MiroHealth
Groundbreaking
science
at your
fingertips
The brain is our most vital organ.
We help you treat it that way.
back to Our Science
CONCURRENT VALIDITY
STUDY RESULTS
Basic Miro variables show significant correlation with basic variables collected by trained clinicians
using standard clinical assessments. In this study, 152 participants were tested on Miro assessments and
a battery of standard comparator tests.
Miro Versions
v3.1.2.6 MiroMind
v2.1.3.0 On-device speech recognition algorithms
v2.1.2.0 Natural language processing algorithms
v2.0 Machine learning algorithms
v3.0 Reference data set
Introduction
This analysis compares individual variable scores produced by Miro’s automated data capture and
scoring approach to individual variable scores produced by the data capture and scoring approach of
trained clinicians using comparator tests. We chose to compare individual variables to establish
performance correlation at the most basic level.

The hi-fidelity data that Miro captures and analyzes, e.g., a person’s speech and movement, cannot be
reliably quantified by a human clinician and is not included in standard clinical testing. Hence, the
majority of Miro’s more than 500 variables lack a traditional counterpart. The variables available for
correlation testing are limited by the constraints of traditional assessments. By providing analogues of
traditional functional assessments, Miro provides a familiar starting point for the interpretation of a
user's functional performance.

To demonstrate test validity, Miro conducts "Concurrent Validity” studies to generate evidence that
determines the relatedness of Miro scores with comparator test scores.
Demographics
Table 1. Demographics
GROUP% F/MMEAN AGE (RANGE)TOTAL N=152
Healthy Controls58% / 42%69 (36 - 82)73
Impaired40% / 60%65 (33 - 89)79
(% F/M = percent female / percent male)
Methods
Location. All participants were assessed in unaffiliated clinics by a trained administrator. Thirty-four
percent were assessed at Johns Hopkins University Medical Center. Sixty-six percent of participants were
assessed at one of three independent clinical research sites.

Assessment. 50% of participants were assessed with comparator tests prior to Miro assessment; 50% were
assessed with Miro prior to comparator assessment. Comparator tests were hand-scored by the
administrator and entered into a spreadsheet. Miro performance data was automatically uploaded from
the iPad to Miro’s HIPAA-compliant cloud-based server..

Missing data. Correlations for each variable were calculated with participants whose results included both
Miro and comparator scores. Participants were excluded from the correlation of individual variables when
missing either Miro or comparator scores (or both) for that variable.

Standardization of scores. Standardization of scores occurred via a two-step process: 1. Subtracting the
mean from the normal participant reference set, 2. Rescaling centered scores by the standard deviation of
reference set scores. The mean score of the normal reference set was set to 0 and the standard deviation
was set to 1.
Results
Pairs of Miro and comparator test scores that are statistically significantly correlated at the nominal
alpha=0.05 level are shown in Table below. Most of the included p-values are well below 0.05, indicating
the demonstrated correlations between Miro and traditional assessments are highly statistically significant.
Table 2. Statistically significant correlations between Miro scores and traditional analogues
MEASUREVARIABLECORRELATIONP-VALUE
Category FluencyAnimals Correct0.672.17E-19
Digit Span ForwardLongest Correct Set0.642.90E-05
Letter FluencyF-Words Correct0.641.56E-17
Trails BTotal Time0.624.76E-14
CodingTotal Correctn0.593.81E-13
Trails ATotal Time0.545.88E-11
Design FluencyFilled Dots Correct0.484.97E-08
Spatial Span Backward Longest Correct Span0.423.32E-07
Delayed Verbal Learning and MemoryFree Recall Correct0.394.71E-05
Finger TappingAverage Taps Left Hand0.395.84E-05
Finger TappingAverage Taps Right Hand0.320.00125
Verbal Learning and MemoryLongest Correct Set0.320.01023
Spatial Span ForwardLongest Correct Span0.304.91E-4
Pairs of Miro and traditional assessment scores that are not statistically significantly correlated in the
currently available data are shown in Table 3. Most of the scores which have been shown not to be
significantly correlated are "error scores" -- counts of various types of errors made in the course of playing a
Miro game or completing a traditional assessment.

The majority of the participants whose data is currently available are cognitively normal. Normal
participants tend to have a narrow range of performance on both the Miro and traditional assessments.
The normal participants who dominate the current data set make very few of the errors demonstrated in
Table 3, so there is little to distinguish the performance of different individuals. This leads to low
correlations with low statistical significance. These scores are likely to be most useful in the identification
of individuals with low functional ability or in conjunction with additional scores that may help corroborate
an emerging story of a participant's capabilities.

By contrast, participants with mild cognitive impairment or other conditions that impact cognitive or
motor function would have a wider range of performance on these assessments which is how their
functional deficits are identified. The relatively low variation in performance among cognitively normal
participants may not be representative of the variance in the prospective user population. Still, with the
currently available data, most of the paired traditional and normal scores are significantly correlated.
Table 3. Correlations between Miro and traditional scores that are not statistically significant
MEASUREVARIABLECORRELATIONP-VALUE
Verbal Learning & MemoryNumber errors: total-0.170.18
Category FluencyNumber errors: set loss-0.020.81
Category FluencyNumber errors: repetitions-0.0030.98
Letter FluencyNumber errors: repetitions0.030.78
Verbal Learning (Delayed Recall)Number errors0.050.62
Design FluencyNumber errors: total0.060.55
Discussion
As expected, concurrent validity between independent variables from Miro’s self-administered,
clinician-supervised tablet assessment and comparator clinician-administered pencil-paper based testing
was moderate, ranging from 0.30 (Spatial Span Forward) to 0.67 (Category Fluency), when excluding the
low-information error scores. Relatively modest correlation is expected given the challenges with
comparator test inter- and intra-rater reliability, bluntness of scores, and low levels of sensitivity and
specificity for disorder differentiation. Statistically significant Spearman correlations between Miro scores
and their comparator analogues suggest that Miro and comparator exams quantify equivalent functional
abilities. Low p-values indicate a high degree of confidence in the correlations. This is notable given the
study’s limited sample size, the large proportion of cognitively normal participants in the study who
demonstrate a narrow range of functional ability, and the large representation of highly functioning but
mildly impaired participants whose difference in cognitive abilities cannot be detected by comparator
tests.

It is important to note that the sample of normal controls exhibit a narrow range of performance on both
Miro modules and traditional neuropsychological tests. The narrow range of scores from the normal
controls dampens potential correlation with comparator scores, whereas the broader performance range
of cognitively impaired participants in the prospective Miro user populations strengthens the potential
correlation between the familiar traditional assessment scores and their analogs in the Miro platform.

Miro module scores demonstrate significant correlation with comparator test scores. Excluding the
relatively low information “error scores”, estimated Spearman correlations for many Miro and comparator
test scores are around 0.5 and are significantly different than zero with p-values of 0.05 or lower. Statistical
results provide preliminary evidence that Miro scores quantify brain function comparably to standard,
traditional, in-depth, clinician-administered neuropsychological assessment methods.
Follow the data
GET STARTED