In the Pennsylvania Medicaid fee-for-service population during the study period, over one-half of recipients with an upper respiratory tract infection received an antibiotic on the same day as or within seven days after a physician visit for a URI. Even among those patients without a concurrent acute illness that may justify an antibiotic, antibiotic use was high (47.8%). These findings are consistent with other similar analyses, which found 44% to 60% prevalence of antibiotic prescribing for colds or URIs[8, 9, 17, 18]. The high prevalence indicates that potentially inappropriate antibiotic use is high and should be a target for antibiotic education and management to reduce excess use.
Antibiotic resistance is a worldwide public health issue, resulting in emerging antimicrobial resistance and costing $1.3 billion in the United States in 1992. Unfortunately, the extent of the problem is not fully realized since surveillance of drug-resistance is limited. Evidence for a causal association between antimicrobial use and resistance is supported by ecological data (changes in use parallel changes in prevalence of resistance). Also, there is evidence of increased prevalence of resistance in nosocomial strains compared to community-acquired strains, association of prior antibiotic use with nosocomial outbreaks and association of increased antibiotic use with increased resistance rates[21, 22].
To address prescribing factors, educational and administrative interventions have been proposed [23–31]. Administrative interventions are generally restrictive in nature and include prior authorization, formulary restriction, ordering forms and cycling. Educational interventions may include dissemination of susceptibility information, use of computer-based algorithms, and academic detailing. An understanding of the factors contributing to potentially inappropriate antibiotic use can help guide policy makers to design an effective educational or administrative intervention.
Our findings suggest that the presence of certain concurrent acute illnesses increase the likelihood of antibiotic use for a URI. It is possible those antibiotics were actually prescribed for these other acute conditions and that the URI was a coincidental diagnosis. Hence, it could be argued that for some patients the prescription of antibiotics may have been indicated.
However, antibiotic use was highly prevalent (47.8%) even among those without a coincidental diagnosis. In the final regression models coincidental diagnoses only had fair to poor ability to distinguish patients who received an antibiotic from patients who did not, as indicated by the C statistics. Although the presence of illnesses such as otitis media or streptococcal pharyngitis was significantly associated with antibiotic use, they did not help to discriminate patients who received an antibiotic during a URI from those who did not. Thus, the reasons why physicians prescribe antibiotics for URIs is complex and is not fully explained by regression models of this study. Several reasons have been cited for inappropriate prescribing of antibiotics. These include patient expectation, [32–35] physicians' perception of patients' expectations,[32, 35–37] patients' lack of education,[35, 38] and economic pressure[35, 38]. Our findings suggest that more research is needed in understanding why antibiotics are being used for URIs in adults, as well as in children.
This paper explored predictive factors of antibiotic use for URI in an administrative database namely a Medicaid population. The concurrent presence of an acute illness such as cellulitis or UTI may explain the observed antibiotic prescription that was contemporaneous to a URI episode. In addition, we chose to explore the importance of comorbidity in predicting antibiotic use. For some chronic comorbid conditions, such as chronic malignancy or immunodeficiency, antibiotic use may be medically indicated for URIs. Thus, we chose a limited number of comorbid conditions, that we believe justified antibiotic use. Other studies have used a wider range of comorbid conditions that included diabetes, cardiovascular disease. These additional comorbid conditions may be important risk factors in that they increase antibiotic prescribing. However, we chose not to include these additional conditions because their presence generally does not justify antibiotic use.
One limitation of this study was that validation of the medical and prescription claims data against the primary data in the actual medical record was not possible due to limited resources. However, Medicaid prescription claims tend to be a reliable source of drug availability.
Reimbursement is linked to accurate billing, though the reliability of the diagnosis on the medical service claim is less certain since diagnosis generally does not influence reimbursement directly. This may result in misclassification of URI-related events due to incorrect data entry or wrong code assignment. For example, we included ICD-9-CM code 490 (bronchitis not otherwise specified) in our definition of URI; this code may also include patients with COPD or acute exacerbation of chronic bronchitis. In addition, the severity of the URI or the actual indication for the antibiotic prescribed cannot be definitively determined from claims data. However, we did attempt to link the antibiotic to the URI episode temporally and control for other potential antibiotic indications in the analysis. Another limitation is that there may be misclassification of chronic disease history. Although we searched the administrative claims for evidence of chronic conditions for one year before the index URI episode, only 64% of subjects were eligible for an entire year. However, when we reran the 'history of chronic conditions' models for only those subjects with one year of eligibility prior to the URI episode, we obtained results to those in Table 2 (row 10), and nearly identical C statistics to the reported models.
Finally, these findings may not be generalizable to populations other than Medicaid, since Medicaid patients are differentiated by low economic status, source of care and other factors that may affect drug utilization.