To the best of our knowledge this is the first study to provide nationally population-based representative estimates of the epidemiological characteristics and recent trends of severe sepsis in Spain. Using information drawn from all acute-care hospitals nationwide from 2006 to 2011, this study demonstrates the growing burden of severe sepsis in Spain in recent years with a remarkable increase in its incidence and mortality rates. On the other hand, it shows a significant decreasing trend in case-fatality rates over time despite an increase in the age and severity of the affected population.
This analysis indicates that severe sepsis represents 1.1% of all hospitalizations in our country within the study period with an overall incidence rate of 87 cases per 100,000 population. Although it is well known that there is a substantial worldwide variability in the occurrence of severe sepsis , our figures are close to those found in related studies from the USA - and Australia . Also, the incidence rate is in accordance with the estimation of 50-100 cases per 100,000 population recently reported for developed countries .
Comparison with similar Spanish studies, while local, show that our figure is greater than that observed in the Community of Valencia between 1995 and 2004  but lower than those described in two studies in the Region of Madrid ,. In one of these studies, we documented a rate of 141 cases per 100,000 population in 2001 . However, that study was limited to a single year and to a very specific region where large hospitals concentrate, which could explain fairly well the differences encountered. Our present work estimates the national incidence and allows us to know the extent and impact of severe sepsis nationwide overcoming the limitations of local data.
Our results also demonstrate that from 2006 to 2011 the standardised incidence rate experienced an upward trend with an annual increase of 8.6%. Though the interpretation of this increasing trend is not easy and may be in part confounded by factors such a greater awareness of severe sepsis, the introduction of specific ICM-9 codes that facilitate its coding in medical records, or other methodological issues , there are possible reasons for a real rise in the incidence of severe sepsis including, among others, the increased age and comorbidities of the population and the greater use of invasive procedures and immunosuppressive drugs ,,. Indeed, factors related to this increase remain an open question that our study is unable to resolve but our figure is rather close to that described by authors such as Martin  and Dombrosvky in the USA . In addition, other Spanish studies, although not focused on severe sepsis, have observed a clear rise in sepsis hospitalizations in the last few years ,.
As for the demographic and clinical characteristics, our results agree, in general, with international data. Thus, we find that severe sepsis is more common in men -, who also develop it at an earlier age than women. Though the explanation for this finding remains uncertain, it is in agreement with Martin’s report  where men were 4-6 years younger than women. Additionally, almost two thirds of the cases are over the age of 65, with a similar percentage presenting underlying comorbidities ,, and regarding the number of organ failures, our results are similar to those found in international literature where about 58%-81% of cases of severe sepsis present one organ dysfunction and just 6%-16% show three or more ,,.
Microbiological data indicate that less than a third of the cases have identified pathogens, with bacteremias coded in 16% of cases overall. Though it has to be recognized that national databases may fail to depict these data accurately, our figures fully agree with that observed in a recent study from our country . In this prospective but short-term study performed in 2003 at three hospitals in Madrid only 29% (95% CI 25-32) of microbiological samples were positive, whereas a total of 12.5% of cases had positive blood cultures.
As regards to the causative organisms, Gram-negative bacteria were the most frequently involved microorganisms and there was a remarkable increase in the number of these organisms from 2006 to 2011. Though these findings could differ from the common literature that indicates that gram-positive organisms have become the most common cause of sepsis in the past 25 years ,, they are in line with more recent data  and fully agree with those found in other studies in our country. In fact, our results confirm the microbiological profile found by Esteban et al in Madrid in 2003  and the rising trend of sepsis caused by gram-negative bacteria found by Andreu et al in the Region of Valencia from 1995 to 2004 .
Additionally, our analysis shows that, over the study period, cases of severe sepsis occur in older people, who have more comorbidities and develop a greater number of organ failures. These changes, similar to those described in reputable USA studies ,, seem to reflect the ageing and chronic morbidity of western societies  and paint a picture of greater frailty and severity of disease over time.
As previously reported -,,,, age, comorbidities and the number of organ dysfunctions are factors significantly associated with the risk of in-hospital death in our cases. However, our fatality close to 43% is high, greater than the 21%-33% described in older studies ,,, or in local studies performed during a very short period . Nevertheless, comparisons with recent studies from our country show that our rate agree with the 42.5% reported by Andreu  in the Region of Valencia between 1995 and 2004, and with the results of Ayala-Rodriguez  (40% in men, 39% in women) in the region of Madrid between 2003 and 2011. Although we realize that we cannot entirely rule out the possibility that the cases diagnosed as severe sepsis are the most critical cases for which mortality rates are higher ,, nor exclude that withholding or withdrawing of treatments have occurred, our figures are visibly lower than those recently reported in the ICU setting ,.
Additionally, our results show a significant decreasing trend in case-fatality rates with an average decrease of 3.5% per year from 2006 to 2011, a pattern that, in the above depicted setting of greater age and frailty of the affected population, we consider particularly important.
These data add to the most recent literature showing a decreasing trend of case-fatality in severe sepsis despite a substantial increase in the number of deaths related to the process -, a finding that our study also confirms. As regards case-fatality rates, Dombrovsky reported annual decreases of 1.4% between the years 1993 and 2003  whereas Lagu observed an average annual decline in adults of 2% from 2003 to 2007 . Our rate of decline is somewhat higher, but differences in the populations analyzed and the methodology used for estimation of trends, joinpoint regression in our work, may at least partially explain the differences. We must also recognize that notable educational clinical initiatives promoting best practices in the management of severe sepsis have been developed in recent years in Spain , which quite possibly have been beneficial in reducing our case-fatality over time.
Lastly, a finding we consider remarkable in our analysis is the gender-related disparities in the incidence and mortality of severe sepsis. In our study, although women suffer less from severe sepsis than men, trend analysis showed a higher increase in both incidence and mortality rates from 2006 to 2011. Additionally, joinpoint analysis showed a lower decline of CFR in women than in men. Prior studies have also found similar disparities -, but the possible impact of gender in the incidence and outcome of severe sepsis is a highly controversial topic and the reasons for these differences remain unclear ,. Although, in our study, age could play an important role, these disparities require a more in depth and detailed analysis which we are in the process of carrying out.
Among the strengths of this study we can emphasize its nationwide nature. While there is not national surveillance for severe sepsis in Spain, this study captures all acute-care, public and private, hospitalisations nationwide and serves as an adequate proxy. The data source used is considered the main information source to carry out epidemiological studies and health services research in our country as it includes data on over 97% of all annual hospital admissions . To avoid errors due to changes in the ICD9-CM system we chose a period where the selected codes for severe sepsis, including the codes 995.92 and 995.91 that became effective in our country in 2004, suffered no change. Furthermore, the joinpoint regression used for analysis is a robust, widely used and validated methodology to evaluate trends. Moreover, we followed the guidelines for reporting observational studies, as outlined by the STROBE Initiative . Thus, we firmly believe this study provides potentially robust data on the characteristics and impact of severe sepsis in our country. This information can be useful in the design and application of educational and therapeutic programmes to promote the early identification and treatment of patients with severe sepsis that can improve the quality of healthcare and enable a cost-effective use of health resources. At the same time, our data represent a fundamental working basis to evaluate the impact of severe sepsis in our country and estimate future needs.
These data are complementary to recent reports of trends in severe sepsis epidemiology published in North America and might be useful in other European countries with a population similar to ours.
We must also acknowledge possible limitations of our work. This study is subject to the limitations inherent in retrospective studies using administrative databases. Data from these databases lack of many measures obtainable only from chart review. Furthermore, these data do not allow causal inferences to be made. Concerning the potential risk for bias based on coding practices although we recognize that the study would be enhanced by clinical validation of diagnostic codes, access to clinical histories in our country is restricted by law , and these may not be consulted where the aim is research, even if it is epidemiological in nature. However, the use of such databases is well-established in severe sepsis epidemiology -,-, and has been shown to furnish valuable information for assessing the need for preventive and therapeutic care and for service planning -,. Further, the results of a recent meta-analysis clearly support the use of administrative data to monitor mortality trends in severe sepsis , reassuring the essential role of a consistent use of national administrative data for epidemiological monitoring of incidence and outcomes . Additionally, the chief advantage of the hospital discharge database that underpins this study is that information is recorded by specialised medical coders based on a detailed review of the medical chart, increasing the likelihood of accurate documentation , and we must emphasize that quality assurance audits are systematically conducted on this national database to verify the adequacy of coding .
However, it should be recognized that although the use of ICD-9 codes has been shown to be highly sensitive for severe sepsis , the extracting coding strategy is not standardized and inconsistencies between various methods remain ,,. Different algorithms used to identify severe sepsis cases in administrative data, select cases of varying disease severity sometimes resulting in estimates, with low specificity, which tend to underrepresent the real cases of severe sepsis . With this data in mind, to carry out this study we have selected and used the combination of infection and organ dysfunction codes initially described by Martin , and later extended by Dombrosky et al. , after the introduction of the new ICD-9 criteria-specific codes for sepsis, severe sepsis and septic shock. The codes employed by Martin et al  have shown a high specificity for capturing severe sepsis cases ,, with a positive predictive value of 97.7% . In addition, code 995.92 has a specificity close to 100% . However, as recent publications , point out, the use of this code is limited and not recommended as a single code for the estimation of national data, but most always be combined with ICD-9 codes for infection plus organ dysfunction. Given the codes selected and the results achieved, both with respect to the incidence and mortality rates as well as case-fatality, we are confident that the cases included in our study are truly severe sepsis cases and not milder conditions.
Eventually our study may be limited by the changes in the ICD-9 system that have taken place in the last few years. Similar to what other authors believe ,,, we should consider the possibility that these changes in the ICD-9 system may have had an influence in the data and trends observed. However, as explained above, we chose a period where the selected codes for severe sepsis suffered no change in our country. In addition, the joinpoint analysis shows a progressive increase in the respective annual percentage changes with no inflection points, suggesting that the potential changes in coding have had little effect.
Lastly, we must recognize that the restriction of the study to hospitalized patients may have introduced a bias, as Linde-Zwirble  points out, and our estimates of incidence rates of severe sepsis really correspond to treated incidence. In line with this, the figures of mortality associated to severe sepsis refer to hospitalized cases. A recent publication by McPherson et al. that analyses the mortality database from the National Statistics Office using the ICD-10 coding systems, points out that in England, between 2001 and 2010, 93.4% of all deaths associated to sepsis occurred within hospitals, while the remaining 6.6% took place outside hospitals . We do not know whether the same occurs in our country, but it is possible that our estimates are conservative and underestimate the impact of severe sepsis in our population and our health system.