A confirmatory factor analysis of an electronic format painDETECT questionnaire for patients with low back pain

Abstract Background The substantial burden of low back pain on patients and healthcare systems is exacerbated by unclear pathology and ineffective diagnostic methods, hindering effective management. The painDETECT questionnaire (PD-Q) has been used to facilitate the evaluation and categorization of low back pain. While preliminary validation and translations of the paper-based format of PD-Q into languages such as Spanish and Dutch have been accomplished, the underlying factor model inherent to the electronic format of the PD-Q remains to be established. Objective The objective of this study was to utilise confirmatory factor analysis (CFA) to investigate the factor structure of an electronic format PD-Q among patients with neuropathic low back pain. Methods This cross-sectional study was conducted at a Spinal Clinic in Sydney between November 2020 and October 2022. Eligible participants were adults over 18 with low back pain and no history of lumbar surgery or systemic co-morbidities. Participants completed the electronic format of the PD-Q, and CFA was employed to assess the validity of the suggested two-factor, nine-item structure. Recommended cut-offs for goodness-of-fit indices were used to evaluate the model fit. Results Of the 236 patients that visited the clinic during the data collection period, 142 (71, 50% female, mean age 51.26 ± 15.28 years) participated in the study. Median pain severity was 9/10 over 4 weeks. CFA indicated strong model fit, with goodness-of-fit and comparative fit indices over 0.9, and overall internal consistency was 0.77. Construct validity analysis demonstrated the PD-Q’s effectiveness in distinguishing neuropathic, mixed, and nociceptive LBP, aiding neuropathic pain evaluation in low back pain patients. Conclusion This study confirms the reliability and two-factor structure of the electronic PD-Q for neuropathic pain assessment in low back pain patients. To enhance comprehension of the clinical applicability of the electronic format PD-Q, future research should conduct clinimetric evaluations.


Introduction
Low back pain (LBP) has been a significant musculoskeletal disorder and the leading contributor to years lived with disability worldwide since 1990 1 .LBP is typically characterized as pain, muscle tension, or stiffness in the region between the 12th rib and the gluteal folds 2 .However, identifying the root cause of LBP may require extensive clinical investigations, as some of the associated changes can be subtle and multifactorial.It is crucial to use validated and universally accepted outcome assessment and screening tools to ensure that clinical diagnoses are accurate and credible.Consequently, there has been a surge in clinical research dedicated to studying LBP and uncovering the root causes of its symptoms.
Evidence suggests that challenges in diagnosing LBP may arise from the presence of mixed pain states involving both neuropathic and nociceptive pain 3 .Consequently, increased clinical research has focused on understanding the underlying mechanisms of LBP symptoms 4 .Since differentiating neuropathic pain from nonneuropathic pain has direct implications for treatment, clinicians need to identify the severity of distinct pain aspects in each patient.Treating pain mechanisms specifically, rather than as a uniform phenomenon, allows tailoring of therapies to specific subgroups of patients.
Internationally, health professionals use the painDETECT questionnaire (PD-Q) as a screening tool for characterizing and quantifying low back pain 5,6 .The PD-Q consists of five sections, including a pain intensity scale of 0-10, graphs depicting the patient's pain course pattern, a body chart to indicate the areas of pain, the presence of radiating pain, and seven pain quality rating questions addressing the extent of neuropathic pain on a 0-5 scale (Supplementary Appendix 1).Rather than treating LBP as a dichotomous pathology, the PD-Q considers chronic pain as a spectrum with pure neuropathic pain on one end and pure nociceptive pain on the other.To estimate the neuropathic component of a patient's LBP, scores from the nine PD-Q items are summed to obtain an overall PD-Q score ranging from −1 to 38.Nociceptive dimension is unlikely for scores � 12, scores between 13 and 18 suggest possible neuropathic pain, and scores � 19 indicate mixed pain.
The PD-Q, originally developed and validated in Germany 7 , has been adapted to various languages worldwide [8][9][10] .During its initial development, rigorous testing was conducted, including item analysis, reliability assessment, and validity testing.The results indicated satisfactory internal consistency and a four-factor structure, with two factors considered determinative.Further work to validate the PD-Q, has resulted in inconsistent results.While a few studies have supported a two-factor model consistent with the original structure of the PD-Q 11,12 , divergent findings have emerged.For example, De Andr� es et al. 13 proposed an alternative four-factor structure, which challenges the primacy of the original subscales.However, such discrepancies do not inherently undermine the validity of the established two-factor structure of the PD-Q.In fact, they underscore the heterogeneity encompassed within prior research endeavours.Validation studies have primarily focused on distinct aspects, including cross-cultural adaptation with limited psychometric evaluation, patient cohorts marked by pain settings differing from low back pain, or populations characterized by anatomical heterogeneity.
Our ongoing research aims to augment the existing body of evidence by assessing the utility of the PD-Q specifically within the context of patients with LBP, aiming to provide valuable insights into its factor structure and validity in this specific population.The Spine Service Clinic in Sydney, Australia, adopted a paper-based English version of the PD-Q in 2015.Recognizing the increasing emphasis on electronic medical records and big data in healthcare, as part of ongoing research efforts to develop evidence-based indicators for LBP, the clinic has started exploring an electronic format of the PD-Q.Preliminary evaluation of this electronic version has revealed its convenience for patients, allowing for portability, real-time data recording, and immediate scoring.
Therefore, building upon the existing groundwork, the aim of this study was to improve the clinical utility of the PD-Q by establishing a clear and robust factor structure through confirmatory factor analysis (CFA).The objective was to accurately assess the distress experienced by patients with neuropathic LBP.Utilising CFA provides a statistical technique that systematically examines the relationships between the observed variables and latent factors within the PD-Q, offering valuable insights into the underlying mechanisms of low back pain.Establishing a clear and robust factor structure for the electronic PD-Q will further facilitate its integration into electronic medical records and leverage its real-time data collection and analysis advantages.

Study design
This study utilised a cross-sectional design.

Setting
The study was conducted at the Spine Service Clinic at the St George Hospital Campus in Kogarah, Sydney.Data collection took place between November 2020 and October 2022.

Participants
Participants in the study were individuals aged 18 years and above who presented to the Clinic with LBP during the study period.The inclusion criteria required participants to be free of lower limb pain and have no prior history of lumbar spine surgery or systemic co-morbidities that could impact the patient's pain experience.Additionally, participants were required to possess the ability to operate a computer, smartphone, or other electronic devices such as a laptop or iPad V R .Conversely, participants who could not use electronic devices were excluded from the study to ensure that all participants could complete the electronic version of the PD-Q (Supplementary Appendix S1), a key component of the study.Approval to adapt and use the electronic version of the PD-Q was granted by the original developers of the questionnaire 7 .

Sample size
A sample of approximately 140-210 patients was anticipated based on an estimate of an average of 10-15 patients presenting with LBP at the Spine Service Clinic each month.Our previous work guided this estimation, suggesting a minimum sample size minimum sample size of 120 participants 14 .Following approval by the Institutional Ethics Committee, of the 236 patients who visited the clinic during the data collection period, a convenience sample of 142 participants were enrolled.

Data collection
All data were collected by administrative staff and a research nurse who provided them access to a secure web-based data management tool, Research Electronic Data Capture REDCap 15 .Completing a paper-based form of the PD-Q was mandatory as part of the routine assessments and care of patients presenting at the Clinic for back pain.However, for those participants who expressed a willingness to participate in the study, completion of the electronic format of the PD-Q was required.All patients who met the inclusion criteria were provided with a Participant Information Statement and a link to the PD-Q questionnaire completed via REDCap.

Ethical considerations
All procedures were approved by the University of Wollongong Human Research Ethics Committee (HREC) Approval number 2020/329.

Data analysis
The present study utilised data directly imported from the web-based data collection system, REDCap, along with the Statistical Package for the Social Sciences Statistical Software package version 22.0 for Windows and AMOS version 22.0 16 for confirmatory factor analysis (CFA).Missing data were minimal, with only 12 missing values (0.93%) replaced by the series mean method.Descriptive statistics were used to summarize the data, including means, standard deviations, and frequencies (percent).For continuous variables, the significance of the differences was analyzed using the one-way analysis of variance (ANOVA) and independent samples ttest.The level of statistical significance was set at p < 0.05.
Several sequential steps were undertaken to evaluate the PD-Q's psychometric properties, considering various standard statistical criteria.Initially, a series of maximum likelihood CFA was conducted to investigate the 9-item two-factor model.Throughout this iterative simulation process, items with low factor coefficients were systematically re-examined using multiple indices of goodness of fit, statistical significance testing of factor loadings, correlations among factors, and computation of confidence intervals for the model.Following the CFA approach, the first item in each subscale was fixed and initially set to 1.00.
In order to evaluate the adequacy of the proposed factor structure and its overall fit to the data, the following criteria were employed to determine a reasonably good model fit: a Goodness of Fit index (GFI) of at least 0.926, a Root Mean Square Error of Approximations (RMSEA) of no more than 0.06, a Comparative Fit index (CFI) of at least 0.95, and a standardized root mean residual (SRMR) of no more than 0.06 17,18 .These estimates were utilized to assess the extent to which the proposed factor structure and its overall fit to the data.

Participants characteristics
A total of 142 patients, with a mean age of 51.26 (SD ¼ 15.28) years, actively participated in the study.The participant distribution across the three groups was approximately balanced, encompassing 37 (26.06%) in the Neuropathic group, 40 (28.17%) in the Mixed group, and 65 (45.77%) in the Nociceptive group.The investigation delved into various aspects, including current pain levels, the most intense pain experienced in the past 4 weeks, and average pain during that period.A comprehensive depiction of patient attributes can be found in Table 1.

Confirmatory factor analysis of the PDQ
Confirmatory factor analysis demonstrated a two-factor model (Figure 1), with factor 1 having seven contributing items and factor 2 having two items.

Reliability analysis
Reliability analysis suggested that the overall internal consistency reliability coefficient for the nine items of the PD-Q was 0.82.These results suggested that the two-factor model was the best fit for the data.Taken together the above results suggest that the 9-item two-factor model of the PDQ offers the best fit for this sample of patients with LBP.

Discussion
The results of CFA reported a two-factor PD-Q model consistent with previous studies' proposed structure.Specifically, the original developer 7 and Matsubayashi et al. 12 , who conducted the Japanese translation of the PD-Q, previously proposed findings suggesting two factors.They categorized all Likert-type items under one factor, while questions related to pain radiation and the course of pain were placed under another factor.The structure was confirmed by significant results in GFI, RMSEA, CFI, and Cronbach alpha coefficients demonstrating sufficient construct validity and high internal consistency of the questionnaire.The values for the indices obtained in the current study are considered to be sufficient by Hu and Bentler's 19 two-index reporting, and consistent with the principles of structural equation modeling 18 .Despite the value falling below the standard range, the CMIN/DF is known to be influenced by low sample size, leading to the rejection of the model even if it accurately describes the data 20 .
The two-factor structure found in the present study is meaningful and useful for summarising and interpreting the questionnaire.All Likert-type items and pain radiation were under one factor, while two items related to the course of pain were under another.More specifically, Factor 1, comprised of the seven items, appears to measure various sensory aspects associated with neuropathic LBP.These include burning sensations, tingling, numbness, sensitivity to touch and pressure, and the occurrence of sudden pain attacks.These items collectively capture the unique sensations and experiences often linked with neuropathic back pain and identified as Neuropathic Sensations and Discomfort.Factor 2, titled Radiation and Distribution of Pain, comprises two items that seem to measure the spread and radiation of pain in the context of back pain.Items in this factor relate to the distribution of pain to other areas of the body and ask about CURRENT MEDICAL RESEARCH AND OPINION the course of pain, which indicates the extent to which the pain radiates and affects different regions beyond the marked area.This factor encompasses the broader characteristics of pain propagation characteristics associated with neuropathic and nociceptive LBP.
Clinicians have predominantly relied on generic unidimensional pain assessment tools such as the Visual Analog Scale (VAS), Numeric Rating Scale (NRS) 21 , and the presence of radiating leg pain to identify and diagnose neuropathic back pain.However, misdiagnosis and suboptimal treatment outcomes often result from this approach.The identification and confirmation of Factor 2 in this study is, therefore, significant.This factor includes two items that facilitate the assessment of neuropathic pain on a spectrum.This underscores the potential utility of the PD-Q in categorising and assessing differentiating neuropathic pain from nociceptive pain or mixed pain.More specifically, recognising the nuanced nature of LBP as a spectrum with varying degrees of neuropathic, nociceptive, or mixed components enables more accurate diagnoses.Consequently, this shift will lead to more effective treatment options targeting specific pathological mechanisms and ultimately improving patient outcomes.
The results of our study indicate that patients with neuropathic back pain consistently reported higher current pain intensity, more severe pain and higher duration of symptoms compared to those with mixed and nociceptive pain.This pattern aligns with the typical characteristics of neuropathic back pain, which is often described as debilitating and involves higher and more intense pain sensations than other types of back pain.Possible explanations for this relationship could be linked to neuropathic pain pathology, including damage or dysfunction in the nervous system, often described as shooting, burning, tingling, or electric shock-like sensations 6 .There is an emerging body of evidence suggesting that a large proportion of patients who complain of LBP suffer from predominantly neuropathic pain 22 .
Alternatively, this finding could reflect the challenges associated with diagnosing neuropathic pain, resulting in less effective treatment options.If this were the case, it would highlight a significant deficiency in our ability to adequately treat patients with neuropathic LBP.Broader recognition of neuropathic pain as a major health problem would allow for further research into the condition and facilitate the development of appropriate management techniques.
Lastly but equally important, the PD-Q has the advantage of being self-reported, distinguishing it from other tools such as diagnostic questionnaires and medical assessments that rely on clinician completion.The electronic format of the PD-Q has the potential to empower patients by enabling them to actively contribute to their own care through independently reporting their back pain experiences and symptoms.Moreover, given the considerable effort put into harnessing electronic medical records for generating comprehensive data, the electronic format of the PD-Q stands out for its portability and capacity for real-time data recording and immediate scoring.This convenience is pivotal for patients, and incorporating automated data entry enhances efficiency while minimizing inaccuracies linked to human error 23 .Simultaneously, this electronic format aids clinicians in diagnosing and addressing LBP, enabling tailored therapies that align with the specific pain mechanism at play.The collaboration between patients and clinicians, supported by the userfriendly approach of the electronic PD-Q, underscores its potential role as a valuable instrument that enhances both the diagnostic process and treatment outcomes.
The study had limitations in terms of sample size, although the sample size was comparable to that of other studies investigating the factor structure of diagnostic instruments in clinical settings.Furthermore, it was constrained by the recruitment of participants from a single center, indicating the potential for improvement in subsequent studies or analyses by incorporating a more diverse recruitment strategy.In this present study, the conclusion drawn is that only two dimensions of the PDQ are valid for this particular sample.However, it is crucial to acknowledge that other studies have demonstrated a different structure.The observed variations in findings could potentially stem from disparities in the demographics of the study populations.

Figure 1 .
Figure 1.Two-factor model of the electronic format painDETECT questionnaire.