Controlling for systematic selection in retrospective analyses: an application to fluoxetine and sertraline prescribing in the United Kingdom

Can we account for selection bias? A comparison between bare metal and drug-eluting stents

OBJECTIVE

In this article we investigate the possibility to account for selection bias in observational data by using econometric techniques.

METHODS

One-year costs of 15,237 patients who received a drug-eluting stent (DES) or a bare metal stent (BMS) in Belgium in 2004 were compared. The treatment choice between DES and BMS could be influenced by patient characteristics; therefore, cost estimates could be biased by overt and/or hidden selection bias. Overt bias was addressed by regression adjustment and propensity score matching. Hidden selection bias was dealt with by using an instrumental variable (IV) approach.

RESULTS

Due to the higher purchase price DES patients incur higher (unadjusted) costs in the short-term; these costs are, however, compensated in the long-term due to less in-stent restenosis and hospitalizations. Analyses indicated that, for the diabetic population, the null hypothesis of similar average 1-year costs of patients receiving a BMS or DES cannot be rejected. For the non-diabetic patients a significant cost difference between BMS and DES patients was found. It cannot be ruled out that the treatment-effect model does not correct for all observable or unobservable characteristics and that the estimated treatment effect is biased, possibly due to weak instruments.

CONCLUSION

It is interesting and necessary to explore the use of econometric tools in cost and cost effectiveness analysis to investigate the effect of a technology in everyday practice and to take into account patient and disease characteristics and uncertainty. Further research is however necessary to investigate how we can fully correct for selection bias when using observational data.

The use of linear instrumental variables methods in health services research and health economics: a cautionary note

OBJECTIVE

To investigate potential bias in the use of the conventional linear instrumental variables (IV) method for the estimation of causal effects in inherently nonlinear regression settings.

DATA SOURCES

Smoking Supplement to the 1979 National Health Interview Survey, National Longitudinal Alcohol Epidemiologic Survey, and simulated data.

STUDY DESIGN

Potential bias from the use of the linear IV method in nonlinear models is assessed via simulation studies and real world data analyses in two commonly encountered regression setting: (1) models with a nonnegative outcome (e.g., a count) and a continuous endogenous regressor; and (2) models with a binary outcome and a binary endogenous regressor.

PRINCIPAL FINDINGS

The simulation analyses show that substantial bias in the estimation of causal effects can result from applying the conventional IV method in inherently nonlinear regression settings. Moreover, the bias is not attenuated as the sample size increases. This point is further illustrated in the survey data analyses in which IV-based estimates of the relevant causal effects diverge substantially from those obtained with appropriate nonlinear estimation methods.

CONCLUSIONS

We offer this research as a cautionary note to those who would opt for the use of linear specifications in inherently nonlinear settings involving endogeneity.

Two-stage residual inclusion estimation: addressing endogeneity in health econometric modeling

The paper focuses on two estimation methods that have been widely used to address endogeneity in empirical research in health economics and health services research-two-stage predictor substitution (2SPS) and two-stage residual inclusion (2SRI). 2SPS is the rote extension (to nonlinear models) of the popular linear two-stage least squares estimator. The 2SRI estimator is similar except that in the second-stage regression, the endogenous variables are not replaced by first-stage predictors. Instead, first-stage residuals are included as additional regressors. In a generic parametric framework, we show that 2SRI is consistent and 2SPS is not. Results from a simulation study and an illustrative example also recommend against 2SPS and favor 2SRI. Our findings are important given that there are many prominent examples of the application of inconsistent 2SPS in the recent literature. This study can be used as a guide by future researchers in health economics who are confronted with endogeneity in their empirical work.

Retrospective analysis of the health-care costs of bupropion sustained release in comparison with other antidepressants

OBJECTIVE

The objective of this study was to evaluate the health care costs associated with the treatment of a new episode of depression with bupropion sustained release (SR) rather than with other antidepressants (selective serotonin reuptake inhibitors [SSRIs], tricyclic antidepressants [TCAs], and serotonin norepinephrine reuptake inhibitors [SNRIs]).

METHODS

This was a retrospective cohort study based on the private-pay, fee-for-service 1997 and 1998 MEDSTAT MarketScan databases. Individuals were included if they were 18 years of age or older, had an initial prescription for an antidepressant under study with an index prescription date between July 1997 and June 1998, and had a claim for a diagnosis of depression diagnosis within 30 days of the index date. All patients' claims from six months before and after receiving their index antidepressant prescription were examined. Total, outpatient, and pharmacy costs were compared among antidepressant groups using an intent-to-treat analysis with exponential regression models and bootstrapped 95% confidence intervals.

RESULTS

A total of 1771 patients were included in the study cohort. The mean age was 41.6 years, and 69.5% of subjects were female. Most patients (75%) continued with the index antidepressant during the 6-month follow-up period. Although the drug acquisition cost was lowest for TCAs, total costs were significantly higher for patients treated with TCAs than for those treated with bupropion SR (p < .05). In comparison with bupropion SR, patients initiating therapy with sertraline had significantly higher mental health payments (p < .05).

CONCLUSIONS

Initiating treatment of depression with bupropion SR was associated with lower total mental health care costs compared with TCAs and with sertraline. This study reaffirms that formulary and medical decision-makers should consider the overall impact of antidepressant treatment, including but not limited to drug acquisition costs, other health care costs, and drug efficacy and safety.