Diagnostic accuracy and predictive validity of combined use of Fototest and Mini-Cog in cognitive impairment

Abstract

Introduction

The Fototest and Mini-Cog include all the domains that are necessary in a cognitive assessment. This study aims to evaluate the diagnostic accuracy of the combined use of both instruments for detecting cognitive impairment.

Methods

We performed a phase III diagnostic accuracy study with 2 independent samples: STUDY, which included 448 participants randomly allocated to 2 datasets (BASE [80%] and TEST [20%]); and EXTERNAL, which included 61 participants. The index test was consecutive administration of the Fototest and Mini-Cog, and the reference test was formal cognitive assessment. We evaluated the diagnostic accuracy of two-step vs consecutive application of the tests and simple (Comb-Simple), logistic regression (Comb-LR), and random decision tree (Comb-RDT) models of their combined use for detecting cognitive impairment (Global Deterioration Scale score ≥ 3). We performed an exploratory analysis of the BASE dataset, selecting criteria that maximise accuracy; a pre-specified analysis was used to evaluate the selected criteria in the TEST and EXTERNAL datasets.

Results

The diagnostic accuracy (95% confidence interval) of the combined models in the BASE dataset (Comb-Simple: 88.3 [88.5−91.4]; Comb-LR: 91.6 [88.2−94.3]; Comb-RDT 95.2 [92.5−97.2]) was significantly higher than the individual values observed for the Mini-Cog and Fototest (81.6 [77.1−85.4] and 84.9 [80.8−88.5], respectively). These results were replicated in the TEST (Comb-Simple: 88.9; Comb-LR: 95.6; Comb-RDT: 92.2) and EXTERNAL datasets (Comb-Simple: 91.8; Comb-LR: 90.2; Comb-RDT: 88.5). Two-step application had the same diagnostic accuracy than consecutive application but required less time (mean [SD] of 197.3 s [56.7] vs 233.9 s [45.2]; P < .0001).

Conclusions

Combined application of the Fototest and Mini-Cog takes less than 4 minutes and improves the diagnostic accuracy of both instruments. Two-step application is more efficient as it requires less time while maintaining the same diagnostic accuracy.