Performance

Success criteria

Hyfe’s performance has been assessed against a set of criteria established through an iterative survey/interview process with key opinion leaders and other stakeholders in the cough space in a study whose details are available here.

Full details on how success criteria were established are available here.

Results

All success criteria were satisfied.

• The Lin’s Concordance Correlation Coefficient has a value of >0.9 for daily cough counts and cough-second counts.
  • Coughs: 0.971
    
  • Cough-seconds: 0.956
• The Lin’s Concordance Correlation Coefficient has a value of >0.9 for hourly cough and cough-second counts.
  • Coughs: 0.9748
  • Cough-seconds: 0.9683
    
• Event-specific sensitivity has a median value of >85% (coughs and cough-seconds)
  • Coughs: 87.34%
    
  • Cough-seconds: 90.48%
    
• Event-specific false positives per hour has a median value of <2.5 (coughs and cough-seconds)
  • Coughs: 1.67
    
  • Cough-seconds: 1.56
    
• At least 75% of patients are correctly classified as low (0-10 per hour), medium-low (10-20 per hour), medium-high (20-30 per hour), or high (30+ per hour) rate coughers per their 24 hour counts.
  • Correctly classified, coughs: (83.87 %)
    
  • Correctly classified, coughs-seconds: (80.65 %)
    
• Using the same categories, >85% of patients are correctly classified or misclassified by only one ordinal level per their 24 hour counts.
  • Correctly classified or misclassified by only one ordinal level, individuals, coughs: 31 (100 %)
    
  • Correctly classified or misclassified by only one ordinal level, individuals, coughs-seconds: 31 (100 %)
    
• Using the same categories, >85% of person-hours are correctly classified or misclassified by only one ordinal level.
  • Correctly classified hours, coughs: 686 (86.29 %)
    
  • Correctly classified hours, cough-seconds: 686 (86.29 %)
    
  • Correctly classified or misclassified by only one ordinal level, hours, coughs: 784 (98.62 %)
    
  • Correctly classified or misclassified by only one ordinal level, hours, cough-seconds: 787 (98.99 %)
    
• Patient-specific trends are correctly captured using the following method: (a) each patient’s hourly cough counts are binned by quartile using ground truth data (low being the 0-25th percentile coughs; medium-low being the 25-50th percentile coughs, etc.); for >75% of patients, >75% of their hours are correctly categorized into quartile bins by HCD, or misclassified by only one ordinal level.
  • Correctly classified hours, coughs: 621 (78.11%)
    
  • Correctly classified hours, cough-seconds: 628 (79.0%)
    
  • Correctly classified or misclassified by only one ordinal level, hours, coughs: 761 (95.72%)
    
  • Correctly classified or misclassified by only one ordinal level, hours, cough-seconds: 758 (95.35%)