Medication Reconciliation Using Electronic Pharmaceutical Record: A Multicenter Study in the Hospitalized Elderly

Context Adverse drug events (ADEs) may occur in community or hospitalized patients. A review of the literature reports an incidence of ADEs varying from 0.1 to 61.4 %. This wide range may be explained by different designs and case-mix in the studies. For instance, the incidence of ADEs is higher in surgical wards or in the elderly, and increases with the number of prescribed drugs. In the elderly, the risk of ADE has been reported as 2-fold higher than that of a younger population. Some ADEs may not have been prevented, for example when the event was associated with an intended or known effect (e.g. neutropenia after cytotoxic drugs administration). However, some other ADEs might have been prevented whenever they were related to medications errors (omission, commission, dosing errors, prescription despite known allergies…). The proportion of preventable ADEs has been reported to range from 25% to 43% of total ADEs.

ADE can lead to extra monitoring because of biological abnormalities but also in more serious cases, to temporary or permanent harms, and even to death.

Prolonged hospitalizations are also attributed to ADE. Gastrointestinal tract events are the most common type of ADEs, followed by electrolyte and renal events, hemorrhages, metabolic and endocrine events. Drugs involved are cardiovascular drugs, followed by analgesics, antibiotics and anticoagulants.

Discrepancies between patient's current medication and the drugs prescribed at hospital admission hospital are frequent (27 to 54%) and have been reported in numerous studies. Potential consequences of such discrepancies have been extensively studied but data dealing with clinical outcomes are sparse.

Medication reconciliation is a process aiming at the creation of the most appropriate list of patient's medications, thus decreasing medication discrepancies and eventually ADE. Medication reconciliation can be performed by physicians, nurses, pharmacists. Usual sources of information for medication reconciliation are patient's prescriptions from the attending physician, patient's interview, hospital records… Additionally, information technology and electronic health records are of great help in this process. In France, drugs dispensed to a patient in community pharmacies during the 4 past months are registered in his/her electronic pharmaceutical record which is available to any pharmacist nationwide. This database is protected and can be consulted after recognition of the patient's healthcare card. To date, the contribution of this record to medication reconciliation has not been studied. The CONCIPAGE study will estimate the impact of a medication reconciliation using patient pharmaceutical record, on the occurrence of ADEs during the hospitalization of patients aged 65 years and over.

Design The CONCIPAGE study is a national, multicenter, cluster-randomized, two-period cross-over study. It will estimate the impact of a medication reconciliation using patient pharmaceutical record, on the occurrence of ADEs during the hospitalization of patients aged 65 years and over.

20 centers will participate in the study. An equilibrated randomization will be applied to the 20 centers.

The recruiting period will last for 26 months and will be divided into four 6-month consecutive periods (A, B, C, D). During the first 6-month period (period A), 5 centers (1 to 5) will apply one strategy (intervention or usual care) while 5 others (6 to 10) will apply the alternative strategy. Centers 11 to 20 will not recruit patients during period A. During the next period (period B), centers 11 to 15 and 16 to 20 will apply the intervention or the control strategy, while centers 1 to 10 will apply a 6-month wash-out period. During period C, centers 1 to 10 will apply the alternative strategy as compared to that applied during period A (cross-over design), while centers 11 to 20 will apply the 6-month wash-out period. During period D, centers 11 to 20 will apply the alternative strategy as compared to that applied during period B, while centers 1 to 10 will not recruit patients during period D.

Description of the intervention During the control period, no intervention will be held. The prior to admission medication list will be made as usual by the hospital physician: interview of the patient and relatives, prescriptions brought by the patient, call to community physicians and pharmacists, hospital medical records, etc. This process will result in a medication prescription list at admission.

During the intervention period, a local hospital pharmacist will independently perform an additional (i.e. to that made by the hospital physician) prior to admission medication list, using the above-mentioned sources and in addition, with the help of the electronic pharmaceutical record. The medical reconciliation in the intervention period consists in the comparison of the pharmacist's and physician's prior to admission medication lists. Such a reconciliation might lead to a modification of the medication prescription that was made by the physician at admission.

Patient follow-up Patients will be included the day of admission and will be followed-up up to 30 days. A study pharmacist will be enrolled and will be in charge of collecting follow-up data for all patients. The study pharmacist will visit centers 2, 4, and 7 months after the beginning of each period. He/she will study the record of each patient included in the study, and will collect all potential biological or clinical abnormalities, prescriptions of antidotes and diagnostic imaging, transfer to ICU, death.

A second study pharmacist, enrolled by the research coordinator, will be responsible for collecting AEs on a sample of files. This second pharmacist (hereinafter referred to as "second pharmacist") will also be independent of the centers and blind to the strategy implemented. The second pharmacist will never be in contact with the study pharmacist and will analyze the files after the latter, completely independently. In particular, the study pharmacist will not know which records will be reviewed by the second pharmacist.

The files seen by the second pharmacist will be drawn at random a priori before the first visit of the study pharmacist in the first center: the corresponding visit (2, 4 or 7 months) will be drawn at random, then within each of the visits, the files will be randomized by random permutation. The numbering of the files will correspond to the order of the inclusions.

The second pharmacist will collect the AEs in the first 2 files of the randomization list thus generated.

The agreement between the judgment of the two pharmacists concerning the 80 files seen independently in duplicate (examination of each file by a given pharmacist will result in a binary judgment: patient presenting at least one AE or patient not presenting an AE) will be reported. , and a corresponding index (eg Cohen's Kappa) will thus make it possible to document the reproducibility of the collection of the primary endpoint of the study.

Objectives The primary objective of the study is to assess the impact of medication reconciliation by a pharmacist using usual sources plus the electronic pharmaceutical record on ADE occurring in hospital settings.

The secondary objectives are the following:

• To assess severity and preventability of ADE in both groups.
• To assess the distribution of ADE in both groups.
• To compare the 30-day mortality rates observed in the two groups
• In the intervention group only, to assess the rate of intentional (treatment modification intended by the physician) and unintentional discrepancies between the prior to admission medication list made by the pharmacist and the hospital physician's prescription at admission.
• In the intervention group only, to assess the rate of discrepancies between the hospital prescription at admission and the prescription after medical reconciliation
• To estimate the cost-effectiveness of the intervention. Analysis set. The primary analysis will be conducted according to the intent-to treat principle (ITT), at the individual level, considering every patient included irrespective of the patient's follow-up and will take into account the clustering nature of data.

Sample size The calculation of number of patients takes into account the cluster cross-over design and results in 44 patients per center and per period. This figure will be rounded to 50 patients per center and per period (to integrate lost to follow-up patients). We therefore intend to include 2000 patients in the study.

Once the required 50 inclusions will be reached in a given center during a given period, the inclusions will be stopped in this center until the end of this period.

Methods for main endpoint analysis. The proportion of patients with at least one ADE observed in the intervention group will be compared to that observed in the control group. A hierarchical logistic regression model will be developed for making this comparison, with the period, the arm, and the cluster, handled as fixed effects, and the order of strategies, and individual residual error handled as random effects. To assess the efficacy of the intervention we will test if the slope of regression line is statistically different from 0.

Methods for secondary endpoint analyses. The analysis of secondary endpoints will be performed using the same methodological framework taking into account the clustering and the period effect but with linear models (for continuous variables) or Poisson (for rates), depending on the type of outcome analyzed.

Any comparison of baseline variables (considering the cluster structure of the study) will be based on Mantel-Haenszel or Wilcoxon signed rank test.

Methods for medico-economic analysis The analysis will be performed according to the payer's perspective with a horizon time of 30 days, and will result in estimating the incremental cost effectiveness of the intervention, expressed as a cost per ADE avoided: [hospital costs (intervention) - hospital costs (control)] / [number of ADEs (intervention) - number of ADEs (control)].