presentation to the innovative member engagement conference oct 22 las vegas

18
Machine Learning and Patient Engagement @timgilchrist

Upload: tim-gilchrist

Post on 22-Jan-2018

157 views

Category:

Healthcare


0 download

TRANSCRIPT

Page 1: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

Machine Learning and Patient Engagement

@timgilchrist

Page 2: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

Big Data

• Many data sources with different formats

• Data with missing values

• Text / Social Media

• Things that don’t fit in Excel

The term for a collection of data sets so large and complex that they become difficult to process

Page 3: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

Artificial Intelligence

3

“Pay no attention to the man behind the curtain”

Page 4: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

Machine Learning

4

The construction and study of systems that can learn from data

Page 5: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

Where did it all Start? Bayes

Thomas Bayes (1701 – 7 April 1761) was an English mathematician and Presbyterian minister, known for formulating the theorem that bears his name: Bayes' theorem

• Bayes theorem uses prior probabilities, combined with new observations to calculate the probability of a hypothesis being true or false

• Bayes is a natural fit to health care due to the presence of hypothesis (diagnosis) and events (tests / observations)

5

Page 6: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

Bayes Example

6

33% 33% 33%10% 80% 10%

Page 7: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

How can we Apply Machine Learning in Healthcare?

7

Identify patterns that humans have trouble seeing

Population Health

Care Optimization

Precision Medicine

R&D Productivity

Page 8: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

What Does this Mean To Patient Engagement?

8

You are Here

Page 9: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

Regression

9

Co

st

ER Visits

Page 10: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

Classification

10

OrthopedicChest PainAbdominal Pain

Co

st

ER Visits

Page 11: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

ExampleSocial Media Text Mining / Diabetes

11

Page 12: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

Twitter Users Self-Described Diabetics

12

•What if you could identify “real” diabetics on twitter?

• You could engage them in diabetes education, etc.

• Cost = $0

• Know things that don’t show up in claims (latency)

• Possibly alert the undiagnosed

“Lets play a game called how many times

will my relatives ask about my diabetes.

#byyyyeeee”

Page 13: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

Results

• 73.5% Accuracy (ability to identify self-described diabetics from spam, people mentioning other people’s diabetes, retweets, bots, etc.)

•Variables in order of importance• #times others favorited tweets

• #followers

• #user statues

13

Page 14: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

Results / Decision Tree

14

# Favorites

# Followers

# Statuses

True (48% / 2%)

<=226 >226

True (44% / 21%)

<=1903 >1903

True (6% / 0%)

<=65.5

Page 15: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

ExampleMI Patients not Taking Beta Blockers

15

Page 16: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

MI Patients not Taking Beta Blockers

16

•What patterns exist in this population?

• You had a heart attack but not taking beta

blockers

• What can we learn to effectively reach these

people

• Are they homogeneous or are there sub groups

• Preemptive activities?

Page 17: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

Results

• 74% Accuracy (ability to predict compliance with rule – take beta blockers)

•Variables in order of importance• #primary diagnoses

• #Evaluation & Management visits

• Prior compliance with other rules

17

Page 18: Presentation to The Innovative Member Engagement Conference Oct 22 Las Vegas

Results / Decision Tree

18

NC = 68%SC = 25%UC = 6%

100%

1,2 0

NC = 42%SC = 46%UC = 12%

35%

NC = 82%SC = 14%UC = 3%

64%

#primary diagnoses

< 1 >= 2

NC = 86%SC = 13%UC = 0%

8%

NC = 29%SC = 55%UC = 15%

8%

E&M Visits

NC = Never CompliantSC = Sometimes CompliantUC = Usually Compliant