Modeling Risk Adjusted Capitation Rates in Regione Umbria

1
Modeling Risk Adjusted Capitation Rates in
Regione Umbria

• Elaine Yuen, PhD Daniel Z. Louis, MS
• Paolo DiLoreto Joseph S. Gonnella, MD
• American Public Health Association Meeting
• October 22, 2001

Thomas Jefferson Center for Research Jefferson Me
dical in Medical Education University College and
Health Care
2
Project Overview

• Purpose to risk adjust per capita reimbursement
  rates
• Age sex adjustment
• Severity of illness adjustment
• Three major tasks
• Collection and compilation of data from Regione
  Umbria
• Risk adjustment using US Medicare PIP-DCGs
• Risk adjustment using Disease Staging

3
Description of Study Database

• Data from Regione Umbria, 1997-1998
• Hospital data
• day and ordinary admissions
• DRGs and DRG based tariffs
• clinical and demographic information
• Umbria residents hospitalized in Umbria and other
  regions
• Pharmacy data
• individual prescription level
• captured drug codes, tariffs and co-pays
• Demographic file
• age, sex, USL

4
Mean Tariffs per Year
5
1998 Tariffs by Age and SexEntire Umbria
Population
6
Disease Staging

• Clinically-based patient classification system
• Over 400 disease categories
• Based upon disease etiology, organ involvement,
  and severity of comorbidity.
• Computerized algorithm uses ICD-9-CM codes
• Severity of illness stages
• Stage 1, conditions with no complications or
  problems of minimal severity
• Stage 2, problems limited to an organ or system,
  with significantly increased risk of
  complications
• Stage 3, diseases with multiple site involvement,
  generalized systemic involvement, and/or poor
  prognosis

7
Use of Staging for Severity Adjustment

• All admissions were aggregated by Disease Staging
  category and severity stage
• Reviewed by clinicians for propensity of
  affecting future year resource use
• Excluded clinical categories
• Acute illnesses that can be cured, e.g. Stage 1
  Appendicitis
• Vague signs/symptoms with no etiology at Stage 1
  or 2
• Chronic diseases that were cured, e.g. Stage 1
  Cholecystitis after cholecystectomy

8
Use of Staging for Severity Adjustment (continued)

• Included clinical categories
• All Cancers (except basal cell)
• All stages of Central Nervous System,
  Cardiovascular, and Respiratory Diseases
• Stage 2 and 3 of Gastrointestinal, Hemapoetic,
  Renal, and Endocrine
• HIV/AIDs
• Impact on future year tariffs of included cases
  were
• Minimum
• Moderate
• Severe

9
Worksheet for Clinical Categories
10
Descriptive Statistics

• Used clinical and demographic information, 1997
  Test database
• Aggregated admissions if there were less than 50
  cases in any one category
• Considered 155 unique clinical categories within
  5 larger categories
• Cancer, HIV, Minimum, Moderate, Severe
• Collapsed admissions into person-level file and
  merged with demographic data
• Test database (N411,539 persons)
• 87.21 (N358,893) were not hospitalized in 1997
• 7.51 (N30,908) were excluded from our severity
  categories
• 5.28 (N21,738) persons were considered in the
  models

11
Included and Excluded Severity CategoriesTest
Database, Regione Umbria
12
Risk Adjustment Models Predicting 1998 Tariffs

• Models were built at the individual person level
• Used a split sample
• One part of the data was used for modeling
• The other part for the testing of model
• TOTAL COSTS in 1998 f (clinical categories in
  1997 age/gender cohorts error)
• 22 age-sex cohorts
• Disease Staging - 133 clinical categories in 1997
• PIP-DCGs - 15 PIP-DCGs in 1997

13
Predicted VS Observed TariffsAge-Sex Adjustment
Only
14
Predicted VS Observed TariffsDisease Staging
Groups
15
Predicted VS Observed TariffsDisease Staging
Groups
16
Predicted VS Observed TariffsPIP-DCG Groups
17
Limitations

• Case finding
• Uses hospitalization data to identify a persons
  severity of illness
• Persons who are ill but may not be hospitalized
  are not captured (for example, someone with
  diabetes who uses only outpatient care)
• Uses only hospitalization and pharmaceutical data
  to calculate tariffs
• Ideally would calculate all costs of medical care
  
• Use of GP and/or outpatient services may vary by
  condition

18
Where do we go from here?

• Refine model
• Outpatient or GP data included in year 2 costs
• Separate models for hospital and pharmacy tariffs
• Re-run with more recent data
• Re-calibrate Disease Staging groupings
• Improve case finding, possibly using
  pharmaceutical data
• Estimate impact
• On USL or distretto within the region
• For different demographic cohorts