Markus Mainka - stock.adobe.com

Using machine learning to address COVID-19 vaccine hesitancy

A look at how Final Mile is trying to fix vaccine hesitancy with its AI model and behavioral science and design, along with the impact of models developed during the pandemic.

A behavioral science consulting firm is working on building a machine learning model that can predict COVID-19 vaccine hesitancy.

What Final Mile is doing

Final Mile -- a subsidiary of AI vendor Fractal Analytics -- works in the public health arena, using behavioral and data science to try to stem the spread of viruses such as HIV, and, most recently, COVID-19.

When the coronavirus pandemic started two years ago, Alok Gangaramany, Final Mile's director of the development sector, said the organization anticipated that when a vaccine became available, people might be hesitant to get it.

After conducting research, Final Mile launched a COVID-19 playbook to look at many of the possible steps to take to limit the spread and impact of the pandemic, including social distancing, masks, vaccines and other actions.

Currently, the firm is trying to build a machine learning model that uses social media data to determine at a granular community level where there are likely to be clusters of COVID-19 and low vaccination rates.

"From social media you can actually gauge who are likely to be more deniers of COVID … not necessarily using social distancing, masks," Gangaramany said.

AI, machine learning and behavior science

Combining machine learning with behavioral science and design can help create the right kind of response to influence people's behavior, he said.

We're trying to build some kind of a public health tool that hopefully people can use.
Alok GangaramanyDirector of development sector, Final Mile

"AI can actually do a far better job than what surveys have done," Gangaramany continued.

Surveys tend to be long, time-consuming and expensive. In comparison, an AI system can look at different data -- such as mail location codes, news articles and social media posts -- to make faster predictions about vaccine hesitancy.

"We're trying to build some kind of a public health tool that hopefully people can use," Gangaramany continued.

Final Mile, which has offices in New York City, Johannesburg and Mumbai, India, now is focusing on places such as Kenya, South Africa and Pakistan. And while the model is focused on vaccine hesitancy, Gangaramany said now with the omicron variant, Final Mile might go back to its investors and donors to expand the project.

Building a vaccine hesitancy model

Much of vaccine hesitancy stems from online misinformation, said Kashyap Kompella, an analyst at RPA2AI Research.

He said that using machine learning, researchers and policymakers can flag and identify that misinformation and also help in crafting the right message and campaigns to help those who are hesitant to get vaccinated.

image of a COVID-19 testing lab
Since the start of the COVID-19 pandemic, different types of machine learning models have been built.

"It will help identify the messaging that needs to be done," said Kompella.

Machine learning models can identify the types of tests people should take, whether to get booster shots and even whether when and where to wear masks.

"A model for public health campaigns is around identifying the messaging and the interventions [the public needs to take]," Kompella said.

Short- and long-term effects of AI on COVID-19

The impact of AI and machine learning on COVID-19 will have both short-term and long-term effects, Kompella said.

Initially, the impact will be limited, particularly on the healthcare side in using machine learning technology to diagnose the virus. However, over the long term it will significantly help advance healthcare technology to fight pandemics, he said.

In the early days of the pandemic, there was a lot of enthusiasm for using AI for diagnosis, prognosis and identifying and developing a drug for COVID-19, Kompella said.

"Literally, hundreds if not thousands of diagnosis models of COVID-19 have been developed," he said. "Similarly on the prognosis side as well."

However, almost all the models have proven to be of limited use in clinical settings, according to the British Medical journal.

That's because the models used low-quality training data, collected when hospitals were overwhelmed.

For example, a model that predicts whether a chest X-ray detects COVID-19 wasn't accurate because most of the data collected was based on a physician's opinion of what COVID-19 looks like on an X-ray. There wasn't enough information to determine whether the X-ray showed signs of pneumonia or COVID-19.

"All the lessons that we are learning are going to be useful to us in fighting the next pandemic and improving the practice of machine learning," Kompella said.

For that to happen, researchers and healthcare providers need to share data, create standardized formats for data collection and build transparent AI models.

Also, healthcare practitioners need to know more about how machine learning models work.

"Both the creators of these models as well as the actual people who will use those things, they need to improve knowledge of each other's disciplines," Kompella said.

While Final Mile's model isn't finished yet, the organization plans to keep updating it with more data once it is released. The company plans to showcase its vaccine hesitancy models to some key nongovernmental organizations once it's completed.

Dig Deeper on AI technologies

SearchBusinessAnalytics
SearchCIO
SearchDataManagement
SearchERP
Close