There are two major findings in our study. One is that the increase in prevalent ARB users is associated with switching from being an ACEI user. This result is consistent with the long-term trend increases in utilization of ARBs, but a long-term trend decrease in utilization of ACEIs. The other is that the increase in expenditures on renin-angiotensin drugs throughout this study primarily resulted from increases in the number of incident ARB users and potential switching from ACEIs to ARBs, but not from changes in the use of overall renin-angiotensin drugs. Ultimately, cost savings have not been achieved in overall use of renin-angiotensin drugs. Our findings differ from those of previous studies that showed cost savings in pharmaceutical expenditures for all PBS listed drugs [5
] and generic substitution driven by the financial incentive for physicians [6
] after price adjustments.
As mentioned, a feature of Taiwan’s health care system is that physicians both prescribe and dispense drugs because they are permitted by Taiwan’s Department of Health to hire pharmacists to work at their on-site pharmacies [2
]. Physicians stand to profit from the gap between the reimbursement price and the market trading price. In 2009, Liu et al. [6
] conducted a study in Taiwan to investigate prescribing preferences between original branded and generic drugs in relation to this potential profit margin, and found that prescribing of generics increased as the reimbursement price decreased. Our results were only consistent with these findings when we examined the change in utilization of off-patent ACEIs. We observed that the utilization of off-patent original branded ACEIs was lower than that of off-patent generic ACEIs since 2001. Although the initial upward trend in the use of off-patent generic ACEIs turned to a decreasing trend after the implementation of PA2005, the use of these drugs is still higher than that of off-patent original branded drugs. However, our findings differ to those of Liu et al. when looking at both ACEIs and ARBs. Since these two classes of drugs work through the same renin-angiotensin pathway and have a similar effect in cardiovascular disease and renal protection [7
], physicians make a prescribing decision between ACEIs and ARBs when the patient needs a renin-angiotensin drugs. We observed that, following PA2006, the initial rising trend of the use of ACEIs turned to a decreasing trend, primarily due to the decreases in the use of off-patent ACEIs. The initial rising trend of the use of ARBs further increased following PA2003 and PA2007, and the use of ARBs exceeded that of ACEIs from 2007. In fact, the profit margin from patented and off-patent drugs varies over time, depending on when a drug turned from patented (applying WAP adjustment) to off-patent (applying GWAP adjustment) as well as the value of the r-zone (the accessible profit margin for physicians) imposed by the BNHI. Because the extent of WAP adjustment for patented drugs is smaller than that of the GWAP adjustment for off-patent drugs, and because the r-zone is larger for patented drugs than for off-patent drugs, patented drugs offer higher financial incentive to physicians through stepwise price adjustments.
Another study conducted by Lee et al. [5
] addressed the issue of Taiwan’s cost containment strategies on pharmaceutical expenditures. The authors of that study also applied Box-Tiao intervention analysis to examine the level change (the immediate effect) after the implementation of drug price adjustments. They found that pharmaceutical expenditures on all PBS listed drugs significantly decreased after the introduction of price adjustments based on generic grouping (PA2001 and PA2003). Our study included a time trend variable in the model, which enabled us to investigate the long-term effects of price adjustments. In fact, we found that the long-term trend in expenditures on overall renin-angiotensin drugs increased after PA2001 and PA2007.
Exploring increases in expenditures on overall renin-angiotensin drugs, we found that annual incident renin-angiotensin drug users declined over time, and no significant trend increases were found in the utilization of renin-angiotensin drugs over the study period. We also found that annual incident ARB users and annual prevalent ARB users increased over time. In particular, the annual prevalent ARB users always exceeded the number of cumulative incident ARB users, indicating that annual prevalent ARB users were not only from the category of cumulative incident ARB users, but also patients who were ever treated with ACEIs. These findings suggest that the increase in expenditures on renin-angiotensin drugs throughout this study primarily resulted from increases in the number of incident ARB users and potential switching from ACEIs to ARBs, but not from changes in the use of overall renin-angiotensin drugs.
Although the health care systems in Canada and European countries are different from that in Taiwan, and the reference pricing (RP) scheme [18
] adopted in these countries control cost from the demand side (patients) but not from the providers (physicians) side, some of the studies conducted reveal that the RP scheme did not have any long-term effects. Evidence from the Netherlands showed that, after the implementation of the RP scheme, the cost of drugs covered by the RP scheme increased less than predicted, but the cost of drugs outside of the RP scheme has increased annually [19
]. In addition, evidence from Germany and Hungary showed that pharmaceutical expenditures are still not well controlled because of volume growth that occurred after the implementation of the RP scheme [20
Previous studies have reported that the use of cheaper, generic drugs may lead to cost savings, but clinical concerns regarding patient safety and therapeutic effectiveness related to treatment discontinuation have also been raised [22
]. Our study demonstrated that prescription switching was from cheaper drugs to more expensive agents, and our patients, with a complexity of clinical conditions, were more likely to be treated with both drugs (subsequent or concurrent use of ACEI and ARB) than ACEIs alone. It is noteworthy that a difference between ARBs and ACEIs may be the persistent coughing caused by ACEIs. Patients who cannot tolerate ACEIs often switch to ARBs. No differences in the clinical recommendations for ARBs and ACEIs were noted during the study period [7
]. Further analyses are warranted to compare the effectiveness and economic outcome for patients treated with ACEIs, those treated with ARBs, and those who switched from ACEIs to ARBs.
This study had several limitations. First, the baseline disease severity which led to initiation of treatment with ACEIs or ARBs was not compared, and we were unable to distinguish the clinical appropriateness of stopping or switching drugs for individual patients. The assessment of clinical conditions using ICD9-CM codes is likely to minimize most, but not all, of the potential bias. Second, there was no control group in this study, because price adjustments were implemented nationwide concurrently. However, the time points in the pre-intervention period served as a control group for the post-intervention period in the intervention analysis. Thus, issues regarding internal validity (such as history and maturation) were taken into consideration [14
]. Finally, because the time period between price adjustments is quite short, incorporating each price adjustment’s level change and trend change into the intervention model would show severe multicollinearity, and it would be difficult to achieve significance for the collinearity parameters. That is why some previous studies only examined the level change, but not the trend change. However, since the trend change represents the long-term effect of policy interventions, its implication is greater than what the level change can explain and so it cannot be ignored. Therefore, we adopted a parsimonious model instead of a full model [14
], keeping only the significant predictors selected by a backward elimination procedure and collinearity diagnostics. In this way, the long-term effect of the price adjustments can be examined and the multicollinearity problem can be avoided.