Background
Adverse drug events (ADEs) are a major cause of morbidity and mortality in hospital practice. An ADE is defined as an injury resulting from medical interventions related to a drug [
1]. A preventable ADE is caused by an error in the medication use process, such as a prescribing error [
2]. The occurrence of ADEs in hospitalised patients described in the literature varies between 2 and 52 ADEs per 100 admissions. An estimated 15% to 59% of these ADEs are considered preventable [
3]. Most research on the occurrence and nature of ADEs focuses on ADEs in general medical care units, medical intensive care units and paediatric units. However, not much specific attention is given to ADEs in surgical patients. This is remarkable since the hospitalisation process of surgical patients differs greatly from that of nonsurgical patients.
Surgical patients inherently have operative interventions, and these interventions cause medication changes [
4]. A period of preoperative fasting is mandatory for surgical patients in order to reduce the risk of pulmonary aspiration during intubation. The period of fasting usually affects preoperative medication intake, and the route of administration frequently needs adjustment [
5]. Surgical patients will also require (preoperative) antibiotics, analgesics and muscle relaxants to which many ADEs can be attributed [
6,
7]. In addition, half of the surgical patients take medication unrelated to surgery. That alone increases the relative risk of a postoperative complication by 2.7 (95% CI 1.76-4.04) [
8].
Taking these issues into account, it was expected that surgical patients would have an increased risk for experiencing an ADE during their admission compared to nonsurgical patients. If the occurrence and nature of ADEs in surgical patients is specified, strategies can be developed to prevent ADEs on surgical wards. Therefore, the objective of this review was to determine the occurrence and nature of ADEs in surgical patients. We furthermore aimed to compare ADE occurrence in surgical patients and nonsurgical patients.
Methods
Search strategy
Two reviewers (EBB, MdB) performed a computer-assisted search of the medical databases Embase and Medline (from 1980 to April 2011) with the aid of a clinical librarian. A combined search term was constructed with keywords in title or abstract, as outlined below. The search was aimed at finding articles that reported observational data on ADEs in surgical hospitalised adult patients. There was no language restriction.
Keywords used to retrieve studies on adverse drug events were: ‘adverse drug events’ , ‘ADE’ , ‘medication related problems’ , ‘adverse drug reaction reporting system’ or ‘Drug therapy/adverse effects’. The terms ‘surgical’ , ‘surgery’ , ‘operation’ , ‘preoperative’ , ‘perioperative’ or ‘postoperative’ were added to specify surgical patients. The terms ‘hospitalized’ or ‘hospitalised’ , ‘hospitalization’ or ‘hospitalisation’ , ‘hospital’ or ‘inpatients’ were included in order to retrieve studies on hospitalised patients. The terms ‘frequency’ , ‘incidence’ or ‘epidemiology’ were used to retrieve studies with epidemiological data.
To exclude children and incidents at presentation on the emergency department, study titles containing the terms ‘child’ , ‘children’ , ‘paediatrics’ or ‘emergency’ were excluded. A manual cross-reference search of the eligible papers was performed to identify other relevant articles.
Study selection and data collection
Two independent reviewers (EBB, MdB) selected the articles based on titles and abstracts. Prospective studies that evaluated ADEs in adult hospitalised surgical patients were included as well as studies from which the data of ADEs in surgical patients could be extracted. The ADE definition was not an inclusion criterion. Articles concerning ADEs in outpatients, children, or in patients at presentation to the emergency department were excluded. Furthermore, studies that did not include surgical patients or studies from which the data of surgical patients could not be extracted were also excluded. Studies that evaluated ADEs related to one specific drug were not included in this review. If the title and abstract justified inclusion, the full text was retrieved.
Each selected article was fully read by two authors (EBB, MdB) and tested for the inclusion and exclusion criteria. If the data of an article was also presented in another previously selected article, the article was to be excluded. Disagreements about inclusion were discussed with a third reviewer (JK) in a consensus meeting.
Summary measures and synthesis of results
Both reviewers independently extracted data using a standardized form. The primary endpoint was: the occurrence and nature of ADEs, i.e. the causality, severity, preventability, and accountable medication, in surgical patients. Additionally, for preventable ADEs, caused by an error in the medication use process, the reviewers aimed to assess the medication error stages (e.g. ordering, transcribing, dispensing, administrating or monitoring errors). The secondary endpoint was to compare the occurrence of ADEs in surgical patients with nonsurgical patients. The occurrence of ADEs was presented separately for surgical and nonsurgical patients. Also an overall occurrence of ADEs was presented, this number included the entire study population of each study. The MINORS checklist developed by Slim et al. was used for the quality assessment [
9]. This checklist was developed to determine the methodological quality of non-randomized studies, for use in comparative studies, i.e. comparing two or more groups, and non-comparative studies. The MINORS checklist has been shown to have good inter-reviewer agreement, high test–retest reliability and good internal consistency [
9]. It has been externally validated for the ability to identify excellent trials, [
9] and has been previously applied in several systematic reviews [
10,
11]. This checklist consists of twelve methodological items including assessment of the risk of bias. For non-comparative studies eight items are scored, and four additional items for comparative studies. The items are scored on a 3-point scale ranging from 0 to 2; the ideal total score for non-comparative studies is 16.
The occurrence of ADEs was described as number of adverse drug events per 100 admissions, or, if the number of admissions was not available, the occurrence was stated per 1,000 patient days or as a percentage of patients with an ADE.
The nature of the ADEs, i.e. causality, severity, preventability, medication error stage and accountable medication, was described in total numbers and their percentage. When the differentiation of medication accountable for the ADEs included more than five medications, the top five most frequent were shown. The nature was presented for surgical and nonsurgical patients separately, if sufficient data were available. If separate data could not be extracted, the nature of the ADEs was presented for the entire population (overall).
Additional analysis
If the number of patient days and the mean length of hospital stay per patient could be extracted from the studies, the number of admissions was estimated by dividing the patient days by the mean length of stay. These data were used to calculate a mean occurrence of ADEs, using a random effects model to account for heterogeneity among studies. This method provided rough figures, but it gave a well founded impression of the difference in occurrences of ADEs between surgical and nonsurgical patients. Missing confidence intervals were calculated manually. The statistical significance of a difference between point estimates was judged in two successive steps by first examining whether either confidence interval contained the other point estimate. Because this approach tends to be anti-conservative, [
12] we subsequently compared the 83% confidence intervals of the seemingly significantly different point estimates from the first step. In case these confidence intervals overlapped, we decided for non-significance at the 0.05 significance level [
13]. The analyses were performed using Microsoft Excel 2003.
Discussion
With a structured literature search six articles were found that provided information on the occurrence and nature of ADEs in surgical patients. The reviewed articles showed that ADEs frequently occur in surgical patients, but were reported using various outcome measures, limiting interstudy comparison and pooling of data. Moreover, apart from preventability, the studies did not provide data that gave a clear insight into the nature of ADEs in surgical patients. For five studies a comparison of the occurrence of ADEs with nonsurgical patients was possible. Head-to-head comparison showed a significantly higher occurrence of ADEs in nonsurgical patients compared to surgical patients in three studies, which contrasts with the initial assumption of this review.
An estimated occurrence of ADEs of 8.5 per 100 surgical admissions, shows that medication related harm is a current problem in surgical patients. The proportion of preventable ADEs furthermore ranged from about a quarter to more than half of all ADEs, which indicates a serious health care problem that needs improvement. Unfortunately, data on the nature of these ADEs were lacking. Future research should focus more on a clear description of the nature of the events in order to better understand the underlying causes of ADEs and develop effective intervention strategies to reduce ADEs.
In five studies the occurrence of ADEs in surgical and nonsurgical patients could be compared head to head [
14‐
17,
19]. In all five studies that allowed comparison between surgical and nonsurgical patients, the occurrence of ADEs was higher in nonsurgical patients (significant in three out of five studies). The grounds of these differences are hard to determine, since detailed information on the surgical and nonsurgical patients was not provided in the included studies. For example information on known patient risk factors for ADEs such as gender, age, kidney function and number of comorbidities could clarify the variation in occurrence of ADEs [
7]. It was expected that the surgical intervention, including the associated patient handovers and medication changes, would have played a major role in the occurrence of ADEs in surgical patients. Current review does not provide information to justify whether the hospitalisation process is a factor in the occurrence of ADEs. From clinical perspective, it is likely that surgical patients overall are less ill than nonsurgical patients, and the admission is usually planned. Moreover, the key reason of admission for the surgical admission is the surgical intervention, whereas the nonsurgical admission usually requires multiple medication interventions to improve the clinical course of the disease.
The percentage of ADEs that was considered preventable varied substantially between the studies and only two studies assessed the preventability of ADEs in surgical and nonsurgical patient populations. Therefore, a solid conclusion on this subject could not be reached. A recent review on methods for assessing the preventability of ADEs also described a large variation of preventability percentages. They concluded that a large variation between studies was due to different instruments used for assessing the preventability [
23].
In studies including surgical patients, analgesics and antibiotics were most frequently accountable for ADEs, analgesics in 5-88% and antibiotics in 16-36% of the ADEs [
14‐
19]. Analgesics (pain management) and antibiotics (prophylaxis or treatment of wound infections) are drugs typically related to surgery. The range in percentage occurrence is wide making interpretation for clinical practice somewhat difficult. Nevertheless, these drug types have been previously identified as high risk for ADEs. A study on ADEs in 937 (surgical and nonsurgical) hospital admissions described analgesics as most frequently accountable for ADEs (26%), antibiotics accounted for 8% of the ADEs [
24]. Different numbers were reported in a study determining the nature of medication related adverse events, including 7889 (surgical and nonsurgical) patients. In that study, anticoagulants were most frequently accountable for the events (18%), analgesics in 5%, and antibiotics in 13% of the events [
25]. A systematic review on preventable ADEs described overdosage as the primary occurring error in analgesic use, resulting in excessive sedation, hypothermia and respiratory distress. Symptoms of a preventable adverse outcome after administration of antibiotics consist of an allergic reaction or rash [
26].
A strength of this review was the high methodological quality of the included studies, tested with the MINORS checklist. However, this systematic review has several limitations as well. The main limitation was the heterogeneity in endpoints used in different studies, being either number of ADEs per 100 admissions, or number of ADEs per 1,000 patient days, or percentage of patients with ADE. Due to this limitation, data could not be pooled, which limits comparison of occurrence of ADEs between surgical and nonsurgical patients. Although all studies identified ADEs with (at least) chart review, the occurrence of ADEs varied greatly (2–29 per 100 admissions). A recent review on ADE assessment methods has reported that direct observation identifies the most events, and voluntary incident reports the least, but does identify high severity events. Trigger tool methods are regarded as the most efficient method [
27]. In the present review, occurrence is higher in studies that performed direct observation, [
14‐
16,
19] than in studies that did not [
17,
18]. Furthermore, the chart reviewers in the various studies had different backgrounds and experiences. The reviewers consisted of nurses, nursing students, physicians or hospital pharmacists. Considering these results, it can be assumed that both the different ADE assessment methods and the backgrounds of the reviewers are significant drivers of the different (numbers of) ADEs identified. The causality assessment also varied greatly. This might be due to the different instruments used to determine the causality, or to the different backgrounds of the reviewers as previously described. Additionally, an aggregated mean occurrence of ADEs per 100 admissions in surgical and nonsurgical patients was calculated. But since the outcome measures varied, the number of admissions based on the length of hospital stay for two studies had to be manually calculated. The results of these means are therefore merely a rough estimation.
Secondly, a limited number of articles that fulfilled all inclusion criteria were found. A logical explanation is the limited amount of studies conducted on the occurrence of ADEs in surgical patients. However, another possible reason could be the search strategy used for this review. The search was performed on Medline and Embase and used explicit search terms to identify literature specifically on ADEs in surgical patients. The latter might have narrowed the view of all available studies. The studies were published in a broad period of time, between 1991 and 2011. In this time frame, it is likely that alterations have been made in medical care. For example, in pharmacological care, electronic prescribing has been introduced; this method reduces the risk of ordering and transcription errors and improves the detection of medication errors by hospital pharmacists.
The included studies described the definition of ADEs in different words, but they all amounted to the same: an injury resulting from medical interventions related to a drug. However, some issues must be noted. In three of the studies an ADE is defined as an event caused by medication in a therapeutic dose, [
16‐
18] and one of these studies explicitly included omission of a medication to the definition [
16]. The other three studies did not include dose restrictions to their definition, it was therefore unclear if omission of a drug was included [
14,
15,
19].
Little information on the was provided in the studies on the usage of medication safety interventions in the study hospitals, such as a computerized ordering system, was provided. We can therefore not explore if such interventions may explain the differences in occurrence of ADEs in the included studies. Customized information technology can reduce the rate of medication errors in hospitals [
28]. Bar-code verification technology with an electronic medication administration system (bar-code eMAR), active participation of an on-ward hospital pharmacy or CPOE are promising strategies [
29‐
32]. No studies on the effect of customized information technology on (preventable) ADEs have been performed in surgical patients. The recent focus of improving safety in surgical patients has been on reducing errors in the perioperative pathway by a system approach using checklists [
33]. For example, a substantial improvement in surgical safety has been achieved by the implementation of the comprehensive checklist, the SURgical Patient Safety System (SURPASS), which reduces the in-hospital mortality by half (from 1.5 to 0.8 per cent) [
34]. To provide optimal care, it is important to gain more insight into the most common medication-related adverse events as well. By implementing a targeted intervention strategy to prevent ADEs, using a system approach focussing on the entire surgical pathway, safety in surgical patients can be improved even more [
35].
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
EB, MdB, MD and MB developed the review protocol. EB and MdB performed searches, selected the studies and extracted the data. EB and MD performed the statistical analysis and interpretation of data. MB and LL are the study’s principal investigators, and responsible for data interpretation and manuscript content. EB drafted the manuscript, all others contributed in reviewing the manuscript. All authors read and approved the final manuscript.