This study revealed a high number of discrepancies (84.3%) between the gold standard Best Possible Medication History and medication profiles found in the PharmaNet provincial medication database. Several of these discrepancies were expected, based on the scope of PharmaNet (e.g. absent over the counter, HIV, and hospital dispensed medications); however, many were not. The most common error was missing medications, followed by medications that did not appear as active (or “Current”) on the PharmaNet profile although they were still being used. Almost half of profiles contained discrepancies that were determined to be potentially clinically significant if not corrected (318/660). That is, if the PharmaNet profile was considered accurate and was solely used to direct prescribing, there was a significant risk for medication errors that could lead to ADEs. There are implications based on the findings in this study that relate to use of PharmaNet for clinicians, researchers and the PharmaNet program.
PharmaNet should be used by clinicians in conjunction with other clinical information (such as a patient history), clinical judgment, and follow up to prevent potential ADEs. The risk of significant error increases with the number of medications a patient is taking. Clinicians should focus attention on the common medication classes listed in Table due to the frequency with which they are discrepant or absent from the PharmaNet profile. Common and accurately recorded medications (~90% accuracy in PharmaNet) tended to be cardiovascular drugs (with the exclusion of ASA, 15% accuracy), thyroid medication, and citalopram (Table ).
Top suggestions for clinicians when using PharmaNet in conjunction with a medication history
For researchers using PharmaNet and potentially other, similar medication repository data, they should be aware of these limitations and not assume that PharmaNet accurately represents medication usage in a given population. Studies have assessed expected completeness of drug profiles, for example in Ontario [27
]. However, researchers should also assess accuracy of instructions (if relevant) as these were also frequently inaccurate. This study highlights that many patients are not taking the medications as they were dispensed. This study has shown several medications are often absent from the profile (e.g. over the counter medications), and many are used differently than was recorded at the time of dispensing (e.g. dose and frequency discrepancies in warfarin, insulin, opiate agonists). When employing PharmaNet data to study medication usage, it would be important to consider validating in PharmaNet the accuracy of medications of interest.
For the PharmaNet program, which is undergoing enhancement [28
], this study suggests that there may be value in allowing trained clinicians (e.g. physicians, pharmacists, nurse practitioners) to update patient profiles in order to correct medication data when medication changes are made or discrepancies are found (e.g. while performing a BPMH). A current or regular medication list is not available in the current PharmaNet design, but could be considered for future updates. As many of the discrepancies were medications that appeared expired that were in fact not expired, the concept of clinicians
flagging medications as “current” or “ongoing” may be clinically relevant. Enhancements such as these would increase the reliability of PharmaNet and thus improve the efficiency of medication reconciliation during a BPMH.
Findings here were similar to a smaller study performed in Saskatchewan, where 39/50 (78%) of patients had at least one discrepancy on their profile [20
]. Also, a study of 493 patients at Veteran Affairs found that their integrated primary care medication record was accurate in only 5.3% of patients [22
], compared to a medication history. This suggests that the findings are not unique to BC and PharmaNet and other regional and jurisdictional repositories may have similar accuracy rates.
This study has added to our knowledge of accuracy of medication repositories by quantifying discrepancies in a large, prospective study of patients whom are typically at risk for medication errors. It has been able to describe, by medication class, some of the more common discrepancies as well as those medications that are accurately described in PharmaNet. Other jurisdictions have or are developing regionalized or national medication records such as in Australia, the US [29
] and the NHS. Where these may rely on dispensing or prescribing information to populate their repositories, similar accuracy issues may exist.
There are several limitations to this study. First, this was not a randomized study of patients in BC; thus, the findings in this study may not be applicable to the entire population of the province. This patient population is likely more complex and on more medications than average. This study was completed within in BC with a single medication repository (PharmaNet). Medication profiles and discrepancy rates may be different in other parts of the world and where other features are incorporated into the tools (e.g. reconciliation). However, other studies suggest that there are significant error rates in other repositories [22
]. We did not perform chart reviews when estimating the potential impact of the observed discrepancies. Having a greater understanding of the patients’ medical history would have allowed the clinical study panel to be more precise in their estimations of severity of potential ADEs. This would not have allowed for the anonymous data collection and would have limited the number of cases that were reviewed. It is not typical that providers would rely solely on the PharmaNet profile for prescribing without follow up, and patients would likely be aware if medications missing or their symptoms were not being managed. Thus, the construct for the Panel assessment, while useful in this study, is somewhat theoretical.
This study was designed as a baseline study to assess the current version of PharmaNet. PharmaNet is being updated, with new features to be deployed in the coming years. A repeat of this study once an enhanced PharmaNet is deployed and stabilized would provide evidence to the impact of changes in the design of PharmaNet. Comparative studies could be completed in other medication repositories. Further studies could assess the rates of medication errors and potential ADEs, comparing patients who receive BPMH on admission/discharge to hospital and those who do not (i.e. those who rely more on the PharmaNet profile). Controlled studies should be performed that look at the addition and use of specific features that support medication reconciliation. These could discover what functions are needed (both social and technical) to improve accuracy of the repositories.