© InfoClin Inc 2006. All Rights Reserved.
Data Standards, Data Cleaning and
Data Discipline
Designing the Next Generation of EMRs
Insight
Agenda
• Why is data quality important?
• Why data quality is poor in the EMR
• The Principles of Data Discipline
The Problem
• If you can’t tell which of your patients has had a pap
smear, how can you do proper cervical cancer screening?
• If you can’t tell which of your patients has bowel disease,
how can you do proper colon cancer screening?
• If you can’t reliably identify patients with chronic disease in
your EMR, how can you have a chronic disease
management program?
• If you don’t know which patients are on warfarin, how can
you make sure that you are prescribing safely?
Current EMRs are not Effective
•
Linder JA, Ma J, Bates DW, Middleton B, Stafford RS. Electronic
health record use and the quality of ambulatory care in the United
States. Arch Intern Med. 2007 Jul 9;167(13):1400-5
– Retrospective study of ambulatory visits across the US
• 18% of visits were recorded in an EMR
– Measured care across 17 quality indicators
• Antibiotic rx, preventive counselling, screening, avoid inappropriate meds in elderly
– 14 measures showed no difference
– 2 areas were better, 1 was worse
•
Overall assessment: EMRs are ineffective in improving patient care
and are probably no better than paper
– A damning indictment
•
EMRs will only be more effective if they can manage information
better –they have the promise, but they need much more to deliver on
the promise
Why Data in EMRs is Poor 1
•
EMRs are optimized for individual patient care
– Documentation within a patient is quite good
– Allows you to view data from a variety of sources in one place
•
Current EMRs are not designed for population-based care
– Data capture is not standardized
• Standard terminology is poorly enforced in most EMRs
• Meta-data is poorly captured (i.e., can put data in the ‘wrong place’)
– Data inconsistency is rampant
• Many patients with HBA1c > 7 or on Insulin are not labelled as diabetic in the EMR
– Data inaccuracies abound
• Many patients with diagnosis code of 250 in the billing system are not diabetic
– There are no standardized data feeds into the EMR
• Laboratory, medications, consult notes, hospital discharges and diagnostic imaging do not come in a consistent and standard way
– Data good enough for individual care are too complex for population care
• To manage a colon cancer screening program, you need to enter information in 4 different places
– Lab data: Stool Occult Blood Test result –if lab doesn’t send it, add it manually! – Procedures: Colonoscopy
– Past medical history: Colon cancer or inflammatory bowel disease – Problem list: current cancer or inflammatory bowel disease
• Any error in where you put the data, will put the patient in the wrong category
• Diabetes patient management is even more complex –requires data in 7 places, not all of which are structured in most EMRs
Why Data in EMRs is Poor 2
•
As humans, we
– Are chronically inconsistent
• We continue to prescribe glyburide and forget to label the patient as being diabetic
– Deviate from standard terms
• CAD, Atherosclerosis, CHD, ASHD all mean the same thing –but computers don’t know that!
– Forget to change the status of information in the EMR
• We tell the patient to stop taking a medication, but don’t actively stop it in the EMR
– Use terms that denote a class, when we really mean an instance
• We say ACE inhibitors or statins, when we really prescribe ramipril and atorvastatin –but computers only know instances, not classes!
•
Medical Knowledge and Terminology evolves over time
– Juvenile vs. Adult onset
– IDDM vs. NIDDM
– Type 1 vs. Type 2
•
Some data is too tedious to structure for non-specialists
– Foot ulcers, retinopathy, diet, exercise, etc.
•
Current EMRs don’t make up for the foibles of humans or the vagaries
of human progress
Principles of Data Discipline
• Data Standardization
– Coding
– Diagnoses, Medications, Labs, Vitals & Physical Exam
• Data Cleaning
– Coverage –all patients are in the system
– Consistency –all data tells the same story
– Completeness –all data is in the system
– Correctness –right patients in, wrong patients out
– Coded –all relevant data is coded or in a single format
• Data Discipline
– Systems thinking
• Templates, reminders and searches work together
• Environmental cues
– System supports humans
• Provides clues that data is incomplete or inconsistent or not
coded
Principles of Data Discipline
• Data discipline should be maintained using
‘systems’
– Using a template for a particular aspect of care should
automatically provide the data to turn a reminder off
• E.g., smoking cessation counselling, any form of in-office
procedure that is not lab related
– Using lab results to turn off a reminder
• E.g., pap smear, HbA1c, LDL, FOBT etc
– Using a scanned report to turn off a reminder
• E.g., mammography
• Importance of Reminders
Managing Colorectal Cancer Screening
• The Task
– Quickly generate a list of patients who need to be screened
– Generate statistics on screening
• Wrinkles
– FOBT results that come in on paper need to be managed carefully
– Patients who have had a colonoscopy need to be excluded for 5
years
– Patients who have had cancer need to be excluded
– Patients with GI bleeding disorders need to be excluded
• Data Issues
– Electronic FOBT results sometimes not consistent
– Colonoscopy needs to be recorded properly and consistently
– GI bleeding disorders and cancers need to be recorded properly,
consistently and in the right place so they can be excluded
Managing Diabetes Care
• The Task
– Be able to find all patients with diabetes reliably
• Wrinkles
– There are different types of diabetes that can confuse the issue –
gestational diabetes, diabetes insipidus
– Patients may not have a diagnosis listed, but may have other signs
of diabetes: on insulin, high blood sugar
• Data Issues
– Different terms are used for diabetes: DM and diabetes mellitus
are the most prevalent
– The diagnosis could be in 2 different places –History of Past
Health and Problem List
– Some signs of diabetes are also a sign of other things: metformin
(a drug for diabetes) is also a drug for another unrelated disease
(polycystic ovarian syndrome)
Implications for EMR Design
• If EMRs want to be better at information management,
they need the following:
• Tools that can help prevent data inconsistencies or detect,
present and resolve data inconsistencies
– E.g., EMR asks you if you want to add diabetes to the problem list
when prescribing insulin
– EMR enforces appropriate words to be used in the right places–
e.g., second hand smoke
– EMR asks if you want to discontinue an active medication after a
few months of non-prescribing
• Tools that automatically add important metadata to the
record or recommend appropriate metadata
– EMR automatically adds metadata for you –e.g., when referring a
patient to Dr. Jones, the system also captures the fact that Dr.
Jones is a cardiologist and that she works at Best General
Hospital.
– EMR suggests putting a diagnosis in the Problem List if it is in the
Past Medical History or in an Encounter Note.
Implications for EMR Design
•
Tools that allow you to manage data at the population level
– Assign a group of patients as having received a particular intervention –e.g., flu shot or pap smear
– Assign a group of patients to a particular registry (group) or treatment – Update status of data without having to enter each chart separately
•
Features that clean up dirty data, inconsistent data and missing data
– Some labs send HbA1c as a % (7%) and some as a decimal (0.07)
• Both are correct, but they can’t coexist in the same database
– Lots of patients have missing dates of birth, addresses and other clinical information
•
Features that allow you to download a computerized guideline from a
respected group
– Currently, you have to do your own programming with complex guidelines
– Doctors are not programmers and get it wrong or spend hours trying to get it right
•
Sophisticated statistical analysis tools
– Physicians expect their EMR to provide them with the information they need –run time charts, uncontrolled patients, etc.
– They don’t want to export data to another application and generate reports –they don’t have the time, the staff nor the capabilities
New Competing Technologies
• Many of the functions that people expect from EMRs are
actually provided in a type of software called disease
management systems
• Disease management systems provide true database
functionality
– The ability to maintain good quality data
– The ability to make changes at the population level
– The ability to have built-in guidelines
– The ability to query and return statistics
• The role of this new technology is still being debated
– Will it work side-by-side with EMRs and provide new functionality
to EMRs?
– Will it replace EMRs as the new wonder tool and recommended
clinical technology?
Summary
• Medical practice is changing rapidly
– Chronic disease management has reached
maturity
– Team-based care is rapidly gaining ground as
the clinical results are outstanding
• The new way of practicing medicine
requires new tools
• EMRs will have to catch up to the new ways
of practicing or lose ground to Disease
Registry and Chronic Disease Management
software
Acknowledgements
• Hamilton Family Health Team for
Dr. Karim Keshavjee
Dr. Karim Keshavjee is a Family Physician with a part-time practice in Mississauga. He
spent five years in the pharmaceutical industry managing clinical trials and managing an electronic drug utilization project. He is currently an Associate Member of the Centre for Evaluation of Medicines, an independent academic research institute affiliated with
McMaster University in Hamilton, Ontario.
At the Centre for Evaluation of Medicines he is the Clinician-Project Director for the
COMPETE (Computerization of Medical Practices for the Enhancement of Therapeutic Effectiveness) series of research studies. The COMPETE research program studies the impact of e-health technologies on the management of patients with diabetes and
vascular disease. You can find out more about COMPETE at www.compete-study.com. Karim was also the physician consultant to Canada Health Infoway for the pan-Canadian
electronic prescribing project (CeRx), the inter-operable electronic health record (iEHR) project and the consumer health architecture project (PAQC). He is also a mentor on a CIHR-funded, pan-Canadian health informatics research training program for post-graduate students.
Karim completed his MBA at the Rotman School of Business in 2004 in technology
commercialization. He now specializes in helping academic researchers disseminate their evidence-based research findings and inventions to primary care physicians who could benefit from them. You can find out more about InfoClin at www.infoclin.ca. You can contact Karim at [email protected].