Abstract
OBJECTIVES: Children treated with chronic medications are at risk of drug-drug interactions (DDIs) when hospitalized with an acute illness and prescribed new medications. We aimed to measure the prevalence of potential DDIs (pDDIs) among hospitalized children treated with antiepileptic drugs (AEDs) and to evaluate the impact of computerized physician order entry (CPOE) on pDDIs.
METHODS: We analyzed a national sample of pediatric hospitalizations from 2005 to 2012 associated with administration of an AED and identified those prescribed a second medication with risk of a DDI. The prevalence of hospitalizations associated with a pDDI was calculated for each AED. We identified the drugs most commonly implicated in pDDIs and factors associated with pDDIs. Rates of pDDIs were measured in pre- and post-CPOE implementation periods.
RESULTS: A pDDI was identified in 181 380 (41.7%) hospitalizations associated with the use of an AED, with 117 880 (27.1%) classified as severe pDDIs. AEDs most often implicated with a pDDI were phenobarbital, valproic acid, and phenytoin. Hospitalizations with pDDIs were associated with increased length of stay and a greater number of medications, ICU admissions, and operating room procedures. The implementation of CPOE did not result in a change in the rate of pDDIs (42.7% before versus 40.8% after; P = .48).
CONCLUSIONS: Children treated with AEDs are at risk of pDDIs while hospitalized. The use of CPOE has not been associated with a significant decrease in the rate of pDDIs. Additional investigation to better define the impact of pDDIs and to advance development of clinical decision support within CPOE systems is warranted.
Drug-drug interactions (DDIs) occur frequently in the hospital setting and lead to substantial hospital-related morbidity.1,2 Children receiving chronic medications who are hospitalized for an acute illness and prescribed a new medication may be at particular risk of an adverse drug reaction related to a DDI unless careful consideration is given to the child’s home medication regimen and his or her potential for a DDI.3
Through embedded physician alerts, computerized physician order entry (CPOE) has the potential to reduce dosing, frequency, and route errors, as well as prevent the prescribing of medications to which a patient is allergic or that may result in a potential interaction with another medication.4 Surveys of physicians using CPOE have found that most believe a large proportion of medication errors could be reduced through the use of these systems and substantial reductions in errors have been shown after the implementation of CPOE.4,5
Epilepsy is a common pediatric disease that is treated primarily with pharmacotherapy and typically requires lifelong drug therapy. In as many as 25% of children, treatment with multiple agents is required to achieve adequate seizure control.6,7 Unfortunately, many antiepileptic agents have significant DDIs, both with other antiepileptic drugs (AEDs) and with other classes of medications.3,8 These interactions may result in changes in the metabolism of AEDs, which have a very narrow therapeutic index, and lead to toxic serum drug concentrations or break-through seizures. Alternatively, drug interactions with AEDs may result in side effects or reduced efficacy for other agents. A number of factors contribute to whether a clinically meaningful interaction occurs, including timing of drug administration, drug dose, and the presence of other drugs, food, or environmental compounds. The coadministration of 2 medications that may lead to a DDI is therefore monitored in terms of potential DDIs (pDDIs).5,9
In this study, we sought to determine the rate of pDDIs among hospitalized children treated with AEDs. Our secondary aim was to evaluate the impact of the implementation of CPOE on the prevalence of pDDIs.
Methods
Study Design and Case Selection
We conducted a retrospective study using data from the Pediatric Health Information System (PHIS) database. PHIS is a proprietary database compiled and managed by the Children’s Hospital Association. Forty-four freestanding children’s hospitals, located throughout the United States and representing a total of 46% of all discharges from pediatric hospitals, contributed de-identified patient data during the study period.10 We analyzed data from 43 hospitals that were providing complete data to PHIS during our study period. This included 3 hospitals that began submitting during this time period. Information available in PHIS includes patient demographic characteristics, admission dates and length of stay, treatments administered, diagnostic tests and procedures performed, illness severity scores, and information on physician specialty. Diagnoses are presented as International Classification of Diseases, Ninth Revision, codes and also in terms of all patient review-diagnostic–related groups. The database includes flags specifying care received in the ICU or in the operating room. Physicians are identified as attending or consulting physicians, and information is provided on their subspecialty. We included all hospitalizations for patients aged ≤21 years who were admitted to a participating PHIS hospital between January 1, 2005, and December 31, 2012, and associated with the administration of an AED. This upper age range was chosen on the basis of frequent admission of patients up to this age among tertiary care pediatric hospitals. The study was deemed exempt by the Institutional Review Board at Boston Children’s Hospital.
Drug Selection and Classification
AEDs were defined as those categorized as antiepileptics in the World Health Organization’s Anatomic Therapeutic Chemical classification system (group N03A). In selecting hospitalizations for inclusion, we did not include those associated with therapy with the AED fosphenytoin alone, because this drug is administered on an acute basis and is not used in chronic therapy for epilepsy. However, we did analyze interactions resulting from the combined use of this agent with other AEDs.
Drugs that interact with AEDs were identified by using Micromedex, a drug database that includes detailed medication-prescribing information, including data on pDDIs.11 For each AED, we identified all of the medications with a potential for interaction. A pDDI was defined as the administration of 2 drugs with the potential for interaction prescribed within a 1-day period, a commonly used time interval for defining pDDIs and triggering medication alerts in CPOE systems.12 The number of pDDIs ranged from 2 among some of the newer AEDs, such as lacosamide and pregabalin, to >200 for carbamazepine. We extracted information on the risks associated with each DDI, which included change in AED serum concentration, change in interacting drug serum concentration, central nervous system effects, and respiratory depression.
We also collected data on the severity of the DDIs. Micromedex categorizes the severity of drug interactions as “contraindicated” “major” (interactions that “may be life threatening and/or require medical intervention to minimize or prevent serious adverse effects”), “moderate” (interactions that “may result in an exacerbation of the patient’s condition and/or require an alteration in therapy”), or “minor” (interactions that would have “limited clinical effects”). We defined severe interactions as those of contraindicated or major severity per Micromedex. This definition is consistent with the definitions used in previous studies assessing the prevalence of pDDIs.12–15
CPOE
To assess the potential impact of CPOE on the prescribing of medications that can result in an interaction, we collected information on implementation of CPOE at each of the PHIS hospitals. Hospitals were individually contacted, and information was obtained on the year of CPOE implementation in the inpatient setting. Pre- and post-CPOE periods were defined for each hospital on the basis of these dates with the year of CPOE implementation considered as a pre-CPOE period in the analysis.
Statistical Analysis
We calculated the number of hospitalizations associated with a pDDI between an AED and another agent and the proportion of those that were nonsevere and severe. Analysis of variance and logistic regression were used to compare the 3 pDDI groups (severe, nonsevere, and none) on characteristics of patients and their hospitalizations, who were chosen a priori. Next, we identified the most common drug combinations implicated in any pDDI and in severe pDDIs. The impact of CPOE was evaluated by using a random-intercept (hospital ID) generalized linear mixed effects model to assess the effect of year (2005–2012) and CPOE status (before/during versus after implementation) on hospitalizations for any pDDI and for severe pDDIs. Both year and CPOE status were treated as categorical variables. The significance level was set at an α of <.05. All analyses were performed by using SAS (version 9.4; SAS Institute, Cary, NC).
Results
A total of 434 794 hospitalizations were associated with an AED, of which 181 380 (41.7%) involved a pDDI. Severe pDDIs were observed in 117 880 (27.1%) of these hospitalizations. The most common potential effects of the interactions included change in AED serum concentration (48.3%), change in non–anti-epileptic drug serum concentration (48.9%), and respiratory depression (41.3%). Respiratory depression may be less relevant in the ICU setting where patients may already be intubated or have other respiratory support. The rate of this potential effect was 29.7% among patients who were not admitted to the ICU and 59.2% for those who were in the ICU.
Demographic and medical characteristics of patients differed between those with pDDIs and those without pDDIs who were treated with an AED (Table 1). Potential severe interactions were observed more frequently among younger patients as well as patients who had longer hospital stays, received a greater number of antiepileptic and non–anti-epileptic medications, were treated in the ICU or operating room, and had a greater mean number of consulting services. Seizure was the most common diagnosis (25.4% of all treated with an AED, 12.3% of potential severe interactions) by all patient-review diagnostic-related groups. This rate was followed by 3.6% of hospitalizations with craniotomy for trauma and 3.1% for chemotherapy.
Characteristics of Hospitalized Patients Treated With an AED and Those With a pDDI
Table 2 provides the number of hospitalizations associated with a pDDI for each of the AEDs examined. Phenobarbital accounted for the greatest proportion of hospitalizations associated with a pDDI (16.6% of all hospitalizations with AED therapy, 39.8% of hospitalizations associated with a pDDI) and was responsible for 52.5% of the severe pDDIs. The most common drugs at risk of interaction with phenobarbital were lorazepam, midazolam, and morphine. Valproic acid and phenytoin were the next most frequent AEDs associated with pDDIs, with 17.6% and 14.2% of hospitalizations associated with a potential interaction, respectively. A variety of medications resulted in pDDIs with these drugs, including sedatives, pain medications, and psychotropic drugs. In terms of severe pDDIs, after phenobarbital, clonazepam was most frequently implicated (13.1% of all severe pDDIs) followed by phenytoin (10.0%).
AEDs and the Most Common Interacting Drugs as a Percentage of Hospitalizations With an AED
Among all hospitalizations with a pDDI, 36.0% were related to an interaction with a pain medication, 12.5% to an interaction with a steroid, and 4.9% to an interaction with an antimicrobial agent (Fig 1). The most common pain medications with an AED interaction were morphine, ketorolac, and fentanyl, representing 31.4%, 23.7%, and 8.3%, respectively, of all AED hospitalizations with a pDDI. The steroids most commonly involved in a pDDI with an AED were dexamethasone (11.5%) and hydrocortisone (7.1%), whereas the antimicrobial agents with the greatest frequency of pDDIs were metronidazole (2.9%) and acyclovir (1.8%).
Types of acute medications associated with a potential drug-drug interaction. Note: there is overlap where some patients have interactions with multiple acute medication types.
Because pain medications were the drug type associated with the highest number of pDDIs, we looked at pDDIs for hospitalizations with (N = 69 635) and without (N = 111 745) an operating room procedure. Morphine, ketorolac, and fentanyl remained the top 3 pain drugs involved in a pDDI, with rates for each drug differing by operating room status. For hospitalizations with an operating room procedure, rates were 49.6%, 34.3%, and 14.8% respectively, for the 3 drugs. For hospitalizations without an operating room procedure, rates were 20.1%, 17.1%, and 4.2%, respectively.
Thirty-six of the hospitals studied had implemented CPOE by the end of the study period, with 25 (58%) initiating CPOE during the study period (Fig 2). The rates of pDDIs were compared in pre- and post-CPOE periods across hospitals and showed a small but nonsignificant decrease after CPOE implementation (Table 3). Among all hospitalizations associated with an AED, 42.7% involved a pDDI in pre-CPOE periods compared with 40.8% in post-CPOE periods (P = .48). Similarly, severe pDDIs were identified in 27.9% of hospitalizations in pre-CPOE periods and in 26.4% of hospitalizations in post-CPOE periods (P = .91). Notably, when controlling for CPOE implementation, the rates of pDDIs decreased over time from 49.6% in 2005 to 36.0% in 2012 (P < .001). The rate of severe pDDIs also decreased significantly over the study period (31.5% in 2005 to 23.8% in 2012; P < .001). The AEDs with the highest rates of pDDIs were the same before and after CPOE implementation (phenobarbital, valproic acid, and phenytoin).
Year of CPOE implementation at study hospitals. Note: 4 hospitals had not initiated CPOE at the time of this study and are not included in the figure.
Hospitalizations With an AED by Status of CPOE and Year of Admission
Discussion
Our findings indicate that children taking AEDs are at high risk of a pDDI while hospitalized, with more than one-quarter of patients exposed to a severe pDDI. Younger patients and those with associated ICU or operating room procedures are at highest risk of potential severe interactions. The majority of these pDDIs are related to the administration of pain medications, steroids, and antibiotics. Although there was a significant overall decrease in the rate of pDDIs over the study period, this decrease was not directly associated with the implementation of CPOE systems.
An estimated 31% of children have chronic medical conditions, of whom 63% are taking medications that will likely be continued when hospitalized with an acute illness. Physicians prescribing medications to hospitalized patients may not always take into account previously prescribed drugs when making treatment decisions. Antiepileptic drugs pose a particular risk for DDIs given the high rate of interactions associated with many of these agents, although high rates of DDIs have been documented among hospitalized patients across a spectrum of drug classes.3,5,13,16,17 Our findings indicate that certain patient characteristics are associated with a higher risk of a pDDI, which may help identify patient populations who require particular vigilance around medication therapies.
A number of resources exist for physicians to verify pDDIs when prescribing an AED. In addition to Micromedex, many physicians also have access to Lexicomp or Epocrates, all of which include sections dedicated to the discussion of drug interactions.18 These resources do not always correlate with regard to pDDIs. The Food and Drug Administration has also developed a teaching module geared specifically toward helping clinicians recognize and prevent drug interactions.19
Health information technology and CPOE have the potential to ensure that pDDIs are detected by physicians at the time of prescribing via decision support systems embedded in electronic health records.20,21 These systems have been associated with increased adherence to guidelines, improved implementation of quality initiatives and measures, and reduction in rates of medication errors.22 CPOE has been observed to decrease medication errors among pediatric patients by as much as 40% in the general ward setting, with a reduction in adverse drug events of 41%.23,24 Studies have also shown that CPOE systems may reduce the number of pDDIs among hospitalized patients, although not all studies showed a significant change.21,24,25
Despite the promise of CPOE, the potential to reduce pDDIs has not been maximized. We observed only a small, nonsignificant decrease in the rates of pDDIs in pre- versus post-CPOE periods. There are a number of factors that may contribute to the potentially limited impact of CPOE on pDDIs. First, many pDDIs may not be detected in current clinical decision support systems, with some estimates indicating that more than half of all pDDIs are not captured.26,27 This finding may be related to a lack of continuous updates to databases of pDDIs or to poor customization of decision support for local use in specific patient populations and settings.28 Second, there is little consensus across institutions as to which pDDIs should trigger an alert to prescribers. Currently, there are no national standards or definitions for critical DDIs among AEDs or other chronic drug classes that would ensure effective screening for DDIs among hospitalized patients.29 Third, the prescriber may have few therapeutic alternatives and may be prepared to respond to the potential interaction. Fourth, the interaction may not be perceived as important relative to the potential benefit of therapy on the basis of previous experience of the patient or clinician.30,31 Finally, other user interface factors or alert fatigue may simply lead clinicians to ignore the alerts.28,30,32–34
The primary limitations of our study are related to the use of retrospective administrative data. We were limited to the information collected in the data set and could not further explore factors contributing to the ordering of medication combinations resulting in pDDIs or to evaluate whether the pDDIs resulted in patient harm. In addition, we relied on a preestablished classification system to identify pDDIs and severe pDDIs (Micromedex). Although this is a rigorous and widely used data set for this information, it is possible that the use of other available pharmacologic databases could have yielded slightly different results. Finally, we selected the cohort of hospitalizations on the basis of treatment with AEDs used in the chronic management of seizure disorders. This approach presents a risk of misclassification because it may include patients treated with AEDs for an acute indication. We examined this specifically and found that 3.6% of patients received AEDs related to a craniotomy. This percentage is a small and unlikely to affect the generalizability of our findings to patients receiving chronic AED therapy.
Conclusions
Children receiving antiepileptic drug therapy are at risk of pDDIs when hospitalized and more than one-quarter receive drug combinations that have the potential for severe interactions. Children with longer hospital stays and a higher mean number of medications, operating room procedures, and admission to the ICU appear to be at increased risk of pDDIs. The use of CPOE has not resulted in a significant decrease in the rate of pDDIs among patients receiving AEDs, highlighting the need for additional investigation to better define the impact of pDDIs and to advance development of clinical decision support within CPOE systems.
Acknowledgments
We thank Dr Shannon Manzi, PharmD, Boston Children’s Hospital Department of Pharmacy, for her contributions and input on pharmacy databases and drug-drug interactions.
Footnotes
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: No external funding.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
References
- Copyright © 2016 by the American Academy of Pediatrics
In this issue
