健保藥費分次調降對口服處方用藥之影響 ~以FELODIPINE心血管用藥為例

Impacts of Taiwan drug price cuts on prescribed oral medication ~An example of cardiovascular drug FELODIPINE

陳贊元(Chan-Yuan Chen)

指導教授 ： 張永源

高雄醫學大學/健康科學院/醫務管理學研究所碩士在職專班/碩士(2010年)

摘要

研究目的
醫療費用逐年上漲是多年來的趨勢，其中藥品費用支出約佔整體醫療費用支出四分之一左右，中央健保局為了抑制不斷上漲的醫療費用，自2000年4月1號第一次實施例行性健保藥價調整政策，迄今已經歷六次藥價調查。因此，為了解藥價調查政策對口服藥品的影響，本研究選擇單一成分單一用途的 Felodipine 心血管用藥來探討前五次藥價調整後藥品耗用的價、量長期變化，以期提供健保局在費用控制上的參考與修正。

研究方法
本研究採回溯性縱貫性研究(Retrospective Longitudinal Study)，推論性統計研究期間為2000年4月1日至2008年12月31日止。利用國家衛生研究院所提供的全民健保資料庫-門診(500:1)及住院(20:1)隨機抽樣次級資料檔，建立國內Felodipine藥品的醫療耗用資料。健保用藥品項符合成分名稱 Felodipine 藥品品項代碼共計有12種，經過適當排除及重整，最後將75,970筆資料納入分析。統計分析使用 SPSS for Windows 套裝軟體，採用 Independent-Samples T-test、 One-Way ANOVA、 Linear Regression、 Logistic Regression 及 Generalized Estimating Equations (GEE) 等統計方法，進行資料分析與驗證研究假設。

研究結果
本研究借重概化估計方程式(Generalized Estimating Equations)來處理時間效應在例行性藥價調查政策於一段時間內量測6次藥品耗用數量與金額的變化，並利用重複量數的數值，來精準估計各預測因子影響結果。

研究結果發現: 在不同的依變項模式內控制其它預測變項的情況下— (1)分別僅有藥品劑量在每醫令平均數量 (p=0.521)、藥價調降幅度在平均數量 (p=0.433)、平均金額 (p=0.906)以及每醫令平均數量 (p=0.425)無統計學上的顯著差異，其餘預測變項皆在藥品耗用的數量與金額有顯著差異。(2)僅有藥品劑量(p=0.983)及醫院權屬別(p=0.789)對藥品醫令開立廠別傾向無統計學上的顯著差異，其餘預測因子皆在藥品廠別傾向有顯著差異。

一、預測因子對平均數量的影響:
1. 醫院權屬別平均數量比較依序為:診所>公立醫院>財團法人醫院>私立醫院，亦即診所在Felodipine藥品耗用平均數量上為最多。
2. 按醫院層級別來比較平均數量依序為:醫學中心>區域醫院>地區醫院>診所，亦即醫學中心在Felodipine藥品耗用平均數量上為最多。

二、預測因子對平均金額的影響:
1. 醫院權屬別平均金額比較依序為:診所>公立醫院>財團法人>私立醫院。
2. 醫院層級別平均金額比較依序為:醫學中心>區域醫院>地區醫院>診所，唯地區醫院對診所並無顯著差異(p=0.55)。

三、預測因子對每醫令平均數量的影響:
1. 醫院權屬別每醫令平均數量比較依序為:診所>財團法人醫院>公立醫院>私立醫院，唯私立醫院對公立醫院並無顯著差異(p=0.209)。
2. 醫院層級別每醫令平均數量比較依序為:診所>醫學中心>地區醫院>區域醫院，唯診所對醫學中心並無顯著差異(p=0.572)。

四、預測因子對每醫令平均金額的影響:
1. 每醫令平均金額原廠藥高出學名藥許多，若使用學名藥則每醫令可在該藥品平均省下70元之多。
2. 醫院權屬別每醫令平均金額比較依序為:診所>財團法人>公立醫院>私立醫院，唯私立醫院則對公立醫院無顯著差異(p=0.082)。
3. 醫院層級別每醫令平均金額比較依序為:診所>醫學中心>地區醫院>區域醫院，診所對醫學中心無顯著差異(p=0.584)。
4. 雖然後期的每醫令Felodipine平均數量較高，但相對的每醫令Felodipine的平均金額卻低了許多。
5. 藥價調降幅度對每醫令平均金額呈現顯著負相關性(p=0.011)，當藥價每調降1%，則每醫令平均金額就少了1.63元(p=0.011)。

五、藥品醫令開立對藥品廠別傾向之預測模式:
1. 藥品劑量並不影響醫令開立的藥品廠別傾向。
2. 醫院層級別開立學名藥傾向比率依序為:診所>地區醫院>區域醫院>醫學中心，唯地區醫院對診所無顯著差異(p=0.156)。
3. 依藥價調整區間來看，可推論區間越往後，則選用學名藥的比率越高。
4. 在考量時間效應及控制模式中其它變項後，當藥價每調降1%，則醫令開立學名藥相對開立原廠藥的機率會減少0.06倍，亦即減少了6% (p=0.021)。

結論與建議
業界有一句話是這麼說的:「價格是藥品最寶貴的生命」。例行性藥價調查一直以來是健保局遏止藥費不斷上揚的利器，但過度強勢大砍藥價，有時反而利基於其它專利內高貴新藥，並可能造成反饋性整體藥費的上揚。天平兩端「價」與「量」上的權衡，有待健保局細思量。

有鑑與此，特對政策擬定者提出以下兩點建議：

一、 考慮採用經建會所建議之合理藥品費用占醫療費用比例 24.3% 為標準，超出目標才調整藥價。
二、 考慮採用費用協定委員會所訂定之建議，藥費應與醫療費用具相同成長幅度，超出目標才調整藥價。

關鍵字

權屬別 ； 層級別 ； 原廠藥 ； 學名藥 ； 藥價調查

並列摘要

Objectives
The trend of medical costs keeps rising over the years. Of the all expenses, drug expenditure accounts for around 1/4 and NHI Taiwan has implemented the routine drug price surveys for 6 times to suppress it since the first time on 04/01/2000. For this purpose, to analyze the impacts of the policy of regular drug price surveys on prescribed oral medication, single –component and single-purpose cardiovascular drug Felodipine is used for probing the drug consumption for the first 5 price adjustments, in order to provide NHI Taiwan the reference and the revision of cost controls.

Methods
The study is of a retrospective longitudinal basis which period of inferential statistics is from 04/01/2000 to 12/31/2008. Utilizing national health insurance research database, OPD & Hospitalization being at the ratio of 500:1 & 20:1 respectively from the randomly sampled secondary data files, yielded from National Health Research Institutes constitutes Taiwan’s medical consumption database of drugs. Prescribed drugs in accordance with ingredient name “Felodipine” are in the grand total of 12, and through proper exclusion and reformation of the database afterwards, 75970 data-sets meet the standards for analyses eventually. SPSS for Windows is used for statistical analyses, which involves Independent-Samples T-test, One-Way ANOVA, Linear Regression, Logistic Regression, and Generalized Estimating Equations for data analyses and verification of hypotheses.

Results
Generalized estimating equations are particularly utilized to deal with time effect through repeated measures of regular drug price surveys for 6 times as well as to examine drug consumption in number and amount during a specific period of time, and by using the results of the repeated measures to precisely compute the model outcomes for each predictor.

Research findings: By control of other predictors in separate models of distinct dependent variables— (1) only the drug doses in average number per physician order (p=0.521), and rate cuts of drug prices in average number (p=0.433) and average amount (p=0.906) show no statistical significance respectively, other predictors of each model all revealing significant differences both in drug consumption of number and amount. (2) only the drug doses (p=0.983) and hospital ownership (p=0.789) prove no statistical significance towards decision making of physician order on brand name or generic drugs, while the other predictors do influence the behavior how a doctor prescribes, presenting significant differences.

1. Average number affected by predictors:
(1) Result of average number affected by hospital ownership shows clinics > public hospitals > profit hospitals > private hospitals, which means clinics consume the most Felodipine drug in average number.
(2) Based on hospital contract type, the consequence of average number presents medical centers > regional hospitals > area hospitals > clinics; in other words, medical center dominates over the others in Felodipine drug consumption of average number.
2. Average amount affected by predictors:
(1) Result of average amount affected by hospital ownership shows clinics > public hospitals > profit hospitals > private hospitals.
(2) Based on hospital contract type, the consequence of average number presents medical centers > regional hospitals > area hospitals > clinics, to clinics however, regional hospitals reveal no significant differences (p=0.55).

3. Average number per physician order affected by predictors:
(1) Result of average number per physician order affected by hospital ownership shows clinics > profit hospitals > public hospitals > private hospitals, but there is no significant differences (p=0.209) of private versus public hospitals.
(2) Based on hospital contract type, the consequence of average number per physician order presents clinics > medical centers > regional hospitals > area hospitals, to medical centers however, regional hospitals prove no significant difference (p=0.572).

4. Average amount per physician order affected by predictors:
(1) Average amount per physician order of brand name drugs is much higher than that of generic drugs; if generic drugs are used, that would be saving 70 NTD compared to the use of brand name drugs.
(2) Based on hospital ownership, the consequence of average amount per physician order indicates clinics > profit hospitals > public hospitals > private hospitals, to public hospitals however, private hospitals show no significant differences (p=0.082).
(3) Result of average amount per physician order affected by hospital contract type indicates clinics > medical centers > regional hospitals > area hospitals, but there is no significant differences (p=0.584) of clinics versus medical centers.
(4) Although higher Felodipine average number per physician order is presented in late periods, but average amount per physician order is much relatively less.
(5) Rate cuts of drug prices to average amount per physician order indicate a significantly negative correlation, in other words, average amount per physician order less 1.63 NTD while drug prices cut 1%.

5. Predictive model of prescribing bias on brand name or generic drugs:
(1) Drug doses do not influence the prescribing bias on brand name or generic drugs.
(2) Based on hospital contract type, the outcome of prescribing bias on generic drugs points out clinics > regional hospitals > area hospitals > medical centers, to clinics however, regional hospitals signify no significant differences (p=0.156).
(3) According to intervals of drug price adjustment, we conclude that the later the interval is, the higher probability the generic drugs will be chosen.
(4) After the consideration of time effect and the control of other variables in the model, probability of prescribing bias on generic drugs will be 0.06-fold reduction while drug prices cut 1%, in another word, a decrease of 6% (p=0.021).

Conclusions and suggestions
A word spreads among people in pharmaceutical industry: “Price is the most precious for drug, just like its life!”. NHI Taiwan utilizes powerful routine drug price surveys to restrain drug expenses from rising up all the time. But excessive cuts of the drug price will result in benefit instead specially for new and costly drugs on patent sometimes, and cause the negative feedback of drug expenses to grow conceivably. It is like the both ends of a balance, and weighing “price” and “quantity” is in need of meticulous deliberation by NHI Taiwan.

In view of the mentioned above, particularly proposing two suggestions for policy makers as the following:

1. To consider the recommendation by “Council For Economic Planning And Development” that a standard of reasonable drug expenses accounts for 24.3% of medical expenditure, and only to adjust drug prices while drug expenses exceed the goal.
2. To consider the recommendation by “Department of Health, Executive Yuan, R.O.C.” that drug and medical costs should have the same degree of growth rate, and only to adjust drug prices while drug expenses exceed the goal.

並列關鍵字

hospital ownership ； hospital contract type ； brand name drugs ； generic drugs ； drug price survey

參考文獻

中文文獻

中央健康保險局 (2004)。中華民國2004年全民健康保險統計。

中央健康保險局 (2005)。全民健康保險住院診斷關聯群支付方案規劃報告2005年。

中央健康保險局 (2005)。全民健康保險健保用藥品項壓縮檔2005年。

中央健康保險局 (2005)。疾病分類代碼及範圍，ICD_9_CM疾病碼一覽表2005年新增。