English publication
Theunissen,
N.C.M., Meulman, J.J., den-Ouden, A.L., Koopman, H.M., Verrips, G.H., Verloove-Vanhorick,
S.P., & Wit, J.M. (2003). Changes can be studied when the measurement
instrument is different at different time points. Health Services and
Outcomes Research Methodology, 4 (2): 109-126 (published may 2004)
Abstract
This paper presents a strategy for the analysis of longitudinal data that
suffer from differences in measurement instruments over time. For illustration,
we used a set of longitudinal Health Status (HS) data from a cohort of
preterm born children. The strategy solved three problems: First, commonly
known HS domains were not explicitly used. Therefore, expert ratings were
used to fit the variables into HS domains. Second, it was unclear which
HS variables at 5 years of age matched which variables at 10 years of
age. Principal component analyses (PCA) was used to match the data between
measurements. Third, the change between points of measurement were estimated
by longitudinal PCAs using the matched HS data. The impact of background
variables such as gender and birth weight on HS changes was studied. This
strategy using an imperfect data set, provided valuable information about
the development of HS in preterm born children.
Keywords
principal component analysis; CATPCA;
longitudinal studies; health status; preterm birth;
