Identifying Social Withdrawal (Hikikomori) Factors in Adolescents: Understanding the Hikikomori Spectrum

Our study targeted psychiatric outpatients, aged 12–15 years (seventh to ninth graders), who visited an adolescent outpatient clinic between December 2014 and November 2015. These participants were being primarily treated for hikikomori (n = 20; 10 of each sex; mean age ± standard deviation (SD) = 14.1 ± 1.1). We also recruited a healthy control group (n = 88; 56 boys; mean age ± SD = 14.0 ± 0.9). Among the clinical patients visiting the hospital chiefly for hikikomori, we targeted those who met the Cabinet Office’s definition; at least 6 months of a person exhibiting either “quasi-hikikomori” (i.e., going out only to engage in hobbies) or hikikomori in the narrow sense (i.e., from almost never going out of one’s room to going out to nearby convenience stores) [16]. Additionally, this definition excludes those diagnosed with schizophrenia and/or physical illness. We only included participants who, according to the DSM-IV-TR, did not meet the criteria for either Axis I or Axis II mental disorders.

For the healthy control group, we primarily recruited the siblings of student volunteers at Kyoto Women’s University, using a snowball sampling method (where respondents recommend additional eligible participants) and matched them with members of the patient group according to gender and age. No significant statistical differences were observed between the groups based on these characteristics (Supplemental Table 1). The exclusion criteria for healthy volunteers were the same as above, namely no diagnosis of schizophrenia, physical illness or Axis I or II mental disorders. All participants in the patient and healthy control groups received information regarding the survey and all parents and children provided consent to participate. Parents signed written consent forms for the participation of minors. Our study was carried out following a review, and permission was granted by the Clinical Study Ethical Review Board of Kyoto Women’s University.

Patients for the hikikomori group were pre-selected (on the basis of the treatment for the condition at an outpatient clinic). Since no DSM criteria currently exist and no formal diagnosis could be issued, the severity of each case needed to be established. We created a novel scale for evaluation (included in the “Appendix 1”), which was administered to the parents of both groups.

Based on the target age (school age) and the definition of hikikomori proposed in the report released by the Cabinet Office [15, 16, 38, 39], we designed our evaluation scale to comprise two items: (a) absenteeism from school and (b) going out; the latter is defined as: “the child went out either alone or with friends (but unaccompanied by family members) to shop, engage in sports, and/or socialized with friends.” The Cabinet Office’s report also identifies people who relate with individuals with hikikomori and also prefer to stay inside their homes and defines them as the “hikikomori affinity group” (Definition of hikikomori. 2016: 9–11) [16]. The survey did not recognize hikikomori as an independent clinical category, but rather as a continuous spectrum that included both the healthy and the affinity groups. This view set the tone for subsequent research on hikikomori. Therefore, in the present study, we attempted to follow the spectrum concept and evaluated our participants, who ranged from healthy to severe, using the same hikikomori scale.

Evaluations were conducted by asking parents to consider the most frequent occurrence of (a) and (b) during the past 6 months. Responses were provided on a 5-point scale, ranging from 0 (“Not at all”) to 4 (“Always”). For item (b), the numerical values were reverse scored and then combined with the scores for item (a). The total score represented the degree of severity of hikikomori, with a higher score indicating more severity. We calculated Cronbach’s α as 0.703 upon conducting a reliability analysis.

To investigate which environmental factors could be related to the occurrence and severity of hikikomori during adolescence, we created another novel evaluation scale to measure the following: (1) parental mental health, (2) parental physical conditions, (3) communication between parents and child, (4) communication between parents, (5) conflict between parent and child, (6) conflict between parents, (7) financial status, (8) communication with the community, (9) overuse of the Internet. Evaluations were conducted by asking parents of both groups to consider the circumstances over the past 6 months. Responses were provided on a 5-point scale (included in the “Appendix 1”).

The parents of participants in the hikikomori and control groups answered all questions in the Child Behavior Checklist (CBCL4-18) [43, 44] to evaluate their child’s psycho-behavioral characteristics. The CBCL was developed by Achenbach and colleagues to comprehensively evaluate children’s emotional and behavioral problems [45]. Based on the raw scores from 118 problem-behavior questions featured in the CBCL4-18, the scores of 11 scales were calculated: eight syndrome subscales (i.e., withdrawn, somatic complaints, anxious/depressed, social problems, thought problems, attention problems, delinquent behavior, and aggressive behavior), an internalizing scale, an externalizing scale, and a total score scale. The scores of these 11 scales were converted into standardized t-scores based on country-specific standard values [46‐50]. The CBCL has become a major research tool widely used in retrospective, cohort, and meta-analysis studies [51‐55]. In the present study, we chose to use the CBCL to identify in detail the subclinical characteristics and symptoms related to hikikomori.

We produced descriptive statistics and t-tests to determine between-group differences on the hikikomori severity scale, nine environmental scales, eight CBCL syndrome subscale t-scores, and the total CBCL t-score. Effect size (Cohen’s d) was calculated to ensure that the sample size was sufficient. We considered d > 0.5 as medium effect size and d > 0.8 as large effect size [56]. The dependent variables were approximately normally distributed within each group.

To identify factors related to hikikomori severity, we conducted multiple regression analysis with the severity of hikikomori as the dependent variable and demographic variables (gender, age), CBCL subscale t-scores, and the nine environmental factors as the predictor variables (n = 108). Two models were calculated: the first model (Model 1) was adjusted for all explanatory variables, and multicollinearity verification was performed using variance inflation factor (VIF) statistics. We then conducted a second multiple regression analysis (Model 2) excluding variables exhibiting VIF > 2.0 and variables with a low contribution (β < 0.01) to the first model. All analyses were performed using SPSS (v22) software for Windows and the significance level was set at p < 0.05.

The results of descriptive statistics are presented in Tables 1 and 2. Hikikomori severity was, as expected, significantly higher in the hikikomori patient group (p < 0.001, d > 0.8). With regard to environmental factors, in the hikikomori patient group, “parental psychiatric disorders” (p < 0.05, d > 0.8), “conflict between parent and child” (p < 0.001, d > 0.8), and “overuse of the Internet” (p < 0.05, d > 0.5) were all significantly higher than the control group, while “communication between parents” (p < 0.01, d > 0.8) was significantly lower. In the hikikomori patient group, mean values for the total CBCL score and syndrome subscales were significantly higher than those of the control group. In the hikikomori patient group, the total CBCL score was in the clinical range, while all syndrome subscale scores were in the subclinical range.

To identify factors related to hikikomori severity, a multiple regression analysis was conducted using “severity of hikikomori” as the dependent variable and demographic variables (gender and age), the eight CBCL syndrome subscale scores, and the nine environmental factors as the independent variables (Model 1, Supplemental Table 2). The variable that most contributed to the severity of hikikomori was the CBCL syndrome subscale “withdrawn.” However, after estimating VIF, “withdrawn” was removed as an independent variable as it exhibited multicollinearity.

Therefore, we attempted to use variables with VIF < 2.0 and identified “independent” factors related to hikikomori severity with multiple regression analysis. The following items were excluded from our explanatory variables: withdrawn, social problems, thought problems, attention problems, delinquent behaviors, aggressive behaviors, and conflict between parent and child. Furthermore, variables with a low contribution (β < 0.01) were also omitted, resulting in conflict between parents also being excluded from the explanatory variables. See Table 3 for Model 2 results. Among the selected independent variables, “somatic complaints,” “anxious/depressed,” “overuse of the Internet,” and “lack of communication between parents” were significantly associated with hikikomori severity.

Previous research has found that individuals who exhibit hikikomori are more likely to be male [12, 14], however, gender was not significantly related to hikikomori severity in our study.

Our investigation of environmental factors that may be associated with the occurrence of hikikomori found that the prevalence of psychiatric disorders among parents was significantly higher in the hikikomori group. This indicated that there may be some genetic predisposition; perhaps related to stress tolerance, coping ability, or resilience; preventing adolescents with hikikomori from adequately coping with stressors such as interpersonal problems at school or poor academic performance. A recent preliminary study has shown blood biomarkers uric acid and high-density lipoprotein cholesterol as possibly correlated with an underlying biological pathology of hikikomori [57]. Individual psychological factors including interpersonal problems [14], coping difficulties, conflicting demands, reduced autonomy [58], low self-esteem [34], and a predisposed introverted personality [31] have been shown to play some role in hikikomori propensity. However, the extent to which these underlying vulnerabilities depend on a biological foundation requires further research. The novel scale we designed to measure the environmental factors also requires further testing and validation.

We also found that the hikikomori group had significantly lower scores for communication between parents and significantly higher scores for conflict between parent and child. Overuse of the Internet was also significantly higher in the clinical group. These could be important risk factors for hikikomori but could also be a result of the hikikomori itself. When personal stress and a negative family environment are added to a nonspecific vulnerability, signs of hikikomori could emerge along with adaptation issues. Similarly, maladaptation (in the form of hikikomori) may increase conflicts between parent and child and perhaps eventually lead to decreased communication between parents should they become overwhelmed. Familial factors, including an absent father, overdependence between mother and child [3], highly educated parents, and maternal panic disorder [59] have all been associated with hikikomori.

Our CBCL results showed that middle school hikikomori patients had significantly higher mean scores for all the syndrome subscales and the total score, as compared to the control group. Although the total mean CBCL score for the hikikomori group was in the clinical range, all eight syndrome subscales were subclinical. This may be interpreted as follows: each of these psychiatric signs associated with hikikomori may not be considered clinically serious when considered alone; however, the combination may warrant psychiatric consultation. Given that there is no distinctive psychiatric sign that is specific to “clinical” hikikomori, as compared to other psychiatric conditions, there may be no single strong predictor that could be used for early detection. Rather, its occurrence will need to be judged by analyzing a combination of features that will change along a spectrum that has “severe hikikomori” at its one extreme [15, 16, 38, 39]. Based on our findings, it is unlikely that a specific vulnerability is the foundation of this condition and it is unclear whether the comorbidities reported thus far [10‐13] may be secondary to the development of hikikomori.

We used multiple regression analyses to investigate environmental factors and psychological characteristics that may be associated with hikikomori severity. The CBCL syndrome subscale “withdrawn” was found to contribute the most to hikikomori. This subscale evaluates the psychological tendencies of hikikomori and is one way to quantify “affinity for hikikomori,” as mentioned in the Cabinet Office reports [15, 16]. However, since we tried to investigate psychological factors that may have played a role in social withdrawal (hikikomori affinity) the “withdrawn” phenotype was too centrally involved to be useful and thus could not function as an independent variable in our model due to multicollinearity issues.

The results from our cross-sectional multiple regression analysis revealed that the following independent variables were correlated with hikikomori severity: “somatic complaints,” “anxious/depressed,” “overuse of the Internet,” and “lack of communication between parents”. It is interesting to note that “lack of communication between parents” was a correlate but “conflict between parents” was not. Could this indicate that regardless of whether parents frequently quarreled, more communication between parents could be a protective factor for adolescents with a tendency toward hikikomori? A more sensitive measure of the quality of the communication, such as the Family Assessment Device [60], would be useful to interrogate this further.

It is also not easy to tell whether anxiety and depression are triggers for hikikomori or simply co-occur. They have been identified as factors in other studies, but the exact relationship remains unclear [11, 12].

The relationship between somatic complaints and hikikomori is also unclear. Somatization could be related to nonspecific genetic vulnerabilities mentioned above (e.g. low stress tolerance). As a result of somatization, those with early hikikomori may frequently visit pediatricians about undefined complaints, which presents an opportunity for early detection. Although early screening for hikikomori may be difficult, the symptom of “school refusal” seems to be highly indicative [34‐36]. One must also consider others on the hikikomori spectrum, who may have no problems attending school but communicate very little with people other than the members of their own families (the “hikikomori affinity group”).

Parents should be encouraged to control Internet use in hikikomori children. These recommendations should be emphasized in support programs for hikikomori that target middle school students. One example is an administrative intervention program in French schools that has reduced the number of adolescent drop outs, by making the school staff focus intensely on any student who is absent for 10 half-days in a month. If absenteeism persists, the case is referred to a public prosecutor, unless the situation is handled medically or socially [61]. Unfortunately, hikikomori sufferers are often concealed by families, stopping judiciary and administrator bodies from intervening, thereby greatly impeding prevention and intervention programs. Such situations could even be viewed as “social neglect.” Social welfare services that encourage parents to address difficulties together with their child, especially through home-visit programs, may be effective for decreasing hikikomori severity and duration [21, 36, 62, 63]. Pre-school developmental-behavioral screening and consecutive support programs may also help prevent early hikikomori [64] but adolescence is a critical period for intervention.