Algorithm automates ACC’s claims process

Return to eHealthNews.nz home page

Picture: Nicholson Consulting senior data scientist Ernestynne Walsh

eHealthNews.nz editor Rebecca McBeth

The Accident Compensation Corporation is using an algorithm to make decisions in its claims process, reducing the time for cover to be confirmed from weeks to seconds in some cases.

Nicholson Consulting senior data scientist Ernestynne Walsh spoke about the automation programme at AI conference Hack Aotearoa held at the Auckland Business School on January 18–19. ACC was a major sponsor of the event.

She said ACC used to manually process more than two million claims every year, which required 60 staff.

However, 90 per cent of ACC claims are considered straightforward and therefore provided the opportunity for automation.

As part of a client transformation programme, Nicholson Consulting worked with the organisation to develop predictive models for these claims and the new algorithm went live in September last year.

It uses data from 12 million anonymised claims lodged between 2010 and 2016 to determine the probability that a given claim would be accepted.

The remaining 10 per cent of claims are more complex, such as sensitive claims, where staff members with expertise need to deal with them.

Any decision to deny cover will always be reviewed and dealt with by a staff member.

Walsh said that in the past claimants could wait weeks for a letter in the mail to say ACC would cover the cost of their treatment.

“Thanks to the services of predictive models, a doctor can now hit submit on the practice management system in a GP surgery and the client could receive a cover notification via text within seconds,” she said.

Walsh said the benefits are that clients can get a decision on cover while they are still with their medical provider and that the system can deal with high volumes of claims, such as occurred after the Christchurch earthquake.

The system is also more transparent, allowing clients to log on and access information about their claim online.

“It’s an algorithm doing automated decision making without a human looking over the decision,” she told conference attendees.

“This could set the scene for other areas of government.”

Walsh’s team also worked with ACC on an accident description service, which automatically converts free text from the accident description on claim forms into structured data.

This creates more consistency in the data and makes it easier for other teams, such as injury prevention, to make use of it.

“Both services could easily be transferred to other areas of healthcare where automation could come in handy,” Walsh said.

Return to eHealthNews.nz home page