Medication reconciliation errors on hospital admission can lead to significant patient harm. A pediatric intermediate care unit initiated a quality improvement project and aimed to reduce errors in admission medication reconciliation by 50% in 12 months.
From August 2017 to December 2018, a multidisciplinary team conducted a quality improvement project with plan-do-study-act methodology. Continuous data collection was achieved by reviewing medications with home caregivers within 18 hours of admission to identify errors. Cycle 1 consisted of nursing training in accurate and thorough medication history documentation. Cycle 2 was aimed at improving data collection. Cycle 3 was aimed at improving pediatric housestaff processes for medication reconciliation. In cycle 4 intervention, the reconciliation process was redesigned to incorporate the bedside nurse reviewing final medication orders with the patient’s home caregivers once the medication reconciliation process was complete. Intermittent maintenance data collection continued for 12 months thereafter.
Cycle 1 and 2 interventions resulted in improvement in the medication reconciliation error rate from 9.8% to 4.7%. In cycle 2, the data collection rate improved from 61% to 80% of admissions sustained. Cycle 3 resulted in a further reduction in the medication error rate to 2.9%, which was sustained in cycle 4 and over the 12-month maintenance period. A patient’s number of home medications did not correlate with the error rate.
Reductions in admission medication reconciliation errors can be achieved with staff education on medication history and process for medication reconciliation and with process redesign that incorporates active medication order review as a closed-loop communication with home caregivers.
Despite The Joint Commission’s designation of inpatient medication reconciliation as a national patient safety goal since 2005, medication reconciliation continues to be a challenging contribution to adverse drug events, especially in pediatric and critical care settings.1,2 Medication reconciliation is a process of “comparing a patient’s medication orders to all of the medications that the patient has been taking.”3 Medication discrepancies occur in up to 70% of patients, with approximately one-third of these potentially causing harm.4 Reduction in medication errors has recently been identified as a priority in pediatric safety research.5
In a previous prospective study, pharmacy gold standard preadmission medication history was compared with the patient’s admission and discharge medication orders, and the results revealed an error rate of 15% of medications reconciled during admissions to our hospital’s intermediate care unit (IMCU) and medical ICU over a 9-month period.2 The sources of error were split between errors in admission history (45%) and errors in reconciling the medication history with admission orders (55%).
Having identified this 15% medication reconciliation error rate, we implemented a quality improvement (QI) project with the aim of reducing admission medication reconciliation errors among patients admitted to the IMCU. Our primary aim was to reduce errors in medication reconciliation on admission to the IMCU by 50% over a 12-month period, starting August 2017. We sought to design interventions that required neither changes in the electronic medical record (EMR) or ordering system nor hiring of additional personnel.
The medication reconciliation QI project was implemented in a 12-bed IMCU in a pediatric tertiary care hospital. The unit cares for patients with high acuity and patients with chronic care needs requiring intensive nursing at baseline; however, it does not provide invasive ventilation, vasopressors, or other invasive therapies usually reserved for ICUs. The unit is staffed with a nurse/patient ratio of 1:2, an on-site attending physician, 4 pediatric housestaff, 2 nurse practitioners, and a respiratory therapist. The IMCU averaged 56 admissions per month over the study period, with an average daily census of 11 patients. There were no policy changes affecting the types of patients admitted over the course of the project.
The baseline IMCU medication reconciliation process consisted of parallel processes of medication history completed by the physician and nurse (Fig 1A). The bedside nurse obtained a medication history and documented it in the EMR. Medications were subsequently reconciled by the physician during admission order entry and then verified by pharmacy staff using the nursing-obtained EMR medication history as the verification source. The bedside nurse electronically reviewed final orders but did not perform another formal reconciliation. Of note, in April 2018, the hospital’s emergency medicine department also sought to improve the accuracy of medication histories for patients treated in the emergency department (ED). The location of the initial medication history documentation had no impact on the medication reconciliation error rate in subsequent analysis.
Our primary outcome measure was admission medication reconciliation error rate, defined as the number of home medications with at least 1 error identified by using a data collection tool (Supplemental Fig 6) divided by the number of home medications of patients for whom the data collection tool was completed. Errors were defined as unintentional discrepancies between home medications, as reported by the home caregivers, and medications ordered on IMCU admission. We chose to count the errors by the number of medications that had an error rather than by patient to reflect actual opportunities for error.
IMCU staff members, including 2 nurse practitioners (coauthors E.J. and D.C.), nurse leadership (coauthors S.S. and J.T.), and 5 daytime charge nurses were trained in standardized data collection. Four night charge nurses were trained later, but minimal data were collected overnight. Data collectors reviewed medication orders with home caregivers within 18 hours of IMCU admission and after the medication reconciliation process was complete. During the medication order review, data collectors used the data collection tool to document any identified errors. Identified discrepancies were reviewed with the prescriber to determine if discrepancies were intentional and to correct any errors. Medications that are meant to be given as needed, except for inhalers, nitroglycerin, opiates, antiepileptics, benzodiazepines, and sedatives, were excluded. We aimed to collect data on 90% of all patients admitted to the IMCU from the ED, home, or an outside facility via continuous data collection. Patients who were transferred into the IMCU from another hospital unit were excluded because their admission medication reconciliation had been completed on the previous unit.
In most instances, “home caregivers” referred to a parent or family member; however, in some cases, the patient’s referring residential care facility was the primary source of information. We chose to use parental confirmation of the home medications on the basis of evidence that parents are the most accurate source of the patient’s current medication list.6 Data collectors sought to perform the review in person with the use of telephone communication when face-to-face reconciliation was not feasible within 18 hours of admission.
The collected data elements included the name of the medication, type of error (classified as omission, dosing, formulation, frequency, timing, and/or route), date and time of review, and number of home medications. As a process measure, the data capture rate was defined as the number of patients admitted for whom medication reconciliation data were collected divided by the number of IMCU admissions each month. As a balancing measure, the time required to review home medications with the caregiver was also tracked. All data were recorded in a Health Insurance Portability and Accountability Act–compliant, password-protected database. The improvement team members performed independent data verification to ensure quality assurance.
The improvement team met monthly to review errors and data trends. Errors were verified by the team to ensure that they did not refer to excluded pro re nata medications, and the team reviewed the medical record or reached out to housestaff to ensure that errors did not reflect intentional discrepancies. Using this review process, we did sometimes identify discrepancies that were not errors, and feedback was provided to the data collector.
Shewhart control charts were used to evaluate process variability and specifically to capture error rates from the baseline and intervention periods with data points for each week.7 P-charts were created by using the SQCpack version 7 software. Given the known high error rate, the team gathered only 5 weeks of baseline data before cycle 1 interventions. The initial mean was calculated on the basis of the first 12 data points.7 Centerlines were recomputed on the basis of the presence of a special cause, when 8 consecutive data points fell above or below the mean, or when 2 of 3 points were above 2 SDs or 4 of 5 points were above 1 SD. The control limits varied on the basis of sample size for each period, set at 3 SDs above and below the mean. The improvement team met regularly to review data and explore special cause variation. Decisions were made at those meetings regarding next plan-do-study-act cycle interventions on the basis of group discussion, special cause analysis, and quantitative methods to draw inferences from data.
The cumulative data on the types of medication reconciliation errors and the error rate compared with the number of home medications were analyzed by using Cochran-Armitage trend tests in Stata 14 (Stata Corp, College Station, TX).
Four plan-do-study-act cycles were implemented throughout this project. The first 3 cycles were focused on education and support for staff to correctly conduct the baseline admission medication reconciliation process as designed and to improve the process measure of the data capture rate. In the fourth cycle, we incorporated a final review of orders with home caregivers by the bedside nurse into a medication reconciliation process.
Given concerns about inadequate documentation of medication histories, cycle 1 intervention consisted of 1-on-1 and small-group nursing training in accurate and thorough medication history documentation by using a reference tool in the EMR for all 42 members of the ICMU nursing staff. The 2-page reference tool outlined the process for eliciting a medication history from the home caregiver and then documenting medication details, including dose, route, frequency, and home administration times (September 2017 to November 2017).
Cycle 2 interventions were focused on improving the capture rate through active tracking of admissions and designation of project interventions as a shift task on the charge nurse handoff form. Project data were shared with unit nurses and housestaff by project leads to raise awareness (October 2017 to April 2018).
On recognition of variation in practice among housestaff in performing medication reconciliation, cycle 3 intervention included targeted education to rotating housestaff to reference the increasingly accurate medication histories for admission order entry while sustaining regular dissemination of project data (May 2018 to July 2018).
Toward the end of our 12-month time frame, the team identified that the data collection process itself, by providing a secondary review at the completion of order entry, served to identify medication reconciliation errors early in the hospital course. Cycle 4 intervention thus consisted of redesigning the reconciliation process to incorporate closed-loop communication by having the bedside nurse and home caregiver review medication orders at the end of the reconciliation process. All 42 unit bedside nurses were trained to complete a review of the final medication orders populated in the EMR. Details reviewed included dose, route, frequency, and home administration times, and any medications missed during initial medication history were queried for. This review was to be completed as soon as feasible pending the ability to communicate with the home caregiver after the medication administration record in the EMR was populated. One-on-one and small-group nursing education was provided by the nurse educator, who explained and demonstrated this additional step. Bedside nurses did not collect data during their augmented reconciliation process, but data collection continued as before with a separate review of medications with families after the completed medication reconciliation process (August 2018 to October 2018) (Fig 1B).
We continued data collection for all IMCU admissions for an additional 3 months to assess for sustained change. In January 2019, we transitioned to a 12-month maintenance phase to audit ongoing error rate using the same methodology, with the data collection tool only used during the first full week of each month. Adherence to the medication reconciliation process was reinforced by noting which newly admitted patients needed a bedside nursing medication review in morning nursing safety huddles and by using monthly emails to announce auditing weeks.
This project met our institutional standards for QI: testing change in a local process by using interventions within standard of care, no patient randomization or risk of harm beyond a minimal loss of privacy related to data collection and analysis, no external funding, and using QI methods with data collection only as needed to meet the goal of local improvement. Thus, the project did not require institutional review board review or informed consent.
The admission medication reconciliation error rate decreased from an initial mean of 9.8% on the basis of the first 12 data points (11.6% in a 5-week baseline period) to 2.9% at the completion of the 16-month project (Fig 2). Nursing education in cycle 1 resulted in a reduction in the error rate, but special cause variation was first achieved in cycle 2 with further efforts to improve data collection, decreasing the error rate from 9.8% to 4.7%. From June 2018 to July 2018, there were 2 data points outside the upper control limit, with a high number of errors on a few admissions performed by new interns. In cycle 3, the team provided focused training for the housestaff on the medication reconciliation process and was able to achieve special cause and reduce the mean error rate to 2.9%.
Implementation of cycle 4 interventions, with a redesign of the process to include home caregivers in the final review of medication orders, allowed for sustained improvement in the medication reconciliation error rate without special cause variation throughout the 12-month maintenance auditing period. One outlier week had an error rate of 12.5% without a clearly identifiable etiology. Errors persisted because of occasional discrepancies in caregiver-provided home medication lists or nonadherence with the reconciliation process, with etiologies such as onboarding of new staff and completion of data collection before the bedside nurse review with the family.
Medication reconciliation error data were collected for 746 of 921 (81%) admissions during the 16-month project period. A 5-week preintervention period revealed an average capture rate of 61%. This rate rose after implementation of cycle 2 to an average of 80% and remained in control through completion of the project (Fig 3). The weekly number of medications reviewed did not vary among project cycles (P = .142; Kruskal-Wallis test).
Patients for whom data were not collected had shorter unit stays (median [interquartile range]: 24 [17–48] hours versus 38 [22–65] hours [P < .001; Kruskal-Wallis test]) and were less likely to prefer the English language (80% vs 86%; P = .023; χ2 test) than patients with data collected; age and sex did not vary significantly between the 2 groups. Data collectors reported that the medication review process with home caregivers required an average of 6 minutes (median: 5 minutes) to complete.
The analysis of cumulative errors identified during the study period revealed that medication timing was the most common error type overall, contributing to 37% of admission medication errors, followed by medication omission, which contributed to 24% of errors (Fig 4). Timing errors were eliminated after cycle 4. Patients with only 1 to 4 home medications had error rates that were not dissimilar from those of patients with ≥20 medications (5.4% and 5.0%, respectively) (Fig 5). Tests for trend revealed neither a significant difference nor a trend of the error rate by the number of home medications.
We were successful in reducing the admission medication reconciliation error rate by >50%, from 9.8% to 2.9%, without changing the EMR or hiring additional personnel. Although educational efforts to improve facility with the EMR and adherence to the initial process did contribute to the decline in the error rate, we sustained the reduction in errors with a redesign of the system to incorporate a closed-loop communication, with the nurse and home caregiver jointly reviewing the final medication orders.
Our baseline medication error rate was similar to that of other studies, including 1 study of pediatric complex care populations in which errors we identified in 21% of admission medications and general pediatric inpatient medication error rates ranged from 5.2% to 13.2% of medication orders.6,8 Interestingly, in our study, we found similar error rates regardless of the number of home medications, which differs from the results of a previous study conducted on our unit.2
We did not reach our aim of a 90% capture rate, but we accept the average 81% capture rate as adequate to support the reduction in the error rate. The project took longer than anticipated, with a 4-month extension to implement cycle 4 and then 12 months of auditing to assess for sustained error rate reduction.
A closed-loop control system in systems engineering is “an automatic control system in which an operation, process, or mechanism is regulated by feedback.”9 Closed-loop communication is a pragmatic feedback loop that “consists of the team’s ability to exchange clear, concise information, to acknowledge receipt of that information, and to confirm its correct understanding.”10 Closed-loop communication was used in the nurse review of medications populated in the medication administration record with the family or caregiver to ensure all home medications were ordered correctly. This step allowed for potential errors to be identified quickly and streamlined tasks, such as retiming of medications, to more closely match home routines. We found that home caregivers were open to this additional review, particularly when it was presented as a safety initiative.
We conducted this QI initiative without changes to our EMR, but our team noted that EMR innovations to improve visualizing active orders adjacent to detailed medication histories would facilitate easier and more accurate medication reconciliation. Such EMR changes have been shown to reduce discrepancies, with 1 study revealing a one-third reduction in potential adverse drug events.4,11 However, the accuracy of electronic preadmission medication lists remains problematic. Patients with complex medical needs may have a large number of medications with intricate dosing regimens that remain dynamic depending on clinical status (eg, antiepileptic medications).
We were able to reduce the error rate without hiring additional personnel, but we acknowledge that there is evidence supporting the designation of pharmacists for obtaining medication histories early in a patient’s course to decrease the error rate.4,6,12–14 In 1 study, medication reconciliation by pharmacists in an ICU reduced medication errors on admission from 45% to 15%.15 At another institution, the addition of a transition of care team comprising registered nurses reduced admission errors from 33.9% to 18.7% and reduced errors on discharge and follow-up visits as well.
Among patients with complex medical care needs, parents provide the most accurate medication history (75% sensitive, 96% specific), even when less available (52%) than other information sources, such as the home pharmacy, last admission electronic health record, or medical admissions history.6 We therefore considered parents and/or home caregivers as a gold standard in identifying errors; however, we acknowledge that we had instances in which home caregivers provided inaccurate medication lists. Although review with home caregivers improved our admission medication error rate substantially, we note that further interventions, such as EMR changes or pharmacy involvement, remain necessary to reach a true gold standard of an accurate medication list.
Although we believe incorporation of feedback loops into our process is generalizable, we recognize that some of our constraints are specific to our institution, particularly the design of the medication history and reconciliation process in our EMR. We also recognize that our nurse/patient staffing ratio of 1:2 may allow our nursing staff more time to review medications for patients with complex medical needs. However, our data collectors reported that the final closed-loop review process took ∼5 minutes and likely saved time and resources by avoiding errors.
We limited our intervention to admission medication reconciliation but recognize that medication errors can occur later in the admission, and additional processes are required to reduce medication errors throughout a hospital course, particularly on transfer and discharge.
Reduction of admission medication reconciliation errors is achievable with staff education and a redesign of the process to include closed-loop communication with home caregivers and results in sustained reduction in the admission medication reconciliation error rate. This concept can be incorporated in other institutions for higher reliability in accurate medication reconciliation.
We thank Michael Agus, Jenny Chan Yuen, Taruna Banerjee, Kate Donovan, and Nadine Straka for their support of this project and the nursing staff of the Intermediate Care Program for their exceptional efforts to reduce medication reconciliation errors.
Dr Russ, Ms Stone, Ms Treseler, and Dr Kelly conceptualized and designed the study, contributed to the design of data collection instruments, coordinated data collection and analysis, and drafted and revised the manuscript; Ms Vincuilla and Ms Partin designed the data collection instruments, assisted with data collection, conducted the initial analysis, developed tables and figures, and reviewed and revised the manuscript; Ms Jones, Dr Chu, and Ms Currier contributed to data collection and data analysis, assisted in drafting the manuscript, and reviewed and revised the manuscript; and all authors approved the final manuscript as submitted.
FUNDING: No external funding.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.