This was a retrospective cohort study that used a Japanese healthcare database managed by MinaCare Co., Ltd. (Tokyo, Japan).
The MinaCare database is an employment-based health-insurance database that includes periodically updated health-checkup data (e.g., individuals’ demographics, smoking status, and vital signs) and medical/pharmaceutical claims data of employees and their dependent family members since 2010. The MinaCare database covers a wide range of age groups up to 74 years who work in large-scale nationwide retailing industries, manufacturers, and food, information, transportation, and energy industries. The database contains data of approximately 6.1 million individuals with medical and pharmaceutical claims and approximately 2.3 million individuals with health checkups (as of February 2020). Subjects aged ≥ 75 years are not included in this database because they are to be included in a different insurance program other than employment-based health-insurance program.
During the health checkup, a recipient undergoes physical examination, such as BP, body-weight measurement, blood test for lipid metabolism and others, and physicians’ examination. Additionally, at the time of the checkup, a questionnaire on one’s lifestyle, such as smoking, and current and past medical treatment is completed. Data regarding follow-ups after checkups (e.g., access to health guidance, contents of health guidance, and other types of follow-ups provided) were not available in this database.
In the present analysis, we used data on BP and demographic and clinical characteristics, such as sex, age, BMI, and self-reported smoking status, in addition to prescription data on antihypertensive drugs.
This study involved data that exist in an anonymized structured format and contained no personal information. Therefore, obtaining informed consent from subjects and approval from the ethical review committee were not required because studies using only unlinkable anonymized data are outside the scope of the “Ethical Guidelines for Medical and Health Research Involving Human Subjects” set by the Japanese government. The study was conducted in accordance with legal and regulatory requirements (e.g., privacy protection laws) and scientific purpose, value, and rigor. MinaCare manages such anonymized data under the data transfer contract with the client health insurers.
Subjects were included if they (1) were aged between 20 and 74 years at the time of the health checkup in 2015, (2) had systolic (SBP) and diastolic BP (DBP) data for 3 consecutive years from 2015 to 2017, and (3) had uncontrolled BP and were not prescribed antihypertensive drugs within 6 months before the health checkup in 2015.
Data Extraction and Management
Health checkup data between 2015 and 2017 (the latest 3 consecutive years available at the time of planning) and prescription claims data between 2014 and 2017 were extracted from the MinaCare database based on Japan’s fiscal year of April through March. Health checkups are usually conducted annually; in case of multiple records in any given year, the first record was used to render three records for each subject corresponding to 2015, 2016, and 2017.
Outcome Measures and Definitions
The primary outcome measures were changes in antihypertensive drug prescription and BP control status based on health-checkup results from 2015 to 2017.
BP control status and HT classification were defined according to the Japanese Society of Hypertension Guidelines for the Management of Hypertension in 2014 that were used in the clinical settings at the time of baseline health checkups in 2015 . “Controlled BP” was defined as SBP < 140 mmHg and DBP < 90 mmHg, and “Uncontrolled BP” was defined as SBP ≥ 140 mmHg and/or DBP ≥ 90 mmHg (Supplementary Table 1). HT was classified as HT grade I (SBP 140–159 and/or DBP 90–99 mmHg), HT grade II (SBP 160–179 and/or DBP 100–109), or HT grade III (SBP ≥ 180 and/or DBP ≥ 110) (Supplementary Table 1).
CV risk-management behavior was evaluated by antihypertensive drug prescription, BMI, and smoking habit. BMI and smoking, considered as secondary outcome measures, were assessed as behaviors for CV risk management because these are major risk factors of CV, and can be potentially changed by the subjects themselves. Antihypertensive drug prescription was defined by prescription records of antihypertensive medications within 6 months before the health checkup. Regarding antihypertensive drugs, all drugs indicated for hypertension and available at the time of this study were included in this study (Supplementary Table 2). There was 99.6% agreement between antihypertensive drug prescription within 6 and 12 months before the health checkup. Thus, antihypertensive drug prescription within 6 months before the health checkup was consistently used in the analyses. For self-reported smoking status, those who reported “yes” were considered current smokers, while those who reported “no” were considered past smokers and nonsmokers.
Metabolic syndrome was defined as waist circumference ≥ 85 cm (males) or ≥ 90 cm (females) and the presence of at least two of the following risk factors: (1) use of antihypertensive medication and/or SBP ≥ 130 mmHg and/or DBP ≥ 85 mmHg; (2) use of antihyperlipidemic medication and/or hypertriglyceridemia (triglyceride level ≥ 150 mg/dL) and/or hypo-HDL-cholesterolemia (HDL-c level < 40 mg/dL); and (3) use of DM medication and/or fasting blood glucose level ≥ 110 mg/dL .
Demographic and clinical characteristics and primary outcome measures were descriptively summarized. Antihypertensive drug prescription (yes/no) in 2016 was summarized, considering the effect of sex, age, and HT stage at baseline (in 2015). Similarly, BP control status (controlled/uncontrolled) in 2016 was summarized, considering the effect of sex, age, baseline HT stage, and antihypertensive drug prescription in 2016.
BMI status in 2016 of the subjects with BMI ≥ 25 kg/m2 at baseline and self-reported smoking status in 2016 of subjects aged ≥ 40 years who reported to smoke at baseline were descriptively summarized. Additionally, to explore the association between changes in BMI and BP control status, the mean changes in BMI (± standard error) were plotted against baseline BMI for each of the three categories of BP changes from 2015 to 2016 (worsened, unchanged, and improved) in subjects aged 40–59 years. The analyses of BMI excluded missing values, which were rare (0.2%). The analysis of BMI and BP control status was performed for subjects aged 40–59 years because (1) there were relatively few subjects aged < 40 and ≥ 60 years, leading to a large variation of mean changes in BMI in these age groups, and (2) individuals aged ≥ 40 years were targeted for a new Japanese national program for lifestyle disease prevention, Specific Health Checkups and Specific Health Guidance, due to the increasing number of individuals strongly suspected of having metabolic syndrome or pre-metabolic syndrome [17, 18]. For smoking, the results were restricted to age ≥ 40 years because a large number of responses were missing for ages < 40 years.
Logistic regression was used to explore factors associated with antihypertensive drug prescription in 2016 and BP control status in 2016. The following variables at baseline were included as explanatory variables: sex, age (categorized by decade), HT stage (I/II/III), obesity status (BMI < 25 or ≥ 25 kg/m2), smoking status (yes/no/missing or not reported), antihyperlipidemic drug prescription within 6 months (yes/no), DM drug prescription within 6 months (yes/no), and metabolic syndrome (yes/no). The cutoff point for obesity (BMI ≥ 25 kg/m2) was based on the Japanese Society of Hypertension Guidelines for the Management of Hypertension in 2014 . For each factor, the odds ratio (OR) and its 95% confidence interval (CI) were calculated. For BP control status in 2016, the analysis was conducted separately for those prescribed and not prescribed antihypertensive drugs in 2016.
No formal hypothesis test was used due to the exploratory nature of the study. Moreover, 95% CI for OR that excluded 1 was sometimes referred to as “statistically significant.”
SAS version 9.4 and R 3.5.0 were used in the statistical analysis.