Usually Available Clinical and Laboratory Data Are Insufficient for a Valid Medication Review: A Crossover Study.

Objectives: To establish the quality of medication reviews performed by nursing home physicians, general practitioners and pharmacists.

Design And Setting: 15 Pharmacists, 13 general practitioners and 18 nursing home physicians performed a medication review for three cases (A, B and C), at three evaluation moments. First, they received the medication list. Secondly, they also received laboratory results and reason for admission and finally, we added medical history. Remarks were divided into 6 categories, i.e. indication without medication, medication without indication, contraindications/ interactions, dosage problems, double medication and wrong medication. Remarks were compared to the remarks made by our expert panel and scored according to our grading model as appropriate (0 to +3) or missed or potentially harmful (-1). For each medication error category, the percentage of participants who made this error was computed.

Results: After the first evaluation moment, the overall estimated mean percentage score was -1.7% for case A, 3.9% for case B, and 8.7% for case C. After the second review, this score was 15.0% for case A, 19.8% for case B, and 22.2% for case C. This further increased to 30.0% for case A, 36.7% for case B and 44% for case C at the final evaluation. The absence of medication where there was an indication (indication without medication) was frequently missed and did not improve after adding the extra information regarding laboratory results, reason for admission and finally medical history.

Conclusion: Increasing clinical information helps physicians and pharmacists to improve their medication reviews, however, additional information was still related with a high margin of error. Detection of certain errors becomes easier with additional information, whereas other errors remain undetected. To achieve a high standard of medication review, we have to change the way medication reviews should be performed.