Aims
Geriatricians are often required to synthesise and interpret cognitive test results from different assessment tools, times and administration modalities (e.g., telephone and face-to-face [F2F]). We aimed to explore the implications of three different analytical techniques to allow comparisons between a 100-point F2F Modified Mini-Mental State (3MS) and a 73-point telephone version.
Methods
We used baseline data from the STAREE (Statin Therapy for Reducing Events in the Elderly) trial, a randomised placebo-controlled trial investigating atorvastatin in older individuals aged ≥70 years in Australia without dementia.
Results
Of the 9,971 participants (mean age 74.7 years, 52% women), 8,924 (89%) had F2F assessment and 1,047 (11%) telephone. The mean telephone score was lower than the F2F score (68.3 vs 93.8, p<0.001). Removing those items from the F2F score that were absent from the telephone score reduced the mean score difference to 1.2 (p<0.001). Linear scaling the telephone score to 100 points also resulted in a reduction in the mean score difference (0.5, p=0.001). Using administration modality-specific z-scores, the mean telephone score (0.00000084) was almost the same as the F2F score (-0.00000025) (p for difference=1.0).
Conclusions
Modality-specific z-scores minimise differences attributable to assessment modality but are difficult for clinicians to interpret. Approaches that use scaling or remove items to improve alignment can lead to more interpretable scores, but risk obscuring potentially important differences related to mode of administration. This work highlights the need to carefully consider how to synthesise different cognitive tests and appropriately balance statistical robustness, generalisability and clinical interpretability.