Abstract
Abstract
Background:
Surgical antibiotic prophylaxis is frequently reported in the literature to be suboptimal, a finding having both clinical and public health implications. This study aimed to calculate rates and patterns of adherence to guidelines at two sites and identify extrinsic contributing factors.
Methods:
A retrospective analysis was conducted over two 12-mo periods during 2013–2014 at the metropolitan Blacktown Hospital and regional Lismore Base Hospital, New South Wales, Australia. A group of 400 patients undergoing abdominal general surgery was selected via simple random sampling (n = 200 per site). Medical records were reviewed, and prophylactic antibiotic regimens were compared with the Australian guideline, Therapeutic Guidelines: Antibiotic (v. 14) with respect to drug choice, dosage, timing of administration, and duration of administration.
Results:
The overall rate of adherence to the guidelines was 16.5% at Blacktown Hospital and 19.5% at Lismore Base Hospital. At each site, prophylaxis was administered to more than 95% of patients and was inappropriately withheld in 4%. Drug choice was the most frequent error type, specifically involving inappropriate omission of metronidazole and use of newer-generation cephalosporins. Errors in the timing of administration also were frequent, with prophylaxis typically occurring excessively early. Logistic regression identified emergency surgery as independently associated with prophylactic errors in both the Blacktown Hospital (p < 0.001) and the Lismore Base Hospital cohorts (p = 0.020).
Conclusions:
Adherence to antibiotic prophylactic guidelines was poor at both the metropolitan and regional sites. Choice of antibiotic and timing of administration were identified as major error types. Consideration should be given to multidisciplinary involvement of anesthetists, implementation of focused interventions with an emphasis on emergency settings, and further research correlating antibiotic use with clinical significance.
I
Several reference guidelines have been produced to provide consistent, judicious, and evidence-based prophylactic regimens that consider local patterns of common bacteria and antibiotic resistance. Such guidelines, including the Australian Therapeutic Guidelines: Antibiotic [4], are updated regularly to respond to new evidence, and clinicians are expected to maintain an evidence-based practice.
Hospitals conduct audits of many fields, one of which is rates of appropriate chemoprophylaxis in surgery. A survey of surgeons in Queensland, Australia, found a significant lack of knowledge regarding prophylactic guidelines [5]. Further, the participants suggested that for reasons including methodology and a lack of standardized measurement, there were inaccuracies in hospital-reported outcomes data, and thus, the true extent of non-adherence remains unclear [5].
Internationally, published audits with sound methodology demonstrate a great range in rates of adherence to guidelines. One key review of 25 publications reported a range between 0–71.9%, with the majority below 50% [6]. Although sparse and predominantly from Victoria, studies from Australia demonstrate a similar pattern, with adherence rates ranging from 8–53.3% (7–9). Further analysis demonstrated that although most patients requiring antibiotic prophylaxis receive it, many are unnecessarily prescribed prophylaxis (7,10). The literature therefore suggests that although the adherence rates are substandard, errors in prophylaxis trend toward over-prescription.
Inappropriate over-administration of prophylactic antibiotics has two main consequences. The most immediate and direct is the avoidable risk of side effects and adverse reactions. The second is contributing unnecessarily to the development of antibiotic resistance. In a published consensus statement from more than 20 years ago, the Journal of the American Medical Association identified optimized antibiotic prophylaxis as a pillar in the global attempt to prevent resistance (11). This contention is still relevant in today's practice, and clinicians have an obligation to consider this in their prophylactic prescribing.
The rates of adherence to prophylactic guidelines at Blacktown Hospital (BH) and Lismore Base Hospital (LBH) in New South Wales (NSW), Australia are unknown, having never been published. Given the importance of optimizing antibiotic prescribing, this study aimed to quantify and characterize patterns of adherence in these two geographically and clinically different settings and to use regression modeling to identify any relevant contributing factors related to errors in prescribing. Comparison between sites was not a desired outcome, and the structure and reporting of the study reflect this.
Materials and Methods
Setting
This retrospective audit was conducted at two hospitals in NSW, Australia. Blacktown Hospital is a metropolitan 400-bed university teaching hospital providing both elective and emergency surgical services across a wide range of disciplines, including general surgery. More than 9,300 surgical procedures are performed annually, servicing a population of more than 260,000 within Western Sydney. Lismore Base Hospital is a regional hospital with a 260-bed capacity, servicing more than 170,000 people in Northern NSW. More than 8,100 surgical procedures are undertaken within the hospital annually, encompassing primarily general, orthopedic, vascular, and gynecologic operations.
Data collection
The study had two sites for data collection, each having a defined time period: BH (1 July 2013–30 June 2014) and LBH (1 January 2014–31 December 2014). In order to meet the needs of a planned subanalysis of data, the BH cohort excluded operations performed between 1 January 2014 and 13 February 2014. All abdominal general surgical procedures performed at the stated sites and within the stated time periods were eligible for inclusion.
There were 1,338 cases from BH and 651 cases from LBH that fulfilled these criteria, and they were compiled in a database. A total of 200 study cases were selected from each cohort via simple random sampling without replacement, utilizing random number generation. Because of the requirements of the planned subanalysis, the BH cohort was split into two subgroups (1 July–31 December 2013 and 14 February–30 June 2014), and 100 cases were selected from each subgroup.
Data were sourced retrospectively from the surgical, anesthetic, and medication records, including electronic and paper documentation. The data recorded were patient demographics (date of birth, gender), surgical characteristics (operation, date and time of procedure, duration, urgency classification), and antibiotic details (drugs and dosages administered, timing, duration of administration, allergies if applicable).
Data evaluation
Prophylactic antibiotic regimens were compared with the widely accepted Australian guideline, Therapeutic Guidelines: Antibiotic v 14 [4]. Cases were considered for adherence to guidelines with respect to four key parameters: Drug choice, dosage, timing of administration, and duration of administration [12]. Redosing was considered necessary when the duration of the surgery exceeded two half-lives of the drug administered, as per the guideline [4]. Where no prescription was recorded, it was assumed that no prophylaxis had been given.
Non-adherence was determined by the presence of one or more errors in prophylaxis, and this was reflected through the sums of error types. Where appropriate, errors were further categorized to identify specific areas of non-adherence.
Statistical analysis
The SPSS v. 21 program (IBM Inc., Chicago, IL, USA) was used to perform all statistical calculations. The study was powered to meet the needs of this analysis as well as a subanalysis of the BH data. Power calculations for the subanalysis used standard parameters, with α = 0.05 and β = 0.2, indicating that a total sample size of 200 was needed. This sample size also was found to have sufficient power for the major arm of the project reported here.
Frequencies and percentages are reported for categorical data, whereas means and standard deviations are reported for continuous data. Comparisons of categorical data were calculated using either χ2 analysis or the Fisher exact test. Comparisons of continuous data were calculated using Student's t-test. Unadjusted odds ratios and binary logistic regression modeling were used to determine the impact of contributing factors. Probability values are not reported extensively, as this study was not primarily comparative and rather aimed to identify trends. Where relevant, significance for analyses was considered when p < 0.05.
Ethics
This study was subject to ethical review and governance under the Western Sydney Local Health District Human Research Ethics Committee.
Results
The medical records of 400 surgical patients were reviewed, 200 each from BH and LBH. One case in the Lismore group did not have timing data recorded and was not included in the analysis of timing of administration. No cases were excluded after randomized selection.
Full descriptive data for both sites are shown in Table 1. Cohorts were similar with respect to gender and age. Laparoscopic procedures were the most frequent type of procedures performed in both groups; however, they were significantly more common in the BH cohort (BH = 85.5%; LBH = 63%; p < 0.01). Laparoscopic cholecystectomy (BH = 39%; LBH =30%) and laparoscopic appendectomy (BH = 28.5%; LBH =21%) were the two most common procedures at both sites. Emergency surgical cases were significantly more frequent at BH (p = 0.016), whereas the mean surgical duration was significantly longer at LBH (p = 0.013).
More than 95% of all patients at both sites received a prophylactic antibiotic regimen. There were 33 cases (16.5%) at BH and 39 cases (19.5%) at LBH that were adherent to the standard prophylactic guidelines. Of the four cases in which prophylaxis was not required, two appropriately received no antibiotics (BH = 1; LBH = 1), whereas two were inappropriately given prophylaxis (BH = 0; LBH = 2). At each site, eight patients (4%) did not receive pre-operative antibiotics despite the guideline recommending prophylaxis. Table 2 summarizes the adherence data into broad error types.
Specific surgical procedures regardless of approach were associated with variable rates of adherence to prophylaxis guidelines. Hernia repairs had the highest rates of adherence in both groups (BH = 68.6%; LBH = 59.6%). Of the most frequent procedures, both appendectomy (BH = 5.1%; LBH = 8.9%) and cholecystectomy (BH = 1.3%; LBH = 4.8%) had poor rates of adherence.
Categories of adherence
The frequency of error for each category of adherence is reported alongside proportions representing comparison with all cases having errors (BH = 167; LBH = 161). These data are shown in Table 3. Timing errors in the LBH group are the only exception to this, where one case had insufficient data. Each case may have more than one contributing error type, and as such, the total number of errors will exceed the number of error cases.
Percentages represent data as a proportion of all cases with errors.
For timing error data in the Lismore Base Hospital cohort (n = 160).
Drug choice was the most frequent error, accounting for 231 of all error cases across both groups (BH = 71.3%; LBH = 69.6%). Specifically, the two most frequent errors in drug choice were inappropriate exclusion of metronidazole (BH = 33.5%; LBH = 27.3%; p = 0.16) and inappropriate substitution of an older-generation cephalosporin by a newer one (BH = 22.2%; LBH = 47.9%; p < 0.01). Table 4 lists the antibiotic agents used at each site and demonstrates that cephazolin and metronidazole were the most common drugs prescribed at both sites.
Errors in dosage were infrequent, although significantly more common in the LBH cohort (BH = 1.2%; LBH = 9.9%; p < 0.01).
Timing of administration produced the second most common error at both sites, occurring in 84 cases (50.3%) at BH and 80 cases (50%) at LBH. Of these, excessively early administration (BH = 37.1%; LBH = 32.5%) was more common than late administration (BH = 13.2%; LBH = 17.5%) at both sites.
In the BH group, the duration of antibiotic administration (hh:mm ± standard deviation [SD]) was prolonged in 20 cases (12%) by an average of 50:24 ± 33:51, while in the LBH group, 48 cases (29.8%) were prolonged by an average of 64:00 ± 43:32.
Redosing of prophylactic antibiotics was necessary in 29 cases across the total cohort but did not occur in 13 cases in each group (BH = 7.8%; LBH = 8.1%). The remaining cases all involved LBH, where redosing was performed too late in two cases and appropriately in one case.
Analysis of contributing factors
Within the BH cohort, emergency surgery was found to be more than 16 times more likely to result in prophylactic errors than was elective surgery (unadjusted odds ratio [OR] 16.09; 95% confidence interval [CI] 4.72–54.90). Similarly, associations with prescribing errors were found when comparing laparoscopic approaches with open procedures (unadjusted OR 2.75; 95% CI 1.12–6.76) and operations performed after hours with those during normal hours (unadjusted OR 4.27; 95% CI 1.25–14.65). Following logistic regression modelling (χ2 = 37.89; Nagelkerke R2 = 0.292), emergency surgery was still found to be independently associated with errors in antibiotic prophylaxis (p < 0.001), whereas the difference in prescribing for the laparoscopic approach was non-significant (p = 0.050).
Analysis of the LBH group was similar, with emergency surgery being more than three times more likely (unadjusted OR = 3.30; 95% CI = 1.43–7.61) and a laparoscopic approach being more than two times more likely (unadjusted OR = 2.38; 95% CI = 1.17–4.83) to result in a higher error rate. Operations performed after hours demonstrated no significant difference from those performed during normal hours (unadjusted OR = 1.55; 95% CI = 0.66–3.62). Logistic regression modelling (χ2 = 13.23; Nagelkerke R2 = 0.102) maintained the independent significance of emergency surgery (p = 0.020) with respect to prophylactic errors.
When combining data from both cohorts and applying logistic regression modelling (χ2 = 44.10; Nagelkerke R2 =0.171), emergency surgery was a significant independent predictor of errors in antibiotic prophylaxis prescribing (p < 0.001). The effect of the laparoscopic approach was non-significant (p = 0.055).
The full results from the logistic regression modelling are shown in Table 5.
Dependent variable = binary antibiotic prophylaxis guideline adherence (adherent/non-adherent).
Discussion
Surgical site infections are one of the most common nosocomial infections [13] and can almost quadruple the risk of surgical mortality [14]. Antibiotic prophylaxis in surgery has strong evidence supporting the reduction of infection-related morbidity and mortality rates of surgery [15]. Good prophylactic agents traditionally have been identified as safe and effective bactericidal drugs with long half-lives and low cost [16]. However, the clinician's decision now extends past balancing the risk of infection and its associated morbidity with the potential side effects of antibiotics. The risk of antibiotic resistance is an equally important consideration in prescribing, and active identification of practice contrary to evidence-based guidelines is warranted.
This study demonstrated poor rates of adherence to the prophylactic guidelines, with overall administration of prophylaxis being appropriate in just 16.5% of cases at BH and 19.5% of cases at LBH. Although this low degree of adherence is clinically concerning, it is not unexpected. The literature demonstrates similar results, with a 2012 review finding that the majority of studies report rates of adherence below 50%, with some values below those described here [6].
However, studies also attribute a large proportion of errors to inappropriate over-prescription [10,17]. A report involving a large public hospital in Greece demonstrated that 19% of patients received prophylaxis inappropriately, with no clinical indication [10]. In contrast, this study identified low rates of over-prescription, involving only 1% of cases at LBH. Part of this difference may be explained by a lack of clarity in the Australian guidelines when defining the types of cases not requiring prophylaxis, resulting in under-estimation of the true figures [4]. However, the data also show that prophylaxis rarely is withheld inappropriately, suggesting that antibiotic resistance, rather than surgical site infection, is the more likely potential outcome of the prophylactic patterns at these sites.
The errors therefore lie within the prophylactic antibiotic regimens themselves; patients are not receiving the correct antibiotic, at the correct time, for the correct duration. Four key parameters were used to evaluate prophylactic regimens, namely the choice of antibiotic agent, dosage, timing of administration, and duration of administration. The choice of antibiotic agent was identified as the most frequent cause of prophylactic error, occurring in 71.3% and 69.6% of error cases at BH and LBH, respectively. This is consistent with the literature, which often reports a high proportion of errors in drug choice [6,18,19].
Of particular concern was the frequency with which the third-generation cephalosporin, ceftriaxone, was used in place of the first-generation cephalosporin, cephazolin. This occurred in 22.2% of error cases at BH and 47.9% at LBH. This finding parallels the current literature, notably in a large American study (total n = 517) that reported this same drug error in 41% of the 179 colorectal surgery cases [6,20]. Ceftriaxone has a different spectrum of activity and is not superior in standard surgical prophylaxis; and its excessive use is contributing to antibiotic resistance [21,22]. This inappropriate choice of drug implies a lack of clinician understanding and may be an effective target for educational interventions.
Extended duration of administration and incorrect timing of administration were two other common error types observed in this study. Guidelines clearly recommend that single doses of antibiotic prophylaxis are appropriate unless the operation extends beyond three hours. Similarly, the clear recommendation is that antibiotics be administered within 60 min prior to the first incision. Both of these error types have been discussed extensively in the literature and are often found to be common in prophylaxis [6–8,10,19,20,23–25].
Monotherapy regimens have been perceived as having superior adherence rates, presumably because of the simplicity of the guideline [20]. Mesh hernia repair is the only operation included in this study in which antibiotic monotherapy is recommended by the Australian guideline [4]. The high rate of adherence in hernia repairs observed at both sites may provide further support for this suggestion. This observation also may be explained by the serious morbidity of infected prosthetic mesh, leading surgeons to be especially diligent in their prescribing for this procedure.
Laparoscopic cholecystectomy was the second most common procedure in the study, and 59 of 138 cases received cephazolin monotherapy, without the recommended metronidazole. The updated version of the Therapeutic Guidelines: Antibiotics (v 15) was released at the end of the study period [26]. Although not relevant for determining adherence in the study, the new guideline for prophylaxis in laparoscopic cholecystectomy is cephazolin monotherapy, albeit at a larger dose of 2 g [26]. It is unclear whether the pattern of prophylaxis observed was attributable to clinicians pre-empting the guideline change, changing clinician practice informing the guideline change, or entirely chance.
The use of logistic regression modeling identified emergency surgery as an independent predictor of prophylactic error in both cohorts. This is not surprising, given the complexity and urgency of such cases, and has been identified as an issue in the literature [20,27]. However, this is a population that generally will be less physiologically able to deal with an infective complication. Rather than accepting this observation, it is important that this group be identified and strategies devised to improve practice. In addition, a laparoscopic approach, when compared with open surgery, approached significance in the BH cohort as a predictor of prophylactic error (p = 0.050), and consideration of further investigation may be justified. It should be noted that only a selection of the potential confounders were included in the modeling, and that the results may be skewed by an unidentified confounder.
This study had limitations that must be taken into consideration when interpreting the results. First, the retrospective design is dependent on the accurate documentation in patient records by relevant clinical staff, notably the anesthetists who administered the prophylaxis. Second, the incorporation of a planned subanalysis impacted the case selection process for the BH cohort. Sampling the cohort as two subgroups has the potential to skew the cases included in the study. Further, the cohort was sampled from 10.5 months of surgical patients, rather than the desired 12-month period from which the LBH cohort was sampled. However, these limitations should not significantly impact the results, and the methodology remains sufficient, allowing the study to present an accurate sample with meaningful data from both sites.
The results of this study indicate that adherence to the guidelines in the surgical prescribing of prophylactic antibiotics continues to be poor and signifies areas for future research and improvement to clinical practice. First, further research on the issue is indicated nationally and internationally, including at each of the study sites in our study. Correlating these results with data on surgical site infections would allow evaluation of the absolute clinical impact of the poor adherence reported both here and worldwide. Second, in light of the poor prophylactic practices at each of our sites, consideration should be given to targeted interventions aimed at improving the specific issues in prophylaxis that were identified. These could include focuses on drug choice and timing of administration, with an emphasis on emergency surgery. Third, few studies have involved Australian hospitals. This study provides Australian clinicians and administrators with a unique insight into the local situation, from which improvements to practice can originate.
Furthermore, broadening the general direction of research and interventions in this field may be justified. When reviewing the literature, it is clear that administration of antibiotic prophylaxis contrary to the relevant guidelines is a problem that has existed for many years. Despite much scientific attention and institutional intervention, the systemic issue remains largely unchanged. There appears to be good surgical engagement on the issue, with much of the research and discussion being driven by surgeons. However, the notable absence in the discussion is anesthetists, who commonly administer antibiotic prophylaxis autonomously on the surgeons' behalf. Although this is a complex and multifactorial issue, it seems wise for anesthetists to join surgeons and become a target for clinician education programs and interventions.
Author Disclosure Statement
The authors report no conflicts of interest, financial or otherwise.