Using Registries in Practice, Quality Improvement, Research, and Education Elizabeth O. Kern, MD, MS, Susan R. Kirsh, MD, and David C. Aron, MD, MS, Center for Quality Improvement Research, VA Medical Center, Cleveland, OH and QUERI-DM

1
Using Registries in Practice, Quality
Improvement, Research, and EducationElizabeth O.
Kern, MD, MS, Susan R. Kirsh, MD, and David C.
Aron, MD, MS, Center for Quality Improvement
Research, VA Medical Center, Cleveland, OH and
QUERI-DM

• Objectives
• To understand the link between Registry data
  structure and its functionality.
• To understand how a Registry can be created from
  the VISTA database.
• To understand how a disease Registry can be used
  to in quality improvement, education, and
  research.

2

• Outline
• Context for registry use Chronic Care Models and
  Systems Redesign based on such models
• Development of the Cleveland VAMC Diabetes
  Registry from the VISTA Database
• Using the Diabetes Registry in Practice
• Identification of patients at high cardiovascular
  risk for targeted interventions
• Identification of patients and provision of
  self-management assistance.
• Using the Diabetes Registry in Quality
  Improvement and Research
• Analyses for managers
• Audit and feedback for staff providers
• Evaluation of quality improvement projects
• Registry as a research data base
• Using the Diabetes Registry in Education
• Audit and feedback for trainees

3
The Context for Registries

• The various models for management of chronic
  illness have one feature common information Rx
  to care for both the sick patient and sick system

WHO
Improving Care for People with Long-Term
Conditions A Review of UK and International
Frameworks. NHS Institute of Innovation, 2006
4
Shared Medical Appointments (Group Visits) Based
on the Wagner Chronic Care Model
5
What are the components of Clinical Information
Systems?

• Patient registries that are organized into a
  database to access important patient information
  easily, track individual patient outcome measures
  and prevention activities, and provide feedback
  to providers.
• Clinical summaries
• Clinical reminders
• Register recall system

6
A Flat File is a Roster Information
Each row represents a unique patient, plus extra
information that can fit within the single row.
7
A Table is Structured by its Attributes and its
Primary Key

• Patient Name
• Patient ID
• Site ID
• Date of Birth
• Primary Care Provider

Primary Key
Attributes are the column headings
8
Tables are Linked to Other Tables by the Primary
Key
9
Linked Tables in the Cleveland VAMC Diabetes
Registry
10
Data Flow from the Database to Web Page
Data Warehouse VISN 10 VISTA
Diabetes Registry Database
Step 1Nightly Data Pull Step 2
SQL Stored Procedures
VA Intranet Web Page Live Data Reports by
User Request
Step 3 ASP.NET platform Step 4 Standard Queries
in C
11
Data Flow Software

• VISTA data VISN 10 SQL
  Data Warehouse
• KB-SQL in a SSIS-SQL Package
• SQL Data Warehouse Diabetes Registry
• SQL Relational Database
• SQL Stored Procedures (helps to run standard
  queries faster)
• Diabetes Registry Web Page
• ASP.NET 2.0 platform
• C programming language to create standard
  queries
• Design tool is Visual Studio 2005
• Web Page reports Clinical User
• Excel Spreadsheets
• Microsoft Mail Merge generates templated
  letters to patients

12
Analytic Software

• To pull data from the Diabetes Registry for ad
  hoc analyses
• SQL Query Analyzer
• To place data in analytic format
• Notepad .txt tab delimited file
• Excel spreadsheet
• For data management and analysis
• SAS statistical analytic program
• SAS datasets
• For security and confidentiality
• All files (including SAS working files) remain
  behind the VA firewall, on a server drive, in
  folders limited to specific users

13
Operational Definitions

• DEFINE patients with diabetes
• Had at least 3 ICD-9 codes indicating diabetes on
  3 separate dates (codes are 250.xx, 357.2,
  362.0, 366.41)
• OR
• Had a diabetes-specific medication dispensed
  from a VISN 10 pharmacy
• Diabetes-specific medication list maintained as
  a look-up table in the Diabetes Registry
  database

14
Operational Definitions

• DEFINE Active versus Non-Active patients
• ACTIVE
• Date of Death null
• AND
• (The patient had a primary care visit within
  the past 18 months
• OR
• The patient had diabetes-specific medications
  dispensed within the past 18 months)
• Non-ACTIVE conditions for ACTIVE not met

15
Operational Definitions

• DEFINE the clinic most responsible for diabetes
  care for each ACTIVE patient
• Find the most recent primary care type visit
  within past 18 months.
• From this visit, assign each patient to the
  facility site and clinic or CBOC associated with
  that visit (i.e., follow the patient trail)
• A novel system was created, mapping each visit
  (also called encounter) to a specific site and
  clinic using the Hospital_Location variable in
  VISTA.
• The 4,200 unique Hospital_Locations were pared
  down to 1,792 associated with encounters in a
  primary care clinic, and categorized as
  definitely indicating primary care (Tier 1) or
  possible indicating primary care (Tier2).

16
Mapping 1,792 Hospital Locations to 51
Different Clinics in VISN 10 (Hospital Location
is a variable included in each visit or
encounter)EXAMPLE
17
Assigning the Primary Care Provider

• From the Primary Care Manager Module database
  (PCMM) most patients are assigned to a primary
  care provider in VISN 10.
• The PCMM database is up dated manually, by a
  person assigned to this task.
• The Diabetes Registry pulls the Primary Care
  Provider (PCP) variable from the PCMM to match
  with each patient in the Registry.
• Approximately 10 of Diabetes Registry patients
  are not assigned to a primary care provider,
  because the PCMM table has not been updated yet,
  or the patient is truly not assigned (e.g., ESRD
  patients, HIV patients, Employee Health patients)
• Some PCPs cover multiple clinic sites therefore
  knowing who is PCP does not necessarily mean the
  clinic site is known

18
Data Cleaning

• Problem Text values appear in what is supposed
  to be a numeric result field
• Example LDL-c comment
• Example HbA1c not done
• Problem Multiple names and codes for the
  same lab test
• Example 14 different names for the A1c test in
  VISN 10
• Example 13 different Test-IDs for the A1c
  test in VISN 10
• Example 3 different National VA Lab Codes for
  the A1c test in VISN 10, or a National VA Lab
  code is not assigned

19
How Many Ways to Name an A1c Test?
20
Using the Diabetes Registry for Population-Based
Disease Management

• Find the patients who are outliers in
• A1c
• LDL-c
• Blood pressure
• Foot exam
• Eye exam
• Group by clinic/provider with primary
  responsibility to these patients for diabetes
  management

21
Using the Diabetes Registry for Population-Based
Disease Management

• Create spreadsheets for patient calls for special
  interventions at clinic level or provider level
• Merge the spreadsheets into templated letters for
  special interventions at clinic level or provider
  level
• Create individualized Diabetes Report Cards
  containing the five parameters used for EPRP to
  send to patients by mail, or to use in group
  classes
• Include the Diabetes Medication Profile in order
  to group patients needing insulin starts or
  titration
• Example patients with A1c gt 9, on 2 oral meds,
  need to start HS NPH

22
Requesting a Report from The Diabetes Registry
Web Page
23
Report Result (fragment) from the Diabetes
Registry Web Page
24
Templated Header to the Birthday Letter (From
the Diabetes Registry web page patients in
Lorain CBOC with high or missing LDL-C, with a
birthday in July ) Underlined text is dropped in
according to links and expert logic.
Cleveland VA July 27, 2007 Dear JOHN
DOE, Happy Birthday! Your VA health care
providers want you to have many more! We are
sending you your latest diabetes test results
because our VA records show that your blood test
for cholesterol is either too high, or needs to
be rechecked. Your LDL-cholesterol (the bad
kind of cholesterol) should be less than 100 to
protect you from stroke or heart attack. Even if
your last test was good, you are due to have it
checked again. Your primary provider at the VA
Lorain clinic would like you to call L W
to go over your results, set up a fasting
blood test, or set up a visit. Please call
[phone] EXT 2247 to schedule. If you come
for a clinic visit, please bring in all of your
medication bottles, your blood glucose meter, and
any glucose records if you have them. Thanks!
25
Individualized Diabetes Report Contained in the
Birthday LetterThe values, messages, and
smiley faces are driven by expert logic.
26
Quality Improvement

• How do we know a change is needed?
• How do we know a change is an improvement?
• How do we know where to put scarce resources?
• A Diabetes Registry can provide data to
• Describe the patient population
• Identify patient sub-groups having the most need
• Identify who is in the sub-groups
• Show the reach of intervention programs
• Show the outcomes of intervention programs

27
Growth in the Patient Population with Diabetes in
VISN 10

• The net growth in live patients with diabetes was
  73 over the 5 year period from 2002 to 2006.
• By the end of 2006, there were 42,499 patients
  with diabetes, representing approximately 21-25
  of the VISN 10 patient population.
• Source VISN 10 Diabetes Registry

28
Almost Half of Patients Do Not Receive
Self-Management Education from the VA

• From 2002-2006
• looking back for
• outpatient notes
• Diabetes Education
• diabetes education class
• glucometer class
• diabetes specialty clinic
• diabetes team program
• Nutrition Education
• any nutrition visit.
• Source
• VISN 10 Diabetes Registry

29
Target Patients with Poor Glycemic Control

• Prioritize by the
• most recent HbA1c
• 27,031 (64) are lt 7.5
• 10,131 (24) are between 7.5-8.9
• 5,278 (12) are 9 or greater
• Source VISN 10 Diabetes Registry

30
Glycemic Control Plus Medication Profiles Can
Guide Interventions

• High A1c, on no diabetes meds from the VA, may
  need VA prescription.
• High HbA1c, on orals only, may need start of
  basal insulin and/or carb counting
• High HbA1c, on insulin, needs insulin titration
  and carb counting
• Source
• VISN 10 Diabetes Registry

31
Drop in HbA1c After DSME classes in the Cleveland
VAMC

• N 436 patients
• Results were same for a subgroup already taking
  insulin.
• Source
• VISN 10 Diabetes Registry

-0.1
-0.3
-0.8
Change in HbA1c
P lt .001 for all strata
-2.4
32
Growth of the Nurse Diabetes Case Manager Program
in Cleveland VAMC

• From 2003 through 2006, the Diabetes Case
  Manager program saw 3,886 unique patients.
• ( 20 of Cleveland VA patient population with
  diabetes).
• The program grew from
• 3 to 10 by 2006.
• 7 achieved CDE after training for case management.

Source VISN 10 Diabetes Registry
33
Diabetes Case Management Resulted in Better A1c
Outcomes than Usual Care

• Case management resulted in greater drops in A1c
  for patients with starting A1c lt 9,
• and an equivalent drop in A1c for patients with
  starting A1c gt 9

0 -0.3
-0.5 -0.7

Change in HbA1c

p lt.05
-1.3 -1.4
Source VISN 10 Diabetes Registry
34
Dataset (from the VISN 10 Diabetes Registry)

• 40,632 patients receiving diabetes-specific
  medications in VISN 10 since Jan 2005, and who
  are alive.
• 9,000 patients in VISN 10 do not receive either
  glucose test strips or hypoglycemic agents from
  the VA, but have an ICD-9 code of diabetes.
  These patients were excluded from this analysis

35
Thiazolidenedione (TZD) and A1c Outcomes Within
VISN 10, by Site
36
Using Registries in Practice, Quality
Improvement, Research, and EducationElizabeth O.
Kern, MD, MS, Susan R. Kirsh, MD, and David C.
Aron, MD, MS, Center for Quality Improvement
Research, VA Medical Center, Cleveland, OH and
QUERI-DM

• Objectives
• To understand the link between Registry data
  structure and its functionality.
• To understand how a Registry can be created from
  the VISTA database.
• To understand how a disease Registry can be used
  to in quality improvement, education, and
  research.

37
Shared Medical Appointments (Group Visits) Based
on the Wagner Chronic Care Model
38
The Patient Encounter

• Personnel
• MD, NP/CDE, RN, Pharmacist, Psychologist
• 8-20 patients/session
• 90 minutes sessions
• Return visit interval 4-8 weeks or until goals
  achieved
• Group activities
• Education
• Patient Centered Discussion
• Review of labs/medications
• Individual activities
• Medication management
• Referrals
• Individualized plan of care outlined and give to
  patient

39
Evaluation of the impact of SMAsKirsh et al.
QSHC 2007 in press.

• Subjects
• Diabetic patients with gt1 of
• A1c gt9
• SBP blood pressure gt160 mmHg
• LDL-c gt130 mg/dl
• Patients largely derived from registry data, few
  referred from pcp
• participated in gt1 SMA from 4/05 to 9/05.
• Study Design
• Quasi-experimental with concurrent, but
  non-randomized controls
• patients who participated in SMAs from 5/06
  through 8/06. A retrospective period of
  observation prior to their SMA participation was
  used.

40
Kirsh et al. 2007 in press. Findings

• Levels of A1c, LDL-c, and SBP all fell
  significantly post-intervention
• A1c decreased 1.4 (0.8, 2.1) (plt0.001)
• LDL-c decreased 14.8 (2.3, 27.4) (p0.022)
• SBP decreased 16.0 (9.7, 22.3) (plt0.001).
• The reductions greater in the intervention group
  relative to the control group
• A1c 1.44 vs -0.30 (p0.002) for A1c
• SBP 14.83 vs 2.54 mmHg (p0.04) for SBP.
• No diff. for LDL-c 16.0 vs 5.37 mg/dl (p0.29).

41
(No Transcript)
42
(No Transcript)
43
Registry use in continuing care

• Track additional patient data hard coded in note
  for future reference
• Monitor progress on patients and give report card
  to providers-pilot
• Birthday letters generated by registry data to
  engage patients in initiating SMA

44
Trainee Participation in SMA

• Internal Medicine residents and third year
  medical students on chronic disease block
• Uses of registry in general to manage population
• Clinical Information System module
• Audit and feedback of residents primary care
  panels and teams

45
Questions?
46
References

• Gliklich RE, Dreyer NA, eds. Registries for
  Evaluating Patient Outcomes A Users Guide.
  (Prepared by Outcome DEcIDE Center Outcome
  Sciences, Inc. dba Outcome under Contract No.
  HHSA29020050035I TO1.) AHRQ Publication No. 07-
  EHC001-1. Rockville, MD Agency for Healthcare
  Research and Quality. April 2007.
• Bodenheimer T, Grumbach K. Electronic Technology
  A Spark to Revitalize Primary Care?
  JAMA. 2003290259-264