In this study we found that significant differences between the three acquisition classifications in relation to demographics, co-morbidities, microbiology, and hospital course and mortality outcomes. On one hand, the characteristics of HCA-BSI were similar to HA-BSI as regards age, the proportion of MRSA, and poly-microbial infections. On the other hand, HCA-BSI resembled CA-BSI in length of stay and the proportion of males. However, features of HCA-BSI were intermediate between HA and CA-BSI as regards infecting species, distribution of co-morbidities, and mortality (Tables 1 and 2). Overall, we consider HCA-BSI to represent a category that is intermediate in clinical, microbiological, and outcome characteristics between HA-BSI and CA-BSI among patients requiring admission to hospital. This study confirms HCA-BSI as a distinct acquisition category and further supports its use in surveillance and in epidemiology studies of infection.
We modeled our study definitions based on those proposed by Freidman et al  and found generally similar findings with their landmark study. Among their cohort of 504 patients with BSI, they found that E. coli and S. pneumoniae were most common in CA, and that S. aureus was the most common in HA and HCA disease. They reported that MRSA occurred with similar frequency among HCA and HA but much lower in CA-BSI. Our results are consistent with these findings with the notable exception that E. coli was found to be the most common cause of HCA-BSI in our study (Table 2). In addition, they found that the distribution of co-morbid illnesses was similar among HA- and HCA-BSI and higher than in CA-BSI, and that hospital length of stay was much longer in HA-BSI but similar in HCA- and CA-BSI. Furthermore, they observed that in-hospital mortality rates were higher in HA-BSI (30%), as compared to HCA-BSI (20%), and CA-BSI (13%). Although clinically significant, they did not find statistically significant difference between HCA-BSI and CA-BSI in-hospital mortality that was likely related to limited statistical power. While our study mirrors these differences in mortality rates between the acquisition categories, our more than 10-fold larger sample size had adequate power to detect statistically significant differences in mortality outcome between each acquisition category both with respect to in-hospital death as well as 28-day mortality.
A number of other studies have investigated alternate definitions for community onset BSI or have evaluated HCA-BSI in different populations of patients . Siegman-Igra proposed four main different categories (with five other subgroups) for community onset BSI and found a number of different demographic, clinical, and outcome differences among 569 patients admitted to a large university hospital in Tel Aviv in 1997 .
Son et al investigated more than 1,144 BSI isolates obtained from 9 university hospitals in Korea . As in our study, they found that E. coli was the most common pathogen in HCA- and CA-BSI and S. aureus in HA-BSI, that co-morbidities were most prevalent in HA- and HCA-BSI, and that there was a decreasing risk for 30-day mortality from HA- (23%), to HCA- (18%) to CA-BSI (10%). Valles et al investigated 1,157 episodes of BSI among adults admitted to three Spanish teaching hospitals and found that as compared to patients with CA-BSI, patients with HCA-BSI were more likely to have MRSA infection and suffered significantly higher mortality . McDonald and colleagues investigated 466 BSI at three North Carolina hospitals and found that as compared to CA-BSI, HCA-BSI was an independent predictor of inadequacy of treatment and this was most strongly related to prior hospitalization within the prior 90 days . Shorr et al investigated a cohort of 6,697 patients admitted to 59 hospitals in the United States and found that HCA-BSI had more co-morbid illness and a higher proportion of MRSA as compared to CA-BSI . In addition, they found that the length of stay was similar between HCA-BSI and CA-BSI but less than for HA-BSI, and that the mortality rates for CA-BSI of 10% was significantly less than for HCA-BSI (15%) and HA-BSI (15%). It is notable that they did not find a difference in mortality among HA-BSI and HCA-BSI and this may be explained in part by their focus on inclusion of patients with cultures obtained in the first five days after admission to hospital. Kollef and colleagues compared HCA-BSI and CA-BSI among 1,143 patients admitted to 7 American hospitals and found that patients with HCA-BSI were older, were more likely to be male, presented with higher severity of illness, had longer lengths of stay, and were more likely to die . They did not examine HA-BSI. A number of other studies have investigated the category of HCA-BSI but these have been limited to restricted populations or infecting etiologies [15–18].
Despite the growing body of evidence supporting the unique nature of HCA-BSI, there is still failure to universally recognize this classification in contemporary BSI definitions and literature, including the current Centers for Disease Control definition for healthcare associated (previously nosocomial) infections . Using these definitions, BSI detected after 48 hours of hospitalization is designated HA, with the remainder of all episodes classified CA-BSI by default, including those that would fall into the proposed HCA-BSI category. An exception is the MRSA classification, which includes a distinct healthcare associated - community onset category analogous to HCA-BSI .
Identification of the HCA-BSI category has a number of implications for surveillance. Hospital-acquired infections are an important measure of quality of care. With the shift towards increasing delivery of complex care in the community setting, decreasing numbers of cases of HA-BSI may be falsely viewed as a marker of success if HCA-BSI's are not tracked to ensure that there is not a concomitant increase in these infections. Similarly, identification of emerging resistant organisms in the community may be falsely attributed to some community based factor (such as travel, food exposures, environmental contamination) if a healthcare associated linkage is not recognized . Perhaps most importantly, failure to recognize that HCA-BSI have higher rates of resistance than CA-BSI may lead to adverse outcomes as a result of increased risk for treatment failure [6, 8].
While our study benefits from large sample size and comprehensive inclusion of all cases occurring in a large well-defined population there are some limitations that are noteworthy. We obtained all of our patient information from our Electronic Surveillance System that is based solely on source data available from regional and provincial databases . We did not have data on specialized home care, immuno-suppressive medication use, and ambulatory parenteral antibiotherapy. As a result, we had to make minor modifications of the definitions of Friedman et al  and did not perform a case-by-case individual chart or patient review in this study. Thus, it is possible that our assignment of patients to acquisition categories may not have been as precise as if we had performed a case-by-case individual review and subsequently patients may have been misclassified. However, our classification of acquisition by the ESS has been previously demonstrated to be highly accurate based on a retrospective comparative review of 306 patients in development , and a further 300 plus cases in validation (manuscript in preparation). It is important to note that while our results are comparable to the study by Friedman because of the similar definitions used , generalization and comparison with other studies that used different definitions should be done cautiously [3, 6–10]. Another limitation of our study is that we did not have information on treatments prescribed as it would be interesting to evaluate whether the HCA-BSI category were at increased risk for inadequate empiric antibiotic therapy. While we report selected resistance rates in this study, we do not present overall group specific resistance rates and therefore may not be recognizing further important differences in acquisition categories. A third limitation is that we did not assess sources of infection, and their may be significant differences in the distribution of foci of infections among the different acquisition classifications. Finally, like all of the other studies conducted to date, we only included patients admitted to hospital. The possibility does exist that HCA-BSI and CA-BSI differ further in this non-admitted population.