Content for IRB proposals

Hyfe has been IRB-approved for dozens of studies around the world. Below we provide examples of content that may be useful as you build your own IRB proposal for studies involving Hyfe's AI products.

Study aims & objectives

This language example comes from a study using Hyfe to screen for COVID-19.

  1. To determine if a cough signature can distinguish between individuals with COVID-19 from individuals without COVID-19.
  2. To characterize the change in frequency or type of cough during treatment or quarantining for COVID-19 infection.
  3. To assess digital cough monitoring for ambulatory remote monitoring of patients with COVID-19.

Procedures & Methods

This language explains how the Hyfe Research app works, and may be useful in the Methods section of your proposal.

Recording of cough sounds using the Hyfe Research app:

The Hyfe Research app is a mobile wellness information tool to monitor cough. The app records and processes sounds continuously, and saves a <0.5 second recording whenever there is an "explosive" sound that is greater than a set threshold above background noise signal. The brief recordings are saved to a server and then classified as a cough using a machine learning algorithm. Each sound recording includes a time stamp and, if users enable location sharing, a GPS coordinate. Daily cough counts are displayed as a bar chart on the app. < We will distribute study-owned smart phones (pre-loaded with the Hyfe Research app and a data bundle) to -- or -- We will assist in downloading the Hyfe Research app on the personal devices of >study participants who provide separate consent to record cough sounds over a 14-day period.

Possible discomforts & risks

This language may be useful in the sections of your IRB proposal that discuss possible risks to participants.

There is no discomfort associated with recording coughs. There are possible minimal risks in the following areas:

a. Inadvertent release of PHI: risk minimized by <details, e.g., use of REDCap for management of clinical data and other PHI >

c. Recording of sounds poses specific ethical concerns regarding privacy of data. Following legal and technical expert consultation, these are addressable at several independent levels.

  • Device level: While prolonged sound recordings can be used to identify participants (e.g. Amazon’s Alexa device [Alexa User Guide 2020]), the short snippets (< 0.5 seconds) recorded by the Hyfe Research app cannot.  Conversations or acoustic environments are not recorded.
  • Study level: At all points, participants can opt out of the study and monitors/phone apps will be removed. Our consent process will explicitly describe exactly what is and is not recorded. We will employ standard data encryption and safety tools at rest and in transit. Specifically, this project will use the Hyfe Research app, which uses the Google Cloud platform for data management and storage. Firebase uses industry-standard encryption and security protocols. Firestore, the Google Cloud storage architecture (where audio files and associated meta-data are stored) is HIPAA-compliant. Both audio files and associated meta-data are encrypted in transit using HTTPS. When the server receives a file, it is processed through the Hyfe algorithms, and sounds likely to be cough are stored in <=0.5 second files on encrypted disks. Servers are physically located in the United States.
  • Data sharing level: Data will be transferred from Hyfe to the research team via a data dump occurring at regular intervals (usually one week). Each research participant/site will be associated with a "Hyfe ID" number; all audio files and meta-data transferred from Hyfe to the research team will be linkable to the study participant/site via this "Hyfe ID" number. Hyfe will not have access to the study participants' identity. Data will be transferred in encrypted form over a secure line (HTTPS); only the PI or authorized delegate on the research team will have the decryption key. The software used for handling transfers will be a custom-made web application, hosted on the Google Cloud platform on servers physically located in the United States.

All data will be stored on secured servers. We adhere strictly to principles of data sharing for public good and will deposit appropriate anonymized acoustic data in the Sage Bionetworks Synapse system (https://sagebionetworks.org/tools_resources/synapse-platform/ ) specifically to ensure that all such shared data is done so in compliance with ethical and legal standards and in service of improving health.

Possible benefits

This language may be useful in the sections of your IRB proposal that discuss possible benefits to participants.

No immediate benefit to participants. Longer term, if we are able to show an association between specific cough patterns and COVID-19, this may help in rapid diagnostics and community surveillance for the infection.

Privacy protection

This language may be useful in the sections of your IRB proposal that discuss the protection of participant privacy.

All participants are provided a unique study ID and will also be provided with a unique Hyfe Research app user identifier (from Hyfe Research app) if they provide consent to participate in this activity. Only local investigators will have access to the key allowing for linkage of cough data and recordings with patient identification and any other metadata collected as part of the study. No patient identifiers, demographic information, or medical data will be entered directly into the app. Putative cough sounds (i.e., "explosive" sounds) recorded using the Hyfe Research app will be <0.5 seconds in duration and uploaded to a secure server. Non-explosive sounds are immediately discarded upon processing. The short length of the recordings means they cannot be used to identify the patient (unlike prolonged sound recordings, like Amazon's Alexa device). Standard data encryption and safety tools will be employed (see details in Possible Risks section).