Poor eating habits and predictors of weight gain during the COVID-19 quarantine measures in Kuwait: a cross sectional study

Study background

This cross-sectional survey was conducted from 2^nd to 12^th April 2020 among the general public in Kuwait City, Kuwait. The country reported its first case on 24^th February 2020, and the country went into partial lockdown since 22^nd March, and on 10^th May, the country was placed in full curfew based on Ministry of Health’s assessment⁷ (at the time of writing, on 4^th August 2020 Kuwait will remain under curfew 9pm-3am). All data were collected through a social media platform (WhatsApp groups), where first participants were chosen through convenience sampling and the rest through snowball sampling methods. Attempts were made to exclude non-Kuwaiti nationals by distributing the questionnaire to Kuwaiti nationals and screening the datasets after collection of data. Inclusion criteria included participants being Kuwaiti nationals residing in Kuwait. There were no specific exclusion criteria pertaining to age, language, gender or occupation of the respondents. After providing written informed consent, the respondents were briefed about the objectives of the survey, ensured anonymity and that the group findings will be reported. Respondents could leave the study at any time if they felt uncomfortable. Approval for this study was provided by the Ethical Review Board of Ministry of Health of Kuwait, Kuwait (Approval #2020/1477).

Survey

The minimum sample size required for this survey was calculated to answer out primary research objective of documenting prevalence of poor eating habits among Kuwaitis. Based on a statistical power of 80%, confidence level 95%, margin of error 5% and a population size of 1.3 million Kuwaitis, a minimum sample size of 385 was required. All data were collected using an electronic version of a bilingual questionnaire in Arabic and English languages. The questionnaire was pre-piloted among a few participants before distribution; however, no factor validity or reliability testing was done due to categorical and open-ended nature of the responses. Participants were encouraged to provide their responses in either of the languages. The questionnaire comprised of three sections²⁰: a) demographic characteristics of respondents, b) eating habits particularly binge eating, consumption of snacks and beverages c) subjective feelings of anxiety and d) weight before and during the pandemic. Demographic characteristics included gender, age, nationality, marital status, parity and housing of the respondents. Several questions were asked regarding binge eating habits documenting change in nutrition habits, time of increased food consumption, number and type of snack foods consumed during the day, water and coffee intake. Physical activity during the day was inquired using two questions: a) a trichotomous question documenting change in physical activity when compared with normal routine before the quarantine and number of hours spent in a sedentary manner per day in the last week. Symptoms of anxiety were measured subjectively using a three-point Likert scale. Participants were asked the extent to which they experienced anxiety every day in last 7 days. Responses were given on a three point scale ranging from never to always. Participants were instructed to respond with their weights measured using a weight scale (digital or analogue) if they had measured it. Similarly, they were instructed to input their current weight status using a digital or analogue weighing scale, if available. The question regarding weight of the respondents before and during the lockdown, yielded 462 (88.50%) valid responses. The remaining participants could not input their weights due to unavailability of weighing scales.

Statistical analysis

All data were analyzed using SPSS (v25). Descriptive statistics were calculated for all variables where quantitative variables were presented as mean (SD) and categorical ones as frequency and percentage. Statistical significance for difference between weight of respondents during the COVID-19 pandemic and before it was assessed using t-test for dependent samples. To delineate predictors of weight gain, it was planned a priori to run multiple regression; however, due to the non-normalized nature of the data, logistic regression analysis was performed.

Thereafter, the respondents were categorized in to two groups: participants with weight gain and participants with negative or no change in weight. To identify the predictors of experience in weight gain among the Kuwaitis, binary logistic regression (stepwise method) was employed. Theoretically sound factors showing relationship with weight gain in previous literature were added as predictors in the logistic regression model⁸. These predictors included demographic characteristics of respondents, eating habits c) subjective feelings of anxiety and d) questions pertaining to physical inactivity. The final model reported significant predictors of weight, with statistical significance set at ≤0.05 and the variance in the model was assessed using the Negelkerke R².

Characteristics of respondents

There was a total of 522 valid respondents, with a mean age of 41.78 (11.75) years. More of the respondents were females (380, 72.8%) than males (142, 27.2%). A higher proportion of the respondents were married (379, 72.6%) than single (97, 18.6%). A total of 201 (38.5%) respondents reported having 1–3 children, followed by >3–6 (170, 32.6%), no children (124, 23.8%) and more than 6 children 27 (5.2%). More than half of the respondents lived in their own houses 297 (56.9%), followed by at parents’ house 128 (24.5%), and apartment 97 (18.6%).

Weight and eating patterns before the COVID-19 pandemic

A higher proportion of the respondents’ families were assisted by a cook (299, 57.3%) to prepare meals than themselves (223, 42.7%). A change in nutritional habits was reported by 387 (74.1%) respondents, with 214 reporting that their eating pattern had improved to a healthier one (214, 41%) than unhealthy one (24.3%). Respondents tend to consume more food at night (232, 44.4%) than during daytime. A total of 352 respondents (67.4%) reported consuming snacks more than twice per day. Water intake was self-reported as enough (287, 55.0%) and 300 (57.5%) respondents reported having 1-3 cups of coffee a day.

For the question regarding weight of the respondents before and during the lockdown, there were 462 (88.50%) valid responses. We did not impute the missing data for responses regarding weight measurements, and excluded participants with missing data in the logistic regression model. Mean weight of respondents before the pandemic was 75.03 (16.85) kg and 75.74 (17.54) kg during the pandemic (Figure 1). There was a significant increase in weight of respondents during the quarantine (mean difference= -1.13, SD 5.39, t= -4.52, p < 0.001). A higher proportion of respondents reported that their weight will stabilize after the pandemic (n=305, 58.4%) while 217 (41.6%) reported that it would increase. A total of 333 (63.8%) of the respondents reported feeling anxious sometimes during the day, 133 (25.5%) always and 56 (10.7%) never. Over 69% of the respondents reported that their physical activity had reduced than before. Mean number of hours spent being sedentary at home were 9.56 (5.64) (Table 1).

Figure 1. Distribution of weight of respondents before and after COVID-19 quarantine.

Predictors of weight increase among the Kuwaitis

Logistic regression (stepwise method) was used to analyze statistically significant predictors of increase in weight among the Kuwaitis during the COVID-19 quarantine. It yielded a significant model (X²= 90.43, df= 8, p< 0.001). The model explained 24.9% of variance in weight changes among Kuwaitis (Nagelkerke R²) and correctly classified 71.6% of the cases, whereas the sensitivity of this particular model was 65.8% and specificity 78% in predicting the increase in weight among Kuwaiti respondents. According to this model, when compared with respondents who changed their diet pattern to healthier, those with reporting unhealthy diets were 4.5 times (95% CI= 2.45 to 8.23) more likely to report an increase in weight. Those reporting having anxiety throughout the day were 2.45 times more likely to have an increase in weight than those never experiencing it. Consuming snacks excessively

(>3 times a day) was associated with 3.27 times higher odds of increase in weight than those not consuming it. Those that consumed moderate amounts of snacks (1–3 times a day) did not differ to their counterparts who never consumed snacks throughout the day (Table 2).

Limitations

Despite its large sample size, this study has several limitations which need to be considered for careful interpretation of its results. Most important limitations include a convenience and snowballing approach for collection of data as opposed to random sampling. Cross-sectional nature of this study limits inferences related to causality and temporality and use of a subjective question for estimating anxiety levels limit validity of these results and also introduce recall bias. Future studies should consider a prospective study design and use of valid and reliable questionnaires like the Generalized Anxiety Disorder 7-item scale¹⁸.

Underlying data

Figshare: Diet Kuwait.csv, https://doi.org/10.6084/m9.figshare.12706223.v1¹⁹.

Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).

Extended data

Figshare: CovidKuwaitWt-Survey.pdf, https://doi.org/10.6084/m9.figshare.12709631.v1²⁰.

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).