The lifetime prevalence of psychotic disorders is estimated to be 1% to 3% worldwide.^1,2 Despite the low prevalence, psychotic disorders like schizophrenia are among the world’s leading causes of disability and morbidity.^3,4,5 Notably, people diagnosed with psychosis are more likely to die around 10 years earlier than the general population.^6,7,8 In contrast to psychotic disorder, psychotic-like experiences (PLE) are much more common in the general population.^9,10 Psychotic-like experiences are transient for most people, but they can be a harbinger of future psychotic disorders.^11,12 Moreover, those experiencing PLEs are at increased risk for developing other psychiatric disorders such as anxiety, mood and substance use disorders.^13,14,15 Hence, PLEs may reflect an underlying susceptibility to a broad range of negative mental health outcomes, highlighting the importance of early detection of PLEs.

The duration of untreated psychosis is associated with unfavourable outcomes, including frequent hospitalisation, inadequate response to treatment and limited functional recovery.^{16,17,18,19,20} Early detection and shorter duration of untreated psychosis improve the treatment outcomes of patients with psychotic illness.^17,18 Screening tools for many psychiatric disorders, including psychotic disorders, aid in early diagnosis, which, in turn, may be associated with a better prognosis.²¹ Unfortunately, existing clinician-administered tools used to detect the presence of psychotic features²² are often not suitable for routine clinical practice and population-based epidemiologic surveys. These measures are typically lengthy and require specialised training.²¹ Thus, screening tools that are practical and easy to administer without the need for clinical training may aid in the early diagnosis and reductions in disability and morbidity from psychotic disorders.

In low- and middle-income countries such as South Africa, screening tools administered by laypersons may be particularly beneficial given the limited mental health workforce.^23,24,25,26 Unfortunately, there are no clinician-administered screening tools for psychosis that have been validated in South Africa, layperson- or clinician-administered. To the best of our knowledge, the only South African study was of a self-report psychosis screening tool (Community Assessment of Psychic Experiences), which performed poorly in screening for psychosis.²⁷

To close the gap in research on screening tools for psychosis in South Africa, we examined the psychometric properties of the psychosis screening questionnaire (PSQ).²⁸ The PSQ is a self-reported measure that has been studied primarily in Western settings.^29,30,31 However, to the best of our knowledge, no validation studies of the PSQ as a screening tool for psychosis have been done in the South African population. There is a published study on the cross-cultural examination of the PSQ across Uganda, Ethiopia, Kenya and South Africa.³² However, this study was focused on a broad comparison of the scales’ performance across the four countries to test its equivalence across settings without including the specifics of the PSQ’s performance from each country. Our study is focused on the PSQ as used in South Africa and will examine the measure performance in detail in our setting, including item-level data with item response theory (IRT).

In this study, we sought to evaluate the construct validity of the PSQ (i.e. factor structure) using data collected from a large South African sample, which is part of a more extensive epidemiological research study on the genetics and phenotypic symptoms of neuropsychiatric disorders across four African countries.³³ We also sought to better understand the latent construct of the PSQ in the South African context using IRT analytic approaches. Item response theory models allow for a better understanding of how the PSQ performs in a specific population. The models relate characteristics of items and attributes of individuals to the probability of selecting various responses of an item on a scale.

The Neuropsychiatric Genetics of African Population-Psychosis Study (NeuroGAP-Psychosis) is a case-control, genome-wide association study (GWAS) aiming to advance the understanding of genetic and environmental risk factors of psychotic disorders in Africa, including Ethiopia, Kenya, South Africa and Uganda.³³ Data for the current research project are based on participants from South Africa.

The characteristics of the study participants are summarised in Table 1. The final analytic sample consisted of 2591 participants. The mean age of the participants was 35 years (standard deviation = 11.7) with slightly more female participants (51.6%). Most of the study participants were single (55.4%) and had secondary education (72.4%). Differences in living arrangements and additional details on demographic information for the sample are depicted in Table 1.

Next, we examined the prevalence of psychotic symptoms (Figure 1). Approximately 11% of the study participants reported psychotic experiences, and of those, 49.1% of them experienced psychotic symptoms within the last 12 months. The prevalence of strange experiences was the highest (5.0%), followed by hallucinations (4.1%), paranoia (3.5%) and thought interference (2.2%). Mania was the least endorsed symptom. The prevalence of psychotic experiences was equally distributed among female participants (n = 154; 52.6%) and male participants (n = 139; 47.4%). Prevalence was highest among the middle-aged (50.5%), followed by young adults (45.1%) and older adults (4.4%). The participants in the study used one of the three languages, namely, English (49.1%), Xhosa (44.8%) and Afrikaans (6.1%). The proportion of psychotic experiences varied between people speaking Xhosa, English and Afrikaans as follows: hallucinations (49.5%, 38.1%, 12.4%), paranoia (54.95%, 42.86%, 2.2%), thought interference (55.2%, 54.8%, 0.0%), strange experiences (46.1%, 50.4%, 3.1%) and mania (0.0%, 87.5%, 12.5%), respectively.

The authors conducted a CFA using the unidimensional factor structure to examine the fit and parameter statistics of the PSQ (see Table 2a and Table 2b). The unidimensional model provided a good fit for the data (RMSEA = 0.023; CFI = 0.977; TLI = 0.954), but the mania item showed only a weak association with the underlying latent factor (standardised factor loading [s.e.] = 0.14). Thus, we re-ran the factor analysis without the mania item and observed an improvement in the fit of the model (RMSEA = 0.025; CFI = 0.991; TLI = 0.972). In addition, Table 2 shows the unidimensional model of psychosis for the PSQ in South Africa with strong factor loadings ranging from 0.69 to 0.79 without the mania item.

In this study, we examined the psychometric properties of the PSQ in a large South African sample of controls who did not have a clinical diagnosis of psychosis. The overall lifetime prevalence of psychotic symptoms was 11%, with strange experiences (5%) as the most prevalent psychotic symptom while mania (0.3%) was the least endorsed. The results of the CFA-confirmed items on the PSQ likely comprise one latent factor based on the CFI (0.977) and root mean square error value (0.023). However, the mania item showed a weak association with the underlying latent trait, psychosis. The IRT analysis showed that the PSQ provided high information only at higher levels of the underlying construct, which indicates that the PSQ will help identify individuals with a high level of psychosis compared to individuals with a low level of psychosis, further supporting the construct validity of the PSQ in South Africa.

The findings on prevalence estimates were difficult to compare to prior research because, in South Africa, there is a lack of reliable incidence data on psychotic disorders. In general, the prevalence of psychotic disorders is relatively low at about 1% – 3%^1,2, and sub-Saharan Africa may have even lower rates of psychotic disorders.^39,40,41 However, PLEs are much more common in the general population than psychotic disorders.^9,10 The prevalence of PLEs varies significantly between countries; for instance, estimates range from as low as 0.8% to as high as 31.4%.⁴² There is some evidence from extensive comparative country studies showing that in some African communities, there tends to be a higher prevalence of PLEs.^43,44 However, other large studies have failed to find a higher prevalence of PLEs in African countries.^42,45 But several other studies conducted in different African countries have found a higher prevalence of PLEs in African communities.^42,46,47,48 However, most of these studies were conducted in adolescents and young adults, a group associated with higher rates of PLEs.^49,50 In our study, about 11% of participants had PLEs, which is on the high end compared to many Western studies, but lower than the previously documented South African prevalence of 16% described in a large cross-national study.⁴² The notable variation in PLEs between studies could be because of the difference in the age of study participants, the content of the scales used, the model of data collection (self-report vs. interviewer-administered) and inherent differences across populations.^51,52 Additionally, culture plays a vital role in the experience, understanding and labeling of PLE.^53,54,55

In our study, the endorsement of psychotic symptoms varied depending on the participant’s language; for instance, Xhosa-speaking participants had the highest prevalence of hallucinatory experiences. Of note, the primary language in South Africa often represents race and ethnicity. There is some evidence showing that performance on the individual items of the PSQ varies between ethnic groups.³¹ Also, there is evidence that the content and associated distress of the psychotic symptoms are influenced by the individual’s culture and the society they live in.^55,56 Hence, it may not be surprising to find higher rates of perceptual disturbances among Xhosa-speaking people considering that interacting with ancestors, including receiving messages from them, is an acceptable practice in their culture. Furthermore, the language used to interview participants may influence the results of the screening tests; for instance, evidence shows that people not interviewed in their primary language may be more likely to endorse psychotic features with the PSQ.³⁴ To counteract these language-related effects, all participants in our study were interviewed in their primary language.

The PSQ performed well as a unidimensional construct on the confirmatory analysis. Our study provides further evidence for the weak association of the mania item with the latent trait.³¹ This is not surprising considering that typically with mania, psychosis occurs in the background of a mood disturbance, and it usually consists of grandiose delusions and disordered speech. This contrasts with the odd ideations, thought disorder and paranoia captured by the PSQ items. Additional studies of this nature are needed to confirm our findings, specifically to evaluate the suitability and possible amendment of the mania item on the PSQ scale, especially in the African context.

The CFA and IRT showed that items assessing strange experiences and hallucinations gave the most precise information regarding psychosis as a measured latent trait compared to other items. The perception of the strangeness of experiences may differ between societies cross-culturally. For example, in non-Western countries, people might be more likely to endorse experiences such as feeling the presence of supernatural forces or communicating with the deceased because such experiences may have a higher value and cultural meaning in these communities, which can easily be recorded as strange on the screening scales.^32,57,58,59 However, as shown in our IRT analysis, the PSQ provides useful information about the psychosis construct at higher levels of the latent trait, which should facilitate detecting mainly the clinical levels of psychosis.

To the best of our knowledge, this is the first study to assess the psychometric properties of the PSQ in South Africa. Our findings suggest good construct validity and a one-dimensional structure for the PSQ in South Africa with a non-clinical population. In addition, using the PSQ to screen for psychosis may be better without the mania item. Future studies that examine the criterion validity of the PSQ are warranted.