Since the beginning of the coronavirus disease 2019 (COVID-19), China policymakers published a series of dramatic containment measures against the COVID-19 pandemic to protect people’s life safety and health. From the strict dynamic zero-COVID policy to the current resumption status, China’s society has gradually returned to normality and resumed work and production. During the COVID-19 pandemic, the medical staff has suffered from a myriad of stressors, such as heavy workload, unstable working setting and hours, strict safety measures, physical exhaustion, and infection risk, which lead to dramatic effects on their work and lifestyles (1). Specifically, poor diet quality not only affects the quality of life, but also leads to adverse health outcomes, such as obesity, metabolic syndrome, diabetes, sarcopenia, and heart diseases (2–10). Consequently, it is necessary to investigate the diet quality of medical staff to provide support to the development of dietary guidelines and nutrition plans for individuals, hospital managers, and national decision-makers.

The Chinese government attaches great importance to people’s health and revised the Dietary Guidelines for Chinese Residents (DGC) in 2022 to guide a healthy diet. Previous studies have pointed out that improving people’s nutrition knowledge could effectively support sound dietary intake and quality among the community or a specific target population (11, 12). Numerous socioeconomic factors including education, income, and occupation influence diet quality. As epidemiologic evidence showed, poor-quality diets were prevalent among individuals from low socioeconomic status (SES) backgrounds (13). Hashimoto et al. found a positive association between education and diet quality and suggested that the diet quality score of high or middle education was higher than low education for both generations (14). Another study showed that diet quality differed by income level, and individuals in the lowest income groups were at greater risk of lower overall diet quality and inadequate nutrient intake (15). A systematic review has provided evidence for the effectiveness of health promotion interventions at the workplace on physical and mental health outcomes (16). The COVID-19 pandemic has brought greater work stress on the medical staff. Work stress and conflict have a great impact on their diet choices and physical health (1). Work–family conflict refers to a lack of overall fit between work and family life, which is a source of stress at the interface of work and family life (17). Participating in multiple roles would inevitably drain substantial time and energy resources, leading to psychological arousal, strain symptoms, emotional eating, and a less healthy diet. As a study showed, work–family conflict and emotional eating were positively related to unhealthy diets (18). A multinational cross-sectional study indicated that job burnout is prevalent among healthcare workers during the COVID-19 pandemic, due to a strong association between SARS-CoV-2 testing, safety attitudes, job role, redeployment, and psychological status (19). The doctors and nurses have to adapt to a new setting during the outbreak, leading to various changes in lifestyles as well as sleep quality (1, 20–24). Smoking, exercise, being overweight or obese, and unhealthy sleep patterns were associated with poor diet quality (25–27). As a type of behavior within a group of various eating behaviors, emotional eating is an important risk factor for poor diet quality and is generally influenced by stress, emotions, lifestyles, and attitudes toward eating (28). Especially in people with obesity, a study has revealed that emotional eaters had lower intakes of vitamins, fiber, folic acid, magnesium, and potassium, while they had a higher intake of sodium, lipids, and fats (29). Additionally, Gacek et al. conducted a study and revealed significant correlations between personality traits and diet quality (30). In practical terms, the pro-healthy diet index showed an increase with increasing extraversion and decreasing neuroticism, while agreeable persons got a lower non-healthy diet index (30). Overall, diet quality is predicated by numerous factors, such as personality traits, socioeconomic status, lifestyles, and individual and working conditions.

Understanding the impact of the COVID-19 pandemic on the diet quality of healthcare workers is crucial. Along with the rapid accumulation of raw information data, machine learning technology plays a key role in modeling algorithms and human life (31, 32). Deep learning technology has demonstrated significant value and outstanding performance in practical applications. Due to their excellent predictability, deep learning models are widely used in various fields, including business, weather, transportation, language recognition, medicine, etc. The nonlinear data transformation performed by hidden layer neurons is the core of deep learning technology. During the training phase, the deep neural network continuously updates the weights of connections between layers, making the results more likely to be close to the target values. Neural networks learn decision functions through a large number of samples. Studies showed that Deep Neural Network (DNN), a powerful tool of machine learning, could provide an automated identification mechanism for various adverse health outcomes, such as body mass index (BMI) changes, diabetic retinopathy, neurological disorders, etc. (1, 33, 34). Given the limited studies focusing on diet and nutrition prediction, applying DNN technology to fill this gap has great benefits. Therefore, this study aimed to develop a diet quality prediction model by applying Deep Neural Network with samples from COVID-19 designated hospitals, and further identify the predictive effects of personality traits, socioeconomic status, lifestyles, and individual and working conditions on diet quality, providing the scientific basis for diet and nutrition improvement among doctors and nurses during the COVID-19 pandemic.

The COVID-19 pandemic has placed unprecedented pressure on front-line healthcare workers, leading to poor health status, especially diet quality. This study recruited 5,013 doctors and nurses from COVID-19 designated hospitals in north China as samples. The mean score of diet quality for all healthcare workers was 46.14 ± 15.08 (the questionnaire score ranging from 0 to 100), indicating poor quality of diets. Specifically, the mean scores for variety, adequacy, moderation, and overall balance were 14.33 ± 3.65, 17.99 ± 5.73, 9.41 ± 7.33, and 4.41 ± 2.98, respectively. Moreover, results showed that diet quality was significantly associated with gender, age, marital status, education, length of employment, income, overtime, frequency of locked down, exercise, BMI, nutrition knowledge, negative emotional eating, positive emotional eating, sleep quality, neuroticism, extraversion, openness, work–family conflict, healthcare worker-patient conflict, emotional exhaustion, depersonalization. Furthermore, the current study developed a Deep Neural Network (DNN) model with a 21–30–28-1 network framework for diet quality prediction, explaining how multiple factors, including work conditions, socioeconomic status, lifestyles, and personality traits, could collectively predict the diet quality of doctors and nurses during the COVID-19 pandemic. In the DNN model, 21 factors were identified as input features, and this deep neural network includes two hidden layers. The DNN model achieved high prediction efficacy, and values of R², MAE, MSE, and RMSE were 0.928, 0.048, 0.004, and 0.065, respectively. Among doctors and nurses in north China, the top five predictors in the diet quality prediction model were BMI, poor sleep quality, work–family conflict, negative emotional eating, and nutrition knowledge.

Results of this study indicated that BMI was the most important predictor of diet quality among doctors and nurses in north China. Of participants in the current study, 83.9% had a normal weight, and 8.9 and 5.2% were overweight and obese, respectively. This study showed that individuals with overweight or obesity had poorer diet quality than individuals with normal weight. Unhealthy eating patterns are one of the important factors for overweight and obesity, and in turn, overweight or obesity impacts diet quality. Previous studies have suggested that the attention bias to food of individuals with overweight or obesity was different from that of individuals with normal weight (43, 44). Food-related attentional bias triggers craving and overeating, especially for individuals with overweight or obesity. People with a higher BMI tend to engage in unhealthy eating habits, such as eating irregularly, a faster eating speed, and a diet high in sugar, etc. (45). It is worth noting that individuals are in a ‘loop’ in the real world, and there is a meaningful self-reinforcing mechanism. Specifically, most obese individuals tend to have a diet with lower quality, which in turn promotes individual weight gain or loss, leading to changes in BMI over time. During the COVID-19 pandemic, healthcare workers were busy on the front line of COVID-19 prevention and treatment, and there was increasing workplace heavy pressure on them. Based on the results of this study, doctors and nurses who had a higher BMI were more likely to eat various high-sugar, high-fat and high-calorie foods to reward themselves, resulting in poor food quality. In addition, due to the irregular working hours, the eating time and eating speed might be inappropriate, which would lead to unconscious overeating. These findings suggest that healthcare workers with a higher BMI should pay more attention to their eating habits and nutritional structure to improve their diet quality.

Among doctors and nurses, poor sleep quality was the second most important factor for diet quality prediction. During the COVID-19 epidemic, a considerable proportion of healthcare workers experience sleep disturbances (46), such as insomnia, short sleep duration, and poor sleep efficiency. The Chinese government established a joint prevention and control COVID-19 team to lead all designated hospitals and healthcare workers’ scientific response to outbreaks; Moreover, doctors and nurses have to adopt 7*24 on-call to keep patients’ life safe and sound. The results of this study determined the important predictive effect of sleep quality on diet quality among doctors and nurses. Indeed, it is well documented that poor sleep quality adversely impacts dietary behaviors (47–49), including higher intakes of energy, fat, and sweets as well as lower intakes of foods shown to have health benefits, including vegetables, fruits, and whole grains. A study has evaluated the cyclical nature of the sleep-diet relationship in adults and revealed various underlying mechanisms of the relationship between sleep and dietary intakes (50). On the one hand, there were increased opportunities to eat and alterations in appetite-regulating hormones (including leptin, ghrelin, adiponectin, glucagon-like peptide 1, and orexin) due to the sleep restriction and added wake time; on the other hand, increased hedonic drive for foods may explain shifts toward poor diet quality following periods of inadequate sleep (50).

Work–family conflict also greatly affected diet quality and ranked third in the DNN model. Increased stress during the pandemic has changed the nature of workplace and family patterns, leading to several work–family conflicts, such as insufficient time to take care of the families and reduced housework completion. An epidemiological study suggested that work–family conflict was implicated in external eating, unhealthy food consumption, poor eating style, and eating quality (18). With the increasingly severe conflict between work and family, unhealthy eating patterns such as irregular meals, fast meals, night meals, and high-energy snacks in healthcare workers are showing a growth tendency. In fact, the ability to balance work and family has a significant impact on healthy eating patterns, family values, work values, and the division of labor in the household, particularly during public health emergencies. Besides, work–family conflict is not only a predictor in explaining unhealthy eating but also a tendency to emotional eating. As the current study indicated, negative emotional eating was the fourth most important predictor of diet quality during the COVID-19 pandemic. Stress and negative emotions may impact appetite, unhealthy eating, and even eating disorders (29, 51, 52). During the COVID-19 pandemic, an Italian survey declared that almost half of the individuals (aged 18–79 years) felt anxious, and consumed comfort food to feel better (53). Individuals with emotional eating tend to increase intakes of energy and macronutrient and have particular food choices, such as fast-food intake, salty snacks, sweet high-fat foods, or energy-dense foods (29). These findings suggest healthcare workers should make efforts to balance work-family, pay attention to their emotional state, reduce emotional eating, and then improve their overall diet quality.

Particularly, nutrition knowledge has a lot of benefits for diet quality, ranking fifth in the DNN prediction model. Nutrition knowledge could play an important role in the nutrition habits and food consumption of an individual, and then change an individual’s diet quality (12). As previous studies showed, an increase in the nutrition knowledge level is related to an increase in the consumption of essential food groups, such as grains, dairy, meats, beans, and so on (54, 55). Especially during the COVID-19 pandemic, nutritional knowledge was significantly associated with the consumption of healthy foods to ensure a robust immune system (56). A healthy eating project in Poland suggested that increasing individuals’ nutrition knowledge was an important target to improve their lifestyles and dietary quality throughout the whole lifespan (57). Results of the current study have identified the predictive effect of nutrition knowledge on diet quality and suggest that interventions of increasing nutrition knowledge can be a promising approach to improving diet quality among doctors and nurses during the COVID-19 pandemic.

This study sheds more light on diet quality prediction in doctors and nurses with a large sample size (N = 5,013) and an effective machine learning technology. Based on all the above discussion, these findings indicate that integrated interventions should be taken to improve diet quality among doctors and nurses, particularly weight management, sleep quality improvement, work-family balance, decreased emotional eating, and increased nutrition knowledge. More specifically, establishing a gym and an emotional release room in the hospital can not only facilitate emotional release but also facilitate weight management, which has a significant effect on reducing emotional eating and unhealthy dietary tendencies. Moreover, it is recommended to set up a lounge within different departments for healthcare workers to have lunch breaks and night shifts. To effectively alleviate work–family conflicts, hospital managers can try the following measures: (1) work arrangements are clearer, including fixed working hours, clear work content and processes, clear division of labor, and performance evaluations that pay more attention to work results; (2) cutting between work and personal environments; (3) improving the construction of hospital affiliated kindergartens, primary schools, and other facilities. In addition to providing nutritious working meals, the hospital cafeteria can also conduct more dietary nutrition education, including promotional slogans, posters, brochures, etc., to further improve the nutritional knowledge of healthcare workers.

Some study limitations need to be noticed. Firstly, all participants were recruited from COVID-19 designated hospitals, leading to a potential sampling bias which might reduce the generalization of these results in other medical settings. Secondly, the self-reporting questionnaires were used in the present study and might cause some evaluation errors. Thirdly, since loss of taste and smell are defining symptoms of COVID-19, the lack of data on infection may be a limitation in developing a diet quality prediction model. The data on COVID-19 infection of hospital workers should be collected in future studies. Finally, future studies should collect more information about the national environment and policy to avoid the limited input characteristics, aiming to predict the diet quality of healthcare workers more accurately.

From the strict dynamic zero-COVID policy to the current resumption status, the COVID-19 pandemic has placed heavy pressure on front-line doctors and nurses in China. The current study has developed a Deep Neural Network (DNN) model for diet quality prediction in doctors and nurses with a large sample size (N = 5,013) in north China. Results showed that poor diet quality was prevalent among doctors and nurses and deep neural networks could provide an automated identification mechanism for diet quality prediction. Personality traits, socioeconomic status, lifestyles, and individual and working conditions have contributed to diet quality prediction. Our DNN model revealed that the top five predictors for diet quality were BMI, poor sleep quality, work–family conflict, negative emotional eating, and nutrition knowledge. The current study supports the scientific basis for sustainable diet and nutrition improvement among doctors and nurses. Integrated interventions can be a promising approach to improving diet quality among doctors and nurses, particularly weight management, sleep quality improvement, work-family balance, decreased emotional eating, and increased nutrition knowledge.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by the Ethics Committee of Harbin Sixth Hospital. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

AL designed the study. QW and HC drafted the protocol, analyzed the data, and wrote the manuscript. HL, CL, SL, HF, DL, TD, and JL participated in the data collection and analyzed and discussed the results. All authors contributed to the article and approved the submitted version.