Session 6

Predictive Analytic Models: A Must in the Journey to Reducing Readmissions

Karen Tomes, RN, MA, PHNVice President, Care Management and CoordinationAllina Health

Learning Objectives

• Recognize the key elements of a care management redesign that help lower potentially preventable readmissions.

• Determine how predictive analytics can help risk-stratify patients to determine an optimal care transition plan.

• Recognize the role of education to engage members of the care team, patients, and families to lower readmissions.

We believe that patients deserve to receive the optimum level of follow-up care and

support after discharge from the hospital—which in turn reduces hospital readmissions

and costs.

Poll Question # 1

Is your organization facing penalties related to high readmission rates?

a) Yesb) Noc) Unsure or not applicable

20%in 30 days

33%in 90 days

$41.3 billion in additional

hospital costs

Much of this is avoidable

Approximately

Allina Health: Region’s Largest Health Care Organization

• 13 Hospitals• 82 Clinic sites• 3 Ambulatory care centers• Pharmacy, hospice, home care,

medical equipment• 26,000 employees • 5,000 physicians• 2.8 million+ clinic visits• 110,000+ inpatient hospital

admissions• 1,658 staffed beds• 3.4B in revenue• 32% Twin Cities market share

Allina Health is dedicated to the prevention and treatment of illness and enhancing the greater health of individuals, families, and communities throughout Minnesota and western Wisconsin.

Background

Potentially preventable readmissions (PPRs) are common and costly.

Hospitals need to manage PPRs to improve quality and avoid penalties.

Many factors along the care continuum influence PPRs.

Managing care transitions at the time of discharge is key to reducing PPRs.

PPRs are readmissions analytically determined to be related by 3M software outputs.

The Problem and Opportunity

Increasing focus on readmissions (state and national)

Lack of readily accessible and relevant data

Significant variation in case finding

Complex discharge needs

40% of readmissions occur within 7 days of discharge.

60% of Allina Health’s patients meet this criteria.

Resulting in poor care transition management at discharge for those who need it most.

Distributed across various locations and across multiple tools.

Patients Are Most Vulnerable in the First Week After Discharge

0

10

20

30

40

50

60

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Num

ber o

f Pat

ient

s Rea

dmitt

ed o

n Gi

ven

Day

Cum

ulat

ive

Read

miss

ion

Prop

ortio

n

Apr. 2015-Mar. 2016 Discharges

Medical SurgicalMedical Cumulative Surgical Cumulative

The Turning Point

1. Allina Health recognized that in order to reduce readmissions it needed to implement systemwide change.

2. Effective change would require a multi-pronged approach: • Care management process—how and at what points in the

care process. • Analytics—predictive models and measurement.• Education and awareness.

Results

10.3% 27% $3.7overall reduction in PPRs

• Reduction in all risk categories

• 4.3% reduction for high-risk patients

• 21.3% reduction percent for moderate/high-risk patients

reduction for patients with clinic follow-up within 5 days

reduction in variable direct costs

million

Results period: January – December 2015

Allina Health Achieved a Sustained Risk-Adjusted Reduction in PPRs

0.60

0.70

0.80

0.90

1.00

1.10

1.206-

2012

7-20

128-

2012

9-20

1210

-201

211

-201

212

-201

21-

2013

2-20

133-

2013

4-20

135-

2013

6-20

137-

2013

8-20

139-

2013

10-2

013

11-2

013

12-2

013

1-20

142-

2014

3-20

144-

2014

5-20

146-

2014

7-20

148-

2014

9-20

1410

-201

411

-201

412

-201

41-

2015

2-20

153-

2015

4-20

155-

2015

6-20

157-

2015

8-20

159-

2015

10-2

015

11-2

015

12-2

015

1-20

162-

2016

3-20

16

Act

ual-t

o-Ex

pect

ed P

PR R

atio

Month-Year

Actual/Expected PPR Ratio Lower Control Limit Mean Upper Control Limit

Allina Health Has Also Seen an Increase in the Care Transition Patient Experience Measure

50

55

60

65

70

75

80

Jan2015

Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan2016

Feb Mar Apr May Jun Jul Aug Sep Oct Nov YearEndGoal

Perc

entil

e

Percentile

The Case for Redesigning Care Management

At Allina Health:• 5% of patients are at a high

risk for a readmission.

• 17% of patients are at a moderate-high risk for a readmission.

• 60% of patients have complex discharge needs based on a combination of clinical, social, and post-acute needs.

Patient Census Analytics Dashboard

How We Achieved Our Results

Care Process Management Process RedesignAnalytics and Predictive Models

Education and Awareness

Care Management Process Redesign

Created complex discharge planning (CDP) teams: Partner with care management team to

determine patient needs during hospital stay and at discharge.

Develop and execute coordinated care management discharge plans.

Conduct care management transition conferences.

Care Management Process Redesign

Care Management Transition Conferences (TC)• Facilitated meeting to outline care management plan

post discharge: Patient, patient’s family, RN care coordinator, and physician.

• Post-acute partner attends when indicated. After TC, care management coordinates care plan and

provides ongoing updates. From April 2012 to March 2016, Allina Health has completed

11,969 TCs.

• Continuous improvement of TC process: Feedback from involved clinicians to identify issues and

barriers.

Analytics and Predictive Models

• Model utilizes EHR data to identify and stratify patient populations into risk categories.

• Predicts a patient’s risk of readmission within 30 days. Risk level (high, moderate-high,

moderate, low) based on: • Prior utilization of Emergency

Department and inpatient services• Medical history• Demographic information• Current clinical data

Developed a predictive model

The Predictive Model Stratifies Patients Accurately When Analyzing Historical DataAllina Health, Potentially Preventable Readmission Rate by Risk Level, Apr. 2015-Mar. 2016 Discharges

0

200

400

600

800

1000

1200

1400

1600

1800

0%

2%

4%

6%

8%

10%

12%

14%

16%

High Moderate High Moderate Low

Rea

dmis

sion

Vol

umes

Pote

ntia

lly P

reve

ntab

le R

eadm

issi

on

Rat

e

30-Day PPR Rate 7-Day PPR Rate Readmission Volumes

Analytics and Predictive Models

• Enables point-of-care identification of patients who might benefit from a TC. Risk levels from predictive model appear in

application 24-48 hours after admission.

• Enables Allina Health to track effectiveness of programs and interventions on readmissions. Based on key clinical variables, they are

able to compare outcomes of patients who received TC against those who did not.

Incorporated readmission metrics into organization-wide scorecard.

• Plan to incorporate risk model into EHR.

Implemented an analytics platform

Measures of Caring ScorecardReadmissions metric added to increase systemwide visibility

Example of Evaluating Effectiveness Based on Risk StratificationNumber needed to treat with a follow-up appointment to avoid readmissions decreases when evaluated at the low readmission risk level (2015 discharges)

14 14 17

90

0102030405060708090

100

High Risk Moderate-HighRisk

Moderate Low

Patie

nts

Seen

Number Needed to Treat with Follow-up Appt Within 5 Days to Prevent a 30-Day Readmission for

Surgical Hospital Discharges

Means, for example,

that 12 high-risk medical patients or 14 surgical

patients need to

typically be seen five days from

discharge to prevent 1

readmission.

12

22

27

35

0

5

10

15

20

25

30

35

40

High Risk Moderate-HighRisk

Moderate Risk Low Risk

Number Needed to Treat with Follow-up Appt Within 5 Days to Prevent a 30-Day

Readmission for Medical Hospital Discharges

Poll Question # 2

Does your organization currently use predictive analytics to drive outcomes improvement initiatives?

a) Yesb) Noc) Unsure or not applicable

Success required active engagement of patients and care management teams.• Used TC to educate and support patients,

family, and outpatient caregivers. • Supported with standardized and easy-to-

understand materials.

• Internal education and marketing plan to engage care teams.• Newsletter, intranet site provide program material

and updates.

Education and Awareness

Summary of How We Achieved Our Results

Using care management process redesign, predictive analytics, and education, we

were able to significantly reduce PPRs and save costs.

Lessons Learned

• Predictive analytics and a readmission risk model are effective in identifying and managing patients at all readmission risk levels.

• Interventions need to be multi-pronged to impact all risk levels.• Example: follow-up appointments for all patients

• A standardized, interdisciplinary care management transition conference is very effective in engaging patients and families at high risk of a readmission.

• Changing the culture requires a lot of heavy lifting.

Future Plans

1. Expand readmissions program and hardwire the process into the care management model to increase reliability and effectiveness.

2. Continuously refine predictive analytics to improve capabilities. Add readmission risk scores to electronic medical record so caregivers have the information front and center.

3. Integrate other sources of data into patient census analytics application.• Claims data

• Patient and family data (bi-directional)

Analytic Insights

AQuestions &

Answers

28

What You Learned…

Write down the key things you’ve learned related to each of the learning objectives

after attending this session

Thank You