Participants of the Japan Public Health Center-based prospective study (JPHC study) in 2014–2015, living in the Saku Public Health Center catchment area in Nagano prefecture, were invited for a mental health survey. The JPHC Study was launched at five public health centres (PHCs) in Japan for Cohort I in 1990 . A self-administered questionnaire on demographic information, lifestyle characteristics, and social factors was distributed to uninstitutionalised residents aged 40–59 years in 1990 and follow-ups of 5 years, ten years, and 15 years after the first survey (response rates: 74–81%) were conducted.
There were 12,219 participants (6172 men and 6047 women) in the baseline survey. After excluding 3392 participants as they moved out of the study area, died, or did not respond to the latter questionnaires during follow-up, we selected the remaining 8827 persons. We invited participants to participate in a mental health survey. A total of 1299 out of 8827 participants (14.7%) responded to the mental health screening. We further excluded 21 participants due to incomplete data for the questionnaires relating to family configuration and 24 participants with a history of depression in the mental health screening questionnaires. The remaining 1254 participants (529 men and 725 women) aged 64–84 were included in the analysis. A flow diagram of the study participants is shown in Fig. 1.
Assessment of living arrangement
The question about the individual’s family configuration in the baseline questionnaire was, “Are you living with someone (spouse, child (ren), parent(s), others, alone) together now?” According to the Japanese culture, “others” are regarded as other family members; siblings, grandparents, uncles, aunts, cousins, in-laws, etc. The same question was repeated for each follow-up survey. In the present study, we used a questionnaire of 1990, and only 15 persons (five men and ten women) living alone participated; hence, we could not analyse them as explanatory variables.
Regarding past histories, we followed up the participants from 1990 until the screening in 2014–2015 and registered participants’ conditions, including survival and medical status, within the catchment areas of Saku City. We assessed the incidence of cancer using medical records from each hospital in the study area. We interviewed past histories of depression, diabetes mellitus, stroke, and myocardial infarction by self-administrated questionnaire of this mental health survey. All other covariates, except for past histories (smoking status, alcohol frequency, sleeping duration, and occupation), were questioned through the baseline survey in 1990.
Confirmation of depression
Certified psychiatrists assessed all participants in this mental health screening. First, we administrated the Center for Epidemiological Scale-Depression (CES-D) [21, 22] and the Patient Health Questionnaire-9 (PHQ-9) [23, 24] screening tests at the same time. Second, well-trained board-certified psychiatrists interviewed the participants by referring to the CES-D and PHQ-9 scores. Finally, the psychiatrists assessed whether the participant was diagnosed with MDD based on the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition (DSM-IV) criteria [19, 20] and reached a consensus for final diagnosis when each psychiatrist’s diagnosis was different.
If a patient experiences depressive symptoms, they are not always diagnosed with depression . Mild cognitive impairment (MCI), dementia, and pseudodementia may be associated with similar symptoms [25,26,–27]. It is often difficult for general doctors to distinguish MDD from others, and patients sometimes have overlapping diseases [25, 28]. Well-trained board-certified psychiatrists met the 1299 participants, confirmed their self-report questionnaires, and assessed whether the participants currently met the DSM-IV criteria for MDD after considering whether their depressive symptoms caused clinically significant distress or impairment .
Logistic regression analyses were performed to calculate the odds ratios (ORs) and 95% confidence intervals (CIs) of MDD associated with the family configuration. We adjusted for age (years, continuous) and sex in the first model and adjusted for other lifestyle-related factors and health history in the second model. These factors included smoking status (never, former, current), alcohol frequency (seldom, 1–3 times per month, 1–2 times per week, 3–4 times per week, 5–6 times per week, every day), sleeping duration (≤4 h, 5–9 h, ≥10 h), occupation (professional, managerial, white-collar, and blue-collar jobs), and education (primary education, lower secondary education, upper secondary education, post-secondary education), history of cancer (yes or no), stroke (yes or no), myocardial infarction (yes or no), and diabetes mellitus (yes or no). The rate of missing values was lower than 0.2–1.4%. Missing data were assumed to be missing at random (MAR), and we used multiple imputations to handle missing data of confounding variables using the ‘mice’ package in R software. All variables in the dataset used in this study were included in the imputation model. Results across five imputed datasets were combined by averaging, and standard errors were adjusted to reflect both within-imputation variability and between-imputation variability in the pooling phase . The level of statistical significance was set α = 0.05 (two-tailed). All statistical analyses were conducted by the R software (version 3.5.3; https://www.r-project.org/).