Psychometric evaluation of a patient- and caregiver-rated early warning signs scale for acute exacerbations in schizophrenia

Abstract

Background

A psychometrically sound instrument is essential for early detection and relapse prevention of schizophrenia by evaluating early warning signs (EWS). This study aimed to modify and shorten the EWS Psychotic Relapse Signature for use by both caregivers and patients in assessing the EWS of individuals with schizophrenia.

Methods

A total of 312 patient-caregiver pairs were recruited from a mental health service network in southern Taiwan.

Results

Caregiver and patient pairs completed the 55-item EWS Psychotic Relapse Signature. Item response theory analysis identified five items with low discrimination and high difficulty across both caregiver and patient groups. Additionally, eight items demonstrated low discrimination and high difficulty exclusively in patient responses, while two items exhibited these properties specifically in caregiver responses. Differential Item Functioning (DIF) analysis further identified seven items with significant uniform DIF and two items with significant non-uniform DIF, indicating potential bias across rater groups. Following the removal of these suboptimal items, the final version of the Early Warning Sign-Caregiver and Patient Version (EWS-CP) was refined to 31 items. The EWS-CP demonstrated strong internal consistency, with Cronbach’s alpha coefficients of 0.915 for caregivers and 0.917 for patients.

Conclusions

The EWS-CP provides a robust, psychometrically sound tool for relapse monitoring in clinical practice, contributing to improved patient outcomes through timely detection and targeted intervention. The dual-informant design allows for comprehensive EWS detection, aiding clinicians and caregivers in proactive relapse prevention, and enhancing early intervention.

Introduction

Schizophrenia is a psychiatric disorder with a lifetime prevalence of 0.3–0.7% [1]. While some individuals with schizophrenia experience periods of remission and even recovery, the majority remain vulnerable to psychotic symptom exacerbations [1]. Acute psychotic relapses can have serious consequences, including biological risks such as gray and white matter destruction and progressive brain atrophy [2], which in turn contribute to altered neuroplasticity, structural abnormalities, and chemical imbalances [3]. As the number of relapses increases, the risk of chronic illness with severe functional impairment escalates [4]. Thus, acute exacerbation not only negatively impacts the course of schizophrenia but may also contribute to further illness progression over time.

Early recognition of early warning signs (EWS) of psychosis provides a crucial opportunity for relapse prevention and early intervention, reducing the likelihood of further psychotic episodes. EWS of psychosis relapses typically manifests as subtle changes in thought, affect, and behavior before the onset of full-blown psychosis. The progression of relapse often follows a pattern: early non-psychotic phenomena, escalating emotional disturbances, and ultimately the development of psychotic symptoms, usually within a span of less than four weeks [5]. Identifying EWS allows for timely clinical interventions, including modifying treatment strategies, reducing environmental stressors, enhancing monitoring, providing additional support, and increasing pharmacological protection, which can effectively disrupt the relapse process [6]. Furthermore, studies have shown that a combination of low-dose maintenance medication and prompt pharmacological intervention at the onset of EWS can reduce psychotic relapse rates to 12–23% over two years [7–10]. A review of randomized clinical trials comparing EWS-based interventions to usual care found that while the time to relapse and re-hospitalization rates did not significantly differ between the two groups, fewer individuals experienced relapses, and the risk of re-hospitalization was significantly lower in those receiving EWS-based interventions compared to those under standard care [11].

Given the variability in the nature and timing of EWS among individuals, Birchwood, Spencer, and McGovern developed a relapse drill designed to help patients recognize and respond to their early warning signs [12]. In this drill, the first step was to identify the patient’s EWS using the Psychotic Relapse Signature, which comprises three dimensions: thinking, feelings, and behavior. The Relapse Signature was designed to help patients explore any noticeable changes in their thoughts, perceptions, feelings, and behaviors leading up to their most recent episode, as well as any events that may have contributed to its onset [12]. The information collected is then used for relapse prevention planning, including role-play exercises to practice appropriate responses when EWS are recognized. Research has shown that EWS has modest predictive validity, and studies incorporating a wide range of EWS monitoring assessments have demonstrated greater sensitivity in predicting relapse risk [13]. Additionally, family caregivers play a crucial role as extensions of the healthcare system, providing continuous support for individuals with schizophrenia [14]. Notably, 65% of caregivers report being able to recognize EWS in their relatives before a relapse occurs, highlighting their potential role in early intervention and relapse prevention [15].

Extending the EWS Psychotic Relapse Signature as an evaluation for caregivers of patients with schizophrenia and family members is important, for typically family members play a crucial role in monitoring illness progression and are often the first to seek medical assistance when needed. As early as 1980, Herz and Melville reported that 70% of patients and 93% of family members were able to identify one or more specific changes in experiences, thoughts, or behaviors that preceded a psychotic episode [16]. Furthermore, patients and family members agree on the most important warning signs [16]. Effective recognition of early warning signs is enhanced through the involvement of patients, family members, other social network members, and healthcare providers [17]. Notably, family members are particularly adept at identifying subtle changes in the patient, thereby making a critical contribution to monitoring the patient’s condition [18]. This is particularly evident in collectivist Chinese culture, where strong family-centered values and kin-based social networks emphasize familial responsibility for caregiving. In China, 90% of patients with schizophrenia live with their families, compared to 60% in Great Britain and 40% in the United States [19]. Moreover, a 14-year follow-up study in China found that individuals with no family caregivers had significantly higher rates of homelessness and lower survival rates than those with family support [20]. Family members are essential partners in collaborating with healthcare providers throughout the recovery process. A meta-analysis further demonstrated that family intervention for psychosis can reduce relapse risk by 58%, decrease hospitalization duration, alleviate psychotic symptoms, and lead to significant functional improvements [21]. Given the critical role of family caregivers, a user-friendly, psychometrically sound, and reliable instrument is essential for detecting EWS. Such a tool would facilitate timely intervention, improving patient outcomes and preventing relapse.

This study aimed to adapt the Birchwood, Spencer, and McGovern EWS Psychotic Relapse Signature [12] by evaluating whether its 55 symptoms could be restructured into a questionnaire usable by both patients and caregivers, thereby enabling patient self-assessment and caregiver recognition of EWS in time-limited clinical settings. To ensure the psychometric quality of the adapted instrument, Item Response Theory (IRT) was employed to evaluate the properties of each item, with a particular focus on assessing item difficulty and discrimination separately for patient and caregiver responses. This approach allows for the identification of items that perform differently across respondent groups and ensures that the final tool remains informative and clinically relevant for both patients and caregivers.

Methods

Participants

The EWS Psychotic Relapse Signature was originally completed as part of routine clinical care. Because primary caregivers are often the first to notice changes and seek medical attention during a relapse, they were invited to complete the EWS Psychotic Relapse Signature. Eligible caregivers were those who lived with the patient or maintained frequent and close contact and had a comprehensive understanding of schizophrenia symptoms and care needs. Once patients’ symptoms stabilized through medical treatment, healthcare personnel also invited patients to complete the EWS Psychotic Relapse Signature as part of standard care, asking them to recall signs experienced prior to their relapse. This dual-report approach aims to enhance patients’ self-awareness of emerging psychotic symptoms and to support caregivers and healthcare professionals in recognizing and responding to each individual’s unique relapse indicators. All EWS Psychotic Relapse Signature were documented in medical records and collected from a mental health service network in southern Taiwan [22, 23].

Eligibility criteria included individuals over the age of 18 who had been diagnosed with schizophrenia and had experienced a relapse within the past year. To ensure the protection of privacy, all data were anonymized and coded at the time of entry, rendering them fully de-identified and untraceable to individual patients. In total, EWS Psychotic Relapse Signature from 312 patient-caregiver pairs were collected for this study. This study adhered to the ethical standards of the 1964 Declaration of Helsinki and its later amendments, and was approved by the Institutional Review Board of Kaohsiung Armed Forces General Hospital (Approval No. KAFGH 111 − 051). As the study involved only secondary analysis of de-identified data collected during routine care and posed no additional risk to participants, the requirement for informed consent was waived by the ethics committee.

Materials

Demographic information was collected, including patient age, sex, marital status, educational level, and age of onset. The 55-item EWS Psychotic Relapse Signature, developed by Birchwood et al. [12], was translated into Mandarin Chinese using a forward–backward translation process to ensure linguistic accuracy. The EWS Psychotic Relapse Signature assesses three core dimensions: Thinking/perception (19 items), feelings (18 items), behaviors (18 items). Caregivers and stabilized patients completed the EWS Psychotic Relapse Signature by recalling symptoms observed prior to the most recent psychotic episode. Checked symptoms were coded as “1,” while unchecked symptoms were coded as “0”.

Statistical analysis

A descriptive analysis was first conducted to summarize participant demographics. Item reduction followed a multi-step procedure integrating psychometric evaluation and clinical judgment. Initially, all EWS items were analyzed separately for patients and caregivers using the two-parameter logistic (2PL) IRT model (ltm package, RStudio; R Foundation for Statistical Computing, Vienna, Austria) to estimate discrimination (a) and difficulty (b) parameters under the assumptions of unidimensionality and local independence. Items with discrimination values < 0.15 were deemed to have negligible ability to differentiate between individuals at different levels of the latent trait, while difficulty values > 35 were treated as estimation artifacts due to sparse responses or near-perfect separation. Items meeting these criteria, or showing persistent misfit after Bonferroni-adjusted S–X² tests, were flagged for removal.

Model fit was evaluated separately for patient and caregiver datasets using the M2 statistic, RMSEA (≤ 0.06), SRMSR (≤ 0.08), and CFI (≥ 0.95). Item characteristic curves and item information curves were inspected, and person ability estimates (θ) were derived for subsequent analyses.

Differential item functioning (DIF) between patients and caregivers was assessed using the likelihood ratio (LR) DIF procedure in the mirt package, which compares constrained and freely estimated multi-group 2PL models. This method, mathematically analogous to logistic regression LR DIF testing, enables detection of both uniform DIF (differences in difficulty) and non-uniform DIF (differences in discrimination). Items with significant DIF (p < 0.05) were considered for removal if they also showed low discrimination or narrow symptom coverage.

Finally, a panel of psychiatrists and experienced mental health professionals reviewed all flagged items to ensure that essential early warning signs were retained. All IRT analyses were performed in RStudio using the ltm and mirt packages, and all other statistical analyses were conducted in SPSS 24.0 for Windows (IBM Corp., Armonk, NY, USA).

Results

Descriptive analysis revealed that 197 of the 312 participants (63.1%) were male, with an average age of 48.16 years (SD = 10.84) (Table 1).

Table 1.

Demographic distribution of patients with schizophrenia (N = 312)

Variables	N (%)
Male	197 (63.1)
Marital status
Single	190 (60.9)
Married	58 (18.6)
Others	64 (20.5)
Level of education
Illiterate	15 (4.8)
Elementary school	39 (12.5)
Middle school	101 (32.4)
High school	122 (39.1)
College/University and higher	28 (9.0)
Missing	7 (2.2)
Illicit drug use	24 (7.7)
Cigarette use	168 (53.8)
Alcohol use	89 (28.5)
Variable	Mean (SD)
Age (range: 19–85)	48.16 (10.84)
Age of onset	27.33 (9.95)
Duration of illness (years)	20.82 (9.89)

IRT analysis was conducted separately for patient and caregiver ratings of the EWS Psychotic Relapse Signature to assess item-level performance, as presented in Table 2. For the caregiver dataset, the 2PL IRT model demonstrated acceptable global fit: M2(1430) = 2282.64, RMSEA = 0.0438, CFI = 0.968, TLI = 0.967. Although the SRMSR value (0.133) exceeded the conventional 0.08 cutoff, the RMSEA, CFI, and TLI indices indicated satisfactory model–data fit. For the patient dataset, the 2PL IRT model also demonstrated acceptable fit: M2(1430) = 2672.45, RMSEA = 0.0529, CFI = 0.952, TLI = 0.950, with SRMSR = 0.115. Among the 55 items, patient-rated discrimination parameters ranged from 0.11 to 0.39, while caregiver-rated discrimination values ranged from 0.10 to 0.43. Difficulty parameters spanned from 24.4 to 39.7 for patient ratings and from 23.3 to 37.6 for caregiver ratings. The evaluation focused on two primary indicators: item discrimination and difficulty. Items identified as having discrimination < 0.15 and difficulty > 35 were excluded. Based on these criteria, five items were removed: Item 19 (patient: discrimination = 0.08, difficulty = 55.99; caregiver: discrimination = 0.13, difficulty = 41.11), Item 24 (patient: discrimination = 0.06, difficulty = 56.79; caregiver: discrimination = 0.11, difficulty = 40.35), Item 37 (patient: discrimination = 0.10, difficulty = 45.14; caregiver: discrimination = 0.06, difficulty = 84.11), Item 47 (patient: discrimination = 0.06, difficulty = 82.11; caregiver: a = 0.13, difficulty = 40.18), and Item 51 (patient: discrimination = 0.07, difficulty = 49.51; caregiver: discrimination = 0.06, difficulty = 56.42).

Table 2.

Item response theory comparison of discrimination and difficulty parameters for the early warning sign scale between patients and caregivers

Item	Discrimination Patient	Difficulty Patient	Discrimination Caregiver	Difficulty Caregiver
Thinking/perception
1 Thoughts are racing	0.16	27.68	0.16	23.63
2 Senses seem sharper	0.11	33.74	0.14	29.31
3 Thinking you have special powers	0.12	36.01	0.12	32.92
4 Thinking that you can read other people’s minds	0.19	27.90	0.14	31.50
5 Thinking that other people can read your mind	0.17	29.30	0.14	31.29
6 Receiving personal messages from the TV or radio	0.14	40.16	0.13	34.81
7 Having difficulty making decisions	0.12	28.33	0.16	21.88
8 Experiencing strange sensations	0.11	39.84	0.13	34.42
9 Preoccupied about 1 or 2 things	0.21	20.97	0.19	21.45
10 Thinking you might be somebody else	0.11	46.02	0.16	40.28
11 Seeing visions or things others cannot see	0.24	24.95	0.16	28.12
12 Thinking people are talking about you	0.17	17.33	0.16	19.42
13 Thinking people are against you	0.24	20.76	0.24	23.44
14 Having more nightmare	0.16	38.83	0.10	37.16
15 Having difficulty concentrating	0.17	14.24	0.21	14.40
16 Thinking bizarre things	0.17	17.58	0.17	21.48
17 Thinking your thoughts are controlled	0.34	15.49	0.22	17.59
18 Hearing voices	0.30	8.20	0.22	9.50
19 Thinking that a part of you has changed shape	0.08	55.99	0.13	41.11
Feelings
20 Feeling helpless or useless	0.12	34.85	0.16	24.54
21 Feeling afraid of going crazy	0.19	28.59	0.17	27.06
22 Feeling sad or low	0.10	35.53	0.14	23.08
23 Feeling anxious and restless	0.24	11.79	0.26	12.07
24 Feeling increasingly religious	0.06	56.79	0.11	40.35
25 Feeling like you’re being watched	0.17	22.52	0.23	19.87
26 Feeling isolated	0.14	33.96	0.23	26.26
27 Feeling tired or lacking energy	0.17	29.52	0.12	36.16
28 Feeling confused or puzzled	0.27	16.46	0.22	18.31
29 Feeling forgetful or far away	0.15	19.90	0.14	23.00
30 Feeling in another world	0.14	35.51	0.19	28.02
31 Feeling strong and powerful	0.14	40.87	0.15	33.37
32 Feeling unable to cope with everyday tasks	0.17	12.36	0.20	15.21
33 Feeling like you are being punished	0.15	41.46	0.16	33.77
34 Feeling like you cannot trust other people	0.28	16.85	0.20	19.90
35 Feeling irritable	0.32	10.16	0.32	8.97
36 Feeling like you do not need sleep	0.23	18.45	0.15	23.86
37 Feeling guilty	0.10	45.14	0.06	84.11
Behaviors
38 Difficulty sleeping	0.21	9.86	0.18	10.61
39 Speech comes out jumbled filled with odd words	0.30	11.56	0.23	16.52
40 Talking or smiling to yourself	0.21	13.13	0.19	16.72
41 Acting suspiciously as if being watched	0.29	15.21	0.29	16.54
42 Behaving oddly for no reason	0.15	17.59	0.16	21.28
43 Spending time alone	0.12	21.40	0.16	18.65
44 Neglecting your appearance	0.24	22.04	0.22	22.65
45 Acting like you are somebody else	0.15	40.48	0.17	31.98
46 Not seeing people	0.14	30.84	0.17	30.15
47 Not eating	0.06	82.11	0.13	40.18
48 Not leaving the house	0.14	29.67	0.11	33.09
49 Behaving like a child	0.24	24.63	0.20	28.48
50 Refusing to do simple requests	0.16	23.19	0.21	22.33
51 Drinking more	0.07	49.51	0.06	56.42
52 Smoking more	0.12	25.26	0.11	28.50
53 Movements are slow	0.13	40.50	0.19	23.00
54 Unable to sit down for long	0.20	13.59	0.18	16.21
55 Behaving aggressively	0.14	20.24	0.13	26.60

In addition, several items demonstrated low discrimination and high difficulty based on patient ratings alone. Eight such items met these criteria: item 3 (discrimination = 0.12, difficulty = 36.01), item 6 (discrimination = 0.14, difficulty = 40.16), item 8 (discrimination = 0.11, difficulty = 39.84), item 10 (discrimination = 0.11, difficulty = 46.02), item 22 (discrimination = 0.10, difficulty = 35.53), item 30 (discrimination = 0.14, difficulty = 35.51), item 31 (discrimination = 0.14, difficulty = 40.87), and item 53 (discrimination = 0.13, difficulty = 40.50).

On the other hand, two items demonstrated low discrimination and high difficulty based on caregiver ratings. These items included Item 14 (discrimination = 0.10, difficulty = 37.16), and Item 27 (discrimination = 0.12, difficulty = 36.16). Since the goal was to modify and shorten the EWS Psychotic Relapse Signature for use by both caregivers and patients, all items that exhibited low discrimination and high difficulty in either group were removed. This refinement resulted in a final set of 40 items retained for further analysis.

DIF analysis was then conducted to examine whether items in the EWS Scale functioned differently between patient and caregiver ratings, as shown in Table 3. Seven items exhibited significant uniform DIF (p < 0.05), 16 (p = 0.026), 20 (p = 0.013), 32 (p = 0.002), 39 (p = 0.002), 40 (p = 0.028), 42 (p = 0.019), and 55 (p = 0.012). Two additional items showed significant non-uniform DIF, Items 17 (p = 0.022) and 36 (p = 0.028).

Table 3.

Differential item functioning (DIF) analysis results for the early warning sign scale comparing patient and caregiver ratings

Item	Uniform DIF p-value	Non-uniform DIF p-value
1	0.083	0.929
2	0.930	0.337
4	0.944	0.216
5	0.792	0.439
7	0.269	0.189
9	0.759	0.524
11	0.803	0.071
12	0.229	0.925
13	0.139	0.954
15	0.319	0.344
16	0.026	0.919
17	0.988	0.022
18	0.881	0.061
20	0.013	0.336
21	0.226	0.582
23	0.506	0.618
25	0.893	0.134
26	0.345	0.068
28	0.548	0.343
29	0.174	0.809
32	0.002	0.309
33	0.105	0.838
34	0.459	0.091
35	0.144	0.980
36	0.359	0.028
38	0.979	0.404
39	0.002	0.101
40	0.028	0.537
41	0.270	0.994
42	0.019	0.864
43	0.686	0.116
44	0.959	0.711
45	0.054	0.798
46	0.321	0.341
48	0.926	0.500
49	0.221	0.507
50	0.321	0.152
52	0.418	0.761
54	0.155	0.549
55	0.012	0.708

The final version of the EWS Scale—Caregiver and Patient Version (EWS-CP) comprised 31 items. The overall scale demonstrated strong internal consistency, with Cronbach’s alpha coefficients of 0.915 for caregiver ratings and 0.917 for patient ratings, indicating high reliability across respondent groups.

Discussions

This study aimed to modify and shorten the EWS Psychotic Relapse Signature for use by both caregivers and patients in assessing the EWS of individuals with schizophrenia. By applying IRT analysis to patient and caregiver ratings, identifying items with low discrimination and high difficulty for patients and caregivers, and items with uniform and non-uniform DIF between patients and caregivers. The final EWS-CP resulted in 31 items, demonstrated high reliability. Intended for use by non-professionals, the EWS-CP is an easy to comprehend, psychometrically sound and efficient instrument for detecting EWS in patients with schizophrenia. By enabling patients to self-evaluate and allowing caregivers to observe EWS symptoms, the EWS-CP facilitates timely identification of EWS, enhancing early clinical intervention to prevent acute exacerbation and reduce the likelihood of full-blown relapse.

The 31-items remained in the EWS-CP are consistent with findings from previous systematic reviews, which have identified sleep disturbances and mood changes as the most reliable predictors of deterioration in the delusion domain, and sleep disturbances and heightened suspiciousness have been consistently associated with worsening symptoms in the hallucination domain [24]. Notably, the item “Difficulty sleeping” was endorsed by both patients and caregivers as a key indicator of EWS, aligning with existing evidence that emphasizes sleep disturbances as a critical target for early warning sign assessment [24]. Items related to suspiciousness and hallucinations include “Feeling like you’re being watched”, “Feeling like you cannot trust other people”, “Seeing visions or things others cannot see”. In contrast, negative affect items such as “Feeling sad or low” (characterized by low discrimination and high difficulty for patients) and “Feeling helpless or useless” (characterized by low discrimination and high difficulty for caregivers) were excluded. A prior systematic review also indicated that low mood may be present but is less frequently reported by patients or observed by caregivers, as it is primarily an internal experience that is not always expressed through observable behavior [17]. Furthermore, this study’s exclusion of these items may be attributed to cultural differences; individuals from collectivist cultures often demonstrate slower emotional development [25] and exhibit higher levels of alexithymia, characterized by difficulties in identifying, differentiating, verbalizing, and communicating emotions [26]. Instead, low mood maybe observable through withdrawal behaviors such as “Not seeing people” and “Not leaving the house”. Suggesting that the EWS-CP captures EWSs that are both clinically salient and more readily detectable in cross‐cultural caregiving contexts, enhancing its practical utility for timely relapse intervention.

The IRT analysis provided critical insights into item-level performance. Some item difficulty estimates in this study exceeded the conventional − 3 to + 3 range expected under a standard normal theta distribution. This occurred for items with very low endorsement frequencies, typically representing severe or infrequent relapse signs that require a high level of the latent trait to be observed. In such cases, the IRT model places the difficulty parameter further along the trait continuum to reflect the rarity of endorsement, which can yield large numerical values. These extreme estimates should be interpreted as indicators of item rarity and severity rather than as attainable theta scores for most respondents. Items were considered to have poor psychometric properties when they met both of the following criteria: low discrimination (< 0.15), indicating minimal ability to differentiate between respondents at different levels of the latent trait; and high difficulty (> 35), suggesting the symptom is rarely endorsed in the sample. Five items met these criteria in both patient and caregiver ratings, showing that they were consistently uninformative across informants. Further item‐level analysis identified eight items meeting these criteria exclusively in patient ratings, while two items did so only in caregiver ratings, reflecting potential perceptual gaps. Comparing discriminative items identified between patients and caregivers showed that, aside from the two behavioral items “not eating” and “drinking more”, caregivers were less likely to identify subtle internalized experiences such as “Thinking that a part of you has changed shape,” “Feeling guilty,” and “Having more nightmares”. The removal of these items contributes to a streamlined EWS‐CP, enhancing its utility for both patients and caregivers.

DIF analysis further elucidated the disparities in symptom reporting between patients and caregivers. The identified items were categorized into uniform DIF and non-uniform DIF, each bearing distinct clinical implications. Seven items exhibited significant uniform DIF, indicating that the probability of endorsement consistently differed between patients and caregivers across the full range of the latent trait, independent of symptom severity. These uniform differences likely reflect systematic rater tendencies, such as differences in symptom awareness or perceptual biases, rather than severity-dependent effects. In contrast, two items demonstrated significant non-uniform DIF, where the magnitude or direction of rater differences varied according to symptom severity. The observed uniform DIF is consistent with previous findings that, although patients and family members often agree on major warning signs, discrepancies are more pronounced for prepsychotic or psychotic symptoms such as incoherent speech and visual hallucinations [16, 17]. In these cases, caregivers consistently rated symptom severity higher than patients, a pattern that may be explained by patients downplaying or minimizing changes in their condition to preserve a sense of functional normalcy, whereas caregivers are more likely to detect and attribute these changes clinical significance [17]. Recognizing such uniform DIF patterns reinforces the importance of incorporating both patient and caregiver perspectives to capture a more comprehensive and balanced assessment of the patient’s clinical condition.

The final 31-item version of the EWS-CP demonstrated strong internal consistency, with Cronbach’s alpha coefficients exceeding 0.90 for both patient and caregiver groups. This high level of reliability reflects robust measurement properties across different rater perspectives, supporting its utility for clinical monitoring.

A limitation of this study is that patients and caregivers were recruited from a single mental health service network in southern Taiwan. Furthermore, because patients completed the EWS after their symptoms had stabilized, their responses may have been influenced by recall bias. Future research should include follow-up assessments to examine whether these EWS remain consistent during subsequent acute exacerbations, as well as recruit a larger and more diverse sample from multiple sites to enhance the generalizability of the EWS-CP. Additionally, cultural differences in emotional processing may influence the structure of EWS. Research suggests that Asian populations exhibit greater emotional restraint and higher levels of alexithymia [25]. In contrast, cultures that place a stronger emphasis on identifying and expressing emotions may be more likely to distinguish between different EWS and dimensional constructs. Therefore, the present findings may not be fully generalizable to populations in other cultural contexts, and future studies should explore cross-cultural variations in EWS perception and assessment to further refine the EWS-CP and improve its applicability across diverse populations.

Conclusions

This study refined the original 55-item EWS Psychotic Relapse Signature into a 31‐item patient–caregiver version with strong psychometric properties. Its dual‐informant format enables comprehensive monitoring and is accessible to non‐professionals, supporting use in routine care. By facilitating timely detection and collaborative intervention, the EWS‐CP has the potential to reduce relapse severity and improve long‐term outcomes in individuals with schizophrenia.

Author contributions

All authors contributed to the study conception and design. Material preparation and data collection were performed by HL and YLT. Data analysis and interpretation were performed by PFC and FWL. The first draft of the manuscript was written by PFC. All authors contributed to and have approved the final manuscript.

Funding

No funding was received for this study.

Data availability

The data supporting the findings of this study are available upon request from the corresponding author.

Declarations

Ethical approval and consent to participate

The study was conducted in accordance with the ethical standards of the 1964 Declaration of Helsinki and its later amendments and was approved by the Institutional Review Board of Kaohsiung Armed Forces General Hospital (Approval No. KAFGH 111 − 051). Because the analysis involved only de-identified secondary data collected during routine care and posed no additional risk to participants, the requirement for informed consent was waived by the ethics committee.

Consent for publication

Not applicable.

Clinical trial number

Not applicable.

Competing interests

The authors declare no competing interests.