Trends in the incidence and outcome of sepsis using data from a Japanese nationwide medical claims database-the Japan Sepsis Alliance (JaSA) study group-

It is estimated that, annually, approximately 48.9 million people worldwide develop sepsis, and of these, 11 million (22.5%) die [1]. Multicenter studies have revealed the characteristics and outcomes of Sepsis-3 [2‐4]; however, limited epidemiological investigational studies using a nationwide database have been performed [5]. Japan, which has universal public health care, developed a comprehensive reimbursement system of medical cost, named the diagnosis procedure combination (DPC) system, in 2003 [6]. The claims-based database has data on diagnosis, examination, and treatment; data from this database have been used in epidemiological studies of brain, orthopedic, and abdominal diseases and cancer [7‐10], but no studies have used these data with regard to sepsis.

Between 1979 and 2012 in North America and Europe, epidemiological sepsis studies documented that mortality decreased; however, the proportion of patients with sepsis across inpatients and the number of deaths in sepsis increased [11‐14]. In contrast, a recent Global Burden of Diseases study in the period of 1990–2017 reported a decreased proportion of patients with sepsis across inpatients and decreased number of deaths in sepsis [1], which is inconsistent with previous studies. Recent trends in sepsis may be changing.

Thus, in the current study, the diagnosis of sepsis was defined as a patient who underwent blood culture with a new antibiotic treatment and presented acute organ dysfunction, which is linked with the third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3) [15]. We tested the hypothesis that there is a trend for an increased incidence of sepsis and death in sepsis in the aging society of Japan using a large nationwide claims-based database including data on over 50 million inpatients in 2010–2017. The primary outcomes were annual incidence of sepsis and death in sepsis per 1000 inpatients.

The analyzed dataset included 50,490,128 adult inpatients in total during the 8-year period between 2010 and 2017. Of these, 2,043,073 (4.0%) patients had sepsis, 70.2% of which were community-onset (Table 1; annual data, Additional file 1: Table S4). The age of patients with sepsis significantly increased (P = 0.0005; annual data, Additional file 1: Table S4). Major comorbidities included malignant tumors (34.9%), hypertension (26.3%), and diabetes mellitus (21.8%), with the number of patients with hypertension or diabetes mellitus increasing significantly every year (P < 0.01; annual data, Additional file 1: Table S4). Respiratory infections were the most common source of infection (41.0%), and significantly increased (P < 0.0001; annual data, Additional file 1: Table S4 and S5). Regarding acute organ dysfunction, patients with respiratory failure who used mechanical ventilation/oxygen therapy were the most common, accounting for 82.3% (293,026 patients) in 2017 (Additional file 1: Table S6). 17.1% of patients with sepsis required ICU admission. The median (IQR) length of hospital stay was 29 (16–55) days and the in-hospital mortality rate was 20.1%.

The annual proportion of patients with sepsis across inpatients significantly increased (R² = 0.98, slope = + 0.30%/year, P < 0.0001; Fig. 2). The annual number of patients with sepsis was 355,833 (4.9% of total inpatients) in 2017 (Additional file 2: Fig. S1). The annual deaths in sepsis per 1000 inpatients significantly increased (R² = 0.90, slope = + 1.8/1000 inpatients year, P = 0.0003; Fig. 2), accounting for 7.8 annual deaths in sepsis per 1000 inpatients in 2017. The annual number of deaths in patients with sepsis was 56,905 in 2017 (a 2.3-fold increase compared to 2010) (Additional file 2: Fig. S1).

The in-hospital mortality rate and length of hospital stay in sepsis significantly decreased (R² = 0.95, slope = − 0.95%/year, P < 0.0001 and R² = 0.92, slope = − 1.70 days/year, P = 0.0002, respectively; Fig. 3). In 2017, the in-hospital mortality rate was 18.3%, and the mean length of hospital stay was 41.9 days in 2017. Repeating analysis in the age subgroups revealed that the decreased in the in-hospital mortality rate and hospital stay were significant in all age subgroups (P < 0.0001 and P < 0.01, respectively; Fig. 4a and 4b) (Additional file 3: Fig. S2a and Additional file 4: Fig. S2b). Additional analysis by dividing into units of 10 years also revealed the decreasing trend of in-hospital mortality in all the subcategories; old groups have more downslope compared to young groups (Additional file 3: Fig. S2a).

In the present nationwide sepsis study, the annual incidence of sepsis and death in inpatients with sepsis significantly increased during the 8-year study period (2010–2017) in Japan. The in-hospital mortality rate and length of hospital stay significantly decreased during this period.

The proportion of patients with sepsis across inpatients increased from 0.99 to 2.38% during 2000–2007 in the US [12], from 1.2 to 2.7% during 2005–2014 in the US [22], from 1.21 to 1.54% during 2007–2013 in Germany [23], and from 1.3 to 2.1% during 2008–2012 in Spain [24]. In accordance with these results, we found the proportion of patients with sepsis across inpatients significantly increased from 2.9 to 4.9% between 2010 and 2017. The incidence of sepsis in the present study is relatively higher compared to those in the US/Europe. Differences in study region or methods of sepsis selection potentially provides the differences. In line with these frequencies per inpatients, population-level incidence rates for sepsis are also increasing but different among regions. In the US, those increased from 346/100,000 to 436/100,000 (2008–2012) [5], in South Korea from 173.8/100,000 to 233.6/100,000 (2007–2016) [25], and from 637.8/100,000 to 772.1/100,000 in Taiwan (2002–2012) [26].

Previous studies of severe sepsis using the definition prior to the publication of “Sepsis-3” reported an increased number of deaths in sepsis. In 2011, Kumar et al. reported an increase in the number of deaths in inpatients with sepsis between 2000 and 2007 (8 years) in the US, from 4.0 deaths/1000 inpatients in 2000 to 6.5 deaths/1000 inpatients in 2007; the annual absolute number of deaths increased 1.8 times (213,124 deaths in 2007) [12]. In 2012, Lagu et al. reported an increased number of deaths between 2003 and 2007 (5 years) in the US, from 4.9 deaths/1000 inpatients in 2003 to 6.3 deaths/1000 inpatients in 2007; the annual absolute number of deaths increased 1.3 times (207,427 deaths in 2007) [12, 27]. In accordance with these results, we found a significant increase in the number of annual deaths in inpatients with sepsis (Slope =  + 1.8/1000 inpatients year, P = 0.0003), an absolute increase of 2.3 times (56,905 deaths in 2017). In contrast, a recent Global Burden of Disease study, which extracted sepsis using the ICD codes, showed a decreased number of deaths in sepsis worldwide [1]. Different regions (worldwide vs. Japan) or different extraction methods (diagnosis data only vs. diagnosis plus examination plus treatment data) may yield different results.

Substantial studies reported that the annual mortality of sepsis decreased. A meta-analysis of 36 multicenter trials of severe sepsis showed a decrease in mortality from 46.9% between 1991 and 1995 to 29% between 2006 and 2009 [28]. A US nationwide study showed decreased mortality of severe sepsis from 39 to 27% during 2000–2007 [12]. Similar trends have been observed in Asian countries, with mortality rate decreasing from 30.9% in 2007 to 22.6% in 2016 in South Korea [25] and from 27.8% in 2002 to 22.8% in 2012 in Taiwan [26]. In accordance with these findings, we found that rate of mortality in sepsis decreased from 25.0 to 18.3% between 2010 and 2017. Furthermore, in age subgroup analysis by dividing into units of 10 years, the elderly had more annual improvement of mortality compared to the young (≥ 80 years: − 1.11%/years, P < 0.0001; 20–29 years: − 0.31%/years, P = 0.011). In Rhee’s study in the US [3], 54.7% of patients with sepsis required ICU, while only 17% of those required ICU in the present study. Japan has a low number of ICU bed (5 per 100,000 people) compared to the US (20 per 100,000 people) [29], which may reflect this low frequency of ICU admission. The length of hospital stay also decreased during the study period. In an epidemiological study of sepsis in Spain, the average length of hospital stay decreased from 15.7 to 14.0 days during 2000–2013 [30]. In the US studies, the average length of hospital stay decreased from 14.0 to 12.5 days during 2000–2007 [12] and 16.1 to 10.7 days during 2005–2014 [22]. In accordance with these findings, we found that the average length of hospital stay decreased from 51.3 to 41.9 days between 2010 and 2017. These improvements in mortality and length of hospital stay in sepsis may be attributed to the widespread use of the international and Japanese sepsis guidelines, which potentially contribute to improvements in the recognition of sepsis, faster treatment, and improvement in care for critically ill patients [31‐34]. In comparison of length of hospital stay in patients with acute care among 34 American, European, and Asian countries in 2019, Japan had extremely long hospital stays (average 16.0 days) compared to all other countries (average ranged from 4.1 to 9.3 days) [35]. In Rhee’s study of patients with sepsis in the US, the median (IQR) hospital length of stay was 10 days (8–12) in 2014 [3]. A Japanese multicenter observational study of severe sepsis (2016–2017, n = 1184) reported long hospital stays [median (IQR) 24 (12–46)] [4]. In accordance with these, we found the length of hospital stay in patients with sepsis was long [median (IQR) 29 (16–55)]. Japanese health care system may involve these longer hospital stays [36].

We acknowledge that our study has some limitations. First, the DPC data did not include laboratory data. Since the definition of sepsis requires laboratory data, including blood levels of creatinine and bilirubin and platelet counts, our extraction approach using diagnosis coding and treatment data on organ supports may underestimate the incidence compared to a more stringent extraction approach using laboratory data according to Sepsis-3. On the other hand, a recent report [37] found that the Rhee criteria [3] using a combination of culture and antibiotic orders approach is significantly less sensitive but more specific compared approaches to defining suspected infection via EHR data. We would expect the current study using suspected infection definition to be relatively specific and identify a more severely ill population. Second, due to the unavailability of blood lactate data, we could not analyze septic shock according to Sepsis-3. Third, the current study is retrospective using administrative database. Fourth, although the DPC is the largest nationwide dataset, the datasets only covered 71.5% of all acute care hospitals. Fifth, the median age of 76 years with the high prevalence of malignancy (35%) limits the generalizability of this study. Sixth, the median length of antibiotic treatment analyzed data between 2010 and 2017 in the present study was 12 days, which was relatively longer than those (median 10 days) in a recent randomized control trial of sepsis in between 2017 and 2019 [38]. This study included substantial patients with respiratory infections (41.0%) and elderly (≥ 65 years old, 78.6%). Since respiratory infections in the elderly have repeated fevers and longer antibiotic treatment [39, 40], these characteristics may partially contribute to the long duration of antibiotic treatment. Older study period may influence the difference in the longer durations of antibiotic treatment. In addition, the incidence of the blood stream infection was low (< 1%). Since we could not obtain blood culture results, we defined the blood stream infection in the diagnosis coding approach, which may underestimate the incidence of the blood stream infection. Seventh, the value presented in the age subgroup analysis related to in-hospital mortality rate was actually a crude rate without any adjustment for patient severity or other factors due to data availability.

The in-hospital mortality rate and length of hospital stay in patients with sepsis significantly improved between 2010 and 2017; however, the incidence of sepsis and death in inpatients with sepsis has been increasing annually in Japan. Sepsis appears to be a significant and ongoing issue. Health systems may consider countermeasure for reduce sepsis incidence.