COVID-19 Triage – Different Perspectives

April 23, 2020


COVID-19 is rampant across India, infecting close to 17,000 people and claiming the lives of 550 people. The country is well into its 4thweek of a nationwide lockdown. While there is a slow-down of the virus in certain areas, we still have a fair amount of red zones across the country. With some parts of the country opening up, in the coming days and weeks, “preventive testing” will become even more important. Being a very populous nation of 1.3 billion people, it would not be possible to test even a small fraction of the populace. Hence there is a definite need for “out of the box” solutions that can help with triaging and assessment, of who needs to get the tests done on a priority.

We, at RoundSqr, have looked at 2 variants of this triage.

STREAM 1: Spectral analysis of cough sample to predict COVID-19 possibility

STREAM 2: End-to-end triage, along with specific “Call For Action”, focussed on frontline workers

Stream 1: Spectral Analysis

We have previously worked on identifying Tuberculosis based on how a person coughs. That was based on both spirometry and spectral analysis of the cough. The spirometry process is a bit cumbersome for both the doctors and the users. However, we were able to achieve close to 85% accuracy using pure cough sounds analysis. The approach and algorithms used are detailed in the following paper:

Leveraging this experience, we intend to validate whether we can re-purpose this TB-solution for COVID-19 detection as well. Using linear and non-linear regression methods, we will conduct a spectral analysis of cough sounds of the subjects (suspected users). These cough sounds will be normalized with other audio samples. This data, when fed into a classification model along with other data (like medical history, demographic data, other symptoms…) might help in providing a COVID-19 prediction with uncertainty measure.

How do we intend to do this

There is no doubt; this is a research project. The principal aspect is to obtain data to run our hypothesis. To facilitate data collection, we intend to use a mobile app (Android-based) to collect relevant data, including

  • Basic Demographic Data: Age, Gender, Nationality, and International Travel (last 2 weeks)
  • Medical Data: Existing Medical History, Temperature, Breathing Difficulty, any Loss of Smell or Taste, and Feeling of Tiredness
  • Audio Data: Cough and few spoken sentences (counting from 1 to 10, and the English / Telugu / Hindi alphabets).
Application Screens

The data collection app itself is a set of simple screens where information about the profile (one-time) is collected and multiple triage performed for each profile.

Where are we right now?

The overall solution is to develop a mobile application (in Android to start with) that can be used to answer a few simple questions, provide audio samples(of a short cough and some spoken words), and calculate the probability of COVID-19 symptoms. This app can be used as the first line of defence before formal testing is carried out, thereby ensuring that the testing resources are optimally utilized as well as making sure that truly at-risk patients are tested as required.

The App is in Google Play Store, and we are in the process of collecting data.

Stream 2: Frontline Workers Triage Portal

As per WHO, close to 22,000 healthcare workers globally have been infected by COVID-19. The Nurse to Patient ratio in India is 1.7:1000. In addition to the healthcare workers, we have essential services – law and order, eCommerce, food delivery, volunteers, ASHA workers, etc. All of them are more susceptible to COVID-19, considering their exposure to patients.

Opening up the economy by selectively removing lockdown restrictions on specific industries like farming, logistics, construction, eCommerce, online food ordering, etc. only means that more people have the risk of exposure and, more importantly, cause the spread of the virus, unless aggressively detected and checked. Apart from the possibility of infection, frontline workers are also under a lot of psychological and physical stress. Unless proper identification of these symptoms and appropriate / timely interventions are carried out, this is just a time-bomb waiting to explode.

We believe a proper triage on a regular basis is a step in the right direction, in terms of understanding the physical as well as the mental health of the frontline workers. With proper analysis, reporting, and exception highlighting, we can equip the supervisors / department heads with the transparency required to ensure that the frontline workers are giving optimal service.

How do we intend to do this

We have built a set of questions (physical and psychological),based on standard COVID-19 symptoms, aimed at assisting the Triage process. The frontline worker (or an operator who can collect this data for each of the workers) will have to answer these questions. The intent is to collect such information diligently and consistently, twice a day (start and end of a shift).

The three steps involved in the process are

The information collected will comprise of:

  • Basic Profile Information – Age, Gender, Nature of Work
  • Health Conditions – Temperature, Sense of Smell, Breathlessness, Fatigue
  • Questions pertaining to mental state, administered over a period of time
Application Screens

The workflow will start from setting up an organization, its departments, and the ability to add a frontline worker. The profile will be set up once (and it could be updated regularly if needed)

We recommend a protocol of collecting triage information from the staff twice a day – Start and End of a shift. Along with the standard health parameters, it is imperative to collect information regarding the Mental Health of the staff.

Outcomes will be monitored by the supervisor, and relevant actions will be taken – Recommend a Day Off, Reduce Shift Timings, Send for screening, Recommend a Counselling session.

Where are we right now?

The core platform with the capability to setup an organization, (optionally) kiosks and operators, along with the ability to enter frontline worker details and carry out triage, is ready and deployed.

All the questions that are administered to determine the Mental Health, and other information collected as a part of triage, are completely configurable.

We are looking at enhancing up the platform by providing the following additional features:

  • Integration to any healthcare device – to collect vitals automatically
  • Ability to configure and determine the ideal “call for action” recommendations based on the triage. A few examples of this could be a supervisor deciding to give a day off, changing the shift or providing counselling to the frontline worker
  • Including a set of training videos, useful links and contact information that can be used by the frontline workers

To get our daily updates on COVID-19 initiatives, please follow us on COVAID-20