Association of childhood family environments with the risk of social withdrawal (‘hikikomori’) in the community population in Japan

Abstract

Aims: Hikikomori is a form of social withdrawal among those who retreat from social interaction for protracted periods of time. This study examines family-related childhood factors for hikikomori using the retrospective data derived from a population-based survey.

Methods:  We derived data from World Mental Health Survey Japan. The subjects of this study were community residents aged 20–49 years (n = 708). Multiple logistic regression was applied to examine the association between the lifetime experience of hikikomori and childhood family environment, adjusting for sex, age, and respondents' history of common mental disorders.

Results:  Father's high educational level (odds ratio [OR] = 6.0, 95% confidence interval [CI] = 1.6–22.9), mother's common mental disorders (OR = 5.9, 95%CI = 1.1–33.3), and mother's panic disorders (OR = 6.6, 95%CI = 1.1–39.1) were significantly and positively associated with hikikomori after controlling for respondents' sex, age, and history of mental disorders.

Conclusions:  Our findings suggest that hikikomori cases are more likely to occur in families where the parents have high levels of education. Maternal panic disorder may be another risk factor for children to develop hikikomori.

HIKIKOMORI IN JAPANESE refers to a behavioral pattern of confining oneself to one's house for a prolonged period of time. Whereas social withdrawal is a general term for withdrawal from social interaction, hikikomori is operationally defined by the Ministry of Health, Labour and Welfare of Japan as a state of confining oneself to one's house for more than 6 months and strictly limiting communication with others.¹ Even though various psychopathologies may underlie hikikomori, 45.5% of hikikomori cases were found not to meet any of the diagnostic criteria for mood disorders, anxiety disorders, substance abuse/dependence, and intermittent explosive disorder at any point in their lifetime.² Without a history of the common mental disorders, these cases are assumed to have risk factors related to the family environment in which they were brought up.^3,4

It has often been noted that hikikomori is more prevalent in middle and upper class families,^5–7 but few studies have ever examined the social and economic status of the family in childhood of hikikomori cases.⁸ Another possible risk factor is parental psychopathology, which was found to increase risks for behavioral and mental health problems among offspring.^9,10 However, no empirical study has investigated the association of parental psychopathology with hikikomori. It was widely acknowledged that parenting styles, such as authoritarian, controlling, rejecting and overprotective attitude, influence the development and stabilization of hikikomori.^11–13 Child abuse was also regarded as a risk factor for hikikomori,¹⁴ while other studies reported that the families of hikikomori cases did not have apparent conflicts at home.¹⁵ Most of these studies were cross-sectional,^12,14 and the samples were restricted to those who had not experienced hikikomori or to those who contacted specialists for consultation.¹⁵

These conflicting results and lack of empirical findings on risk factors hinder the development of effective programs for hikikomori. Thus, this study examines the association of family environments with hikikomori, using self-reported retrospective data on the following areas of childhood family environment: social class, parental psychopathology, parents' childrearing style, and family-related adverse events. Lifetime experience of mental disorders was added in the analysis in order to control for possible effects of mental disorders on hikikomori.

METHODS

Survey areas

The data were derived from the World Mental Health Japan (WMHJ) survey, which was part of the cross-national World Mental Health (WMH) survey conducted worldwide. In Japan, the data were collected at 11 sites in six prefectures from 2002 to 2006.¹⁶ The sites included a metropolitan city, two urban cities, and eight rural municipalities. The survey sites were selected based on geographic variation, cooperation of local governments, and availability of site investigators. Although not randomly selected, the participating sites were considered to cover the diverse Japanese population in terms of municipality size and geographic area variations.

Survey instrument

Face-to-face interviews were conducted using the Japanese computer-assisted personal interviews (CAPI). The questionnaire was translated from the World Health Organization Composite International Diagnostic Interview (WMH-CIDI),¹⁷ a fully structured diagnostic interview that was designed to be conducted by trained lay interviewers.

WMH-CIDI was composed of two parts. Part 1 interview contained a core diagnostic assessment and basic sociodemographic data. The diagnostic assessments of common mental disorders were based on the definition and criteria of the DSM-IV.¹⁸ Part 2 interview included potential correlates and disorders of additional interest. Questions related to childhood factors were included in this second part. This instrument demonstrated acceptable reliability and validity.¹⁹

Study sample

Subjects for the WMHJ survey were randomly selected from voter registration lists or resident registries at each site. Those who (i) were dead or institutionalized, (ii) had moved from the survey site, or (iii) could not speak Japanese were excluded from the recruitment. All respondents who consented to participate completed Part 1 interview (n = 4134, response rate = 55.1%). Respondents who met the criteria for any mental disorders assessed in Part 1 interview and a probability of approximately 25% of the remainder were administered Part 2 interview (n = 1682), which contained variables on childhood factors. Subjects in this study were limited to respondents who were aged 20–49 years (n = 708). The age range of subjects was determined based on a survey result which demonstrated that many hikikomori cases started hikikomori before the age of 35 years.¹

The participation in this study was completely voluntary, and anonymity and confidentiality were assured. Written consent was obtained from each respondent. The human subject committees of the following institutions approved the recruitment, consent, and field procedures: Okayama University Graduate School of Medicine, Dentistry, and Pharmaceutical Sciences; Japan National Center of Neurology and Psychiatry; Nagasaki University Graduate School of Biomedical Sciences; Yamagata University Graduate School of Medical Science; and Juntendo University Graduate School of Medicine.

Measures

1. Assessment of hikikomori

Hikikomori was defined as a state of social withdrawal for more than 6 months, communicating only with family members and hardly leaving the house for work or school.^1,20 The same types of withdrawal situations that were caused by physical illness or injury were defined as non-hikikomori cases.

2. Social class in childhood

Social class in childhood included the educational status of parents and family economic adversity. Parents' completing more than 12 years of education was coded as high education, which was equal to graduation from a higher level of institute than high school. Family economic adversity was defined as: (i) a respondent's family ever received money from a governmental assistance program; (ii) either male or female head of the family was absent and the other head did not work for all or most of the respondent's childhood; or (iii) there was neither female nor male head of the family. A positive response to either item was coded as having family economic adversity.

3. Parental psychopathology

The following parental mental disorders were assessed using a modified version of the Family History Research Diagnostic Criteria Interview:²¹ major depression, general anxiety disorder, panic disorder, substance abuse and dependence, and antisocial personality disorder. The ‘any mental disorder’ category was defined if the respondents reported any of these parental mental disorders, separately for fathers and mothers or for any of them.

4. Parents' childrearing style

Parenting style was measured by items from the Parental Bonding Instrument (PBI).²² The present study adopted three PBI items: the affection item (‘How much love and affection did she/he give you?’), the overprotection item (‘How much did she/he stop you from doing the things that other kids your age were allowed to do?’), and the authoritarian item (‘How strict was she/he with her/his rules for you?’). Each item was scored ‘a lot’, ‘some’, ‘a little’, and ‘not at all’, and dichotomized as having or not having the childrearing style. The affection item was scored in reverse to calculate OR for not having the affectionate parenting style.

5. Other childhood adversity

Childhood adversity included the experience of abuse, parental loss, and family violence while respondents were growing up. Three types of child abuse were assessed; physical abuse, sexual abuse, and neglect. Parental loss involves a parent's death, divorce, or any other type of parental loss. Family violence refers to witnessing violence at home. The assessment and classification of child adversity have been described elsewhere in detail.²³

6. Sociodemographic factors

Demographic variables included age, sex, and education. Age at the time of the survey was categorized into three groups: 20–29, 30–39, and 40–49 years old. Educational level was also categorized into three groups: junior high school, senior high school, and college or university graduates.

7. Respondents' mental disorders

The respondents' histories of common mental disorders were also evaluated based on the DSM-IV adopted in WMH-CIDI.¹⁷ The disorders assessed were: mood disorders (bipolar I and II disorders, dysthymia and major depressive disorder), anxiety disorders (agoraphobia, generalized anxiety disorder, panic disorder, post-traumatic stress disorder, social phobia and specific phobia), and substance abuse and dependence (alcohol and drug abuse and dependence).

Statistical analysis

The sample was standardized to match the age and sex structure of the survey site population in order to rectify the effect of under-coverage, non-response, and sampling variability on the results. The sample was also weighted by the inverse of a participant's selection probability for Part 2 interview (Part 2 weight) in order to adjust for differential probability of selection for Part 2 interview.¹⁶

First, the crude prevalence of each childhood family environment was calculated for both hikikomori and control groups. Second, a logistic regression model was used to examine the bivariate association of each childhood familial factor with hikikomori. The analysis could not be conducted when no hikikomori cases were found in a group classified by childhood factors, because the estimation unadjusted for possible confounders was not precisely convergent. In the first model, a childhood factor and two demographic variables (age and sex) were entered in order to control for positive effects of demographics. History of mental disorder was added in the second model in order to control the possible effect of a respondent's mental disorder. The analysis was conducted using the sas statistical package (sas Institute Inc, Cary, NC, USA).

RESULTS

Sample characteristics

Among the 708 respondents who were eligible for this study, 15 experienced hikikomori during their lifetime (n = 9.7 or 1.2% in a weighted analysis). The characteristics of the total sample and hikikomori cases are shown in Table 1. Hikikomori cases were more likely to be male, but the difference was not significant (P > 0.05). Age at survey and educational level of individuals indicating a previous hikikomori case in this study were not evidently different from those of the control group (P > 0.05). Over 80% of the first onset were in their teens and twenties, and almost half of the hikikomori durations were less than 1 year.

Table 1.  Demographic characteristics of the total sample under 50 years old and of hikikomori cases: The World Mental Health Japan Survey 2002–2006

Variables	Total sample (n = 708)			Hikikomori cases (n = 9.7)^†
	Crude	Weighted^†	Weighted^†	Weighted^†	Weighted^†	P-value**
	n	n	%	n	%
** Significance for the difference between respondents with and without lifetime experience of hikikomori (χ^2-test with Part 2 weight). ^† Weighted for the Part 2 weight (see text for details).
Sex (male)	290	383.3	49.2	6.8	70.0	0.167
Age at survey (years)	Mean = 30.1, SD = 8.6			Mean = 36.3, SD = 11.1
20–29	188	244.7	31.4	5.0	51.7
30–39	232	269.4	34.6	1.3	13.5
40–49	288	264.5	34.0	3.4	34.8	0.229
Education
Junior High	33	32.2	4.1	0.4	3.8
Senior High	488	513.4	65.9	6.9	71.4
University	187	233.0	29.9	2.4	24.8	0.895
Age at first onset of hikikomori (years)				Mean = 25.2, SD = 11.2
10–19				5.0	51.3
20–29				3.1	31.9
30–39				–	–
40–49				1.4	14.1
Unknown				0.3	2.7
Duration of hikikomori (months)				Mean = 7.8, SD = 4.7
6–11				4.7	48.8
12 or longer				4.1	42.4
Unknown				0.9	8.8

The distribution of family-related childhood factors is presented in the Appendix. For the latter statistical analysis, neglect, physical abuse, and sexual abuse were integrated into a ‘child abuse’ category, and maternal and paternal death were combined into a single ‘parental death’ category due to the small number of instances.

Effect of childhood family environment on lifetime experience of hikikomori

The associations of childhood factors with hikikomori are shown in Table 2. After adjusting for age and sex (Model 1), high education of fathers and mothers were significantly and positively associated with hikikomori (odds ratio [OR] = 4.9, 95% confidence interval [CI] = 1.3–18.2; OR = 6.7, 95%CI = 1.3–35.4, respectively). No significant association was found in parental psychopathology with the exception of maternal mental disorders (OR = 9.1, 95%CI = 1.5–56.8) and maternal panic disorder (PD) (OR = 9.9, 95%CI = 1.5–63.8).

Table 2.  Association of retrospectively assessed childhood social class, parental psychopathology, and childhood adversities with lifetime prevalence of hikikomori: The World Mental Health Japan Survey 2002–2006

Childhood risk factors			Total sample weighted n	Hikikomori cases weighted n (%^†)	Model 1		Model 2
					Adjusted for sex & age Odds (95%CI)	P-value	Adjusted for sex, age, & mental disorders of respondents Odds (95%CI)	P-value
*P < 0.05, **P < 0.01. ^† Row % (Proportion of hikikomori cases to the total sample with/without each childhood factor).
I. Social class in childhood
Father	High education	Yes	85.3	3.9 (4.5)	4.90 (1.32–18.23)*	0.018	6.00 (1.57–22.94)**	0.009
		No	693.3	5.8 (0.8)	1		1
Mother	High education	Yes	48.1	2.9 (6.1)	6.65 (1.25–35.41)*	0.026	6.20 (0.98–39.35)	0.053
		No	730.4	6.8 (0.9)	1		1
Family	Economic adversity	Yes	35.7	0.3 (0.8)	0.67 (0.07–6.27)	0.723	0.46 (0.05–4.63)	0.510
		No	742.9	9.4 (1.3)	1		1
II. Parental psychopathology
Parents	Any mental disorder	Yes	20.2	1.2 (5.7)	5.76 (0.75–44.38)	0.093	3.38 (0.44–25.87)	0.242
		No	758.3	8.6 (1.1)	1		1
Mother	Any mental disorder	Yes	16.7	1.2 (6.9)	9.13 (1.47–56.84)*	0.018	5.94 (1.06–33.32)*	0.043
		No	762.7	8.6 (1.1)	1		1
Mother	Panic disorder	Yes	15.1	1.2 (7.6)	9.87 (1.52–63.84)*	0.016	6.64 (1.13–39.08)*	0.036
		No	763.4	8.6 (1.1)	1		1
III. Parental childrearing style
Mother	Affectionless	Yes	88.2	1.3 (1.4)	1.04 (0.24–4.50)	0.960	0.87 (0.20–3.73)	0.854
		No	690.3	8.5 (1.2)	1		1
	Overprotective	Yes	180.7	1.4 (0.8)	0.55 (0.12–2.56)	0.446	0.56 (0.12–2.50)	0.444
		No	597.9	8.3 (1.4)	1		1
	Authoritarian	Yes	457.4	4.6 (1.0)	0.66 (0.18–2.42)	0.533	0.65 (0.17–2.47)	0.524
		No	321.1	5.1 (1.6)	1		1
Father	Affectionless	Yes	138.0	2.8 (2.0)	1.96 (0.56–6.88)	0.293	1.63 (0.45–5.90)	0.457
		No	640.6	7.0 (1.1)	1		1
	Overprotective	Yes	180.8	1.0 (0.6)	0.38 (0.08–1.92)	0.244	0.44 (0.09–2.15)	0.307
		No	597.7	8.7 (1.4)	1		1
	Authoritarian	Yes	411.5	5.1 (1.2)	1.00 (0.31–3.24)	0.998	1.14 (0.35–3.69)	0.833
		No	367.0	4.6 (1.3)	1		1
IV. Childhood adversity
Any child abuse		Yes	71.4	0.3 (0.5)	0.35 (0.05–2.57)	0.300	0.24 (0.03–1.92)	0.177
		No	707.2	9.4 (1.3)	1		1
Parental death		Yes	51.9	1.2 (2.4)	2.66 (0.54–13.07)	0.227	1.93 (0.28–13.52)	0.508
		No	726.7	8.5 (1.2)	1		1
Other parental loss		Yes	11.3	0.3 (2.3)	1.28 (0.11–15.19)	0.847	1.56 (0.15–16.09)	0.709
		No	767.2	9.4 (1.2)	1		1
Family violence		Yes	99.7	0.3 (0.3)	0.20 (0.02–1.90)	0.161	0.17 (0.02–1.61)	0.121
		No	678.9	9.4 (1.4)	1		1

When the mental disorders of respondents were added to the model (Model 2), the educational level of fathers increased the OR (OR = 6.0, 95%CI = 1.6–22.9), while mothers' mental disorders and PD decreased the OR (OR = 5.9, 95%CI = 1.1–33.3; OR = 6.6, 95%CI = 1.1–39.1, respectively). Mother's educational level did not remain significant, even though the 95%CI was still marginally high (OR = 6.2, 95%CI = 1.0–39.4).

No significant association was found for parents' childrearing styles or other childhood adversities. There were no hikikomori cases in the groups reporting, parental divorce, maternal depression, paternal PD, and paternal or maternal generalized anxiety disorder, substance abuse/dependence or anti-personality disorder. The analysis was thus not conducted for these variables.

DISCUSSION

In the current cross-sectional study, hikikomori in Japan was more prevalent among families with little socioeconomic disadvantages, and maternal PD was associated with a higher risk of hikikomori. These findings expand our insight on the unique mechanisms of the development of hikikomori.

Our results showed that the higher educational level of parents was significantly associated with a higher risk of hikikomori; the association of fathers' educational level remained significant even after adjusting for respondents' history of mental disorders. A possible explanation for this association is that highly educated parents have better income and can afford to have their children jobless at home.⁵ Also, unreasonably high expectations of their children from highly educated parents may enhance the children's dependent and over-optimistic attitudes, which could result in a coping pattern of avoiding interpersonal conflicts and failures in social interaction.^4,13 However, neither childhood poverty nor parenting style had a significant association with hikikomori in the current study. Further study is needed on these explanations, preferably using a longitudinal design.

Among parental history of mental disorders, maternal PD was significantly associated with hikikomori. This is concordant with previous studies reporting an association between maternal PD and child's internalizing problems, especially anxiety disorders.^24–27 This association is possibly due to parenting behaviors which are specific to mothers with PD, who tend to reinforce children's anxiety and their avoidant coping strategy by failing to decrease children's attentional bias to threats and interfering with their exploration and social activity.^28–32 Another possible explanation is a common genetic factor underlying maternal PD and child internalizing problems.^33–35 However, further research is needed to warrant the association, because the assessment of parental PD in this study may not be consistent with diagnosis by a psychiatrist. In addition, we did not find a significant association between paternal PD and child hikikomori.

None of the parenting styles were significantly associated with hikikomori. The findings are in agreement with previous research reporting a small effect of general parenting style on children's behavioral and mental health problems.^10,36,37 However, the lack of a significant association may be partly attributed to the small number of hikikomori cases in this study. In addition, items that we used for measuring parenting styles have not been fully validated, and thus may have failed to capture entire parenting styles adequately.

Our study should be interpreted in light of the following limitations. First, none of our respondents was in a hikikomori state at the time of the survey, and the average hikikomori duration was shorter than that of another community-based survey.³⁸ The traits of our sample might have resulted in underestimating the associations between variables, and might have limited the generalization of the findings. Second, the use of retrospective reports on childhood variables might have caused an information bias. Furthermore, recall errors or conscious unwillingness to disclose painful memories may lead to overlooking the association between child adversities and hikikomori. Third, many of the childhood variables assessed in this study were evaluated by a single item, which may not accurately measure these factors. The interview measure of hikikomori was not fully validated for epidemiological studies either, even though it is widely used in other studies. Last, we statistically adjusted for respondents' history of common mental disorders. To clarify the risk factors with and without mental disorders, a stratified analysis is needed with a larger sample. Future study is needed to examine other possible risk factors, such as adverse experience at school, child temperament and developmental disorder. Our findings also suggest that parents' response to children's reluctance and anxiety to social interaction need more attention in future prevention/intervention strategies.

ACKNOWLEDGMENTS

The WMH-J 2002–2006 Survey Group members other than those listed in the author byline are as follows: Yutaka Ono, MD (Health Center, Keio University), Yoshibumi Nakane, MD (Dejima Shinryosho [Dejima Mental Clinic], Nagasaki, Japan [Prof. Emeritus of Nagasaki University]), Yoshikazu Nakamura, MD, MPH, FFPH (Department of Public Health, Jichi Medical School), Akira Fukao, MD (Department of Public Health, Yamagata University, Graduate School of Medical Science), Itsuko Horiguchi, PhD (Department of Public Health, Juntendo University Graduate School of Medicine), Hisateru Tachimori, PhD (National Institute of Mental Health, National Center of Neurology and Psychiatry), Noboru Iwata, PhD (Department of Clinical Psychology, Hiroshima International University), Hidenori Uda, MD (Health, Social Welfare, and Environmental Department, Aira-Isa Regional Promotion Bureau, Kagoshima Prefecture), Hideyuki Nakane, MD (Department of Psychiatric Rehabilitation Science, Unit of Rehabilitation Sciences, Nagasaki University Graduate School of Biomedical Sciences), Makoto Watanabe, MD, PhD (Department of Preventive Cardiology, National Cerebral and Cardiovascular Center), Masatsugu Oorui, MD (Yamagata Prefectural Tsuruoka Hospital), Kazushi Funayama, MD, PhD (Yokohama City Turumi Public Health and Welfare Center), Yoichi Naganuma, PSW, MSc (National Institute of Mental Health, National Center of Neurology and Psychiatry), Toshiaki A. Furukawa, MD (Department of Health Promotion and Human Behavior, Kyoto University Graduate School of Medicine), Masayo Kobayashi, MD (Department of Public Health, Jichi Medical School), Tadayuki Ahiko, MD (Yamagata Prefectural Institute of Public Health, Yamagata Prefecture), Yuko Yamamoto, PhD candidate (Department of Public Health, Juntendo University Graduate School of Medicine), Tadashi Takeshima, MD (National Institute of Mental Health, National Center of Neurology and Psychiatry), Takehiko Kikkawa, MD (Seisen Jogakuin College).

The study was supported by the Grant for Research on Psychiatric and Neurological Diseases and Mental Health from the Japan Ministry of Health, Labour, and Welfare (H13-SHOGAI-023, H14-TOKUBETSU-026, H16-KOKORO-013, H19-KOKORO-IPPAN-011). We would like to thank staff members, field coordinators, and interviewers of the WMH Japan 2002–2006 Survey. The WMH Japan 2002–2006 Survey was carried out in conjunction with the World Health Organization World Mental Health (WMH) Survey Initiative. We also thank the WMH staff for assistance with instrumentation, fieldwork, and data analysis. These activities were supported by the US National Institute of Mental Health (R01MH070884), the John D. and Catherine T. MacArthur Foundation, the Pfizer Foundation, the US Public Health Service (R13-MH066849, R01-MH069864, and R01 DA016558), the Fogarty International Center (FIRCA R01-TW006481), the Pan American Health Organization, Eli Lilly and Company, Ortho-McNeil Pharmaceutical, Inc., GlaxoSmithKline, and Bristol-Myers Squibb. A complete list of WMH publications can be found at http://www.hcp.med.harvard.edu/wmh/publications.php.

Appendix

Distribution of childhood social class, parental psychopathology, and parent childrearing style in the total sample under 50 years old: The World Mental Health Japan Survey 2002–2006

Childhood risk factors		Crude (n = 708)		Weighted (n = 778.6)^†
		n	%	Weighted^†n	Weighted^† %
I. Social class in childhood
Father	High education (higher than high school graduate)	59	8.3	85.3	11.0
Mother	High education (higher than high school graduate)	43	6.1	48.1	6.2
	Economic adversity (yes)	34	4.8	35.7	4.6
II. Parental psychopathology (present)^†
Parents	Any mental disorder	29	4.1	20.2	2.6
Father	Any mental disorder	30	4.2	18.4	2.4
	Depression	3	0.4	0.9	0.1
	General anxiety disorder	5	0.7	1.8	0.2
	Panic disorder	7	1.0	5.0	0.6
	Substance abuse/dependency	15	2.1	9.6	1.2
	Antisocial personality disorder	11	1.6	5.7	0.7
Mother	Any mental disorder	25	3.5	16.7	2.1
	Depression	5	0.7	1.8	0.2
	General anxiety disorder	4	0.6	3.3	0.4
	Panic disorder	20	2.8	15.1	1.9
	Substance abuse/dependency	1	0.1	0.7	0.1
	Antisocial personality disorder	3	0.4	1.3	0.2
III. Parents' childrearing style (higher tendency)^‡,§
Father	Affectionless	145	20.5	138.0	17.7
	Overprotective	163	23.0	180.8	23.2
	Authoritarian	368	52.0	411.5	52.9
Mother	Affectionless	87	12.3	88.2	11.3
	Overprotective	155	21.9	180.7	23.2
	Authoritarian	430	60.7	457.4	58.8
IV. Childhood adversity (yes)
Child abuse	Any abuse	83	11.7	71.4	9.2
	Physical abuse	69	9.7	64.4	8.3
	Neglect	11	1.6	5.0	0.6
	Sexual abuse	9	1.3	5.4	0.7
Loss	Parent died	36	5.1	51.9	6.7
	Father died	24	3.4	35.9	4.6
	Mother died	12	1.7	16.0	2.1
	Parental divorce	13	1.8	8.5	1.1
	Other parental loss	14	2.0	11.3	1.5
Other	Family violence	111	15.7	99.7	12.8