Among the 350 first-visit patients, 309 patients were under 65 years old. 304 patients were included in the study analysis excluding 5 patients who did not answer the question identifying Hikikomori status. The number of “current” Hikikomori patients was 60 (19.7%). Excluding these patients, 81(26.6%) had experienced past Hikikomori. The percentage of patients who presented in-person at the first visit was significantly smaller (p < .001) in current Hikikomori patients (n = 34, 56.7%) compared to other patients (past Hikikomori patients, n = 75, 92.6%; others, n = 151, 92.6%).
Table 1 shows the comparison of basic demographics among all Hikikomori patients and “others.” The distribution of age was unusual, with the distribution of age of “current” Hikikomori patients being bimodal, the most frequent ages being 20, 40 - 45, whereas those of past Hikikomori patients and others were left-skewed and the peak frequency was at 22.5 and 20, respectively. (Figure1) The percentages of patients with past school refusal were significantly higher in current (n=31, 59.6%) and past Hikikomori patients (n=35, 50.0%) than in others (n = 28, 17.8%) (p < .001). The number of clinic visits was significantly larger in past Hikikomori patients (median = 1.0, interquartile range (IQR) = 0 - 3.0) than in others (median = 1.0, IQR = 0 - 1.0) (p = .002). Bonfferoni correction for multiple comparison setting the significance threshold as p<0.0125resulted in no significant difference in the number of clinic visits among groups.
Table 2 shows the comparison of the basic demographics of Hikikomori patients and others who visited the clinic in person at the first visit. The between-group difference of all patients was replicated in this analysis; there was no significant difference between groups in age and percentage of females; the percentage of patients with past school refusal was significantly higher in the “current” and “past” Hikikomori patients, and the number of clinics visited was significantly higher in “past” patients than others. “Current” and “past” Hikikomori patients were significantly more anxious than others as measured by the OASIS scale (“current” Hikikomori mean (standard deviation; SD) = 12.3(4.1);” past” Hikikomori 10.8 (4.7); “others” 8.0 (4.9); p < .001). There is no significant difference in diagnosis between groups. Bonfferoni correction for multiple comparison setting the significance threshold as p<0.008 resulted in no significant difference in the number of clinic visits among groups.
We compared clinic attendance history as of the one-year follow-up between the groups (Table 3). The percentage of “home visits by doctors” and “continuous consultation by family” were significantly larger in “current” Hikikomori patients (home visit by doctor n = 6, 10%; continuous consultation n = 4, 6.7%) than “past” Hikikomori patients (n = 0, 0%; n = 0, 0%) and “others” (n = 0, 0%; n = 1, 0.6%) (p < .001). Percentage of patients who had recovered and finished visiting the clinic was larger in “others” (n = 37, 22.7%) than “current” (n = 4, 6.7%) and “past” hikikomori patients (n = 4, 4.9%) (p < .001).
Table 4 indicates support from a “nurse,” “PSW,” “OT/PT,” and “psychologist” as of the one-year follow-up. The percentage of support from nurses was significantly higher in “current” Hikikomori patients (n = 9, 15%) than “others” (n = 6, 3.7%) (p = .003). There was no significant difference in support from other medical staff members.
Table 5 indicates social function at pre- and post- follow-up. One person of the “current” Hikikomori patients in the “work/school/homemaker” category was a housewife. Over half of the “past” Hikikomori patients and “others” were in the “work/school/homemaker” category at pre- (past Hikikomori patients n = 43, 53.1%; others n = 112, 69.1%) and post- (past Hikikomori patients n = 45, 57.0%; others n = 128, 79.0%) follow-up, whereas small number of “current” Hikikomori were in the work/school/homemaker category in pre- (n = 1, 1.7%) and post- (n = 14, 23.3%) follow-up. Around half of the “current” Hikikomori patients (n = 31, 51.7%) were restricted in indoor activity without going out in the pre-follow-up period, which was reduced in post-follow-up (n = 14, 23.3%). Some “current” patients restricted their activity in their room prior to the the-follow-up period (n = 3, 5.0%).
We analyzed factors contributing to clinic attendance and social function, which were the primary outcomes. Table 6 shows the results of uni- and multivariate logistic regression for the clinic attendance variable. The results of the multivariate logistic regression indicated that “current” Hikikomori status at baseline predicted significantly less regular attendance/completed treatment at the clinic (odds ratio(OR) 0.43, 95% confidence interval (CI) 0.22–0.83, p = .012), and the support from PSW and OT/PT significantly predicted more regular attendance/completed treatment (PSW OR = 2.35, 95% CI 1.14–4.86, p = .021; OT OR = 6.07, 95% CI 1.28–28.71, p = .023). Being female, and having support from a nurse, PSW, or OT/PT significantly related to maintained contact with patients at the clinic or at home (female OR 1.66, 95% CI 1.01–2.73, p = .046; support from nurse OR = 8.44, 95% CI 1.05–68.11, p = .045; PSW OR 2.44, 95% CI 1.12–5.32, p < .025; OT/PT OR 10.06, 95% CI 1.25–80.89, p = .030).
Table 7 shows the results of uni- and multivariate logistic regression for social function. The result of multivariate logistic regression showed that “current” status and “past” Hikikomori experience at baseline significantly decreased and “female” significantly increased the work/school status at the one-year follow-up (current Hikikomori OR = 0.30, 95% CI 0.13–0.68, p = .004; past Hikikomori OR = 0.40, 95% CI 0.20–0.82, p = .012; female OR = 1.96, 95% CI 1.04–3.70, p = .037).
Bonfferoni correction for multiple comparison setting the significance threshold as p<0.025 for two primary outcomes did not change the significance according to the Hikikomri status.
We conducted the same analysis with the “current” Hikikomori patients. The “attendance in person at the first visit” was included as an independent variable this time. Factors that significantly predicted the regular visit/completed treatment by the logistic regression were as follows: the attendance in person at the first visit (univariate OR = 6.97, 95% CI 2.18–22.26, p = .001; multi-variate OR = 21.59, 95% CI 3.1– 150.30, p=.002), support by nurse (univariate OR=11.43, 95%CI 1.33–98.34, p=.027; multi-variate OR=10.00, 95% CI 0.73–137.51, p = .085), and support by PSW (univariate OR = 3.55, 95% CI 0.97–13.03, p = .056; multivariate OR = 15.24, 95% CI 1.62–143.26, p = .017). The percentage of patients who had attended the clinic regularly as of their one-year follow-up was 79.3% (n = 23) for those whom had “attended in person at first visit” and 35.5% (n = 11) for “others.”
Factors that significantly predicted the maintained contact with patients at clinic or home were as follows: the “attendance in person at the first visit” (univariate OR=3.79, 95%CI 1.26 to 11.46, p=.018; multivariate OR=4.74, 95%CI 1.04 to 21.58, p=.044) and “support by PSW” (univariate OR=4.62, 95%CI 0.93 to 23.01, p=.062; multivariate OR=14.43, 95%CI 1.29 to 161.42, p=.03).
The “female” factor significantly increased patients’ work/school/homemaker status at a one-year follow-up (univariate OR = 4.16; 95% CI 1.06–19.50; p = .041, adjusted for baseline social function; multivariate OR=6.23, 95%CI 0.95 to 40.83, p=.057). No other factors significantly predicted the work/school/homemaker social function. No factors predicted work/school status using the multivariate logistic regression.