Implementation of Unique Device
Identification Demonstration Project –
A Report from MDEpiNet
Joseph P. Drozda, Jr., M.D., F.A.C.C.
Director, Outcomes Research
Mercy Health
Mercy Health
34
4,396
36,917
185
4,659
1,235
$4.05
ACUTE CARE HOSPITALS
LICENSED BEDS
CO-WORKERS
PHYSICIAN PRACTICE LOCATIONS
MEDICAL STAFF MEMBERS
INTEGRATED PHYSICIANS
OPERATING REVENUE (Billions USD)
UDI Demonstration Project Aims
1. Implement a coronary artery stent UDIbased surveillance system in the EHR in a
multi-hospital system (Mercy)
2. Identify obstacles to implementation of UDI
in clinical information & to characterize the
effectiveness of interventions to overcome
them;
3. Assess the validity and utility of data
obtained from the EHR and incorporated
UDIs for purposes of post-market
surveillance
4
UDI Demonstration Timeline
• Began work on system design: April, 2012
• Identify 1 device and a multi-stakeholder group: Sep, 2012
• Develop IT infrastructure : Oct, 2012
• Processes & systems for surveillance: Feb, 2013
• Demonstrate surveillance capabilities: May, 2013
• Complete demonstration of surveillance capabilities: Oct,
2013
• Final report: Dec, 2013
Key Components of UDI Demonstration
•
•
•
•
•
•
Create Draft UDIs & associate with base
attributes in the FDA’s Global UDI Database
(GUDID
Create clinically meaningful supplemental
attributes to be stored in a reference database
Create UDI data flow through ERP to cath lab to
EHR to UDI data set
Create UDI fields in the CathPCI Registry
Perform studies to demonstrate validity and
reliability of data
Identify obstacles to incorporating UDIs into
EHR and explore solutions
6
The Expert Work Group
• The Expert Panel: Five interventional
cardiologists appointed in conjunction with
ACC and SCA&I
• “Ex officio” members
•
•
•
•
FDA representatives
Coronary Stent manufacturer representatives
HTG system representatives
NCDR representatives
7
Tasks for Expert Work Group
• Develop a constrained list of coronary stent clinical
attributes to supplement the GUDID attributes
(Expert Panel)
• Propose a permanent home for UDI clinical attribute
database (SUDID)
• Recommend a governance structure for the SUDID
• Develop a proposal for an organization and
processes for ongoing maintenance of the SUDID
8
Table 3: SUDID Clinical Attributes and Parameters
Attribute
Definition
Parameter
Length
Nominal length per manufacture specification
Fractional dimension in mm
Diameter
Non-conventional
Property
Nominal (inner) diameter per manufacturer
specification
Stent having nonconventional design,
variable or multiple length/diameter
parameters
Structural Material
Composition of principal structural element
Coating(s)
Non-Structural material covering surface of
structural element
Drug(s)
Active agent released from stent
Strut Thickness
Surface to Artery
Ratio*
Expansion Method
MRI Compatibility
Maximum nominal thickness of stent struts on
a radius from the center of the stent
Percentage of the surface area of the artery
covered by the stent at nominal expansion of
the stent
Method used to achieve nominal stent
deployment
MRI compatibility category per testing
Fractional dimension in mm
Covered stent
Bifurcation Stent
Tapered Stent
Constrained list
e.g. L605 cobalt chromium
-- Constrained list to be developed
Constrained list
-- Constrained list to be developed
--Need to handle multiples
--name that would be mostly referenced
--start with what is in the IFU
--accommodate multiple coatings
NDC directory (default)
--Use name if no applicable NDC code—do
it uniformly
Dimension in microns
Data Type
4 significant digits, w/1
precision
4 significant digits, w/2
precision
Alphanumeric
Alphanumeric
Alphanumeric
Alphanumeric
4 integer digits
3 significant digits, w/1
precision
Balloon
Self
4 categories per existing standard:
--Safe
--Conditional
--Unsafe
--Not tested
*This attribute was originally selected by the Expert Panel but subsequently withdrawn
SUDID = Supplemental Unique Device Identifier Database; IFU = Instructions for Use; NDC = National Drug Code; MRI = Magnetic Resonance Imaging
Alphanumeric
4 Categories
UDI Demonstration Project
High Level Architecture
Single EHR UDI Tracking System Data Flow
The Reference Data Set of Procedures
and Coronary Stents
• Patients and stents included in dataset if
found in any one of the following datasets
•
•
•
•
Cath Lab Data table in UDIR
Inventory Data table in UDIR
Billing Data table in UDIR
Registry Data from Registry submission
• Completeness of each data set measured
against the reference dataset
Unique Coronary Stent Procedure Counts (percent
of procedures in reference datasets)
St. Louis
Springfield
Rogers
(All Sites)
Registry Data
291 (88%)
566 (87%)
-
857
Cath Lab Data
281 (85%)
546 (84%)
236 (83%) 1063
Inventory Data
301 (91%)
608 (93%)
253 (89%) 1162
Billing Data
291 (88%)
565 (87%)
212 (75%) 1068
Reference
Dataset
330
653
283
1266
Patient Counts by Month and System for
St. Louis and Springfield
250
Patient Counts
200
150
CathPCI
Registry
Registry
Cath
Lab
Merge
Cath Lab
Optiflex
Inventory
Inventory
Billing
Billing
Billing
100
50
0
November
December
January
2012
February
2013
Month
March
Percent Agreement by System Combinations for St. Louis and Springfield
System
%
Registry Data
Cath Lab Data
Inventory Data
Billing Data
Y
Y
Y
Y
73%
Y
N
Y
Y
9%
N
Y
N
N
4%
N
N
Y
N
3%
Y
Y
Y
N
3%
N
Y
Y
N
2%
N
N
N
Y
2%
Y
Y
N
Y
1%
N
Y
Y
Y
1%
N
N
Y
Y
0.4%
Y
N
Y
N
0.3%
Y
N
N
Y
0.1%
Y
N
N
N
0.1%
N
Y
N
Y
0.1%
Y
Y
N
N
0.1%
Stent Counts by Site and System
(percent of stents in reference datasets)
St. Louis
Springfield
Rogers
Cath PCI
514 (76.3%)
877 (81.1%)
--
All
Sites
1391
Merge
603 (89.5%)
902 (83.4%)
412 (80.6%)
1917
OptiFlex
539 (80.0%)
929 (85.9%)
432 (84.5%)
1900
Billing
496 (73.6%)
768 (71.0%)
344 (67.3%)
1608
Reference
Dataset
674
1082
511
Coronary Stent Count Comparison at
Mercy Hospital St. Louis and Springfield
400
Stent Counts
350
300
250
200
CathPCI
Registry
150
Merge
Cath Lab
100
Optiflex
Inventory
50
Billing
Billing
0
November
December
January
2012
February
2013
Month
March
Overall Scan Compliance
Inventory Data
Rogers
Springfield
St Louis
Total
Cath Lab Data
Inventory/Cath Lab
744
856 86.9% (744/856)
1996
2279 87.6% (1996/2279)
897
1097 81.8% (897/1097)
3637
4232 85.9% (3637/4232)
Scan Compliance – by Month
Rogers
Inventory /Cath Lab (%)
100
95
93
92
80
95
90
90
83
85
96
85
85
77
75
70
65
60
Time
88
83
81
Scan Compliance – by Month
Springfield
Inventory/Cath Lab (%)
100
95
90
85
80
89
91
88
90
88
85
83
81
75
70
65
60
Time
93
89
88
85
Scan Compliance – by Month
Inventory/Cath Lab (%)
St Louis
100
95
90
85
80
88
84
82
85
81
78
80
79
75
70
65
60
Time
79
81
82
77
Scan Compliance – Non Regular Hour vs Regular Hour
Inventory/Cath Lab (%)
100
90
83.2 82.6
80
88.3
78.4 81.2
70
61.7
60
50
Non regular hour
40
Regular hour
30
20
10
0
Rogers
St Louis
Mercy’s Sites
Springfield
Scan Compliance – Emergent (AMI) vs Non-emergent
Inventory/Cath Lab (%)
100
90
87.1 86.1
81.4 83.8
86.8
90.6
80
70
60
50
Not AMI Emergency
40
AMI Emergency
30
20
10
0
Rogers
St Louis
Mercy’s Sites
Springfield
Options for Device Comparisons in the
UDIR
• Baseline patient characteristics
• Demographic
• Clinical (lab, diagnoses, etc.)
• Devices
• Product identifier +/- production identifier
• Attributes
• Outcomes (MACE)
• Individual
• Combined
Statistical Analyses
• Detect safety signal (a) with different drug attributes, (b)
Step 1 combined drug attributes
Step 2
Step 3
Step 4
Step 5
• Investigate MACE events in different follow-up periods
• Identify selection bias
• Reduce selection bias
• Examine match-pair samples
Case Selection Flow Diagram
2484 procedures performed
between 1-Nov-12 and 26-Oct-13
(total 2250 patients)
676 procedures excluded:
482 no stent attributes
189 blank drug attribute in
SUDID
5 invalid medical records
1808 procedures with 1963 stents
(total 1657 patients)
1640 procedures with 1772 stents
(total 1545 patients)
Stents:
1166/1545: Everolimus drug
176/1545: Zotarolimus drug
19/1545: Paclitaxel drug
184/1545: Bare metal
112 patients excluded:
73 had >= 2 stents with
different drug attributes at initial
procedure
39 had stents with different
drug attribute at subsequent
procedure
Step 1a:
Device attribute: Drug
Patient characteristics: All
Outcome: Mortality
Paclitaxel: 19 patients with 0 deaths
Everolimus: 1166 patients with 28 deaths
Zotarolimus: 176 patients with 7 deaths
Bare metal: 184 patients with 18 deaths
Mortality
Step 1b:
Device attribute: DES Combined
Patient characteristics: All
Outcome: Mortality
Drug eluting stent: 1361 patients with 35 deaths
Bare metal: 184 patients with 18 deaths
Mortality
Step 2: Investigate Mortality in Different Follow-up
Periods
Mortality (Total 53)
DES
BMS
P-value
30 days (N=1405, 35 deaths)
1.6% (20/1230)
8.6% (15/175)
<0.0001
60 days (N=1246, 40 deaths)
2.2% (24/1096)
10.7% (16/150)
<0.0001
90 days (N=1111, 43 deaths)
2.8% (27/982)
12.4% (16/129)
<0.0001
120 days (N=947, 45 deaths)
3.4% (28/832)
14.8% (17/115)
<0.0001
150 days (N=798, 49 deaths)
4.6% (32/702)
17.7% (17/96)
<0.0001
180 days (N=665, 50 deaths)
5.6% (33/586)
21.5% (17/79)
<0.0001
210 days (N=539, 51 deaths)
7.2% (34/472)
25.4% (17/67)
<0.0001
240 days (N=374, 52 deaths)
10.6% (34/322)
34.6% (18/52)
<0.0001
270 days (N=281, 52 deaths)
14.2% (34/240)
43.9z% (18/41)
<0.0001
300 days (N=202, 52 deaths)
20.4% (34/167)
51.4% (18/35)
0.0004
330 days (N=129, 52 deaths)
33.3% (34/102)
66.7% (18/27)
0.0035
360 days (N=61, 53 deaths)
85.4% (35/41)
90.0% (18/20)
NA
Step 3: Identify Selection Bias
Baseline Characteristics DES (n=1230)
(N=1405)
BMS (n=175)
P-value
Standardized Diff
(DES-BMS) %
Female
32.4% (398/1230)
32.6% (57/175)
0.9549
-0.43
Age > 65
53.1% (653/1230)
53.1% (93/175)
0.9894
0
Caucasian
95.9% (1177/1227) 92.5% (161/174)
0.0505
14.58
Married
69.6% (854/1227)
59.2% (103/174)
0.0069
21.95
Alcohol used (Y)
37.4% (440/1178)
32.9% (53/161)
0.2964
9.44
Illicit drug used (Y)
6.0% (66/1106)
10.9% (16/147)
0.0319
-17.69
Acute MI (Y)
35.0% (431/1230)
53.1% (93/175)
<0.0001
-37.08
Cardiac arrest (Y)
0.3% (4/1230)
1.1% (2/175)
0.1652
-9.61
Shock (Y)
1.9% (23/1230)
9.7% (17/175)
<0.0001
-33.84
COPD (Y)
12.9% (158/1230)
18.9% (33/175)
0.0339
-16.46
Diabetes mellitus (Y)
37.0% (455/1230)
27.4% (48/175)
0.0144
20.66
Dialysis (Y)
1.9% (23/1230)
0.6% (1/175)
0.3482
11.72
EF < 30% (Y)
2.0% (25/1230)
5.7% (10/175)
0.0078
-19.32
Step 4: Reduce Selection Bias
Baseline Characteristics DES (n=145)
(N=290)
BMS (n=145)
P-value
Standardized Diff
(DES-BMS) %
Female
33.1% (48/145)
34.5% (50/145)
0.9012
-2.96
Age > 65
46.9% (68/145)
42.8% (62/145)
0.5550
8.25
Caucasian
95.9% (139/145) 94.5% (137/145) 0.7853
6.55
Married
57.2% (83/145)
56.6% (82/145)
0.9056
1.21
Alcohol used (Y)
28.3% (41/145)
31.0% (45/145)
0.6998
-5.91
Illicit drug used (Y)
10.3% (15/145)
10.3% (15/145)
NA
0
Acute MI (Y)
53.1% (77/145)
52.4% (76/145)
0.9064
1.4
Shock (Y)
9.0% (13/145)
8.3% (12/145)
0.8343
2.49
COPD (Y)
20.7% (30/145)
20.7% (30/145)
NA
0
Diabetes mellitus (Y)
28.3% (41/145)
29.7% (43/145)
0.8971
-3.09
EF < 30% (Y)
4.8% (7/145)
6.2% (9/145)
0.7980
-6.14
Step 5: Examine Matched-pair Samples
BMS
Matched-pair
DES
Death
Alive
Total
Death
2
5
7
Alive
9
129
138
Total
11
134
145
McNemar’s Test
Statistic (1 degree of freedom)
Simple Kappa Coefficient
1.1429 (p=0.2850)
Kappa (ASE)
0.1735 (0.1348)
Test for Kappa = 0
Statistic (two-side p-value)
2.1494 (p=0.0316)
Step 1a:
Device attribute: Drug
Patient characteristics: All
Outcome: Myocardial Infarction (MI)
MI
Paclitaxel: 19 patients with 0 MACE MI
Everolimus: 1166 patients with 34 MACE MIs
Zotarolimus: 176 patients with 6 MACE MIs
Bare metal: 184 patients with 5 MACE MIs
Step 1a:
Device attribute: Drug
Patient characteristics: All
Outcome: Stent Thrombosis (ST)
ST
Paclitaxel: 19 patients with 1 MACE ST
Everolimus: 1166 patients with 27 MACE STs
Zotarolimus: 176 patients with 6 MACE STs
Bare metal: 184 patients with 4 MACE STs
Step 1a:
Device attribute: Drug
Patient characteristics: All
Outcome: Total Revascularization (TR)
TR
After excluding 25 patients who had unmatched procedure dates:
Paclitaxel: 19 patients with 1 MACE TR
Everolimus: 1147 patients with 113 MACE TRs
Zotarolimus: 172 patients with 17 MACE TRs
Bare metal: 182 patients with 18 MACE TRs
Obstacles
• Technical: Biggest problem so far is Merge
• Agreeing on:
• Industry-wide standards
• Device attributes
• Organizational infrastructure and support for
designing and maintaining a UDI system
• The business case for all stakeholders
38
UDI Phase 2
• Purpose:
Build a national network of UDI enabled EHR data
sets around national registries for device surveillance
and research.
• The UDI Alliance:
• HTG Health Systems (Mercy, Mayo, Geisinger,
Intermountain, and Kaiser)
• National medical societies and registries
(ACC/NCDR)
• Industry (Medtronic, Abbott, Boston Scientific
• Consumer groups/patient representatives
Coronary Stent Distributed Data Sharing
Network
Thanks!
Joseph P. Drozda, Jr., M.D., F.A.C.C.
Director, Outcomes Research
Mercy
14528 South Outer Forty
Chesterfield, MO 63017
314-628-3864
Mobile: 314-308-1732
Joseph.Drozda@Mercy.net
41