Article Text

Hospital admission medication reconciliation in medically complex children: an observational study
1. Bryan L Stone1,2,
2. Sabrina Boehme1,
3. Michael B Mundorff1,
4. Christopher G Maloney1,2,
5. Rajendu Srivastava1,2,3
1. 1Primary Children's Medical Center, Intermountain Healthcare, Salt Lake City, Utah, USA
2. 2Division of Inpatient Medicine, Department of Pediatrics, University of Utah, Salt Lake City, Utah, USA
3. 3Institute for Healthcare Delivery Research, Intermountain Healthcare, Salt Lake City, Utah, USA
1. Correspondence to Dr Bryan L Stone, Division of Pediatric Inpatient Medicine, Department of Pediatrics, University of Utah School of Medicine, 100 N Mario Capecchi Drive, Salt Lake City, UT 84113, USA; bryan.stone{at}hsc.utah.edu

## Abstract

Objective To evaluate admission medication reconciliation in children with medically complex conditions (MCC) by determining the availability and accuracy of five information sources and characterising admitting order errors.

Design Prospective quality improvement cohort study.

Setting Tertiary care free-standing children's hospital in the Intermountain west, USA.

Participants 23 children with MCC identified from 219 admissions between 16 December 2004 and 7 January 2005.

Intervention Medication reconciliation at hospital admission using information from five sources.

Main outcomes The accuracy of information sources was determined by sensitivity and specificity compared with verified outpatient medication lists. Errors were determined by comparing admitting orders with reconciled inpatient medication lists and categorised by frequency, type and clinical risk.

Results Children with MCC averaged 5.3 chronic medications. The reconciliation process took an average of 90 min. Availability/sensitivity/specificity respectively were parents 52%/0.75/0.96, pharmacy 61%/0.64/0.74, primary provider 43%/0.25/0.86, last admission electronic health record 87%/0.74/0.33 and admitting history 65%/0.31/0.94. Thirty-nine errors were identified in 182 admission medications (21%) including 17 omissions, affecting 13 patients (57%). The estimated clinical risk, if an adverse drug event had occurred, was serious or life-threatening in five instances.

Conclusions In children with MCC admitted at our institution during the study period, no medication information source was optimally available, sensitive and specific. Admitting order medication errors affected more than half of patients, the most common being omissions. Efforts to improve medication reconciliation at hospital admission in this population must account for availability and accuracy of information sources and medication omissions at the time of hospital admission.

View Full Text

## Background

Approximately 20% of fatal medical errors have been found to be associated with medicationrelated adverse events, a subset of all medication errors.1,2 A significant percentage of medication errors occur at transitions in care, including hospital admission.R3,5 Preventable adverse drug events (ADEs) due to medication errors occur during the ordering, interpretation, dose formulation or administration stages of medicating hospitalised patients.6,7 Admission order writing should include acquisition and review of an accurate outpatient medication list to decide which medications to continue or hold and to evaluate for interactions with new medications. This process is known as medication reconciliation, a Joint Commission (JC) patient safety goal for accreditation since 2006.8,9

### What is already known on this topic

• Errors are common in hospital admission orders, and medication reconciliation has been shown to effectively minimise these errors.

• Patients with complex chronic medical illness are typically taking multiple chronic medications and are at higher risk for medication errors at transitions in care.

• Children with complex chronic medical illness are likely to experience medication errors at the time of hospital admission, including a significant contribution from omissions.

• Parents are a reliable source of accurate medication information at hospital admission but tend to be less available in children with complex chronic illness.

Most paediatric inpatient medication safety studies have examined errors and ADEs, but not the medication-reconciliation process.R10,12 Studies suggest that risk for ADEs in children is higher than in adults. Medication safety in children is more complex than in adults due to weightbased dosing, custom medication formulations and inability of children to communicate adverse effects.13 Studies further demonstrate that most ADEs are due to unexpected and therefore nonpreventable drug reactions, and that children on multiple prescriptions are at a higher risk of preventable ADEs.10,14 The objectives of this study were to examine the process of medication reconciliation at admission by describing (1) availability and accuracy of five medication information sources and (2) errors in admitting orders. We focused on a population of children characterised as having medically complex conditions (MCC).15 They have frequent hospitalisations, are on multiple medications and therefore are at a higher risk for errors and ADEs.10,13,14,16,17

## Methods

### Setting

Subjects were admitted to two paediatric teams at Primary Children's Medical Center (PCMC), a 253-bed tertiary care children's hospital in Salt Lake City, Utah, drawing patients from five western states. PCMC is owned and operated by Intermountain Healthcare, a not-for-profit vertically integrated healthcare system.18

### Design and study population

We conducted a prospective quality improvement cohort study from 16 December 2004 to 7 January 2005. Attending hospitalists identified new admissions as a child with MCC or not, based on number and complexity of medical problems, number of involved organ systems, specific diagnoses, clinical experience and published guidelines.15 We identified two enrolled children with newly diagnosed chronic illness not on any medications prior to admission, prohibiting reconciliation. They were excluded from the study (see figure 1).

Figure 1

Subject enrollment process

### Intervention

The quality improvement team included a hospitalist physician, a pharmacist and a project manager in the hospital quality improvement department. The team was not available on the weekends, and these admissions were excluded from the study population. Once identified as a child with MCC, the pharmacist was alerted and collected medication information on eligible participants. Five information sources were queried: parents (and patient if appropriate), primary care provider, community-based pharmacy, current admission history and physical exam note, and last admission electronic medical record. Our hospital electronic medical record includes information from past hospital admissions only. The number of attempts required to contact a source was tracked. At our institution, paediatric trainees write all admitting orders, including all medication orders. Attending physicians contribute information and guidance to the admitting orders and are involved in the order-writing processes. The standard practice for establishing medication lists for all patients at our hospital at the time of this study was not formalised and most commonly involved a review of the most recent admission and discussion with the parents if they were available. The same process was used for MCC.

### Medication information source accuracy and availability

All information provided by the five sources was included in a preliminary outpatient medication list. Source information comparison allowed refinement of the preliminary outpatient medication list, which was then rigorously verified utilising outpatient and subspecialty providers, pharmacists and families, as available, to generate a final ‘verified’ outpatient medication list. This verified outpatient medication list included correct information about name, dose, formulation, frequency of administration and route for each medication identified in the full preliminary outpatient medication list, categorised as either a true positive (if verified) or a true negative (if not verified). We also attempted to collect medication information from each source with regard to name, formulation, dosage, frequency and route. These were likewise categorised as a true positive, a true negative, a false positive (if erroneously added information) and a false negative (if erroneously excluded information). Missing data were categorised as if the source had been contacted but provided no information. Each source was evaluated independently against the verified outpatient medication list. Analysis was completed using the entire study cohort (unadjusted for availability and including missing data) and restricting the analysis to those patients successfully contacted within each source category (adjusted for availability), which may better reflect the actual accuracy of a source. Sensitivity and specificity were calculated by analysing individual 2×2 tables for each source (STATA Intercooled V.10.1, StataCorp, College Station, Texas). Availability of an information source was represented as the percentage of successful source contacts.

The verified outpatient list was compared with the admitting orders and all discrepancies recorded. Discrepancies were discussed with the admitting team including attending physician to determine if they were planned medication changes or if they represented admitting order errors. This process generated a final reconciled inpatient medication list. The total time required to complete the process for each patient, including resolving each discrepancy, was estimated. Errors in the admitting orders were categorised as omission, incorrect dosage, incorrect formulation, incorrect frequency of dosing or incorrect route of administration errors by comparing with the reconciled inpatient medication list. Expert consensus (BS, SB) methodology was used to rank each error's ADE risk as not significant (NS), significant (1), serious (2), life-threatening (3) or fatal (4).6 Since all errors were recognised and corrected, these were estimates of potential risk; no actual ADEs occurred in the study population during the study period. Descriptive statistics were calculated for errors.

### Other data sources

Patient age, gender, cost, length of hospital stay, number of diagnoses, severity of illness, number of medications, and admission type and source were obtained from the hospital's administrative database.19

### Ethics review

The study was approved by the Institutional Review Board from Intermountain Healthcare and the University of Utah.

## Results

During the study period, attending hospitalists prospectively evaluated 219 admissions; 32 (15%) were identified as children with MCC, and 23 underwent the full reconciliation process (see figure 1). Children with MCC were older (7.5 vs 4.8 years), used more resources (mean costs $11 265 vs$7851), had more diagnoses (8.1 vs 4.9) and were on more medications (7.5 vs 3.2) than non-MCC children.

Medication information collected from the five sources for the 23 children with MCC generated a preliminary list of 217 medications. After accounting for discontinued medications reported as active medications, there were 173 medications in the verified outpatient medication list. During the process of admission, 28 outpatient medications were held, 17 were erroneously omitted, and 31 new medications were initiated. When the omissions were corrected, there were 182 medications in the reconciled inpatient medication list (see figure 2). The estimated time required to complete the entire process including correcting errors was approximately 90 min per patient.

Figure 2

Medication reconciliation process

The availability ratio of the five sources is reported in table 1. Forty-two attempts to contact parents were made, with only 12 successes in our 23 subjects for a success rate of 52%. In contrast, the electronic medical record was available 20 out of 23 times for a success rate of 87%.

Table 1

Availability of medication information sources

The sensitivity and specificity of each source are listed in table 2 with 95% CI, both adjusted for source availability (excluding missing data) and unadjusted (including missing data), along with the actual total number of medications identified by each source. Adjusted sensitivity and specificity were highest for parents as the desired information source (0.75 and 0.96, respectively).

Table 2

Sensitivity and specificity of medication information sources

There were 39 admission order errors (including separately three medications with two separate errors) affecting 21.4% of all admitting order medications. These included 17 omissions (43.6% of errors), 12 dosage, two formulation and eight frequency errors, with no route errors. Errors affected 13 of 23 patients (56.5%); 21 of 39 (53.8%) errors were classified as potential ADEs. Estimated risk, if an ADE had occurred and error details for the 39 errors, are listed in table 3. Five out of 39 (12.8%) errors affecting 4/23 patients (17.4%) may have led to life-threatening ADEs (no steroid stress-dosing in dependent patients, inadvertent lowering of a tacrolimus dose despite a low level), and one out of 39 (2.6%) errors affecting 1/23 patients (4.3%) may have led to a potentially serious ADE (aspirin omission in a liver transplant patient with prior portal vein thrombosis); the remaining 15 were felt to be ‘significant’ but not serious or life-threatening. The intervention design was expected to prevent ADEs from occurring and appeared to fulfil this expectation, as no ADEs were observed.

Table 3

Medication error detail

## Discussion

This is the first published study to our knowledge that examines admission medication reconciliation in MCC and corrects errors in real time in a paediatric population. Almost one in five admission medication orders were in error, affecting more than half of children with MCC. Omission was the most frequent error type.20 Potential ADE risk was judged to be serious or life-threatening in almost one in six patients in this population. No single information source exhibited ideal availability, sensitivity and specificity.

Our findings are similar to other paediatric studies regarding frequency and type of errors and number of affected patients, and again demonstrate the importance of omission errors at transitions in care.3,20,,23 We provide novel information about the actual sources of medication information at the time of admission, including their availability and sensitivity and specificity of each source for medication information compared with rigorously verified information obtained in the reconciliation process. Parents provide accurate and only minimally incomplete information when available; the electronic medical record is the most accessible and data-rich source but, unless kept up to date, introduces opportunities for error.24 Parent availability in the first 18 h of admission was surprisingly low and may reflect a need for respite after dealing with the progressively worsening health status of their chronically ill and fragile child at home. Identification of outpatient pharmacy and primary care provider depended, in part, on parents and suffered from their unavailability. Although a rare occurrence, we experienced a malfunction of our electronic medical record on three occasions during the study period. More often than expected, we also found the history and physical exam record missing due to house staff use in morning rounds, or the chart accompanying a patient off the floor for a procedure.

Time and professional workforce required for reconciliation using this method were excessive but similar to those of other studies.21,25 We went to great lengths to verify the outpatient medication list and reconcile the admission medication list and to einsure that we did not misinterpret as errors, changes in medications by the admitting team. We studied a higher-risk population on multiple medications with significant chronic illness and examined the process when the system was stressed by the holiday season. Our approach to medication reconciliation was robust and does provide a method of creating ‘gold standard’ medication lists for comparison studies to evaluate future less labour-intensive interventions to improve medication reconciliation. Given the complexity and expertise required, we believe, as have others, that the medication-reconciliation process is more efficiently managed by pharmacists.11 26 27

Study limitations include the timing during the holiday season, which may have contributed to difficulty contacting both parents and outside sources. We noted that parents of this patient population often arrived at the hospital at the time of admission in need of respite. They would frequently ensure that their child was settled, and then go home to rest, leaving them absent during the intervention window of our study.28 This phenomenon impacted the ability of study personnel to obtain medication information from parents during the allotted time frame; parental absence also impacted our ability to ascertain both the primary care physician and the outpatient pharmacy, thereby interfering with querying of these sources. Because of poor availability of some sources, we were left with missing data. In attempting to adjust for this missing data, we completed an analysis based on successful contact with sources (adjusted analyses), in addition to our primary analyses that included the missing data (unadjusted analyses). We recognise that this might introduce bias in the adjusted analysis. We studied a small number of patients given our limited resources and the intensity of the reconciliation intervention and in only one tertiary care institution. We did not include over-the-counter medications or allergies in our error determinations. Our study was not designed to demonstrate a reduction in ADE occurrence through this intervention. Our findings cannot be extended to paediatric patients outside our study population definition of MCC.

Our population choice reflects a challenge for medication reconciliation, given the complexity of care and multiple medications in MCC, with its attendant higher risk of medication errors. We chose this group, to demonstrate the vulnerability of children who most rely on our healthcare delivery system. We believe that any systematic intervention of medication reconciliation in this group will be more likely to improve measurable outcomes and simultaneously benefit other populations at less risk.

## Conclusions

In this setting and with this patient population, errors were found in one in five admission medication orders. Almost one in six patients may have experienced a serious or lifethreatening ADE; the intervention was able to identify and rectify these. Parents provided the most accurate, although still incomplete, information, but were often unavailable. The electronic medical record was easily accessible, but accuracy was variable, and inaccuracies with the potential of leading to an error were higher. Ideally, parents must assume responsibility for providing accurate medication information at transitions in care. Short-term efforts should focus on providing parents the education and electronic or paper resources needed to maintain and communicate accurate medication information. An electronic medical record capable of capturing and continuously updating medication information may be a long-term solution; however, it must be demonstrated to be timely and accurate, and to account for medication changes that may occur outside the current pharmacy/payor/provider systems.

View Abstract

## Footnotes

• Funding RS is the recipient of a Eunice Kennedy Shriver National Institute of Child Health and Human Development career development award K23 HD052553, and this project was supported in part by the Children's Health Research Center at the University of Utah and Primary Children's Medical Center Foundation.

• Competing interests None.

• Ethics approval Ethics approval was provided by the University of Utah; Intermountain Healthcare.

• Provenance and peer review Not commissioned; externally peer reviewed. Patient consent Not obtained.

## Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.