Search | Search by Center | Search by Source | Keywords in Title
Steinman MA, Rosenthal GE, Landefeld CS, Bertenthal D, Sen S, Kaboli PJ. Conflicts and concordance between measures of medication prescribing quality. Medical care. 2007 Jan 1; 45(1):95-9.
BACKGROUND: Several instruments commonly are used to assess the quality of medication prescribing. However, little is known about the relationship between these instruments or the concordance of their quality assessments when applied to the same group of patients. METHODS: We assessed 3 indicators of prescribing quality in a cohort of 196 veterans age 65 and older who were taking 5 or more medications. These 3 indicators assessed whether each patient was (1) taking any medication from the drugs-to-avoid criteria of Beers et al, (2) taking any medication with a score of 3 or more on the Medication Appropriateness Index (MAI), and/or (3) taking 9 or more medications (polypharmacy). Kappa statistics were used to assess agreement between measures. RESULTS: Mean age was 74.6 years, and patients used a mean of 8.1 medications. Six percent of drugs were rated inappropriate by the Beers drugs-to-avoid criteria, whereas 23% of drugs received an MAI score of 3 or more. Overall agreement between these metrics was 78%, little more than expected by chance (kappa statistic 0.14, P < 0.01). At the level of the patient, the proportion of subjects taking one or more inappropriate drugs was 37% by drugs-to-avoid criteria and 82% by MAI, whereas 37% had polypharmacy of > or = 9 drugs. Prescribing was classified as inappropriate by all 3 metrics in 18% of patients and as appropriate by all 3 metrics in 13%. Together, this level of agreement was slightly better than chance (3-way kappa statistic 0.08, P = 0.03). Agreement remained low in sensitivity analyses using different cutoffs for the Beers criteria, a range of thresholds for MAI scores, and different definitions of polypharmacy, with kappa statistics < or = 0.30 for all comparisons. CONCLUSIONS: Commonly used measures of drug prescribing quality yield widely discordant results. Because the overall quality of prescribing may not be readily inferred from a single measure, multidimensional approaches will likely be necessary for robust assessment of prescribing quality.