We initiated this multicenter, prospective observational study to learn about the current epidemiology, clinical presentation, treatment and outcomes for patients with cSSTIs. The study protocol did not recommend or mandate any initial or subsequent interventions, but CRFs recorded each of the local physicians’ management decisions. The prospective design of the study made it possible to capture a more comprehensive patient profile, including data on risk factors for healthcare-associated infections, wound measurements, and infection parameters. These enabled us to score wound severity, to clinically assess the infection, to note the reasons investigators initiated and discontinued each agent, and to evaluate quality of life measures. Furthermore, we enrolled a large number of patients with varied types of infections admitted to hospitals representing a wide-range of types of institutions and geographic regions. Thus, these data are likely to reflect usual current clinical practice for managing cSSTIs in hospitalized patients in the US.
Our study showed that, among the 1,033 patients enrolled, 73.5% had a healthcare-associated infection. This is nearly identical to the 73.6% reported in a single-center retrospective cohort study of patients hospitalized with cSSTIs
. Compared to our study, their criteria for defining healthcare-associated infection used a longer time frame for previous hospitalization (1 year versus 6 months) and previous antibiotic use (90 days versus 30 days) and did not include immunosuppressed state as a criterion. This rate of healthcare associated infection is considerably higher, however, than the 27% reported in a recent retrospective cohort study of more than 12,000 patients hospitalized with skin, soft-tissue, bone or joint infection
. This difference could be attributed to that study including bone and joint infections, and using a definition of healthcare associated infection with a substantially shorter timeframe for previous hospitalization compared with ours (30 days versus 180 days) and that did not include previous antibiotic treatment, a criterion met by 47.6% of our patients. Physicians treating patients with cSSTI may find the high prevalence rate of healthcare-associated infections noted in this study to be clinically important, as previous studies
[15, 24] have reported that these patients had longer lengths of stay, higher mortality, and higher hospital costs.
Our evaluation of initial antibiotic treatment revealed that vancomycin was used considerably more often (in 61.0%) than has been reported in previous studies. Two retrospective multicenter studies performed in the past decade, both of which used an inpatient claims database, eported initial treatment with vancomycin in 17.6% and 27.8% of cSSTI patients, respectively
[14, 25]. The much higher use of vancomycin in our study is likely attributable to the steep rise in the prevalence of MRSA as a pathogen in cSSTIs over the past several years
. It is noteworthy that the initial intravenous therapy selected was discontinued in over a fifth of patients because it was either not needed or inappropriate. Also of note is that a substantial percentage of patients were treated with relatively broad-spectrum agents that are typically used to cover gram-negative bacilli (e.g., fluoroquinolones, 3rd generation cephalosporins, aminoglycosides), which is probably unnecessary in most cases of cSSTI.
The mean LOS we observed in this study is longer than might be observed in some practices; this is probably related to the fact that we enrolled patients with more severe infections as they needed to be hospitalized for at least 48 h. Our LOS is, however, similar to those previously reported in patients hospitalized for cSSTIs
[15, 24, 27]. Despite differences in patient population, the overall mean LOS in our study (7.1 days, ranging from 5.8 to 8.1 days) was similar to that reported by Lipsky et al. (7.3 days, ranging from 5.1 to 8.9 days depending on infection type and whether or not the infection was complicated)
. The LOS for patients with cellulitis or SSI, the two infection types examined in our study that were also included individually in the Lipsky study
, were slightly higher in our study, probably because many of our patients would have been considered to have a complicated infection. The LOS found in our study is also similar to that found by Zilberberg et al.
, (9.4 days for patients with healthcare associated infection and 5.5 days for community associated infection) and Edelsberg et al.
 (5.2 days for those successfully treated and 10.6 days for those who failed treatment). The in-hospital mortality rate in the overall cohort in this study (0.4%) is, however, considerably lower than the rates reported in the literature (0.8%-5.2%)
[14, 15, 24, 27, 28]. This could be related to differences in patient populations, infection types or severity, or how the infections were managed.
Our study has several limitations. First, there is a possibility of selection bias in the patients enrolled or hospitals selected, notwithstanding our considerable efforts to minimize this by establishing a priori definitions and standardized data collection procedures. Secondly, we did not recognize the need to provide the investigators with specific enough criteria to identify DFI or SSI until after patient enrollment had begun. Initially, we intended for the investigators to use their discretion when enrolling patients. While using this method would have added to the naturalistic design of the study and resulted in enrollment of a more heterogeneous population, several sites asked us questions about what constituted a DFI or SSI. Thus, we sent out clarification to all the sites on how to define these infection types (see the appendix). Third, although the data from this study underwent a rigorous review and query process, the data accuracy and completeness ultimately depended on the investigators at the study sites. To minimize this problem, we made efforts to sign up experienced clinicians and trained them on data collection in advance of the study. Fourth, clinical outcome was based on the investigators’ subjective assessment and not on a change in the wound score or microbiological response at the end of therapy. This was necessary, as there are no agreed upon criteria for the wound score and the presence and resolution of cSSTIs are based on clinical (not microbiological) criteria. Finally, our findings may not be generalizable to non-hospitalized patients, to patients hospitalized in other countries, to other types of cSSTIs not included in our study, or to patients who had conditions that were excluded from our study.
Strengths of this study include the prospective observational design, the variety of hospital types, the training of the investigating clinicians, the attempt to define types of cSSTIs, the development of standardized CRFs, the use of a validated wound scoring system and the large number of patients enrolled.