Mental health problems are a significant source of distress for millions of people across the world and are amongst the largest contributors to disability (WHO). Traditionally, the classification of mental health problems has distilled symptoms down into discrete disorder categories, despite common disorder co-occurrence and heterogeneity within disorders. Recent advances in psychiatry have supported a move away from this type of categorisation and towards a more biologically plausible, mechanistic and dimensional understanding of mental health issues (cf. RDoC, Insel & Cuthbert, 2013; HiTOP, Kotov et al 2017).
The move towards a more dimensional approach to psychopathology has spurred novel research questions and methodologies. The general psychopathology (p)-factor was discovered through the attempt to understand the high comorbidity between disorders using questionnaires that assess multiple mental health domains (Lahey et al., 2012; Caspi et al., 2014). Recent research has further linked transdiagnostic factors such as ‘anxious-depression’ and ‘compulsive behaviour and intrusive thought’ to different aspects of cognition including goal-directed behaviour (Gillan et al., 2016), metacognition (Rouault et al., 2018) and aversive learning (Wise & Dolan, 2020). Similarly, transdiagnostic mechanisms such as the ability to adapt to volatile environments have been revealed to be important for internalising psychopathology including depression and anxiety, but not either disorder separately (Gagne et al., 2020). Crucially, this line of research has relied on the exploration of a diverse range of symptomatology, requiring a large number of questionnaires to be used for each study.
A significant disadvantage emerging from this is the huge time and effort cost arising from administering a large questionnaire set. The total number of items are large and routinely surpass 200 items, rendering a large time and effort cost to participants and a large time and financial burden to researchers. Reducing the set of questions, but keeping the specificity and span of the symptomatology represented would be greatly beneficial for mental health researchers and would reduce the cognitive and emotional burden for participants. We propose to conduct a study that will reduce these questionnaire items, outlined below.
This study has a very practical and useful goal, which is to create a short-form questionnaire that reliably captures a range of mental health problems. The end result of this study will therefore produce a useful tool for mental health researchers who want to quickly investigate a wide range of transdiagnostic symptom measures.
The minimum sample size needed is based on factor analysis guidelines for the recommended minimum number of people per total number of items (Cattell, 1978) and previous studies in this field (Gillan et al., 2016). The total number of items from the Gillan et al. (2016) questionnaire set is 233, and the full set is 291. Cattell guidelines (Cattell, 1978) recommend 3-6 times as many participants for the number of variables, so we aimed for the discovery sample to have 1200 participants to ensure the retest sample is large enough after accounting for attrition.
The calculated study costs (including services fees) are given in the table below. We are asking for a total of £9702.20 to complete all stages of the study.
|Sample 1.1 (discovery sample)||1200||45||0.75||£ 6,282.00|
|Sample 1.2 (retest validation, assuming 20% attrition)||960||15||0.25||£ 1,675.20|
|Sample 2 (out-of-sample validation)||1000||15||0.25||£ 1,745.00|
Participants will fill out a number of questionnaires spanning a range of psychiatric problems. The questionnaire set will include the full set of questionnaires reported in Gillan et al. (2016) as well as additional questionnaires included to broaden the span of the psychiatric disorders represented (see Materials).
The first sample of participants (discovery sample, n = 1200) will fill out all questionnaire items. Following the dimension reduction analysis (see analysis plan), the same participants will be invited back to fill out the reduced questionnaire set. We expect an attrition rate of approximately 25/30% (Palan & Shiter, 2018), which provides a sample of n = 960 for the reduced questionnaire set. The second (confirmation) sample of participants (n = 1000) will also complete the reduced set of questionnaires.
The original questionnaire set reported in Gillan et al. (2016) includes measures of alcohol use (AUDIT), apathy (AES), impulsiveness (BIS), eating attitudes (EAT-26), social anxiety (LSAS), OCD (OCI-R), schizotypy (SSMS), anxiety (STAI) and depression (SDS). We will additionally include questionnaires to assess callous and unemotional traits (ICU), ADHD (ASRS), autism spectrum symptoms (AQ), anhedonia (TEPS) and intolerance of uncertainty (IUS).
An exploratory factor analysis will be performed on the discovery sample. We will define the best number of factors using a combination of approaches, including parallel analysis and Cattell’s criterion (Cattell, 1966) parallel analysis, then perform dimension reduction to obtain the key items for each factor. The reduced set will then be validated within the same participant sample. The key validation measure will be the correlation between the original participant factor scores and the predicted factor scores from the reduced questionnaire set.
Out-of-sample validation will be performed on the reduced set of questionnaires. The distributions of the factor scores from the discovery sample will be compared to the distributions obtained from the confirmation sample.
A preprint of the research findings will be posted in an open access repository (PsyArXiv) and published in an open access peer-reviewed journal. All data will be fully anonymised and openly-shared on OSF. All analysis code will be uploaded to Github.
A link to our AsPredicted OSF preregistration can be found here: https://osf.io/gzn86