Customize your JAMA Network experience by selecting one or more topics from the list below.
Tamma PD, Avdic E, Li DX, Dzintars K, Cosgrove SE. Association of Adverse Events With Antibiotic Use in Hospitalized Patients. JAMA Intern Med. 2017;177(9):1308–1315. doi:10.1001/jamainternmed.2017.1938
What is the likelihood of developing antibiotic-associated adverse drug events (ADEs) for hospitalized patients receiving antibiotic therapy?
In this cohort study, medical records of 1488 adult inpatients were examined for 30 days after antibiotic initiation for the development of the following antibiotic-associated ADEs: gastrointestinal, dermatologic, musculoskeletal, hematologic, hepatobiliary, renal, cardiac, and neurologic; and 90 days for the development of Clostridium difficile infection or incident multidrug-resistant organism infection. Twenty percent of patients experienced at least 1 antibiotic-associated ADE.
These findings underscore the importance of judicious antibiotic prescribing to reduce the harm that can result from antibiotic-associated ADEs.
Estimates of the incidence of overall antibiotic-associated adverse drug events (ADEs) in hospitalized patients are generally unavailable.
To describe the incidence of antibiotic-associated ADEs for adult inpatients receiving systemic antibiotic therapy.
Design, Setting, and Participants
Retrospective cohort of adult inpatients admitted to general medicine wards at an academic medical center.
At least 24 hours of any parenteral or oral antibiotic therapy.
Main Outcomes and Measures
Medical records of 1488 patients were examined for 30 days after antibiotic initiation for the development of the following antibiotic-associated ADEs: gastrointestinal, dermatologic, musculoskeletal, hematologic, hepatobiliary, renal, cardiac, and neurologic; and 90 days for the development of Clostridium difficile infection or incident multidrug-resistant organism infection, based on adjudication by 2 infectious diseases trained clinicians.
In 1488 patients, the median age was 59 years (interquartile range, 49-69 years), and 758 (51%) participants were female. A total of 298 (20%) patients experienced at least 1 antibiotic-associated ADE. Furthermore, 56 (20%) non–clinically indicated antibiotic regimens were associated with an ADE, including 7 cases of C difficile infection. Every additional 10 days of antibiotic therapy conferred a 3% increased risk of an ADE. The most common ADEs were gastrointestinal, renal, and hematologic abnormalities, accounting for 78 (42%), 45 (24%), and 28 (15%) 30-day ADEs, respectively. Notable differences were identified between the incidence of ADEs associated with specific antibiotics.
Conclusions and Relevance
Although antibiotics may play a critical role when used appropriately, our findings underscore the importance of judicious antibiotic prescribing to reduce the harm that can result from antibiotic-associated ADEs.
Antibiotic use is common in the inpatient setting. Approximately 50% of hospitalized patients receive at least 1 antibiotic during their hospital stay,1 with an estimated 20% to 30% of inpatient days of antibiotic therapy considered unnecessary.2-6 The reasons for antibiotic overuse are myriad, including administration of antibiotics for nonbacterial or noninfectious syndromes, treatment of conditions caused by colonizing or contaminating organisms, and durations of therapy that are longer than indicated. Unnecessary use of antibiotics is particularly concerning because antibiotics may be associated with a number of adverse drug events (ADEs), including allergic reactions, end-organ toxic effects, subsequent infection with antibiotic-resistant organisms, and Clostridium difficile infections (CDIs).7-12
Estimates of the incidence of antibiotic-associated ADEs in hospitalized patients are generally unavailable. Previously, Shehab and colleagues13 conducted a retrospective analysis of ADEs among patients presenting to emergency departments and found that antibiotics were implicated in 19% of all emergency department visits for ADEs. It is unclear whether these data are generalizable to hospitalized patients for a number of reasons: (1) acutely ill hospitalized patients may be predisposed to certain ADEs, such as antibiotic-associated nephrotoxic effects, particularly those admitted with acute renal failure for non–antibiotic-related reasons; (2) hospitalized patients are frequently administered intravenous antibiotic therapy, often at high doses, which may have different adverse event profiles than the oral regimens more commonly prescribed in the outpatient setting14; (3) hospitalized patients are commonly administered multiple medications concurrently, causing a potentially synergistic increase in the risk of ADE development15; and (4) hospitalized patients are more likely to be elderly or have multiple medical conditions, resulting in impaired drug elimination and an increased risk of ADE development.16,17 Previous studies evaluating antibiotic-associated ADEs in the inpatient setting have used administrative databases and have not accounted for antibiotic-associated ADEs that occurred after hospital discharge.18,19 Additionally, they have limited their evaluation of ADEs to single antibiotic classes or single infectious syndromes.18-21 A comparative analysis of the incidence of ADEs across all classes of antibiotics has yet to be performed. Therefore, in the present study, we sought to describe the incidence of antibiotic-associated ADEs for adult inpatients receiving systemic antibiotic therapy while hospitalized in general medicine wards.
This study was conducted at the Johns Hopkins Hospital, a 1194-bed tertiary care facility in Baltimore, Maryland. This study was approved by the Johns Hopkins University School of Medicine Institutional Review Board, with a waiver of informed consent due to the retrospective nature of the study. The data were retrospectively collected on patients 18 years and older admitted to 4 general medicine services between September 2013 and June 2014.6 All patients who received antibiotics for at least 24 hours were included. Exclusion criteria included prophylactic antibiotic use with no clear stop dates, antibiotics used for noninfectious indications (eg, rifaximin for hepatic encephalopathy, erythromycin for intestinal motility), topical or inhaled antibiotics, and antituberculosis regimens.
Demographic data, preexisting medical conditions, antibiotic regimens, and ADEs were collected via patient medical record review. Both inpatient and outpatient medical records were reviewed to obtain follow-up data for patients in the Johns Hopkins Health System. In addition, the Epic Care Everywhere Network, a secure health information exchange, was accessed to view patient data from a large number of health care facilities throughout the United States.22 This enabled the identification of patients presenting to outside emergency departments, hospitals, or primary care clinics with antibiotic-associated ADEs, if these facilities were in the Epic system.
All antibiotic regimens were adjudicated for appropriateness and associated ADEs by at least 2 infectious diseases physicians or pharmacists (P.D.T., E.A., K.D., and S.E.C.). Days of therapy (DOTs) were defined as the number of days from antibiotic initiation until the completion of antibiotic courses. A single DOT was recorded for each individual antibiotic administered to a patient on a given calendar day. Unnecessary antibiotic days were defined as DOTs that were not clinically indicated based on recommendations in the Johns Hopkins Hospital Antibiotic Guidelines.23 For calculations of overall rates of ADEs, the denominator included all patients receiving antibiotics (n = 1488). For calculations involving a single antibiotic, the denominator included only patients receiving that particular antibiotic.
Avoidable ADEs were defined as the proportion of overall ADEs that occurred in patients for whom antibiotic therapy was considered not indicated. Nonindicated antibiotic regimens did not include patients with prolonged durations of therapy because our goal was to determine the incidence of adverse reactions for patients for whom no antibiotic therapy was necessary. For example, if a patient received ciprofloxacin for 15 days for pyelonephritis when 7 days would have been sufficient and the patient developed tendinitis on day 16, one would be unable to attribute the adverse event to the 7 indicated days of ciprofloxacin use or the additional 8 days of unnecessary ciprofloxacin use. We also did not consider overly broad spectrum antibiotic therapy prescribed for valid indications as not indicated because of the impossibility of knowing whether the patient would or would not have developed an ADE with a narrower choice, particularly in the same class of antibiotics.
Criteria used to define antibiotic-associated ADEs are summarized in Table 1. These definitions were derived from available literature, package inserts, and/or consensus opinions prior to any data collection related to the present work. Patients were observed for 30 days from the date of antibiotic initiation for most ADEs (gastrointestinal, dermatologic, musculoskeletal, hematologic, hepatobiliary, renal, cardiac, and neurologic events) and for 90 days from the date of antibiotic initiation for CDI and the development of multidrug-resistant organism (MDRO) infections not previously identified. All ADEs other than CDI or incident MDRO infections were censored at 30 days due to concerns for underestimating the incidence if a longer evaluation period was used because these ADEs generally occur during exposure to particular antibiotics or shortly thereafter. In contrast, data suggest that CDI and the emergence of MDRO infections can become clinically apparent several weeks to months after discontinuing antibiotic therapy.26,27
All potential ADEs were adjudicated in the context of the patient’s medical history and clinical course to ensure that each event was likely to have been antibiotic associated, both to rule out alternative explanations and to appropriately categorize ADEs. Each ADE was then attributed to a single antibiotic, based on the likelihood of that antibiotic causing the specific ADE and the temporal relationship of the antibiotic’s administration to the ADE. For example, acute kidney injury in a patient receiving vancomycin and cefepime would have been attributed to vancomycin use only. This step was performed to avoid overestimating the incidence of ADEs because most patients in our cohort received multiple antibiotics during their hospital stays. However, because virtually all antibiotics can cause CDI or the emergence of MDRO infections, the development of either of these 90-day ADEs was attributed to all preceding antibiotic used.
Rates per 10 000 person-days and 95% confidence intervals were calculated for each ADE and antibiotic class. For 30-day ADEs, the numerator was the number of ADEs attributed to each antibiotic or class of antibiotics. The denominator was the person-time at risk for all patients who received that particular antibiotic or class of antibiotics, computed as the time, in days, from antibiotic initiation to the ADE for patients who experienced the ADE, with censoring at 30 days for patients who did not experience the ADE. The proportion of 30-day antibiotic-associated ADEs per antibiotic or antibiotic class and the proportion of patients receiving a particular antibiotic or antibiotic class who developed a 30-day ADE were also calculated. For 90-day ADEs, the numerator accounted for all preceding antibiotics rather than only a single antibiotic. The denominator was the person-time at risk for all patients who received antibiotics, computed as the time, in days, from antibiotic initiation to ADE onset, with censoring at 90 days. Hazard ratios were calculated to identify the incremental risk of an ADE conferred by each additional day of antibiotic use. All analyses were performed using Stata 13 (StataCorp).
Of the 5579 patients admitted to the 4 included medicine wards during the study period, 1488 (27%) patients received antibiotics for at least 24 hours and were included in the analysis. Previous work describes the demographic data, preexisting medical conditions, sources of infection, and “appropriateness” of antibiotic use of the included population in more detail.6 In brief, the median age was 59 years (interquartile range [IQR], 49-69 years) and 758 (51%) participants were female. The most common underlying medical conditions were diabetes (491 [33%]), structural lung disease (327 [22%]), and congestive heart failure with an ejection fraction of less than 40% (178 [12%]). The median length of hospital stay was 4 days (IQR, 2-9 days). The most common indications for antibiotic therapy were urinary tract infections (179 [12%]), skin and soft-tissue infections (119 [8%]), and community-acquired pneumonia (104 [7%]).
The most frequently prescribed antibiotics were third-generation cephalosporins (607 [41%] regimens), parenteral vancomycin (544 [37%] regimens), and cefepime (414 [28%] regimens) (Table 2). The majority of patients (1176 [79%]) received more than 1 antibiotic during the hospitalization. The median DOTs per patient was 7 days (IQR, 4-14 days). A total of 324 unique ADEs occurred; 298 (20%) patients experienced at least 1 antibiotic-associated ADE. The overall rate of antibiotic-associated ADEs was 22.9 per 10 000 person-days.
Every additional 10 antibiotic DOTs conferred a 3% increased risk of an ADE. A total of 236 (73%) antibiotic-associated ADEs occurred during hospitalization and the remaining 88 (27%) occurred after hospital discharge including 33 (18%) 30-day ADEs, 11 (20%) CDIs, and 44 (52%) MDRO infections. The study investigators determined that 287 (19%) of antibiotic regimens were not clinically indicated, most commonly because of treatment of asymptomatic bacteriuria or treatment of noninfectious lower respiratory tract conditions (eg, aspiration pneumonitis, congestive heart failure).6 Of the 287 nonindicated antibiotic regimens, 56 (20%) were associated with an ADE.
Of the 324 overall ADEs, 186 (57%) were 30-day ADEs. The median time to development of a 30-day ADE was 5 days (IQR, 3-8 days). The median times to 30-day ADEs for the various organ systems were as follows: cardiac, 11 days (IQR, 4-18 days); gastrointestinal, 5 days (IQR, 2-9 days); hematologic, 12 days (IQR, 6-24 days); hepatobiliary, 8 days (IQR, 4-12 days); renal, 5 days (IQR, 2-10 days); and neurologic, 3 days (IQR, 2-4 days). The most common ADEs were gastrointestinal, renal, and hematologic abnormalities, accounting for 78 (42%), 45 (24%), and 28 (15%) 30-day ADEs, respectively (Table 2). Table 3 and Table 4 outline the proportions of 30-day ADEs attributable to specific antibiotics or antibiotic classes and the proportion of patients receiving a specific antibiotic or antibiotic class who developed 30-day ADEs, respectively.
Aminoglycosides, parenteral vancomycin, and trimethoprim-sulfamethoxazole were associated with the highest rates of nephrotoxic effects at 21.2 (95% CI, 12.5-66.0), 12.1 (95% CI, 7.7-19.0), and 13.2 (95% CI, 5.9-29.3) episodes per 10 000 person-days, respectively (Table 2). Two patients experienced QTc prolongation—1 receiving azithromycin and 1 receiving ciprofloxacin after 4 and 18 days of therapy, respectively. Seven patients (6.7 [95% CI, 2.7-12.0] episodes per 10 000 person-days) receiving cefepime developed neurotoxic effects, including encephalopathy or seizures. Less frequent 30-day ADEs, all occurring in single patients, included cefepime-associated anaphylaxis, piperacillin-tazobactam–associated drug fever, daptomycin-associated myositis, ciprofloxacin-associated tendinitis, trimethoprim-sulfamethoxazole–associated pancreatitis, linezolid-associated peripheral neuropathy, vancomycin-associated hives, and a trimethoprim-sulfamethoxazole–associated nonhives rash.
There were 138 ADEs occurring within 90 days, accounting for 43% of all ADEs. Of these 138 ADEs, 54 (39%) were CDI and 84 (61%) were MDRO infections. The median time to development of a 90-day ADE was 15 days (IQR, 4-34 days). The rate of CDI was 3.9 (95% CI, 3.0-5.2) per 10 000 person-days for patients receiving antibiotics, corresponding to 54 (4%) study patients developing CDI within 90 days of antibiotic initiation. The antibiotics most frequently associated with CDI were third-generation cephalosporins (present in 28 [52%] regimens preceding CDI), cefepime (26 [48%] regimens), and fluoroquinolones (19 [35%] regimens).
The rate of emergence of incident MDRO infections was 6.1 (95% CI, 4.9-7.6) per 10 000 person-days, corresponding to 84 [6%] study patients developing an infection with a new MDRO within 90 days of antibiotic initiation. Subsequent gram-positive resistance was observed in 60 (4%) patients, at a rate of 4.8 (95% CI, 3.7-6.1) cases per 10 000 person-days. Forty (67%) of the MDRO cases were related to vancomycin-resistant enterococci infections. Gram-negative resistance occurred less frequently at a rate of 1.7 (95% CI, 1.2-2.6) cases per 10 000 person-days, or in 30 (2%) patients, with extended-spectrum β-lactamase production being the most common resistance mechanism identified.
Antibiotic-associated ADEs were then categorized into clinically significant and non–clinically significant categories. Only 1 category was selected per patient, with the more severe category selected when multiple categories were met. A total of 314 (97%) of the 324 antibiotic-associated ADEs were considered clinically significant because of the following reasons: new hospitalization(s) (n = 10 [3%]), prolonged hospitalization (n = 77 [24%]), additional clinic or emergency department visits (n = 29 [9%]), and additional laboratory tests, electrocardiograms, or imaging (n = 198 [61%]). There were no deaths attributable to any antibiotic-associated ADE.
We found that 20% of hospitalized patients receiving at least 24 hours of antibiotic therapy developed an antibiotic-associated ADE. Moreover, 20% of ADEs were attributable to antibiotics prescribed for conditions for which antibiotics were not indicated. Every 10 DOTs conferred an additional 3% risk of an ADE. Our findings underscore the importance of avoiding unnecessary antibiotic prescribing to reduce the harm that can result from antibiotic-associated ADEs.
Previous studies on antibiotic-associated ADEs in the inpatient setting have largely been limited to single infectious syndromes or single antibiotic classes.18-21,28 For example, Lin and colleagues18 evaluated the incidence of antibiotic-associated ADEs using an administrative database of hospitalized patients with pneumonia. They found that even though less than 1% of patients developed ADEs, the presence of an antibiotic-associated ADE was an independent predictor of prolonged hospital lengths of stay and total hospital charges. Werner et al20 evaluated the frequency of adverse events related to unnecessary fluoroquinolone use in hospitalized patients based on medical record review. They found that approximately 40% of days of fluoroquinolone therapy were unnecessary and 27% of regimens were associated with adverse events including gastrointestinal events (14%), MDRO colonization (8%), and CDI (4%). Finally, Macy and Contreras19 evaluated the incidence of cephalosporin-associated ADEs using an administrative database and found that the most frequently reported serious ADEs were CDI, occurring in approximately 1% of patients.
We believe that our study enhances these investigations in a number of ways. First, unlike previous studies, we evaluated antibiotic-associated ADEs that occurred in both the inpatient setting as well as the outpatient setting after hospital discharge, enabling us to produce a more global picture of the overall incidence of antibiotic-associated ADEs.13,18,19,29 Our previous work suggests that approximately 40% of antibiotics prescribed for hospitalized patients represent antibiotics prescribed at the time of hospital discharge that are to be continued after leaving the hospital.6 We believe that it is important to include these antibiotic days in estimates of antibiotic-associated adverse events for hospitalized patients. Second, in our cohort, infectious diseases physicians and pharmacists reviewed all patient medical records to identify ADEs and to determine whether they were most likely attributable to recent or current antibiotic use using strict, predefined criteria. In contrast, previous studies have generally used administrative databases, in which relevant events are commonly miscoded and through which attributable risk cannot always be assigned.13,18 Furthermore, we did not limit our evaluation to specific antibiotic classes but, rather, included all antibiotic classes.
Our study has a number of limitations. This was a single-center study at an academic hospital with a medically complex patient population. Replication of our results at other institutions and in other patient populations is necessary to enhance the generalizability of our findings. This would also allow for ADE estimates for antibiotic agents not included on our hospital formulary. Furthermore, because prescriptions of some antibiotics were so infrequent (eg, penicillin, ceftaroline fosamil, tigecycline), accurate estimates of some drug-specific ADEs could not be calculated. Our approximations of antibiotic-associated ADEs are likely underestimations for a number of reasons. First, our hospital has had a robust antibiotic stewardship program since 2002 that remained active during the study period, likely reducing overall antibiotic prescriptions, durations of antibiotic therapy, and consequently antibiotic-associated ADEs. Second, we were unable to evaluate data from patients who had follow-up medical care outside the Epic Care Everywhere network, for example those who presented to primary care clinicians, emergency departments, or urgent care centers not using the Epic electronic medical record system.22 Of note, only 119 (8%) patients were considered lost to follow-up with no subsequent inpatient or outpatient visits documented in the Epic Care Everywhere network. Additionally, it is plausible that a portion of patients in this cohort may have previously experienced serious antibiotic-associated ADEs, leading to future avoidance of these agents (eg, hives from penicillin use as a child), also potentially underestimating the incidence of antibiotic-associated ADEs. Finally, we did not include excessively prolonged durations of antibiotic therapy or inappropriately broad antibiotic use toward our calculation of avoidable antibiotic-associated ADEs, likely underestimating this value.
In summary, antibiotic-associated ADEs are common among inpatients receiving antibiotics, some of which may be avoidable with more judicious use of antibiotics. The frequency of antibiotic-associated ADEs may not be recognized by clinicians because ADEs have varied manifestations, clinicians may be unaware of the risks associated with specific antibiotic agents, or because they may occur after patients are discharged from the hospital. Our findings provide quantitative data about the risk of ADEs that clinicians should consider when weighing decisions to initiate or discontinue antibiotic therapy and lend further credence to the importance of antibiotic stewardship to optimize patient safety.
Accepted for Publication: April 6, 2017.
Corresponding Author: Pranita D. Tamma, MD, MHS, Division of Pediatric Infectious Diseases, Department of Pediatrics, Johns Hopkins University School of Medicine, 200 N Wolfe St, Ste 3149, Baltimore, MD 21287 (firstname.lastname@example.org).
Published Online: June 12, 2017. doi:10.1001/jamainternmed.2017.1938
Author Contributions: Dr Tamma had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Tamma, Avdic, Li, Cosgrove.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Tamma, Avdic, Li, Dzintars.
Critical revision of the manuscript for important intellectual content: Avdic, Li, Dzintars, Cosgrove.
Statistical analysis: Tamma, Li.
Obtained funding: Tamma, Avdic, Cosgrove.
Administrative, technical, or material support: Avdic, Dzintars.
Supervision: Dzintars, Cosgrove.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was made possible by an investigator-initiated grant from Pfizer Independent Grants for Learning and Change and The Joint Commission.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We thank Yuan Zhao, MPH, Johns Hopkins University, and John Keenan, MD, Johns Hopkins University, for their assistance with data collection. Dr Keenan received a portion of his salary from Pfizer/The Joint Commission.
Create a personal account or sign in to: