Purpose: The medications taken by a typical patient, particularly elders, are generally recorded in a number of distinct medication lists corresponding to different facilities, i.e., a clinic, pharmacy and a residential facility. Ideally, these lists would be identical, but in practice they may be significantly discordant, and must be reconciled. We have developed techniques for assessing these discordances and are building tools to aid the work-intensive reconciliation process. The resulting metrics take into account the probability of the correspondence of the prescriptions across several lists, the clinical treatment implications of the discordance (hazard) and the probability that the discordance would lead to harmful outcomes or events. Methods: The theoretical discordance metrics were defined in terms of the “computational” complexity of the transformations in the reconciliation process and in terms of the expected clinical treatment effects. Medication lists of 23 randomly selected elder patients residing in a residential facility were collected from two, and whenever possible from all three, of the organizations. These lists were transcribed, reordered and then scored with respect to discordance. Each item was described in terms of the drug identifier, dose, timing, and route. For each patient, the residential facility list was selected as the reference list. After finding the best match for each item, the corresponding items were classified as identical, equivalent, missing or not the same for each of the attributes. Each “error” discordance, i.e., not identical or equivalent, was judged with respect to the severity (hazard) and the likelihood of the adverse outcome. Results: The medication lists averaged 14.3 items per list representing the average number of prescriptions per patient. The average proportion of errors between the pharmacy and the residential facility was 17%, but between the clinic and the residential facility the proportion of errors reached 37%. Despite the large magnitude of these errors, the expected hazard was much smaller, because only approximately 7% of the 37% of errors were likely to have serious consequences. Conclusions: The overall results suggest that the multiattribute classification of discordances and evaluation based on expected utility provide an effective approach to the evaluation of medication list discordances. This method offers the potential for identifying and mitigating possible adverse events at a much earlier stage.