Challenge of missing data in observational studies: investigating cross-sectional imputation methods for assessing disease activity in axial spondyloarthritis

Stylianos Georgiadis; Marion Pons; Simon Rasmussen; Merete Lund Hetland; Louise Linde; Daniela di Giuseppe; Brigitte Michelsen; Johan K. Wallman; Tor Olofsson; Jakub Zavada; Bente Glintborg; Anne G. Loft; Catalin Codreanu; Daniel Melim; Diogo Almeida; Sella Aarrestad Provan; Tore K. Kvien; Vappu Rantalaiho; Ritva Peltomaa; Bjorn Gudbjornsson; Olafur Palsson; Ovidiu Rotariu; Ross MacDonald; Ziga Rotar; Katja Perdan Pirkmajer; Karin Lass; Florenzo Iannone; Adrian Ciurea; Mikkel Østergaard; L. M. Ørnbjerg

doi:10.1136/rmdopen-2024-004844

Challenge of missing data in observational studies: investigating cross-sectional imputation methods for assessing disease activity in axial spondyloarthritis

Stylianos Georgiadis^*, Marion Pons, Simon Rasmussen, Merete Lund Hetland, Louise Linde, Daniela di Giuseppe, Brigitte Michelsen, Johan K. Wallman, Tor Olofsson, Jakub Zavada, Bente Glintborg, Anne G. Loft, Catalin Codreanu, Daniel Melim, Diogo Almeida, Sella Aarrestad Provan, Tore K. Kvien, Vappu Rantalaiho, Ritva Peltomaa, Bjorn GudbjornssonOlafur Palsson, Ovidiu Rotariu, Ross MacDonald, Ziga Rotar, Katja Perdan Pirkmajer, Karin Lass, Florenzo Iannone, Adrian Ciurea, Mikkel Østergaard, L. M. Ørnbjerg

^*Corresponding author for this work

Department of Clinical Medicine - The Section for Rheumatology

Research output: Contribution to journal/Conference contribution in journal/Contribution to newspaper › Journal article › Research › peer-review

Abstract

Objectives We aimed to compare various methods for imputing disease activity in longitudinally collected observational data of patients with axial spondyloarthritis (axSpA). Methods We conducted a simulation study on data from 8583 axSpA patients from ten European registries. Disease activity was assessed by the Axial Spondyloarthritis Disease Activity Score (ASDAS) and the corresponding low disease activity (LDA; ASDAS<2.1) state at baseline, 6 and 12 months. We focused on cross-sectional methods which impute missing values of an individual at a particular time point based on the available information from other individuals at that time point. We applied nine single and five multiple imputation methods, covering mean, regression and hot deck methods. The performance of each imputation method was evaluated via relative bias and coverage of 95% confidence intervals for the mean ASDAS and the derived proportion of patients in LDA. Results Hot deck imputation methods outperformed mean and regression methods, particularly when assessing LDA. Multiple imputation procedures provided better coverage than the corresponding single imputation ones. However, none of the evaluated methods produced unbiased estimates with adequate coverage across all time points, with performance for missing baseline data being worse than for missing follow-up data. Predictive mean and weighted predictive mean hot deck imputation procedures consistently provided results with low bias Conclusions This study contributes to the available methods for imputing disease activity in observational research. Hot deck imputation using predictive mean matching exhibited the highest robustness and is thus our suggested approach.

Original language	English
Article number	e004844
Journal	RMD Open
Volume	11
Issue	1
ISSN	2056-5933
DOIs	https://doi.org/10.1136/rmdopen-2024-004844
Publication status	Published - Feb 2025

Keywords

Axial Spondyloarthritis
Epidemiology
Interleukin-17
Tumour Necrosis Factor Inhibitors

Access to Document

10.1136/rmdopen-2024-004844

Cite this

Georgiadis, S., Pons, M., Rasmussen, S., Hetland, M. L., Linde, L., di Giuseppe, D., Michelsen, B., Wallman, J. K., Olofsson, T., Zavada, J., Glintborg, B., Loft, A. G., Codreanu, C., Melim, D., Almeida, D., Provan, S. A., Kvien, T. K., Rantalaiho, V., Peltomaa, R., ... Ørnbjerg, L. M. (2025). Challenge of missing data in observational studies: investigating cross-sectional imputation methods for assessing disease activity in axial spondyloarthritis. RMD Open, 11(1), Article e004844. https://doi.org/10.1136/rmdopen-2024-004844

@article{6e75df3574c8458ebc6a73a9680320bf,

title = "Challenge of missing data in observational studies: investigating cross-sectional imputation methods for assessing disease activity in axial spondyloarthritis",

abstract = "Objectives We aimed to compare various methods for imputing disease activity in longitudinally collected observational data of patients with axial spondyloarthritis (axSpA). Methods We conducted a simulation study on data from 8583 axSpA patients from ten European registries. Disease activity was assessed by the Axial Spondyloarthritis Disease Activity Score (ASDAS) and the corresponding low disease activity (LDA; ASDAS<2.1) state at baseline, 6 and 12 months. We focused on cross-sectional methods which impute missing values of an individual at a particular time point based on the available information from other individuals at that time point. We applied nine single and five multiple imputation methods, covering mean, regression and hot deck methods. The performance of each imputation method was evaluated via relative bias and coverage of 95% confidence intervals for the mean ASDAS and the derived proportion of patients in LDA. Results Hot deck imputation methods outperformed mean and regression methods, particularly when assessing LDA. Multiple imputation procedures provided better coverage than the corresponding single imputation ones. However, none of the evaluated methods produced unbiased estimates with adequate coverage across all time points, with performance for missing baseline data being worse than for missing follow-up data. Predictive mean and weighted predictive mean hot deck imputation procedures consistently provided results with low bias Conclusions This study contributes to the available methods for imputing disease activity in observational research. Hot deck imputation using predictive mean matching exhibited the highest robustness and is thus our suggested approach.",

keywords = "Axial Spondyloarthritis, Epidemiology, Interleukin-17, Tumour Necrosis Factor Inhibitors",

author = "Stylianos Georgiadis and Marion Pons and Simon Rasmussen and Hetland, {Merete Lund} and Louise Linde and {di Giuseppe}, Daniela and Brigitte Michelsen and Wallman, {Johan K.} and Tor Olofsson and Jakub Zavada and Bente Glintborg and Loft, {Anne G.} and Catalin Codreanu and Daniel Melim and Diogo Almeida and Provan, {Sella Aarrestad} and Kvien, {Tore K.} and Vappu Rantalaiho and Ritva Peltomaa and Bjorn Gudbjornsson and Olafur Palsson and Ovidiu Rotariu and Ross MacDonald and Ziga Rotar and Pirkmajer, {Katja Perdan} and Karin Lass and Florenzo Iannone and Adrian Ciurea and Mikkel {\O}stergaard and {\O}rnbjerg, {L. M.}",

note = "Publisher Copyright: {\textcopyright} Author(s) (or their employer(s)) 2025.",

year = "2025",

month = feb,

doi = "10.1136/rmdopen-2024-004844",

language = "English",

volume = "11",

journal = "RMD Open",

issn = "2056-5933",

publisher = "BMJ Publishing Group",

number = "1",

}

Georgiadis, S, Pons, M, Rasmussen, S, Hetland, ML, Linde, L, di Giuseppe, D, Michelsen, B, Wallman, JK, Olofsson, T, Zavada, J, Glintborg, B, Loft, AG, Codreanu, C, Melim, D, Almeida, D, Provan, SA, Kvien, TK, Rantalaiho, V, Peltomaa, R, Gudbjornsson, B, Palsson, O, Rotariu, O, MacDonald, R, Rotar, Z, Pirkmajer, KP, Lass, K, Iannone, F, Ciurea, A, Østergaard, M & Ørnbjerg, LM 2025, 'Challenge of missing data in observational studies: investigating cross-sectional imputation methods for assessing disease activity in axial spondyloarthritis', RMD Open, vol. 11, no. 1, e004844. https://doi.org/10.1136/rmdopen-2024-004844

Challenge of missing data in observational studies: investigating cross-sectional imputation methods for assessing disease activity in axial spondyloarthritis. / Georgiadis, Stylianos; Pons, Marion; Rasmussen, Simon et al.
In: RMD Open, Vol. 11, No. 1, e004844, 02.2025.

Research output: Contribution to journal/Conference contribution in journal/Contribution to newspaper › Journal article › Research › peer-review

TY - JOUR

T1 - Challenge of missing data in observational studies

T2 - investigating cross-sectional imputation methods for assessing disease activity in axial spondyloarthritis

AU - Georgiadis, Stylianos

AU - Pons, Marion

AU - Rasmussen, Simon

AU - Hetland, Merete Lund

AU - Linde, Louise

AU - di Giuseppe, Daniela

AU - Michelsen, Brigitte

AU - Wallman, Johan K.

AU - Olofsson, Tor

AU - Zavada, Jakub

AU - Glintborg, Bente

AU - Loft, Anne G.

AU - Codreanu, Catalin

AU - Melim, Daniel

AU - Almeida, Diogo

AU - Provan, Sella Aarrestad

AU - Kvien, Tore K.

AU - Rantalaiho, Vappu

AU - Peltomaa, Ritva

AU - Gudbjornsson, Bjorn

AU - Palsson, Olafur

AU - Rotariu, Ovidiu

AU - MacDonald, Ross

AU - Rotar, Ziga

AU - Pirkmajer, Katja Perdan

AU - Lass, Karin

AU - Iannone, Florenzo

AU - Ciurea, Adrian

AU - Østergaard, Mikkel

AU - Ørnbjerg, L. M.

PY - 2025/2

Y1 - 2025/2

N2 - Objectives We aimed to compare various methods for imputing disease activity in longitudinally collected observational data of patients with axial spondyloarthritis (axSpA). Methods We conducted a simulation study on data from 8583 axSpA patients from ten European registries. Disease activity was assessed by the Axial Spondyloarthritis Disease Activity Score (ASDAS) and the corresponding low disease activity (LDA; ASDAS<2.1) state at baseline, 6 and 12 months. We focused on cross-sectional methods which impute missing values of an individual at a particular time point based on the available information from other individuals at that time point. We applied nine single and five multiple imputation methods, covering mean, regression and hot deck methods. The performance of each imputation method was evaluated via relative bias and coverage of 95% confidence intervals for the mean ASDAS and the derived proportion of patients in LDA. Results Hot deck imputation methods outperformed mean and regression methods, particularly when assessing LDA. Multiple imputation procedures provided better coverage than the corresponding single imputation ones. However, none of the evaluated methods produced unbiased estimates with adequate coverage across all time points, with performance for missing baseline data being worse than for missing follow-up data. Predictive mean and weighted predictive mean hot deck imputation procedures consistently provided results with low bias Conclusions This study contributes to the available methods for imputing disease activity in observational research. Hot deck imputation using predictive mean matching exhibited the highest robustness and is thus our suggested approach.

AB - Objectives We aimed to compare various methods for imputing disease activity in longitudinally collected observational data of patients with axial spondyloarthritis (axSpA). Methods We conducted a simulation study on data from 8583 axSpA patients from ten European registries. Disease activity was assessed by the Axial Spondyloarthritis Disease Activity Score (ASDAS) and the corresponding low disease activity (LDA; ASDAS<2.1) state at baseline, 6 and 12 months. We focused on cross-sectional methods which impute missing values of an individual at a particular time point based on the available information from other individuals at that time point. We applied nine single and five multiple imputation methods, covering mean, regression and hot deck methods. The performance of each imputation method was evaluated via relative bias and coverage of 95% confidence intervals for the mean ASDAS and the derived proportion of patients in LDA. Results Hot deck imputation methods outperformed mean and regression methods, particularly when assessing LDA. Multiple imputation procedures provided better coverage than the corresponding single imputation ones. However, none of the evaluated methods produced unbiased estimates with adequate coverage across all time points, with performance for missing baseline data being worse than for missing follow-up data. Predictive mean and weighted predictive mean hot deck imputation procedures consistently provided results with low bias Conclusions This study contributes to the available methods for imputing disease activity in observational research. Hot deck imputation using predictive mean matching exhibited the highest robustness and is thus our suggested approach.

KW - Axial Spondyloarthritis

KW - Epidemiology

KW - Interleukin-17

KW - Tumour Necrosis Factor Inhibitors

UR - http://www.scopus.com/inward/record.url?scp=85218791081&partnerID=8YFLogxK

U2 - 10.1136/rmdopen-2024-004844

DO - 10.1136/rmdopen-2024-004844

M3 - Journal article

C2 - 39979039

AN - SCOPUS:85218791081

SN - 2056-5933

VL - 11

JO - RMD Open

JF - RMD Open

IS - 1

M1 - e004844

ER -

Challenge of missing data in observational studies: investigating cross-sectional imputation methods for assessing disease activity in axial spondyloarthritis

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this