Microsoft PowerPoint - NYP talk.ppt [Read-Only]
Post on 23-Jun-2015
414 Views
Preview:
Transcript
®
© 2008 University HealthSystem Consortium
Steve Meurer PhD, MBA/MHSUniversity HealthSystem ConsortiumVice President, Clinical Data & InformaticsNovember 6, 2008
Data’s Role in Driving Sustained Change
©2008 University HealthSystem Consortium 2
University HealthSystem ConsortiumUniversity HealthSystem Consortium
• A member owned and governed consortium of academic medical centers
– This relationship is what makes us unique– Approximately 95% of all major not for profit academic medical
centers are UHC members (n = 102)– Affiliate hospitals are welcome and increasing in numbers (we
currently have over 190 associate member hospitals)
• UHC provides comparative databases, benchmarking and other associated services, a Group Purchasing Organization, and networking opportunities
©2008 University HealthSystem Consortium 3
The Difficulty of ChangeThe Difficulty of Change
Not an innate drive
Against current societal norms
Current healthcare environment where quality has become scrutinized
©2008 University HealthSystem Consortium 4
DrivenDriven
Drive to Learn and make sense of the world and ourselvesDrive to Acquire objects and experiences that improve our status relative to othersDrive to Bond with others in long-term relationships of mutual care and commitmentDrive to Defend ourselves, our loved ones, our beliefs, and our resources from harm
By Paul R Lawrence and Nitin Nohria
©2008 University HealthSystem Consortium 5
All improvement must result from a changeIt’s difficult to change without first admitting that you don’t know somethingThe community believes that physicians know what is needed to improve health, and they don’t understand variationPhysicians / Nurses are taught that they are the sole purveyors of clinical medicine and that they should never admit to not knowing
Earnest Codman, MDThe Dartmouth Atlas
The Bell Curve
©2008 University HealthSystem Consortium 6
Don’t KnowWhat You Don’t Know
You KnowWhat You Don’t Know
You KnowWhat You
Know
Don’t KnowWhat You
Know
Selfawarenessthroughexperience/persuasion
Formal or informal education
Practice
Learning/Changing Process
C.R.A.P. Detector
©2008 University HealthSystem Consortium 7
Other Improvement TipsOther Improvement TipsThe Power of One is as important as the support of Leadership!The fallacy of CQI, Six Sigma and LeanThe Tipping Point
The Law of the Few (Mavens, Connectors and Salesmen)
Paul Revere’s Ride
The Stickiness FactorTetanus Shots (small but significant changes – 3% to 30%)
The Power of ContextNew York City at night (Broken Windows theory)
©2008 University HealthSystem Consortium 8
Quality Data ConcernsQuality Data Concerns• You need a physician to talk to a physician
• Risk Adjustment
• Transparency
• Data Quality / Accuracy
• Clinical vs. Administrative Data
• Timeliness
©2008 University HealthSystem Consortium 9
Motivating a Clinician to ChangeMotivating a Clinician to Change• Data in quality and variation of care is now mature
enough to use for improvement, but its still not perfect
• This is not about physicians vs. others, rather its about the process of approaching a clinician with data
– Step 1 – Provide the data, an explanation and support in a non-threatening manner
– Step 2 - Answer all questions regarding the data– Step 3 – Wait. If no change, then Medical Staff
• A physician not trained in the data performing steps one and two are much more likely to dismiss variation when the answer is ‘My patient’s are sicker, and the methodology does not adequately adjust…’
©2008 University HealthSystem Consortium 10
12.6the data
the analysis
the information
the tools and resources
the relationship
the understanding& innovation
The Improvement5.3
©2008 University HealthSystem Consortium 11
• 80 year old female, 1 day LOS ED Visit• MSDRG 389 – GI Obstruction w/ CC (Gastroenterology)• Moderate ROM and expected value of 1.7%
Diagnoses – intestinal obstr nos; senile depressive; rheumatoid arthritis; esophageal reflux; depressive disorder;hypopotassemia; palliative care
©2008 University HealthSystem Consortium 12
1.7% may seem low on face value, but it is higher than the model group observed of 1.4%
©2008 University HealthSystem Consortium 13
Oncology Product Line Analysis
qtr3 ’07 through qtr
2 ’08Includes the
Gyn/Onc, Medical Oncand Surgical
OncProduct Lines
To a physician not trained in data management, two responses are likely: 1) what’s wrong with the Oncologists;
& 2) the data is bad
©2008 University HealthSystem Consortium 14
Mortality ReductionMortality Reduction• 100% monthly mortality review using
the case profiles• Begin with mortality in minor or moderate
risk of mortality• Determine if a patient’s death:
1. Can be explained2. Needs further coding analysis (low expected)3. Needs further documentation analysis (low
expected)4. Needs further analysis from a clinical
committee
©2008 University HealthSystem Consortium 15
Risk AdjustmentRisk Adjustment
©2008 University HealthSystem Consortium 16
Key QuestionsKey Questions• Do I feel comfortable with the following items:
– What is the model group being used?– Are the model groups homogeneous enough to
minimize the effect of outliers?– Are the model groups inclusive enough (e.g. only
MedPar vs. all payor)?– Are the sample sizes of the model groups
appropriate?Too low = no statistical significanceToo high = overfitting
– Do the models use the appropriate variables
©2008 University HealthSystem Consortium 17
UHC Risk AdjustmentUHC Risk Adjustment
Each Patient’sCPDF Data
3M MS-DRGGrouper Each Patient
Assigned to a MS-DRG
Each PatientAssigned a ROM / SOI
Literature
Member Feedback
Last 2 Years of Pts in the DRG
CovariatesDeterminedBy DRG
Multiple Regression LOS Model
Multiple Regression Cost ModelLogistic Regression Mortality Model
Expected LOSExpected CostExpected Mortality
Base MS DRGs = model groupUse only AMC patients
©2008 University HealthSystem Consortium 18
©2008 University HealthSystem Consortium 19
High % of patients with low severity & above expected on a few levels
©2008 University HealthSystem Consortium 20
©2008 University HealthSystem Consortium 21
61.6% 62.1%
90.4%
98.9%
3.6%0.3%0.1% 0.1%
30.0%
3.9%0.3% 0.7%0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
60.0%
70.0%
80.0%
90.0%
100.0%
1 2 3 4
Severity of Illness (SOI) class
Pred
icte
d m
orta
lity
% (r
ange
)Why not just use the APRWhy not just use the APR--DRG to predict mortality? DRG to predict mortality?
What does the UHC methodology add?What does the UHC methodology add?If we predicted mortality using only the SOI class, all patients within that class will get the same expected mortality (see red squares). UHC models refine this prediction by using additional variables in risk adjustment. Note the range of expected mortality within each class when UHC models are used.
CDB CY2006. DRG 127 for all UHC hospitals.
©2008 University HealthSystem Consortium 22
OverfittingOverfitting & Palliative Care& Palliative Care
1. Significant Variables and Coefficients 1. Sample Size2. Prevalence
1. In a model group of 1 million, shoe size will be statistically significant, but definitely not relevant
2. Palliative Care / Hospice Care as a variable
©2008 University HealthSystem Consortium 23
TransparencyTransparency
Who are you being compared against?
Remember same questions from risk adjustment regarding model groups
©2008 University HealthSystem Consortium 24
Cross Hairs are national norms for cost & LOS Bubbles are physicians
©2008 University HealthSystem Consortium 25
Build your own custom group from a number of hospital demographics or choose a pre defined group and automatically populate the CDB
©2008 University HealthSystem Consortium 26
Gastroenterology Mortality by Quarter and Year
©2008 University HealthSystem Consortium 27
Cardiology Product Line AnalysisCompared to US News Heart Honor Roll
qtr3 ’07 through qtr 2 ’08
Includes Cardiology and Cardiac Surgery Product Lines
©2008 University HealthSystem Consortium 28
Data Quality ReportsData Quality Reports
CPDF txto UHC
Is it ‘clean’?Does it meet the specs?Are there major errors?
PASS (comparable)ORFAIL & RETURN
Adm Reg Data
Billing Data
Medical Staff Data
CDBCRM
Core MeasuresManagement Reports
Q & A Study
©2008 University HealthSystem Consortium 29
• Will receive after each submission• Can update individual records online
©2008 University HealthSystem Consortium 30
• Tolerance report displays items of most concerned
• Someone at each hospital can go into the records and change these from this tool
©2008 University HealthSystem Consortium 31
Z Score Report
This submission has more patient’s race categorized as Other than Hospital X’s last year’s data
©2008 University HealthSystem Consortium 32
Clinical vs. Administrative DataClinical vs. Administrative Data• In theory, only specificity of certain comorbid
conditions is enhanced by clinical data– Currently, there are approximately 63 codes one can
use for diabetes and 100 for infections– ICD 10 will provide more even more specificity
• Who do you want to be abstracting data?– Coders have extensive training in understanding how
to place certain words and phrases into a code• Clinical data can improve accuracy of risk
models
©2008 University HealthSystem Consortium 33
More Timely DataMore Timely Data
CM Pt AdmittedCM Pt Discharged
CM Pt Chart CodedCM Pt Coded Data Sent to IntermediaryIntermediary Determines CM Pt to be abstractedCM Pt chart abstracted
Day14
35404145
CM Vendors Attempting to Close this Space
The Holy Grail for Timely Core Measures Data
©2008 University HealthSystem Consortium 34
More Timely Clinical Comparative DataMore Timely Clinical Comparative DataEMR, decision support
warehouse, automated pt intake
UHC
data sent when entered
Daily pt expected valuesUpdated comparisonsTransparent utilization
Flags on potential issues
©2008 University HealthSystem Consortium 35©2007 University HealthSystem Consortium 35
Don BerwickDon Berwick
Healthcare’s single most important issue is its inability to improve
One major hurdle to improvement is that very little quality data is perfect
Imperfect data can be very useful in providing direction for improvement efforts … only if you understand the imperfections
©2008 University HealthSystem Consortium 36
Steve Meurer630.954.6677meurer@uhc.edu
CDP Info Line630.954.3792cdpinfo@uhc.edu
Use us like an extensionof your staff
Give us the opportunity to respond to an issue / inconsistency you may have or find
Contact us regularly
top related