Data source, model inputs and patient population
The AirView database in the United States provides the data used to train the model and make predictions. We source a de-identified copy of the AirView database to train the model. A tool filters this data to remove all protected health information (PHI) and create an anonymous patient population that includes all active AirView patients in their first 90 days of therapy who use non-ventilation PAP devices. The model then analyzes device usage data from these patients to create the following inputs (or features):
Device usage: measures the total hours that the patient used their device on the treatment date.
Average device usage: displays the patient's average device usage over the last 3, 7, 14 and 28 days.
Number of compliant days: displays the number of days where a patient used their device for four hours or more over the most recent 10, 20 and 30 day periods.
Days since setup: displays the number of days that have passed since the patient’s HME-assigned setup date.
Approximately 90% of patient genders are unknown. Patient ages range between 0 and 99, with the majority in the 50-74 age group. Other demographic related information, such as ethnicity and disease or health concern, is not available in the data set used to train and evaluate the model.
Note: model performance does not vary by age or gender as explained in Model Performance.