Assessment of score- and Rasch-based methods for group comparison of longitudinal patient-reported outcomes with intermittent missing data (informative and non-informative)
The purpose of this study was to identify the most adequate strategy for group comparison of longitudinal patient-reported outcomes in the presence of possibly informative intermittent missing data. Models coming from classical test theory (CTT) and item response theory (IRT) were compared.
Two groups of patients’ responses to dichotomous items with three times of assessment were simulated. Different cases were considered: presence or absence of a group effect and/or a time effect, a total of 100 or 200 patients, 4 or 7 items and two different values for the correlation coefficient of the latent trait between two consecutive times (0.4 or 0.9). Cases including informative and non-informative intermittent missing data were compared at different rates (15, 30 %). These simulated data were analyzed with CTT using score and mixed model (SM) and with IRT using longitudinal Rasch mixed model (LRM). The type I error, the power and the bias of the group effect estimations were compared between the two methods.
This study showed that LRM performs better than SM. When the rate of missing data rose to 30 %, estimations were biased with SM mainly for informative missing data. Otherwise, LRM and SM methods were comparable concerning biases. However, regardless of the rate of intermittent missing data, power of LRM was higher compared to power of SM.
In conclusion, LRM should be favored when the rate of missing data is higher than 15 %. For other cases, SM and LRM provide similar results.
KeywordsIRT CTT Rasch models Longitudinal data PROs Missing data
This study was supported by the Ligue Nationale Contre le Cancer and the Comité de Loire-Atlantique de la Ligue Contre le Cancer.
- 1.Garcia, S. F., Cella, D., Clauser, S. B., Flynn, K. E., Lad, T., Lai, J. S., et al. (2007). Standardizing patient-reported outcomes assessment in cancer clinical trials: a patient-reported outcomes measurement information system initiative. Journal of Clinical Oncology: Official Journal of the American Society of Clinical Oncology, 25, 5106. doi: 10.1200/JCO.2007.12.2341.CrossRefGoogle Scholar
- 2.Falissard, B. (2001). Mesurer la subjectivité en santé: Perspective méthodologique et statistique. Paris: Masson.Google Scholar
- 3.Fairclough, D.L., Peterson, H.F., Chang, V. (1998). Why are missing quality of life data a problem in clinical trials of cancer therapy? Statistics in Medicine. doi: 10.1002/(SICI)1097-0258(19980315/15)17:5/7<667::AID-SIM813>3.0.CO;2-6.
- 4.Bernhard, J., Cella, D.F., Coates, A.S., Fallowfield, L., Ganz, P.A., Moinpour, C.M., Mosconi, P., Osoba, D., Simes, J., Hürny, C. (1998) Missing quality of life data in cancer clinical trials: serious problems and challenges. Statistics in Medicine. doi: 10.1002/(SICI) 1097-0258(19980315/15)17:5/7 < 517::AID-SIM799 > 3.0.CO;2-S.
- 8.Curran, D., Molenberghs, G., Fayers, P.M., Machin, D. (1998). Incomplete quality of life data in randomized trials: missing forms. Statistics in Medicine. doi: 10.1002/(SICI)1097-0258(19980315/15)17:5/7<697::AID-SIM815>3.0.CO;2-Y.
- 10.Curran, D., Bacchi, M., Schmitz, S.F.H., Molenberghs, G., Sylvester, R.J. (1998). Identifying the types of missingness in quality of life data from clinical trials. Statistics in Medicine. doi: 10.1002/(SICI)1097-0258(19980315/15)17:5/7<739::AID-SIM818>3.0.CO; 2-M.
- 11.Fayers, P.M., Curran, D., Machin, D. (1998). Incomplete quality of life data in randomized trials: missing items. Statistics in Medicine. doi: 10.1002/(SICI) 1097-0258(19980315/15)17:5/7 < 679::AID-SIM814 > 3.0.CO;2-X.
- 13.Blanchin, M., Hardouin, J.-B., Le Neel, T., Kubis, G., & Sébille, V. (2011). Analysis of longitudinal patient reported outcomes with informative and non-informative dropout: Comparison of CTT and Rasch-based methods. International Journal of Applied Mathematics and Statistics, 24(SI-11A), 107–124.Google Scholar
- 14.de Bock, E., Hardouin, J-B., Blanchin, M., Le Neel, T., Kubis, G., Dantan, E., Bonnaud-Antignac, A., Sébille, V. (2013). Rasch-family models are more valuable than score-based approaches for analysing longitudinal PRO with intermittent missing data. Statistical Methods in Medical Research: Published online.Google Scholar
- 22.Peyre, H., Leplège, A., & Coste, J. (2011). Missing data methods for dealing with missing items in quality of life questionnaires. A comparison by simulation of personal mean score, full information maximum likelihood, multiple imputation, and hot deck techniques applied to the SF-36 in the French 2003 decennial health survey. Quality of Life Research, 20, 287. doi: 10.1007/s11136-010-9740-3.PubMedCrossRefGoogle Scholar
- 23.Tenenhaus, M. (1999). Statistique et logiciels: Analyse de la variance à effets mixtes utilisation de la proc MIXED: Mais que reste-t-il à la proc GLM ? La Revue de Modulad, 23, 53–67.Google Scholar
- 26.Davier, M., Meiser, T. (2007). Rasch models for longitudinal data. New York: Springer, Statistics for Social and Behavioral Sciences.Google Scholar
- 27.Zheng, X., & Rabe-Hesketh, S. (2007). Estimating parameters of dichotomous and ordinal item response models with gllamm. Stata Journal, 7(3), 313–333.Google Scholar
- 28.Dempster, A. P., & Rubin, D. B. (1983). Overview, in incomplete data in sample surveys, Vol. II: Theory and annotated bibliography. New-York: Academic Press.Google Scholar