Over the nine-year study period, we found that state-level clindamycin and linezolid rates are correlated with national MRSA wound infection rates, but not correlated with zip code or state level MRSA wound infection rates. These results suggest that national, not local, MRSA data may be influencing clinicians’ decisions when prescribing empiric antimicrobials in the outpatient setting. This is at odds with CDC recommendations that facilities keep track of local rates of resistant infections in order to fight the spread of MRSA [4–6]. Thus, efforts by hospitals or health departments to share state or local MRSA rates with clinicians may represent an opportunity to improve antimicrobial prescribing and antimicrobial stewardship in general.
Local knowledge of MRSA infection and resistance trends can improve outpatient antibiotic prescribing. For example, a study by Marra et al., found that outpatient clinicians increased their use of clindamycin and trimethoprim-sulfamethoxazole after a regional (Western Canada) epidemic of community-associated MRSA. Yet the clinicians in that region also frequently prescribed clindamycin even though the predominant MRSA strain in that region was resistant to clindamycin [7]. Similarly, Gupta et al. found that although local rates of community-associated MRSA skin and soft tissue infections were increasing, the majority of patients with these infections were still receiving inactive antimicrobial therapy [8]. A good example of reporting MRSA rates at a regional level is the European Antimicrobial Resistance Surveillance System (EARSS) in which individual countries share their MRSA rates [15]. If other regions such as U.S. states or Canadian provinces similarly shared their MRSA rates, local clinicians would be able to use those rates to enhance antimicrobial prescribing.
This study found no correlation between zip code level antimicrobial prescribing and rates of MRSA wound infections. This could be due to two factors. First, clinicians may not have knowledge of zip code level rates of MRSA wound infections. However, it is likely that the zip code was too small of a unit of analysis. Zip code level rates of antimicrobial prescribing and infection data varied greatly from month to month. For instance, in one zip code the MRSA infection rate varied from 6.0 MRSA infections per 10,000 people in one month to 2.7 per 10,000 people in the following month. Fluctuations such as this would be difficult to correlate with any other measure. The zip code data are also limited because we assessed pharmacy zip codes and laboratory zip codes but we did not have access to the patients’ zip codes. It is possible that these zip codes may not coincide with a patient’s own zip code.
Rates of trimethoprim-sulfamethoxazole prescribing were not associated with MRSA wound infection rates at the zip code, state and national level. This is most likely because trimethoprim-sulfamethoxazole is prescribed for a variety of pathogens in the outpatient setting and is not as well correlated with MRSA infections as linezolid or clindamycin, which are prescribed for a fewer number of indications [16]. It is interesting to note that when time was adjusted for in the statistical models, the correlation between national MRSA infection rates and linezolid prescribing remained, but the correlation between national MRSA infection rates and clindamycin prescribing decreased substantially. This demonstrates that the correlation between MRSA infection rates and linezolid prescribing is not solely due to increased use of linezolid after linezolid was FDA approved. However, further research is needed to determine why the correlation between clindamycin prescribing and national MRSA outpatient infection rates was influenced by time. Potential reasons for changes in clindamycin prescriptions over time include replacement of clindamycin by other antimicrobials, use of clindamycin for infections other than MRSA, concern in regards to the association between antimicrobial use and Clostridium difficile infections, or improved antimicrobial stewardship in the outpatient setting.
Our findings are consistent with those of other studies and the recent CDC Active Bacterial Core findings that community-onset MRSA rates remained steady over time, even though rates of healthcare-onset MRSA have been declining in recent years [7–10, 17]. In fact, that CDC study found that for the first time, community-onset MRSA infection rates have surpassed healthcare-onset MRSA rates [17]. Thus, it is more important than ever to provide outpatient clinicians with the resources they need to prescribe appropriate antibiotics in the outpatient setting.
This study is limited by its ecologic study design. Some patients may have received MRSA-directed antimicrobial therapy without having a culture sent to the laboratory, thus they would be represented in the IMS Health XponentTM database but not in TSN database. However, the IDSA guidelines recommend that cultures be collected and tested from abscesses and other purulent skin and soft tissue infections in patients treated with antimicrobial therapy [2, 3]. Thus, patients should be included in both databases. Additionally, MRSA infection was measured by positive culture results rather than the more stringent Centers for Disease Control and Prevention (CDC) National Healthcare Safety Network (NHSN) criteria. However, a study by Harris et al., found that 82% of MRSA positive clinical cultures were MRSA infections as defined by CDC NHSN criteria [18]. Therefore, the majority of clinical cultures in our dataset should be true MRSA infections, not laboratory contaminants. Also, the laboratories included in TSN were chosen to be nationally representative of hospitalized patient populations but may not be representative of all patients receiving outpatient care for wound infections. These laboratories may be located in more urban settings in which MRSA is more likely to be suspected compared with rural settings.
Although national rates of MRSA infections were statistically significantly associated with state-level prescribing of clindamycin or linezolid, the R2 values were not exceptionally high (22% and 17%), meaning that the majority (78% to 83%) of variation in antimicrobial prescribing was due to factors other than national rates of MRSA infection. Other factors that would contribute to variation in antimicrobial prescribing include patient signs and symptoms, patient risk factors and severity of infection [2, 3]. All of these factors should contribute to therapy decisions. Clinician access to geographic variations in rates of MRSA infections would be just one more tool to assist in the determination of which empiric therapy to prescribe.
We performed this study in order to generate a hypothesis as to whether clinicians rely on local or national MRSA data when prescribing antimicrobials. Future studies should survey clinicians in order to determine what factors are considered when prescribing empirical antimicrobials for suspected wound infections. In summary, clinicians may be relying on national, not local data, when prescribing antimicrobials for suspected wound infections. Local efforts to disseminate local MRSA rates to clinicians may improve empiric prescribing of antimicrobials and should be considered for study in future intervention trials.