The Use of Antimicrobial Agents after Diagnosis of Viral Respiratory Tract Infections in Hospitalized Adults: Antibiotics or Anxiolytics?

Abstract

OBJECTIVE

Because extensive antibiotic use by inpatients has been associated with the development of multidrug-resistant organisms, we aimed to determine which variables were associated with the use of antibiotics after viral respiratory tract infection diagnosis among adult patients admitted to the hospital with respiratory symptoms.

METHODS

A retrospective cohort study was conducted at 2 affiliated urban hospitals in Pennsylvania. We identified all adult patients admitted to the hospital during the period from November 1, 2005, through August 1, 2007, with a viral assay positive for influenza A or B, parainfluenza, adenovirus, or respiratory syncytial virus. Among these patients, we identified those who received antibiotics after the diagnosis of viral RTI. Data on demographics; comorbidities; and physical examination, laboratory, and radiographic findings were ascertained to identify risk factors for antimicrobial use among these patients.

RESULTS

A total of 196 hospitalized patients with positive viral assay results were included; 125 of 131 patients administered antibiotics continued to receive them after viral RTI diagnosis. Among 52 patients with an abnormal chest radiograph, 46 continued antibiotic therapy. An abnormal chest radiograph was independently associated with continued antibiotic use (adjusted odds ratio, 4.28 [95% confidence interval, 1.71–10.77]; P = .002). However, the majority of patients (79 of 125 [63%]) who continued antibiotic therapy had normal chest imaging findings. Eight patients (6%) who continued antibiotic therapy and no patients who stopped developed C. difficile infection (95% CI, 1.5–∞; P = .05), but there was no significant difference in length of stay or mortality.

CONCLUSIONS

Antibiotics are commonly used to treat hospitalized patients with known acute viral RTIs. Continued use is strongly associated with abnormal radiograph findings at admission. However, the reasons for continuation of antibiotics in the treatment of the majority of patients with normal radiographs are unclear and may represent inappropriate use.

Widespread antimicrobial use in the treatment of hospitalized patients creates an ideal environment for the development of antimicrobial resistance. Numerous studies have revealed an association between antibiotic use and natural selection of antibiotic-resistant organisms.^1,2 Despite this association, unnecessary antibiotic use remains commonplace, with estimates that approximately 30% of antibiotic-days for hospitalized adults are unnecessary.^3,4 One area where antibiotic use could be limited is in the treatment of patients with viral respiratory tract infections (RTIs).⁵ Recently, highly sensitive and specific nucleic acid detection assays have become available for the diagnosis of several respiratory viruses, including influenza A and B, parainfluenza, adenovirus, and respiratory syncytial virus.⁶ The use of nucleic acid detection assays for the diagnosis of viral RTIs offers the potential advantage of allowing clinicians to make a diagnosis of viral infection for illnesses that might otherwise be mistaken for bacterial RTIs. Despite this theoretical advantage, it is unclear whether improved diagnosis of viral RTIs has actually led to reduced antibiotic use in the treatment of hospitalized adults. The pediatric literature has shown some advantage to early diagnosis of viral RTIs,^7-10 but surprisingly, there is a paucity of literature regarding this question for hospitalized adults. Prior research was limited by heterogeneous viral diagnostic techniques, limitation to a single viral diagnosis, or small numbers of subjects with a viral diagnosis.^11,12 It is particularly important to understand the extent of antibiotic use by inpatients after diagnosis of viral RTI, because extensive antibiotic use by inpatients has been associated with the development of multidrug-resistant organisms.^1,2,13-15 In addition, there are limited data available on the risk factors associated with the continuation of antibiotic therapy despite a diagnosis of viral RTI. Such data could prove useful in targeting interventions to curtail the use of unnecessary antibiotics for the treatment of inpatients.

In an effort to address these issues, we performed a retrospective analysis of antibiotic use among newly hospitalized adults admitted with acute respiratory symptoms who subsequently received a diagnosis of viral RTI. The primary aim of the study was to determine which variables were associated with the use of antibiotics after viral RTI diagnosis. Secondary aims included determining whether use of antibiotics after viral RTI diagnosis conferred any benefit in terms of patient outcome (ie, mortality, length of stay, and readmission rates) or adverse events (ie, diagnosis of Clostridium difficile infection or allergic reactions associated with antibiotic use).

METHODS

Study Design and Setting

This study was approved by the institutional review board at the University of Pennsylvania and was compliant with the Health Insurance Portability and Accountability Act. We conducted a retrospective cohort study to evaluate antibiotic treatment of patients who had received a diagnosis of viral RTI (influenza A or B, adenovirus, parainfluenza virus, or respiratory syncytial virus) after hospitalization for acute-on-set respiratory complaints.

All patients were admitted to 1 of 2 affiliated hospitals located in Philadelphia, Pennsylvania. The Hospital of the University of Pennsylvania is a 725-bed academic medical center that serves as the tertiary care center for the University of Pennsylvania Health System, and Penn Presbyterian Medical Center is a 324-bed urban community hospital.

Study Population

Adult patients hospitalized with viral RTIs diagnosed during the period from November 1, 2005, through August 1, 2007, were identified by searching the records of the health system’s molecular diagnostics laboratory, which performs respiratory virus diagnostics for both study institutions. The database was queried for all polymerase chain reaction (PCR) assays positive for influenza A or B, adenovirus, parainfluenza virus, or respiratory syncytial virus among patients aged 18 and older seen in either the emergency department or hospital inpatient units. Detection of influenza A, influenza B, parainfluenza virus, adenovirus, and respiratory syncytial virus infection was based on nucleic acid testing of nasopharyngeal swab samples obtained within 72 hours of presentation. Nucleic acid amplification was performed as a partially multi-plexed real-time reverse transcription PCR for RNA viruses (influenza, parainfluenza virus, and respiratory syncytial virus) or real-time PCR for DNA viruses (adenovirus). All assays were performed on an ABI 7900 with TaqMan probes (Applied Biosystems). β₂-microglobulin was used as a control gene to control for the presence of inhibitors or poor sample quality. Viral testing for both study locations was performed in the same central laboratory.

Patient medical records were reviewed, and patients meeting the following criteria were included in the study: (1) admission to the hospital, (2) evidence of acute viral RTI on presentation (defined as new onset cough, sputum production, or coryza), and (3) positive result on a viral PCR assay. Positive viral PCR assay results from patients who were evaluated in the emergency department but not admitted were excluded. Patients whose assays were performed on samples obtained more than 72 hours after initial presentation were excluded to minimize the potential inclusion of nosocomial cases. Because this study assessed antibiotic use in patients without documented bacterial infections, patients were excluded if they had evidence of concurrent bacterial infection (defined as respiratory cultures positive for organisms known to cause disease in the community or in healthcare facilities, blood cultures positive for Streptococcus pneumoniae, or urine cultures positive for antigens for Legionella pneumophila or S. pneumoniae). Patients who had received a diagnosis of concurrent infection outside the respiratory tract, as determined by means of positive microbiological results from a normally sterile body site (eg, blood) or a positive urine culture result with at least 10⁵ colony-forming units from clean catch or at least 10³ colony-forming units from straight catheterization, were also excluded.

Data Collection

The primary outcome measure of antibiotic use after viral diagnosis and duration of antibiotic use was ascertained through the hospitals’ electronic medication ordering system (Sunrise Clinical Manager, Eclypsis Corporation). Antibiotic durations were calculated from the date of the first dose to the date of the last dose delivered. If patients were discharged while still receiving antibiotics, the last dose delivered was calculated by extrapolating from the total number of doses prescribed at discharge (documented in hospital discharge records). Antimicrobials prescribed for chronic suppressive therapies or prophylaxis for preexisting conditions at the time of admission were not considered in the primary outcome measure. The use of antiviral medications was ascertained in the same manner as described above. Outcome measures of death during hospitalization, length of hospital stay, development of C. difficile infection, allergic reaction ascribed to use of an antibiotic, and readmission within 30 days (for any reason) were also determined from patient records. C. difficile infection was defined as the onset of diarrhea in conjunction with a concurrent stool toxin assay result positive for C. difficile. Allergic reaction to an antibiotic was determined by review of the inpatient record. Any rash or swelling diagnosed by an attending physician that was attributed to an antibiotic given during the hospitalization was considered to be an allergic reaction. Hospital medical records of patients who met the inclusion criteria were reviewed to collect demographic characteristics (age, sex, and race), vital signs (maximum temperature, minimum systolic blood pressure, maximum pulse rate, and maximum respiratory rate during the first 24 hours of hospitalization), symptoms (cough, dyspnea, sputum production, and coryza), clinical examination findings (wheezing and acute need for supplemental oxygen), and need for admission to an intensive care unit. Definitions for clinically important tachycardia (heart rate, at least 125 beats per minute), hypotension (systolic blood pressure, 90 mm Hg or less), hypoxia (oxygen saturation, less than 90%), and tachypnea (respiratory rate, at least 30 breaths per minute) were based on parameters from the pneumonia severity index scoring system.¹⁶ The presence of the following comorbid conditions was recorded for each patient: asthma, chronic obstructive pulmonary disease, congestive heart failure, diabetes mellitus, active malignancy, receipt of a transplant, human immuno-deficiency virus–positive status, long-term use of corticoste-roids (equivalent to at least 20 mg/day of prednisone for more than 30 days), and residence in a nursing home.

The following presenting laboratory and radiologic information was also collected: total white blood cell count, serum creatinine level, and results of the admission chest radiograph. Chest radiograph reports with evidence of new “infiltrate,” “consolidation,” “pneumonia,” or “opacity” (in the lung fields) were considered to be evidence of new abnormal findings.

Statistical Analysis

The proportion of patients given antibiotics after viral diagnosis was calculated. Bivariable analyses were then conducted to compare demographic variables, comorbid conditions, presenting vital signs and symptoms, laboratory data, radiographic data, need for admission to an intensive care unit, and outcomes between patients given antibiotics and patients not given antibiotics after viral diagnosis. The Fisher exact test was used to compare categorical variables. The Wilcoxon rank-sum test or the Student t test was used to compare continuous variables. For the primary aim, a multivariate logistic regression model was created to identify variables independently associated with antibiotic use after viral diagnosis. Variables with a P value of .20 or less in bivariable analysis were included in the initial model. Alternative models were compared using the Akaike information criterion to determine the optimal final model.¹⁷

For the secondary aims of outcome-related events, bivariable analyses were conducted. P values of less than .05 were considered to reveal a significant difference. Statistical analysis was performed using Stata, version 11 (StataCorp).

RESULTS

A total of 375 patients were identified with a positive result on a respiratory virus assay (Figure 1). Of these patients, 224 (60%) were admitted to the hospital. The majority of the 196 patients with viral RTI included in the final analysis had influenza (142 patients [72%]), followed by respiratory syncytial virus (30 patients [15%]), adenovirus (13 patients [7%]), and parainfluenza virus (11 patients [6%]).

FIGURE 1.

Identification of patients admitted to the hospital with viral respiratory tract infections. RTI, respiratory tract infection.

Of the 196 included patients, 131 (67%) began antibiotic therapy, and 125 (64%) continued to receive antibiotics after the diagnosis of viral RTI. There was no difference in baseline demographic characteristics (age, sex, and race) or distribution of viruses diagnosed between patients who received or did not receive antibiotics after viral diagnosis (Table 1). The mean time from admission to viral RTI diagnosis was 1 day (range, 1–3 days [95% confidence interval {CI}, 1.1–1.3 days]). The time to diagnosis did not differ between patients who received antibiotics and those who did not (P = .93).

TABLE 1.

Bivariable Analysis of Risk Factors for Antibiotic Use after Diagnosis of Viral Respiratory Tract Infection

Variable	Antibiotics used after viral diagnosis (n = 125)	Antibiotics not used after viral diagnosis (n = 71)	OR (95% CI)	P
Demographic characteristics
Age, years, mean (range)a	56 (18–91)	52 (18–92)	1.15 (0.98–1.36)	.08
African American race	78 (62)	47 (66)	0.85 (0.44–1.59)	.60
Male sex	49 (39)	29 (41)	0.93 (0.49–1.77)	.88
Respiratory virus identified
Influenza	86 (67)	56 (82)	1.55 (0.36–9.35)	.75
Respiratory syncytial virus	22 (18)	8 (11)	0.59 (0.28–1.22)	.14
Adenovirus	8 (6)	3 (4)	1.68 (0.67–4.63)	.30
Parainfluenza virus	9 (7)	4 (6)	1.30 (0.35–6.00)	.77
Presentation
Temperature ≥37.8°Cb	87 (70)	47 (66)	1.12 (0.61–2.05)	.71
Pulse rate ≥125 beats/min	13 (10)	7 (10)	1.23 (0.43–3.55)	.70
SBP <90 mm Hg	2 (2)	2 (3)	0.47 (0.08–2.91)	.42
Respiratory rate ≥30 breaths/min	10 (8)	3 (4)	1.28 (0.36–4.58)	.39
Oxygen saturation <90%	24 (19)	8 (11)	1.56 (0.63–4.25)	.42
Cough	116 (93)	60 (86)	1.87 (0.67–5.30)	.23
Dyspnea	106 (85)	58 (82)	1.23 (0.50–3.01)	.61
Sputum	65 (52)	29 (41)	1.52 (0.82–2.80)	.18
Coryza	32 (26)	32 (45)	0.47 (0.25–0.90)	.02
Wheezing	36 (29)	20 (28)	0.97 (0.52–1.97)	.93
Comorbidities
COPD	23 (18)	4 (6)	3.75 (1.21–11.60)	.02
Asthma	16 (13)	13 (18)	0.89 (0.38–2.03)	.78
Malignancy	33 (26)	7 (10)	2.82 (1.19–6.66)	.02
Receipt of transplant	20 (16)	7 (10)	1.05 (0.42–2.61)	.91
Congestive heart failure	28 (24)	22 (30)	0.68 (0.34–1.33)	.27
Diabetes mellitus	50 (40)	19 (27)	1.67 (0.87–3.20)	.12
HIV-positive status	9 (7)	6 (8)	0.84 (0.27–2.60)	.76
Long-term use of corticosteroids	24 (19)	14 (19)	0.84 (0.39–1.81)	.66
Nursing home residence	14 (11)	1 (1)	6.21 (1.37–28.04)	.02
Evaluation
Creatinine level, mg/dL, mean (range)	1.5 (0.5–11.6)	1.4 (0.6–11.0)	1.02 (0.83–1.25)	.82
WBC count, thousand cells/μL, mean (range)c	8.4 (0.2–18.3)	7.6 (2.5–21.5)	1.04 (0.97–1.11)	.25
Abnormal chest radiograph result	46 (37)	6 (8)	4.58 (1.89–11.06)	.001
ICU admission	32 (26)	6 (8)	3.11 (1.19–8.13)	.02

NOTE. Data are no. (%) of patients unless otherwise specified. CI, confidence interval; COPD, chronic obstructive pulmonary disease; HIV, human immunodeficiency virus; ICU, intensive care unit; OR, odds ratio; SBP, systolic blood pressure; WBC, white blood cell.

^aAge was divided into 10-year increments for comparative analysis.

^bMaximum recorded temperature during first 24 hours of hospitalization.

^cInitial laboratory value.

For patients given antibiotics, the median duration of antibiotic use after viral diagnosis was 8 days (range, 2–27 days). The mean number of antibiotics used was 2 (range, 1–6), with fluoroquinolones (66 [53%] of 125 patients), cephalosporins (38 [30%] of 125 patients), and macrolides (25 [20%] of 125 patients) used most commonly (Figure 2). Oseltamivir was administered to 30 (21%) of 142 patients with influenza; 16 (53%) of the 30 patients receiving oseltamivir continued to receive antibiotics after influenza diagnosis. No other antiviral medications were used. Twenty-one patients who continued to receive antibiotics had sputum cultures that grew “normal oral flora.” The median duration of antibiotic use after diagnosis for these patients was 9 days (range, 2–18 days).

FIGURE 2.

Distribution of antimicrobials used after a diagnosis of viral respiratory tract infection. One hundred twenty-five (64%) of 196 adult inpatients continued to receive antibiotics after the diagnosis of viral respiratory tract infection was made. Antibiotics that are typically used for community-acquired pneumonia were most commonly prescribed. Multiple different antibiotics were often used at different times during the hospitalization. In contrast, the anti-influenza drug oseltamivir was used to treat only 30 (21%) of 142 patients with influenza. PCN, noncephalosporin β-lactams.

Risk Factors for Use of Antibiotic Therapy

Several variables were associated with use of antibiotics after viral RTI diagnosis in bivariable analysis (Table 1). In multivariable regression analysis, the presence of an abnormal admission chest radiograph finding was the only significant risk factor for use of antibiotics after viral RTI diagnosis (odds ratio, 4.28 [95% CI, 1.71–10.77]; P = .002) (Table 2).

TABLE 2.

Mutivariable Analysis of Risk Factors for Antibiotic Use after the Diagnosis of Viral Respiratory Tract Infection

Risk factor	Antibiotics used after viral diagnosis (n = 125)	Antibiotics not used after viral diagnosis (n = 71)	Adjusted OR (95% CI)	Adjusted P
Sputum	65 (52)	29 (41)	1.3 (0.66–2.57)	.45
COPD	23 (18)	4 (6)	3.13 (0.95–10.37)	.06
Malignancy	33 (26)	7 (10)	2.36 (0.94–5.95)	.07
Nursing home residence	14 (11)	1 (1)	5.09 (1.02–25.3)	.05
Abnormal chest radiograph result	46 (37)	6 (8)	4.28 (1.71–10.77)	.002
ICU admission	32 (26)	6 (8)	2.56 (0.92–7.14)	.07

NOTE. Data are no. (%) of patients unless otherwise indicated. CI, confidence interval; COPD, chronic obstructive pulmonary disease; ICU, intensive care unit; OR, odds ratio.

Outcomes and Complications of Antibiotic Therapy

The median length of hospital stay was 5 days (95% CI, 4–6 days) for patients given antibiotics, compared with 3 days (95% CI, 2–3 days) for patients not given antibiotics after receiving a diagnosis of viral RTI (P < .001). In-hospital all-cause mortality was significantly higher in the antibiotic continuation group (10 vs 0 deaths; P = .01). The frequency of rehospitalization within 30 days of discharge was higher among patients who continued to receive antibiotics after viral RTI diagnosis (28 [22%] of 125 vs 9 [13%] of 71) but was not statistically significant (P = .13). Eight patients (6%) who continued antibiotic therapy and no patients who were not exposed to antibiotics after diagnosis developed C. difficile infection (95% CI, 1.5–∞; P = .05). One patient who continued to receive antibiotics developed a rash attributed to ceftriaxone. No other allergic reactions were reported.

DISCUSSION

Administration of antibiotics was commonly continued after the diagnosis of viral RTI in this cohort of hospitalized adults. One hundred twenty-five (64%) of 196 patients continued to receive antibiotics after the diagnosis of viral RTI. The only antiviral used, oseltamivir, was prescribed for a small percentage (21%) of patients with influenza; one-half of these patients continued to receive antibiotics. In multivariate analysis, an abnormal chest radiograph finding was independently associated with antibiotic use after viral RTI diagnosis. The use of antibiotics was associated with C. difficile infection.

The patients included in this study arrived at the hospital with complaints of acute respiratory illness compelling enough for healthcare providers to consider and test for viral RTIs. Despite positive results on viral RTI testing, a majority of the patients were prescribed antibiotics and the decision to stop antibiotic therapy was rare. Among 131 patients who received antibiotic therapy, only 6 patients stopped receiving antibiotics after a viral diagnosis was made; 125 patients continued to receive antibiotics after viral diagnosis. One explanation for continuing antibiotic therapy after a viral diagnosis is that clinicians were concerned about bacterial coinfection. In an earlier retrospective survey limited to hospitalized patients with influenza, concern for secondary bacterial infection was the most common reason for continuing antibiotic therapy.¹² In our study, the duration of antibiotic use after diagnosis (median, 8 days) and the spectrum of antibiotics used (fluoroquinolones, cephalosporins, and macrolides) further suggest that clinicians harbored concerns for concomitant bacterial pneumonia.¹⁸

For the treatment of patients with an abnormal chest radiograph finding, the decision to use antibiotics is not unreasonable. Several studies of community-acquired pneumonia have revealed bacterial coinfection with respiratory viruses, with estimates as high as 47%.^19-22 In clinical practice, extensive testing for bacterial origins of disease is not recommended outside the intensive care unit. One can therefore understand why some clinicians might choose to treat for possible bacterial coinfection when chest radiography shows evidence of pneumonia.¹⁸ However, only 46 (37%) patients had abnormal chest radiograph findings. It is less clear why the remaining 79 (63%) patients with normal chest imaging results were prescribed antibiotics. Possible factors that might have led clinicians to use antibiotics, such as age, presenting signs, symptoms, and comorbid conditions, were not associated with antibiotic use in multivariate analysis. The use of antibiotics for nonpulmonary infections is also unlikely, because patients with evidence of any concurrent bacterial infection were excluded from the analysis (Figure 1).

Residency in a nursing home was also associated with continued antibiotic use after viral diagnosis (Table 2). Although other individual comorbidities were not significant risk factors for antibiotic use, nursing home residency may reflect overall physical debilitation related to multiple comorbid conditions and advanced age. As such, this particular patient population is at increased risk for complications related to antibiotic exposure, including C. difficile infection and infection due to multidrug-resistant pathogens.²³ Therefore, limiting the use of unnecessary antibiotics is perhaps a greater priority in this population.

Although antibiotics may have been used as a prophylactic measure against secondary bacterial pneumonia, there are few data to support such practice.^24-26 In contrast to antibiotic use, oseltamivir, an effective treatment for influenza, was used for only 21% of patients with influenza. Animal models and clinical trials have revealed neuraminidase inhibition to be effective in reducing bacterial complications and the subsequent need for antibiotics after a diagnosis of influenza.^27-32 The duration of symptoms prior to influenza diagnosis may have contributed to the decision regarding when to use oseltamivir, because early treatment has been associated with improved outcomes.³² Unfortunately, data on the duration of symptoms were not collected for this cohort.

This study did not reveal an obvious clinical benefit with antibiotic use after viral RTI diagnosis. Antibiotic use may have led to harm in some cases. Indeed, a significantly higher frequency of C. difficile infection—a well-established risk of antibiotic use—was observed among patients who continued to receive antibiotics after viral diagnosis. The higher frequency of C. difficile infection observed here underscores only one of the potential problems associated with administering antibiotics without a clear reason. The higher all-cause mortality and longer hospital stays observed in the antibiotic group suggest that antibiotic use did not provide substantial clinical benefit. However, because of the small number of observed outcomes, it is difficult to draw conclusions regarding whether antibiotic use directly affected mortality and length of stay or merely reflected empirical antimicrobial use in a subset of patients with progressive clinical decline.

Other serious public health issues associated with antibiotic use, such as the emergence of antibiotic-resistant pathogens, were not the subject of this study. Nonetheless, these data reveal at least one area where antibiotics are commonly used in hospitalized patients without a clear reason. Recognition of this may be helpful in developing interventions to limit inappropriate antibiotic use in the future.

Potential limitations of the study should be noted. Because of the study’s retrospective nature, we were not able to survey providers’ reasons for continuing antibiotic therapy after a diagnosis of viral RTI. One can hypothesize, however, that clinicians’ very use of the viral diagnostic test argues against a deep mistrust of the test’s specificity. Second, this study was not powered to detect differences in outcomes, including mortality, hospitalization durations, readmission rates, and C. difficile infections. However, the well-established association between C. difficile infection and antibiotic use makes this outcome a plausible and concerning risk that should be underscored when treating a known viral illness. Third, the study was conducted at 2 affiliated academic hospitals, thereby limiting the generalizability of results to other settings. Last, we excluded patients with concurrent positive microbiological findings from respiratory sources on the assumption that these results reflected bacterial RTI; however, some cases may have only reflected colonization, which would not justify antibiotic use.

In conclusion, antibiotics are commonly prescribed to hospitalized patients with known acute viral RTI. Radiographic evidence that caused concern for pneumonia was the strongest predictor for continuation of antibiotic therapy after a diagnosis of viral RTI. However, the majority of patients who were prescribed antibiotics had normal chest radiograph findings. The reasons for continuing antibiotic therapy for patients with normal chest radiograph findings were not immediately apparent. One hypothesis that could explain antibiotic use after viral diagnosis is clinician anxiety over the possibility of concurrent or developing bacterial RTI. An antimicrobial stewardship program might be beneficial in targeting patients who receive a diagnosis of viral RTI to limit unwarranted antibiotic use.