Risk Adjustment: Concepts, Choices, and Challenges

1
Risk Adjustment Concepts, Choices, and
Challenges Henry T. Ireys, Ph.D. Associate
Professor, Department of Population and Family
Health Sciences School of Hygiene and Public
Health Johns Hopkins University Baltimore,
Maryland
2
What is risk adjustment? Attempt to distribute
financial resources for medical care in a just
fashion Pay more money prospectively
to cover persons likely to have high service
needs
3
What else is risk adjustment? Strategy designed
to minimize anticipated financial problems for
insurers Method for minimizing corporate efforts
to enroll only healthy populations
4
Whose risk is being adjusted? Ans Insurers
who are at risk for financial insolvency if
expenses exceed income
5
Adjusted risk increase payment rates for
persons who have high service needs
6
The challenge Determine who will have high
service need/cost Create groups along a
continuum of expected need/cost Determine
whether groupings predict actual expenditures
7
How to create groups? Prior use
Demographic variables age, gender,
residential location Clinical descriptors
diagnosis, symptoms Self-reported health
status Functional status
8
Limitations At best 20-25 of future
expenditures can be predicted but a few
percentage points means lots of
money Financing health services A very
dynamic process
9
Risk adjustment methods similarities and
differences Similarities Use ICD-9 codes,
other clinical information Use
expenditures from one year to set
capitation rates in the subsequent
year Try to beat demographic model
in predicting future costs
10
Differences Different method of grouping
ICD-9 codes Different developmental
histories, mostly excluding
children
11
Why so much attention now? Growing number of
elderly Recognition of chronic conditions/ need
for long term care in children Penetration of
managed care into Medicaid populations
12
State of the Art Rapid technical
improvements Expanding field experience
Very recent application to children
13
REPORT ON A STUDY COMPARISON OF RISK
ADJUSTERS FOR CHILDREN WITH CHRONIC
ILLNESSES by Wenke Hwang, PhD Henry T. Ireys,
PhD Gerard F. Anderson, PhD Johns Hopkins School
of Hygiene and Public Health
14
INTRODUCTION Increasing numbers of children with
chronic illnesses and disabilities in capitated
payment systems Serious threats to access and
quality of care
15
What is needed Appropriate capitation rates that
cover additional services Several methods are
available
16
Key question How well can existing methods
predict service costs for children with
chronic illnesses?
17
Capitation Methods/Models Ambulatory
Diagnostic Groups (ADGs) Adjusted Clinical
Groups (ACGs) Hierarchical Coexisting
Conditions (HCC) Disability Payment System
(DPS)
18
Ambulatory Diagnosis Groups (ADGs) 32
categories, not mutually exclusive Based on
primary and secondary diagnoses Grouping
criteria whether the condition persisted over
time, need for continued treatment, other
clinical indicators
19
Adjusted Clinical Groups (ACGs) Derives from
ADGs by using ADG categories as building
blocks 83 mutually exclusive categories
20
Hierarchical Coexisting Conditions (HCC)
Categories based on presence of 1 or 1
conditions and interrelations among coexisting
conditions Adds predicted costs for each
condition to calculate total predicted costs
21
Disability Payment System (DPS) Designed for
child and adult Medicaid recipients with
disabilities 18 categories based on body
system or type of illness/disability 43
subcategories based on association with higher
future costs
22
Study Methods Maryland Medicaid claims data,
1995 - 96 Children lt 18, enrolled in the
Medicaid program for 6 months in 1995 and 6
months in 1996 Included newborns with 2 months
of eligibility in 1995 and 6 months of
eligibility in 1996.
23
Selected Chronic Conditions Asthma
Cystic Fibrosis ADD/ADHD Muscular
Dystrophy Diabetes Neoplasms,
Malignant Epilepsy Hemophilia Sickle
Cell Anemia Arthritis Cerebral Palsy
Congenital Anomalies Chronic Respiratory
Disease
24
Objective Estimate each of the models
predictive accuracy (PA) PA Defined as
extent to which model predicts actual costs
in year 2 based upon information available in
year 1 Data from year 1 used to create
groups according to specifications of each
model
25
Expected expenditures for each person were
calculated based upon the model Data from year
2 were used to calculate actual
expenditures Actual and predicted expenditures
were compared in second year to calculate PA
26
Two approaches to estimate predictive accuracy of
capitation methods Individual-level analysis
Uses entire sample population in a
regression equation Tests how well models
predict expenditure for population as a
whole
27
Group-level analyses Divides population into
strategic subgroups (random or deliberately
skewed) Tests how well models predict
expenditure for each subgroup
28
Results Our sample 111,769 children in
the Maryland Medicaid program, 1995
8.1 of these had at least 1 of 13 chronic
conditions
29
Individual-level
analysis Model Predictive Accuracy Demograph
ic 3.0 ADG 9.5 ACG
8.7 HCC 15.5 DPS 12.5
30
Initial Findings on Under or Over Prediction by
Model for Selected Simulations One percentage
point of over prediction means that model will
overpay an MCO by 1
31
Model No CC 20 Demographic 61
-32 ADG 16 -10 HCC 12
2 DPS 16 - 9
32
Model GP TEACH HOSP Demographic
27 -50 ADG 9 -
7 HCC - 1 - 5 DPS -
2 - 6
33
Summary Diagnosis-based models outperform
demographic model by wide margin Overall, HCC
generates best PA smallest range of
fluctuation in PA with varying types of
random and skewed groups
34
Limitations Important factors not accounted for
administrative feasibility protection of
sensitive patient data ability to resist
gaming Simulations never capture
complexities of health system NACHRI model not
included
35
Conclusions Demographic model, now most common
method, is problematic Emerging risk
adjustment methods account for health
status of enrollees hold promise to reduce
inappropriate incentives
36
Financing health care Rapid evolution in
response to changes in industry
experience consumer demands in
political priorities
37
Child health professionals Understand key
concepts in rate setting Participate in
continuing debates on care for vulnerable
populations