在藥物不良反應中，要偵測的是少見的藥物及與其相對應的不良反應。在尋找時，藥物與症狀是分析中的基本單位。有多種不同的方法被用來尋找這種少見的症狀，統計的方法中勝算比是其中之一，它可告訴我們資料庫中有可能造成危險的不良反應。在藥品調配錯誤被發現及被通報的過程中，與藥物不良反應的通報有多處類似，主要的相似處為尋找藥物不良反應中使用「藥品－症狀」的配對去尋找有潛在危險的不良反應，而我們使用於尋找有可能的調配錯誤配對。因此研究的目的在借用這種方法將配方中的錯誤「應配藥品－所配藥品」運用於評估藥品調配錯誤的相對危險性，發現有高機率造成錯誤的藥品，這有助於我們預防錯誤的發生。
我們檢查了一個醫學中心自93年2月至6月的門診處方箋，以一個藥品為一個處方計算，該期間共有1,326,770筆處方。其中有960筆（0.0724%）錯誤經複檢時發現。經整理後，有730對「應配藥品-所配藥品」，在其中有620對只出現一次，66對出現2次，44對出現等於或大於3次。我們以出現等於3次或大於3次的配對計算勝算比來評估其配錯的危險性。結果顯示，最常發生配錯的藥品為具相同藥名，但單位劑量不同或是英文藥名相近似的藥品，其次為具有相同外觀的藥品，再其次為具有相關治療用途的藥品，最後為藥品儲存位置相近者。
這個研究使配方錯誤的研究對象確定為「應配藥品-所配藥品」，有效的運用該研究對象，有助於發現配方錯誤。其次，我們成功的運用藥物不良反應中偵測少見的不良反應的相對通報勝算比的方法於篩選有高機率的調配錯誤，相信該方法有助於偵測各個藥局中自己有的調配錯誤的藥品配對。
 

In detecting adverse drug reactions (ADR), signal detection is based on the appearing of a rare symptom that is associated with a particular drug. Therefore, the couple of drug and symptom is the basic unit of signal detection. Different methodologies have been developed to explore these rare units of adverse drug reaction. One of the approaches is the relative odds ratio. For the detection of dispensing error, we applied the couple of “prescribed drug-dispensed drug” as the analysis unit, as in ADR signal detection. The error signal was generated when reported by a patient or a staff. The process is quite similar to the automatic reporting system of adverse drug reaction. In the present study, we sought to apply the couple of odds ratio in quantifying the potential risk of certain types of dispensing error. 

We reviewed 1,326,770 outpatient prescriptions (one drug for one prescription) in a medical center during February to June 2004. Before handing it to patient, 960 dispension errors (0.0724%) were found. Among these errors, 730 different types of error pairs were found which included 620 pairs appeared once; 66 pairs appeared twice, and 44 pairs appeared no less than 3 times in reported data. The potential risk of dispensing error was estimated according to the odds ratio having frequency greater than 3 times. The most frequently appeared errors were (1) those drugs that sound or spell similarly; (2) those that are similar in appearance; (3) those that are therapeutically related and those that are located in close proximity to each other.
In conclusion, we successfully applied the signal detection technique of ADR to examine the medication error of dispensing. It is an easy way to estimate the potential risk of error during dispensing.