Patients diagnosed with FEP (not >6 mo onset of illness) by experienced psychiatrists according to the DSM-IV-TR criteria (high diagnostic stability)²² were selected randomly by 1 researcher (W.T.C.) from the patient lists (in alphabetical order of surnames) of 2 regional outpatient clinics (OPCs) for early psychosis service. After initial screening, 460 of about 1000 FEP patients attending the OPCs met the study inclusion criteria below; 116 of them on the patient lists were selected using computer-generated random numbers. With their consent, informed consent was also obtained from 1 main family caregiver to participate.

The inclusion criteria for caregivers were: (a) aged 18 years or above, (b) able to communicate in written and conversational Chinese/Cantonese, (c) first-time caregiver (for mental illness), and (d) lived with and provided most of the daily care and support for the patient with FEP. Exclusion criteria for caregivers were suffering from an acute episode of mental illness and recipients of specialized family interventions. Inclusion criteria for patients were: (a) diagnosis of FEP²² and without co-morbidity of another mental illness at baseline; and (b) first contact with mental health services. Exclusion criteria for patients included relapse/re-hospitalization before randomization and receiving another family intervention.

A computer-generated randomization list was obtained from a statistician, in blocks of 10 and given to a research assistant for recruitment. One researcher allocated the next available number on the trial entry, and the code was concealed until completion of the baseline measurements to avoid allocation bias. After random selection and baseline measurement, caregivers were randomly assigned into CSPSB or UOFS using computer-generated random numbers by the statistician.

Sample size was calculated a priori. Based on 2 clinical trials of family intervention in FEP (China and Australia),^17,23 the effect sizes on family burden were 0.60 and 0.58 and re-hospitalization (days of hospital-stay) were 0.68 and 0.56, respectively. Therefore, we estimated that 58 family dyads per intervention with 15% expected attritions were required to provide 80% power (2-sided, P < .05) to detect a statistically significant difference on changes in family burden of 1.5 points (SD = 2.7)/re-hospitalization rate of 4.5 d/mo (SD = 8.7),²⁴ indicating a more conservative effect size (0.50) than the above-mentioned trials.

Study outcomes were assessed by a research assistant who received 1 day’s training by the researchers and assessed for inter-rater reliability of scale ratings (ICC = 0.88–0.95).

Ethical approval was obtained from the Research Ethics Committee of The Hong Kong Polytechnic University and the OPCs (HSEARS20140218003). After obtaining written consent from family dyads, the trained research assistant (who was blind to group assignment) administered the baseline measurements (Time-1) before group allocation, and 3 outcome measurements in the OPCs at 1 week (Time-2), 6 months (Time-3) and 12 months (Time-4) post-intervention. Patients’ re-hospitalization rates (frequency and d/mo) and psychotropic medications and their adherence rates were also examined.

All outcome and sociodemographic data were screened and analyzed using IBM’s SPSS, version 20.0. Homogeneity of the 2 study groups’ sociodemographic characteristics and baseline outcome measure scores was checked with independent sample t (2-tailed) or Chi-square tests. Based on an intention-to-treat basis and no violation of assumptions of multivariate analyses,²⁴ a repeated-measures mixed-model MANOVA test was performed for the outcome variables (FBIS, ECI, SPSI-R:S, SLOF, PANSS, total amount [dosage] of psychiatric medications, and re-hospitalization rates) to determine the interactional (Group × Time) treatment effects, and univariate between-group effects across time. If between-group effects were found significant on the outcome measure(s), Helmert’s contrasts tests were used to identify where the significant differences on each outcome mean score(s) were located.²⁴ For outcomes showing significant between-group differences over time, the mean scores of each outcome in the CSPSB were compared between clinics, using 1-way ANOVA test. The level of statistical significance was set at 0.05, except univariate ANOVA tests for between-group effects at 0.006 (using a Bonferroni adjustment).²⁴

One hundred thirty of the 460 eligible patients were randomly selected and contacted by the first author; 116 patients agreed to participate (ie, response rate = 89%). One hundred ten families (94.8%) completed the intervention and follow-up; 2 families in the CSPSB and UOFS withdrew, or were lost to contact at the first post-test (Time-2), and were not included in the final data analysis (figure 1). While 2 caregivers discontinued their participation from the CSPSB, 54 (93.1%) completed ≥4 modules, 2 orientation and ≥2 review sessions (>80% of the CSPSB). Reasons for withdrawal or discontinuing from CSPSB participation/the study were: insufficient time to attend (n = 2), patient’s mental state worsened (n = 1), and lack of interest in participation (n = 2). During the telephone calls, 4 caregivers in the CSPSB were referred to family counseling (n = 2 at the 8th and 10th week of intervention, respectively) and suicidal prevention service (n = 2 at the 16th week of intervention).

The sociodemographic characteristics of caregivers and patients (n = 58 in each group) are summarized in table 1. Caregivers’ mean age was about 50 years (CSPSB: M = 49.1, SD = 9.8; UOFS: M = 50.2, SD = 9.0) and about two-thirds were female. Their relationships with patients were mainly parent (43.0% and 44.8%), spouse (both 19.0%), or child (both 19.0%).

The mean age of the patients was about 26 years. More than half (57.0% and 58.6%) were male and the majority (82.8% and 84.5%) were taking a low/medium dosage of antipsychotics (haloperidol equivalents = 2.5–4.9mg/d).³³ Over 80% were on oral medication only, mainly typical (32.8% and 31.0%) and atypical/blended (48.3% and 50.0%) antipsychotics. On average, 2 family members lived with the patients, and the duration of illness was 4 months (ranged 1–6 mo). The duration of untreated psychosis was about 2 months (range 0–4 mo).

There were no significant differences between the 2 groups in any sociodemographic characteristics (P > .11) and no significant correlations (Spearman’s r < .12) between any outcome variables at baseline.

There were no significant differences in any of the outcome mean scores between the 2 groups at baseline (P > .13), thus negating the need for co-variance analysis. Results of the MANOVA indicated a statistically significant difference between groups on the combined outcome variables, F(6,110) = 6.82, P = .001 (Wilks’ λ = 0.91; a large effect with partial η² = 0.24).²⁴ As indicated in table 2 (with independent F values for MANOVA), there was a statistically significant interaction (Group × Time) of treatment effects for the CSPSB (with large effect sizes),²⁴ with significantly greater improvements in caregivers’ ECI [F(1,110) = 7.21, P = .0008, partial η² = 0.28] and its subscale scores [Positive experiences: F(1,110) = 5.68, P = .005; Negative experiences: F(1,110) = 7.80, P = .0005] and FBIS [F(1,110) = 6.86, P = .001, partial η² = 0.23] and its 5 subscales [F(1,110) = 5.12–7.13, P = .005–.0008]. The CSPSB group also had significant greater improvements in patients’ SLOF score [F(1,110) = 6.95, P = .003, partial η² = 0.25], PANSS score [F(1,110) = 5.98, P = .005, partial η² = 0.20], and duration of re-hospitalizations [F(1,110) = 5.78, P = .005, partial η² = 0.19].

The Helmert’s contrasts tests results indicated that there were statistically significant differences between the CSPSB and UOFS on the changes in mean scores of:

There were no significant differences in the dosages of antipsychotics (P = .09; table 2), types of family/patient services utilized (P = .13) and medication adherence rates (P = .15) across measurements between groups. The patient and family support services utilized by the CSPSB and UOFS participants over the follow-up included: individual patient (n = 30 and 21; median = 6 and 8 times, range 3–10 and 2–11, respectively) and family counseling (n = 32 and 24; median = 8 and 5 times, range = 4–13 and 2–9, respectively), supportive employment (n = 30 and 31; median = 7 and 5 times, range = 4–10 and 2–7, respectively), social skills training (n = 28 and 26; median = 8 and 6 times, range = 4–11 and 2–10, respectively), and social welfare service (n = 45 and 48; median = 9 and 8 times, range = 4–14 and 4–16, respectively); all patients (n = 56 in both groups) received psychiatric consultation and brief education sessions on mental illness (median = 4 and 7 times, range = 2–6 and 5–10, respectively). The average medication (antipsychotic) adherence rates at Times 1–4 were 69.5%–89.8% (range 60.1%–95.2%) in the CSPSB and 70.8%–83.3% (range 58.2%–89.1%) in the UOFS.

Comparisons of mean scores on caregiver and patient outcomes at Times 2–4 between the CSPSB subgroups in terms of availability of other family caregiver(s) and study sites (OPCs) indicated no significant differences on any outcome scores at all post-tests between the subgroups (P > .15). The CSPSB indicated consistent greater reduction in the percentages of patients being re-hospitalized over the 4 measurement periods than the UOFS group (55.2% and 51.0% [Time-1], 39.0% and 40.4% [Time-2], 37.3% and 43.1% [Time-3], and 32.5% and 40.0% [Time-4]). However, these percentages were not significantly different between-group across time (Kruskal-Wallis test, P = .10).

Given relatively less available usual family support services than many Western countries were provided at the OPCs, the 5-month CSPSB demonstrated very positive effects on family caregiving at 12-month follow-up. The primary hypotheses concerning the effects of bibliotherapy on the caregiving experience and burden were supported, with the CSPSB group reporting significantly greater and substantive improvements in positive appraisals of caregiving experiences (ECI) and aspects of family burden at 12-month follow-up, than the UOFS. The CSPSB also experienced a greater reduction in negative appraisals of caregiving and more positive experiences of family relationships/communication with their FEP relative than the UOFS group. The McCann and colleagues’¹⁸ RCT, using a similar bibliotherapy manual, indicated partial benefits for caregivers’ appraisals of caring for their FEP relative but the benefits were only on 2 aspects of appraisals (“positive personal experiences” and “need to back-up” subscales), and with short-term follow-up (16wk). In addition, Gleeson and colleagues’³⁵ RCT of family CBT-based relapse prevention program demonstrated a similar positive effect in a wide range of caregivers’ appraisals of their caring experiences as shown in the present study over 2-year follow-up. Nevertheless, the relapse prevention program was more resource-intensive and lengthy and did not have any effect on caregiving burden.

The bibliotherapy in the present study also demonstrated significant benefit for caregivers’ burden of care. The CSPSB group indicated significantly greater improvements in their family burden on finance, social activities/relationships and general health, than the UOFS group over the 12-month follow-up. In contrast with other bibliotherapy programs,^18,35,36 the CSPSB could provide prompt post-intervention benefits to the caregivers in coping with their initial negative perception of the illness and family burden associated with their relative’s FEP diagnosis and newly-adopted caregiving role, which may be due to the strong interconnectedness of families in Hong Kong; and these benefits could help sustain them for a longer period of follow-up.

The CSPSB also demonstrated a number of positive effects for patients with FEP, including reduced severity of psychotic symptoms, improved functioning and shorter re-hospitalizations at 6- and 12-month follow-up. Whilst few family interventions for people with FEP produce significant and consistent beneficial effects on patients’ functioning and psychotic relapse,^1–3 these favorable findings may be attributable to the experiential learning, self-help and user-friendly approach of the bibliotherapy in improving caregivers’ experience of, and skills in, caring. It may also equip caregivers to better manage patients’ FEP and detect the signs of relapse and engage in help-seeking earlier. Nevertheless, it is noteworthy that over one-third of patients were hospitalized during follow-up, although slight reductions in the average number of re-hospitalizations occurred in the CSPSB. This may suggest the usual outpatient care provided at the clinics is not adequate/effective in these patients’ community-based rehabilitation.

There was no significant difference between the 2 study groups on social problem-solving ability, although a gradual improvement in this ability was noted in the CSPSB. A possible explanation for this non-significant result is that the problem-solving orientation/style, which is correlated with the individual’s complex social situation and family environment, can be changed and thus evaluated only in the longer-term.^29,34 To establish a rational and positive problem-solving ability in caregivers, more emphasis may need to be added to interpersonal/communication skills training in order to strengthen their social interactions and trusting relationships with patients.

Caregivers’ rates of completion of the CSPSB (93%) and overall attrition (5%) in this study are much more favorable than other family/caregiver intervention studies in FEP.^10,18,19,35 There are 4 possible explanations for this high participation/low attrition rate. First, it could be explained by the benefits induced by the problem-solving approach and easy-to-read manual, shorter duration and briefer content of the intervention, regular review sessions and telephone calls, and self-directed approach to module completion at the caregivers’ convenience and pace. Second, first-time caregivers might have perceived they were mandatory or had no choice but to participate and complete the intervention, even though the researchers were not employees of the clinics and had discussed the trial to ensure participation was voluntary. In addition, the caregivers with high income and education might have more time and tend to participate more in the care of their loved one. Third, it might be attributable to a culture of trust and respect for people perceived to be in authority, common in Chinese/Asian populations.^8,16,20 Finally, the high participation rate might be due to the fact that very few tailored or brief and user-friendly family support services have been provided to Chinese caregivers of people with FEP.^7,10,36

Based on our previous guided self-help family intervention studies,^11,16–18 and feedback from the CSPSB group during the review sessions, completion of 1 module of bibliotherapy per month was found to be appropriate and desirable for caregivers’ learning and rehearsals of positive appraisal of caregiving. The review sessions could also help them consolidate what they had learned and clarify questions about the reading materials.

This trial was 1 of very few to test the effectiveness of a clinician-supported bibliotherapy in providing family support for patients with FEP,^37,38 the first in Chinese/Asian family caregivers of these patients, and was found to be effective over a medium-term (12 mo) follow-up. The study design, such as clear procedures for recruitment, randomization and data collection and analyses and blinding of the outcome assessor, researchers and clinical staff throughout the study, ensured a high-quality RCT.

There are 3 potentially important limitations to this study. First, the sample was recruited from 2 of 18 outpatient clinics in Hong Kong and the patients experienced FEP for <6 months. Therefore, this selective sample may not be representative of the broader FEP population.^23,39 Second, despite random selection of caregiver participants, most were young (aged 20–49 y) and literate with secondary or higher level of education (>80%) and higher monthly incomes than the local population’s average value. The participants’ sociodemographic characteristics and/or baseline outcome scores should have been compared with those of the nonparticipants, in order to reveal the representativeness of the sample (ie, generalization of the findings). Last, the use of self-report measures for the caregiver outcomes might induce participants’ biased responses to the treatment effects, thus requiring objective measurements to enhance the internal validity.⁴⁰