1
00:00:00,000 --> 00:00:02,159
Hi, everyone. This is Lucas Voss with Becker's

2
00:00:02,159 --> 00:00:03,839
Healthcare. Thanks so much for tuning in to

3
00:00:03,839 --> 00:00:06,879
the Becker's Healthcare podcast series, and welcome to

4
00:00:06,879 --> 00:00:09,139
today's episode from detection to prevention,

5
00:00:09,679 --> 00:00:11,539
AI's role in payment

6
00:00:11,919 --> 00:00:13,779
integrity. And I'm very excited

7
00:00:14,105 --> 00:00:16,105
to welcome Steve Southern. He's the senior vice

8
00:00:16,105 --> 00:00:18,204
president information systems at Cirrus

9
00:00:18,585 --> 00:00:20,904
for our conversation today. Steve has close to

10
00:00:20,904 --> 00:00:21,724
thirty years

11
00:00:22,024 --> 00:00:24,425
of experience in IT and joined Cirrus in

12
00:00:24,425 --> 00:00:25,704
1996.

13
00:00:25,704 --> 00:00:26,605
He's been passionate

14
00:00:26,980 --> 00:00:29,779
and very involved in proactively building efficiencies and

15
00:00:29,779 --> 00:00:32,899
leveraging technology to keep health care organizations ahead

16
00:00:32,899 --> 00:00:34,659
of the curve. Steve, I did not call

17
00:00:34,659 --> 00:00:36,600
you old. Welcome to the podcast.

18
00:00:38,340 --> 00:00:40,054
Thank you. Thank you, Lucas. Thank you very

19
00:00:40,054 --> 00:00:43,015
much for the introduction. Yeah. Thirty years, time

20
00:00:43,015 --> 00:00:44,695
flies when you're having fun as they say.

21
00:00:44,695 --> 00:00:47,254
Right? Exactly. I'm so excited to have you

22
00:00:47,254 --> 00:00:49,335
just because of all of the experience that

23
00:00:49,335 --> 00:00:51,655
you bring to our conversation today. And and

24
00:00:51,655 --> 00:00:53,034
I wanna start off with

25
00:00:53,619 --> 00:00:55,320
putting things into perspective

26
00:00:55,700 --> 00:00:56,600
for our audience.

27
00:00:57,619 --> 00:00:58,679
As we're looking

28
00:00:59,379 --> 00:01:02,280
at our topic today, how do you define

29
00:01:02,820 --> 00:01:05,479
payment integrity today, and how is that definition

30
00:01:06,019 --> 00:01:06,519
evolved

31
00:01:07,114 --> 00:01:09,515
as AI has become more embedded in payer

32
00:01:09,515 --> 00:01:10,015
operations?

33
00:01:11,275 --> 00:01:12,575
Sure. That's a great question.

34
00:01:13,355 --> 00:01:15,995
In today's environment, I would say payment integrity

35
00:01:15,995 --> 00:01:19,295
really spans the entire claim life cycle. So

36
00:01:19,840 --> 00:01:22,239
end to end set of processes from claim

37
00:01:22,239 --> 00:01:22,739
submittal

38
00:01:23,120 --> 00:01:25,120
all the way to final payment for in

39
00:01:25,120 --> 00:01:26,420
a prepayment setting

40
00:01:26,719 --> 00:01:28,180
as well as the traditional

41
00:01:28,959 --> 00:01:30,180
post payment reviews.

42
00:01:31,200 --> 00:01:33,540
Along the way, there are many tools, analytics,

43
00:01:33,680 --> 00:01:36,504
technologies out there used to ensure the health

44
00:01:36,504 --> 00:01:39,004
care payments are getting paid or getting adjudicated

45
00:01:39,144 --> 00:01:41,004
paid accurately compliant

46
00:01:41,625 --> 00:01:42,444
and timely.

47
00:01:43,545 --> 00:01:45,625
AI is obviously at the forefront of all

48
00:01:45,625 --> 00:01:47,405
of that. Just about every organization's

49
00:01:48,265 --> 00:01:49,724
strategy and road map

50
00:01:50,390 --> 00:01:52,629
are intertwined, wrapped around that, if you will,

51
00:01:52,629 --> 00:01:54,870
these days. So I feel we're really just

52
00:01:54,870 --> 00:01:56,870
starting to scratch the surface with the impacts

53
00:01:56,870 --> 00:01:59,109
that AI can and will have down the

54
00:01:59,109 --> 00:01:59,609
road.

55
00:01:59,989 --> 00:02:02,150
We're seeing we're already seeing some major impacts

56
00:02:02,150 --> 00:02:03,290
with process improvements,

57
00:02:04,545 --> 00:02:05,685
automation, obviously.

58
00:02:06,625 --> 00:02:08,645
However, I do think that some organizations

59
00:02:08,944 --> 00:02:11,425
might be taking a slightly more conservative approach

60
00:02:11,425 --> 00:02:13,685
or slower pace, let's say, to

61
00:02:14,224 --> 00:02:15,905
to roll some of these things out just

62
00:02:15,905 --> 00:02:18,004
due to governance, regulatory,

63
00:02:18,305 --> 00:02:21,000
data use challenges, just making sure those those

64
00:02:21,000 --> 00:02:22,139
boxes are checked.

65
00:02:22,840 --> 00:02:23,819
And you've seen

66
00:02:24,199 --> 00:02:27,560
so many trends evolve over your time in

67
00:02:27,560 --> 00:02:29,180
health care IT, and there's certainly

68
00:02:29,639 --> 00:02:32,199
a lot of change that's happening right now,

69
00:02:32,199 --> 00:02:35,715
especially 2025 and then now obviously in 2026.

70
00:02:36,194 --> 00:02:38,914
From your work with these organizations across the

71
00:02:38,914 --> 00:02:39,414
country,

72
00:02:39,955 --> 00:02:42,354
how has the integration of AI and machine

73
00:02:42,354 --> 00:02:42,854
learning

74
00:02:43,314 --> 00:02:47,155
changed the strategic approach then to payment integrity

75
00:02:47,155 --> 00:02:49,655
for leaders? And is there anything that surprises

76
00:02:49,715 --> 00:02:50,854
you about that process?

77
00:02:51,830 --> 00:02:53,669
Yeah. Sure. I think the biggest change over

78
00:02:53,669 --> 00:02:55,110
the last year or so has been the

79
00:02:55,110 --> 00:02:57,590
shift from a traditional post payment or a

80
00:02:57,590 --> 00:02:58,810
pay and chase model

81
00:02:59,430 --> 00:02:59,830
to,

82
00:03:00,230 --> 00:03:00,730
prioritization

83
00:03:01,189 --> 00:03:03,129
of of prepayment solutions.

84
00:03:03,830 --> 00:03:05,665
So as I mentioned previously,

85
00:03:06,365 --> 00:03:08,685
audits that are happening during the payment cycle

86
00:03:08,685 --> 00:03:10,224
as opposed to after,

87
00:03:11,085 --> 00:03:13,004
AI has really helped fuel this due to

88
00:03:13,004 --> 00:03:15,645
the acceleration of development tools and the speed

89
00:03:15,645 --> 00:03:17,965
at which some of these solutions sophisticated solutions,

90
00:03:17,965 --> 00:03:19,264
if you will, can be deployed.

91
00:03:19,800 --> 00:03:21,400
I would say I'm most surprised by the

92
00:03:21,400 --> 00:03:23,979
rapid adoption rate that we've already experienced.

93
00:03:24,280 --> 00:03:26,840
Everybody is using these tools. They're everywhere. You're

94
00:03:26,840 --> 00:03:27,900
hearing about them

95
00:03:28,360 --> 00:03:28,860
constantly.

96
00:03:29,560 --> 00:03:30,060
Also,

97
00:03:30,680 --> 00:03:32,594
on the flip side of that really is

98
00:03:32,835 --> 00:03:35,895
the significant amount of resources and effort required

99
00:03:35,955 --> 00:03:37,415
for oversight and governance.

100
00:03:38,115 --> 00:03:40,275
This is an area that sometimes can lag

101
00:03:40,275 --> 00:03:41,814
behind the development cycle,

102
00:03:42,354 --> 00:03:45,094
but it really cannot be minimized or ignored.

103
00:03:45,395 --> 00:03:47,890
You know, the question today is not is

104
00:03:47,890 --> 00:03:50,370
no longer are you using AI, it's how

105
00:03:50,370 --> 00:03:52,450
are you using AI and what are the

106
00:03:52,450 --> 00:03:55,330
guardrails and framework that you have established around

107
00:03:55,330 --> 00:03:55,830
it.

108
00:03:56,290 --> 00:03:58,550
Yeah. And you mentioned the the fast adoption

109
00:03:58,610 --> 00:04:00,629
rates and everybody sort of jumping

110
00:04:01,325 --> 00:04:03,885
to or on that trend, so to speak,

111
00:04:03,885 --> 00:04:05,724
but that doesn't necessarily mean there is an

112
00:04:05,724 --> 00:04:07,665
impact right away for a lot of organizations.

113
00:04:08,284 --> 00:04:10,685
Where are you seeing the most meaningful impact

114
00:04:10,685 --> 00:04:12,605
for AI today? And is there an example

115
00:04:12,605 --> 00:04:14,224
that sticks out to you

116
00:04:14,605 --> 00:04:16,625
where AI really surfaced

117
00:04:16,970 --> 00:04:19,470
patterns that traditional rule based systems

118
00:04:19,850 --> 00:04:22,350
that we've touched on likely would have missed?

119
00:04:23,529 --> 00:04:24,029
Definitely.

120
00:04:24,649 --> 00:04:26,729
The most meaningful impact I'm seeing today is

121
00:04:26,729 --> 00:04:27,949
the ability to convert

122
00:04:28,250 --> 00:04:30,110
and process very large,

123
00:04:30,894 --> 00:04:31,394
complicated,

124
00:04:32,014 --> 00:04:32,514
unstructured

125
00:04:32,814 --> 00:04:35,235
documents and and even datasets into

126
00:04:35,615 --> 00:04:36,675
useful structured,

127
00:04:37,055 --> 00:04:40,435
easily machine readable data, which can be processed

128
00:04:40,654 --> 00:04:42,675
much more efficiently and accurately.

129
00:04:43,535 --> 00:04:44,754
So I guess

130
00:04:45,169 --> 00:04:47,750
an example I would say is so traditional

131
00:04:47,810 --> 00:04:48,870
rules based systems

132
00:04:49,410 --> 00:04:52,790
usually only detect specific patterns. Right? Programmatic,

133
00:04:53,649 --> 00:04:54,149
deterministic

134
00:04:54,529 --> 00:04:58,085
algorithms that, you know, humans have have developed

135
00:04:58,384 --> 00:05:00,384
as where on the flip side, where machine

136
00:05:00,384 --> 00:05:02,384
learning and large language models are able to

137
00:05:02,384 --> 00:05:05,904
analyze ambiguous notes, narrative patterns across these very

138
00:05:05,904 --> 00:05:08,625
large datasets that those traditional rules would have

139
00:05:08,625 --> 00:05:09,125
missed.

140
00:05:09,759 --> 00:05:12,180
A great example of this is medical records.

141
00:05:12,720 --> 00:05:15,220
So these documents are typically very large,

142
00:05:15,839 --> 00:05:18,480
many, many pages, sometimes hundreds, if not thousands

143
00:05:18,480 --> 00:05:19,220
of pages

144
00:05:19,520 --> 00:05:22,259
in these documents, very cumbersome to handle,

145
00:05:22,964 --> 00:05:25,764
challenging to process, and for, you know, humans

146
00:05:25,764 --> 00:05:26,904
to to review.

147
00:05:27,365 --> 00:05:30,105
So using a model to summarize that information,

148
00:05:30,324 --> 00:05:32,964
extract key data points can really assist clinical

149
00:05:32,964 --> 00:05:33,464
auditors

150
00:05:34,004 --> 00:05:36,745
by increasing their efficiency, accuracy, consistency

151
00:05:37,045 --> 00:05:38,779
to go through this information.

152
00:05:39,879 --> 00:05:42,519
Yeah. And we're seeing already seeing that impact

153
00:05:42,519 --> 00:05:43,259
on organizations,

154
00:05:43,639 --> 00:05:45,720
I think. I I wanted to come back

155
00:05:45,720 --> 00:05:48,199
to the governance piece that you've mentioned and

156
00:05:48,199 --> 00:05:50,039
sort of a couple of related topics in

157
00:05:50,039 --> 00:05:51,720
regards to governance because I think this is

158
00:05:51,720 --> 00:05:52,939
a very important topic.

159
00:05:53,354 --> 00:05:55,754
Again, you've talked about it. Automation can improve

160
00:05:55,754 --> 00:05:58,235
speed and consistency. It's very crucial. But there

161
00:05:58,235 --> 00:06:01,995
is concern around fairness, false positives, provider trust.

162
00:06:01,995 --> 00:06:03,854
Again, we're sort of getting into that governance

163
00:06:03,914 --> 00:06:05,854
territory here. How can organizations

164
00:06:06,314 --> 00:06:07,454
balance efficiency

165
00:06:07,914 --> 00:06:10,750
with that clinical accuracy and credibility at scale,

166
00:06:10,750 --> 00:06:11,250
really?

167
00:06:12,189 --> 00:06:14,110
Yeah. I I think finding this balance is

168
00:06:14,110 --> 00:06:16,589
really one of the most urgent challenges that

169
00:06:16,589 --> 00:06:19,009
we face today in the payment integrity space.

170
00:06:19,310 --> 00:06:21,629
So just as with the development tools and

171
00:06:21,629 --> 00:06:22,289
the technology,

172
00:06:22,875 --> 00:06:24,814
the appropriate resources and prioritization

173
00:06:25,194 --> 00:06:26,254
must be dedicated

174
00:06:26,555 --> 00:06:28,814
specifically to these areas

175
00:06:29,355 --> 00:06:31,694
these areas as in, you know, the governance,

176
00:06:31,835 --> 00:06:32,335
security,

177
00:06:32,795 --> 00:06:33,775
those things. So

178
00:06:34,154 --> 00:06:36,095
building a strong governance program,

179
00:06:36,395 --> 00:06:36,895
transparent

180
00:06:37,275 --> 00:06:37,775
controls,

181
00:06:38,180 --> 00:06:39,879
always keeping humans in the loop,

182
00:06:40,259 --> 00:06:41,000
and continuous

183
00:06:41,300 --> 00:06:41,800
modeling,

184
00:06:42,660 --> 00:06:44,980
not all not only model monitoring, but, you

185
00:06:44,980 --> 00:06:47,860
know, solution monitoring to make sure, you know,

186
00:06:47,860 --> 00:06:50,259
of model drift and the things that are

187
00:06:50,259 --> 00:06:52,439
of concern are are at the forefront.

188
00:06:53,254 --> 00:06:55,254
I would say those are some of the

189
00:06:55,254 --> 00:06:57,254
keys the the main keys, I would say,

190
00:06:57,254 --> 00:06:59,915
to ensure trust and credibility in this space.

191
00:07:00,774 --> 00:07:03,095
Now turning some of these topics that you've

192
00:07:03,095 --> 00:07:05,995
mentioned into strategic elements. Right?

193
00:07:07,060 --> 00:07:10,100
What should payer IT and operations leaders be

194
00:07:10,100 --> 00:07:11,319
doing right now

195
00:07:12,100 --> 00:07:14,979
to prepare for that next phase of AI

196
00:07:14,979 --> 00:07:16,199
driven payment integrity?

197
00:07:17,220 --> 00:07:18,979
Sure. I can share a couple of, at

198
00:07:18,979 --> 00:07:21,365
least, points that I think are important here.

199
00:07:21,365 --> 00:07:21,865
So

200
00:07:22,324 --> 00:07:25,044
IT leaders should be assessing and improving their

201
00:07:25,044 --> 00:07:25,544
teams,

202
00:07:26,884 --> 00:07:28,824
teams from a staffing standpoint,

203
00:07:29,285 --> 00:07:29,785
experience,

204
00:07:30,245 --> 00:07:30,745
technology,

205
00:07:31,979 --> 00:07:33,360
governance, and data infrastructure,

206
00:07:33,660 --> 00:07:35,439
as we just talked about a little bit.

207
00:07:35,899 --> 00:07:38,879
AI success really hinges on high quality, accessible,

208
00:07:39,019 --> 00:07:41,919
interoperable data. So data is key here.

209
00:07:42,220 --> 00:07:45,360
So having a clear, concise, consolidated data foundation,

210
00:07:45,955 --> 00:07:48,115
again, is the most important thing when it

211
00:07:48,115 --> 00:07:49,415
comes to AI.

212
00:07:50,595 --> 00:07:53,335
Preparing to support the shift to prepaid solutions

213
00:07:53,715 --> 00:07:56,435
by proactively considering some of the core systems,

214
00:07:56,435 --> 00:07:57,095
the foundational

215
00:07:58,129 --> 00:08:00,149
applications and and data and workflows

216
00:08:00,850 --> 00:08:02,149
that would support that.

217
00:08:02,930 --> 00:08:03,430
Considering

218
00:08:04,209 --> 00:08:05,189
integrated ecosystems

219
00:08:05,730 --> 00:08:06,470
and prioritizing

220
00:08:06,850 --> 00:08:10,310
API integrations with partners over, you know,

221
00:08:10,675 --> 00:08:13,735
more traditional fragmented point to point solutions.

222
00:08:14,835 --> 00:08:17,395
And finally, building equity and fairness testing into

223
00:08:17,395 --> 00:08:18,215
model governance,

224
00:08:19,074 --> 00:08:21,014
always maintaining a human in the loop,

225
00:08:21,555 --> 00:08:24,935
creating clear and transparent documentation along with defensible

226
00:08:25,074 --> 00:08:27,149
policies and procedures to back everything up.

227
00:08:28,250 --> 00:08:31,310
The future of AI driven payment integrity isn't

228
00:08:31,610 --> 00:08:34,649
really just about more automation. It's about trusted

229
00:08:34,649 --> 00:08:35,149
transparent

230
00:08:36,009 --> 00:08:39,210
human aligned intelligence that's embedded across the entire

231
00:08:39,210 --> 00:08:40,269
claim life cycle.

232
00:08:41,034 --> 00:08:42,495
Three decades of experience

233
00:08:42,955 --> 00:08:44,335
in fifteen minutes.

234
00:08:44,794 --> 00:08:47,195
Steve, thanks so much for for being here.

235
00:08:47,195 --> 00:08:48,554
Thanks so much for your time. It's so

236
00:08:48,554 --> 00:08:49,534
great to have you.

237
00:08:50,154 --> 00:08:52,315
You bet. Hey. It's it's truly exciting time

238
00:08:52,315 --> 00:08:53,995
to be working in this industry and at

239
00:08:53,995 --> 00:08:56,820
the intersection of payment integrity and technology.

240
00:08:57,440 --> 00:09:00,000
We're watching AI reshape not just health care,

241
00:09:00,000 --> 00:09:01,700
but nearly every industry,

242
00:09:02,160 --> 00:09:04,740
and the pace of innovation is just accelerating

243
00:09:04,879 --> 00:09:05,379
daily.

244
00:09:05,919 --> 00:09:07,759
So the work we're doing now will define

245
00:09:07,759 --> 00:09:10,915
the next generation of accuracy, fairness, efficiency in

246
00:09:10,915 --> 00:09:12,795
the health care space. And I'm just grateful

247
00:09:12,795 --> 00:09:14,434
to be a part of the transformation and

248
00:09:14,434 --> 00:09:16,375
working with the Saris team and our partners.

249
00:09:16,674 --> 00:09:19,075
I also appreciate the opportunity to collaborate with

250
00:09:19,075 --> 00:09:21,794
Becker's and share these conversations, so I appreciate

251
00:09:21,794 --> 00:09:22,774
you having me.

252
00:09:23,269 --> 00:09:24,790
Absolutely. It's great to have you. Thanks for

253
00:09:24,790 --> 00:09:26,230
your time. And we also want to thank

254
00:09:26,230 --> 00:09:28,470
our podcast sponsor, Cerus. You can tune in

255
00:09:28,470 --> 00:09:31,029
to more podcasts from Becker's Healthcare by visiting

256
00:09:31,029 --> 00:09:34,649
our podcast page at beckershospitalreview.com.