Rong Yi, Senior Research Associate, DxCG Inc'

1
Better Provider Profiling Adding Patient Risk
Adjustment to Episodes
Rong Yi, Senior Research Associate, DxCG Inc.
Academy Health Conference June 6, 2004
2
Questions Keeping Us Up at Night
The credibility of a provider profiling
initiative depends largely on how well we can
respond to the my patients are sicker
objection.

• How much do patient comorbidities affect episode
  costs?
• Could we use the DCG Relative Risk Score that
  measures patient risk along with Medstats
  Episode Group method to more accurately determine
  expected costs?
• How do severity and risk-adjusted episodes change
  our provider profiling results?

3
Presentation Outline

• Project objectives
• Proposed methodology
• Application to BCBS of South Carolinas provider
  profiles
• Conclusions

4
MEGsMedstats Episodes Grouper
Look-back
Episode
Clean Period
Office Visit
Office Visit
Prescription
Lab
Hospital Admission
Office Visit

• Links together a patients claims into a
  clinically meaningful episode across care
  settings
• Calculates summary episode cost and utilization
  metrics
• Assigns a managing physician to the episode to
  support profiling
• Determines the disease stage of the episode
  (highest)

5
DxCGs DCG/HCC Models
Clinical Categories
Diagnosis Information
Age/Sex
Patient Risk Scores

• Developed using regression methods on Medstats
  MarketScan database (commercial model)
• Model input includes demographic information and
  all diagnosis information (and/or drug
  information) for a patient for a period
  (typically a year)
• Assigns a set of risk scores to the patient that
  measures current and future risk (used for
  adjustment in profiling and predictive modeling).

6
DCG Calculating a Patients Risk Score

• 0.45 54 year old male
• Condition Categories
• 5.71 Diabetes with renal manifestation
• 1.84 Congestive heart failure
• 0.90 Acute myocardial infarction
• 0.89 Vascular disease with complication
• 0 Vascular disease hierarchy
• 18.09 Dialysis status
• 0.46 Diabetes congestive heart failure
• interaction
• ______
  
  
• 29.34 Relative Risk Score

Member ID 00001Name John SmithAge
54 Sex MRel Risk Score 29.34
7
How Much Should an Episode Cost? - Depends on
the patient!
8
Cost of Pneumonia and Patient Risk (DCGs)
DCG Risk Score
Healthier.. Sicker
9
Cost of Chronic Diabetes Patient Risk (DCGs)
DCG Risk Score
Healthier.. Sicker
10
Overall Relationship between Episode Disease
Stage and Patient Illness Burden
Stages are not comparable across MEGs, but
broadly higher stages go with higher risk scores.
11
Risk-Adjusted Episodes in Provider Profiling

• Problem While episodes can be
  severity-adjusted, without adjusting for patient
  risk, there is the potential to unfairly reward
  physicians who care for patients with few
  co-morbid diseases and penalize those who
  effectively care for patients with significant
  disease burden.
• Proposed solution Marry the patient-level risk
  scores from DxCG with the severity score within
  Medstats Episodes Grouper to fairly evaluate
  physicians and pay for the best performance.

Medstats Episodes Grouper Severity-adjustment
within Episode
DxCGs HCC Model Whole-patient Relative Risk
Score
Risk-adjusted Episodes
12
Developing the Risk-Adjusted Episode Model

• MarketScan database
• Only complete episodes with enough time for claim
  run-out
• 20 million episodes in 2002
• Regression models incorporating
• MEG
• Disease Stage
• DCG/HCC Prospective Relative Risk Score
• Predict episode cost within each MEG

13
Improvements in Predictive Power (R2)
14
Overview of BCBS of South Carolinas Profiling
Efforts

• 1 million enrolled members
• Provider Contracting deals with over 1600
  physicians
• Profiling effort began in 1998, using the MEGs.
• Profile specialists with more than 100 members on
  episode cost and use information and compare to
  specialist norms

15
Risk-Adjusted Episode Profile for Internal
Medicine / General Practice
16
Performance Ratios by Physician

• Performance ratios for most physicians (the ratio
  of actual / expected ) are similar between the
  two methods.
• Some physicians performance ratio changes
  significantly when we add patient risk to the
  adjustment.

Adjusted using episode group and stage Adjusted
using episode group, stage and patient risk
Physician H moved from an outlier to practicing
within expected range Physician D is even
more of an outlier
17
Physician D Drill Down
This physicians patients have consistently lower
illness burden than expected.
18
Physician H Drill Down
Physician H moves from an outlier to within the
norm due to treating more severely ill patients.
19
Conclusions

• Episode costs increase with the severity of the
  disease (MEG) and disease burden of the patient
  (RRS)
• Considerable variation in episode costs leaves
  room for risk adjustment
• Organizations can improve the accuracy of
  provider performance assessments using
  risk-adjusted episodes. This is important for
  ensuring equitable pay-for-performance.
• Plans are underway to incorporate risk-adjusted
  episodes into the Medstats standalone episode
  grouper and Advantage Suite.