Typing behaviour to remotely monitor clinical MS status

Real-world smartphone typing behaviour (‘keystroke dynamics’; KD) can effectively distinguish between MS patients and healthy controls, and between MS patients with different levels of disability. These findings show the potential of KD as a digital biomarker to remotely monitor clinical MS status.

Physical and cognitive functions required for typing are affected in MS patients. Dr Ka-Hoo Lam (Amsterdam UMC, the Netherlands) and colleagues studied KD data in people using an app called Neurokeys [1]. The information collected included typing speed features (alphanumeric keys) and processing speed features (backspaces and punctuation marks), based on hold time, release time, or both. The study aimed to verify if KD collected by smartphone could discriminate between MS patients and healthy controls and between MS patients with different levels of disability, as defined by Expanded Disability Status Scale (EDSS) scores.

A logistic regression and random forest algorithm were combined in a model to capture linear and non-linear trends. It was tested in 97 MS patients and 22 healthy controls, from whom 2 weeks of KD were aggregated per hour. MS patients were divided into 2 sub-groups: low disability (EDSS ≤3.5; n=61) and higher disability (EDSS >3.5; n=36). In the low disability group, 44 patients had relapsing-remitting MS, 12 had secondary progressive MS, and 5 had primary progressive MS. In the higher disability group, 14 had relapsing-remitting MS, 17 had secondary progressive MS, and 5 had primary progressive MS.

The best performing model for distinguishing between MS patients and controls had an area under the curve (AUC) of 0.78, indicating good sensitivity and specificity. This model included 2 typing speed features, number of suggestions, age, and gender. Similarly, the AUC for distinguishing between low and higher disability groups was 0.78. This model included 3 typing speed features, a mental processing feature, and age. Both models were primarily driven by KD.

1. Lam K, et al. Real-world smartphone keyboard interactions discriminate between different levels of disability in multiple sclerosis. EPR-114, EAN 2021 Virtual Congress, 19–22 June.

 

Copyright ©2021 Medicom Medical Publishers



Posted on 18 August 202112 January 2022