Applying electronic health record data to quality of care

Report
Applying electronic health record data to
quality of care improvement and practice
based research initiatives
Cecil Pollard, Director
West Virginia University Office of Health Services Research
2014 KBPRN Collaborative Conference
5/9/2014
"The project described was supported by the
National Institute of General Medical Science,
U54GM104942. The content is solely the
responsibility of the authors and does not
necessarily represent the official views of the NIH."
2014 KBPRN Collaborative Conference
5/9/2014
Overview

Office of Health Services Research

Era of Big Data

Introduction of EHR’s

Concerns with Big Data

Repurposing of EHR’s

Practical applications using EHR’s

Where are we and where might we be
going
West Virginia University Office of Health Services
Research






30 years collaborating with primary care and public
health
Past 15 years focusing on quality improvement in
chronic disease
Provider and patient education
Collaborating with about 50 community based
primary care sites
Focus on underserved and rural populations
Also working with Caribbean and Latin American
nations and U.S. Territories in the So. Pacific
2014 KBPRN Collaborative Conference
5/9/2014
By 1985 it had evolved into this
2014 KBPRN Collaborative Conference
5/9/2014
Concerns over Data Accuracy
 1985-Devin and Murphy of IBM
 Development of architecture for
data
warehousing
 Focusing on high quality, historically complete
data, and accurate data
2014 KBPRN Collaborative Conference
5/9/2014
“Big Data”
 July
1997
 The Problem of Big Data
 The term "big data" was used for the first
time in an article by NASA researchers Michael
Cox and David Ellsworth.
 The computer processing cold not keep up
with the increase in the large amount of data
being generated.
2014 KBPRN Collaborative Conference
5/9/2014
Big data in health care

Knowledge translation between health analytics and the
realities of patient care

The statement ‘There are right ways to analytics’
implies we may not be doing analytics correctly

Health care seems to think that big data will improve
patient care and population health management

It isn’t about the data and how much you have, but
about data management

We are creating data landfills

Turning data into useful information
The beginning of Electronic Health Records-1964
http://www.youtube.com/watch?v=t-aiKlIc6uk
So what were the promises
from this 1964 experiment

Relieve doctors and nurses of some of
their paperwork

Better management of diseases

Eliminate errors in medication and tests
What is current status

The promise of EHR’s

Have reduced paperwork

Reduced errors in patient medications and
testing

Are we making best use of the data

Do we have good tools-software and
skilled analyst
2014 KBPRN Collaborative Conference
5/9/2014
Some examples of using EHR data
 Example
1 – Patients with last HbA1c >=9
 Example
2 – Losing QI incentive pay
 Example
3 – Identifying patients with
hypertension
2014 KBPRN Collaborative Conference
5/9/2014
Example 1 – Patients with last HbA1c >=9 (HRSA report)
 Report showed 85%
 Nurse responsible for QI at site questioned data
 We found that only the hand-entered results from
their in-house labs were picked-up (HRSA treats
patients with missing HbA1c as >=9; missing data
treated as non-compliant)
 Lab reports from outside vendor were missed
 True statistic = 7%
2014 KBPRN Collaborative Conference
5/9/2014
Example 2 – Excess prescription of antibiotics among
children without proof of bacterial infection
Automated report on children receiving
antibiotics showed excess prescribing among
providers
 Prescribing antibiotics for viral infections
 Report was missing the diagnoses that should have
been tied to the prescription


Automated report did not match the appropriate
diagnoses with the prescriptions
 Loss
of $20,000 in incentive pay
2014 KBPRN Collaborative Conference
5/9/2014
Example 3 – Identifying patients with
hypertension
 Worked
with 11 primary care centers on underdiagnosis of hypertension
 Identified patients based on ICD-9 coding
 Noticed
significant use of free text coding (the EHR
allowed providers to use free text)
 Found significant amount of patients with
consistently high blood pressure readings but no
diagnosis of hypertension (EHR missed this
biomarker)
 Found
sites
nearly 2000 patients missed across all
2014 KBPRN Collaborative Conference
5/9/2014
Increase in HTN patients
Search Criteria
Number
Number added
Cumulative
Percent
ICD-9 code
12,919
---
86.7
ICD-9 code plus
free text
13,817
898
92.3
ICD-9 code plus
free text plus
BP measures
14,893
1,078
100
Total
1,974
John Snow and the Broad Street pump
 John
Snow’s chemical and microscopic
examination was not able to conclusively
prove the danger of the Broad Street
pump.
 Snow created a map to show how the
cholera cases were clustered around the
pump.
 Pump
handled removed upon new conclusion
2014 KBPRN Collaborative Conference
5/9/2014
John Snow Revisited
 How
could electronic health records have
help…?
 EHR
identifies all cases of cholera
 Look
at location indicators (addresses)
 Create
thematic map
 Removed
pump handle
2014 KBPRN Collaborative Conference
5/9/2014
Identifying patients at-risk for diabetes


Previously, relied on provider intervention at point of
care to identify diabetes risk and think/make effort to
refer the patient
 One patient at a time
 Inefficient
Identify at-risk patients using existing data
 Clinic-wide
 More efficient
2014 KBPRN Collaborative Conference
5/9/2014
Using de-identified data from 14 WV primary
care centers, we did the following:
 Standardized
the data in a common format (CDEMS)
 Identified established patients by site (those receiving
care for 12 months of more)
 Excluded patients with a diagnosis of diabetes or prediabetes
 Identified persons at risk for pre-diabetes based on
CDC’s Group Lifestyle Balance criteria:
Age > 45 with last recorded BMI >25
 Age < 45 with last recorded BMI >25, with HTN, hyperlipidemia, gestational
diabetes, family history of diabetes, or cardiovascular disease
 Last fasting blood glucose in the range of 100-125

2014 KBPRN Collaborative Conference
5/9/2014
Identifying patients
Identified persons at risk for prediabetes based on CDC’s Group Lifestyle
Balance criteria:



Age > 45 with last recorded BMI >25
Age < 45 with last recorded BMI >25, with HTN,
hyperlipidemia, gestational diabetes, family
history of diabetes, or cardiovascular disease
Last fasting blood glucose in the range of 100-125
Results
 14
primary care centers:
 130,021
active patients
 106,367 (81.8%) are
established (receiving care
for 12 months or more)
 94,283 (88.6%) do not have
a diagnosis of diabetes or
pre-diabetes
130,021 active
106,367 established
 Those
patients are the focus
of the analysis
94,283 no dx of DM
or pre-DM
Results-Identifies 10,673 (11.3%)
Discussion
 Patients
at-risk for pre-diabetes and in need of
targeted screening can be identified using EHR
data
 Streamlines
opportunity for patient identification,
screening, and referral
 No need for additional data collection at the sites
 Making meaningful use of existing data
2014 KBPRN Collaborative Conference
5/9/2014
Discussion
 Early
identification and intervention 
opportunity for prevention
 Improving
outcomes and quality of life, and reducing
long-term costs of care
 Implementation
highlights
 Algorithms
built using de-identified data
 Identified data used to create lists of at-risk patients at individual
sites
 Each site contacted patient in an effort to recruit them for
intervention
2014 KBPRN Collaborative Conference
5/9/2014
Questions or comments
Some closing comments

At one time there were 450 different EHR’s in
country

EHR’s need better Import/Export functions

Common Import/Export data formats

Should EHR’s be permitted to charge extra for
analytics

EHR’s charge for each site to be connected to
state Information Exchanges ($10,00)

similar documents