In our study population of mainly older patients, approximately 50% were affected by errors in the medication history at admission to hospital. The most common error was the erroneous omission of a drug from the hospital EHR medication list. Predictors of the occurrence of medication errors included an increased number of preadmission drugs and living in one's own home without community care service.
Our findings are similar to those of other studies [2
], but there are also reports of lower [15
] or higher [2
] rates of error in medication histories at hospital admission. Different definitions of medication discrepancies and errors and variability in methods of data collection could explain the differences between studies and make it difficult to compare rates of error across studies.
The association between the number of prescribed drugs at admission and the occurrence of medication errors was not surprising. Previously, some researchers have found associations between errors in the medication history and the number of drugs at admission [5
], but some have not [6
]. To the best of our knowledge, the association between absence of any care service before admission and medication errors has not been previously suggested. It is likely that the type of care service is not, in itself, important. Rather, the availability of a current medication list at hospital admission might be the important underlying factor. Patients in community care or care homes often take a current medication list with them to hospital, possibly facilitating the recording of the initial medication history by a physician or nurse and subsequently lowering the risk for medication errors. However, the absolute difference between the groups (48% vs 47% of patients with an error, as seen in Table ) was small and the value of this predictor in clinical practice would be limited. Also, the influence of this predictor is likely to vary substantially between settings and above all between countries; it will depend on the level of communication between community care services and the hospital, and on the routines for the patient taking their medication lists or medications with them when attending the emergency department.
There are varying results from other research on predictors for errors in the medication history. In accordance with our results, Gleason and colleagues found that there were few predictors associated with medication errors at admission and they suggested that well-designed processes for medication history verification were more important than patient characteristics [5
]. In contrast to our results, some researchers have found that higher age [5
] is a significant predictor. However, the patients in our study wards were older than those in previous studies [5
], and our results might have differed if the patient cohort had been younger. Previous studies have identified significant predictors for errors in the medication history which were not included in our regression model, e.g. certain "high-risk" drugs, many outpatient visits during the previous year, and staffing levels [6
The relative importance of the medication reconciliation by a pharmacist in terms of added value compared to standard care was also of interest. Optimally, this should be studied in a randomized, controlled trial. Because we were unable to carry out a randomized trial, we used an indirect measure to evaluate the degree to which standard care corrected medication errors. The pharmacists did not conduct the medication reconciliation until 4-11 days after admission for 20% of the patients due to time constraints or lack of personnel. If standard care had not identified and corrected any medication errors at all, the probability that the pharmacists identified medication errors on days 4-11 would have been as high as that on days 0-1. However, regression analysis suggested that the probability that there would still be an error in a patient's medication history after 4 or more days in the ward was lower compared to days 0-1. This implies that standard care had had partly corrected the errors in affected patients by that time, but a considerable proportion of the errors made in the initial EHR medication history at admission remained undetected by standard care. Two to three days after admission, the probability that patients would still be prescribed the wrong drug or dose was as high as the first day after admission. The potentially severe nature of some of the errors in medication history [5
] underlines the importance of reconciling the medication list soon after admission to avoid patient harm as a consequence of error, preferably within 24 hours of admission.
We believe it is necessary for medication reconciliation processes to be well designed and systematic, and aided by structured forms and detailed guidelines. Clinical pharmacists, as key members of a multidisciplinary team, are very well suited to perform such systematic medication reconciliations. A review by the British National Institute for Health and Clinical Excellence [1
] showed that there is evidence that pharmacist interventions are the most effective among the studied medication reconciliation interventions. However, it was commented that the current evidence is poor and further comparative studies of different medication reconciliation programs will be needed to reveal which approach is most effective from a clinical and economic perspective. There are many promising and emerging technologies that may be effective in medicines reconciliation as well. Nonetheless, our results highlight the benefits of structured reconciliations by pharmacists over occasional reconciliations as part of standard care.
The physicians' acceptance of the pharmacists' recommended changes to drug therapy is often used in studies of clinical pharmacy services as a measure of quality. In this study, 94% of the recommendations from the pharmacist concerning errors in medication history were accepted and implemented by the physicians, which suggests that the process was effective. In a number of cases, there was no information about the measures taken by the physicians after the pharmacists' recommendations, or about the pharmacists' reasons for not recommending changes to the medication list. More precise documentation might have provided even better insight into the effectiveness of the process.
This study adds to the evidence that LIMM-based patient care in hospital offers a positive contribution. We detected and corrected medication errors in almost half of the study patients. Although this study did not evaluate possible harm from these errors, a study including a sample of our study patients reported the clinical significance of pharmacists' recommendations [26
]. Recommendations were ranked by two physicians on a six-point scale from 1 (adverse significance) to 6 (extremely significant). Of 70 recommendations, 59% were ranked somewhat significant, 23% significant and 10% very significant. Seven percent had no significance and one recommendation was judged to have adverse significance. However, this case did not result in documented patient harm. Fifty-six of our study patients were also followed up as part of an intervention study [9
]. That study showed that LIMM-based medication reconciliation at admission and medication reviews in hospital improve the appropriateness of drug therapy and may also decrease drug-related revisits to hospital. The results of the admission process (i.e. the correction of medication errors) in the present study are therefore very likely to be at least partly responsible for these positive clinical outcomes [9
This study had several limitations. Firstly, it was conducted in an internal medicine population in a single hospital, which limits the generalisability. Secondly, acceptance of the pharmacist-acquired medication list as the most accurate preadmission drug list available could be questioned. It is possible that some medication discrepancies escaped our detection. However, studies have shown that pharmacists appear to be especially suited and more effective than physicians when obtaining medication histories [22
] and the methods used by pharmacists to obtain medication histories are well established [5
]. Our method was strengthened by the fact that the pharmacists used a number of different information sources apart from the patient interview; for example, pharmacy records are known to improve the accuracy of medication lists [27
]. The pharmacists were also well informed of the requirements and followed a strict protocol for the medication reconciliation process. Thirdly, the classification of discrepancies into medication errors partly relies on subjective judgment and is therefore subject to bias.