continuously increasing over time.⁴ As such, it is important to examine whether the above-mentioned food economics mediate certain commonly observed associations, such as the link between lower socioeconomic circumstances and increased mortality.⁵ Indeed, dietary costs have been found to serve as a mediator between socioeconomic position and diet quality.⁶

To address this ‘market failure’, many countries have enacted taxes on various foods that are considered unhealthy in an attempt to lower their consumption (eg, sugar-sweetened beverages). The effect of taxing, however, remains debatable and often difficult to assess,⁷ although some consider this approach promising.⁸ An alternative potential approach is public education regarding the complex relationship between food cost and healthy dietary choices. However, dietary cost considerations are seldom discussed in national healthy eating recommendations.^{9 10} As such, identifying dietary patterns that are both healthy and affordable could be beneficial to public health, yet further research is needed, as are improved diet cost assessment tools.

Monetary diet value can be derived either at the household level or at the individual level. Estimating dietary costs at the individual level, however, can be especially challenging, as different people in the same household often share purchased food, and food that has been purchased can be stored and consumed over large time intervals. Three typical approaches have been previously used to evaluate individual dietary costs11,14: (1) linking a Food Frequency Questionnaire (FFQ) with a price database; (2) linking a food diary to a price database; and (3) conducting a 24-hour recall linked to a price database. The most common methods used to create price databases are supermarket prices, receipt collection or various forms of national price registries. There are, however, limitations that are inherent to such nutritional and cost assessments. The first and third approaches mentioned above are based on participants’ long and short-term memories, respectively, and perceptions. While the second and third approaches, are time restricted (usually less than 7 days) and thus are not always representative of usual consumption. Moreover, price databases often contain a small portion of the true variation in prices, with little to no consideration being given to food that is purchased and eaten away from home.11,14 As a result, errors in estimating dietary costs and nutrition are difficult to avoid.

Cash-register receipts items, an alternative to the price database approach, can be used without implementing the process of linking foods to an estimated price, as they provide a direct record of actual food purchases and their costs. Also, by capturing details on food items purchased over a longer period, this method has the potential to reduce information bias (such as recall/report bias) associated with traditional dietary intake assessment methods and better reflect true consumption patterns.

While previous research has focused primarily on using cash-register receipts to estimate household food costs,¹⁵ their use at the individual level to assess both dietary costs and nutrient intake has yet to be fully explored. Thus, this study seeks to fill this gap by assessing the feasibility and accuracy of using cash-register receipts to estimate individual dietary costs and nutritional intake. We compare this method to the commonly used FFQ and supermarket price methods, focusing on both cost estimation and nutrient intake assessment.

Methods

Study design, sample size and population

This research is a substudy of the Vegan Israeli Study (VIS), an initiative aimed at exploring the relationship between plant-based diets and health.¹⁶ The VIS compares dietary patterns of vegans and omnivores and also has an emphasis on assessing differences in diet quality and cost. As part of the ongoing study, we initiated this pilot project to validate the use of the FFQ in estimating individual dietary costs, while also testing the feasibility of cash-register receipts as an alternative assessment tool.

The required sample size was calculated to detect a medium-strong correlation (α=0.6) between the FFQ and cash-register receipts methods, with a significance level of 5% (α=0.05) and a statistical power of 80%. Using G*Power Statistical Power Analyses for Windows (V.3.1.9.6, Heinrich Heine Universität Düsseldorf), we applied the ‘Correlation: Bivariate normal model’ function to determine that a sample size of 44 participants would be sufficient to detect a strong correlation. This approach ensures adequate power to identify a meaningful association between the two methods.

A total of 45 participants, aged 20–65 years, were recruited to the current study. All study procedures followed the ethical standards of the Declaration of Helsinki, with approval from the Institutional Review Board of Tel-Aviv University. Written or verbal informed consent was obtained from each participant, and verbal consent was formally recorded.

Recruitment and study procedures

Participants were recruited through social media platforms, including targeted Facebook groups, with the main eligibility criterion being that participants regularly purchase the majority of their own food. No financial compensation was provided for participation.

Participants began by completing a short background questionnaire and the appropriate FFQ, tailored to their dietary lifestyle (vegan or omnivore). They were then asked to collect cash-register receipts for all food purchases intended for personal consumption over a 28-day period. Any missing receipts were supplemented by manually recording prices, and participants were also asked to report purchases made for others (eg, family members). In households where two participants resided, each individual classified the items on their receipts according to who consumed them.

Receipts were submitted physically at the end of the collection period or sent digitally via photos through a messaging app, offering flexibility and ease for participants. This comprehensive approach allowed for a more accurate and detailed capture of both dietary costs and food intake patterns over an extended period.

Exclusion criteria

Participants were excluded from the study if they were less than 20 years old, lived outside Israel or had an energy consumption (calculated either based on the receipts or the FFQ) that was outside the range of 500–4000 kcal/day for women and 800–4500 kcal/day for men. Moreover, participants were also excluded if they did not provide receipts for at least 3 weeks (out of the required 28 days), were not the sole purchaser of their food, were unable to separate which food they had purchased for themselves or were on a medically prescribed diet.

Dietary intake and cost assessment

Dietary questionnaires, as previously presented by Avital et al,¹⁶ were used in the study; participants were asked to complete either a 118-item FFQ for omnivores or a 140-item FFQ for vegans, prior to collecting their cash-register receipts. The participants’ reported daily dietary energy intake was calculated based on the data reported in their FFQs as follows:

$${\displaystyle {\displaystyle \begin{array}{l}DailyEnerg{y}_{reported}=\sum \frac{consumptionfrequency\left(\frac{times}{year}\right)\cdot servingsize\left(kg\right)\cdot \left(Caloriesper100gr\cdot 10\right)}{365}\end{array}}}$$

The weight of the items that appeared on the cash-register receipts was either extracted from the receipt if stated or, alternatively, based on similar packets of food that were validated by the participants. Items on receipts were classified by researchers as either ‘for-home-consumption’ or ‘eating-away-from-home’ purchases. The former were defined as purchases made in a supermarket, market, local grocery store or convenience store; all other sources were defined as ‘eating-away-from-home’ purchases (eg, restaurants or fast-food chains). This was accomplished by identifying the business name written on the receipt and using an online search to identify it, and if needed, the participant was asked regarding the specific purchase. Eating at a restaurant, for example, includes not only the raw cost of the food items but also other expenses, such as fees for a chef, service, restaurant rent and other possible costs. This led us to separate both costs, as we hypothesised that the conventional method of assessing diet costs by using supermarket prices linked to an FFQ would not represent the costs actually paid for this component of diet costs. The sum of each item’s weight was divided by the number of collection days (21–28 days) to calculate the average daily consumption. This item weight (g/day) was then entered into the ‘Tzameret’ dietary analysis programme developed by the Israel Ministry of Health¹⁷ for the purpose of nutritional intake analysis. Programme outputs included daily energy (kcal), carbohydrates (g), protein (g), fat (g), calcium (mg) and iron (mg):

$${\displaystyle \begin{array}{ll}{\displaystyle DailyEnerg{y}_{purchased}=}& \sum i=0n\frac{\sum Itemweight{\left(g\right)}_i}{28}\ast \left(Caloriesper1g\right)\\ n=totalnumberofitemsincollectionperiod\end{array}}$$

We included only iron and calcium, as we did not intend, due to time constraints, to assess all micronutrients; iron and calcium are included because they are two well-studied micronutrients.

Only foods containing calories were included, with water and other non-caloric beverages excluded from the cost analysis, as the cost assessment objective was to estimate calorie-based food costs and not hydration. A database containing average food prices for 118 and 140 items in each FFQ was created based on average yearly prices from four large supermarket chains in Israel (Shufesal, Rami- Levi, Victory, Yeinot Bitan).

Assessment of metabolic equivalents and expended energy

The metabolic equivalents (METS)-hours/day variable was estimated according to participants’ answers regarding their leisure-time exercise and work exercise. An assumption was made of 8 hours of work per participant per day. Low-level work exercise was defined as 2 METS/hour, medium 5 METS/hour and high 7.5 METS/hour.^{18 19} Leisure-time exercise level was calculated according to published METS exercises by type of exercise (eg, soccer was considered to be 10.3 METS/hour). Any other hours not reported as exercise were considered 1 MET/hour. METS-hours/day per participant was converted into expended energy (kJ)/day by multiplying METS-hours/day by the participant’s weight in kg.

Statistical analysis

Descriptive statistics were calculated regarding the participants’ characteristics. The distribution of variables was examined for normality using density plots and the Shapiro-Wilk normality test. The Pearson correlation coefficient was computed to assess linear correlations between the two-diet cost/nutritional intake estimation methods (FFQs versus cash-register receipts) and between nutritional estimations. We used the residuals method for energy adjustments.²⁰ We used an exchange rate of 0.2848 for converting new Israeli Shekel (NIS) to US$, representing the exchange rate at the date of the start of the recruitment phase of the study. The strength of the correlation was interpreted according to Dancey and Reidy.²¹ Moreover, the Bland-Altman method was used to evaluate the agreement between both methods,²² with differences in 28-day diet costs and their means as well as 95% limits of agreement computed. Sensitivity analysis was conducted to examine whether additional factors may change diet cost correlations. Due to the small sample size, to avoid very small group sizes, we only used variables that stratify the groups into equal groups (n=15) (eg, median body mass index (BMI) was used). All analyses were conducted using SPSS V.22.0 for Windows or RStudio V.1.1.383.

Results

Study population and participants’ characteristics

Initially, 45 participants volunteered to participate in the study, but 15 were subsequently excluded from the study for a number of reasons: one participant reported improbable energy values, five collected receipts <3 weeks, four decided to drop out of the study, three were unable to separate their food purchases from others in the same household (eg, family members/ flat mates) and two contacts were lost after signing consent forms. Thus, the overall study completion rate was 67% (30/45), with a total of n=30 participants included in the analysis (vegans, n=15; omnivores, n=15). No statistical differences were found between the baseline characteristics of the vegan participants and those of the omnivores (table 1).

Table 1. Participant's demographic and socioeconomic characteristics by questionnaire type.

Characteristic	All (n=30) n/mean	SD(%) range	Vegans (n=15) n/mean	SD(%) range	Omnivores (n=15) n/mean	SD(%) range	P value
Female gender, n, n%	21	70%	9	60%	12	80%	0.69
Age, mean, range	31.9	20.8–50.6	33.5	26.2–47.9	30.3	20.8–50.6	0.22
Marital status, n, n%							0.68
Married/with partner	8	27%	3	20%	5	33%
Other	22	73%	12	80%	10	67%
Smoking status, n, n%							0.60
Current smoker	1	3%	0	0%	1	7%
Past smoker	4	13%	2	13%	2	13%
Non-smoker	25	83%	13	87%	12	80%
Education, n, n%							0.30
High school or less	3	10%	1	7%	2	13%
Associate degree	2	7%	2	13%	0	0%
Academic degree	25	83%	12	80%	13	87%
Income level (NIS/28 days), n, n%							0.51
<4000	8	27%	4	27%	4	27%
4000–7999	12	40%	6	40%	6	40%
8000–12 000	5	17%	2	13%	3	20%
>15 000	5	17%	3	20%	2	13%
BMI (mean±SD; $\frac{kg}{m^2}$), mean, SD	22.6	2.8	22.5	3.0	22.8	2.8	0.79
Energy reported ($\frac{Kcal}{day}$), mean, SD	2298	911	2296	857	2299	993	0.99
Energy purchased ($\frac{Kcal}{day}$), mean, SD	1816	612	1956	643	1676	567	0.22
METS/day, mean, SD	35	6	34	5	37	8	0.21
Energy expended ($\frac{Kcal}{day}$), mean, SD	2243	546	2099	399	2387	642	0.15
‘eating-away-from-home’ food cost ($\frac{NIS}{month}$), mean, SD	426	386	486	371	358	402	0.37

^*vegans versus omnivores.

BMI, body mass index; METs, metabolic equivalents; NIS, new Israeli Shekels.

Since our original goal of n=44 was not met due to time constraint considerations, we emphasise a statistically non-significant result could merely be a result of inadequate sample size, a type 2, false negative error.

According to the questionnaires, the participants’ mean age was 34 years for vegans and 30 years for omnivores, with ages ranging from 21 to 51 years. Most vegans and omnivores were not married (80% and 67%, respectively), had never smoked (80% and 67%, respectively) and held an academic degree (80% and 87%, respectively). Overall, both groups were healthy in weight, with a mean BMI of 22.5 kg/m² for vegans and 22.8 kg/m² for omnivores. In both groups, 67% reported a low-to-medium income (< 8000 NIS/28 days, which is equivalent to 2278 US$/28 days).

Energy consumption and expenditure were as follows: (a) based on the FFQ, the mean reported energy consumed among vegans was 2296±857 (kcal/day) and 2299±992 (kcal/day) among omnivores; (b) based on items listed on the cash-register receipts, the energy purchased was 1956±642 (kcal/day) among vegans and 1676±567 (kcal/day) among omnivores; (c) based on the leisure-time exercise and work exercise questionnaire included with the FFQ and converted into daily expended energy, the reported expended energy was 2099±399 (kcal/day) for vegans and 2387±642 (kcal/day) for omnivores.

Diet cost correlations between methods

Pearson correlations between the two different methods of cost assessment are presented in table 2. Following the exclusion of one outlier (>2 SD), the Pearson correlation between total diet cost according to the ‘cash-register-receipts-items’ method and the ‘FFQ-and-supermarket-prices’ method was weak (r=0.29 (p=0.13) and r=0.23(p=0.23) after energy adjustment). However, a moderate correlation was observed when the ‘cash-register-receipts-items’ method excluded ‘eating-away-from-home’ diet costs, with only ‘for-home-consumption’ being considered (r=0.52 (p<0.004) and r=0.58 (p=0.001) after energy adjustment). The mean total diet cost estimated by the ‘cash-register-receipts-items’ method, 1267 (SD=459) NIS/28 days (361 US$/28-days, SD=131 US$/28 days), was higher than the mean of the reported ‘FFQ-and-supermarket-prices’ method, 985 (SD=454) NIS/28 days (281 US$/28 days, SD=129 US$/28 days). Finally, the mean of the partial ‘for-home-consumption’ diet cost based on the ‘cash-register-receipts-items’ method was 832 (SD=376) NIS/28 days (236 US$/28 days, SD=107 US$/28 days).

Table 2. Comparison between reported and purchased estimates of various dietary nutrients and diet cost (n=29).

	Estimation by purchases on receipts	SD	Reported FFQ estimation	SD	Pearson correlation coefficients	P value	Energy-adjusted Pearson correlation coefficients	P value
Energy (kcal/day), mean, SD	1843	605	2310	925	0.58	<0.001
Protein (g/day), mean, SD	75	39	108	44	0.46	0.01	0.32	0.095
Fat (g/day), mean, SD	68	26	79	39	0.5	0.005	0.35	0.061
Carbohydrates (g/day), mean, SD	208	82	322	157	0.76	<0.001	0.77	0.001
Calcium (mg), mean, SD	824	373	1016	445	0.46	0.01	0.47	0.009
Iron (mg), mean, SD	18	7	28	15	0.37	0.04	0.16	0.413
Diet cost, mean, SD	1267	459	985	454	0.29	0.13	0.23	0.229
‘For-home-consumption’ diet cost, mean, SD	832	376	1267	459	0.52	0.004	0.58	0.001

FFQ, Food Frequency Questionnaire; METS, metabolic equivalents.

Diet cost agreement between methods

Figure 1 presents the differences and averages between each of the two diet cost measurement methods. For 95% of the participants, the total diet costs estimated by the ‘cash-register-receipts-items’ method were 1453 NIS/28 days (414 US$/28 days) higher to 1010 NIS/28 days (288 US$/28 days) lower than those estimated by the ‘FFQ-and-supermarket-prices’ method. However, when stratified below and above the median of ‘eating-away-from-home’ expenses, the 95% limits of agreement range were wider for the above-median subgroup, and the mean differences in diet costs between methods were more extensive for this subgroup. In the group with below-median ‘eating-away-from-home’ expenses, the mean difference in diet costs was −23 NIS (95% CI: −309, 263) (US$7, 95% CI −88, 75), and the 95% limits of agreement were 701 NIS/28 days (199 US$/28 days) higher to 506 NIS/28 days (144 US$/28 days) lower. In the above-median group, the mean difference in diet costs was 466 NIS (95% CI: 107, 827) (133 US$, 95% CI: 30, 236), and the limits of agreement were 1798 NIS/28 days (512 US$/28 days) higher to 838 NIS/28 days (239 US$/28 days) lower (figure 2).

Figure 1. Bland-Altman graph showing differences in 28-day diet cost (‘cash-register-receipts-items’ method A or B minus ‘FFQ-and-supermarket-prices’ method) by monthly mean diet costs for the two methods. (A) Cash-register-receipts method using 28-day total diet cost. The solid line represents the mean of the differences in diet costs: 222 NIS (95% CI: −13, 457). The dashed lines above and below represent 95% limits of agreement between the two diet cost methods. (−1010, 1453). (B) Cash-register-receipts method with 28-day ‘for-home-consumption’ diet cost only. The solid line represents the mean of the differences in diet costs of −202 NIS (95% CI: −384, −19). The dashed lines above and below represent 95% limits of agreement between the two diet cost methods (−1159, 756). FFQ, Food frequency questionnaire; NIS, new Israeli Shekels.

Figure 2. Bland-Altman graph showing differences in 28-day diet cost (‘cash-register-receipts-items’ method using 28-day total diet cost minus ‘FFQ-and-supermarket-prices’ method) by 28-day mean diet costs for the two methods. Graph is stratified by (A) ‘eating-away-from-home’ expenses ≤ median and (B) ‘eating-away-from-home’ expenses > median. (A) The solid line represents the mean of the differences in diet costs: −23 NIS (95% CI: −309, 263). The dashed lines above and below represent 95% limits of agreement between the two diet cost methods (−1035, 989). (B) The solid line represents the mean of the differences in diet costs 466 NIS (95% CI: 107, 827). The dashed lines above and below represent 95% limits of agreement between the two diet cost methods. (−807,1741). FFQ, Food frequency questionnaire; NIS, new Israeli Shekels.

Accuracy of nutritional intake: correlations of macro and selected micronutrients between methods

Pearson correlations between the two methods of nutritional assessment are presented in table 2. The correlation between the energy estimations of both methods was moderate r=0.58 (p<0.001). The average reported energy (2310 kcal/day) was higher than the purchased energy (1843 kcal/day). Moderate correlations were also found between the two methods for assessing macronutrients: protein with an unadjusted r=0.46 (p=0.012) and r=0.32 (p=0.095) after energy adjustment, fat with an unadjusted r=0.5(p=0.005) and r=0.35 (p=0.061) after energy adjustment. Carbohydrate measurements showed a strong correlation between the two methods with an unadjusted r=0.76 (p<0.001) and r=0.77 (p<0.001) after energy adjustment. Among micronutrients, calcium measurements showed a moderate unadjusted correlation r=0.46 (p=0.012) with r=0.47 (0.009) after energy adjustment, and iron measurements were found to have a weaker unadjusted correlation r=0.36 (p=0.049 and r=0.16 (p=0.413) after energy adjustment.

Sensitivity analysis

To further explore whether additional factors may change the diet cost correlations observed between ‘for-home-consumption’ diet cost by the ‘cash-register-receipts-items’ method and the ‘FFQ-and-supermarket-prices’, we stratified the participants by the following: (1) questionnaire type (vegan or omnivore), which showed overall similar results (0.59, 0.021 vs 0.44, p=0.10); (2) BMI <22.5 kg/m² and BMI ≥22.5 kg/m², which showed overall similar results in the BMI <22.5 kg/m² group compared with the other group (R=0.40, p=0.13 vs R=0.38, p=0.18), as shown in table 3.

Table 3. Pearson correlation between purchased ‘for-home-consumption’ diet cost and reported FFQ diet cost, stratified by subgroups.

Subgroup	Pearson correlation coefficients	P value
Questionnaire type
Vegan (n=15)	0.59	0.02
Omnivore (n=15)	0.44	0.10
BMI category
BMI<22.5 (n=15)	0.40	0.13
BMI≥22.5 (n=15)	0.38	0.18

BMI, body mass index; FFQ, Food Frequency Questionnaire.

Discussion

In this paper, we assessed the feasibility and agreement between two dietary cost and nutritional intake assessment methods (FFQ compared with cash-register receipts) in 30 young, healthy adults who follow two different dietary patterns (omnivore vs vegan). Our main findings showed a stronger correlation between the partial ‘for-home-consumption’ dietary cost and the conventional ‘FFQ-and-supermarket-prices’ method (r=0.52, p<0.004) than the correlation between the full costs considered by the ‘cash-register-receipts-items’ method and the conventional method (r=0.29, p=0.13). We also found improved limits of agreement between these two methods in the analysis containing the partial ‘for-home-consumption’ diet cost (−1159, 756 NIS/28 days) compared with including the total dietary cost in the analysis (−1010, 1453 NIS/28 days). These findings could be explained in a number of ways. First, this may stem from differences between ‘eating-away-from-home’ prices and the supermarket prices used in the ‘FFQ-and-supermarket-prices’ method. In addition, this may have been impacted by possible under-reporting of the FFQ of foods eaten away from home. Stratifying below or above the median of ‘eating-away-from-home’ diet costs shows that on average, ‘eating-away-from-home’ spending below the median was closer to the FFQ estimate than dietary costs above the median. The mean difference in dietary costs between methods was 466 NIS (95% CI: 107, 827) in the above-median subgroup compared with −23 NIS (95% CI: −309, 263) in the below-median subgroup—an observation that lends support to the prior explanation. Finally, the average total dietary cost estimate according to the ‘cash-register-receipts-items’ method was higher than the average derived by the ‘FFQ-and-supermarket-prices’ method, which may be a reflection of higher food costs for ‘eating-away-from-home’ food—if true, this provides further support in favour of the first explanation. If confirmed, these explanations could be beneficial for investigating dietary costs in future studies.

This finding suggests that the FFQ method may under-represent total dietary costs, particularly when food consumed outside the home is not accurately captured. In addition to the primary aim of this study, we further investigated the possibility of using the cash-register receipts instrument to capture dietary intake by extracting nutritional information from receipt items. For nutrients, the energy-adjusted macronutrient correlations ranged from as high as 0.77 for carbohydrates to a medium strength of 0.32 for protein. Micronutrients showed somewhat weaker correlations, ranging from a medium strength of 0.47 for calcium to a weak strength of association of 0.16 for iron. These observations support the use of the ‘cash-register-receipts-items’ instrument as a possible alternative to commonly used tools, such as the FFQ, 24-hour-diet-recall and food diaries. Future nutrition researchers may find this method of extracting nutrient information from items on cash-register receipts to be practical, as it does not rely on participants’ memory or perceptions.

According to the Israeli Central Bureau of Statistics household expenditure survey,²³ in 2019, the average monthly per capita expenditure on food in Israel is about 800 NIS/28 days (228 US$) and the average monthly household expenditure is 2260 NIS (644 US$/28days). In our study, individuals spent 1267 NIS/28 days (361 US$) on average.

To the best of our knowledge, this is the first time that the ‘cash-register-receipts-items’ method has been used to evaluate individual-level dietary costs in this manner. Although we believe it offers a close ‘true’ estimate of dietary costs, no comparable methods currently exist that could be considered the gold standard. This approach, however, has a number of limitations. First, to accurately assess individual-level estimations, we primarily recruited single participants and participants who could separate purchases made for their own consumption. This limits the external validity of our findings, which require further testing in additional settings. In addition, most of our participants appeared to be health conscious and educated, as reflected by their relatively low BMI and high prevalence of academic education. Moreover, our approach of using items on receipts differs greatly from common methods that entail the recording of food consumption through standard tools such as the FFQ, 24-hour-diet-recall and dietary diary.^{12 24 25} In this study, we did not record the food that was actually consumed but rather only food that was purchased for the purpose of future consumption. Therefore, this instrument may be vulnerable to food purchase patterns, such as participants who buy large quantities of long shelf-life food in advance. Such behaviour may lead to the overestimation or underestimation of monthly food purchases, as participants may buy surplus or incomplete purchases in the period of receipt collection. Further research could therefore benefit from conducting such receipt collection studies over a longer period of time to overcome possible bias caused by infrequent large purchases of long-life items.

However, this method offers the advantage of decreasing the burden on participants who are otherwise asked to keep a written dietary diary over time or implement another tool for diet ascertainment. This method also provides the ability to capture longer periods of food consumption with greater ease—a feature that is lacking in many dietary assessment tools. Further research is also warranted to test the benefits of our approach to separate food purchases by self-reporting, in contrast to the method used by Ransley et al,²⁶ in which food purchased was divided by factors representing each member of the household (eg, a child was considered to purchase 0.62 of the energy of an adult). Finally, this method also decreases the burden on the researchers themselves, as the labour-intensive task of linking dietary data to receipt costs, as previously reported by Monsivais et al,¹² is not needed.

Conclusion

Although an essential public health endeavour, capturing individual dietary costs in epidemiological research poses numerous challenges. By recognising that dietary costs are composed of two major components—‘for-home-consumption’ and ‘eating-away-from-home’ expenses—we were able to identify that dietary costs based on the ‘FFQ-and-supermarket-prices’ method are more agreeable and strongly correlated with objectively documented food purchases when ‘eating-away-from-home’ costs are excluded. Moreover, the current study demonstrated that collecting food receipts can be used not only to estimate dietary costs but also to obtain nutritional information. Although understanding how to properly measure dietary costs is critical, our results suggest that the ‘FFQ-and-supermarket-prices’ method does not adequately capture ‘eating-away-from-home’ diet costs. As such, future studies should measure ‘eating-away-from-home’ costs or acknowledge non-measurement as a weakness in their analysis. To our knowledge, this is the first study to validate the use of cash-register receipts as a tool for estimating both dietary costs and nutrient intake at the individual level. This method offers a unique opportunity to minimise recall bias, capture actual food expenditures and provide a more direct measurement of nutrient consumption compared with traditional methods.

Data availability statement

Data are available upon reasonable request.