Consumption of home-prepared meal at workplace as a predictor of glycated haemoglobin among people with type 2 diabetes in Hong Kong: a mixed-methods study

The study addresses how workplace-conditioned eating patterns relate to glycaemic control among employed people with type 2 diabetes, a growing demographic with increasing incidence and notable risks of complications and mortality. Work environments heavily shape daily routines, including eating, and factors such as long hours, shift work, and psychosocial stress can affect diabetes control. While global shifts toward eating out and irregular meals have been linked to cardiometabolic outcomes, little research has focused specifically on employed patients whose eating patterns are constrained by workplace conditions. The authors aim to identify key workplace eating patterns associated with glycaemic control (HbA1c) in employed type 2 diabetes patients in Hong Kong, where average working hours are long.

Prior research has associated specific eating behaviors (e.g., breakfast skipping, lower meal frequency, late-night dinner) with glycaemic control in diabetes, but has not focused on employed patients or workplace-constrained patterns. Eating out is increasing globally and is associated with higher energy intake, higher fat, lower micronutrients, increased diabetes risk, and insulin resistance, whereas frequent home-prepared meals are associated with lower risk of type 2 diabetes. Work-related factors such as hours worked, shift work, and psychosocial stress have been linked to diabetes outcomes, yet evidence is limited and variable in quality. This study builds on these findings to examine workplace-specific eating patterns and HbA1c among working patients.

Design: Exploratory sequential mixed-methods study with independent samples. Stage 1 qualitative focus groups identified relevant workplace eating patterns; Stage 2 developed and piloted a contextualized survey; Stage 3 cross-sectional quantitative study tested associations with HbA1c. Participants: Inclusion for both parts: age 18–65, type 2 diabetes ≥6 months, same employment ≥6 months. Qualitative (Apr–Jul 2019): Semi-structured focus groups recruited via diabetes complication screening at Prince of Wales Hospital (PWH), Hong Kong, using heterogeneous purposive sampling for diversity (occupation, regimen, age, gender). Sessions (50–75 min) were moderated by the first author using a discussion guide, audio-recorded, transcribed, anonymized; data saturation assessed with field notes. Analysis: Inductive thematic analysis at semantic level (Braun and Clarke), managed with NVivo 12. Quantitative (Jan 2020): Cross-sectional survey at Fanling Family Medicine Center (FLFMC), a large public clinic. Random sampling from the appointment list; consented participants completed a self-administered questionnaire on demographics, socioeconomic status, employment/working conditions, disease conditions/regimens, and workplace eating patterns. Latest HbA1c (%) within 3 months post-survey was retrieved from electronic medical records. Variables: Outcome: HbA1c (%). Key predictors: workplace eating patterns identified from focus groups—consumption of home-prepared meals (HPM) at work (regular/occasional/never) and meal hours (regular/irregular). Confounders: age, sex, education (primary/below; secondary; tertiary), personal monthly income (HK$0–9,999; HK$10,000–29,999; HK$30,000+), occupation (non-/medium-/highly-skilled), smoking (current or not), exercise outside home/work per week (yes/no), diabetes duration (years), comorbidities (yes/no). Sample size was calculated (G*Power 3.1.9.4) for multiple linear regression assuming small effect size, 95% CI, SD 0.5, power 80%, and 12 predictors. Ethics: Approved by The Joint CUHK-NTEC CREC; reporting followed STROBE and COREQ. Statistical analysis: Descriptive statistics profiled participants. Linear-by-linear association tests and ANOVA profiled patients with HPM patterns. Spearman correlations assessed HbA1c vs continuous covariates (age, disease duration, working hours). Group comparisons of HbA1c used Kruskal-Wallis H-test or Mann-Whitney U-test. Hierarchical multiple linear regression (enter method) examined predictors of HbA1c with sequential blocks: (1) socio-demographics, (2) lifestyle factors, (3) disease conditions, (4) workplace eating patterns. Assumptions (linearity, homoscedasticity, multicollinearity, normality of residuals, outliers) were checked. Sensitivity analyses excluded insulin-treated patients and excluded outliers (studentized deleted residuals > ±3 SD).

Qualitative: Two workplace eating patterns were most relevant to glycaemic control: (1) consumption of home-prepared meals (HPM) versus eating out at restaurants; (2) regular versus irregular meal hours. Barriers to HPM included job nature (e.g., lack of fixed work location, professional drivers), impractical dietary advice not tailored to working conditions, and long working hours. Quantitative sample: 185 employed T2D patients (73.5% male), mean age 56.7 (SD 6.0) years, mean diabetes duration 6.7 (SD 5.7) years. HPM frequency at work: 21.1% regular HPM (n=39), 11.4% occasional HPM (n=21), 67.6% restaurant meals only (n=125). Factors associated with HPM consumption: Female sex (p=0.03), non-skilled occupation (p=0.002), no diabetic medication (p=0.032), shift work (p=0.046), fixed work location (p<0.001), and having work breaks (p=0.004). HbA1c comparisons: Regular HPM associated with lower HbA1c compared to restaurant-only eating (Kruskal-Wallis H(2)=9.579, p=0.008). Diabetes duration correlated positively with HbA1c (Spearman r=0.299, p<0.001). Meal hour regularity was not significantly associated with HbA1c in group comparisons (p=0.418). Regression: Hierarchical multiple linear regression showed: Model 3 (adding diabetes duration) increased R² by 0.055 (F(2,172)=5.293, p=0.006); Model 4 (adding HPM consumption) increased R² by 0.044 (F(2,170)=4.339, p=0.015). Full model: R²=0.146, adjusted R²=0.076; F(14,170)=2.075, p=0.015. Regular HPM at work predicted lower HbA1c (B=-0.475, β=-0.221, p<0.05) independent of covariates; occasional HPM was not significant. Sensitivity analyses excluding insulin-treated patients (n=181) and excluding outliers (n=182) yielded similar results, supporting robustness.

The findings indicate that among employed patients with type 2 diabetes, regularly consuming home-prepared meals at work is associated with lower HbA1c independent of socio-demographic, lifestyle, and disease factors. Although certain working conditions (fixed location, breaks, shift work) were associated with the likelihood of consuming HPM, they were not directly associated with HbA1c in this sample. The qualitative component identified practical barriers to adopting HPM, including job-related constraints and dietary advice perceived as impractical for working contexts. These results suggest actionable strategies: providing practical, work-adapted dietary counseling (e.g., simple recipes, consideration of energy needs by job demands) and workplace accommodations (space and facilities for storing, heating, and eating HPM). In the COVID-19 context, HPM may concurrently reduce restaurant-based exposure risks while improving glycaemic control. The study advances understanding of workplace-specific eating behaviors in diabetes management, a previously underexplored area, and underscores the potential value of interventions supporting HPM for working patients.

Regular consumption of home-prepared meals during work hours is associated with better glycaemic control among employed individuals with type 2 diabetes. Interventions should promote HPM through practical, context-sensitive dietary advice and workplace accommodations (storage and food preparation facilities). For those unable to bring HPM due to job constraints, restaurant-based strategies (e.g., menu labeling, healthier options) should be considered. Future research should include larger-scale cohort studies to assess causality and explore the impact of specific nutritional content of HPM and tailored workplace interventions.

- No data on nutritional content, ingredients, or cooking methods of HPM; dietary quality could not be directly assessed. - Potential unmeasured confounding (e.g., genetic determinants) may affect HbA1c. - Self-reported predictors (eating patterns, working conditions) may contain measurement error. - Relatively small cross-sectional sample, especially the subgroup regularly consuming HPM, limiting generalizability. - Differences in diabetes regimens across HPM groups (more on medications among non-HPM) could influence HbA1c; regimen effectiveness implications were not assessed. - Observational, cross-sectional design precludes causal inference between HPM at work and HbA1c.