Smartphone Software Application-mediated At-home Telespirometry (AHT) Artificial Intelligence(AI)-algorithm Identifies Similar Proportion of Valid Erect (eFVC) and Supine (sFVC) Forced Vital Capacity Measurements and Comparable per Subject Test-ReTest Variability Above and Below Forced Vital Capacity 60% Predicted in an Amyotrophic Lateral Sclerosis (ALS) Clinic Population
George Slavinski1, Eufrosina Young2, Dongliang Wang3, Benjamin Brooks4
1Pulmonary, 2Neurology, 3Statistics, SUNY Upstate University Hospital, 4Clinical Trials Planning LLC
Objective:
Validate adherence to American Thoracic Society(ATS) 2019 Spirometry Guidelines when measuring eFVC/sFVC in ALS subjects with AHT monitored by AI-algorithm assessing FVC flow-volume-curves(Graham 2019).
Background:
Smartphone software-mediated AHT is being deployed world-wide to permit timely management with earlier treatment of ALS respiratory associated co-morbidities(Young Muscle&Nerve 2022, 2023; Moran Muscle&Nerve 2023). Each eFVC/sFVC measurement, previously evaluated by respiratory therapist(RT)can now be assessed by AI-algorithm that classifies eFVC/sFVC by flow-volume-curve characteristics and number of acceptable repeatable measurements within 0.150L(3-A, 2-B)|0.200L(2-C)|0.250 L(2-D)|> 0.250 L(2-E) or (1- acceptable-E)|none acceptable by prior repeatability criteria but usable based on at least one acceptable flow volume-curve(1-U).
Design/Methods:
120 data sets representing the best eFVC/sFVC measurement in 22 ALS subjects participating in a pandemic pilot feasibility deployment of the ZephyRx-Breath-Easy-software application with MIR-turbine-spirometer and ZephyRx-cloud-database-dashboard to permit longitudinal AHT over median 194 days[95%CI=92, 223 days].
Results:
eFVC/sFVC ATS 2019 standards[classifications A,B,C,D,E]were met for mean-82.6%[95%CI=69.9%,95.2%]of measurements in ALS subjects increasing to mean-87.7%[95%CI=73.9 %,101.5 %]for classifications A,B,C,D,E,U]. eFVC/sFVC ATS 2019 standards[classifications A,B,C,D,E,U]were met for mean-93.2%[95%CI=78.5%,107.8%]of measurements for ALS subjects with eFVC-BL>60 %p, not statistically different from mean-82.5%[95%CI=-9.8%,174.4%]of measurements for ALS subjects with eFVC-BL<60 %p. eFVC/sFVC ATS 2019 failure(F)classification stratified by eFVC-BL>or<60 %p is not significantly increased in the eFVC-BL<60%p cohort in this observational study. ALS subject Test-ReTest repeatability was similar for eFVC/sFVC measurements despite stratification by eFVC-BL>or<60 %p.
Conclusions:
This is the first report evaluating an ATS 2019 standards-based AI-algorithm in an ALS subject population monitored by AHT. eFVC/sFVC demonstrated high comparable attainment of benchmark indicators of appropriate eFVC/sFVC measurement when stratified by eFVC-BL>or<60 %p. We recommend that each iterative change AI-algorithm release be evaluated in the appropriate patient population to determine whether the benchmark indicators are maintained overtime. Future studies will include comparative assessment of best versus not-best assessments in his ALS clinic population.