The shaded region in panels A-D represents the 95% CI. In panel E, the bars represent the order amount (ie, how many orders were placed by customers each day). The lines are local polynomial smooth lines of order amount. The left and right vertical lines represent the beginning and end of the Circuit Breaker period.
eFigure. Empirical Strategy
eTable. Dictionaries of Outcome Variables
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Agarwal S, Huang P, Luo C, Qin Y, Zhan C. Assessment of Online Food Ordering and Delivery in Singapore During the COVID-19 Pandemic. JAMA Netw Open. 2021;4(9):e2126466. doi:10.1001/jamanetworkopen.2021.26466
In 2020, the COVID-19 outbreak was declared to be a pandemic. Governments implemented various measures, such as social distancing, quarantine, and lockdowns, to minimize the spread and effects of COVID-19. Previous studies1,2 based on questionnaires showed that respondents reported unhealthier eating habits during lockdown. Unhealthy eating behavior formed during the pandemic could result in undesirable long-term health consequences, such as coronary heart disease,3,4 chronic diseases,5 and Alzheimer disease.6
This cohort study was deemed exempt from full review by the faculty ethics review committee at the National University of Singapore, and informed consent was waived because of the deidentified and encrypted nature of the data. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
In this cohort study, we used an order-level data set from a Singapore-based online food ordering and delivery platform to investigate the association between Singapore’s lockdown (ie, Circuit Breaker [CB]) and eating habits. Each observation with food items was an order placed by an encrypted customer. We coded Chinese New Year (CNY) as day 1. Day −89 to day 220 of 2020 CNY was the treatment period, while day −89 to day 220 of 2019 CNY was the control period.
We estimated the difference-in-differences (DID) model, which assumed similar dietary trends in the 2 periods, to test whether customer diet changed during CB and post-CB period. In the first equation, dietrct denoted the outcome (ie, vegetable, barbecue or fried food, beverage, or dessert) of restaurant r and customer c on date t. Each of the outcome variables equaled 1 if any item could be matched with keywords specified in our dictionaries (eTable in the Supplement), otherwise, 0. Dummy treatt denoted 1 for the treatment period or 0 for the control period. Dummies cbperiodt and aftert represented the CB period (day 74 to day 129) and the post-CB period (day 130 to day 220), respectively. Coefficients β1 and β2 were dietary changes during the 2 periods. Day −89 to day 73 was the benchmark period.
In the second equation, the sample was further divided into 31 ten-day periods (represented by j) and interacted with treatt to analyze the dynamic dietary changes.
The first 10-day period was the benchmark. eFigure in the Supplement further indicated research design and model details. The software was Stata 15 (StataCorp, LLC). We performed 2-sided t tests, and statistical significance was set at P < .001. The statistical analysis was performed from August 2020 to July 2021.
This cohort study included 42 495 obversions involving 11 372 customers and 462 restaurants. Relative to the benchmark, during CB, the probability that the order contained vegetables decreased by 15% (95% CI, 12%-19%), while the probability of an order in the barbecue/fried food or beverage category increased by 11% (95% CI, 8%-14%) and 4% (95% CI, 2%-6%), respectively (P < .001) (Table). During the post-CB period, these patterns persisted. Dynamic changes in the Figure also showed that the probability of ordering vegetables decreased by 21% (95% CI, 6%-37%) during day 61 to day 70 when the advance work-from-home order was implemented and the lockdown was announced (P < .001), relative to the benchmark, and this trend persisted till the end of the sample period.
After excluding customers who placed their first orders during the treatment period to eliminate the concerns about new customers’ distinct tastes, our results were consistent with baseline results. No significant pre-CB changes were found when we included a dummy beforet, which indicated the pre-CB period between the confirmation of the first infection to the day before CB began.
We used a data set of real food orders instead of surveys, and this ordering data could reflect the actual changes in eating habits. Our results suggested an association between customers’ diets and unhealthier eating habits during the COVID-19 lockdown, and changes in eating habits persisted after lockdown measures were removed.
This study has limitations. We could not measure exact proportions of nutritional ingredients. More data is needed to study the long-term dietary changes and health outcomes. Our findings suggest that dietary changes occurred during and after the pandemic. Governments should properly guide the public regarding their dietary choices when implementing lockdown policies.
Accepted for Publication: July 21, 2021.
Published: September 23, 2021. doi:10.1001/jamanetworkopen.2021.26466
Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2021 Agarwal S et al. JAMA Network Open.
Corresponding Author: Changwei Zhan, MS, Department of Real Estate, National University of Singapore, 15 Kent Ridge Dr, Singapore 119245 (email@example.com).
Author Contributions: Dr Agarwal had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs Agarwal, Luo, and Qin and Mr Zhan equally contributed to this paper.
Concept and design: Agarwal, Qin, Zhan.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Zhan.
Critical revision of the manuscript for important intellectual content: Agarwal, Huang, Luo, Qin.
Statistical analysis: Qin, Zhan.
Obtained funding: Agarwal, Qin.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was funded by grant MOE2019-SSRTG-024 from the Ministry of Education, Singapore.
Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.