While most of us were sharing and retweeting SOS messages about #urgentneed for hospital beds and oxygen cylinders on social media, a group of techies from Telangana did something different. They slogged on weekends for three months to work out a solution that predicts how many hospital beds the state needs in the near future.
“Being forewarned is being forearmed. States can use this information to be Covid-ready, perhaps convert schools and colleges as hospitals to deal with emergencies,” say the team members, comprising Hyderabad-based working professionals Sateesh Kumar Talupuri, Shruti Galande, Mahita GM, Vasundhara Konanki, and Chiru Hasini Tondapu.
They used last year’s data sourced from the Covid care website of Telangana, to predict the requirements. They developed this as part of their project for Great Learning, a professional higher education firm. “We learnt to leverage time series analysis to estimate bed occupancy based on the Covid-19 positive and active cases,” explains Talupuri. However, the predictions should be treated as guidelines and not as the final word, the team warns. Predictions need years of data and currently, the team has enough only to predict the situation June 2021 onwards.
According to their model, first the number of active cases is predicted. They have seen that an average of 22 percent of active cases are hospitalised. Of these, the share of private oxygen beds is 27.6 percent and the government oxygen beds 20.4 percent. The students used the modelling and predictions for April 2021 and forecasted that the number of active cases were doubling every week. This model can be applied to the data available in the other states as well.
Says Talupuri, “Based on recent past data collected post the lockdown, our model predicts that in the next two to four weeks, cases will come down to almost 0.” And interestingly, Telangana did record 1.5 percent positive cases in mid-June, and the state decided to end the lockdown. The team is keen to develop a complete interface and translate it into a mobile/cloud/web app.
The model also takes into account triggers that can change the requirement. For example, in July the state celebrates Bonalu and it could lead to crowding. The festival could be a trigger. “We introduce these triggers into our AI-model and it throws up a new set of predictions. Similarly, we can predict the number of beds we may need if public spaces such as cinema halls are open,” says Mahita, who works for Google.
They think that as Covid-19 itself is an unprecedented phenomenon, they are not going to make any predictions that could cause panic in the public and would rather submit their academic project to the state government for further action. “We are looking at translating this project into something that makes us third wave-ready,” Talupuri says. We shall ride and conquer this wave.
According to their model, first the number of active cases is predicted. An average of 22 percent of active cases are hospitalised.