New Drugs vs. Old Drugs

New drugs are (typically) more expensive than old drugs. If a hospital decides to use the latest drug technology, this will increase costs…right?  

This may not always be the case.  If new drugs are more effective than old drugs, then giving patients the new expensive drug may cuts costs in other areas (such as decreased hospitalizations). Lichtenberg (2009) examines whether utilizing new cardiovascular drugs has cut costs for 20 OECD countries between 1995 and 2003.


Lichtenberg uses a diference in difference framework as follows:

  • ln Yit = β(RxVintageit) + γZit + αi + δt + εit

The dependent variable, Yit, measures either hospital use or mortality due to cardiovascular disease.  The vector Z represents a host of variables which could explain cardiovascular related morality or hospital admission. The variables αi, δt are country and year dummy variables effects. 

The variable of interest is the RxVintage.  This variable is the weighted average of the launch year of the cardiovascular drugs used.  For instance, let us assume that drug A was launched in 1996 and drug B was launched in 2002.  If the U.S. used 2/3 drug B and 1/3 drug A, then RxVintage =1998.  If on the other hand, they used 1/3 drug A and 2/3 drug B, then RxVintage=2000. [Lichtenberg actually uses a simpler technique where a drug gets a value of 1 if it was launched after 1995 and a value of 0 otherwise.]

In essence, this specification compares changes in hospitalization rates among countries that increased RxVintage over time with those who did not.


Of all the countries, Australia, Canada and Norway were mostly likely to use newer vintage drugs.  The U.S. placed 5th.  The Czech Republic, Poland, Hungary, and the Slovak Republic were lease likely to use newer drugs among OECD countries.  

Countries with larger increases in the share of cardiovascular drug SUs [standard units] that were post-1995 SUs had smaller increases in the cardiovascular disease hospital discharge rate, controlling for the quantity of cardiovascular medications consumed per person, the use of other medical innovations (CT scanners and MRI units), consumption of calories, tobacco, and alcohol, and demographic variables (population size and age structure, income, and educational attainment)...The estimates indicate that if drug vintage had not increased during 1995–2004, hospitalization and mortality would have been higher in 2004.  We estimate that per capita expenditure on cardiovascular hospital stays would have been 70% ($89) higher in 2004 had the drug vintage not increased during 1995–2004.

For cardiovascular drugs, it looks like new drugs win the day.


  1. The controls in the association studies conducted by Dr. Lichtenberg are completely arbitrary . . . the only other medical innovations in cardiovascular disease besides drugs were CT scans and MIRs? Also, in its demographic data, it completely misses the point that differentials in income and educational attainment within groups is far more important than absolute levels in determining health status in general and cardiovascular disease in particular.

    You also fail to report that Dr. Lichtenberg’s research is almost always funded by either the drug industry and promoted by conservative think tanks. The journal Health Economics is a long-time bastion of such industry-funded research.

    I am not a trained economist, but I can spot a flawed association study when I see one. Double-blind controlled trials has always been considered the gold standard for medical research. And the massive controlled trials conducted on cardiovascular medicines — the ALLHAT study, for instance — have almost always come to the opposite conclusion from Dr. Lichtenberg, i.e. old drugs are just as effective as newer medications in preventing heart disease.

  2. Dr. Lichtenberg uses a difference and difference specification. Thus, he controls for country-specific characteristics. If one country has more income or education inequality, this will not effect the estimates (since he includes country-specific dummy variables). If, however, changes income and education inequality are correlated with the a countries takeup of new drugs, the could affect the results. However, Lichtenberg does control directly for average income and education levels.

    A more realistic issue would be if there are significant expansions of insurance coverage or insurance benefits. Increased insurance coverage/benefits can lead to newer drug adoption, but can also cause other health benefits (through better care in other areas). It is possible that increased insurance coverage is driving these results, but Lichtenberg does not document any significant changes in health insurance policies in the OECD countries under study.

    I do not have any information concerning whether or not Dr. Lichtenberg is an unbiased researcher.

Leave a Reply

Your email address will not be published. Required fields are marked *