Aim
The aim of this study was to determine the effect of a medication reconciliation program performed by pharmacists on the proportion of patients with MTEs both at ICU admission and ICU discharge. In addition, the severity of potential harm of these MTEs was measured, based on a potential adverse drug event score (pADE = 0; 0.01; 0.1; 0.4; 0.6). Furthermore, a cost–benefit analysis was performed.
Study design
The TIM (Transfer ICU and Medication reconciliation) study was a prospective 8-month intervention study with a before and after design in two Dutch hospitals. The pre-intervention phase consisted of 14 weeks of usual care [General Teaching Hospital (GTH): January–April 2013 and University Hospital (UH): February–May 2014]. After a 2-week implementation period, the intervention program with medication reconciliation by a pharmacist at both ICU admission and ICU discharge started. The post-intervention phase consisted of 14 weeks (GTH: May–September 2013, UH: July–October 2014). A detailed description of the study protocol is published elsewhere [19].
Setting and study population
The study was carried out in the Haga Teaching hospital in The Hague (GTH; 18 ICU beds) and the Erasmus University Medical Center in Rotterdam (Erasmus MC; UH; 32 ICU beds).
Patients were included when they used at least one medicine at home and when the ICU length of stay exceeded 24 h. An ICU discharge and readmission within 24 h was counted as the same ICU admission.
At discharge, patients were included if they were included in the admission part of the study and if they survived until at least 24 h after ICU discharge.
Exclusion criteria were: transfer to another hospital, both admission and discharge within the same weekend (Friday 17:00 until Monday 8:30) and patient’s inability to be counseled in Dutch or English. None of the patients of the pre-intervention group were part of the post-intervention group.
Since this study did not affect patients’ integrity, a waiver from the Zuid Holland Medical Ethics committee (METC) and the Erasmus MC METC was obtained. This waiver is in line with Dutch trial legislation. Data collection complied with privacy regulations. Figure 1 gives an overview of the study procedures.
Pre-intervention phase: usual care
Upon ICU admission, the ICU physician collected information about pre-admission medication and registered this in the patient data management system of the ICU (PDMS). The GTH ICU used Metavision (Itémedical BV, Tiel, The Netherlands) and the UH ICU used Care Suite 8.2 (PICIS Inc., Wakefield, MA, USA). The ICU discharge letter contained information about medication in use at discharge. Sometimes pre-admission medication and/or suggestions for medication use after discharge were registered.
After transfer, the physician of the admitting ward had to transcribe medication orders from the discharge letter to the hospital electronic patient records.
The intervention
After ICU admission, a best possible medication history (BPMH) was constructed, based on a medication history of 6 months from the community pharmacy, available hospital medication information and a medication verification interview with the patient and/or a representative. On the medication history of the community pharmacy, the latest date of filling was documented, as well as the date of the medication was due to be finished. Based on this list, we interviewed the patient and/or caretaker asking for all medication currently in use, the used dose, etc. By combining pharmacy record information with the patient information, we were able to get the best possible medication list. This is common practice in medication reconciliation, based on the WHO High 5 s program [21]. The BPMH included drug name, dosage, frequency and route—as well as an analysis of discrepancies between the medication used at home and prescribed at ICU admission [22]. The BPMH was documented in the PDMS and presented to the ICU physician responsible for the patient, helping him or her by explaining the effect of the medicine. We supported the physician to make the right decision on stopping or continuing. The ICU pharmacists also used the BPMH during their patient rounds at the ICU.
Shortly before ICU discharge, the ICU pharmacist made a discharge medication summary based on the BPMH and medication used prior to ICU discharge. For each medicine, the ICU physician was prompted for possible recommendations (i.e., restart, stop and continue). During reconciliation of this list with the doctor, the pharmacist helped the doctor to make the right advice for the ward. As a result, a best possible ICU medication discharge list (BPMDL-ICU) was made. This was sent as an annex of the ICU discharge letter to the physician of the receiving ward.
The medication was pre-registered by the pharmacist in the Computerized Physician Order Entry/Clinical Decision Support (CPOE/CDS) system of the general ward the patient was sent to [20]. By doing so, the ward doctor was supported by the pharmacist in the transcribing process. To prescribe the proper after ICU medication, at the right frequency, the right dose and route, the ward doctor only had to check the already pre-registered medication and, if found appropriate, simply authorize this pre-registered medication.
Outcome measures
The primary outcomes were the proportions of patients with ≥ 1 MTE 24 h after ICU admission and 24 h after ICU discharge.
An MTE at admission was defined as an unintentional discrepancy between BPMH and medication prescribed 24 h after admission to the ICU. An MTE at discharge was defined as an unintentional discrepancy between the actual medication chart of the patient and the best possible general ward medication list best possible general ward medication list 24 h after the ICU discharge (BPML-GW24). This BPML-GW24 was based on the BPMH, on information in the electronic patient records of the hospital and the PDMS, on medication prescribed in the CPOE/CDS and, whenever necessary, on interviewing the physician on the ward afterward.
Data collection was performed by trained ICU pharmacists. Whether a discrepancy was intentional or not was based on information documented in the HIS or the PDMS, information given during the medication reconciliation, the ICU standards of care and the ICU pharmacist’s interpretation of the situation. Whenever necessary, the physician on the ward was interviewed afterward. In this way, we gave the doctor the opportunity to correct the error made. Two pharmacists performed a crosscheck on the data. Subsequently, all identified MTE underwent a validity check during the pADE assessment of the MTEs (see below).
The secondary outcomes were the proportions of patients with a pADE score ≥ 0.01 due to an MTE at ICU admission and at ICU discharge. A pADE was defined as an MTE that could potentially cause harm and/or clinical deterioration and was based on the methodology described by Nesbit et al. [23, 24] using the following categories for pADE scores: 0 (zero likelihood of an ADE expected by the MTE), 0.01 (very low likelihood of an ADE), 0.1 (low likelihood of an ADE), 0.4 (medium likelihood of an ADE) or 0.6 (high likelihood of an ADE).
All MTEs at ICU admission and discharge were presented blinded and in randomized order to two assessors: one hospital pharmacist/clinical pharmacologist and one internist/clinical pharmacologist in training, both with ICU experience, who independently from each other, gave a pADE score for each MTE, based on clinical data of the patient. For MTEs that were given a different pADE severity score in the assessments, the assessors reached consensus in a meeting.
We measured a total pADE score for every patient by summing up the individual pADE scores. These pADE scores reflected potential harm per patient in the following way: pADE = 0 (no harm expected), 0.01 ≤ pADE > 0.1 (very low likelihood of an ADE), 0.1 ≤ pADE > 0.4 (low likelihood of an ADE), 0.4 ≤ pADE > 0.6 (medium likelihood of an ADE) and pADE ≥ 0.6 (high likelihood of an ADE).
Cost–benefit analysis
The cost avoidance of the TIM program was determined by subtracting the average pADE score per patient post-intervention from the average pADE score pre-intervention. This difference was multiplied by the number of patients post-intervention and the relative cost of an ADE.
The relative ADE cost price was set at € 1079. This was derived from a study by Rottenkolber and was indexed to 2014 [25].
Costs incurred by the reconciliation process were restricted to labor costs of the pharmacist. The direct time spent on this intervention was calculated using the bottom-up approach, i.e., measuring the number of minutes spent per patient by the pharmacist in a representable group of patients. These minutes were multiplied with the cost price of one minute of labor and a marginal markup percentage to account for indirect labor time (43%) [26]. The cost price of one minute was valued €1.18, based on standardized costs per minute [27].
The costs per patient were multiplied with the total number of included patients and the percentages of availability of the BPMH and the BPMDL-ICU, respectively. All costs were based on 2014 Euro cost data.
Sensitivity analysis
A one-way sensitivity analysis was performed for known variables in order to determine the effect of varying these estimates on the cost–benefit analysis.
The time spent on the intervention was varied by ± 50%. Salary costs were varied by using the highest senior hospital pharmacist scale, the lowest point on a basic pharmacist scale and the salary costs of a pharmacy technician with 7 years of experience. For ADE costs, we used the study by Bates et al. [28] as alternative to the study by Rottenkolber [25], thus varying the costs to €7177 per ADE. Finally, the ADE probability was varied by ± 50% [23, 24].
Data collection
Data were collected from the hospital electronic patient records, PDMS records, CPOE/CDS medication charts, BPMH, BPMDL-ICU and BPML-GW24. All data were collected in MS Access 2007 (version 2007, Microsoft Nederland BV, Amsterdam).
We collected the following TIM intervention characteristics: availability and quality of the BPMH and the BPMDL-ICU and the used sources (i.e., patient list, electronic patient file, medication brought from home and pharmacy medication history). The quality of the BPMH and the BPMDL-ICU was set at A, B or C [20]. Quality A was defined as a reliable reconciliation (based on a recent, reliable community pharmacy medication list and a reliable verification with patient and/or his representative), quality B as an intermediate and quality C as a sub-optimal reconciliation.
The following medication information was collected: name of medicine, dose form, medication group [30], dose and frequency; prescribed in the PDMS within 24 h after admission; prescribed in the CPOE/CDS within 24 h after the ICU discharge. All discrepancies had an intended or non-intended score, a pADE score and a discrepancy type (omission, medication added, different dose or substitution).
Data analysis
Sample size
The primary outcome of this study was the proportion of patients with ≥ 1 MTE at admission and discharge from the ICU. Based on the literature, the expected proportion of patients with MTE between wards within one hospital is 62% [4]. Based on a conservative interpretation of this study, we took a proportion of 30% in our study. With an estimated 50% reduction in errors due to the intervention, an alpha of 0.05 and a power of 0.80 calculated the sample size was 133. With an estimated mortality of 35%, in each measurement phase 205 patients should be included. We estimated extra loss of 30% due to the ICU stay less than 24 h and another 35% loss due to weekend ICU stay. Based on the number of ICU admissions per year, this resulted in a study period per intervention arm of 7 weeks for Erasmus MC and 8 weeks for Haga. To be on the safe side and to measure during a robust intervention period, we doubled the number of weeks. Therefore, a pre- and post-intervention period of 14 weeks was chosen. Based on an alpha of 0.05 and a power of 0.80, the calculated sample size was 205 patients per measurement phase for the primary outcome of this study. Based on the number of admissions per year and the potential loss due to ICU stays of less than 24 h and admission and discharge in one weekend, a pre- and post-intervention period of 14 weeks was chosen [20].
Statistical analysis
All data were analyzed with SPSS Statistics (version 24, IBM Corp., New York).
Patient characteristics pre- and post-intervention were compared using the two sample t test for continuous normally distributed variables, Mann–Whitney U test for continuous non-normally distributed variables and Chi-square test for categorical variables.
For the primary (the proportions of patients with ≥ 1 MTE at ICU admission and at ICU discharge) and secondary outcomes (the proportions of patients with a pADE score of ≥ 0.01 at ICU admission and ICU discharge), adjusted odds ratios and 95% confidence intervals (95% CI) were calculated by using a multivariate logistic regression analysis. Potential confounders were selected based on a univariate analysis (p < 0.20) and were retained in the multivariate model when they changed the beta-coefficient with more than 10%.