Erica Burry and Angela Punnett contributed equally to this work.
Antibiotic administration within 60 minutes of presentation for medical care may be used as a treatment target for febrile neutropenia (FN); however, anecdotal evidence suggests this target is often missed. Few studies have examined the prevalence or causes of delay. We describe the median time to antibiotic administration at our institution, predictors of delay, and barriers to prompt administration to inform quality improvement strategies.
A random sample of 50 episodes of FN presenting to the emergency department (ED) between 2008 and 2009 were reviewed. Times between triage, MD assessment, lab results, and antibiotic administration were recorded. Patient and ED variables were examined as possible predictors of delay. In parallel, lean methodology was used to identify system inefficiencies. A trained moderator conducted group interviews with interdisciplinary representatives involved in the emergency care of neutropenic patients to identify process barriers to prompt antibiotics.
The median time from triage to antibiotics was 216 minutes (interquartile range [IQR] = 151–274 minutes). The greatest delay occurred following the reporting of lab results (152 minutes, IQR = 84–210 minutes). Only fall season predicted a longer time to antibiotics (P = 0.03). The lean process identified unnecessary areas of delay between departments.
Febrile neutropenia (FN) is a common treatment complication in pediatric oncology patients, with a low but identifiable mortality rate 1. Prompt administration of broad-spectrum antibiotics has been considered standard of care for FN for decades 2. However, most guidelines do not define prompt administration. The most recent guidelines published by the Infectious Diseases Society of America briefly cite administration within 2 hours of presentation as ideal, but do not discuss this in detail 3. While evidence is lacking to support a particular target, some institutions have implemented a target as low as 30–60 minutes from arrival 4, 5.
Though the literature has not definitively correlated a delay in antibiotics to adverse outcomes in children with FN, such an association has been shown in other clinical situations 6. Battleman et al. 6 showed a correlation between the timing of antibiotic administration and the length of stay in hospital in adults with community-acquired pneumonia, suggesting that antibiotic delay may lead to increased morbidity and financial expenditure. Rivers et al. 7 showed better outcomes and improved survival with early goal-directed therapy for adults with severe sepsis and septic shock. The Surviving Sepsis Campaign advocates for antibiotic initiation within an hour for sepsis requiring intensive care 8, 9. Although these studies were not conducted in oncology patients, each advocates for early treatment of infectious disease. In the absence of an accepted standard for time to antibiotic administration in the setting of FN, a target of 60 minutes may be ambitious but is not unreasonable and not without precedent.
Very few studies have examined the prevalence and predictors of antibiotic delay in pediatric FN. Such information is necessary to design appropriate interventions. Our primary objectives were therefore to describe the median time to antibiotic administration at our institution and to identify barriers to prompt administration. Our secondary objective was to determine predictors of delayed antibiotic administration.
A mixed methodology approach was used to investigate the delays in and barriers to antibiotic administration in pediatric oncology patients with FN in the emergency department (ED). A chart review of oncology patients presenting to the ED was conducted to determine the prevalence and predictors of antibiotic delay. Concurrently, focus groups including representatives from all disciplines involved in FN management were completed using lean methodology to identify barriers to prompt treatment.
As per institutional guidelines, quality improvement projects are reviewed and approved by the institution's Quality and Risk Management Department, ensuring compliance with institutional policies and ethical guidelines for projects defined as quality improvement. As the collection of patient data was limited to retrospective chart review, informed consent was not required.
A chart review was completed using a random sample of 50 episodes of FN in the ED over a 1-year period. Inclusion criteria included any patient with a known diagnosis of malignancy aged 0–18 years who presented to the ED with a complaint of fever and who subsequently was found to meet the local definition for FN. An ED electronic patient tracking system was used to identify all such patients who presented between December 1, 2008 and November 30, 2009. Fever was defined as an oral temperature of ≥38°C or an axillary temperature of ≥37.5°C. Neutropenia was defined as an absolute neutrophil count (ANC) ≤0.5 × 109/L. Patients appearing toxic receive antibiotics regardless of ANC, and were therefore also included. According to local protocol, all patients meeting FN criteria receive broad-spectrum antibiotics and are admitted.
Eligible patients were stratified by season by date of presentation: December 1 to February 28 (winter), March 1 to May 31 (spring), June 1 to August 31 (summer), and September 1 to November 30 (fall). Thirteen charts were randomly selected from each of the winter and fall groups, while 12 charts were randomly selected from each of the spring and summer groups, for a total of 50 charts. Only one episode per patient was used in a given season.
The time of triage, initial medical assessment, entry of bloodwork orders, receipt of bloodwork, reporting of lab results, entry of antibiotic orders and first antibiotic administration were recorded for each episode. Possible predictors of delayed administration were also recorded, including both patient-level (gender, age, and diagnosis), and episode-level (season, time of presentation, febrile at presentation, training level of first assessing physician, and use of ED-stocked antibiotics) variables. The definition for season can be found above. Time of presentation was subdivided into day, evening, and night, with day defined as 8:00 to 17:00, evening as 17:00 to 23:00, and night as 23:00 to 8:00. The training level of first assessing physician was subdivided into junior versus senior, with junior defined as a medical student, first or second year pediatrics resident and senior defined as a third or fourth year pediatrics resident, fellow or attending physician.
Median times between events were calculated. The episodes were then categorized into below median or above median time to antibiotic administration, using the median time from triage to antibiotic administration (216 minutes) as the cut off between groups. The median time was used as the cut off for the purposes of statistical analysis rather than the 60-minute benchmark described in the introduction, as none of the 50 episodes reviewed in this study received antibiotics within 60 minutes. Logistic regression was used to examine predictors of delayed antibiotic administration. Statistical analyses were performed using SAS-PC software (version 9.2; SAS Institute, Cary, NC). Statistical significance was defined as P < 0.05.
Independently, lean methodology was used to conduct group interviews made up of representatives from all disciplines involved in the care of oncology patients with FN. Lean principles were first applied in the automotive industry to streamline manufacturing processes, and have since been adapted for use in health care to identify system inefficiencies. The main goal of lean methodology is to distinguish value-added and non-valued-added tasks in order to identify and eliminate waste 10–12. This allows front-line workers to improve flow and better serve customers' needs by ensuring that every step adds value to the process. In health care, waste is described as an action that is not required for patient care, decreasing quality of care and increasing wait times 10–12. Lean methodology focuses on bringing front-line staff together to identify areas of waste using value stream mapping 12, 13. Group participants outline step-by-step the events that occur in the care of a patient. These events are the actual steps taken on a daily basis, and not what ideally should occur. Once this has been completed, members then identify actions that do not add value to the patient's care. Interventions are put into place to eliminate the identified waste to streamline patient care, improve patient experience and decrease wait times 12, 13. The lean process has the very specific purpose of streamlining a process without additional resources, and aims at producing tangible and feasible results 12.
In this study, a moderator trained in lean techniques (A.M.) conducted interviews with a group consisting of ED and oncology nurses, ED and oncology staff and trainee physicians, laboratory personnel, and pharmacy personnel. The group first developed a flow diagram outlining the events that occur in the treatment of FN from triage to antibiotic administration. Consensus was then sought on barriers to prompt treatment. Finally, possible interventions to improve delays caused by these barriers were discussed. The results of the chart review were discussed throughout the lean process.
One hundred seventy-nine episodes of FN among known oncology patients presented to the ED during the study period. Fifty-three presented in winter, 39 in spring, 42 in summer, and 54 in fall. Demographic and episode-level information for the 50 charts selected for review are seen in Table I. The median time between triage and antibiotic administration was 216 minutes (interquartile range [IQR] 151–274 minutes). The longest delays occurred between the lab report of neutropenia and antibiotic administration, with a median time of 152 minutes (IQR 84–210 minutes). Other end points, such as timing of medical assessment, bloodwork orders, and antibiotic orders were also investigated, but were not reliably noted in patient charts. It was felt that time periods that included these end points could not be measured with confidence.
Table I. Characteristics of Study Cohort
Cohort (n = 50)
IQR, interquartile range; MD, medical doctor; N, number.
Gender, N (%)
Age (years), median (IQR)
5 (3, 12)
Diagnosis, N (%)
Season, N (%)
Time of day, N (%)
Febrile at presentation, N (%)
Training of first assessing MD, N (%)
Ward stock antibiotics, N (%)
The results of univariate analyses of predictors of delayed antibiotic administration are presented in Table II. Presentation to the ED during the fall season was significantly associated with delayed antibiotic administration (odds ratio [OR] 10, 95% confidence interval [CI] 1.6–63, P = 0.03). No other variable examined significantly predicted delay. When a sensitivity analysis was run excluding the fall season, the median time to antibiotics was 199 minutes (IQR 142–257 minutes); no variable examined significantly predicted delay.
Table II. Univariate Analysis of Predictors of Delayed Antibiotics
CI, confidence interval; MD, medical doctor; OR, odds ratio.
Age (per year)
Time of day
Febrile at presentation
Training of first assessing MD
Ward stock antibiotics
The focus group consisted of eight individuals representing the groups noted above. The interviews first resulted in the development of a flow diagram outlining each event that occurred in FN management from triage to antibiotics, illustrated in Figure 1. This diagram outlined the actual steps that occurred, despite any problems that may have been associated with an individual step or the order of steps.
During the process of developing the flow diagram, a number of assumptions and misunderstandings were uncovered. ED personnel incorrectly assumed that bloodwork sent to the lab from the ED was automatically considered to be STAT and, in some cases, were not labeling them as such. Lab personnel were therefore not processing these samples immediately. In addition, ED personnel were unaware that different sizes of specimen vials had different processing times. Although larger tubes allowed for faster processing times, smaller tubes were frequently used to minimize the amount of blood drawn. Though inpatient wards used a computerized system to print orders on colored paper for easy identification of urgent requests, written antibiotic orders from the ED were sent to the pharmacy as paper orders. Consequently, pharmacy personnel were not able to easily identify the urgency of these orders.
Additionally, ED nursing personnel identified that on parental request, bloodwork was often delayed until after topical anesthetic had been applied and taken effect. Very few families had applied their own topical anesthetic prior to arrival.
Finally, differences in opinion on the most appropriate timing to access the implanted central venous catheter (port) were uncovered. Hospital protocol dictated that both peripheral and central blood cultures be drawn. However, due to patient comfort, time management, and often parental request, ED nurses preferred to wait for peripheral bloodwork results before accessing the port. ED personnel reported that patients who were not neutropenic could have their port accessed for cultures and then the access needle could be immediately removed, thus decreasing patient and parent anxiety and avoiding the discomfort associated with dressing removal.
Based on these results, a number of interventions were developed to overcome the identified barriers. These interventions are indicated in boxes labeled interventions in Figure 1. The first intervention was the development of a pre-printed order sheet for oncology patients with fever. This will standardize the process of collecting peripheral and central bloodwork and blood cultures of any oncology patient presenting with a fever. The pre-printed order sheet will also have a section with the recommended antibiotics and dosing to facilitate ordering.
Other interventions focused on bloodwork and antibiotic ordering. All bloodwork from patients with suspected FN will be labeled as STAT. As well, larger tubes will be encouraged. Non-ward stock antibiotic orders will continue to be sent as paper orders, but marked to indicate STAT to pharmacy personnel. Finally, targeted nursing education will occur. One area of focus will be that of central access; ports will be accessed with peripheral bloodwork and before lab results are reported. This will now be mandated by the pre-printed order sheet. Parents will also be instructed to use their own supply of topical anesthetic before arrival.
We demonstrated that the median time from arrival to antibiotic administration in our institution was close to 3.5 hours. Interestingly, in many cases, a large portion of this wait occurred after the lab report of neutropenia. As all neutropenic patients receive antibiotics and are admitted in our institution, a median time of 2.5 hours was spent waiting for treatment after it was determined to be indicated.
Only season predicted differences in treatment times, with the longest times occurring during the fall. However, the study period coincided with the H1N1 outbreak in the fall of 2009. While treatment times during this period may not be representative of those of an average fall season, excluding children who presented during the fall did not have a major impact on the analysis. It is relevant to note that periods of increased ED volume may be associated with longer treatment delays. It has been previously shown in the literature that influenza season is associated with increased length of stay in the ED, increased admissions, and increased numbers of patients leaving without being seen 14.
Although FN treatment delay has received little attention in the literature, some investigators have found similar delays and have developed effective means to improve treatment times. One study conducted among adult oncology patients found a median time of 210 minutes between presentation and antibiotic administration; interestingly, patients with more co-morbidities had longer delays 4. In the pediatric setting, Corey and Snyder 5 found that average waiting times for antibiotics ranged from 80 to 120 minutes; no one met their aggressive target of 30 minutes. Through various multi-disciplinary educational initiatives, they were able to decrease average wait times to 25 minutes, with 80–100% of patients meeting the 30-minute target. While children in this study were known to be neutropenic prior to arrival and were evaluated and treated on the pediatric ward instead of the ED, limiting generalizability to our institution, the positive results nonetheless support the use of infrastructural and educational interventions.
During focus group interviews, it became apparent that the flow of events was not optimal and that identified misunderstandings and assumptions were impeding prompt care. As a result, the potential interventions noted above were developed through consensus from these interviews. For example, the previous practice of accessing ports only after bloodwork results were available was instituted by ED personnel in order to prevent unnecessary taping and patient discomfort, in keeping with ED and institutional philosophies of family-centered care to minimize and manage painful procedures. Education of ED and oncology staff and of parents on early application or home application of topical anesthetic will allow for port access prior to obtaining bloodwork results without compromising patient comfort. This situation represents the kind of issue that can only be identified by combining analysis of quantitative data with the qualitative input of front-line workers.
At a time of cost-consciousness in healthcare, improving care through strategies requiring increased resources is not always feasible. Lean methodology provides an alternative solution by focusing on improving process efficiency and minimizing waste 10–13. The interventions that have been developed in this study focus on changing the order and flow of events that occur in FN treatment. Implementing these interventions should result in a decrease in treatment times without increased resources.
This study has a number of implications for the future. The ED electronic patient tracking system used to identify cases of FN in this study will continue to be used to prospectively monitor antibiotic administration times. This information will also be used in future studies to assess the effectiveness of the proposed interventions, and to identify future targets. Future value stream mapping to continuously improve the care provided to FN patients is planned. Indeed, repeated PDSA (Plan-Do-Study-Act) cycles with ongoing monitoring of treatment times allow for continuous improvement 11.
Several limitations deserve mention. With a sample of only 50 episodes of FN, our power to meet our secondary objective of determining predictors of delay was limited. However, this was primarily a quality improvement study designed to determine treatment times and describe the lean process used to identify barriers and potential interventions. Future studies with improved power to evaluate predictors of delays are warranted. The chart review component of this study was a retrospective analysis, and was therefore limited to the data present in the electronic medical record. Time periods that were recorded by hand rather than computer, such as the time of medical assessment and time of bloodwork and antibiotics orders, were not regularly available. We were not able to evaluate time periods between each step in FN patient care as initially intended due to gaps in data. Thus, we could not determine how much of the delay after the reporting of neutropenia was due to delays in physician ordering. There was, however, complete data on time of antibiotic administration, allowing us to fulfill our primary objective of determining median treatment times. Finally, the group interviewed during the lean focus group consisted of a small number of individuals. However, while initial input was solicited from these individuals alone, the proposed interventions were taken back to and approved by the groups they represented.
In conclusion, the median treatment time for FN in pediatric oncology patients at our institution was 216 minutes, which is likely to be longer than ideal even in the absence of evidence for a standard target. Lean methodology identified areas of waste in order to develop educational and infrastructural interventions, and was informed by a chart review documenting where the greatest periods of delay occurred. Ongoing monitoring of treatment times will allow for continuous quality improvement to occur. The methodology used in this project can act as a template for other institutions wishing to identify their own site-specific barriers to prompt treatment, while the interventions developed will be evaluated in future studies for their effectiveness on decreasing delays within this institution. This quality improvement study provides a mixed methodology model for improving process efficiency and developing “best-practice” targets in the treatment of FN.
This study was funded by an education grant from the C17 Council. The funding agency had no role in the conduct of the study or the writing of this manuscript. The authors would like to acknowledge Ms. Marie Pinard for her thoughtful review of the project and manuscript. The authors would also like to acknowledge Dr. Stephen Porter and Dr. Jim Whitlock for their comments.