Pharmacoepidemiology Roundtables  1998-1999


EVALUATING DIFFERENTIAL COST SHARING OF DRUGS USING ADMINISTRATIVE DATABASES:
HYPOTHESES, COUNTERFACTUALS AND CAUSAL GRAPHS WITH IMPLICATIONS FOR THE ANALYSIS.

Sebastian Schneeweiss M.D., S.M.

ABSTRACT
Background: Differential cost sharing (DCS) of prescription medications  is of growing importance in pharmaceutical benefits plans. However there is little evidence of the effects on drug- and health care utiliza-tion and even less is known about the impact on health outcomes and overall costs.
Objective is to develop models to evaluate DCS form a policy maker’s and a clinician’s perspective using the recent introduction of DCS for ACE inhibitors in British Columbia as an example.
Results: Benefits plan managers are mainly interested in the overall performance of their plan. A Policy Model tests, whether there are changes in the time trends of outcomes after the introduction of DCS compared to before the DCS policy. This will lead to better estimates of the overall effect and practical consequences of the policy maker’s decision. However, estimates are summary effects and null effects might be the average of a benefit in non-compliers and harm in compliers. Results apply to a specific policy implementation and are limited in their generalizability. Segmented linear regression or generalized linear models for repeated measurements can be used for analysis. Clinical decision maker and patients are interested in the consequences, given their actual compliance to the policy. A Clinical Model assesses the effects of DCS as-treated, isolated from program specific effects such as exemption rules. However, the model must make partly unprovable assumptions of the appropriate control of selection processes. More complex generalized linear models controlling for selection factors must be employed. Instrumental variable estimation should be considered to overcome the limitations in measuring these factors.
Conclusions: Both, Policy and Clinical Models should be tested with a clear understanding of their interpretation, using quasi experimental time-series designs and corresponding analysis techniques to evaluate the effects of DCS.

Keywords: Differential cost sharing, Drugs, Causal graphs, Health policy analysis, Epidemiologic methods

1) INTRODUCTION
2) BACKGROUND TO DIFFERENTIAL COST SHARING OF PRESCRIPTION MEDICATIONS
    Differential Cost Sharing in the United States
    Existing Evidence on the Effects of Differential Cost Sharing
    The Natural Experiment of Reference Pricing in British Columbia as a Model for Evaluating Differential Cost Sharing
3) HYPOTHESES IN THE EVALUATION OF DCS AND THEIR INTERPRETATIONS
    The Policy Model (PM):
    The Clinical Model (CM):
4) UNDERSTANDING SELECTION PROCESSES USING CAUSAL GRAPHS
    Measurement of Intervention and Outcomes in the Clinical Model
    Intermediate Factors that originate Policy Compliance
    Patient Related Factors
    Physicians Related Factors
    Health System Related Factors
    Unknown or Unmeasurable Factors
5) IMPLICATIONS FOR THE ANALYSIS
    Policy Model: Modeling Time
    Clinical Model: Controlling for Selection Factors
    Clinical Model: Instrumental Variable Estimation to control for Unmeasurable Selection Factors
 
 


If you have questions or comments about the Seminars or want to be included in a mailing list, please send an e-mail to Kathy Putnam.

Return to the top of this document

Return to DDIL HOMEPAGE

This document has been provided through the Drugs and Devices Information Line, http://www.hsph.harvard.edu/Organizations/DDIL/ddil.html
If you have questions or comments about the DDIL, please write to: sschneew@hsph.harvard.edu