Joint Effects of Physical Activity and Diet Quality on Depression and Mortality in Diabetes

Depression will be measured using the Patient Health Questionnaire-9 (PHQ-9), a validated 9-item self-report instrument scored from 0 to 27. Participants scoring ≥5 will be classified as having depressive symptoms. The outcome is the percentage of participants who meet this criterion at baseline.

Physical activity will be assessed using NHANES physical activity questionnaires and expressed in metabolic equivalent minutes per week (MET-min/week). Participants will be categorized into three PA levels: inactive (<600 MET-min/week), insufficiently active (600-1,199 MET-min/week), and sufficiently active (≥1,200 MET-min/week). Depression will be measured using the Patient Health Questionnaire-9 (PHQ-9), a validated 9-item self-report instrument scored from 0 to 27, with scores ≥5 indicating the presence of depressive symptoms. Logistic regression models will be used to estimate odds ratios for depression across PA categories, using the inactive group as the reference.Odds Ratio (95% Confidence Interval)

Diet quality will be assessed using the Healthy Eating Index-2015 (HEI-2015) score, derived from two 24-hour dietary recall interviews. Scores range from 0 to 100, with higher scores indicating better diet quality. Participants will be categorized into two groups: unqualified diet quality (HEI-2015 <60) and qualified diet quality (HEI-2015 ≥60). Depression will be measured using the Patient Health Questionnaire-9 (PHQ-9), a validated 9-item self-report instrument scored from 0 to 27, with scores ≥5 indicating the presence of depressive symptoms. Logistic regression models will be used to estimate odds ratios for depression between diet quality groups, using the unqualified group as the reference.Odds Ratio (95% Confidence Interval)

Physical activity (PA) will be assessed using NHANES physical activity questionnaires and expressed in metabolic equivalent minutes per week (MET-min/week). Diet quality (DQ) will be assessed using the Healthy Eating Index-2015 (HEI-2015) score, derived from two 24-hour dietary recall interviews. Participants will be classified into four PA_DQ groups:

Inactive_Unqualified (PA <600 MET-min/week and HEI-2015 <60) Inactive_Qualified (PA <600 MET-min/week and HEI-2015 ≥60) Active_Unqualified (PA ≥600 MET-min/week and HEI-2015 <60) Active_Qualified (PA ≥600 MET-min/week and HEI-2015 ≥60) Depression will be measured using the Patient Health Questionnaire-9 (PHQ-9), a validated 9-item self-report instrument scored from 0 to 27, with scores ≥5 indicating the presence of depressive symptoms. Logistic regression models will be used to estimate odds ratios for depression across the four PA_DQ groups, using the Inactive_Unqualified group as the reference.Odds Ratio (95% Confidence Interval)

All-cause mortality will be determined from the NHANES Linked Mortality Files, which are linked to the National Death Index (NDI). Participants will be classified into four combined PA_DQ categories:

Inactive_Unqualified (PA <600 MET-min/week and HEI-2015 <60) Inactive_Qualified (PA <600 MET-min/week and HEI-2015 ≥60) Active_Unqualified (PA ≥600 MET-min/week and HEI-2015 <60) Active_Qualified (PA ≥600 MET-min/week and HEI-2015 ≥60) Subgroup analyses will examine whether hazard ratios for all-cause mortality associated with these PA_DQ categories differ across age (<65 vs ≥65 years), sex, BMI (<29 vs ≥29 kg/m²), and sedentary time (<4 vs ≥4 hours/day). Cox proportional hazards models with interaction terms will be used, and results will be reported as stratified hazard ratios relative to the Inactive_Unqualified group.

Shapley Additive exPlanations (SHAP) analysis will be used to identify and rank the most important predictors of depression risk, defined as Patient Health Questionnaire-9 (PHQ-9) score ≥5. Predictors assessed will include physical activity (PA), diet quality (DQ), body mass index (BMI), education, poverty-income ratio (PIR), age, sex, and sedentary behavior. Results will be reported as variable importance rankings with SHAP values.Relative importance score (SHAP value).

Causal mediation analysis will be used to estimate the indirect effect of body mass index (BMI, kg/m²) on the association between combined physical activity and diet quality (PA_DQ) and depression, defined as Patient Health Questionnaire-9 (PHQ-9) score ≥5. Participants will be classified into four PA_DQ categories:Inactive_Unqualified (PA <600 MET-min/week and HEI-2015 <60),Inactive_Qualified (PA <600 MET-min/week and HEI-2015 ≥60),Active_Unqualified (PA ≥600 MET-min/week and HEI-2015 <60)

,Active_Qualified (PA ≥600 MET-min/week and HEI-2015 ≥60).The outcome will be reported as the average causal mediation effect (ACME) and the proportion mediated (%), using the Inactive_Unqualified group as the reference category.Average Causal Mediation Effect (ACME) and proportion mediated (%)