Background & Aims

Approximately 11% of the population have chronic pain which is considered high impact and has wide-ranging consequences on peoples’ lives1. Research is needed to understand the factors that contribute to high impact chronic pain so that targets for intervention can be implemented. The 2019 UK Biobank (UKB) experience of pain (EOP) survey2 collected information from 167,184 participants, including health status and outcomes, and the presence, duration, severity, and location of pain. This dataset is a uniquely rich and large resource for potential pain research. However, due to a coding error in the EOP survey, participants with chronic pain all over the body (here called widespread pain) did not answer questions about the impact of their pain. This limits the generalisability of pain impact-based analyses in the EOP survey. The aim of this study was to investigate the suitability of the EuroQol 5 Dimension 5 Level index (EQ-5D-5L)3 as a measure of pain impact in the EOP survey.

Methods

94,048 (56.3%) EOP participants had chronic pain. 83,219 (89.5%) did not have widespread pain and completed the Brief Pain Inventory (BPI) interference scale4 and the EQ-5D-5L, and 9878 (10.5%) had widespread pain and completed the EQ-5D-5L only. EQ-5D-5L items were weighted according to the official guidelines for England5 and combined into a single index (Range = -0.281 – 1.000). For those without widespread pain, severe pain interference was defined as a mean BPI score over 46. Pearson’s correlation coefficient was calculated for the mean BPI score and EQ-5D-5L index. Receiver operating characteristic curves, area under curve (AUC), Youden’s index, sensitivity and specificity assessed the ability of the EQ-5D-5L index to predict severe pain interference and to choose the optimal EQ-5D-5L cut-off. For all participants, the ability of the EQ-5D-5L index to discriminate between persons with different pain conditions was examined visually. Analysis was performed in R version 4.1.1.

Results

The mean BPI interference score and EQ-5D-5L were significantly negatively correlated (r = -0.695, 95% CI: -0.698 – -0.691, p < 0.001). The EQ-5D-5L had very good predictive ability (AUC = 0.873, 95% CI: 0.869 – 0.876) when used to estimate the mean BPI score cut-off for severe pain interference. The optimal cut-off for predicting severe pain interference using the EQ-5D-5L was 0.809 (95% CI: 0.808 – 0.821), determined by the maximum Youden’s index (J = 0.605, 95% CI: 0.598 – 0.613). The EQ-5D-5L at this cut-off had a sensitivity of 78.29% (95% CI: 77.75 – 82.76%) and specificity of 82.18% (95% CI: 77.87 – 82.44%). In 94,048 EOP participants with chronic pain, the EQ-5D-5L discriminated between those with and without widespread pain, those with chronic pain in different locations, and those with different numbers of pain locations. In all 167,184 participants, the EQ-5D-5L discriminated between EOP participants with no pain, acute pain, and chronic pain.

Conclusions

We present a solution for defining the impact of pain in the UK Biobank EOP survey using the EQ-5D-5L index as an alternative to the BPI interference scale. The EQ-5D-5L cut-off to differentiate severe pain interference defined by the BPI had very good overall performance in this population according to common measures of diagnostic accuracy. Imputation of ordinal data that is missing not at random is technically possible when some data in the target population are present7. However, the missing EOP BPI data did not fit this criterion. Using the EQ-5D-5L index is advantageous over imputation as it does not present technical challenges, rely on assumptions, or increase the uncertainty of the data. Our future research will validate these results in external cohorts. Our research will enable researchers to define pain impact in people with widespread pain in the EOP survey and potentially in other datasets where EQ-5D-5L items are available but other measures of pain impact are not.

References

1.Von Korff M, DeBar LL, Krebs EE, et al. Pain 2020;161(3):651-661
2.UK Biobank. 2022 Mar. Available from: https://biobank.ndph.ox.ac.uk/showcase/ukb/docs/pain_questionnaire.pdf
3.Herdman M, Gudex C, Lloyd A, at al. Qual Life Res. 2011 Dec;20(10):1727-36
4.Cleeland CS, Ryan KM. Ann Acad Med Singap. 1994 Mar;23(2):129-38
5.Devlin NJ, Shah KK, Feng Y, et al. Health Econ. 2018 Jan;27(1):7-22
6.Shi Q, Mendoza TR, Dueck AC, et al. Pain. 2017 Jun;158(6):1108-1112
7.Hammon A. AStA Adv Stat Anal. 2023 (107), 671–692

Presenting Author

Charlotte Woolley

Poster Authors

Charlotte Woolley, PhD

BSc (Hons), MSc, PhD

Centre for Epidemiology Versus Arthritis, University of Manchester

Lead Author

John McBeth

MA (Hons)

Centre for Epidemiology Versus Arthritis, University of Manchester

Lead Author

Danielle Van der Windt

Keele University

Lead Author

Annick De Paepe

Ghent University

Lead Author

Barbara Nicholl

BSc (Hons)

University of Glasgow

Lead Author

Emma Fisher

PhD

University of Bath

Lead Author

Emma Parry

MRCGP

Primary Care Centre Versus Arthritis, School of Medicine, Keele University

Lead Author

Gary Macfarlane

MBChB

University of Aberdeen

Lead Author

Geert Crombez

Ghent University

Lead Author

Taylor Robert

Research User Group, Primary Care Centre Versus Arthritis, Keele University

Lead Author

Chris Eccleston

PhD

University of Bath

Lead Author

Marcus Beasley

BSc (Hons)

Institute of Applied Health Sciences, School of Medicine, University of Aberdeen

Lead Author

Milica Blagojevic-Bucknall

PhD

Primary Care Centre Versus Arthritis, School of Medicine, Keele University

Lead Author

Edmund Keogh

PhD

University of Bath

Lead Author

Payam Amani

PhD

Primary Care Centre Versus Arthritis, School of Medicine, Keele University

Lead Author

Anna Gibby

University of Bath

Lead Author

Laura Lisboa

PhD

Institute for Mathematical Innovation, University of Bath

Lead Author

Beate Ehrhardt

PhD

Institute for Mathematical Innovation, University of Bath

Lead Author

Matthew Nunes

PhD

School of Mathematical Sciences, University of Bath

Lead Author

Topics

  • Epidemiology