Skip to main content
. 2021 Nov 14;29(2):522–534. doi: 10.1111/ene.15162

TABLE 3.

Area under the curve values of the clinical benchmarks and the top five keystroke features for each clinical anchor

Clinical anchor

Clinical benchmark

 Keystroke dynamics
Measure AUC (95% CI) Feature a AUC (95% CI) MCID SRCind SRCgroup n b
Disease activity
Radiological and/or clinical EDSS 0.52 (0.34–0.70) PPLlast maximum 0.76 (0.60–0.91) 0.16 0.22 0.02 2
APPfirst maximum 0.75 (0.58–0.92) 0.24 0.26 0.03 2
RRLlast maximum 0.71 (0.53–0.89) 0.16 0.21 0.02 2
ES‐neutrality c 0.71 (0.53–0.89) 0.01 0.29 0.04 841
Session durationabsolute energy 0.71 (0.53–0.89) 0.70 0.86 0.09 2
Radiological EDSS 0.54 (0.35–0.72) PPLlast maximum 0.78 (0.62–0.95) 0.20 0.23 0.02 2
RRLlast maximum 0.74 (0.56–0.92) 0.18 0.22 0.02 2
Session durationabsolute energy 0.74 (0.55–0.92) 0.70 0.87 0.10 2
APPfirst maximum 0.73 (0.52–0.93) 0.28 0.26 0.03 1
ES‐polarityfirst maximum 0.72 (0.54–0.91) 0.22 0.25 0.03 2
Fatigue
CIS‐F FSS 0.54 (0.40–0.68) ES‐negativityfirst maximum 0.70 (0.56–0.84) 0.19 0.28 0.03 3
MFIS 0.45 (0.31–0.58) Session duration d 0.70 (0.56–0.84) 0.01 0.09 0.01 81
ES‐negativitylast maximum 0.69 (0.54–0.83) 0.25 0.43 0.05 3
ES‐negativityapproximate entropy 0.69 (0.51–0.86) 0.09 0.14 0.02 3
ES‐positivitystrike below mean 0.67 (0.52–0.83) 1.00 1.68 0.20 3
Clinical disability
EDSS TWT 0.50 (0.34–0.66) ES‐neutralityapproximate entropy 0.79 (0.63–0.96) 0.11 0.14 0.02 2
NHPT‐D 0.56 (0.42–0.71) Word lengthfirst maximum 0.72 (0.60–0.83) 0.11 0.19 0.02 3
ES‐positivityapproximate entropy 0.70 (0.52–0.88) 0.11 0.16 0.02 3
FTapproximate entropy 0.68 (0.56–0.80) 0.04 0.08 0.01 4
ES‐neutralitymean absolute change 0.66 (0.52–0.80) 0.16 0.54 0.06 12
Ambulation FS TWT 0.80 (0.70–0.90) RRLabsolute energy 0.73 (0.62–0.84) 3.44 6.86 0.73 4
Pre‐CSabsolute sum of changes 0.70 (0.58–0.81) 2.70 5.77 0.62 5
FTstrike above mean 0.69 (0.58–0.80) 1.50 1.23 0.13 1
FTabsolute energy 0.69 (0.57–0.80) 2.72 7.30 0.78 8
ES‐polaritystrike above mean 0.69 (0.55–0.83) 1.00 1.34 0.16 2
AMSQ NHPT 0.58 (0.36–0.80) Pre‐CSmedian 0.75 (0.59–0.92) 113.10 259.54 27.67 6
Pre‐CSmean 0.74 (0.54–0.95) 186.42 281.46 30.00 3
ES‐negativitymean 0.74 (0.62–0.86) 0.03 0.05 0.01 3
Post‐CSmedian 0.73 (0.53–0.93) 135.75 126.24 13.46 1
ES‐neutralitystrike above mean 0.72 (0.54–0.90) 1.00 1.50 0.18 3
SDMT EDSS 0.51 (0.39–0.63) ES‐positivityfirst maximum 0.68 (0.56–0.81) 0.09 0.28 0.03 10
ES‐polarityfirst maximum 0.66 (0.53–0.79) 0.14 0.26 0.03 4
PPLstrike below mean 0.66 (0.54–0.78) 3.00 5.67 0.60 4
CDstrike above mean 0.63 (0.51–0.75) 1.50 12.90 1.37 74
Keystroke count 0.63 (0.51–0.75) 3.50 13.29 1.42 15

Abbreviations: AMSQ, Arm function in MS Questionnaire; APP, After Punctuation Pause; AUC, area under the curve; CD, Correction Duration; CI, confidence interval; CIS‐F, Checklist Individual Strength Fatigue subscale; EDSS, Expanded Disability Status Scale; ES, Emoji Sentiment; FSS, Fatigue Severity Scale; FT, Flight Time; MCID, minimal clinically important difference; MFIS, Modified Fatigue Impact Scale; NHPT‐D, Nine‐Hole Peg Test dominant hand; Post‐CS, Post‐Correction Slowing; PPL, Press‐Press Latency; Pre‐CS, Pre‐Correction Slowing; RRL, Release‐Release Latency; SDMT, Symbol Digit Modalities Test; SRCgroup, smallest real change group; SRCind, smallest real change individual; TNR, true negative rate; TPR, true positive rate; TWT, Timed 25‐foot Walk Test.

a

Five keystroke features with the highest AUC‐values.

b

Number of patients or repeated measures needed to detect clinical relevant change beyond measurement error on the individual level (i.e., MCID = SRC).

c

Time reversal asymmetry statistic.

d

Mean value of a central approximation of the second derivative.