a US based study

Report
Using a Standardized Cost Master to Study
Variation in Hospital Care - a US based study
Raj Srivastava, MD, FRCP(C), MPH
Associate Professor of Pediatrics, University of Utah School of Medicine
Chair, Executive Council, Pediatric Research in Inpatient Settings (PRIS) Network
Fellow, Institute for Healthcare Delivery and Research, Intermountain Health Care
Activity Based Conference 2013, Sydney, Australia
May 15th, 2013
Outline
• Some of the problems we’re facing in the U.S. Health
Care System
• A joint venture between a large hospital research
network and the CEOs of children’s hospitals
• Relevance of pediatric specific methodology and
implications for setting a National Efficient Price in
pediatric hospital care
What are the some of the problems in the US
Health Care System?
Our health care system is complex
We have some waste in health care
Delivering the right care to the right patient at the
right time is challenging
What are some of the pediatric challenges facing
children’s hospitals?
• What are the important conditions/diseases?
• Too much variation in resource utilization for clinical care
• Difficult to track outcomes
• Disconnect between hospital administration and
clinicians and investigators
• Where is the data we need to make the right decisions?
What are the solutions?
• Prioritize conditions
• Communicate to stakeholders
• Measure variation and track outcomes
• Get the right additional data
• Include patient-centered outcomes
• Incorporate trainees
Pediatric Research in Inpatient Settings (PRIS) Network
• PRIS is an independent hospitalist research network
founded through a collaborative effort of three
organizations: the Academic Pediatric Association (APA),
the American Academy of Pediatrics (AAP), and the
Society for Hospital Medicine (SHM)
• Over 700 hospitalists from 86 centers
Core Principles
Perform comparative
effectiveness research
aimed at defining best
practices
Disseminate results to
healthcare institutions
Implement best
practices and measure
patient/cost outcomes
Mission
• Improve the health of and healthcare delivery to
hospitalized children and their families
Membership
What are the solutions?
• Prioritize conditions
• Communicate to stakeholders
• Measure variation and track outcomes
• Get the right additional data
• Include patient-centered outcomes
• Incorporate trainees
• Almost two-fold
differences in
spending across
hospital referral
region
• Explained almost
entirely by
differences in volume
of health care
services received by
similar patients
• Problem with
overutilization
There is variation in Pediatrics
• High variation in resource utilization across children’s
hospitals for a variety of inpatient conditions:
– Osteomyelitis
– Pneumonia
– UTI
– Appendicitis
Overutilization leads to opportunity
• In setting of strong evidence base, excessive
variation signals a need for standardization of care
• Reduce unwarranted variation  reduce costs,
improve outcomes
• Excessive variation may be a symptom of
evidence gaps, and thus signals a need for more
comparative effectiveness research (CER)
Overutilization leads to opportunity
• In setting of strong evidence base, excessive
variation signals a need for standardization of care
• Reduce unwarranted variation  reduce costs,
improve outcomes
• Excessive variation may be a symptom of
evidence gaps, and thus signals a need for more
comparative effectiveness research (CER)
Aims
To develop a screening methodology for identifying
conditions that should be prioritized for CER or QI
– Cumulatively expensive and/or
– Prevalent
– Display high inter-hospital variation in resource
utilization
– Strong evidence base  QI
– Weak evidence base  CER
Study Design and Data Source
Design:
• Retrospective descriptive study
Data Source:
• Pediatric Health Information System (PHIS)
– Hospitals located in 17 of the 20 major metropolitan areas in the
US
– Detailed billing data from 43 freestanding children's hospitals
– Calendar years 2004-2009
– n=3,482,709 admissions
Pediatric Health Information System (PHIS)
hospitals
• Detailed billing data 43 children’s hospitals
• 3.5 million admissions from 2004 – 2009
What’s collected on each patient encounter in
PHIS
Creating a clinically meaningful condition grouper
• Generated lists of ICD-9 CM primary discharge
diagnosis codes associated with hospitalizations
(inpatients, ambulatory surgery, and observation unit)
that accounted for:
– 80% of admissions (high prevalence) and/or
– 80% of total hospital charges (cumulatively expensive)
– 701 ICD-9 CM codes
•
Grouped into 502 distinct “conditions”
“Was the initial clinical evaluation and management of the
diagnoses are the same across individual codes?”
Further Cohort Refinement
• Stratified conditions in ICD9 Grouper into
– Medical: >80% with no surgical procedure
– Surgical: >80% with surgical procedure
– Med/Surg: 20-80% with surgical procedure
• Excluded admissions for medical condition if unrelated
surgical procedure performed
• Included only admissions for surgical condition if related
surgical procedure performed
• Conducted by team of 8 pediatric hospitalists in PRIS
Executive Council
22
Variation in unit costs persists using usual health
services methods
• High variation in unit costs across hospitals  excess
noise in analysis
Cost Master Index
• Created a “cost master index”—a standard cost for
EVERY CTC code—i.e. every billable item (n>22,000)
• Use hospital data (charges and RCCs) to calculate item
costs in every record
• Standardized item cost = median of hospital median
costs (based on RCC)
• Recalculate hospital bills: # Units x CMI Cost
• All costs inflated to 2009 US $ CPI M
Standardizing unit costs
• Median cost for CBC = $32
• 2 CBC’s: cost = 2 x $32
Analysis
Identify conditions with extreme variation in inter-hospital
standardized costs
•
Look at overall distribution of cost/admission
•
Exclude admissions that account for top 1% of hospital admission costs
•
Define overall quintiles of cost/admission
•
Compare distribution (by quintile) of cost/admission across hospitals (binplot)
•
Compare range of cost/admission across hospitals (box-plot), adjusted for
patient factors
•
Count number of hospitals that have >30% of admissions in the lowest
quintile or >30% of admissions in highest quintile: Summary measure of
outlier hospital count
•
Calculate intra-class correlation coefficient cost/admission across hospitals:
Summary measure of inter-hospital variation
Results
• Conditions
– 255 medical
– 231 surgical
– 16 medical/surgical
Example
• ACUTE APPENDICITIS WITHOUT PERITONITIS
– 540.9 : ACUTE APPENDICITIS WITHOUT MENTION
OF PERITONITIS
Appendicitis without Peritonitis
99th %ile = $16,836
Appendicitis without Peritonitis
20th %ile = $ 4,571
40th %ile = $ 5512
60th %ile = $ 6,502
80th %ile = $ 7,889
Distribution of costs for ALL
hospital admissions
Bin Plot, Standardized Cost/Admission
7
8
Box Plot, Standardized Cost/Admission
ICC = 0.19
Out of 50 most
prevalent and 50 most
costly conditions (77 in
total), 26 had ICCs >
0.10 and 5 had ICCs >
0.30
Conditions Sorted by Standardized Cost Rank
Conditions Sorted by Prevalence Rank
Prevalence Rank, Cost Rank, and Variation in Adjusted Standardized Cost Per
Admission for 50 Most Prevalent and 50 Most Costly Conditions*
* Only conditions with an ICC >0.1 or more than 10 outlier hospitals are numbered
Limitations
• “Mixed bag” conditions need to be tossed or broken out
• No clinical outcomes data: High resource utilization may
be warranted if it leads to better outcomes
• Residual data issues
– Missing billing data
– Missing RCCs
– Unreasonably high unit count at specific hospitals resulting in high mean
standardized cost/admission
Conclusions
• Novel approach for prioritizing conditions for CER and
quality improvement work
• Directionally accurate but still requires validation for
individual hospitals and specific conditions
• Have identified a limited set of conditions that are high
cost, high prevalence, and for which there is a lot of
variability in resource utilization across hospitals
Next Steps
• Select conditions for further investigation of extreme
variation: “drill down”
– Identify and correct coding issues
– Identify sources of variation (by clinical category e.g. radiology)
– Identify practice patterns resulting in high resource utilization
• Explore relationship between resource utilization/practice
patterns and outcomes (e.g. complications,
readmissions)
Implications
• Relevance of pediatric specific methodology and
implications for setting a National Efficient Price in
pediatric hospital care
Acknowledgements
PRIS Executive Council
PRIS Network Manager
PRIS Advisory Members
Funding Sources
Patrick Conway
Jaime Blank
Brent James
R&D Grant - Children’s Hospital Association
Ron Keren
Lucy Savitz
Chris Landrigan
CHOP
Charlie Homer
Samir Shah
Xianqun Luan
J. Michael Dean
Sanjay Mahant
Russell Localio
Don Berwick
Karen Wilson
Lisa McCleod
Nate Kupperman
Joel Tieder
Debbie Hillmann
Tamara Simon/Jay Berry

similar documents