Management by Primary Care Clinicians of Patients Suspected of Having Community-Acquired Methicillin-Resistant Staphylococcus Aureus Infections
The intervention was designed to optimize treatment for SSTIs consistent with the CDC CA-MRSA guidelines. In particular, we looked for changes in rates of I&D procedures performed, cultures obtained, and use of systemic antibiotics. When 3,112 pre-intervention SSTI cases were compared to 1,406 intervention-period cases, overall procedure and culture rates were low. Based on the qualitative findings, many of the abscesses were spontaneously draining when the patient presented at clinic, and these generally do not require that a procedure be performed.
A trend toward a small decrease in the overall procedure rate was observed. Clinical considerations, such as abscess size or depth degree of fluctuance may drive decisions about drainage procedures more than the suspected etiology of the infection. If this is true, then an intervention aimed at increasing MRSA awareness would not be expected to increase procedure rates. It is possible that increasing awareness of MRSA across the population may have resulted in patients presenting earlier in the course of their illness over time. Very early abscesses may be less likely to need drainage procedures; this might explain the overall decrease in procedure rates during the intervention. In this context, it is interesting to note that men were two times more likely to receive a drainage procedure; perhaps this is because men are less likely to see a doctor for any complaint and may present later in the course of an SSTI. Among midlevel providers, culture rates were significantly higher despite lower procedure rates. Cultures obtained from spontaneously draining lesions and from aspirated lesions may account for cultures when procedures are not done or detected. However, it is unclear why these circumstances would be more common among midlevels.
It is likely that we were unable to capture the majority of procedures in the electronic dataset, as suggested by the much higher procedure rate in the manual audit. Reasons for not capturing procedures include that the clinician does not bill for it (this may be more common for aspiration procedures, which are a more minor procedure) or the patient is referred to a specialist or emergency room where a procedure is done but not captured in the electronic dataset. When the majority of procedures are not captured, the trend toward a decrease in the procedure rate may be a spurious finding. Finally, the intervention period coincided with the peak of the 2009 H1N1 influenza epidemic when practices reported very high patient volumes. Amid high patient volume, providers may have felt they did not have enough time to perform procedures.
The CDC recommends that providers culture all purulent infections, but the intervention did not increase culture rates significantly. Cultures are important for MRSA disease surveillance, yet the culture may have little impact on the care of an individual patient, especially if MRSA-covering antibiotics will be given anyway. The providers who agreed to participate in the QI evaluations were very interested in MRSA surveillance and performed cultures in all patients where there was purulent drainage, yet none of these culture results actually changed the clinical management for those patients. Although MRSA prevalence is increasing (or providers are becoming more aware of increasing MRSA prevalence), it is reasonable to expect little or no change in clinician behavior on drainage (and therefore culture, which is tied to drainage) because that is already standard of care. Finally, similar to the discussion of procedure rates above, for unclear reasons, cultures were not fully captured in the electronic dataset, and therefore significant changes in the culture rate could be missed. Supporting this is the finding of a much higher culture rate in the manual audit.
CDC guidelines recommend I&D as the primary treatment for abscesses and that systemic use of antibiotics, if used, cover MRSA. Antibiotics are recommended for the treatment of cellulitis; however, the CDC notes that the role of MRSA in cellulitis is uncertain. Compared to the pre-intervention period, during the intervention period antibiotic use increased significantly in both systems for all 680.x, 681.x, and 682.x skin infections, and the proportion of prescribed antibiotics that covered MRSA also increased significantly. It is possible that the significant increase in antibiotics and MRSA-covering antibiotics for 681.x-682.x can be accounted for by temporal trends. The piecewise GEE model indicated that the intervention generated an increase in antibiotic and MRSA antibiotic prescriptions that exceeded the natural rate of change that occurred over the period of the study. However, this increase was not statistically significant (p=0.0539 for antibiotics and p=0.1220 for MRSA antibiotics). The increase in antibiotics and MRSA-covering antibiotics for 680.x was also significant and, as opposed to 681.x-682.x cases, temporal trends do not explain it.
Patients with diabetes among 680.x cases appeared to receive fewer antibiotics. It is possible that diabetics were referred to specialists more often for I&D since primary care clinicians may be reluctant to perform procedures in higher risk diabetic patients. If this was the case, any antibiotics would not appear associated with the provider at the index visit.
If we postulate that the proportion of cellulitis cases in diagnostic codes 681.x and 682.x remained approximately the same across the duration of the study, then most of this increase in antibiotic prescribing during the intervention may have resulted from the increased use of antibiotics for abscesses. In the focus groups and qualitative interviews, providers frequently noted that they were uncomfortable not treating possible MRSA with antibiotics. Some clinicians commented that using antibiotics in abscess patients with suspected MRSA was the "community standard of care" with the implication that deviating from the community standard of care may increase malpractice risk. From the increasing prevalence of MRSA, or increased awareness of its prevalence, it is reasonable to expect that clinicians would change their prescribing behaviors, including possibly prescribing more often, and certainly prescribing more often to cover MRSA, as the results suggest.
The patient follow-up data did not reveal significant differences from pre-intervention to intervention on key outcomes of hospitalization, ED or urgent care visits, proportion of cases that resolved, medication adherence, the need for a second course of antibiotics, or time to infection resolution. This suggests that the intervention did not have any untoward effects in each of the systems nor did it improve these important outcomes. It may be that unmeasured effects, such as the severity of the SSTI, were more important for these outcomes. There were some notable differences between the two health systems. In the multivariate analyses, a previous MRSA infection was associated with more unplanned care events, which may be due to greater difficulty in successfully treating recurrent infections. Otherwise, the multivariate analysis for time to resolution and unplanned care events did not yield any new significant findings.
Although we do not have data to explain the variation in reported visits to ED or urgent care, it is possible that differences in clinic access for follow-up care, availability and access to urgent care centers, or open-access scheduling would account for these differences between the systems. Finally, almost all cases in WHA were resolved at the time of follow-up, whereas only about half of the cases had resolved in MCNT. This is likely because the average days to followup differed significantly in the two health systems, with WHA completing calls about 36 days later on average than MCNT.
Provider QI Case Reports
Clinicians providing QI reports performed significantly more drainage procedures for abscesses and obtained cultures 100 percent of the time a procedure was done or when the abscess drained spontaneously. This could reflect increased concern about MRSA in providers who volunteered to be interviewed or a reporting bias in which providers were more likely to agree to be interviewed about cases that were more difficult to manage and, therefore, were more likely to require drainage. These providers were more likely to prescribe antibiotics for abscesses, and when antibiotics were prescribed, they reported using MRSA-covering antibiotics 100 percent of the time. This could be a result of the intervention, or it could reflect greater provider knowledge of MRSA and the guidelines from a baseline higher level of concern about MRSA. As noted before, these providers frequently expressed the idea that they were uncomfortable not treating abscesses with antibiotics because they perceived that this was the community standard of care. Only a few of the interviewed providers were willing to forgo antibiotics in completely drained abscesses regardless of the CDC guidelines and research data showing that adding antibiotic treatment to drainage procedures does not change outcomes for completely drained abscesses less than 5 cm in diameter. It should be emphasized that the provider QI reports and interviews were not done on randomly selected cases. When all cases were considered, no Hawthorne Effect was seen.
The qualitative analysis of the QI reports also added information not available in any other part of the study about:
- Widely varying provider views about and use of packing.
- Provider concerns about using doxycyline or clindamycin for SSTI treatment.
- High rates of spontaneous drainage in abscesses and how this changed treatment decisions.
- Whether culture results changed clinical management.
- Whether providers thought about covering MRSA and/or strep in cellulitis patients.
- Providers' thoughts about presenting characteristics of MRSA infections.
- Providers' assessment of MRSA risk factors.
- How the intervention materials were used in clinical practice.
Manual Chart Audit
The manual audit indicates that the 680.x code is sensitive and specific for purulent infections (as compared to cellulitis). The manual audit also suggests limitations in the electronic dataset for identifying cultures and procedures, as discussed above.
Although data were extracted electronically from EHRs, there are possible limitations with regard to the accuracy and completeness of the data. The focus of this project was on purulent infections, which are likely to be caused by CA-MRSA, as compared to cellulitis, in which the etiology is less clear. Although the 680.x ICD-9 codes reliably identify purulent infections, the majority of purulent infections are coded as 681-682.x. Yet within the 681-682.x codes, the majority of cases appear to not be purulent (cellulitis only). Thus a major limitation of this project is identifying with reasonable certainty the cases that are purulent based on the diagnostic code. Billing codes were used to determine if a procedure was done, but there may be inaccuracies in the data due to billing data being separate from EHR data. Based on our interviews with clinicians, we learned that some procedures are not billed for any number of reasons. In one health system with a surgical department, the billing system is separate from the EHR. Because there is a manual process for following up on missing Current Procedural Terminology, or CPT, codes, physicians may feel less urgency to enter these at the same time as the procedure; thus billing codes for procedures might not appear with the ICD-9 classification code at the time of visit. Although EHRs have codified prescriptions for most medications and dosages, they still allow for manual entry by clinicians. Because manual entries contained misspellings, incomplete words, or non-standard entries, we have missed detecting some antibiotics. Also, we would have missed handwritten prescriptions for antibiotics. While we do not believe these data limitations were different during the pre-intervention and intervention periods, it is possible that providers who were not billing as much for procedures did increase their coding during this QI project.
With heavy reliance on servers for storing and transferring electronic data, server outages can have potentially negative consequences for a study such as this. At one point, a server in WHA was down, so new data feeds were not updating records. This outage meant that patient followups could not be conducted for several weeks. While WHA attempted to remedy the problem by calling patients who had been missed, it may have significantly decreased the number of completed patient follow-up evaluations WHA could conduct. In addition, calls to patients were delayed because of this, increasing the possibility of a recall bias and leading to data that could not be used for time to resolution of infection. These delays also biased the comparison between the two systems of whether an infection was resolved, since the longer the delay in contacting the patient, the more likely the infection would have resolved independent of clinical care factors. Finally, at WHA, the calls to patients were made by a staff person with other work responsibilities, resulting in delays even when the servers were fully operational (this staff person also changed mid-study due to health reasons). In MCNT, the calls to patients were made by an outside company contracted by MCNT, and this resulted in calls to patients occurring much sooner after the initial SSTI encounter. The small sample size of 680.x cases in the patient follow-up calls unfortunately made it difficult to discern trends in these ICD-9 codes that have the most specificity for purulent infections.
Late in the study period, servers at CINA were out for a period of time, which halted provider QI case reporting in both health systems. It is not known how many case reports were delayed or incomplete due to the outage, but the number is likely very small.
The study team for this QI project was based in Kansas City and Denver, remote from where the intervention was conducted in North Carolina and Texas. Although there was a key champion for each system, there was no identified champion at each site; thus, we are uncertain of the fidelity of the intervention. This concern was confirmed in the clinician interviews, during which several clinicians stated they were unaware of the intervention.
There was a potential ceiling effect for benefit from the intervention, because the prevalence of MRSA was already in two-thirds of S. aureus cases prior to the start of the intervention. This intervention may demonstrate greater benefit in clinics or systems where MRSA is less prevalent.
The proportion of pediatric patients with SSTIs was somewhat lower in the intervention period. This could be due to a lower proportion of pediatric providers in one health system. It could also be that seasonality played a role as the intervention period did not include the summer months.
Finally, the intervention consisted of provider education, a ready-made I&D kit, and point-of-care informational materials for providers and patients. The intervention resulted in an increase in antibiotic use and the use of MRSA antibiotics specifically. Yet because of the multiple components to the intervention, we cannot state with certainty which component(s) contributed to these findings.
Management of Skin and Soft Tissue Infections
Initial management of SSTIs in these primary care practices is mostly concordant with CDC recommendations; however, management of specific cases can vary significantly based on the provider's assessment of the initial presentation of the skin infection.
While earlier focus group results with clinicians suggested that providers would like I&D kits readily available, in practice, it appears that some providers prefer to use their own routine to gather supplies. Patient handouts were valuable, regardless of the utility of I&D kits.
Electronic Data Extraction
- Lessons learned from Distributed Ambulatory Research in Therapeutics Network (DARTNet)
- The use of DARTNet ultimately provided the QI effort and the project team with the use of high-quality data across two very large clinical enterprises. It is difficult to access the practicality of this effort without such an automated system.
- The "certification" of data extracted from EHRs by DARTNet is an important, yet demanding endeavor. Because data entry into EHRs and billing systems is still a human process, mapping the process for study-specific codes and criteria in each clinic system would help understand early data extraction problems and plan for improving programming to find and extract the correct data. Multiple data extractions and analyses were required to assure the fidelity of data capture and the development of a continuity of care record that met the analytic needs of the QI effort as well as the project team.
- Reliance on a single programmer as the dependent person to extract and organize the required data can be problematic.
- Frequent surveillance of multiple data systems is needed for early detection of any problems. Weekly or monthly audits of all data sources by at least two separate entities may help detect data problems earlier. Such an effort would have provided an early warning to the project team concerning missing patient information.
- Study teams should cover a portion of person's time in each health system to have "on-the-ground" help for planning, troubleshooting, and monitoring the project.
In this project, the intervention resulted in an increase in antibiotics use and an increase in the proportion of prescribed antibiotics that covered MRSA. It did not change the rate of obtaining a culture, and there was a trend toward a decrease in the procedure rate. Although the specific drainage kit may not be generalizable, the overall intervention is, including having a ready-to-go drainage kit at the point of care, as well as the educational materials for providers and patients. Having an educational session for providers on SSTIs and MRSA is also a generalizable intervention, and may have contributed to the outcomes.
Based upon our experiences and learning from this project, consideration should be given to replicating this intervention in a health system where the prevalence of CA-MRSA is lower, such as <20 percent of S. aureus positive cultures. Also, it may be more generalizable to not use the pre-packaged procedure kits due to the wide variance in how procedures are done throughout the country and to only use the printed point-of-care patient and provider materials, as well as clinician education.
Finally, the efforts of the project team using Government funds have resulted in the ability to certify DARTNet for large-scale data extractions and aggregations related to CA-MRSA care in primary care practices. The Government may have an interest in conducting larger scale comparative effectiveness studies (comparing multiple intervention or practice redesign strategies) using these newly developed capabilities of DARTNet to build upon what was learned in the project funded by this task order and to provide actionable evidence to guide the provision of care for CA-MRSA.
1. Agency for Healthcare Research and Quality. Medical Expenditure Panel Survey. Available at: http://meps.ahrq.gov/mepsweb/. Accessed May 28 2008.
2. Fridkin SK, Hageman JC, Morrison M, et al. Methicillin-resistant Staphylococcus aureus disease in three communities. N Engl J Med 2005 Apr;352(14):1436-44. Erratum in N Engl J Med 2005 Jun 2;352(22):2362.
3. Miller LG, Perdreau-Remington F, Rieg G, et al. Necrotizing fasciitis caused by community-associated methicillin-resistant Staphylococcus aureus in Los Angeles. N Engl J Med 2005 Apr 7;352(14):1445-53.
4. Martinez-Aguilar G, Avalos-Mishaan A, Hulten K, et al. Community-acquired, methicillin-resistant and methicillin-susceptible Staphylococcus aureus musculoskeletal infections in children. Pediatr Infect Dis J 2004 Aug;23(8):701-6.