The number of people with diabetes in China has reached 121 million. The Bayesian sparse latent factor model can include multiple influencing factors at the same time to obtain the food patterns of a study population.

This study investigates the relationship of diabetes with food items and dietary patterns among residents of Zhongshan Street, Songjiang District, Shanghai.

The study population were the residents of Zhongshan Street, Songjiang District, Shanghai, who were Shanghai registered residents, aged 20-74 years, and had lived in the community for more than half a year. A total of 3 587 participants were included into the analysis. They were divided into healthy people (n=458, who did not self-report diabetes, coronary heart disease, hyperlipidemia, etc., and whose on-site health examination results did not show diabetes), diabetes patients (n=458, who self-reported diabetes, and whose on-site health examination results confirmed diabetes), and new diabetes patients (n=276, who self-reported to be healthy, but whose on-site health examination results confirmed diabetes). Food intake data were collected using food frequency questionnaire (FFQ) which included 29 food items. Participants' daily intake was assessed by examining how often each food was consumed and how much was consumed each time. All food intake was standardized by subtracting the mean and then dividing by the standard deviation to incorporate subsequent Bayesian latent factor models to derive dietary patterns. Dietary patterns were analyzed in both healthy people and diabetes patients. Then the healthy people and the new diabetes patients were included in the third model to analyze the relationship of incidence of diabetes with single food factors and dietary patterns.

Except marital status, there were significant differences in gender, age, and retirement among the three groups (P=0.0025, < 0.000 1, and < 0.000 1, respectively). Comparisons of food intake showed that there were significant differences among the three groups in the intakes of rice (mean values were 288.52, 256.88, and 304.48 g·d^-1, respectively), fruit (mean values were 127.52, 79.77, and 95.15 g·d^-1, respectively), yogurt (mean values were 35.45, 17.20, and 19.09 g·d^-1, respectively), soy milk (mean values were 26.01, 16.24, and 17.83 g·d^-1, respectively), carbonated beverages (mean values were 13.58, 3.00, and 6.38 g·d^-1, respectively), pure fruit and vegetable beverages (mean values were 9.16, 2.67, and 5.09 g·d^-1, respectively), confectionery chocolate (mean values were 2.11, 0.16, and 0.99 g·d^-1, respectively), and pastries (mean values were 11.05, 8.09, and 8.61 g·d^-1, respectively). Five dietary patterns and three dietary patterns were derived from the healthy people and diabetes patients respectively. There were no relationships between dietary patterns and the incidence of diabetes after including the healthy people and the new diabetes patients in the third model, but relationships between some food items and diabetes were observed. The intakes of dark vegetables, milk, yogurt, other livestock meat (except pork), freshwater fish, sea fish, and shrimps, crabs, and shellfish had negative relationships with the incidence of diabetes (the factor loadings were -0.45, -0.12, -0.16, -0.13, -0.23, -0.48, and -0.14, respectively), and the intakes of fresh vegetables and processed meat had positive relationships with the incidence of diabetes (both factor loadings were 0.12).

Diabetes will change dietary patterns. In the selected new diabetes group and healthy group in the study, dietary patterns have no relationships with diabetes. The intakes of certain food items may be associated with the incidence of diabetes.