 |
 |
 |
 |
 |
 |
 |
|
|
|
|||||||||
|
|||||||||||
Professor, Melbourne School of Health Sciences, Faculty of Medicine Dentistry and Health Sciences, The University of Melbourne, Carlton, Victoria, Australia
Senior Research Fellow, Melbourne School of Health Sciences, Faculty of Medicine Dentistry and Health Sciences, The University of Melbourne
Honorary Senior Fellow, Department of Medicine, Faculty of Medicine Dentistry and Health Sciences, The University of Melbourne
Statistical Consultant, Statistical Consulting Centre, Department of Mathematics and Statistics, The University of Melbourne
Reprints: Dr. Manias, Melbourne School of Health Sciences, Faculty of Medicine Dentistry and Health Sciences, The University of Melbourne, Level 5, 234 Queensberry St., Carlton, Victoria 3053, Australia, fax 61 3 9317 4375, emanias{at}unimelb.edu.au.
 |
Abstract |
|---|
 Top Abstract MethodsResultsDiscussionReferences |
|---|
OBJECTIVE: To identify the patient-, environment-, and medication-related factors involving unexplained medication discrepancies across transition points after ED presentation.
METHODS: Using a retrospective chart review design, a stratified, random sampling of data was undertaken over a 12-month period. Information was obtained from an electronic administrative database and medical records as patients moved from the ED to another transition point of care. Medication discrepancies were classified into 2 outcome groups: (1) no discrepancies and situations in which discrepancies were adequately explained and (2) discrepancies that had no adequate explanation.
RESULTS: For the 12-month period, 210 randomly selected patients were included; 73 (34.8%) had at least one unexplained medication discrepancy. Binary logistic regression modeling showed 4 factors that were statistically significant in determining the incidence of at least one unexplained medication discrepancy. Benefit card holders (individuals who receive benefits from government insurance programs comparable to the US-based Medicare and Medicaid initiatives, which include the elderly, the disabled, low income earners, and unemployed persons) had 3.73 greater odds of experiencing an unexplained medication discrepancy (95% CI 1.72 to 8.07; p = 0.001). Patients prescribed 5 or more drugs at discharge from the ED had 12.22 greater odds of having at least one unexplained medication discrepancy (95% CI 5.52 to 27.08; p < 0.001). Patients who were first seen by a physician within 1 hour of a change in working shift had 3.70 greater odds of having an unexplained medication discrepancy (95% CI 1.67 to 8.18; p = 0.001). For each additional minute of wait time for a physician, the odds of having an unexplained medication discrepancy increased by a factor of 1.01 (95% CI 1.00 to 1.01; p = 0.042).
CONCLUSIONS: Patient-, environment-, and drug-related factors contribute to the risk of medication discrepancies across transition points from the ED.
Key Words: care transition, communication, emergency department, medication discrepancy, medication reconciliation
Published Online, October 20, 2009. www.theannals.com, DOI 10.1345/aph.1M206
Past investigations of medication discrepancies have focused on admission and discharge practices in general medical and surgical settings.5-13 Little consideration has been given to the emergency department (ED) context of care; however, a recent survey of ED physicians has shown that it is an environment that is prone to medication errors.14,15 Patients who present to the ED with an unplanned medical event are most vulnerable to medication mismanagement because they have just sustained a critical illness and often cannot communicate adequately. The ED is also an environment characterized by the need for decisions to be made rapidly under high levels of stress. In view of the great vulnerability of patients in this context, it is important that medication management be examined comprehensively.16
The aim of this study was to measure the incidence of patients having at least one unexplained medication discrepancy after presentation to the ED. The objectives were to identify patient-, environment-, and medication-related factors associated with these discrepancies during patients' time in the ED.
|
Methods |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
DATA COLLECTION
A stratified random sampling technique was undertaken for data collection.
For each month of the review period, all patients who had presented to the ED
were assessed for eligibility. The variable used for stratification related to
discharge destination of patients. Attempts were made to ensure that equal
numbers of patients admitted to a surgical, medical, or short-stay emergency
unit were obtained. Remaining patients recruited over the study period were
discharged home. Patients were eligible for inclusion if they were aged 18
years or older, if they presented to the ED for an unplanned admission, and if
they were seen in the ED by a physician. Patients were excluded if they
transferred from another hospital, were dead on arrival, were readmitted
during the study period, or were previously included in the study. A random
numbers table was used to select which eligible patients' records were to be
accessed for each month, and between 16 and 18 patient records were accessed
each month to derive the sample.
The sample size was calculated using the confidence intervals method, based on an estimated prevalence of the outcome.18 It was hypothesized that the outcome, which pertained to patients having had at least one medication discrepancy without appropriate clinical explanation, would have an estimated prevalence of 30%. Using this estimated prevalence, a sample size of 191 would be required to produce a confidence interval of ±6.5% precision or 0.13 for the desired interval width of the 95% confidence level. The sample was obtained by collecting approximately 16–18 patient medical records per month for the review.
Information was sought from the Patient Admission System, which is an electronic means of collecting demographic and clinical data about patient care. Medical records were also manually examined. Collectively, data information helped to identify patient-, environment-, and medication-related factors that could influence the occurrence of medication discrepancies as patients moved from the ED to another transition point, such as a surgical or medical ward, or home (Appendix I).
|
Patient-related factors included the patient's age, sex, chief type of complaint for admission to the ED, need for an interpreter, visual deficit, hearing deficit, and benefit card status. Benefit card holders involved individuals receiving benefits from government insurance programs, which are comparable to the US-based Medicare and Medicaid initiatives, and included the elderly, disabled, low income earners, and unemployed. Medication-related factors included the number of medications taken by the patient at discharge from the ED and presence of a medication allergy. Environment-related factors included the discharge destination, position of the attending physician, mode of arrival to the ED, arrival time at the ED, discharge time from the ED, duration of stay in the ED, time first seen by the ED physician, time first seen being within 1 hour of change of working shift, wait time before an ED physician examined the patient, and allocation of category on the Australasian Triage Scale. The Australasian Triage Scale classifies the urgency of presentations to Australian EDs and comprises 5 categories.19,20 A pharmacist was employed in the ED halfway into the review period, and the effect of this environmental factor was also examined.
Medications were audited and compared at 2 time points. At the first time point, a description of all medications at the point of entry to the ED was collected as determined by the ED physician at first interview. This description included the dose and frequency of all medications taken, the method of administration and type of formulation, and the presence of any known allergies. This was a routine component of the initial patient assessment by the ED physician and was documented in patients' ED admission notes in their medical records. At the second time point, a description of the drugs ordered for patients at discharge from the ED was documented by the admitting physician of the inpatient department, in consultation with the ED physician. If patients were not admitted, the ED physician documented the medications to be prescribed for patients upon discharge from the hospital. The description included the dose and frequency of all medications taken, the method of administration and type of formulation, and the identification of any new medication allergies as well as other known allergies. This information was documented on the patient's medication chart. Medication discrepancies between the 2 time points were examined. Sources of information for drugs at the point of entry included ED discussions with patients, ambulance officers, and family members or other people accompanying the patient; outpatient notes; ED nurses' notes from communication they had with patients and accompanying people; staff pharmacists' notes; and community physicians' notes.
MEASURES
A medication discrepancy was defined as any difference observed between the
drugs listed on the medication chart by the attending physician at the time of
discharge from the ED and those contained in the medication history obtained
at admission. The discrepancies were systematically identified and categorized
into 1 of 8 forms by a pharmacist not involved in the patient's care. These
categorizations have been validated in past
research.9,21
DATA ANALYSIS
Statistical analysis was primarily carried out on patient-level data using
SPSS version 16 (SPSS Inc., Chicago, IL). The outcome (dependent) variable of
consideration was defined to be binary: (1) patients who had no medication
discrepancies or whose medication discrepancies were explained and (2)
patients who had at least one unexplained medication discrepancy. For
statistical analysis, categories of discrepancy were collapsed into 2 outcome
groups: no discrepancies, which also included situations in which
discrepancies were adequately explained (categories 1–4), and
discrepancies that had no adequate explanation (categories 5–8).
Descriptive statistics analysis was undertaken for all variables. Univariate
associations with the medication discrepancy outcome were first investigated
for all patient-, medication-, and environment-related factors using
cross-tabulations and 
2 tests for categorical variables
and binary logistic regression for continuous variables. Binary logistic
regression modeling was then undertaken with all factors, using a forward
stepwise approach, which is an appropriate method if little previous work has
been completed in the particular context under investigation. In the forward
stepwise method, the model begins with a constant and tests whether added
variables have a substantial effect in terms of how well the model fits the
observed
data.22
Unexplained discrepancies were assessed as having an unlikely, possible, or
probable potential to cause patient discomfort or clinical harm, using an
instrument described in past work (Appendix
II).5
This process of examining the clinical impact was undertaken independently by
2 researchers. A p value of less than 0.05 was considered statistically
significant. Interrater reliability for assessing the potential of unexplained
discrepancies to cause patient discomfort or patient harm was analyzed using
Cohen's 
measure of
agreement.22
|
|
Results |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
|
|
CHARACTERISTICS OF MEDICATION DISCREPANCIES
A total of 1087 medications were prescribed at the time of discharge from
the ED across the sample of 210 patients, according to the 8 categories of
discrepancies (Table 2). For
37.6% of the drugs, no discrepancy was detected. For all 210 patients, 73
(34.8%) had at least one unexplained medication discrepancy. Most notably,
13.2% (143/1087) of medications were associated with unexplained discrepancies
for which there was a lack of clinical explanation for changes made. Of the
unexplained discrepancies (n = 143), 84 (58.7%) were unlikely to cause patient
discomfort or clinical harm, while 55 (38.5%) and 4 (2.8%) discrepancies had
the potential to cause either possible or probable patient discomfort or
clinical harm, respectively. The score obtained for agreement between
the 2 researchers assessing clinical impact of medication discrepancy was
0.97. Table 3 shows the
formulation type and medication type in relation to no discrepancy or an
explained discrepancy, and unexplained discrepancy determined for all 1087
medications.
|
|
UNIVARIATE ANALYSES
Univariate analyses for patient-related factors revealed that increased
patient's age, the patient's need to have an interpreter present in the ED,
benefit card status, and presence of a visual deficit were statistically
significantly related to the incidence of an unexplained medication
discrepancy (Table 4). Two
drug-related factors—having 5 or more medications ordered at the time of
discharge from the ED and having a known medication allergy—were
statistically significant relative to the incidence of an unexplained
medication discrepancy in the univariate analysis
(Table 4). Univariate analysis
also revealed that 3 environmental factors were statistically significantly
associated with the incidence of an unexplained medication discrepancy:
patients' admission to a hospital inpatient setting, the type of discharge
destination, and being seen by an ED physician within 1 hour of change of the
working shift (Table 4).
|
EXPLANATION OF PREDICTOR EFFECTS BY LOGISTIC REGRESSION MODELING
Binary logistic regression modeling involving a forward stepwise variable
selection approach was used to identify the best combination of patient-,
medication-, and environment-related factors for predicting the occurrence of
an unexplained medication discrepancy. All factors were considered for
inclusion in the model regardless of whether they showed a significant
relationship in univariate analysis. One patient-related factor was
independently associated with an unexplained medication discrepancy. Patients
who were benefit card holders had 3.73 greater odds of experiencing an
unexplained medication discrepancy compared with those who were not benefit
card holders (95% CI 1.72 to 8.07; p = 0.001)
(Table 5). Regarding
medication-related factors, patients who were prescribed 5 or more drugs at ED
discharge had 12.22 greater odds of having at least one unexplained medication
discrepancy compared with those who were prescribed fewer than 5 medications
at ED discharge (95% CI 5.52 to 27.08; p < 0.001). Two environment-related
factors were significant: the patient being seen by an ED physician within 1
hour of change of the working shift and the physician wait time were
independently related to the incidence of an unexplained medication
discrepancy. Patients who were first seen by a physician within 1 hour of a
change in working shift in the ED had 3.70 greater odds of having an
unexplained medication discrepancy compared with those who were first seen by
a physician at other times (95% CI 1.67 to 8.18; p = 0.001). As the physician
wait time increased, patients had 1.01 greater odds of having an unexplained
medication discrepancy (95% CI 1.00 to 1.01; p = 0.042).
|
The multivariate logistic regression model had a sensitivity of 70%, which is the proportion of the sample with at least one unexplained medication discrepancy that was correctly identified by the model. The specificity was 86%, which relates to the percentage of the sample without at least one unexplained medication discrepancy that was correctly identified by the model. The positive predictive value was 73%, while the negative predictive value was 84%. Using this model, we were able to correctly classify 81% of cases overall.
|
Discussion |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
The unexplained medication discrepancy rate of 34.8% (73/210 patients) was lower than that reported by Lau et al.21 (67.0%) and Cornish et al.5 (53.6%), but was higher than that observed by Miller et al.23 (4.3%). Variations in discrepancy rates are likely due to the different environments and time periods in which studies were undertaken. In the Lau et al.21 study, patient records were accessed from general internal medicine wards of 2 acute care hospitals over a 2-year period from 1993 to 1995. Cornish et al.5 accessed records during 3 months in 2003 from patients admitted to the general internal medicine clinical teaching units of a tertiary care teaching hospital. In Miller et al.'s23 study involving patients admitted to a trauma center during 2005, the urgency of treatment decisions that needed to be made was likely to have been similar to that of the ED in our study. The large difference in rates in the 2 settings may have been due to the formalized medication reconciliation process instituted by Miller et al.
Previous studies have also shown that omission of medication was the most common form of unexplained discrepancy, which corresponds with our findings.13,24 In our study, the omission of medication pertained to the transition period from the ED to another setting, which occurred during the course of 1 day. The clinical impact of this finding is not clear because of the short time involved. Further clarification is needed as to what happens to these omissions at the time of discharge from inpatient hospital environments.
The logistic regression modeling process identified a combination of 4 factors that significantly statistically predicted the incidence of at least one unexplained medication discrepancy. The patient factor that was shown to be significant was the patient's status as a benefit card holder. The medication factor shown to be significant was patients who had 5 or more medications ordered at the time of discharge from the ED. Patients presenting to the ED with these 2 characteristics can be targeted by ED pharmacists to pre-empt possible medication problems at the time of discharge from the ED. Two significant factors were of an environmental nature: the ED physician seeing a patient within 1 hour of the change of shift and the physician wait time.
Benefit card holders, who included the elderly, disabled, unemployed, and low income earners, had significantly increased odds of an unexplained medication discrepancy. It is possible that being a benefit card holder is a confounding factor in our study because, typically, these individuals are prescribed many medications and are therefore at high risk of experiencing an unexplained medication discrepancy. Nevertheless, benefit card holders have been shown to have multiple chronic diseases and are regular users of EDs.25 In our study cohort, benefit card holders were prescribed significantly more medications compared with those without benefit card status. We are not aware of prior studies that evaluated the influence of insurance status on medication discrepancy. Social and economic disadvantage plays an enormous role in healthcare access and equity, as individuals without insurance are often least able to advocate for their needs. It is therefore possible that, due to the complexity of their situation, benefit card holders may not receive adequate health care. Such a disadvantage has an impact on how medications are managed in EDs; clearly, this is an area for consideration in future work.
Patients prescribed 5 or more drugs at the time of discharge from the ED had greatly increased odds of experiencing an unexplained medication discrepancy. The mean number of medications prescribed for patients before ED presentation was 3.2, while the mean number prescribed at discharge from the ED was 5.3. Hence, medication regimens were significantly altered during a patient's stay in the ED. An acute crisis can cause physicians to withhold certain medications, cease medications, or change the dosage regimen.25 These alterations become even more pronounced as the number of drugs taken by patients increases.8 Furthermore, at many hospitals, the presence of closed formularies and the stocking of certain preparations lead to automatic substitution of one drug for another in the same class.26
A notable factor that had significant impact on unexplained medication discrepancy was patients being seen by an ED physician within 1 hour of change of a working shift. Interestingly, discharge time of the patient was not a significant factor for the incidence of an unexplained medication discrepancy. A possible reason for lack of effect of the timing of discharge is that all treatments in the ED are undertaken at different times depending on the patient's urgency of need and severity of illness. Once a patient's condition has stabilized, there may be considerable delay in organizing appropriate transfer arrangements from the ED. On the other hand, as ED physicians near the end of their working shifts and are still required to complete a number of tasks, there may be opportunities for errors to occur.
Another environment-related factor, physician wait time in the ED, demonstrated that as wait time increased, the odds of experiencing an unexplained medication discrepancy also increased. Many reasons could account for this. ED physicians can delay making a definitive diagnosis as they wait for pathologic and radiologic tests to be completed.14,27 Ambiguous responsibilities in the designation of a definitive diagnosis may also be associated with a prolonged wait time to see an ED physician, thereby contributing to communication problems and medication discrepancies.14,28 A lengthy wait time may also be related to high workloads, which occur when an ED is crowded or when admitting physicians receive several new patients simultaneously. High workload can have an enormous impact on the quality of communication. When clinicians are busy, handoffs around the point of transfer from the ED can be rushed and not interactive.14 While the physician wait time was shown to be a statistically significant factor affecting medication discrepancies, further research is needed on larger samples of patients to determine the clinical significance of this effect.
Problems related to transitions in the ED are likely to have complex causes that are resistant to single interventions in promoting medication reconciliation. The ED has been identified as an environment associated with a number of vulnerable areas, including lack of effective communication, high workload, ineffective physical design, insufficient access to changes occurring in patient status and treatment orders, ineffective patient flow, and lack of follow-up on pending information.14 Targeted strategies and systems-based interventions that focus on the actual context of the ED environment could help to ameliorate medication discrepancy problems and improve patient safety.29,30
Limitations of this study include its retrospective nature; thus, reasons for changing medication regimens may not have been properly documented. In addition, the single-center design may limit the applicability to institutions with a different focus. Medication reconciliation was not a formalized process in the hospital during the time when data were collected. In addition, no attempt was made to examine how unexplained medication discrepancies that occurred in the ED were resolved or identified by clinicians as patients progressed through the hospital system.
The novel approach used in this study identifies the complexity of patient-, medication-, and environment-related factors that have an impact on medication discrepancies across transition points of care from the ED. It is important to examine the influence of these factors in different ED environments. Confirmation is required about the influence of various factors by direct observation of clinical practice where patients are mapped across various environments as they move through the hospital from admission to discharge.
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractMethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
 |
 |
 |
 |
 |
 |
