Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
9/5/2016
1
DATA ANALYTICS IN THE CLINICAL LABORATORY: ENHANCING DIAGNOSTIC VALUE AND REDUCING COSTS
Brian Jackson, MD
Jason Baron, MD
Anand Dighe, MD, PhD
DISCLOSURES
• Brian Jackson– ARUP Laboratories (Nonprofit entity of University
of Utah): Consultant/Salary
• Jason Baron– Research support to his instituion (MGH) from
IBM
• Anand Dighe– No relevant disclosures
Data Analytics in the Clinical Laboratory Part 1:Measurement Philosophy
Brian Jackson, MD, MS
VP, CMIO, ARUP Laboratories
Assoc Prof of Pathology (Clinical), University of Utah
Flexner Report (2010)
• Established science as foundation of medical training
• Created cultural divide between physicians (science) and administrators (business)
9/5/2016
2
Myths: Business versus Medicine
• Myth #1: Business management is primarily about money
– Reality: Business management is primarily about organizational effectiveness
• Myth #2: Mixing medicine and business is “dirty”
– Reality: If we really care about our patients, we’ll use all useful tools to improve the effectiveness of our healthcare organizations
Is your laboratory successful?
Prove it.
Ideal laboratory metrics:
• Support learning and improvement
• Summarize major dimensions of
performance
• Be balanced (unbiased)
• Include both outcomes (past performance)
and processes (future performance)
3 Key Questions
• How much are we accomplishing?
• What does it cost us to accomplish it?
• How reliable are our processes?
9/5/2016
3
3 Key Questions
• How much are we accomplishing?
– Most businesses: Revenue
– Value-based healthcare: Patient benefit
• What does it cost us to accomplish it?
– Resources ($)
• How reliable are our processes?
– Quality
Organizational Learning and Improvement
• Senior leadership needs:
– Big picture
– Representative
– Don’t get lost in the details
• Front line needs:
– Directly tied to day-to-day activities
– Under their control
Putting it all together (Factory)
Quality Revenue Costs
Key rollupmeasure(s)
Sales Total costs
Defects by category
Market shareRevenue by segment
LaborSuppliesDepreciation
Defects by process
Revenue per sales rep
Detailedcosts per department
Executive
Front line
Putting it all together (Clinical Laboratory)
Quality Patient Benefit Costs
Overall health system reliability
Global benefit Total cost to lab
Per clinical practice unit
Benefit per test Cost per test
Per test:• TAT• Accuracy• Process
quality
Benefit per case• Variation• Consistency
with guidelines• Consistency
w/expert opinion
Cost per case
Executive
Front line
9/5/2016
4
Where Do We Have Good Metrics Today?
Executive
Front line
Quality Patient Benefit Costs
Overall health system reliability
Global benefit Total cost to lab
Per clinical practice unit
Benefit per test Cost per test
Per test:• TAT• Accuracy• Process
quality
Benefit per case• Variation• Consistency
with guidelines• Consistency
w/expert opinion
Cost per case
Where Are the Opportunities?
Executive
Front line
Quality Patient Benefit Costs
Overall health system reliability
Global benefit Total cost to lab
Per clinical practice unit
Benefit per test Cost per test
Per test:• TAT• Accuracy• Process
quality
Benefit per case• Variation• Consistency
with guidelines• Consistency
w/expert opinion
Cost per case
Quality Patient Benefit Costs
Overall health system reliability
Global benefit Total cost to lab
Per clinical practice unit
Benefit per test Cost per test
Per test:• TAT• Accuracy• Process
quality
Benefit per case• Variation• Consistency
with guidelines• Consistency
w/expert opinion
Cost per case
Managing Diagnostic Test Utilization
Executive
Front line
Total Cost of Laboratory Operations
• Labor
• Reagents
• Instruments
• Facility overhead
– Space, utilities, IT, etc.
9/5/2016
5
Cost per Test
• Proper Approach
– Labor, reagents, instruments, overhead
• Do not use 3rd party fee schedule!
• Do not use chargemaster!
Cost per Case
• Assumes you have valid costs at component level
• Overhead allocation is tricky
• Dependent on the clinical algorithms
Harvard Business Review Sept 2011
Managing Diagnostic Test Utilization
Executive
Front line
Quality Patient Benefit Costs
Overall health system reliability
Global benefit Total cost to lab
Per clinical practice unit
Benefit per test Cost per test
Per test:• TAT• Accuracy• Process
quality
Benefit per case• Variation• Consistency
with guidelines• Consistency
w/expert opinion
Cost per case
9/5/2016
6
Global Measures of Healthcare Quality?
Program # Measures # Diagnostic # Lab
HEDIS 74 20 9
CMS ACO 33 13 4
Choosing Wisely
135 90 21
Patient Benefit per Test
• Function of how the test is used
– NOT an intrinsic quality of the test itself
• Example: H pylori testing
– Do stool Ag and breath test provide more patient value than serology?
– Answer: Depends on the rate of endoscopy
Holmes et al. BDM Health Services Research 2010, 10:344
Patient Benefit (of a Test) per Case
• Outcomes
– Generally not practical in this setting.
• Normative (Evidence Based Medicine)
– Guidelines
– Other clinical literature
– Local expert opinion
• Non-normative/Descriptive
– Variation
Managing Diagnostic Test Utilization
Executive
Front line
Quality Patient Benefit Costs
Overall health system reliability
Global benefit Total cost to lab
Per clinical practice unit
Benefit per test Cost per test
Per test:• TAT• Accuracy• Process
quality
Benefit per case• Variation• Consistency
with guidelines• Consistency
w/expert opinion
Cost per case
9/5/2016
7
Data Analytics in the Clinical Laboratory Part 2
Using Analytics to Guide Operational and
Clinical Decisions
Jason Baron, MD
Medical Director, Core Laboratory
Massachusetts General Hospital
Assistant Professor
Harvard Medical School
Overview
Lab Informatics in Quality Management
1. Using metrics to guide clinical and operational decisions
• Ask useful questions
• Maintain usable data
• Address the question and appreciate limitations
• Variation Analysis
• Take a panoramic view
2. Computational Pathology and emerging
• Example: Spurious glucose identification
• Example: Computational Pathology and ferritin prediction
Overview
Lab Informatics in Quality Management
1. Using metrics to guide clinical and operational decisions
• Ask useful questions
• Maintain usable data
• Address the question and appreciate limitations
• Variation Analysis
• Take a panoramic view
2. Computational Pathology and emerging
• Example: Spurious glucose identification
• Example: Computational Pathology and ferritin prediction
Metrics should be designed around a
specific question
• Effective metrics are designed purposefully and thoughtfully–
to answer a specific question
• Excessive use of metrics not designed around key questions
may just lead to data overload without real added value
9/5/2016
8
Areas to ask questions
Domains
• Operational
• Analytic and pre-analytic performance
• Financial performance
• Clinical performance
• Utilization management
• Computational pathology
Circumstances
• Before vs. after an intervention
• Trends over time
• Comparisons to benchmarks
• Goal-based
• Standards-based
Operational Metrics Examples
Examples
• Turn around time
• Tech productivity
• Results per hour
• Can be broken down in many ways (shift, section, etc.)
Collect to result time (days) for a test
Analytic and Pre-analytic Performance
Examples
Examples
• Result distribution
• Frequency of anomalous results reported
• Corrected results
• Frequency of outlier results
Plasma Potassium Example
Example : Babesia Serologies by Physician
0
20
40
60
80
100
120
A B C D E F G H I J K L M N O P Q R S T U V W X Y Z a b c d e f g h
Test
s
Provider
Positive
Negative
9/5/2016
9
Example: Goal and Standards based
Metrics
March April
ER 83 84
IP 87 87
OP 90 89
Percent of Specimens
with a Valid Collect Time
• Can be metrics that have a
right answer—or best
performance
• E.g. no specimens received
without a valid collect time or
no corrected reports is optimal
• Can set realistic “red, yellow,
green” or other cutoffs to
define realistic goals
Overview
Lab Informatics in Quality Management
1. Using metrics to guide clinical and operational decisions
• Ask useful questions
• Maintain usable data
• Address the question and appreciate limitations
• Variation Analysis
• Take a panoramic view
2. Computational Pathology and emerging
• Example: Spurious glucose identification
• Example: Computational Pathology and ferritin prediction
Data: The Lifeblood of Analytics and Metrics
Raw
DataMetrics
Question Information
Proper Interpretation
•Metrics and analytics require a source of data
•Obtaining this data is often a frustration for laboratories
•Laboratories should advocate for direct data access
Key: Data Accessibility
Text
File
LIS
Nightly
reports
Datamart
Digest and
Import
Pathologist
ODBCSQL
Queries
•Data readily available
•Can easily link data from multiples sources for analysis
EHR
Nightly
reports
9/5/2016
10
Overview
Lab Informatics in Quality Management
1. Using metrics to guide clinical and operational decisions
• Ask useful questions
• Maintain usable data
• Address the question and appreciate limitations
• Variation Analysis
• Take a panoramic view
2. Computational Pathology and emerging
• Example: Spurious glucose identification
• Example: Computational Pathology and ferritin prediction
Metrics by themselves are just numbers
• Need to ensure that the metric truly addresses the intended question/hypothesis/problem
• Almost all metrics are subject to limitations
• Standard must not be perfection
• Remember that not everything is measureable
• Be aware of key limitations
Confounding and Effect Misattribution
• Correlation ≠ causation
• Frequently, metrics are designed to assess the impact of an initiative
or track performance over time
• Generally multiple factors vary before/after an initiative or over time
• Changes in a metric may be misattributed to the wrong cause
• Normalization
– E.g. tests per visit or tests per admission
• Adjust for seasonal variation
• Leverage statistics
• Sometimes it is sufficient to be aware of and accept limitations
Strategies to Controlling Confounding
9/5/2016
11
Seasonality Example
•Initiative in the summer of 2014 to restrict babesia serologies
•Although maybe these data speak for themselves, capturing the full effect
requires considering seasonality
Overview
Lab Informatics in Quality Management
1. Using metrics to guide clinical and operational decisions
• Ask useful questions
• Maintain usable data
• Address the question and appreciate limitations
• Variation Analysis
• Take a panoramic view
2. Computational Pathology and emerging
• Example: Spurious glucose identification
• Example: Computational Pathology and ferritin prediction
Variation Analysis: An Important Time to Consider
Confounding• Test ordering patterns can be compared between physicians and leveraged as a
utilization management tool
• Can be used as a tool both to identify utilization improvement opportunities and to encourage thoughtful utilization
• Clinicians identified as outliers may adjust test ordering practice to better mimic colleagues
• Important to qualitatively or quantitatively account for factors that appropriately impact utilization (e.g. subspecialty or patient mix)
Inter-Specialist Variation in Sendout Costs
9/5/2016
12
Sendout Adjusted Expenditure, Accounting for Diagnosis Other Variation Metrics
Yield Analysis
•Variation analyses can look at metrics besides cost
•Examples
•Counts
•Yields
•Appropriateness
•Outcomes
Remember
• Not all outcomes can be measured
• Some metrics are qualitative
Overview
Lab Informatics in Quality Management
1. Using metrics to guide clinical and operational decisions
• Ask useful questions
• Maintain usable data
• Address the question and appreciate limitations
• Variation Analysis
• Take a panoramic view
2. Computational Pathology and emerging
• Example: Spurious glucose identification
• Example: Computational Pathology and ferritin prediction
9/5/2016
13
Ensure Metrics are Balanced
• Management often entails balancing competing priorities
• It is important to track competing goals within the lab in parallel
• Otherwise management initiatives may too strongly favor one goal at the expense of others
Cost
Reduction
Quality
Improvement
NB: Cost and quality are not necessarily
in competition but they can be
Potential examples of balancing
Cost Quality/ Service
Full time Staff Overtime expense
Utilization reductions Clinical outcomes
Capital expenses Operating expenses
Data accessibility Data security
Important to consider downstream effects
Overview
Lab Informatics in Quality Management
1. Using metrics to guide clinical and operational decisions
• Ask useful questions
• Maintain usable data
• Address the question and appreciate limitations
• Variation Analysis
• Take a panoramic view
2. Computational Pathology and emerging
• Example: Spurious glucose identification
• Example: Computational Pathology and ferritin prediction
Raw
Most Processed
“Given this patient’s test results and clinical data,
administration of vancomycin will improve odds of
survival from 37% to 87%.”
“EBV-positive immunoblastic reaction,
consistent with infectious mononucleosis”
Present
Comp
Path
Atomic data
Interpretive
comments
Diagnoses
Integrated
Information
Predictive
Information
103127
3.8
101
21 1.86
19
9/5/2016
14
Overview
Lab Informatics in Quality Management
1. Using metrics to guide clinical and operational decisions
• Ask useful questions
• Maintain usable data
• Address the question and appreciate limitations
• Variation Analysis
• Take a panoramic view
2. Computational Pathology and emerging
• Example: Spurious glucose identification
• Example: Computational Pathology and ferritin prediction
Computational Derivation of Knowledge: Example
Spurious Glucose Identification
Spuriously
elevated glucose
result
•Commonly problem at many hospitals
•We were seeing spurious critically elevated glucose
results about once per day
•Fewer than 10% of these spuriously elevated
glucoses were being identified
Goal: Develop an Algorithmic Protocol to Distinguish
Spurious from Real Critically Elevated Glucose Values
Spurious Glucose Identification, Methods
Annotated Training Data (glucose >500 mg/dl)
Patient Glucose Na Additional Predictors
(K, CO2, AG, etc.)
Gold Standard Annotation
A 670 119 … Spurious
B 710 141 … Real
C 721 138 … Real
… … … … …
Supervised
Machine Learning
(Recursive
partitioning
decision trees)
Decision Tree
Test data or Un-annotated patient data
Patient Predictors
(glucose, Na, K, AG, etc.)
X …
Prediction as to whether
result is real or spurious
AJCP 2010 133:860
Tree built Using: Na, K, Cl, Bicarbonate, Anion Gap,
Glucose, and 30 day mean glucose
Spurious Glucose Identification, Results
Training Data
Test Data
Spurious
Correctly
Classified57 32
Total
Spurious61 37
Sensitivity
(95% CI)93%
(84-98%)86%
(72-95%)
Real
Correctly
Classified68 5
Total Real 77 6
Specificity
(95% CI)88%
(79-9%)
83%
(42-99%)
AJCP 2010 133:860
9/5/2016
15
Overview
Lab Informatics in Quality Management
1. Using metrics to guide clinical and operational decisions
• Ask useful questions
• Maintain usable data
• Address the question and appreciate limitations
• Variation Analysis
• Take a panoramic view
2. Computational Pathology and emerging
• Example: Spurious glucose identification
• Example: Computational Pathology and ferritin prediction
Example: Ferritin Prediction
We use Ferritin as an analyte of focus in an early proof-of-
concept
•Ferritin
•A marker of iron stores
•Used in the diagnosis of iron deficiency
•must be interpreted in the setting of other clinical and
laboratory data
•thus a good problem for data integration
•Decreased in iron deficiency
•Increased in inflammation
Ferritin Methods OverviewRaw Data (3 months outpatient
ferritin values)
•Transform ferritin values
•Divide at random into training and
test partitions (7:3 ratio)
•Mask ferritin results for test partition
Raw Data with
ferritin masked for
test patients
•Impute missing data
•Use 4 different imputation
methods
“Completed”
dataset
Performance
Metrics
•Predict ferritin results
•4 regression methods
•1 classification method
•Pair each with each
imputation method
Predicted ferritin
results and
classifications
•Compare predicted ferritin
results to measured results
•Compare predicted to
“masked” results for test
partition
•Review selected cases to
determine clinical
significance
AJCP (2016). 145:778-88.
Ferritin Classification PerformanceTest Data
Negative Control
AJCP (2016). 145:778-88.
9/5/2016
16
Ferritin Case Review
Test dataset (N=1538)
Ferritin
Results
Predicted
Ferritin
Results
Ferritin result differs
from prediction by a
factor of 10 or more
Highly discrepant
results
N=26 (1.7%)
4 Cases
Predicted ferritin
≤30 ng/ml (male ref
limit)
Case Ferritin Predicted
Ferritin
Impression Comment
1 230 21 Iron deficiency,
not clinically
identified
Ferritin increased
secondary to inflammation
2 197 19 Recovering iron
deficiency
Receiving IV iron therapy
3 1768 9 Limited predictive
data
•Only two predictor tests
available
•Decision support will
likely require a minimum
number of predictor tests
4 197 19 Complex
hematologic
picture
Referral to hematology
would have likely been
useful had the testing been
ordered by a non-specialist
Conclusion: Predicted ferritin may more accurately reflect
underlying iron status in some patients� signals potential
application to clinical decision support
AJCP (2016). 145:778-88.
Ferritin Summary
•Coexisting data can discriminate normal from abnormal ferritin
results with a high degree of accuracy (AUCs as high as 0.97)
•Predictions of numerical ferritin results were moderately
accurate
•In at least certain cases, predicted ferritin may better represent a
patient’s underlying iron deficiency status.
Final Conclusions
1. Analysis of existing data can identify quality improvement opportunities
and strategies
2. Metrics must be developed thoughtfully and strategically
3. Artificial intelligence, machine learning and big-data analytics provide
emerging opportunities to enhance diagnostic precision and clinical quality
Acknowledgements
• Anand Dighe
• John Gilbertson
Ferritin Prediction
• Peter Szolovits
• Yuan Luo
• MGH eCore– MGH-
MIT Grand Challenge
Grant
• Kent Lewandrowski
• Joseph Rudolf
Many Aspects
Spurious Glucose Identification
• Craig Mermel
Variation Analysis
• Jeffrey Weilburg
• Michael Hidrue
Babesia analysis
• Vikram Pattanyak
9/5/2016
17
Data Analytics in the Clinical Laboratory Part 3
Using Data Analytics and the Electronic Heath Record to Enhance Quality and Safety
Anand Dighe MD, PhD
Director, MGH Core LaboratoryDirector, Laboratory and Molecular Medicine Informatics
Associate Professor, Harvard Medical School Massachusetts General Hospital
Boston, MA
Inter-pretation
ReportingProcessing/AnalysisCollectionOrdering
Post-AnalyticAnalyticPre-analytic
Middleware
• Interference checking
• Rules-based auto-dilution
• Automated add-ons
Test Result Auto-verification
Info Buttons
• Guidelines
• Literature
• Online resources
PathologyInterpretative
Services
PROCESS
Computerized Provider Order Entry (CPOE)
• Test panels
• Redundancy alerts
• Clinical guidelines
Automated Specimen Collection Process
RFID/bar coding
• Enhanced Electronic Medical Record systems
• Actionable result reporting
Institutional Reflex
Algorithms
Enhanced Result
Generation
Informatics in the Laboratory Testing Process
Why Emphasize Informatics?
• Acts as a force multiplier for many projects– Orders. Growth of the EHR and CPOE provides the
opportunity for informatics approaches to improve order entry
– Results. Enables high levels of robotics and auto-validation to be safely and efficiently implemented
– Results Management. Informatics approaches to results management can provide added quality and safety
• The increased growth of informatics-based approaches requires new strategies to measure lab and EHR outcomes
2010: in the U.S. only 12% of hospitals utilize computerized provider order entry for laboratory testing
2016: “Meaningful use” initiatives and other health care reforms have dramatically changed this to an estimated 50-70% usage in just 5 years
Advantages of CPOE• Opportunity to interact with the ordering clinician in
real time• Can present information to the clinician at the time of
the decision• Education is much less effective before or after the decision
• Links ordering physician tightly with the order, simplifying utilization audits
Computerized Provider Order Entry (CPOE)
9/5/2016
18
Inpatient/ED/Outpatient Provider Order Entry
Providers
CPOE: Pathology’s Perspective
Laboratory Information System
Laboratory Staff
Lab Orders
Pathology must avoid from being “shut out” from the CPOE system
• Improved diagnosis
• Error reduction
• Utilization control
• Laboratory efficiency
Providers
MGH PathConnect Middleware
Laboratory Staff
Permits Pathology to have control over Provider Order Entry screens
Inpatient/ED/Outpatient Provider Order Entry
Laboratory Information System
Lab Orders
MGH PathConnect Middleware
Web
services
MGH PathConnect MiddlewareKnowledge Management
• Synchronizes with Laboratory Information System (LIS)
• Receives data from LIS regarding each test
• LIS test data can be augmented with ordering messages, alerts, search terms, related tests
• Allows cataloging of Pathology data such that it can be shared with other parts of the organization via web services
POE LISMGH PathConnect
• Provider order entry calls middleware web service to build test dictionaries in a “just in time” manner
• Screen content can be updated in real time by Pathology via a web service
• Critical to control the content of menus to control utilization
POE Lab Ordering ScreenThe order entry group leaves us “white space” that the lab fills in.
User Interface Built Entirely from MGH PathConnect Data
9/5/2016
19
Order Entry: the Importance of Search
• Support synonyms
• Support misspellings
• Provide key test information (TAT, cost)
With 1,500 tests on the menu a robust search engine must be a part of all test order entry applications.
Provide information to guide appropriate utilization
Improving Vitamin D Utilization with CPOE
Search provides more than a list of possible matches �
Provides information to guide appropriate utilization
Middleware enables rapid (minutes, to author and update test) responses to utilization issues
• Adding non-interruptive ordering message dropped 1,25 OH vitamin D orders by 70% (p < 0.001)
• Cost savings = $20K/yr
MGH PathConnect Middleware(5 minutes to create and test new alert with no IS involvement)
MGH Order Entry Screen
Passive, non-interruptive ordering message
9/5/2016
20
CK-MB Additional Info Screen (Interruptive Alert)
Inpatient CK-MB Results Per Day
0
20
40
60
80
100
120
140
1/18
/201
1
1/25
/201
1
2/1/20
11
2/8/20
11
2/15
/201
1
2/22
/201
1
3/1/
2011
3/8/
2011
3/15
/201
1
3/22
/201
1
3/29
/201
1
4/5/
2011
4/12
/201
1
4/19
/201
1
4/26
/201
1
• Sustained 80% reduction in CK-MB orders within 3 weeks
• Cost savings of $30,000 per year
Added interruptive alert to POE
Building systems that get smarter
1) A strong vocabulary and data model can make getting to version two unnecessary
2) Monitoring actual use is critical to understanding how to improve the underlying data model
– Create monitoring reports for use on Day 1
– Expect and embrace failure!
– If you have a system that can learn you need content
1
Provider Order Entry
Cache
Lab Orders
Providers
Sunquest LIS
Laboratory Staff
MGH PathConnect Middleware
ODBCXMLMS SQL
Web
services
Daily reports of all searches and orders
Expect Failure: Monitoring Reports for Orders and Search
2a) Non-productive searches
b) Free text orders
9/5/2016
21
Analysis of User Search Productivity
Daily reports of all user searches, results, and orders available to lab
Reasons for Nonproductive Searches
FIX: Update middleware with misspellings
FIX: Update middleware with synonyms
FIX: Add test via middleware
(e.g. “insulin”)
(e.g. “celiac”)
(e.g. “syfillis”, “ferritan”)
1
Provider Order Entry
Cache
Lab Orders
Providers
Sunquest LIS
Laboratory Staff
MGH PathConnect Middleware
ODBCXMLMS SQL
Web
services
Daily reports of all searches and orders
Expect Failure: Monitoring Reports for Orders and Search
2a) Non-productive searches
b) Free text orders
Free Text Orders (e.g. “workarounds”)
0.0%
2.0%
4.0%
6.0%
8.0%
10.0%
12.0%
14.0%
6/2
5/2
008
7/2
5/2
008
8/2
5/2
008
9/2
5/2
008
10/2
5/2
008
11/2
5/2
008
12/2
5/2
008
1/2
5/2
009
2/2
5/2
009
3/2
5/2
009
4/2
5/2
009
5/2
5/2
009
6/2
5/2
009
7/2
5/2
009
8/2
5/2
009
9/2
5/2
009
• Free text (directly typed) orders monitored on a daily basis by Pathology
• Free text is essential to eliminate before orders are communicated electronically
• Permits near real time intervention with non-compliant physicians and/or system changes (adding tests, improving search)
• Current free text percentage 0.32% (4 sigma)
Order communication start
9/5/2016
22
Provider Order Entry
• Provider Order Entry is a key leverage point for Pathology to improve ordering practices, prevent ordering errors, and avoid pre-analytic error
• Pathology should have input/control over all laboratory order entry modules (IP/ED/OP) to permit rapid responses to ordering issues
• Centralization of laboratory knowledge combined with an understanding of ordering behavior is essential to improve quality and control utilization
Clinicians
OrderingClinicalInterpretation
Collection
Receiving
Resulting
InterpretiveReporting
Processing
Lack of robust electronic order entrysystems for all hospitals and all sites of care (inpatient, outpatient, ED) and for all pathology areas (AP, core, micro, blood bank)– Resulting manual processes
create huge amount of rework and inefficiency
– Hampers innovation and customer service by consuming staff resources
– Prevents implementation of decision support tools
Lack of closed loop results acknowledgment. – From ordering to resulting to
action taken need to ensure lab results are having their desired outcome
Order Entry and the Enterprise EHR
Inpatient/ED/Outpatient Provider Order Entry
Providers
CPOE: Pathology’s Perspective
Laboratory Information System
Laboratory Staff
Lab Orders
Pathology must avoid from being “shut out” from the CPOE system
Clinical Service BWH FH DFCI MGH1 NSMC NWH
Clinical Laboratories-Chemistry-Hematology-Microbiology
BICS2 MEDITECH
v5.64
Sunquest
v6.3; upgrading to v7.1 in 2013
Sunquest
7.1
Sunquest
v6.2; upgrading to v7.0 in 2013
MEDITECH
v5.64
Phlebotomy Collection Lattice MEDITECH
v5.64
Sunquest
Collection Manager
Sunquest Collection Manager
Will be Meditech
Point-of-Care Testing Abbott/Sybase
ISTAT
Roche Does Not Perform
Telcor Abbott/Sybase3 RALS
Anatomic Pathology-Surgical Pathology-Cytology-Cytogenetics-Molecular Diagnostics-Autopsy
Sunquest
- Powerpath
Cerner/Co-Path
v3.2
N/A – send specimens to
BWH
Sunquest
- Co-Path
Sunquest
CoPath
V4; upgrading to V6 in 2013
MEDITECH
LABLION Specimen Tracking
Blood Bank &
Transfusion Medicine-Donor Center-Blood Bank
Mediware
-Lifetrak
-HCLL
MEDITECH
v5.64
Sunquest
v6.3 for Cell Therapies only (limited use).
Use BB at BWH
Mediware
-Lifetrak
-HCLL
Sunquest MEDITECH
v5.64
Tissue Typing G4 N/A N/A – send specimens to
BWH
mTilda N/A5 N/A5
Reference Lab Mayo7
LIS Systems Across PHS (2013)
9/5/2016
23
Clinical Service BWH FH DFCI MGH1 NSMC NWH
Clinical Laboratories-Chemistry-Hematology-Microbiology
Sunquest
7.1
Sunquest
7.1
Sunquest
7.1
Sunquest
7.1
Sunquest
7.1 in Jan 2017
Sunquest
7.1
Phlebotomy Collection Sunquest
Collection Manager
Sunquest
Collection Manager
Sunquest
Collection Manager
Sunquest
Collection Manager
Sunquest Collection Manager
Sunquest Collection Manager
Point-of-Care Testing Abbott/Sybase
ISTAT
Roche N/A Telcor Abbott/Sybase RALS
Anatomic Pathology-Surgical Pathology-Cytology-Cytogenetics-Molecular Diagnostics-Autopsy
Sunquest
PowerPath
Sunquest
PowerPath
N/A Sunquest
CoPath
Sunquest
CoPath
Sunquest
CoPath
Blood Bank &
Transfusion Medicine-Donor Center-Blood Bank
Mediware
-Lifetrak
-HCLL
Sunquest 7.1 BTS Module
Sunquest
v6.3 for Cell Therapies only (limited use).
Use BB at BWH
Mediware
-Lifetrak
-HCLL
Sunquest 7.1 BTS Module
Sunquest 7.1 BTS Module
Tissue Typing G4 N/A N/A – send specimens to
BWH
mTilda N/A N/A
Reference Lab Mayo
LIS Systems Across PHS (2016)Lab Impacts: Adapting to Enterprise Information Systems
(Epic go live April 2016)
Homegrown system Enterprise system
Ordering favorites Not permitted Allowed
Order sets Reviewed by lab Uncommonly reviewed by lab
Collection process Simple (but manual) Complex (but electronic)
Menu size Limited (95%) Most tests available (99%)
Lab test search Provides decision support, CDS visible when searching
Search capabilities primitive, does not store search results or provide visible CDS when searching
Decision support availability Custom, fast, not requiring programming
Extensive possibilities but requires many levels of approvals, implementation complex
In lab processing Manual steps, slow Rapid, efficient
Moving from Reports to Real-time, Interactive Dashboards
• Allow access to variety of lab and support staff to track key metrics for EHR usage, efficiency, and quality and safety• Much of data is downloadable into Excel for process improvement projects• Limits number of custom reports needed
Lab Hospital Metric
Percentage of samples received in core and microbiology labs that were ordered in Epic
93.8%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
% Epic All Patients
Week 4 details:Inpatient = 96.1%Emergency room = 97.5%Outpatient = 91.1%
9/5/2016
24
8 PM 10 PM
80/3March 23
April 26
Outpatient Specimen Processing Changes
April Post-Epic (last 2 weeks): Collect to Receive = 153 min
March Pre-Epic: Collect to Receive = 237 min
Basics first: Ensure your EHR/LIS queues are worked
• EMR and LIS interfaces have numerous error queues that need a technical and sometimes a clinical eye on a periodic basis
• Numerous reasons for results failing to post, posting in wrong location, or failing to bill properly
• Failure to ensure that these are being reviewed in a timely manner can impair clinical care and billing
Application ReportsApplication Reports
Epic System and Reporting Tools
Clarity ReportsClarity Reports
Chronicles
(PRD)
Clarity
Data mart Data mart
Reporting
ShadowETL Process
Data mart
Hyperspace
Reporting WorkbenchReporting Workbench
• Real time & small data set
• More flexibility in customization
• Actionable to patient record
• Refreshed every night
• Analytical trending data
• Highly customizable
• Quick, real time data
• Little flexibility in customization
• Actionable to access patient record
Daily report of all Epic orders including problem list, diagnosis, order source (order set or search)r
Daily monitoring needed to assess for utilization issues
Key Fields for an EMR CPOE Report
• Ordering provider and department• Origin of order (Order set, facility list, personal preference list, database search)• EHR test code• Order number sent to LIS (to tie order to result)• Order attributes (Future/Standing, STAT/routine, etc)• Responses to questions asked during order entry• Provider comments
A few examples how this report can be used…
9/5/2016
25
If it’s on the menu it will be ordered… RBC Folate
• With our Epic daily report
we were quickly able to localize the orders to a single enterprise anemia panel• We first swapped out RBC folate for serum folate and later removed the RBC folate from the menu
Limiting Options is Often the Best Approach: SPEP with Immunofixation
• Allowing clinicians to order SPEP + immunofixation led to large increase in lab workflow for a highly manual test• Removed test from menu and only allowed clinicians to order SPEP with reflex immunofixation• Used an automated tool to change > 100 personal preference lists to swap out inappropriate order for preferred order • We are back to our pre-Epic baseline now.
Tracking Miscellaneous Test Requests
• Prevent workarounds that are not easily trackable (e.g. paper requisitions)• Provides data for determining if additions to menu are needed
Regular Review of Misc Test Report for Trends and Outliers
Appropriate
WHOLE EXOME SEQUENCING
MATERNAL CELL CONTAMINATION FOR SNP MICROARRAY
CFTR SEQUENCING, REFLEX TO DEL/DUP IF NEGATIVE
ZIKA
Already built
MATERNI T21 PLUS
BORRELIA MIYAMOTOI AB
SERUM IRON, TIBC.
BUPRENORPHINE LEVEL
Don't build
BABESIA PCR
• 0.5% of orders (20/4,000 per day) are Misc Tests• Possible outcomes of review: Build new tests, educate providers, or keep as misc test
9/5/2016
26
Summary
• Laboratory patient safety and efficiency starts in the EHR
• Standardization of Lab Information systems is needed to lower complexity of upstream and downstream processes
– Ordering, interfaces with EHR, interfaces with reference labs, collection systems, test codes
• Detailed monitoring of lab results and orders provides needed information for process improvement and utilization control
Thanks!