using data for evidence based decision making: a davies story › sites › himss365 › files ›...
TRANSCRIPT
1
Using Data for Evidence Based Decision Making: A Davies Story
Session 196, March 8th, 2018
DuWayne Willett, MD, MS, CMIO, UT Southwestern
Mujeeb A. Basit, MD, MMSc, ACMIO, UT Southwestern
Suzanne Fink, MSN, RN, CCRN-K, Lead Clinical Analyst, Cleveland Clinic
Kathleen Burns, DNP, APRN, ACCNS-AG, ACNS-BC,
Sr. Director Nursing Education, Cleveland Clinic
2
Conflict of Interest
Has no real or apparent conflicts of interest to report:
DuWayne Willett, MD, MS
Mujeeb A. Basit, MD, MMSc
Reported conflict: VitalScoutSM
is a proprietary algorithm and service Cleveland
Clinic is working to take to market. The purpose of this session is to articulate
conceptually how dynamic visualization can drive adaptive care and measurable,
tangible outcomes.
Suzanne Fink, MSN, RN
Kathleen Burns, DNP, APRN
3
Agenda• University of Texas Southwestern
• Cleveland Clinic
• Questions
4
Learning Objectives
• Explain why conforming a relatively small set of master dimensions (date, patient, provider, location, organization, diagnosis, treatment, etc.) enables analytic interoperability across a wide variety of data sources
• Justify providing data visualizations to front-line clinicians as well as managers in support of an enterprise initiative
• Articulate how dynamic visualization can drive adaptive care and measurable, tangible outcomes
5
UT Southwestern Health System
6
The Challenge
+ Every Department
= Enormous Amount of Work
Quality Metrics
7
Source
of Truth
Governing Functions
Data Quality
Data Approval
Compliance
Interoperability
Unified Architecture
Dictionary
Technology
• Executive Steering Committee
• Business Rules Workgroup
• Pre-approved and Approved Environments
• Kimball Star Schema
• Reusable Data Model
• Data Lake• Data Model
• Business Definitions• Technical
Documentation
• Power BI• Write-back to Epic
Data Warehouse Architecture
8
Applications1
99
5 E
MP
I an
d C
linic
al
Dat
a R
ep
osi
tory
20
02
Ep
icC
are
Am
bu
lato
ry2
00
3 C
lari
ty a
nd
HIM
20
04
Cad
en
ce a
nd
P
relu
de
20
05
MyC
har
t
20
08
Ep
icC
are
Inp
atie
nt,
W
illo
w,
and
ASA
P
20
09
Re
solu
te P
B a
nd
St
ork
20
11
Hai
ku/C
anto
, C
are
Eve
ryw
he
re, B
eac
on
20
12
Rad
ian
t, P
relu
de
, A
DT,
Re
solu
te H
B
20
13
An
est
he
sia,
K
ale
ido
sco
pe
, Ph
oe
nix
20
15
Nu
rse
Tri
age
, Sw
oo
p R
efi
ll W
izar
d
19
84
ID
X1
99
0 P
yxis
20
05
Xce
lera
20
08
CP
OE
20
10
Su
nQ
ue
stan
d C
oP
ath
20
11
Pio
ne
er
Rx
20
12
Ph
ilip
s iS
ite
20
14
In
tou
ch H
eal
th
Tele
stro
ke
20
15
Ph
arm
Trac
20
16
Pro
vati
on
and
C
love
rle
af W
eb
Serv
ice
s
20
17
db
Mo
tio
n
20
16
Lif
eIm
age
, N
eu
ro W
ork
be
nch
, R
adim
etr
ics
20
16
Ep
ic L
ink,
Vid
eo
V
isit
s an
d S
ingl
e B
illin
g O
ffic
e
Futu
re H
eal
thy
Pla
ne
t, C
up
id,
Be
ake
r
9
Architecture
65Epic Modules
Applications
241
10
Data Warehouse Cube
12Terabytes
Data
33Source
Systems
Fact
Dim
Dim
Dim
Dim
Dim
CubesCubes
Cubes
Research
Descriptive, Predictive & Prescriptive Analytics
BI To
ols
Qu
ery To
ols Self-Service
Ad-Hoc Reporting
Self-ServiceData Visualization
Dashboards
Standard Reports
Co
mb
ined
Too
ls
Epic
Write
-back
11
Quality Measure Reporting Framework
FACT_PatientQualityScorecard --1 row per patient per measure per reporting period per attributed provider
12
by
Date
by
Patient
by
Location
by
Diagnosis
by
Encounter
Type
by
Provider
by
Funding
Source
by
Sponsor
by
Academic
Dept.
by
Procedure
by
Measure
Professional
Charges Fact
Charge Amt.,
RVU
Grant Funding
Fact
Budget,
Expense
Patient
Measure
Fact
Numerator /
Denominator
Data Model and Analytic Interoperability
13
Data Model and Analytic Interoperability
Dimensions
Fact Grain (1 row per:) Metric Dat
e
Prov
ider
Patie
ntEn
coun
ter /
Enc
Typ
e
Clin
ical
Org
Uni
tLo
catio
nD
iagn
osis
Trea
tmen
t/ P
roce
dure
Med
icat
ion
Qua
lity
Mea
sure
Risk
Eve
ntCo
ncep
t
Encounter (Row Qty), Enc Qty, +/- Duration, LOS X X X X X X X X
Order (Test/Procedure) (Row Qty), Order Qty, +/- TAT X X X X X X (X) X
Results Result Qty, Value X X X X X X X X
Medication Order Order Qty, Cancelled Qty X X X X X X (X) X
Problem/Diagnosis Dx Qty X X X X X X X
OR Procedure OR Case Qty, Case Duration X X X X X X X X
Note Note Qty X X X X X X
Professional Charges Prof Charge Qty, Charge Amt, RVU X X X X X X X X
Hospital Charges Hosp Charge Qty, Charge Amt X X X X X X X (X) (X)
Hospital Account Hosp Acct Qty, Total Charges, Costs X X X X X X X (X)
Patient-Entered
Questionnaire AnswersAnswer Qty, Likert Value X X X X X
Custom Data Concept
Value Row Qty, Value X X (X) X
Clinical Decision Support
Alert DisplayAlert Display Qty, Override Qty X X X X X X
Clinical Quality Measure
ValueDenom Qty, Numer Qty, Exception Qty X X X X
Risk Score Risk Score X X (X) X
Observation (i2b2) Value X X X X X
Twenty Conformed Dimension Tables
Fift
y Fa
ct T
able
s
14
Item # 2016
Specialties 30
Registries 43
EHR Tools Built 111
Measures 163
Dashboards 32
Unique patients tracked on registries >60,000
Patient-reported outcome questionnaires completed
>2,100
Project recognized with Healthcare Informatics magazine’s Innovator Award for 2016, published at:http://www.healthcare-informatics.com/article/2016-healthcare-informatics-innovator-awards-first-place-winning-team-ut-southwestern
Item # currently
Registries 52 with CQMs; 81 total
Unique patients >80,000 actively managed;
222K on a registry
Pt-reported outcomequestionnaires
>8,500(by 3,000 pts)
Results
15
Rheumatoid Arthritis: Quality Measurement & Improvement
0
20
40
60
80
100
Mea
sure
Per
form
ance
Rat
e (%
)
DMARD Compliance
CDAI Assessment
CDAI Score Severity: LowDisease Activity (> 2.8and <= 10)
CDAI Score Severity: HighDisease Activity (> 22)
96.3 → 98.63.7 → 1.4 Defect
42.7 → 80.9
38.6 → 42.3
13.7 → 10.1
16
PowerBI Dashboard
17
Provider Dashboard
18
Clinic Medical Director & Quality Leadership Dashboard
19
Reducing Variation in Provider Clinical Practice
20
Summary
21
Summary
Plan
DoStudy
Act
Agile Modeling
Agile Development
As a ___, I want to be able to ____ so that ____.
Data Warehouse
Automated Error Detection
22
Question:Scenario: Two health care organizations are affiliating to form a clinically-integrated network (CIN), to accelerate their journey to providing and measuring high-value care to the patients they serve.
Question: Which of the following will be most helpful in creating analytic interoperability between the two organizations' data for a 360-degree view of each patient and for population health analytics?
A. Co-locate transaction data from both organizations on a single physical server
B. For key analytic dimensions (patient, provider, diagnosis, treatment), map local codes in each system to a conformed set of standard codes
C. Co-locate analysts working on the data in the same physical location
D. For key analytic fields, update each source system at both health care organizations to use the exact same descriptions/labels for the same data attributes (such as provider specialty: "Cardiology" vs "Cardiovascular Disease")
23
Additional Questions?
DuWayne Willett, MD, MS
Mujeeb A. Basit, MD, MMSc
Friendly reminder - please complete online session evaluation!
24
VitalScoutSM
| Cleveland Clinic
Patient Story
Our mission is to provide
better care of the sick,
investigation into their
problems, and further
education of those who
serve.
25
Nurse A Physician PCNA Nurse B
Assessment Diagnosis
Planning
Implementation
Evaluation• Communication• Chain of command• Transition to practice gap• Preponderance of technology• Workflow variance
Challenges Faced
26
Situational Awareness (SA)
Key Components:
• Awareness of information• Comprehension of its meaning• Projection of future status
Under stress, SA is adversely effected.
SA is difficult to measure : You can’t process what you’re not aware of!
27
VitalScout SM Program
Automated Early Warning System
VitalScoutSM
Screensavers
Shared ReportAccountability tools
and templates
VitalScoutSM
Cleveland Clinic © 2015. All Rights Reserved
28
DNR-CC – Comfort Care Only (designated by color purple)
*Nursing Staff can always escalate to higher level (i.e from green to red)
29
Life (Live) Screen “Saving” & Enhancing Situational Awareness
30
CSL DATA LAYER
EPICANALYTICS
DBIMAGING
APPLICATION APPLICATION
Clinical Data• Request
Routing• Data
Validation• Standardized
API• Composite
Services• Batching
Access Management• Security• Logging• App
Management• Reporting• Caching• Documentation
CSL BROKER LAYER
APPLICATION
31
* HIPAA de-identified information outside
of Epic (location only) with the screen
saver facilitates rapid recognition of patient
changes; full workflow remains within EMR
© 2017 Epic Systems Corporation. Used with permission.
32
Nurse A PhysicianPCNA Nurse B
Assessment Diagnosis
Planning
Implementation
Evaluation
VITAL SCOUT SCREEN SAVER
Shared Report
Algorithm
Documentation Tools
© 2017 Epic Systems Corporation. Used with permission.
33
Cumulative Patient Hours at High Acuity
0%
1%
2%
3%
4%
5%
6%
7%
8%
9%
10%
1 2 3 4 5 6 7 8 9 10 11 12
% P
atient
Hours
Month
Pre-Implementation
(2015-2016)
Post-
Implementation
(2016-2017)
34
VitalScout
Pre Post
Re
as
se
ss
me
nt
Tim
e
≤ 2
hours37.7% 83.9%
> 2
hours62.3% 16.1%
p = 4.67 x 10-92
Reassessment Times of High Acuity Patients
35
36
37
PATIENT STORY
38
Questions
With a CDS that is visible in real-time (i.e. screen saver), which is the best target audience?
A. Clinical providers
B. Non clinical Support Staff
C. Patients and Family Members
D. All of the above
39
Questions
Designing analytics for complex CDS cannot be accomplished until the CDS is live and initial data is collected (True/False)?
40
Additional Questions ?
Suzanne Fink, MSN, RN
Kathleen Burns, DNP, APRN
Friendly reminder - please complete online session evaluation!