Critique of the Week: A Chronic Disease Score with Empirically Derived Weights'

1
Critique of the Week A Chronic Disease Score
with Empirically Derived Weights.

• Daniel Clark, Michael Von Kroff, Kathleen
  Saunders, William Baluch and Gregory Simon.

2
Critique of the Week

• Objective
• To develop a Chronic Disease Score with
  Empirically Derived Weights
• To compare the Revised Chronic Disease Score
  (empirically derived weights) with Original
  (physician assigned weights) CDS and ADGs
• To assess associations between the Predicted
  Costs, Utilization, Hospitalization and Death
  with total costs, outpatient cost, primary visits.

3
Critique of the Week

• Significance
• To develop a reliable and more predictive Risk
  Adjuster for determining Capitation payments.
• A wider range of medications are covered compared
  to the original CDS.
• To asses the correlation between both the CDS and
  ADG to each of the outcome variable (predicted
  costs, utilization, hospitalization and death)

4
Critique of the Week

• Hypotheses
• Not explicitly Conveyed
• Probable Hypotheses
• The Revised CDS explains more variance in
  Predicted Costs, Utilization, Hospitalization and
  Death than the Original CDS and the ADGs.
• The Revised CDS predicts the Costs, Utilization,
  Hospitalization and Death, to a higher degree
  than the Original CDS and the ADGs.

5
Critique of the Week

• Data Collection
• Group Health Cooperation of Puget Sound in
  Western Washington State.
• Adults (18 yrs or older).
• Enrolled continuously at GHC for last six months
  of 1992.
• 245,694 subjects were selected for the study.

6
Critique of the Week

• Results
• Proxy measures for Disease severity (CDS ADGs)
  considerably improve prediction of health care
  utilization and costs.
• Both the Revised CDS and ADGs were equal in
  predicting primary visits and costs.
• Revised CDS was best in prediction mortality.

7
Critique of the Week

• Results

8
Critique of the Week

• Results
• Variance explained in concurrent models was
  higher then the Variance explained for the
  Prospective period.
• There was no clear better predictor (CDSs or
  ADGs) for the outcome variables.

9
Critique of the Week

• Results

10
Critique of the Week

• Results
• For Concurrent Model
• No improvement over the revised CDS was found for
  total cost.
• There was only marginal improvement in variance
  explained with the incorporation of ADGs and the
  original CDS into the same model.

11
Critique of the Week

• Results
• For the Prospective Model
• Improved prediction was achieved by the revised
  CDS relative to the original CDS
• ADGs performed better on all outcomes in the
  concurrent model but the revised CDS performed
  slightly better in explained variance in total
  and outpatient costs

12
Critique of the Week

• Results
• Odds Ratio for Death and Hospitalization were
  estimated from predicted costs and Utilization.
• The Revised CDS had stronger correlation with
  death than ADGs, Original CDS had the weakest
  correlation.

13
Critique of the Week
Results Odds Ratio.

• The odds ratio is a way of comparing whether the
  probability of a certain event is the same for
  two groups.
• An odds ratio of 1 implies that the event is
  equally likely in both groups.
• An odds ratio greater than one implies that the
  event is more likely in the first group.
• An odds ratio less than one implies that the
  event is less likely in the first group.

14
Critique of the Week

• Level of Study Exploratory and
  Descriptive
• Time Frame Ex post facto (retrospective)

15
Critique of the Week

• Operational definitions
• Not clearly given.
• Assumed knowledge of various standardized Disease
  Severity Scores (CDSs, ADGsetc).
• Methods used were well documented and were
  adequate for reproducibility.

16
Critique of the Week

• Methods
• A Revised CDS was created using Regression model
  using medication usage and age and gender, for a
  random half of the population.
• Parameters associated with each medication
  variable were estimated for each of three
  outcomes outpatient cost, primary care visits
  and total costs.

17
Critique of the Week

• Methods
• These weights were then used to calculate a
  predicted score for total cost, Outpatient cost
  and primary care visits for the other half of the
  population.
• The weights were also used to predict the outcome
  parameters for a Prospective Cohort.
• The Predicted parameters were then correlated
  with the actual parameters. Also the Scores were
  correlated with risk of Hospitalization and
  Death.

18
Critique of the Week

• Controls
• Randomization
• Testing the Regression model Scores on both
  Concurrent and Prospective gorups.

19
Critique of the Week

• Identified Sources of Bias.
• Revised CDS obtained from six months data
  compared to twelve data for Original CDS.
• Not all enrollees had medications filled in the
  six month period.
• Manipulation.
• Endogeneity.

20
Critique of the Week

• Generalizability That particular Health
  Maintenance Organization (HMO).
• Future Study Testing and Validating the Revised
  CDS at different sites than the one used here.

21
Critique of the Week

• Weak Research Methodology
• Typical of a Pharmacoeconomic Study
• Strong yet Simple statistical methods
• Least Square Regression
• Odds Ratio