Can Newer Prescription Drugs Pay for Themselves Comparing Olanzapine and Lithium to Treat Bipolar Di

1
Can Newer Prescription Drugs Pay for Themselves?
Comparing Olanzapine and Lithium to Treat Bipolar
Disorder

• Yuting Zhang
• November 28th, 2005

2
Research Question

• Can the increased use of newer and more expensive
  drugs offset spending for other types of health
  care services?

3
Bipolar Disorder - Comparing Olanzapine and
Lithium

• A disease that is associated with a high rate of
  hospitalization
• Medication is very effective so drug and non-drug
  treatment are substitutable
• A new drug recently came into market, equivalent
  or more effective than an existing drug
• Not too rare disease

4
Study Design and Analysis

• Data and sample selection
• Empirical strategies
• Interrupted time series
• Differencing strategies
• Instrumental variables
• Results
• Policy implications

5
Eligibility for Bipolar Disorder
Study-MarketScan database 1998-2001

• Had a maintenance drug monotherapy, defined as an
  initial use of a drug continuously for at least
  three months (13,736)
• Continuously enrolled in the network one year
  prior to and one year following the onset of
  maintenance monotherapy (3,073)
• At least one inpatient diagnosis or two
  outpatient diagnoses of bipolar disorder in the
  1998-2001 period (1,492)
• Never diagnosed as schizophrenia and no use of
  any other atypical antipsychotics and other
  bipolar-disorder related drugs 6-month before
  (935)

6
Comparison of Patient Characteristics
7
Identification Strategies

• Propensity score matching
• Empirical strategies
• Interrupted time series
• Differencing strategies
• Instrumental variable

8
Re-Comparison of Patient Characteristics After
Matching
9
(No Transcript)
10
(No Transcript)
11
Empirical Strategies to Address Endogeneity
Problem

• Population-level estimate
• Interrupted time series
• Individual-level estimate
• Differencing strategies
• Fraction of zero outcome
• Highly skewed and heavily tailed
• Hetergeneity in residual
• Instrumental variable

12
Interrupted Time Series
13
Generalized Linear Model (Gamma, Log)
14
Empirical Strategies to Address Endogeneity
Problem

• Population-level estimate
• Interrupted time series
• Individual-level estimate
• Differencing strategies
• Fraction of zero outcome
• Highly skewed and heavily tailed
• Hetergeneity in residual
• Instrumental variable

15
Instrument Variable
16
Two-stage Least Square
17
Comparison of Results (YTotal Non-drug Medical
Spending)
Note monthly average in the year after drug
treatment, significant at 1 level , number
of observation is 9366, robust s.e. clustered by
patients 1st F-test for IV is 15
18
Implications

• Drug benefit coverage design
• Drug approval procedure
• Standards of economic evaluation of drug
  intervention
• Optimal prescribing behavior
• The effect on drug pricing

19
Are You Convinced?Why or why notThank you!
20
Change in Selected National Health Expenditures,
1970-2002
21
Factors Contributing to Increase in Prescription
drug, 1997-2002