Hypothesis / aims of study
Screening for psychological factors as part of a biopsychosocial assessment provides an individualised patient-centred approach to care. The Pelvic Pain Psychological Screening Questionnaire (3PSQ) is a new short 15-item tool developed to screen for the presence of 13 psychological factors in individuals with persistent pelvic pain (PPP). The 15 items in the 3PSQ were derived from 9 original validated questionnaires, however, the 3PSQ requires psychometric testing before it can be used as a valid and reliable screening questionnaire. The aim of this study was to assess the convergent and construct validity as well as the reliability of the 3PSQ.
Study design, materials and methods
A quantitative, observational, cross-sectional study was conducted to assess validity and reliability of the 3PSQ.
Participants from around the world with self-reported pelvic pain, with or without a formal diagnosis, were invited to participate in the study. Participants over the age of 18 years who identified as any gender were deemed eligible. Anyone who was pregnant or had experienced childbirth or pelvic surgery in the preceding 6 months was excluded from the study.
Participants were recruited through social media platforms, flyers displayed at various pelvic health clinics and an email circulated directly to clinicians who manage clients with pelvic pain.
Qualtrics (Qualtrics, Provo, UT) was used to develop and conduct an online survey, which commenced with questions relating to participants' demographic details and symptom profile. The 3PSQ was then completed, followed by all 9 original questionnaires, in a randomised order. Questions relating to sexual function within the 3PSQ were optional. Table 1 presents the relevant 3PSQ items, the psychological domains being screened and the original questionnaire from which they were derived.
To determine convergent validity, the association between each Likert scale question of the 3PSQ and the comparative questionnaire scores were examined using linear regression models. Results were summarised with standardised beta (β) coefficients, with target coefficients of ≥0.30 considered the cut-off for demonstrating convergent validity. Positive correlations were hypothesised for all questions except for the negatively phrased self-efficacy and sexual assertiveness questions, which were hypothesised to correlate negatively. A binomial logistic regression model was used to examine the association between the binary 3PSQ question and corresponding questionnaire with score results summarised using Odd ratios (OR) and 95% CIs.
Construct validity was tested by running a Principal Components Analysis (PCA) on 13 out of 15 items of the 3PSQ. Two items related to sexual function were excluded as they were unforced responses on the survey and contained missing data. The eigenvalue-one (Kaiser) criterion was used to retain components within the PCA (components with eigenvalues >1.0 were retained).
Cronbach’s alpha was used to assess internal consistency of the clusters of items loaded onto each retained component. An alpha value of 0.7 was considered acceptable while >0.8 was considered good.
G*Power 126.96.36.199 was used for a priori power calculations. It was indicated that 485 participants were required to detect a small effect size, with 80% power, in a linear regression model. Sample size for the PCA was deemed appropriate for a subject-to-item ratio of 20:1. With 15 items in the 3PSQ, a minimum of 300 participants was required. Therefore, a sample of at least 485 was targeted and p-values <0.05 were considered statistically significant.
1052 participants commenced the survey, with 761 (84.4%) completing the 3PSQ. The total sample population (n=789) had a mean age of 39.1 years (range 18-76 years) and standard deviation (SD) of 13.3. Majority of participants (n=737) were female (93.4%), with 48 male participants (6.1%) and 4 who identified as transgender, gender variant, or non-binary (0.5%).
All fourteen linear regression models were in the hypothesised direction and reached the target coefficient of >0.30 (standardised β 0.32-0.73). The logistic regression indicated a positive relationship between the binary 3PSQ question and the comparative questionnaire (OR 1.049). Linear and logistic regression models showed statistically significant associations for all questions with p<0.001 (Table 1). Items 1-13 underwent PCA, revealing two constructs with eigenvalues >1.0. Items loading highest on each component were deemed representative of that component. Component 1 was interpreted as representing pain-related psychological factors and component 2 as representing global psychological factors. Component 1 demonstrated ‘good’ internal consistency and component 2 demonstrated ‘acceptable’ internal consistency (Cronbach’s α = 0.890, Cronbach’s α = 0.773 respectively). Further inspection of the data indicated removal of item 13 (trauma) produced a higher Cronbach’s alpha for component 2 cluster (Cronbach α = 0.810).
Interpretation of results
The results from this study demonstrate that the 3PSQ is psychometrically robust and is representative of the nine original questionnaires it was derived from. These findings support the use of the 3PSQ as a clinically useful tool to screen for psychological factors in people with pelvic pain.
All associations within the regression models were in the hypothesised direction and were of sufficient magnitude to demonstrate convergent validity for all 15 items within the 3PSQ. The PCA showed that items 1-13 represented two components, with the Cronbach’s alpha being good and acceptable on each component, illustrating adequate internal consistency. The Cronbach’s alpha improved with item 13 (trauma) removed, indicating it was statistically redundant. However, trauma identification and trauma-informed care is essential for people with PPP and previous research has identified trauma as being a contributing factor to complex pelvic pain presentations (1). This question was considered clinically important and therefore was retained.
The selection of comparative questionnaires may have also impacted on the degree of correlation. All comparative questionnaires have demonstrated reliability and validity, however not all within the persistent pain population, and none within a pelvic pain population. Of the 3PSQ items with the highest correlations, item 10 (helplessness) and item 7 (rumination) were both compared to questionnaires that had been validated in a persistent pain population, whilst item 4 (depression) was compared to a questionnaire which was not validated in the persisten pain population. However, items with the lowest correlations were compared to questionnaires not validated in a persistent pain population. It therefore appears questions were less likely to be well correlated if they were compared to a questionnaire not validated in the persistent pain population.
Interestingly, this study reported lower rates of diagnosed anxiety (26.7%) and depression (24.8%) in participants with PPP when compared to wider literature. This may suggest that participants with diagnosed mental health conditions were underrepresented in our sample. However, there were potentially a number of participants with undiagnosed mental health conditions, given that 37.8% and 31.2% scored ≥ 3 for the anxiety and depression subscales of the Pelvic Health Questionnaire (PHQ) respectively, suggesting presence of anxiety or depressive disorders. Notably participants completed the survey during a global pandemic (COVID-19), which may have limited access to healthcare to obtain a mental health diagnosis or contributed to a higher level of psychological distress.