Sepsis is one of the major causes of in-hospital death worldwide [3–5]. Hospital discharge codes from administrative data are extensively used for studying the epidemiology of sepsis or other diseases and have a substantial impact on epidemiological estimates.
The real burden of sepsis in Italian hospitals is largely unknown. A recent paper that analysed all the conditions mentioned in the death certificates in Italy showed an increase in mortality over the years 2003–2015 [12].
Our study highlighted a high burden of hospitalizations for sepsis and septic shock in Sicily and the percentage of sepsis hospitalization compared with hospital admission for all the causes increased over the observed period, such as the septic shock diagnosis and the mortality for sepsis.
Our finding showed elevated mortality for ICU-admitted adult patients (74%) and septic shock diagnosis (67.7%). An Italian study involving ICUs showed a similar septic shock mortality (69.3%) but a lower overall ICU mortality (19.2%) [14]. In China, another study conducted in the period from 2017 to 2019 estimates over 10 million hospitalizations for sepsis with a neonatal (< 1-year-old) incidence of 8.7% [6], lower than ours (16.2%, 2495/15373). A Japanese retrospective cohort study involving adult patients admitted to ICUs with sepsis from April 2010 to March 2020 showed that in-hospital mortality was higher for hospital-onset than community-onset sepsis (35.5% vs. 19.2%) [10]. The reported mortalities were lower than those of our patients admitted to the ICU (74.0%), however, our H-SDF does not consent to distinguish between in-hospital and community-acquired sepsis.
While in the literature most studies analysed only ICU-admitted patients, we described the epidemiology of sepsis both in ICUs and non-ICUs and evidenced significant differences in mortality, hospital length of stay, and charges, that were higher for ICU-admitted patients.
Over half of the hospitalizations involved Medicine Units, and 12% of admissions regarded paediatric patients (< 16 years).
Recent studies on sepsis epidemiology considered both explicit and implicit cases of sepsis [4–7]. The explicit codes contained the word “sepsis”, “septicaemia”, or “septic shock”, while the implicit were defined by the association of an organ failure code with an infection code. In our work, we preferred to consider only explicit diagnosis codes to avoid the risk of over-coding the sepsis diagnosis. For example, a patient hospitalized for myocardial infarction could develop cardiac heart failure and the prolongation of hospitalization could determine hospital-acquired pneumonia. In this case, the H-SDF present the codes for organ failure and infection, but it is not sepsis.
Valent et al. supposed the probability of under-coding of sepsis by using ICD-9 codes. They did a comparison of hospital discharge diagnoses of ICD-9 codes with laboratory results and found that only half of the patients with laboratory-confirmed bloodstream infections were discharged from the Italian University Hospital of Udine with a diagnosis suggestive of sepsis [15]. However, although it is known that not all cases of bacteraemia are associated with sepsis, the use of ICD codes for epidemiological estimation may be subject to bias.
A prospective multicentre Italian study showed an incidence of sepsis in ICUs of 11.4%, with 32.5% of septic shock. As in our study, the main associated infections were pneumonia, intrabdominal and urinary tract infection, while a slightly minor incidence of septic shock was observed [13].
The above-mentioned studies and others [4, 13, 14], were concord with the high in-hospital mortality for sepsis and septic shock. Patients who died of sepsis tended to be older adults with multiple comorbidities and recent hospitalizations [3].
By statistical analysis based on H-SDF data, we tried to identify factors associated with mortality and reported them by distinguishing the patients according to age and ICU admission. Age, Klebsiella pneumoniae infection, aspergillosis, peritonitis, chronic respiratory failure, and hypernatremia were independent predictors of mortality in adult patients hospitalized for sepsis in non-ICU departments while only age for ICU-admitted adults. In a previous work (2006–2014), septic patients with intra-abdominal infection had the highest mortality, followed by lower respiratory tract infection [17]. The findings were partially conformed with ours, which showed peritonitis as an independent predictor of mortality. For non-ICU-admitted paediatric patients, gastrointestinal infection, diabetes, and anaemia were associated with mortality, while In ICU patients, gastrointestinal infection remained a predictor of mortality. To support our findings, a global burden sepsis analysis mentioned diarrheal diseases among the main causes of sepsis in children younger than 5 years [4].
Septic shock was an independent predictor of mortality regardless of age and admission units. Some previous studies tried to realize a score and detect factors to previse the death risk in patients with sepsis. Lu et al demonstrated that reduced serum cholinesterase activity, elevated total bilirubin, lactic acid, creatinine, pro-BNP levels, and respiratory failure were the independent risk factors for the death of sepsis patients within 24h following hospitalization [18]. In our study, only exam alterations reported in H-SDFs were considered. The lack of exams was a limitation. However, we have posed our attention to comorbidities and infections and individuated some conditions associated with a major risk of death and then we have identified a potential group of patients needing even closer monitoring, such as adults with Klebsiella pneumoniae infection, peritonitis, hypernatremia and chronic respiratory failure and children with diabetes, anaemia, and gastrointestinal infections.
Mirijello et al identified respiratory failure, chronic obstructive pulmonary disease, and haemodialysis as predictors associated with the risk of death or admission to the intensive care unit in internal medicine patients with sepsis [16]. Pieroni et al using an Italian multicentre database showed prognostic models for the three main types of sepsis: pulmonary, urinary, and abdominal sepsis [19].
Consistent with data reported in the literature, sepsis hospitalization had a high charge for the National Health System, with higher costs for dead patients, ICU admission, and septic shock. Other studies showed that expenditures increased with severity [20, 21], conforming with our data highlighting elevated charges associated with septic shock diagnosis.
Agodi et al identified Pseudomonas aeruginosa, Acinetobacter baumanii and Klebsiella pneumoniae as the most isolated microorganisms from sepsis episodes in ICUs [14].
Our data did not provide information about microbiological isolations and infection responsible for sepsis, except for the ICD-9 codes that reported the bacterium. According to these codes, we individuated a pathogen in a quarter of the case, with a slightly higher percentage of Gram-negative bacteria. We also reported the absolute and relative frequencies of infections inserted in the H-SDFs, and pneumonia was the most represented, as described in the literature [4, 13, 14].
Limitations
There were several limitations of this study. Identifying sepsis using the ICD-9 CM codes algorithm may not be as precise as screening with clinical criteria with a risk of over-coding or under-coding. Furthermore, although we considered multiple factors that could influence hospital mortality, there may be other confounding factors that we did not account for, such as comorbidities non-reported in H-SDF. Finally, our research did not present blood exams and complete microbiological data except for some reported in H-SDF, which are probably under-coded.