Towards individualized monitoring of cognition in multiple sclerosis in the digital era: a one-year cohort study

Abstract (link to article)

Highlights

  • Frequent cognition tests in the ambulant setting could enhance clinical assessment
  • Weekly smartphone tests enable higher temporal resolution and individual curve fits
  • Curve fitting improves the detection of statistically reliable change in cognition
  • Thus enabling the remote monitoring of cognition on the individual patient-level

Background

Cognitive impairment is frequent in multiple sclerosis (MS), but reliable, sensitive and individualized monitoring in clinical practice is still limited. Smartphone-adapted tests may enhance the assessment of function as tests can be performed more frequently and within the daily living environment. The objectives were to prove reproducibility of a smartphone-based Symbol Digit Modalities Test (sSDMT), its responsiveness to relevant change in clinical cognitive outcomes, and develop an individual-based monitoring method for cognition.

Methods

In a one-year cohort study with 102 patients with MS, weekly sSDMTs were performed and analyzed on reproducibility parameters: the standard error of measurement (SEM) and smallest detectable change (SDC). Responsiveness of the sSDMT to relevant change in the 3-monthly clinically assessed SDMT (i.e. 4-point change) was quantified with the area under the receiver operating characteristic curve (AUC). Curve fitting of the weekly sSDMT scores of individual patients was performed with a local linear trend model to estimate and visualize the de-noised cognitive state and 95% confidence interval (CI). The optimal assessment frequency was determined by analyzing the CI bandwidth as a function of sSDMT assessment frequency.

Results

Weekly sSDMT showed improved reproducibility estimates (SEM=2.94, SDC=8.15) compared to the clinical SDMT. AUC-values did not exceed 0.70 in classifying relevant change in cSDMT. However, utilizing weekly sSDMT measurements, estimated state curves and the 95% CI were plotted showing detailed changes within individuals over time. With a test frequency of once per 12 days, 4-point changes in sSDMT can be detected.

Conclusion

A local linear trend model applied on sSDMT scores of individual patients increases the signal-to-noise ratio substantially, which improves the detection of statistically reliable changes. Therefore, this fine-grained individual-based monitoring approach can be used to complement current clinical assessment to enhance clinical care in MS.