Data Source
A cross-sectional study was conducted using data from the Administrative claims database and specific check-up data obtained from the Japan Health Insurance Association, Osaka branch. The population in the Japan Health Insurance Association covers almost 40 million people and is the largest medical insurer in Japan. A flow diagram of the study participants is shown in Fig. 1. Of the 1,621,252 individuals from the database of the Japan Health Insurance Association, Osaka branch, our final analytic sample consisted of 226,363 participants (from April 2016 to March 2017). Among them, 29.5% were smokers. According to propensity score matching, 62,573 participants were selected for each group.
Questionnaire and clinical parameters in specific medical check-up
The smoking status was determined during the medical consultation. Information of medication (hypotensive, insulin or hypoglycemic, hyperlipidemia drugs) and medical history (stroke, heart disease, chronic renal failure or artificial dialysis and anemia) were obtained from personal questionnaire. Further, we included some other items such as the following questions “Have you gained over 10 kg from your weight at age 20?” (Yes/No) “Are you in the habit of exercising to sweat lightly for over 30 minutes a time, 2 times weekly, for over a year?” (Yes/No) “In your daily life, do you walk or do an equivalent amount of physical activity more than one hour a day?” (Yes/No) “Is your walking speed faster than the speed of those of your age and sex?” (Yes/No) “Have you gained or lost over 3 kg from your weight in a year?” (Yes/No) “Is your eating speed faster than others?” (Fast/ Normal/ Slow) “Do you skip breakfast more than 3 times a week?” (Yes/No) “Do you have any snacks or sweet beverages after dinner more than 3 times a week?” (Yes/No) “Do you have an evening meal within 2 hours before bedtime more than 3 times a week?” (Yes/No) “How often do you drink alcohol?” (Everyday/ Sometimes/ Rarely) “Do you feel refreshed after a night’s sleep?” (Yes/No) Furthermore, the check-up variables included waist circumstance (WC), body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), total cholesterol, triglycerides, high-density lipoprotein cholesterol (HDL), low-density lipoprotein cholesterol (LDL), fasting blood glucose (FBG), HbA1c, glutamic-oxaloacetic transaminase (GOT), glutamic pyruvic transaminase (GPT), ɤ-glutamyl transpeptidase (ɤ-GTP), uric acid (UA), and estimated glomerular filtration rate (eGFR). Status of smoking (smokers versus non-smokers) and metabolic syndrome (Applicable, pre-applicable, and not applicable) were obtained from the doctor’s perspective.
Dental and medical utilization and cost.
Each data was linked with the claimed database file by participants’ id numbers. We also calculated dental and medical care utilizations with the total annual cost of each procedure from April 2016 to March 2017. The annual number of medical visits was calculated for each participant. We counted the number of all diagnosed diseases and procedure codes of each group and calculated the differences between smokers and non-smokers.
Propensity score matching and statistical analysis
We calculated propensity score, defined as the conditional probability of each participant having a smoking habit given several confounders, such as age, sex, exercise, physical activity, walking speed, eating habit (speed, breakfast, snacks and evening meal), alcohol intake, and sleep using logistic regression. Standardized differences were calculated to assess the balance of covariates between smokers and non-smokers. If the standardized differences were less than 10%, the covariates were considered balanced. The propensity score of the smoking group and the non-smoking group were compared to create matched pairs (smokers and non-smokers as reference), 1:1 PSM method. In this propensity score model, goodness of fit was secured (Hosmer-Lemeshow test; P < 0.001 and C-index was 0.699). For continuous measures, we calculated Wilcoxon-Mann-Whitney test. The 2-sided significant level was set at P < 0.05. All statistical analysis were performed using SPSS version 23 (IBM Corp.Armonk,NY, USA).