Senate Community Affairs Legislation Committee, , 50:16 to 1:06:19
Senators inquire about the algorithms used to prioritize aged care access and the tracking of mortality data for those on waiting lists.
Topics: Artificial intelligence, Health
Speakers: Hon Bridget McKenzie, Greg Pugh, Sonja Stewart, Jasmine Snow, Blair Comley PSM, Cathy Milfull, Dorinda Cox, Rachel Blackwood
>> Thank you so much, Chair. I have a question that I'm asking on behalf of the member for Gibsland, Darren Chester. Uh but I think it's an issue that will concern a lot of Australians. Um two of his constituents, Frank and Valma, are 99 years of age. They still live together in their own home uh with a level two home care package down in East Gibsland. And after a recent medical incident, the couple was reassessed and they were told they'd be eligible for a higher package, but that the wait time would be between 11 and 12 months, by which time both of them will be over 100 years old. The couple was categorized as standard and have sought a reassessment for higher priority given their age and their increased needs. My questions are as follows. Once an assessment is made supporting a need for a higher level package, how is it prioritized? And would the fact that the applicants are still in their own home and 99 years old actually impact the prioritization of the increased Who's answering that?
>> Senator, can I confirm you're asking about essentially the prioritization system and how that would work?
>> Because I think Frank and Valma and anybody who has heard of their plight wants to know what's going on in CRA. That means we've got a 12 month wait to go from a level Who's going to answer it? How is it
Potentially it's worth then because there are several components to your question, Senator, that uh cuts across some some various areas. I might ask Miss Blackwood just to quickly cover off uh how the assessment process works and then what people >> I'm very happy with the assessment
process. They've been assessed. They've been assessed as being eligible for a higher care package. They've been told it's a 12 month weight by which both of them will be over aundred years old. So they rightly have said is there any way we can have us our package increase uh prioritized given things like our age and our increased needs after a medical incident. So my question is once an assessment is made supporting the need for a higher level package, how is it prioritized? What is the process?
>> Thank you for clarifying that. Uh Senator McKenzie will give the officials time to answer your question. Uh
>> so I'm happy to apolog So the support home priority system algorithm relies on six criteria to assign an older person to a priority level. the individual's assigned points out of a maximum of seven which are determined by adding together the points specified in section 875 of the age care rules. So they consider things such as does the individual live alone? Do they have a cognitive impairment? Is the individual homeless or at risk of homelessness uh and that's essentially what what feeds the information around their prioritization. I would say if the individual circumstances have changed at any point that we would encourage them to seek a reassessment.
>> Well, when I look at cognitive impairment, risk of homelessness, and do they live alone, they're a married couple that are about to turn They're not about to be homeless. They clearly don't have cognitive [snorts] impairment. They know to contact Aaron Chester for assistance. So on the list of things that impact a prioritization, none of that would have changed. Are they the three things?
>> No, apologize. I can come.
>> I want a comprehensive list because
Senator McKenzie, I believe that Miss Snow was still continuing her answer and so um maybe if we can get the comprehensive list um >> appreciate it >> that uh we might let her complete her
answer. So the other three criteria whether or not the individual is an ainal touristra islander person if the individual has a need for urgent access to ongoing inhome age care and has the individual waited at least 6 months for an assessment and resides in MM5 6 or 7.
>> Okay. When we're examining whether any of those criteria have been met for prioritization, is it um a specialist or do you use artificial intelligence to make a recommendation on prioritization?
>> It's not um artificial intelligence.
>> Sorry, I can't hear.
>> Sorry. Uh it's not artificial intelligence. It's an algorithm. And so what that uh does is it takes the outputs of the assessment uh basically the level of care that you're assessed as needing and then categorizes you into a class for which uh funding level you should receive.
>> So in other words, you have an a human assessor form a judgment against those six criteria high, medium, low or some equivalent and then it gets put into you know a weight for each of the six criteria. So in the case that you're talking about you have mentioned a number of times that Frank and Valerma are 99 as you see that's not a criteria the age um I think all of us who've dealt with the age care system know that the support needs of someone at particular ages can age can be a pretty poor proxy for support needs. you have very healthy 95 year olds and you have 75 year olds that have um you know very very acute needs. So they're the six criteria that would go into the vulnerability that would then lead to the question of assessment. >> And if I could sorry I just reinforce it's not AI it's essentially assessor makes a judgment against those six criteria. We use the term algorithm, but I think sometimes that can be misleading. That's, you know, in the oldfashioned way, it's like a scoring system that leads to try and ensure we have comparability across the weight lists.
>> And I was just going to add that the rules um that the secretary and um and others have talked about it were they were created by looking at uh 22,000 uh individual cases and uh reviewing those. So, it's based on um evidence. So you've got the clinical assessment, then the application of rules that look at have looked at 22,000 um cases to form
So I just want to be very very clear that it is um one of your staff Mr. Comley in the department that considers the merits of the case that that themselves determines the recommendation. It's not a suite of data that's put through a program um that says given the data points uh age no one's homeless no one's aboriginal tourist rate islander etc according to the criteria therefore our recommendation is this
>> we um we
>> I just want to be really clear
>> sorry um so in terms of the department we actually um work with assessors across Australia um then there's contracts for those assessor to do it in the community or in a hospital setting. So, it's um just to respond to your question, Senator, it's not technically the department. Of course, we've got grant agreements with assessors and you know, they've got training and manuals and um and regular contact and expectations around delivery, but it's not technically the department that makes the decision. Um >> commission them effectively,
>> right? So, we've got people out in the field that are providing recommendations back to the department about um prioritization of of packages.
>> Senator, if I may, and this might just go some way to the opening of the question where we were working out which was the best way to approach it. There are there are three parts to this. You've got an assessment workforce that Miss Stewart referenced. They rely on an integrated assessment tool that is a set of clinically validated independently verified tools that are embedded into one tool. The assessment workforce undertakes the assessment. Then an algorithm is run for the integrated assessment tool classification algorithm. And what that does is uses a set of rules to objectively assign um each participant to a funding class based on their clinical characteristics. It does that because we want to make sure that people who have a same or a similar level of acuity, needs, conditions uh are you know matched with similar levels of support. So you don't have uh the two folks that you've mentioned in your in your case study uh receiving one level and then two folks who are in exactly the same position receiving something marketkedly different. So it aims to achieve consistency and then as u snow and Mr. Comia have pointed out you've then got the support at home priority algorithm which then assigns you uh into uh your priority level for support at home funding and that's based on the criteria that are set out in the rules. So all of this has been um we we've got contracts in place with the assessment workforce. The tool the tools have been independently and clinically verified uh and the algorithms have been extensively trial on tens if not hundreds of thousands of uh test cases to make sure that the uh answers that are being provided are as cons as as consistent as they possibly can be. So um this was the culmination of years of work. This is not um uh as you know some suggestions might be the work of AI trying to curb government spending. This is about making sure that people can get access to the care that they need at the level that they're assessed at and in a consistent manner across all care and condition types.
>> Thank you so much. Mr. Q. Does the department track the number of people who qualify for a package but pass away before the support is provided?
>> It's go please.
>> In 2024 25 4,8 >> apologies. In 2024 25 4,812 people uh exited for reason of death from the MPS.
>> Yep. 25. Oh, 24 25. Obviously, we don't have 25 26. >> Do we have a year-to- date figure? >> At the last count, more than 100,000 older Australians were waiting for packages and another 100,000 waiting just to be assessed. So, the issue is fast approaching a national emergency. And that was a quote from uh Aging Australia chief executive Tom Simmons. Are these numbers accurate?
Those numbers are accurate. Yes.
>> Can you confirm that older Australians are actually dying whilst waiting for at home while they wait for inhome care packages to be actually delivered?
I would and I was given the the figure you know the passing of any older person across our country is um a very sad occurrence and we're delivering a program at the um for people who are elderly um and uh and the chief medical officer is here and I think it's better to attest to me that um you know as you age you uh you come across a range of different medical conditions and become unwell. So um you know the the by the very fact that we're dealing with an age care population um who who have got uh uh fragility and um medical issues there are people who are sadly uh passing
>> That's life. >> In the end of the day, but I I would argue and given the tens of thousands, hundreds of thousands of case studies of which your integrated assessment tool has been um able to have data input into having a couple living at home at the age of 19, I would suggest would be an outlier of the data. I don't think there'd be many situations um that that this type of um that this would be normal. I just wanted to a couple of more questions on the integrated assessment tool. Um and then I know Senator Ruston's got a whole lot what happens if a human assessor comes to a different determination around um the integrated assessment tool when it um comes to determining outcomes? Is there a a check and balance there? And how's that resolved if there's a
So if an assessment de we're giving clear advice to assessment delegates around the level of discretion that they have under the the law as it applies around determining ongoing home support classifications and priority levels. Uh there they need to accept the system generated outcomes on those classification and priority levels. We are closely monitoring the operation of those algorithms and receiving feedback on case studies from assessment organizations and other stakeholders. And in addition, if a client is uh not satisfied with their classification or prioritization outcome, they can lodge a reconsiderations request with the department.
>> Good. That was my second question. Um if you disagree, what sort of remedy is there? So lodging a reclassification. Um, I do know the member for Gibsland has written to the minister twice about this issue. Um, has the department received or responded to correspondence from the minister's office replying to Darren Chester?
>> Call that one particularly
>> the Coro register someone
>> could we take that on notice perhaps? I
would I'm happy to come back or for you to read it into Hansa because I know there will be someone here >> with a Coro register of Cororo that's come down from the min office that you've had to write the reply to sent back up. >> Um I'd like to know >> he's written twice. >> I'd like to know that it's been replied on. I often know sometimes the reply can get stuck in a minister's office. Um I'd like I'm not worried about that. It's a matter for her. Um I'd like to know if the department has sent the response back to the min office. >> Thank you.
>> We can absolutely find that out for you. Just to flag um because of the way that we separate um the assessment process with the reconsiderations rightly so. It's a different area of our group that looks after the reconsiderations process. So we can get that answer for you throughout the course of today.
>> Fantastic. Thank you. That that would be wonderful. Thanks Frank and Valma.
>> Thank you Senator McKenzie. Uh, Senator