The study population included all persons who accessed care from May 1 to December 31, 2009, at Harbor–UCLA Medical Center (HUMC) in Los Angeles, California. HUMC is a 538-bed, urban, academic, teaching hospital; it serves a diverse population, which is ≈55% Latino, 11% Caucasian, 24% black, 4% Asian, and 4% Pacific Islander.

We conducted a retrospective cohort study to evaluate 3 issues: 1) adherence of clinicians to national recommendations for use of osteltamivir among patients with suspected or confirmed influenza virus infection; 2) appropriateness of patient selection for diagnostic testing; and 3) the likelihood of clinicians to prescribe antiviral drug therapy for persons with known influenza-like illness (ILI) or lower respiratory tract infection (LRTI), 2 conditions for which CDC specifically recommended antiviral drug therapy. For the first 2 objectives, we identified child and adult inpatients and those seen in the emergency department with A(H1N1)pdm09 disease by using 4 overlapping data sources, including the following: 1) prospectively collected electronic A(H1N1)pdm09 virus laboratory-based surveillance data obtained by the HUMC clinical virology laboratory and the Infection Prevention and Control Department; 2) electronic, pharmacy-based oseltamivir utilization data; and 3) data on point-of-care testing performed in the emergency department. These data were combined, and we reviewed the medical records of all patients with a positive laboratory test for influenza in the outpatient setting and of inpatients who had a laboratory test that was positive for influenza virus or were prescribed oseltamivir. Approval for human subjects research was obtained from the Los Angeles Biomedical Research Institute.

We performed a comprehensive review of medical records by using a standardized data collection instrument to identify demographic information and clinical characteristics of patients with the illness, including symptoms and signs and results of viral diagnostic testing and chest radiographs. Use of and indications for oseltamivir, including dose and duration of use, were recorded and, if oseltamivir was not prescribed, reasons for not using the drug were noted. We also recorded whether the patient exhibited risk factors for complications and death (from a preselected list that included concomitant cardiopulmonary, renal, liver, endocrine, blood, or metabolic disorders; immunosuppressive conditions; aspirin therapy; and neurologic conditions), diagnoses at admission or discharge, and length of stay.

We defined suspected influenza as illness in any patient for whom oseltamivir was prescribed by the treating clinician. We defined confirmed influenza disease as illness in a patient with a positive laboratory test result for the virus. To evaluate adherence to guidelines, we used the contemporaneous CDC definition for ILI (fever and cough with or without sore throat) and defined severe illness as requiring intensive care, a documented oxygen saturation of <92%, or both.

To assess the likelihood of clinicians to prescribe antiviral drug therapy for persons with known ILI or LRTI, we identified all inpatients and outpatients with possible upper or lower respiratory tract influenza disease by using International Classification of Diseases, Ninth Revision (ICD-9) diagnostic codes as follows: 079.89 (viral infection), 079.99 (viral infection not otherwise specified [NOS]), 460 (nasopharyngitis, acute), 462 (pharyngitis, acute), 465.8 (infectious upper respiratory, multiple sites, acute), 465.9 (infectious upper respiratory, multiple sites, acute NOS), 466.0 (bronchitis, acute), 466.19 (bronchiolitis, acute, due to other infectious organism), 478.9 (disease, upper respiratory /NOS), 480.1 (pneumonia caused by respiratory syncytial virus), 480.8 (pneumonia caused by virus), 480.9 (viral pneumonia unspecified), 484.8 (pneumonia in other infectious disease), 485 (bronchopneumonia, organism NOS), 486 (pneumonia, organism NOS), 487.0 (influenza with pneumonia), 487.1 (influenza with respiratory manifestation), 487.8 (influenza with manifestation), 488.1 (influenza caused by identified novel H1N1 influenza virus), 490 (bronchitis NOS), 780.6 (fever), 784.1 (pain, throat), 786.2 (cough) (21). The validity of the ICD-9–based ascertainment was assessed by using prospective emergency department triage ILI surveillance data collected beginning October 21, 2009, through the end of the study period.

From this group, we randomly selected 100 persons, stratified by age (50 persons <18 and 50 >18 years of age) by using SAS 9.2, Proc Samplesurvey (SAS Institute, Cary, NC, USA). Using medical record review, we then identified persons with ILI (defined above) or LRTI, defined by the presence of at least 1 specific lower respiratory tract sign, including tachypnea, retractions, or hypoxia (oxygen saturation <92%), and/or abnormal auscultatory findings (crackles/crepitations or wheezing), and/or unequivocal and abnormal radiographic findings.

We performed descriptive analyses of the above variables by using SAS version 9.2. Testing of proportions was performed by using χ² or Fisher exact test as appropriate. All reported p values are 2-tailed and were considered significant if p<0.05.

We believe that this study provides useful information with regard to the diagnostic and therapeutic behaviors of clinicians caring for patients with possible influenza virus infection. Although our data reflect physician behavior during the 2009–10 influenza A (H1N1) pandemic, the findings are likely applicable to any influenza year because diagnostic test performance, disease intensity, antiviral agent resistance, and virus strain affect clinical decision making each year.

We were interested in 2 general concepts: practice performance when influenza was clinically suspected and the potential for missed therapeutic opportunities when it was not. For the former, we found that providers’ practices were often consistent with CDC guidelines but notable deficiencies were also identified. In particular, a substantial proportion of potentially high-risk patients were not empirically treated, and a reason to withhold therapy could not be documented. This dynamic is similar to that for other medical conditions for which clinical practice guidelines are available: provider behavior at variance with the guideline may reflect available patient-level information or other immediate concerns (22,23). In any case, we have identified potential areas for targeted education of healthcare providers that should be supplemented by rapid dissemination and follow-up of national guidelines if and when they change over time.

We also found inconsistencies in the use of antiviral drug therapy, which was often at variance with contemporaneous guidelines. In our population, 25% of patients who received oseltamivir did not have ILI or another clear indication for treatment. During the pandemic, the drug was recommended for inpatients with ILI and outpatients with ILI and risk factors for severe illness if they had sought treatment within 48 hours of symptom onset (5). However, although too many outpatients without ILI received oseltamivir, too few (32%) received the drug despite having LRTI, a consistent indication for therapy. For most patients with LRTI, we could not identify a reasonable justification for withholding therapy. Not surprisingly, all of these patients received antibacterial agents, yet it remains unclear whether the clinicians actively considered influenza virus as a primary pathogen or risk factor for the presumed bacterial superinfection. Influenza virus infection and its association with secondary bacterial infection is well documented with influenza A(H1N1)pdm09 virus infection and with interepidemic disease (24–27). Treatment with antiviral drugs in this setting may lessen illness when superinfection exists (26,28). In this circumstance, greater recognition of the possibility of influenza virus infection and use of antiviral drug therapy may mitigate illness and lessen hospital costs (29,30).

We found that diagnostic practices were often inconsistent with contemporaneous guidelines. Nearly one third of patients were tested for influenza virus, despite the lack of ILI and ≈20% had no other indication for which testing might otherwise be justified (e.g., headache, myalgia). Previous work has shown that relatively few patients with influenza virus infection have systemic signs without fever, sore throat, or cough (31). Although changes in CDC recommendations were quickly disseminated to hospital clinicians by management memo, email, or face-to-face meetings, even more rapid communication and follow-up reminders may have enhanced adherence to guidelines.

We found that the dosage and duration of oseltamivir were generally consistent with CDC guidelines in ≈90% of all treated patients, and specifically for all outpatients. HUMC required the use of a preauthorization drug form that noted the appropriate age- and weight-based dose; an outpatient prescription for oseltamivir would not have been released without a completed form. Such tools have been shown to limit dosing errors (32,33). Also consistent with the CDC guidelines, >90% of hospitalized patients and patients with severe illness in our study received oseltamivir. Among outpatients, we noted that for ≈50%, an appropriate rationale for not providing oseltamivir was documented in the medical record.

Among the small number of dosing errors identified, >40% were related to inappropriate adjustment for renal insufficiency. More than 90% of oseltamivir is metabolized to oseltamivir carboxylate, 99% of which is eliminated by renal excretion, thus requiring dosage adjustment in this setting. Antimicrobial drug dosing errors are common (34,35), and a failure to adjust for renal impairment is a frequent underlying reason (36,37). Although controlled data are not available, oseltamivir has been associated with the development of thrombocytopenia, particularly when renal clearance is artificially lowered by concomitant administration of the drug probenicid (38). Attention should be given to patients’ renal function, particularly in the elderly (diminished renal clearance) and in those for whom higher doses may be recommended, such as the severely ill or obese (8).

We identified clinical management differences between how clinicians prescribed treatment for adult patients and how they prescribed treatment for children. Children who were inpatients were significantly less likely to have ILI at the time of testing and to receive treatment for ILI. When testing was carried out, children were also more likely to test positive for influenza virus than were adults, possibly because of the higher virus load in this population. These data also may reflect more overall testing of children, particularly young children who are more likely than adults to have nonspecific signs and symptoms (lethargy, poor feeding, abdominal pain) (39,40). In addition, infants and young children may not articulate symptoms of ILI (e.g., sore throat), leading to increased nonspecific testing and treatment of this population.

The main strength of this study is the comprehensive nature of case ascertainment, which included laboratory-based information and review of all prospectively collected logs for emergency department point-of-care testing. However, some patients who underwent testing for influenza virus may not have been noted in the outpatient log system. We appreciate that ICD-9 code data for ILI and LRTI may be nonspecific, but our prospectively collected ILI data (albeit for a limited portion of the surveillance period) validated the temporal trends for this diagnosis in the outpatient setting. We also did not include data from medical outpatient (nonemergency department) clinics where other patients with influenza may have been identified and treated, perhaps skewing our data to those who were more ill. As a retrospective study, our conclusions depend solely upon information documented in the medical record, which may be incomplete. Also, the use of an antiviral agent authorization form most likely improved the dosing practice, as has been shown in other settings (32,33). Last, our study population includes only a single academic medical center and therefore may not be representative of the region or the nation.

To our knowledge, similar studies of physician behavior with regard to influenza disease, and for A(H1N1)pdm09 disease in particular, have not been reported. We have identified variations in clinical practice in relation to national guidelines that suggest potential areas of education for future influenza seasons.