Using latent trajectory analysis of residuals to detect response shift in general health among patients with multiple sclerosis article

Individuals experiencing a change in health may experience a response shift that may attenuate HRQoL change estimates. The
objective of this study was to estimate the proportion of individuals with multiple sclerosis (MS) who experienced a response
shift as detected by the Latent Trajectory of Residuals approach.

Participants in the NARCOMS Registry were included if they responded to the general health (GH) question of the SF-12 in at
least 3 surveys. Linear growth modeling was used to identify predictors of self-reported GH, and the residuals from this model
were used to determine group-based trajectories. Dual trajectories of GH and a measure of disability (PDSS) were used to examine
convergence in change patterns over time.

A total of 1,566 individuals were included in this study. The predictive GH model explained 55% of the variance; 99.7% of
subjects did not experience a response shift as indicated by flat trajectories, and 0.3% lowered their rating of health as
compared to predicted indicating a potential response shift. Among 13% of subjects with flat trajectories of PDDS, 5% had
GH decreasing most strongly showing some discordance between symptoms and GH.

A lower percentage of individuals experienced response shift than previous research on smaller samples. These results may
indicate the true absence of response shift, or may be limited by using a categorical outcome of GH, and GH predictors that
may have also been amenable to response shift, which decreases the appropriateness of using the LTA approach. Future work
will include the use of growth curve latent class analyses to assess response shift.