Design, setting, and participants
We used the data collected from the National Usual Source of Care Survey (NUCS) conducted in May 2021. The NUCS was a nationwide mail survey that aimed to collect data on the usual source of primary care, health care use, health conditions, health-related quality of life, and sociodemographic characteristics in a representative sample of the Japanese adult population. In the NUCS, a national representative panel in Japan administered by the Nippon Research Center was used to select possible participants. This panel is composed of approximately 70,000 residents who were selected from the Japanese general population using a multistage sampling method and who participated in a previous survey from the Nippon Research Center.16 From the panel, 2,000 possible participants aged 20–75 years were selected through stratified sampling by age, sex, and residential area. The survey participants received 500 JPY gift certificates. Among residents responding to the NUCS, eligible participants in this study were individuals who responded to the survey item regarding health care use. The institutional review board of the Jikei University School of Medicine approved this study [approval no. 32-416(10505)].
Measures
Health-seeking behaviors
We collected data on participants’ health-seeking behaviors using a structured questionnaire. Participants were asked to answer the following questions about health care use for symptoms or health-related events in the last month: OTC drug use, physician’s office visit, hospital outpatient clinic visit, university medical center visit, emergency room visit, home health care use, CAM use, and hospitalization. We excluded telemedicine visits because telemedicine for new symptoms or health-related events was not widespread in Japan as of May 2021.17
Sociodemographic and clinical factors
We collected data on the sociodemographic and clinical factors of the participants. The questionnaire measured the age, sex, years of education, annual household income, social isolation, and the number of chronic conditions.
We used the Japanese version of the abbreviated Lubben Social Network Scale (LSNS-6)18 to assess social isolation. The LSNS-6 score is an equally weighted sum of six items, and the scores range from 0 to 30 points, with higher scores indicating a better quality of the social network. The reliability and validity of the Japanese version of LSNS-6 have been assessed in a previous study in Japan.18 As suggested in the previous study, we classified patients with a score of <12 points as being socially isolated.19
We used a validated list of 20 chronic conditions that were created based on previous multimorbidity literature and relevance to the primary care population20: hypertension, depression/anxiety, chronic musculoskeletal conditions that cause pain or limitation, arthritis/rheumatoid arthritis, osteoporosis, chronic respiratory disease (asthma, chronic obstructive pulmonary disease, or chronic bronchitis), cardiovascular disease, heart failure, stroke/transient ischemic attack, stomach problems, colon problems, chronic hepatitis, diabetes, thyroid disorder, any cancer in the past 5 years, kidney disease/failure, chronic urinary problem, dementia/Alzheimer’s disease, hyperlipidemia, and obesity.
Statistical analysis
Descriptive analyses were performed using the ecology of medical care model.7 We estimated the number of persons per 1,000 residents who had experienced different health care use during a one-month period and calculated 95% confidence intervals (CI) for event rates.
Subgroup analyses for variables of interest in health care use were conducted by age, sex, years of education, annual household income, social isolation, and the number of chronic conditions. In addition, to investigate the associations of sociodemographic and clinical factors with each health care use, we performed multivariable logistic regression analyses. In the multivariable analyses, we defined a hospital visit as a composite outcome that included hospital outpatient clinic visits, university medical center visits, and emergency room visits.
For each analysis, we used a two-sided significance level of P = 0.05. For missing independent variables in the regression model, we performed a complete case analysis. Statistical analyses were performed using R version 4.1.0 (R Foundation for Statistical Computing, Vienna, Austria; www.R-project.org).