Sepsis and septic shock case identification from electronic health records: an open-source workflow and comparison of cohorts by criteria Authors: Seth R. Bauer, Lyla Mourany, Paul R. Gunsalus, Alex Milinovich, Sandra L. Kane-Gill, Xiaofeng Wang, Yasir Tarabichi, Vidula Vachharajani, Jarrod E. Dalton Published: 2025 ABSTRACT BackgroundDeploying clinical criteria to identify patients with sepsis from the electronic health record is challenging. We sought to build an open-source data workflow for sepsis and septic shock case identification using three distinct criteria and assess the identified cohorts.MethodsWe developed a workflow from raw data queries through case identification using the Centers for Disease Control and Prevention Adult Sepsis Event (CDC ASE), Sepsis-3, and International Classification of Diseases criteria. The workflow was then applied to patient encounters between 2012 and 2024 across 25 emergency departments and hospitals in an integrated health system. Separate cohorts of sepsis and septic shock patients by criteria were derived, restricted to a patient’s first sepsis encounter, and compared. Pairwise agreement was evaluated with Cohen’s kappa (κ) and Jaccard index.ResultsOur programming code for the complete workflow is available on a public platform. We identified 302,112 patients with sepsis and 59,043 with septic shock who met one or more of the three criteria. Multiple criteria were satisfied for only 35% of patients meeting at least one sepsis criterion and 45% of patients meeting at least one septic shock criterion. Pairwise agreement was the highest between CDC ASE and Sepsis-3 criteria for both sepsis (κ = 0.39, Jaccard index = 0.31) and septic shock (κ = 0.53, Jaccard index = 0.37). Characteristics and outcomes of patients differed across cohorts identified with the three criteria. Patients meeting both CDC ASE and Sepsis-3 criteria for sepsis and septic shock were more severely ill and had high in-hospital mortality incidence.ConclusionsOur open-source workflow is useful for individuals seeking to build and analyze sepsis and septic shock cohorts. Agreement between case criteria was fair to moderate. Clinicians, researchers, and quality improvement practitioners who identify patient cohorts with sepsis and septic shock retrospectively should consider applying multiple criteria.Graphical abstract Supplementary InformationThe online version contains supplementary material available at 10.1186/s13054-025-05749-1. Background The 2021 Surviving Sepsis Campaign Guidelines strongly recommend healthcare systems use a performance improvement program for sepsis, including sepsis screening [1]. Identification of patients (cases) with sepsis underpins these programs. Yet, multiple clinical criteria exist to operationalize the definitions for sepsis and septic shock [2–4]. International Classification of Diseases (ICD) coding is commonly used for retrospective studies due to its simplicity for classification, but it is subject to provider documentation, interpretation by coders, and bias from increasing sepsis awareness [4, 5]. For these reasons the US Centers for Disease Control and Prevention (CDC) recommend using Adult Sepsis Event (ASE) criteria [4]. Yet, the Third International Consensus Definition for Sepsis and Septic Shock (Sepsis-3) clinical criteria are often considered the gold standard by researchers [2, 3]. While consensus is lacking for the optimal case identification approach [6], some use the CDC ASE and Sepsis-3 clinical criteria interchangeably, particularly for sepsis surveillance [7]. Importantly, case identification criteria have varying diagnostic performance compared with manual chart review, which is typically not considered by quality improvement programs, researchers, or consumers of the literature [5, 8–11]. Deploying clinical criteria to identify sepsis case patients from the electronic health record is challenging [6]. Data flow architecture (i.e., a processing workflow) is required for extraction, transformation, and loading into a repository for application of criteria logic. Despite the availability of detailed guidance for sepsis case identification [3, 4, 12], variations and inadequate methodological descriptions are common. Published manuscripts utilizing electronic health record data for sepsis cohort identification typically do not describe the detailed data workflow steps, which are crucial for replication. Further, variations in the case identification approach can lead to substantial differences in mortality and criteria predictive validity [12]. Although exemplar publications exist for workflow descriptions, the programming code employing the workflow for cohort identification is unavailable in a public repository or accompanying the manuscript [3, 5, 13, 14]. Crucially, even in exemplar publications and an existing open-source pipeline for Sepsis-3 case identification (“OpenSep”) [15], programming code for data extraction from the electronic health record is not included. The availability of programming code for all components of a workflow is essential not only for reproducibility, but also for understanding data processing steps and the management of decision points for ambiguous criteria components. Consequently, there were two co-primary goals for this work. First, we sought to derive sepsis and septic shock cases from the electronic health record, openly describe all process steps, and provide publicly available programming code to locally reproduce our workflow. Accordingly, we outline in full transparency the complete data workflow methodology used, starting from extraction through criteria application, to ensure rigor and reproducibility. Second, we sought to describe and compare sepsis and septic shock cohorts identified with CDC ASE, Sepsis-3, and ICD criteria. Methods Data source We queried patient encounters from 2012 to 2024 across 25 emergency departments and hospitals in the Cleveland Clinic integrated health system. Cleveland Clinic uses Epic® (Epic Systems Corporation, Verona, WI) as its electronic health record vendor, and data were extracted from Epic’s analytical reporting server Clarity®. Patient encounters were included in the query if they had one or more microbial culture resulted and one or more antimicrobial dose administered during the encounter. Patients were excluded if they had one or more encounter with inconsistent time stamps across linked data tables. This exclusion criterion and other methodological details are explained further in the Supplementary Material 1: eMethods. Queries of data elements required for sepsis case identification were first performed on encounters meeting inclusion criteria. Subsequently, the exclusion criterion was applied, and sepsis cases were identified utilizing a data workflow. The study was approved by the Cleveland Clinic Institutional Review Board (approval number 22–212) with a waiver of informed consent. Open-source data workflow development We built a data workflow for retrospective sepsis case identification consisting of three core components: (1) queries across electronic health record data tables, (2) transformation of extracted data, and (3) application of sepsis and septic shock case identification criteria. The data flow architecture from raw query output through case identification can be visualized in Supplementary Material 1: eFigure 1. Data extraction steps are outlined in the Supplementary Material 1: eMethods and Supplementary Material 1: eTable 1. Extracted data were ingested, transformed, cleaned, and stored in a relational database using R software (version 4.3.2, R Foundation for Statistical Computing, Vienna, Austria) with the Quarto markdown system (Posit Software, Boston, MA) on Posit Workbench (Posit Software, Boston, MA). The rationale for data transformation decisions is integrated as annotated comments through the programming code along with validation checks for data transformations. Supplementary Material 1: eTable 2 outlines the number of observations (range 37,404 to 134,853,957), encounters, and patients in each query output file that were transformed and loaded for sepsis criteria evaluation. Subsequently, we applied the case identification procedures. We primarily identified sepsis and septic shock cases based on CDC ASE and Sepsis-3 clinical criteria (Table 1). For completeness and comparison, we queried explicit sepsis and septic shock ICD codes (Supplementary Material 1: eTable 1). To optimize fidelity with CDC ASE and Sepsis-3 criteria we closely aligned our case identification procedures with criteria source documents without adaptations [2–5, 16]. Our approach to deriving and assessing each component of the CDC ASE and Sepsis-3 criteria is delineated in detail in annotated Quarto markdown (programming code) documents. For transparency, specific elements that are inconsistently applied or inadequately described in the published literature are outlined in the Supplementary Material 1: eMethods. We used an open science framework to facilitate deployment at external sites, reduce variability, and enhance adoption [17, 18]. All programming code files for data extraction, transformation, and cohort identification with accompanying annotations as well as the analytic code used for reporting in this manuscript are publicly available on GitHub (GitHub, Inc., San Francisco, CA; https://github.com/ClevelandClinicQHS/sepsis-case-identification-workflow). Table 1Overview of the applied sepsis and septic shock criteriaCriteriaInfectionOrgan dysfunctionTimingCommentsCDC ASEPresumed infection: Blood culture obtained and at least four qualifying antimicrobial daysaThe first qualifying antimicrobial day must be within the time epoch from two calendar days before through two calendar days after blood culture collectionAcute onset of organ dysfunction within one or more of six organ systems per criteria (also called eSOFA criteria)Criteriab• Cardiovascular (vasopressor initiation)• Respiratory (mechanical ventilation)• Kidney (creatinine ↑≥100% from baseline)• Liver (bilirubin > 2 mg/dL and ↑≥100% from baseline)• Coagulation (platelets < 100 cells/mcL and ↓≥50% from baseline)• Lactate ≥ 2 mmol/L (optional criterion)Septic shock criteria are met by fulfillment of cardiovascular criteria (vasopressor initiation)Organ dysfunction criteria met within the time epoch from two calendar days before through two calendar days after blood culture collectionEligible antimicrobials for presumed infection include antibiotics, antifungals, and antiviralsBaseline kidney, liver, and coagulation organ function is determined using values throughout the encounter, with differing approaches for community- and hospital-onset eventsSepsis-3Suspected infection: Microbial culturec obtained and > 1 dose of antibiotic administeredFirst antibiotic dose must be within the time epoch from 24 h before through 72 h after microbial culture collectionIf suspected infection criteria are met, onset of infection is defined as the date/time of the first of either microbial culture collection or antibiotic administrationAcute increase in total SOFA score ≥ 2SOFA sub-scores are measured within cardiovascular, respiratory, kidney, liver, coagulation, and central nervous systems, with each organ system scored from 0 to 4, and higher scores indicating greater dysfunction. Total SOFA score is the sum of sub-scores from each organSeptic shock criteria (all components must be met):• Vasopressor administration• Mean arterial pressure < 65 mm Hg• Lactate > 2 mmol/LOrgan dysfunction criteria met within the time epoch from 48 h before through 24 h after the onset of infectionOnly antibiotics are eligible for suspected infection adjudicationBaseline SOFA can be assumed to be zero in patients not known to have pre-existing organ dysfunctionExplicit ICD codingAdjudicated by medical coding professionals based on chart documentation and application of coding guidelinesAdjudicated by medical coding professionals based on chart documentation and application of coding guidelinesAdministrative codes applied at admission, within the encounter, and at dischargeDiagnosis date in the electronic health record matches admission dateaQualifying antimicrobial days include specific criteria that must be met (e.g., a new antimicrobial [not administered in the prior two calendar days] must be administered to be regarded as the first qualifying antimicrobial day). Presumed infection criteria can be met with fewer than four qualifying antimicrobial days in specific scenarios. See the CDC ASE criteria document (reference [4]) for detailsbOutlined organ dysfunction criteria are representative examples and may not reflect the complete criteriacMicrobial culture of fluids from specific anatomical sites (e.g., blood, lung, urinary tract) are considered. See the Sepsis-3 criteria documents (references [2, 3]) for detailsCDC ASE = US Centers for Disease Control and Prevention Adult Sepsis Event; eSOFA = Electronic Sequential Organ Failure Assessment; SOFA = Sequential Organ Failure Assessment; ICD = International Classification of Diseases Table 1Overview of the applied sepsis and septic shock criteriaCriteriaInfectionOrgan dysfunctionTimingCommentsCDC ASEPresumed infection: Blood culture obtained and at least four qualifying antimicrobial daysaThe first qualifying antimicrobial day must be within the time epoch from two calendar days before through two calendar days after blood culture collectionAcute onset of organ dysfunction within one or more of six organ systems per criteria (also called eSOFA criteria)Criteriab• Cardiovascular (vasopressor initiation)• Respiratory (mechanical ventilation)• Kidney (creatinine ↑≥100% from baseline)• Liver (bilirubin > 2 mg/dL and ↑≥100% from baseline)• Coagulation (platelets < 100 cells/mcL and ↓≥50% from baseline)• Lactate ≥ 2 mmol/L (optional criterion)Septic shock criteria are met by fulfillment of cardiovascular criteria (vasopressor initiation)Organ dysfunction criteria met within the time epoch from two calendar days before through two calendar days after blood culture collectionEligible antimicrobials for presumed infection include antibiotics, antifungals, and antiviralsBaseline kidney, liver, and coagulation organ function is determined using values throughout the encounter, with differing approaches for community- and hospital-onset eventsSepsis-3Suspected infection: Microbial culturec obtained and > 1 dose of antibiotic administeredFirst antibiotic dose must be within the time epoch from 24 h before through 72 h after microbial culture collectionIf suspected infection criteria are met, onset of infection is defined as the date/time of the first of either microbial culture collection or antibiotic administrationAcute increase in total SOFA score ≥ 2SOFA sub-scores are measured within cardiovascular, respiratory, kidney, liver, coagulation, and central nervous systems, with each organ system scored from 0 to 4, and higher scores indicating greater dysfunction. Total SOFA score is the sum of sub-scores from each organSeptic shock criteria (all components must be met):• Vasopressor administration• Mean arterial pressure < 65 mm Hg• Lactate > 2 mmol/LOrgan dysfunction criteria met within the time epoch from 48 h before through 24 h after the onset of infectionOnly antibiotics are eligible for suspected infection adjudicationBaseline SOFA can be assumed to be zero in patients not known to have pre-existing organ dysfunctionExplicit ICD codingAdjudicated by medical coding professionals based on chart documentation and application of coding guidelinesAdjudicated by medical coding professionals based on chart documentation and application of coding guidelinesAdministrative codes applied at admission, within the encounter, and at dischargeDiagnosis date in the electronic health record matches admission dateaQualifying antimicrobial days include specific criteria that must be met (e.g., a new antimicrobial [not administered in the prior two calendar days] must be administered to be regarded as the first qualifying antimicrobial day). Presumed infection criteria can be met with fewer than four qualifying antimicrobial days in specific scenarios. See the CDC ASE criteria document (reference [4]) for detailsbOutlined organ dysfunction criteria are representative examples and may not reflect the complete criteriacMicrobial culture of fluids from specific anatomical sites (e.g., blood, lung, urinary tract) are considered. See the Sepsis-3 criteria documents (references [2, 3]) for detailsCDC ASE = US Centers for Disease Control and Prevention Adult Sepsis Event; eSOFA = Electronic Sequential Organ Failure Assessment; SOFA = Sequential Organ Failure Assessment; ICD = International Classification of Diseases Overview of the applied sepsis and septic shock criteria Presumed infection: Blood culture obtained and at least four qualifying antimicrobial daysa The first qualifying antimicrobial day must be within the time epoch from two calendar days before through two calendar days after blood culture collection Acute onset of organ dysfunction within one or more of six organ systems per criteria (also called eSOFA criteria) Criteriab • Cardiovascular (vasopressor initiation) • Respiratory (mechanical ventilation) • Kidney (creatinine ↑≥100% from baseline) • Liver (bilirubin > 2 mg/dL and ↑≥100% from baseline) • Coagulation (platelets < 100 cells/mcL and ↓≥50% from baseline) • Lactate ≥ 2 mmol/L (optional criterion) Septic shock criteria are met by fulfillment of cardiovascular criteria (vasopressor initiation) Eligible antimicrobials for presumed infection include antibiotics, antifungals, and antivirals Baseline kidney, liver, and coagulation organ function is determined using values throughout the encounter, with differing approaches for community- and hospital-onset events Suspected infection: Microbial culturec obtained and > 1 dose of antibiotic administered First antibiotic dose must be within the time epoch from 24 h before through 72 h after microbial culture collection If suspected infection criteria are met, onset of infection is defined as the date/time of the first of either microbial culture collection or antibiotic administration Acute increase in total SOFA score ≥ 2 SOFA sub-scores are measured within cardiovascular, respiratory, kidney, liver, coagulation, and central nervous systems, with each organ system scored from 0 to 4, and higher scores indicating greater dysfunction. Total SOFA score is the sum of sub-scores from each organ Septic shock criteria (all components must be met): • Vasopressor administration • Mean arterial pressure < 65 mm Hg • Lactate > 2 mmol/L Only antibiotics are eligible for suspected infection adjudication Baseline SOFA can be assumed to be zero in patients not known to have pre-existing organ dysfunction aQualifying antimicrobial days include specific criteria that must be met (e.g., a new antimicrobial [not administered in the prior two calendar days] must be administered to be regarded as the first qualifying antimicrobial day). Presumed infection criteria can be met with fewer than four qualifying antimicrobial days in specific scenarios. See the CDC ASE criteria document (reference [4]) for details bOutlined organ dysfunction criteria are representative examples and may not reflect the complete criteria cMicrobial culture of fluids from specific anatomical sites (e.g., blood, lung, urinary tract) are considered. See the Sepsis-3 criteria documents (references [2, 3]) for details CDC ASE = US Centers for Disease Control and Prevention Adult Sepsis Event; eSOFA = Electronic Sequential Organ Failure Assessment; SOFA = Sequential Organ Failure Assessment; ICD = International Classification of Diseases Study eligibility criteria and specifications Sepsis criteria adjudication was initially performed within each encounter. Sepsis onset time was specified as the first instance during an encounter the respective sepsis clinical criteria, whether CDC ASE or Sepsis-3, were met. Because ICD codes were applied to an encounter as a whole, sepsis onset time was unable to be reliably discerned in cases identified with ICD coding. To meet CDC ASE or Sepsis-3 septic shock criteria, cases were required to meet the vasopressors criterion (and additional criteria for Sepsis-3) during the requisite time epoch for organ dysfunction. For study inclusion, all cases were restricted to the first sepsis encounter per patient. We limited our analysis to the first qualifying sepsis encounter per patient to remove the possibility of correlated outcomes across multiple sepsis encounters per patient. Further, sepsis survivors may represent a distinctly different pathobiological and outcome risk cohort compared with patients presenting with an initial sepsis encounter [19, 20]. In a sensitivity analysis, we evaluated all sepsis and septic shock encounters without restriction to the first sepsis encounter per patient. Because previous studies assessed construct validity of the CDC ASE and Sepsis-3 clinical criteria with comparisons to manual chart review we did not pursue this approach [5, 8–10]. Statistical analysis We describe sepsis and septic shock cohorts by individual criteria as well as their set union and intersection. We report descriptive statistics without null hypothesis testing; because individual patient encounters may fulfill multiple sepsis criteria the cohorts are not independent, complicating inferences from comparisons. Case identification criteria pairwise agreement was assessed with Cohen’s kappa (κ) and the Jaccard index [21, 22]. Agreement interpretation was based on the original kappa value proposal [22]. R version 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria) was used for all data handling, visualizations, and statistical analyses. Missing data were characterized and imputation was not used. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines [23]. Results Over 8.7 million encounters for 2.3 million patients occurred at study locations between 2012 and 2024. A total of 1,293,782 encounters (665,956 patients) were included in the data extraction query after applying study eligibility criteria (Fig. 1). Of the 25 encounter admission locations, the average location was medium size (median [interquartile range] beds: 203 [100, 323]), 10 (40%) were teaching facilities, 19 (76%) were in Ohio, and 6 (24%) were in Florida. Fig. 1Patient flow diagram. CDC = US Centers for Disease Control and Prevention; ICD = International Classification of Diseases; MAP = mean arterial pressure; Sepsis-3 = Third International Consensus Definition for Sepsis and Septic Shock Patient flow diagram. CDC = US Centers for Disease Control and Prevention; ICD = International Classification of Diseases; MAP = mean arterial pressure; Sepsis-3 = Third International Consensus Definition for Sepsis and Septic Shock After application of case identification criteria and restriction to the first sepsis encounter for each patient, we included 139,267 patients with sepsis per CDC ASE criteria, 234,601 patients with sepsis per Sepsis-3 criteria, and 64,461 patients with sepsis per ICD coding. Within first sepsis encounters, septic shock criteria were met for CDC ASE in 38,387 patients, Sepsis-3 in 21,253 patients, and ICD coding in 35,437 patients. Of patients with sepsis per CDC ASE criteria, 79% had only one encounter where CDC ASE sepsis criteria were met, but 14% had two encounters and 7% had three or more encounters where CDC ASE criteria were met. Similarly, of patients with sepsis per Sepsis-3 criteria, 69% had only one encounter where Sepsis-3 criteria were met, but 17% had two encounters and 14% had three or more encounters where Sepsis-3 criteria were met. There was fair to moderate overlap of patients with sepsis and septic shock identified by CDC ASE, Sepsis-3, and ICD criteria (Fig. 2, Supplementary Material 1: eTable 3). Two or three criteria were met for only 35% of patients meeting at least one sepsis criterion and 45% of patients meeting at least one septic shock criterion. Pairwise agreement was the highest between CDC ASE and Sepsis-3 criteria for both sepsis (κ = 0.39, Jaccard index = 0.31) and septic shock (κ = 0.53, Jaccard index = 0.37) (Supplementary Material 1: eFigure 2). Both CDC ASE and Sepsis-3 sepsis criteria were met for 63% and 38%, respectively, of patients with sepsis identified by the two individual criteria. Further, both CDC ASE and Sepsis-3 septic shock criteria were met for 42% and 76%, respectively, of patients with septic shock identified by the two individual criteria (Supplementary Material 1: eFigure 3). Of the patients meeting either the CDC ASE or Sepsis-3 clinical criteria, explicit ICD coding identified only 17% of those with sepsis and 46% of those with septic shock. Overlap and agreement remained fair to moderate in the sensitivity analysis of all identified sepsis and septic shock encounters without restriction to the first patient encounter (Supplementary Material 1: eFigure 4 and Supplementary Material 1: eFigure 5). Fig. 2Overlap of sepsis and septic shock case identification criteria. Proportional Venn diagrams representing sepsis (A) and septic shock (B) case identification overlap between the three evaluated criteria. Cases were restricted to the first sepsis encounter on the patient level. Septic shock cases are nested within sepsis cases and reflect the occurrence of septic shock within the first sepsis encounter for a patient. In total, 302,112 cases met one, two, or three sepsis criteria. Sepsis criteria were met for only one criterion in 195,410 (65%) of cases and 106,702 (35%) cases met multiple criteria during their encounter (77,187 [26%] cases met two criteria and 29,515 [10%] cases met all three criteria). Sepsis criteria agreement was the highest between CDC ASE and Sepsis-3 criteria (κ = 0.39, Jaccard index = 0.31). In total, 59,043 cases met one, two, or three septic shock criteria. Septic shock criteria were met for only one criterion in 32,235 (55%) of cases and 26,808 (45%) cases met multiple criteria during their encounter (17,582 [30%] cases met two criteria and 9,226 [16%] cases met all three criteria). Septic shock criteria agreement was also the highest between CDC ASE and Sepsis-3 criteria (κ = 0.53, Jaccard index = 0.37). CDC ASE = US Centers for Disease Control and Prevention Adult Sepsis Event; ICD = International Classification of Disease; Sepsis-3 = Third International Consensus Definition for Sepsis and Septic Shock Overlap of sepsis and septic shock case identification criteria. Proportional Venn diagrams representing sepsis (A) and septic shock (B) case identification overlap between the three evaluated criteria. Cases were restricted to the first sepsis encounter on the patient level. Septic shock cases are nested within sepsis cases and reflect the occurrence of septic shock within the first sepsis encounter for a patient. In total, 302,112 cases met one, two, or three sepsis criteria. Sepsis criteria were met for only one criterion in 195,410 (65%) of cases and 106,702 (35%) cases met multiple criteria during their encounter (77,187 [26%] cases met two criteria and 29,515 [10%] cases met all three criteria). Sepsis criteria agreement was the highest between CDC ASE and Sepsis-3 criteria (κ = 0.39, Jaccard index = 0.31). In total, 59,043 cases met one, two, or three septic shock criteria. Septic shock criteria were met for only one criterion in 32,235 (55%) of cases and 26,808 (45%) cases met multiple criteria during their encounter (17,582 [30%] cases met two criteria and 9,226 [16%] cases met all three criteria). Septic shock criteria agreement was also the highest between CDC ASE and Sepsis-3 criteria (κ = 0.53, Jaccard index = 0.37). CDC ASE = US Centers for Disease Control and Prevention Adult Sepsis Event; ICD = International Classification of Disease; Sepsis-3 = Third International Consensus Definition for Sepsis and Septic Shock Characteristics and outcomes of patients differed across the three sepsis case identification criteria (Table 2). Patients meeting Sepsis-3 sepsis criteria had the lowest frequencies of hospital-onset sepsis, intensive care unit (ICU) admission, and in-hospital mortality (7%), as well as lower Sequential Organ Failure Assessment (SOFA) scores. Patients meeting ICD coding criteria had the highest frequencies of admission to a teaching hospital, ICU admission, and in-hospital mortality (19%). Patients meeting CDC ASE sepsis criteria had characteristic and outcome measure values predominantly between those of the other two groups, including in-hospital mortality (14%). Characteristics and outcomes of patients also differed across the three septic shock case identification criteria, with comparative patterns generally similar to those for patients with sepsis (Table 3). Patients meeting Sepsis-3 septic shock criteria had the lowest in-hospital mortality (27%), whereas patients meeting CDC ASE septic shock criteria had the highest in-hospital mortality (32%). Notably, documentation of continuous infusion vasopressor administration was missing for 21% of patients with ICD coding for septic shock (Supplementary Material 1: eTable 4). Table 2Characteristics and outcomes of patients meeting sepsis case identification criteriaVariablesAll encounters in query(N = 1,293,782)Either CDC ASE or sepsis-3 criteria(N = 285,624)CDC ASE(N = 139,267)Sepsis-3(N = 234,601)ICD coding(N = 64,461)Patient and hospital characteristics Age, years63 (45, 76)69 (58, 80)68 (57, 78)69 (58, 80)69 (58, 79) Female sex761,299 (58.8)138,246 (48.4)66,354 (47.6)112,317 (47.9)30,481 (47.3) Race and ethnicity Asian104,58 (0.8)2,442 (0.9)1,218 (0.9)2,015 (0.9)591 (1.0) Black282,869 (22.5)52,318 (19.0)27,115 (20.2)41,857 (18.5)12,724 (20.5) Hispanic or Latinx81,135 (6.5)13,308 (4.8)6,229 (4.6)10,940 (4.8)3,173 (5.1) Multiracial17,216 (1.4)2,887 (1.0)1,452 (1.1)2,315 (1.0)715 (1.2) Other2,441 (0.2)535 (0.2)274 (0.2)443 (0.2)119 (0.2) White863,544 (68.7)203,956 (74.0)97,781 (72.9)168,358 (74.5)44,794 (72.1) Admission year2019(2016, 2023)2020(2016, 2023)2020(2016, 2023)2019(2016, 2023)2021(2018, 2023) Hospital state Ohio1,110,683 (85.8)243,249 (85.2)120,756 (86.7)198,600 (84.7)54,077 (83.9) Florida183,099 (14.2)42,375 (14.8)18,511 (13.3)36,001 (15.3)10,384 (16.1) Number of hospital beds < 200244,454 (18.9)51,059 (17.9)24,293 (17.4)40,679 (17.3)10,701 (16.6) 200–499565,190 (43.7)115,684 (40.5)54,796 (39.3)94,883 (40.4)24,861 (38.6) ≥ 500484,138 (37.4)118,881 (41.6)60,178 (43.2)99,039 (42.2)28,899 (44.8) Teaching hospital status1,003,369 (77.6)228,865 (80.1)113,131 (81.2)188,929 (80.5)55,691 (86.4)At sepsis onset Onset settinga Community onsetNA251,248 (88.0)119,011 (85.5)207,067 (88.3)35,978 (78.5) Hospital onsetNA34,376 (12.0)20,256 (14.5)27,534 (11.7)9,880 (21.5) Unit location Emergency departmentNA189,021 (66.2)98,446 (70.7)143,820 (61.3)NA WardNA64,165 (22.5)22,221 (16.0)57,096 (24.3)NA ICUNA32,278 (11.3)18,505 (13.3)33,566 (14.3)NA Prior surgeryb190,198 (14.7)41,636 (14.6)18,453 (13.3)36,114 (15.4)8,086 (12.5) SOFAcNA2 (1, 4)3 (1, 6)3 (2, 4)NA Organ system dysfunctiond CardiovascularNA45,464 (15.9)38,387 (27.6)30,351 (12.9)NA RespiratoryNA64,032 (22.4)35,438 (25.4)62,408 (26.6)NA KidneyNA70,470 (24.7)44,268 (31.8)50,341 (21.5)NA LiverNA26,000 (9.1)13,777 (9.9)28,569 (12.2)NA CoagulationNA33,670 (11.8)14,903 (10.7)36,921 (15.7)NA LactateNANA99,041 (71.1)NANA NeurologicNANANA57,181 (24.4)NAThroughout encounter Maximum SOFA1 (0, 3)3 (2, 6)5 (3, 9)3 (2, 6)6 (3, 11) Mechanical ventilation96,877 (7.5)58,833 (20.6)44,895 (32.2)48,226 (20.6)22,261 (34.5) Blood cultures collected641,157 (49.6)231,278 (81.0)139,267 (100.0)180,255 (76.8)62,928 (97.6) SARS-CoV-2 PCR positive25,294 (2.0)8,897 (3.1)5,242 (3.8)7,164 (3.1)3,257 (5.1)Hospital outcomes Mortality33,617 (2.6)22,184 (7.9)19,055 (13.9)16,103 (7.0)11,858 (18.6) Length of stay, days3.3 (0.4, 7.0)6.3 (3.8, 11.7)8.2 (4.9, 15.1)6.2 (3.7, 11.6)8.0 (4.2, 15.2) ICU admission241,329 (18.7)125,334 (43.9)83,333 (59.8)102,784 (43.8)45,065 (69.9) Discharge disposition Home969,115 (75.7)162,444 (57.7)65,010 (47.3)136,030 (58.8)26,602 (41.7) Other health care facilitye218,890 (17.1)77,059 (27.4)41,003 (29.8)63,775 (27.6)18,954 (29.7) Death33,617 (2.6)22,184 (7.9)19,055 (13.9)16,103 (7.0)11,858 (18.6) Acute care facility17,041 (1.3)4,983 (1.8)3,305 (2.4)3,515 (1.5)1,204 (1.9) Hospice24,229 (1.9)12,387 (4.4)8,244 (6.0)9,458 (4.1)4,429 (6.9) Against medical advice17,955 (1.4)2,642 (0.9)884 (0.6)2,309 (1.0)688 (1.1)Data presented as median (interquartile range) or count (percentage). Percentages for categorical variables may not sum to 100 due to rounding. Missingness for each variable is characterized in Supplementary Material 1: eTable 4aCommunity onset sepsis events were within two calendar days of presentation for CDC ASE cohort, within 48 h of presentation for Sepsis-3 cohort, and “present on admission” for ICD Coding cohort. Encounters not meeting criteria for community onset events were adjudicated as hospital onset events (the two categories are mutually exclusive)bFrom 30 days prior to admission through 48 h after sepsis onset. Surgery from 30 days prior to admission through encounter discharge in the All Encounters in Query and ICD Coding cohortscSOFA ranges from 0 to 24, where 0 to 4 points are assigned for each of six organ systems: cardiovascular, respiratory, kidney, liver, coagulation (hematologic), and neurologic (central nervous system). Greater values correspond to greater severity. Maximum value on the day of sepsis onset. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met SOFA on the day of sepsis onset per CDC ASE criteria was used; otherwise, SOFA on the day of sepsis onset per Sepsis-3 criteria was useddBased on CDC ASE organ dysfunction criteria (also termed eSOFA criteria) during the sepsis “window period” for CDC ASE study cohort. Based on individual organ system SOFA sub-score ≥ 2 during the sepsis “window period” for Sepsis-3 study cohort. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met CDC ASE organ dysfunction criteria were used; if not, individual organ system SOFA sub-score ≥ 2 criteria were usedeIncludes patients discharged to a nursing home, intermediate care, or skilled nursing facility (including long-term acute care facilities). Does not include patients discharged to assisted living care or assisted living facilitiesCDC ASE = US Centers for Disease Control and Prevention Adult Sepsis Event; PCR = polymerase chain reaction; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; eSOFA = Electronic Sequential Organ Failure Assessment; ICD = International Classification of Diseases; ICU = intensive care unit; NA = not applicable or not available; SOFA = Sequential Organ Failure Assessment Table 2Characteristics and outcomes of patients meeting sepsis case identification criteriaVariablesAll encounters in query(N = 1,293,782)Either CDC ASE or sepsis-3 criteria(N = 285,624)CDC ASE(N = 139,267)Sepsis-3(N = 234,601)ICD coding(N = 64,461)Patient and hospital characteristics Age, years63 (45, 76)69 (58, 80)68 (57, 78)69 (58, 80)69 (58, 79) Female sex761,299 (58.8)138,246 (48.4)66,354 (47.6)112,317 (47.9)30,481 (47.3) Race and ethnicity Asian104,58 (0.8)2,442 (0.9)1,218 (0.9)2,015 (0.9)591 (1.0) Black282,869 (22.5)52,318 (19.0)27,115 (20.2)41,857 (18.5)12,724 (20.5) Hispanic or Latinx81,135 (6.5)13,308 (4.8)6,229 (4.6)10,940 (4.8)3,173 (5.1) Multiracial17,216 (1.4)2,887 (1.0)1,452 (1.1)2,315 (1.0)715 (1.2) Other2,441 (0.2)535 (0.2)274 (0.2)443 (0.2)119 (0.2) White863,544 (68.7)203,956 (74.0)97,781 (72.9)168,358 (74.5)44,794 (72.1) Admission year2019(2016, 2023)2020(2016, 2023)2020(2016, 2023)2019(2016, 2023)2021(2018, 2023) Hospital state Ohio1,110,683 (85.8)243,249 (85.2)120,756 (86.7)198,600 (84.7)54,077 (83.9) Florida183,099 (14.2)42,375 (14.8)18,511 (13.3)36,001 (15.3)10,384 (16.1) Number of hospital beds < 200244,454 (18.9)51,059 (17.9)24,293 (17.4)40,679 (17.3)10,701 (16.6) 200–499565,190 (43.7)115,684 (40.5)54,796 (39.3)94,883 (40.4)24,861 (38.6) ≥ 500484,138 (37.4)118,881 (41.6)60,178 (43.2)99,039 (42.2)28,899 (44.8) Teaching hospital status1,003,369 (77.6)228,865 (80.1)113,131 (81.2)188,929 (80.5)55,691 (86.4)At sepsis onset Onset settinga Community onsetNA251,248 (88.0)119,011 (85.5)207,067 (88.3)35,978 (78.5) Hospital onsetNA34,376 (12.0)20,256 (14.5)27,534 (11.7)9,880 (21.5) Unit location Emergency departmentNA189,021 (66.2)98,446 (70.7)143,820 (61.3)NA WardNA64,165 (22.5)22,221 (16.0)57,096 (24.3)NA ICUNA32,278 (11.3)18,505 (13.3)33,566 (14.3)NA Prior surgeryb190,198 (14.7)41,636 (14.6)18,453 (13.3)36,114 (15.4)8,086 (12.5) SOFAcNA2 (1, 4)3 (1, 6)3 (2, 4)NA Organ system dysfunctiond CardiovascularNA45,464 (15.9)38,387 (27.6)30,351 (12.9)NA RespiratoryNA64,032 (22.4)35,438 (25.4)62,408 (26.6)NA KidneyNA70,470 (24.7)44,268 (31.8)50,341 (21.5)NA LiverNA26,000 (9.1)13,777 (9.9)28,569 (12.2)NA CoagulationNA33,670 (11.8)14,903 (10.7)36,921 (15.7)NA LactateNANA99,041 (71.1)NANA NeurologicNANANA57,181 (24.4)NAThroughout encounter Maximum SOFA1 (0, 3)3 (2, 6)5 (3, 9)3 (2, 6)6 (3, 11) Mechanical ventilation96,877 (7.5)58,833 (20.6)44,895 (32.2)48,226 (20.6)22,261 (34.5) Blood cultures collected641,157 (49.6)231,278 (81.0)139,267 (100.0)180,255 (76.8)62,928 (97.6) SARS-CoV-2 PCR positive25,294 (2.0)8,897 (3.1)5,242 (3.8)7,164 (3.1)3,257 (5.1)Hospital outcomes Mortality33,617 (2.6)22,184 (7.9)19,055 (13.9)16,103 (7.0)11,858 (18.6) Length of stay, days3.3 (0.4, 7.0)6.3 (3.8, 11.7)8.2 (4.9, 15.1)6.2 (3.7, 11.6)8.0 (4.2, 15.2) ICU admission241,329 (18.7)125,334 (43.9)83,333 (59.8)102,784 (43.8)45,065 (69.9) Discharge disposition Home969,115 (75.7)162,444 (57.7)65,010 (47.3)136,030 (58.8)26,602 (41.7) Other health care facilitye218,890 (17.1)77,059 (27.4)41,003 (29.8)63,775 (27.6)18,954 (29.7) Death33,617 (2.6)22,184 (7.9)19,055 (13.9)16,103 (7.0)11,858 (18.6) Acute care facility17,041 (1.3)4,983 (1.8)3,305 (2.4)3,515 (1.5)1,204 (1.9) Hospice24,229 (1.9)12,387 (4.4)8,244 (6.0)9,458 (4.1)4,429 (6.9) Against medical advice17,955 (1.4)2,642 (0.9)884 (0.6)2,309 (1.0)688 (1.1)Data presented as median (interquartile range) or count (percentage). Percentages for categorical variables may not sum to 100 due to rounding. Missingness for each variable is characterized in Supplementary Material 1: eTable 4aCommunity onset sepsis events were within two calendar days of presentation for CDC ASE cohort, within 48 h of presentation for Sepsis-3 cohort, and “present on admission” for ICD Coding cohort. Encounters not meeting criteria for community onset events were adjudicated as hospital onset events (the two categories are mutually exclusive)bFrom 30 days prior to admission through 48 h after sepsis onset. Surgery from 30 days prior to admission through encounter discharge in the All Encounters in Query and ICD Coding cohortscSOFA ranges from 0 to 24, where 0 to 4 points are assigned for each of six organ systems: cardiovascular, respiratory, kidney, liver, coagulation (hematologic), and neurologic (central nervous system). Greater values correspond to greater severity. Maximum value on the day of sepsis onset. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met SOFA on the day of sepsis onset per CDC ASE criteria was used; otherwise, SOFA on the day of sepsis onset per Sepsis-3 criteria was useddBased on CDC ASE organ dysfunction criteria (also termed eSOFA criteria) during the sepsis “window period” for CDC ASE study cohort. Based on individual organ system SOFA sub-score ≥ 2 during the sepsis “window period” for Sepsis-3 study cohort. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met CDC ASE organ dysfunction criteria were used; if not, individual organ system SOFA sub-score ≥ 2 criteria were usedeIncludes patients discharged to a nursing home, intermediate care, or skilled nursing facility (including long-term acute care facilities). Does not include patients discharged to assisted living care or assisted living facilitiesCDC ASE = US Centers for Disease Control and Prevention Adult Sepsis Event; PCR = polymerase chain reaction; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; eSOFA = Electronic Sequential Organ Failure Assessment; ICD = International Classification of Diseases; ICU = intensive care unit; NA = not applicable or not available; SOFA = Sequential Organ Failure Assessment Characteristics and outcomes of patients meeting sepsis case identification criteria 2019 (2016, 2023) 2020 (2016, 2023) 2020 (2016, 2023) 2019 (2016, 2023) 2021 (2018, 2023) Data presented as median (interquartile range) or count (percentage). Percentages for categorical variables may not sum to 100 due to rounding. Missingness for each variable is characterized in Supplementary Material 1: eTable 4 aCommunity onset sepsis events were within two calendar days of presentation for CDC ASE cohort, within 48 h of presentation for Sepsis-3 cohort, and “present on admission” for ICD Coding cohort. Encounters not meeting criteria for community onset events were adjudicated as hospital onset events (the two categories are mutually exclusive) bFrom 30 days prior to admission through 48 h after sepsis onset. Surgery from 30 days prior to admission through encounter discharge in the All Encounters in Query and ICD Coding cohorts cSOFA ranges from 0 to 24, where 0 to 4 points are assigned for each of six organ systems: cardiovascular, respiratory, kidney, liver, coagulation (hematologic), and neurologic (central nervous system). Greater values correspond to greater severity. Maximum value on the day of sepsis onset. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met SOFA on the day of sepsis onset per CDC ASE criteria was used; otherwise, SOFA on the day of sepsis onset per Sepsis-3 criteria was used dBased on CDC ASE organ dysfunction criteria (also termed eSOFA criteria) during the sepsis “window period” for CDC ASE study cohort. Based on individual organ system SOFA sub-score ≥ 2 during the sepsis “window period” for Sepsis-3 study cohort. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met CDC ASE organ dysfunction criteria were used; if not, individual organ system SOFA sub-score ≥ 2 criteria were used eIncludes patients discharged to a nursing home, intermediate care, or skilled nursing facility (including long-term acute care facilities). Does not include patients discharged to assisted living care or assisted living facilities CDC ASE = US Centers for Disease Control and Prevention Adult Sepsis Event; PCR = polymerase chain reaction; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; eSOFA = Electronic Sequential Organ Failure Assessment; ICD = International Classification of Diseases; ICU = intensive care unit; NA = not applicable or not available; SOFA = Sequential Organ Failure Assessment Table 3Characteristics and outcomes of patients meeting septic shock case identification criteriaVariablesEither CDC ASE or sepsis-3 criteria(N = 43,559)CDC ASE(N = 38,387)Sepsis-3(N = 21,253)ICD coding(N = 35,437)Patient and hospital characteristics Age, years68 (58, 77)68 (58, 78)67 (57, 76)69 (58, 78) Female sex19,652 (45.1)17,642 (46.0)9,382 (44.2)16,588 (46.8) Race and ethnicity Asian393 (0.9)330 (0.9)225 (1.1)325 (1.0) Black7,798 (18.8)7,077 (19.4)3,666 (18.2)6,806 (20.1) Hispanic or Latinx1,640 (4.0)1,427 (3.9)883 (4.4)1,397 (4.1) Multiracial451 (1.1)387 (1.1)216 (1.1)398 (1.2) Other87 (0.2)75 (0.2)46 (0.2)66 (0.2) White31,054 (75.0)27,229 (74.6)15,123 (75.0)24,953 (73.5) Admission year2020 (2016, 2023)2020 (2016, 2022)2020 (2017, 2023)2021 (2018, 2023) Hospital state Ohio38,567 (88.5)34,020 (88.6)18,606 (87.5)30,873 (87.1) Florida4,992 (11.5)4,367 (11.4)2647 (12.5)4,564 (12.9) Number of hospital beds < 2005,733 (13.2)5,421 (14.1)2,474 (11.6)4,270 (12.0) 200–49915,274 (35.1)13,928 (36.3)7,221 (34.0)11,818 (33.3) ≥ 50022,552 (51.8)19,038 (49.6)11,558 (54.4)19,349 (54.6) Teaching hospital status37,671 (86.5)32,866 (85.6)18,638 (87.7)31,578 (89.1)At sepsis onset Onset settinga Community onset32,380 (74.3)29,170 (76.0)16,202 (76.2)19,822 (70.0) Hospital onset11,179 (25.7)9,217 (24.0)5,051 (23.8)8,475 (30.0) Unit location Emergency department22,439 (51.6)21,178 (55.2)9,746 (45.9)NA Ward8,706 (20.0)6,690 (17.4)3,692 (17.4)NA ICU12,366 (28.4)10,477 (27.3)7,801 (36.7)NA Prior surgeryb11,179 (25.7)8,101 (21.1)6,856 (32.3)5,504 (15.5) SOFAc7 (4, 10)7 (4, 10)9 (5, 12)NA Organ system dysfunctiond Cardiovascular42,484 (97.5)38,387 (100)19,800 (93.2)NA Respiratory25,436 (58.4)22,042 (57.4)14,770 (69.5)NA Kidney16,354 (37.5)15,299 (39.9)7,209 (33.9)NA Liver5,470 (12.6)4,503 (11.7)4,772 (22.5)NA Coagulation7,940 (18.2)6,355 (16.6)6,080 (28.6)NA LactateNA28,366 (73.9)NANA NeurologicNANA15,609 (73.4)NAAt shock onset (vasopressor initiation) SOFAc9 (7, 12)9 (6, 12)10 (7, 12)9 (6, 12) MAP, mm Hge62 (55, 71)61 (55, 70)63 (56, 72)61 (55, 68) Lactate, mmol/Le2.6 (1.5, 4.8)2.5 (1.5, 4.7)3.4 (2.2, 5.7)2.7 (1.6, 4.8) Initial vasopressor Norepinephrine34,520 (79.2)30,916 (80.5)16,910 (79.6)24,322 (87.1) Dopamine1,274 (2.9)1,179 (3.1)457 (2.2)395 (1.4) Phenylephrine4,217 (9.7)3,775 (9.8)1,764 (8.3)2,239 (8.0) Epinephrine2,019 (4.6)1,183 (3.1)1,468 (6.9)456 (1.6) Vasopressin1,529 (3.5)1,334 (3.5)654 (3.1)519 (1.9) Mechanical ventilation20,513 (47.1)17,119 (44.6)11,010 (51.8)11,040 (39.5)Throughout encounter Maximum SOFA11 (8, 13)11 (8, 13)11 (8, 14)9 (6, 13) Mechanical ventilation29,395 (67.5)25,184 (65.6)15,537 (73.1)18,503 (52.2) Blood cultures collected41,577 (95.4)38,387 (100.0)19,271 (90.7)34,673 (97.8) SARS-CoV-2 PCR positive1,984 (4.6)1,840 (4.8)907 (4.3)2,064 (5.8)Hospital outcomes Mortality12,579 (29.3)11,952 (31.6)5,565 (26.5)10,231 (29.3) Length of stay, days11.7 (6.2, 21.1)11.4 (6.0, 20.9)11.4 (6.1, 20.7)10.3 (5.4, 19.5) ICU admission42,922 (98.5)37,803 (98.5)21,105 (99.3)32,751 (92.4) Discharge disposition Home13,042 (30.4)10,095 (26.7)7,551 (36.0)10,005 (28.6) Other health care facilityf13,400 (31.2)12,135 (32.1)6,116 (29.1)10,953 (31.4) Death12,579 (29.3)11,952 (31.6)5,565 (26.5)10,231 (29.3) Acute care facility1,039 (2.4)958 (2.5)450 (2.1)743 (2.1) Hospice2,645 (6.2)2,484 (6.6)1,184 (5.6)2,758 (7.9) Against medical advice226 (0.5)186 (0.5)135 (0.6)244 (0.7)Data presented as median (interquartile range) or count (percentage). Percentages for categorical variables may not sum to 100 due to rounding. Missingness for each variable is characterized in Supplementary Material 1: eTable 4aCommunity onset sepsis events were within two calendar days of presentation for CDC ASE cohort, within 48 h of presentation for Sepsis-3 cohort, and “present on admission” for ICD Coding cohort. Encounters not meeting criteria for community onset events were adjudicated as hospital onset events (the two categories are mutually exclusive). For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met the CDC ASE sepsis onset date was used; if not, the Sepsis-3 sepsis onset date was usedbFrom 30 days prior to admission through 48 h after sepsis onset. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met the CDC ASE sepsis onset date was used; if not, the Sepsis-3 sepsis onset date was used. Surgery from 30 days prior to admission through encounter discharge in the ICD Coding cohortcSOFA ranges from 0 to 24, where 0 to 4 points are assigned for each of six organ systems: cardiovascular, respiratory, kidney, liver, coagulation (hematologic), and neurologic (central nervous system). Greater values correspond to greater severity. Maximum value on the day of sepsis onset or on the day of shock onset, respectively. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met SOFA on the day of sepsis onset per CDC ASE criteria was used; otherwise, SOFA on the day of sepsis onset per Sepsis-3 criteria was used. For the ICD Coding cohort, SOFA on the day of first vasopressor initiation was useddBased on CDC ASE organ dysfunction criteria (also termed eSOFA criteria) during the sepsis “window period” for CDC ASE study cohort. Based on individual organ system SOFA sub-score ≥ 2 during the sepsis “window period” for Sepsis-3 study cohort. Because Sepsis-3 criteria for septic shock include administration of phenylephrine and vasopressin (among other criteria), patients could meet cohort criteria without cardiovascular organ system dysfunction since the original SOFA description does not assign a score for phenylephrine or vasopressin administration. Further, a SOFA score is only assigned for vasopressors given for ≥ 1 h, but administration duration is not a component of Sepsis-3 clinical criteria for septic shock. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met CDC ASE organ dysfunction criteria were used; if not, individual organ system SOFA sub-score ≥ 2 criteria were usedeClosest value prior to shock onset and within 6 h prior to shock onset. For the ICD Coding cohort, the closest value prior to vasopressor initiation and within 6 h prior to vasopressor initiation was usedfIncludes patients discharged to a nursing home, intermediate care, or skilled nursing facility (including long-term acute care facilities). Does not include patients discharged to assisted living care or assisted living facilitiesCDC ASE = US Centers for Disease Control and Prevention Adult Sepsis Event; eSOFA = Electronic Sequential Organ Failure Assessment; ICD = International Classification of Diseases; ICU = intensive care unit; MAP = mean arterial pressure; NA = not applicable or not available; PCR = polymerase chain reaction; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; SOFA = Sequential Organ Failure Assessment Table 3Characteristics and outcomes of patients meeting septic shock case identification criteriaVariablesEither CDC ASE or sepsis-3 criteria(N = 43,559)CDC ASE(N = 38,387)Sepsis-3(N = 21,253)ICD coding(N = 35,437)Patient and hospital characteristics Age, years68 (58, 77)68 (58, 78)67 (57, 76)69 (58, 78) Female sex19,652 (45.1)17,642 (46.0)9,382 (44.2)16,588 (46.8) Race and ethnicity Asian393 (0.9)330 (0.9)225 (1.1)325 (1.0) Black7,798 (18.8)7,077 (19.4)3,666 (18.2)6,806 (20.1) Hispanic or Latinx1,640 (4.0)1,427 (3.9)883 (4.4)1,397 (4.1) Multiracial451 (1.1)387 (1.1)216 (1.1)398 (1.2) Other87 (0.2)75 (0.2)46 (0.2)66 (0.2) White31,054 (75.0)27,229 (74.6)15,123 (75.0)24,953 (73.5) Admission year2020 (2016, 2023)2020 (2016, 2022)2020 (2017, 2023)2021 (2018, 2023) Hospital state Ohio38,567 (88.5)34,020 (88.6)18,606 (87.5)30,873 (87.1) Florida4,992 (11.5)4,367 (11.4)2647 (12.5)4,564 (12.9) Number of hospital beds < 2005,733 (13.2)5,421 (14.1)2,474 (11.6)4,270 (12.0) 200–49915,274 (35.1)13,928 (36.3)7,221 (34.0)11,818 (33.3) ≥ 50022,552 (51.8)19,038 (49.6)11,558 (54.4)19,349 (54.6) Teaching hospital status37,671 (86.5)32,866 (85.6)18,638 (87.7)31,578 (89.1)At sepsis onset Onset settinga Community onset32,380 (74.3)29,170 (76.0)16,202 (76.2)19,822 (70.0) Hospital onset11,179 (25.7)9,217 (24.0)5,051 (23.8)8,475 (30.0) Unit location Emergency department22,439 (51.6)21,178 (55.2)9,746 (45.9)NA Ward8,706 (20.0)6,690 (17.4)3,692 (17.4)NA ICU12,366 (28.4)10,477 (27.3)7,801 (36.7)NA Prior surgeryb11,179 (25.7)8,101 (21.1)6,856 (32.3)5,504 (15.5) SOFAc7 (4, 10)7 (4, 10)9 (5, 12)NA Organ system dysfunctiond Cardiovascular42,484 (97.5)38,387 (100)19,800 (93.2)NA Respiratory25,436 (58.4)22,042 (57.4)14,770 (69.5)NA Kidney16,354 (37.5)15,299 (39.9)7,209 (33.9)NA Liver5,470 (12.6)4,503 (11.7)4,772 (22.5)NA Coagulation7,940 (18.2)6,355 (16.6)6,080 (28.6)NA LactateNA28,366 (73.9)NANA NeurologicNANA15,609 (73.4)NAAt shock onset (vasopressor initiation) SOFAc9 (7, 12)9 (6, 12)10 (7, 12)9 (6, 12) MAP, mm Hge62 (55, 71)61 (55, 70)63 (56, 72)61 (55, 68) Lactate, mmol/Le2.6 (1.5, 4.8)2.5 (1.5, 4.7)3.4 (2.2, 5.7)2.7 (1.6, 4.8) Initial vasopressor Norepinephrine34,520 (79.2)30,916 (80.5)16,910 (79.6)24,322 (87.1) Dopamine1,274 (2.9)1,179 (3.1)457 (2.2)395 (1.4) Phenylephrine4,217 (9.7)3,775 (9.8)1,764 (8.3)2,239 (8.0) Epinephrine2,019 (4.6)1,183 (3.1)1,468 (6.9)456 (1.6) Vasopressin1,529 (3.5)1,334 (3.5)654 (3.1)519 (1.9) Mechanical ventilation20,513 (47.1)17,119 (44.6)11,010 (51.8)11,040 (39.5)Throughout encounter Maximum SOFA11 (8, 13)11 (8, 13)11 (8, 14)9 (6, 13) Mechanical ventilation29,395 (67.5)25,184 (65.6)15,537 (73.1)18,503 (52.2) Blood cultures collected41,577 (95.4)38,387 (100.0)19,271 (90.7)34,673 (97.8) SARS-CoV-2 PCR positive1,984 (4.6)1,840 (4.8)907 (4.3)2,064 (5.8)Hospital outcomes Mortality12,579 (29.3)11,952 (31.6)5,565 (26.5)10,231 (29.3) Length of stay, days11.7 (6.2, 21.1)11.4 (6.0, 20.9)11.4 (6.1, 20.7)10.3 (5.4, 19.5) ICU admission42,922 (98.5)37,803 (98.5)21,105 (99.3)32,751 (92.4) Discharge disposition Home13,042 (30.4)10,095 (26.7)7,551 (36.0)10,005 (28.6) Other health care facilityf13,400 (31.2)12,135 (32.1)6,116 (29.1)10,953 (31.4) Death12,579 (29.3)11,952 (31.6)5,565 (26.5)10,231 (29.3) Acute care facility1,039 (2.4)958 (2.5)450 (2.1)743 (2.1) Hospice2,645 (6.2)2,484 (6.6)1,184 (5.6)2,758 (7.9) Against medical advice226 (0.5)186 (0.5)135 (0.6)244 (0.7)Data presented as median (interquartile range) or count (percentage). Percentages for categorical variables may not sum to 100 due to rounding. Missingness for each variable is characterized in Supplementary Material 1: eTable 4aCommunity onset sepsis events were within two calendar days of presentation for CDC ASE cohort, within 48 h of presentation for Sepsis-3 cohort, and “present on admission” for ICD Coding cohort. Encounters not meeting criteria for community onset events were adjudicated as hospital onset events (the two categories are mutually exclusive). For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met the CDC ASE sepsis onset date was used; if not, the Sepsis-3 sepsis onset date was usedbFrom 30 days prior to admission through 48 h after sepsis onset. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met the CDC ASE sepsis onset date was used; if not, the Sepsis-3 sepsis onset date was used. Surgery from 30 days prior to admission through encounter discharge in the ICD Coding cohortcSOFA ranges from 0 to 24, where 0 to 4 points are assigned for each of six organ systems: cardiovascular, respiratory, kidney, liver, coagulation (hematologic), and neurologic (central nervous system). Greater values correspond to greater severity. Maximum value on the day of sepsis onset or on the day of shock onset, respectively. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met SOFA on the day of sepsis onset per CDC ASE criteria was used; otherwise, SOFA on the day of sepsis onset per Sepsis-3 criteria was used. For the ICD Coding cohort, SOFA on the day of first vasopressor initiation was useddBased on CDC ASE organ dysfunction criteria (also termed eSOFA criteria) during the sepsis “window period” for CDC ASE study cohort. Based on individual organ system SOFA sub-score ≥ 2 during the sepsis “window period” for Sepsis-3 study cohort. Because Sepsis-3 criteria for septic shock include administration of phenylephrine and vasopressin (among other criteria), patients could meet cohort criteria without cardiovascular organ system dysfunction since the original SOFA description does not assign a score for phenylephrine or vasopressin administration. Further, a SOFA score is only assigned for vasopressors given for ≥ 1 h, but administration duration is not a component of Sepsis-3 clinical criteria for septic shock. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met CDC ASE organ dysfunction criteria were used; if not, individual organ system SOFA sub-score ≥ 2 criteria were usedeClosest value prior to shock onset and within 6 h prior to shock onset. For the ICD Coding cohort, the closest value prior to vasopressor initiation and within 6 h prior to vasopressor initiation was usedfIncludes patients discharged to a nursing home, intermediate care, or skilled nursing facility (including long-term acute care facilities). Does not include patients discharged to assisted living care or assisted living facilitiesCDC ASE = US Centers for Disease Control and Prevention Adult Sepsis Event; eSOFA = Electronic Sequential Organ Failure Assessment; ICD = International Classification of Diseases; ICU = intensive care unit; MAP = mean arterial pressure; NA = not applicable or not available; PCR = polymerase chain reaction; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; SOFA = Sequential Organ Failure Assessment Characteristics and outcomes of patients meeting septic shock case identification criteria Data presented as median (interquartile range) or count (percentage). Percentages for categorical variables may not sum to 100 due to rounding. Missingness for each variable is characterized in Supplementary Material 1: eTable 4 aCommunity onset sepsis events were within two calendar days of presentation for CDC ASE cohort, within 48 h of presentation for Sepsis-3 cohort, and “present on admission” for ICD Coding cohort. Encounters not meeting criteria for community onset events were adjudicated as hospital onset events (the two categories are mutually exclusive). For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met the CDC ASE sepsis onset date was used; if not, the Sepsis-3 sepsis onset date was used bFrom 30 days prior to admission through 48 h after sepsis onset. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met the CDC ASE sepsis onset date was used; if not, the Sepsis-3 sepsis onset date was used. Surgery from 30 days prior to admission through encounter discharge in the ICD Coding cohort cSOFA ranges from 0 to 24, where 0 to 4 points are assigned for each of six organ systems: cardiovascular, respiratory, kidney, liver, coagulation (hematologic), and neurologic (central nervous system). Greater values correspond to greater severity. Maximum value on the day of sepsis onset or on the day of shock onset, respectively. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met SOFA on the day of sepsis onset per CDC ASE criteria was used; otherwise, SOFA on the day of sepsis onset per Sepsis-3 criteria was used. For the ICD Coding cohort, SOFA on the day of first vasopressor initiation was used dBased on CDC ASE organ dysfunction criteria (also termed eSOFA criteria) during the sepsis “window period” for CDC ASE study cohort. Based on individual organ system SOFA sub-score ≥ 2 during the sepsis “window period” for Sepsis-3 study cohort. Because Sepsis-3 criteria for septic shock include administration of phenylephrine and vasopressin (among other criteria), patients could meet cohort criteria without cardiovascular organ system dysfunction since the original SOFA description does not assign a score for phenylephrine or vasopressin administration. Further, a SOFA score is only assigned for vasopressors given for ≥ 1 h, but administration duration is not a component of Sepsis-3 clinical criteria for septic shock. For the Either CDC ASE or Sepsis-3 Criteria cohort, if CDC ASE criteria were met CDC ASE organ dysfunction criteria were used; if not, individual organ system SOFA sub-score ≥ 2 criteria were used eClosest value prior to shock onset and within 6 h prior to shock onset. For the ICD Coding cohort, the closest value prior to vasopressor initiation and within 6 h prior to vasopressor initiation was used fIncludes patients discharged to a nursing home, intermediate care, or skilled nursing facility (including long-term acute care facilities). Does not include patients discharged to assisted living care or assisted living facilities CDC ASE = US Centers for Disease Control and Prevention Adult Sepsis Event; eSOFA = Electronic Sequential Organ Failure Assessment; ICD = International Classification of Diseases; ICU = intensive care unit; MAP = mean arterial pressure; NA = not applicable or not available; PCR = polymerase chain reaction; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; SOFA = Sequential Organ Failure Assessment Patients identified with sepsis and septic shock by both CDC ASE and Sepsis-3 criteria, compared with those meeting only one of the criteria, had differing characteristics, signs of critical illness, and outcomes (Table 4 and Supplementary Material 1: eTable 5). Patients meeting both sepsis criteria more frequently were admitted to a large hospital, more commonly had hospital-onset sepsis, had higher SOFA scores, more dysfunctional organ systems, and the highest in-hospital mortality incidence (15%). Those patients meeting Sepsis-3 sepsis criteria without meeting CDC ASE sepsis criteria had less severe illness and better outcomes. In-hospital mortality incidence was similar between patients meeting both CDC ASE and Sepsis-3 septic shock criteria (31%) and those meeting only CDC ASE septic shock criteria (32%), whereas patients meeting only Sepsis-3 septic shock criteria had a lower mortality incidence (12%). Patients identified with sepsis by only ICD coding, compared with those meeting only CDC ASE or Sepsis-3 criteria (and not ICD coding), were more frequently admitted during a later year and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) positive, had more severe illness, and had higher mortality incidence (Supplementary Material 1: eTable 6). Those patients identified with septic shock by only ICD coding, compared with those meeting only CDC ASE or Sepsis-3 criteria (and not ICD coding), were also more frequently admitted during a later year and SARS-CoV-2 positive, but less frequently had prior surgery, and had similar mortality incidence (Supplementary Material 1: eTable 7). Table 4Characteristics and outcomes of patients meeting and not meeting CDC ASE and sepsis-3 sepsis case identification criteriaVariablesBoth CDC ASE and sepsis-3 criteria met(N = 88,244)CDC ASE only(N = 51,023)Sepsis-3 only(N = 146,357)Neither criteria met(N = 1,008,158)Patient and hospital characteristics Age, years68 (57, 78)69 (57, 79)70 (58, 81)61 (41, 75) Female sex40,425 (45.8)25,929 (50.8)71,892 (49.1)62,3053 (61.8) Race and ethnicity Asian791 (0.9)427 (0.9)1,224 (0.9)8,016 (0.8) Black16,654 (19.7)10,461 (21.1)25,203 (17.8)230,551 (23.5) Hispanic or Latinx3,861 (4.6)2,368 (4.8)7,079 (5.0)67,827 (6.9) Multiracial880 (1.0)572 (1.2)1,435 (1.0)14,329 (1.5) Other182 (0.2)92 (0.2)261 (0.2)1,906 (0.2) White62,183 (73.5)35,598 (71.9)106,175 (75.1)659,588 (67.2) Admission year2019 (2016, 2022)2020 (2017, 2023)2019 (2016, 2023)2019 (2016, 2023) Hospital location (US state) Ohio76,105 (86.2)44,651 (87.5)122,493 (83.7)867,434 (86.0) Florida12,139 (13.8)6,372 (12.5)23,864 (16.3)140,724 (14.0) Number of hospital beds < 20014,387 (16.3)9,906 (19.4)26,766 (18.3)193,395 (19.2) 200–49934,174 (38.7)20,622 (40.4)60,888 (41.6)449,506 (44.6) ≥ 50039,683 (45.0)20,495 (40.2)58,703 (40.1)365,257 (36.2) Teaching hospital status72,710 (82.4)40,421 (79.2)115,734 (79.1)774,504 (76.8)At sepsis onset Onset settinga Community onset74,043 (83.9)44,968 (88.1)132,237 (90.4)NA Hospital onset14,201 (16.1)6,055 (11.9)14,120 (9.6)NA Unit location Emergency department58,722 (66.6)39,724 (77.9)90,575 (61.9)NA Ward14,938 (16.9)7,283 (14.3)41,944 (28.7)NA ICU14,515 (16.5)3,990 (7.8)13,773 (9.4)NA Prior surgeryb12,652 (14.3)5,801 (11.4)23,183 (15.8)148,562 (14.7)Organ system dysfunction at sepsis onset SOFAc4 (2, 6)1 (0, 4)2 (1, 3)NA Organ system dysfunctiond Cardiovascular29,071 (32.9)9,316 (18.3)7,077 (4.8)NA Respiratory27,283 (30.9)8,155 (16.0)28,594 (19.5)NA Kidney32,722 (37.1)11,546 (22.6)26,202 (17.9)NA Liver11,005 (12.5)2,772 (5.4)12,223 (8.4)NA Coagulation11,528 (13.1)3,375 (6.6)18,767 (12.8)NA LactateNA39,259 (76.9)NANA NeurologicNANA24,257 (16.6)NAThroughout encounter Maximum SOFA6 (3, 10)3 (1, 7)3 (2, 4)1 (0, 2) Mechanical ventilation34,288 (38.9)10,607 (20.8)13,938 (9.5)38,044 (3.8) Blood cultures collected882,44 (100.0)51,023 (100.0)92,011 (62.9)409,879 (40.7) SARS-CoV-2 PCR positive3,509 (4.0)1,733 (3.4)3,655 (2.5)16,397 (1.6)Hospital outcomes Mortality12,974 (14.9)6,081 (12.0)3,129 (2.2)11,433 (1.1) Length of stay, days9.2 (5.3, 17.1)6.8 (4.1, 11.8)5.0 (3.0, 8.7)2.5 (0.2, 5.7) ICU admission60,783 (68.9)22,550 (44.2)42,001 (28.7)115,995 (11.5) Discharge disposition Home38,596 (44.4)26,414 (52.3)97,434 (67.6)806,671 (80.7) Other health care facilitye27,719 (31.9)13,284 (26.3)36,056 (25.0)141,831 (14.2) Death12,974 (14.9)6,081 (12.0)3,129 (2.2)11,433 (1.1) Acute care facility1,837 (2.1)1,468 (2.9)1,678 (1.2)12,058 (1.2) Hospice5,315 (6.1)2,929 (5.8)4,143 (2.9)11,842 (1.2) Against medical advice551 (0.6)333 (0.7)1,758 (1.2)15,313 (1.5)All data presented as median (interquartile range) or count (percentage). Percentages for categorical variables may not sum to 100 due to rounding. Missingness for each variable is characterized in Supplementary Material 1: eTable 4aCommunity onset sepsis events were within two calendar days of presentation for CDC ASE and within 48 h of presentation for Sepsis-3. Encounters not meeting criteria for community onset events were adjudicated as hospital onset events (the two categories are mutually exclusive). For the Both CDC ASE or Sepsis-3 Criteria cohort the CDC ASE sepsis onset date was usedbFrom 30 days prior to admission through 48 h after sepsis onset. For the Both CDC ASE and Sepsis-3 Criteria Met cohort the CDC ASE sepsis onset date was used. Surgery from 30 days prior to admission through encounter discharge in the Neither Criteria Met cohortcSOFA ranges from 0 to 24, where 0 to 4 points are assigned for each of six organ systems: cardiovascular, respiratory, kidney, liver, coagulation (hematologic), and neurologic (central nervous system). Greater values correspond to greater severity. Maximum value on the day of sepsis onset. For the Both CDC ASE and Sepsis-3 Criteria Met cohort SOFA on the day of sepsis onset per CDC ASE criteria was useddBased on CDC ASE organ dysfunction criteria (also termed eSOFA criteria) during the sepsis “window period” for CDC ASE study cohorts. Based on individual organ system SOFA sub-score ≥ 2 on the day of sepsis onset for Sepsis-3 study cohort. For the Both CDC ASE and Sepsis-3 Criteria Met cohort CDC ASE organ dysfunction criteria were usedeIncludes patients discharged to a nursing home, intermediate care, or skilled nursing facility (including long-term acute care facilities). Does not include patients discharged to assisted living care or assisted living facilitiesCDC ASE = US Centers for Disease Control and Prevention Adult Sepsis Event; ICU = intensive care unit; NA = not applicable or not available; PCR = polymerase chain reaction; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; SOFA = Sequential Organ Failure Assessment (score) Table 4Characteristics and outcomes of patients meeting and not meeting CDC ASE and sepsis-3 sepsis case identification criteriaVariablesBoth CDC ASE and sepsis-3 criteria met(N = 88,244)CDC ASE only(N = 51,023)Sepsis-3 only(N = 146,357)Neither criteria met(N = 1,008,158)Patient and hospital characteristics Age, years68 (57, 78)69 (57, 79)70 (58, 81)61 (41, 75) Female sex40,425 (45.8)25,929 (50.8)71,892 (49.1)62,3053 (61.8) Race and ethnicity Asian791 (0.9)427 (0.9)1,224 (0.9)8,016 (0.8) Black16,654 (19.7)10,461 (21.1)25,203 (17.8)230,551 (23.5) Hispanic or Latinx3,861 (4.6)2,368 (4.8)7,079 (5.0)67,827 (6.9) Multiracial880 (1.0)572 (1.2)1,435 (1.0)14,329 (1.5) Other182 (0.2)92 (0.2)261 (0.2)1,906 (0.2) White62,183 (73.5)35,598 (71.9)106,175 (75.1)659,588 (67.2) Admission year2019 (2016, 2022)2020 (2017, 2023)2019 (2016, 2023)2019 (2016, 2023) Hospital location (US state) Ohio76,105 (86.2)44,651 (87.5)122,493 (83.7)867,434 (86.0) Florida12,139 (13.8)6,372 (12.5)23,864 (16.3)140,724 (14.0) Number of hospital beds < 20014,387 (16.3)9,906 (19.4)26,766 (18.3)193,395 (19.2) 200–49934,174 (38.7)20,622 (40.4)60,888 (41.6)449,506 (44.6) ≥ 50039,683 (45.0)20,495 (40.2)58,703 (40.1)365,257 (36.2) Teaching hospital status72,710 (82.4)40,421 (79.2)115,734 (79.1)774,504 (76.8)At sepsis onset Onset settinga Community onset74,043 (83.9)44,968 (88.1)132,237 (90.4)NA Hospital onset14,201 (16.1)6,055 (11.9)14,120 (9.6)NA Unit location Emergency department58,722 (66.6)39,724 (77.9)90,575 (61.9)NA Ward14,938 (16.9)7,283 (14.3)41,944 (28.7)NA ICU14,515 (16.5)3,990 (7.8)13,773 (9.4)NA Prior surgeryb12,652 (14.3)5,801 (11.4)23,183 (15.8)148,562 (14.7)Organ system dysfunction at sepsis onset SOFAc4 (2, 6)1 (0, 4)2 (1, 3)NA Organ system dysfunctiond Cardiovascular29,071 (32.9)9,316 (18.3)7,077 (4.8)NA Respiratory27,283 (30.9)8,155 (16.0)28,594 (19.5)NA Kidney32,722 (37.1)11,546 (22.6)26,202 (17.9)NA Liver11,005 (12.5)2,772 (5.4)12,223 (8.4)NA Coagulation11,528 (13.1)3,375 (6.6)18,767 (12.8)NA LactateNA39,259 (76.9)NANA NeurologicNANA24,257 (16.6)NAThroughout encounter Maximum SOFA6 (3, 10)3 (1, 7)3 (2, 4)1 (0, 2) Mechanical ventilation34,288 (38.9)10,607 (20.8)13,938 (9.5)38,044 (3.8) Blood cultures collected882,44 (100.0)51,023 (100.0)92,011 (62.9)409,879 (40.7) SARS-CoV-2 PCR positive3,509 (4.0)1,733 (3.4)3,655 (2.5)16,397 (1.6)Hospital outcomes Mortality12,974 (14.9)6,081 (12.0)3,129 (2.2)11,433 (1.1) Length of stay, days9.2 (5.3, 17.1)6.8 (4.1, 11.8)5.0 (3.0, 8.7)2.5 (0.2, 5.7) ICU admission60,783 (68.9)22,550 (44.2)42,001 (28.7)115,995 (11.5) Discharge disposition Home38,596 (44.4)26,414 (52.3)97,434 (67.6)806,671 (80.7) Other health care facilitye27,719 (31.9)13,284 (26.3)36,056 (25.0)141,831 (14.2) Death12,974 (14.9)6,081 (12.0)3,129 (2.2)11,433 (1.1) Acute care facility1,837 (2.1)1,468 (2.9)1,678 (1.2)12,058 (1.2) Hospice5,315 (6.1)2,929 (5.8)4,143 (2.9)11,842 (1.2) Against medical advice551 (0.6)333 (0.7)1,758 (1.2)15,313 (1.5)All data presented as median (interquartile range) or count (percentage). Percentages for categorical variables may not sum to 100 due to rounding. Missingness for each variable is characterized in Supplementary Material 1: eTable 4aCommunity onset sepsis events were within two calendar days of presentation for CDC ASE and within 48 h of presentation for Sepsis-3. Encounters not meeting criteria for community onset events were adjudicated as hospital onset events (the two categories are mutually exclusive). For the Both CDC ASE or Sepsis-3 Criteria cohort the CDC ASE sepsis onset date was usedbFrom 30 days prior to admission through 48 h after sepsis onset. For the Both CDC ASE and Sepsis-3 Criteria Met cohort the CDC ASE sepsis onset date was used. Surgery from 30 days prior to admission through encounter discharge in the Neither Criteria Met cohortcSOFA ranges from 0 to 24, where 0 to 4 points are assigned for each of six organ systems: cardiovascular, respiratory, kidney, liver, coagulation (hematologic), and neurologic (central nervous system). Greater values correspond to greater severity. Maximum value on the day of sepsis onset. For the Both CDC ASE and Sepsis-3 Criteria Met cohort SOFA on the day of sepsis onset per CDC ASE criteria was useddBased on CDC ASE organ dysfunction criteria (also termed eSOFA criteria) during the sepsis “window period” for CDC ASE study cohorts. Based on individual organ system SOFA sub-score ≥ 2 on the day of sepsis onset for Sepsis-3 study cohort. For the Both CDC ASE and Sepsis-3 Criteria Met cohort CDC ASE organ dysfunction criteria were usedeIncludes patients discharged to a nursing home, intermediate care, or skilled nursing facility (including long-term acute care facilities). Does not include patients discharged to assisted living care or assisted living facilitiesCDC ASE = US Centers for Disease Control and Prevention Adult Sepsis Event; ICU = intensive care unit; NA = not applicable or not available; PCR = polymerase chain reaction; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; SOFA = Sequential Organ Failure Assessment (score) Characteristics and outcomes of patients meeting and not meeting CDC ASE and sepsis-3 sepsis case identification criteria All data presented as median (interquartile range) or count (percentage). Percentages for categorical variables may not sum to 100 due to rounding. Missingness for each variable is characterized in Supplementary Material 1: eTable 4 aCommunity onset sepsis events were within two calendar days of presentation for CDC ASE and within 48 h of presentation for Sepsis-3. Encounters not meeting criteria for community onset events were adjudicated as hospital onset events (the two categories are mutually exclusive). For the Both CDC ASE or Sepsis-3 Criteria cohort the CDC ASE sepsis onset date was used bFrom 30 days prior to admission through 48 h after sepsis onset. For the Both CDC ASE and Sepsis-3 Criteria Met cohort the CDC ASE sepsis onset date was used. Surgery from 30 days prior to admission through encounter discharge in the Neither Criteria Met cohort cSOFA ranges from 0 to 24, where 0 to 4 points are assigned for each of six organ systems: cardiovascular, respiratory, kidney, liver, coagulation (hematologic), and neurologic (central nervous system). Greater values correspond to greater severity. Maximum value on the day of sepsis onset. For the Both CDC ASE and Sepsis-3 Criteria Met cohort SOFA on the day of sepsis onset per CDC ASE criteria was used dBased on CDC ASE organ dysfunction criteria (also termed eSOFA criteria) during the sepsis “window period” for CDC ASE study cohorts. Based on individual organ system SOFA sub-score ≥ 2 on the day of sepsis onset for Sepsis-3 study cohort. For the Both CDC ASE and Sepsis-3 Criteria Met cohort CDC ASE organ dysfunction criteria were used eIncludes patients discharged to a nursing home, intermediate care, or skilled nursing facility (including long-term acute care facilities). Does not include patients discharged to assisted living care or assisted living facilities CDC ASE = US Centers for Disease Control and Prevention Adult Sepsis Event; ICU = intensive care unit; NA = not applicable or not available; PCR = polymerase chain reaction; SARS-CoV-2 = severe acute respiratory syndrome coronavirus 2; SOFA = Sequential Organ Failure Assessment (score) Discussion Summary of key findings We provide an open-source workflow for the complete process of sepsis and septic shock case identification from the electronic health record, from data extraction through data pre-processing and application of case identification criteria. In our large, diverse dataset of patients with sepsis derived from patient encounters at 25 locations over a 13-year period, there was only modest agreement between three commonly used criteria for sepsis case identification. Cohorts developed based on criteria for CDC ASE, Sepsis-3, and explicit ICD coding were disparate in characteristics and outcomes of identified patients. Further, patients meeting both CDC ASE and Sepsis-3 criteria for sepsis and septic shock generally had higher illness severity and in-hospital mortality incidence compared with patients meeting only one of the criteria. Considering the crucial role of rigor and reproducibility in the biomedical sciences, transparency in the data workflow process for cohort creation using electronic health record data is imperative [24]. Through our use of an open science framework and detailed approach descriptions we provide the necessary information for reproducibility. While other workflows and sepsis cohort identification descriptions exist, ours is the first to include all required components from beginning (data extraction) to end (criteria application and case identification). Our workflow can be immediately implemented by Epic® users, the electronic health record vendor holding 78% of medical records in the United States. While data transformation is a separate pre-processing step for other published sepsis data workflows [5, 15], our approach seamlessly integrates this necessary step. For users of other electronic health record vendors our approach remains applicable since the variables outlined in Supplementary Material 1: eTable 1 can be considered a minimum clinical dataset for sepsis cohort identification. After data extraction our programming code for data pre-processing and case identification can be readily deployed. Relationship to previous studies Our approach to sepsis case identification according to the Sepsis-3 criteria overcomes limitations of the OpenSep pipeline [15]. OpenSep utilizes the quick SOFA score for patients outside of an ICU, which has suboptimal performance [25]. Additionally, for ICU patients OpenSep assesses for the presence of any SOFA score ≥ 2, which assumes no organ dysfunction at baseline. Our workflow evaluates for a SOFA change from baseline ≥ 2 in all patients, consistent with the Sepsis-3 clinical construct [2, 3]. Studies evaluating agreement between sepsis case identification methods have reported varying findings. Our results are consistent with a study that found fair to moderate agreement between CDC ASE, Sepsis-3, and ICD criteria (maximum κ and Jaccard index 0.42 and 0.30, respectively), and different patient demographics and clinical outcomes by cohort [26]. Our data build upon their findings in medical patients to show similar results in a cohort 11-fold larger in size and inclusive of surgical patients. The key differentiator leading to disagreement is the suspected infection component of the criteria, with CDC ASE requirements having greater specificity than Sepsis-3 clinical criteria. Yet, our findings contrast other studies reporting good agreement between criteria, meriting further discussion of methodology [7, 14, 27, 28]. Although there was high case identification agreement in a study comparing organ dysfunction criteria for CDC ASE and Sepsis-3, the use of the same standards for presumed infection and the same time epoch for concomitant criteria fulfillment within the comparison limit the results [7]. Indeed, while CDC ASE and Sepsis-3 criteria both have two core components (infection and organ dysfunction) and time epoch requirements surrounding microbial cultures for concomitant occurrence, each of these components differs between criteria. Our findings likely diverge from this study because we incorporated each of these elements into the case identification procedures for our comparison. Other studies finding high agreement either did not describe case identification methods in adequate detail for comparison [27], or used infection criteria inconsistent with the framework for implementing Sepsis-3 criteria for coding and research [2, 14]. Our results are also consistent with a previous study showing increasing application of sepsis ICD (administrative) codes over time and higher mortality in patients identified by ICD codes compared with those identified with CDC ASE criteria [5]. A novel finding is the higher prevalence of SARS-CoV-2 positivity in patients identified with sepsis and septic shock only by ICD coding compared with those identified only with CDC ASE or Sepsis-3 clinical criteria. This finding suggests patients with viral sepsis may not be well captured by CDC ASE and Sepsis-3 clinical criteria in their current form. Implications of the findings Our findings of modest agreement between sepsis and septic shock criteria and different patient characteristics and outcomes in isolated cohorts by criteria have important implications. Most notably, conclusions from observational studies assessing sepsis epidemiology, prediction models for outcomes and treatment effects, and observational causal inference are based on the context of the case identification criteria used. As such, consensus criteria for retrospective identification of sepsis and septic shock patients are urgently needed. Machine learning tools may prove useful for this work [6]. In the interim, the use of several approaches for retrospective case identification and inclusion of only patient cases meeting multiple criteria appears most prudent. Whether cohorts identified by one or multiple criteria represent unique sepsis subclasses (i.e., subphenotypes) should be further explored [29]. Each of the assessed criteria has advantages and disadvantages for sepsis case identification from the electronic health record, which have been detailed elsewhere [6]. Briefly, the CDC ASE criteria have high specificity but lower sensitivity, and the septic shock criteria omit elements of the Sepsis-3 clinical construct (e.g., lactate > 2 mmol/L) [6, 9]. The Sepsis-3 clinical criteria have high sensitivity but lower specificity, and the reliance on SOFA scores in their original form may not adequately capture current management (e.g., vasopressin administration or high-flow oxygen therapy) [6, 9, 30]. The ICD coding criteria are internationally standardized and easily extracted from the electronic health record, but are subject to provider documentation and have suboptimal diagnostic accuracy [11]. Using only explicit ICD codes for retrospective sepsis and septic shock case identification appears perilous. The sole application of ICD codes in our dataset would have omitted 83% of patients with sepsis and 54% of patients with septic shock cases identified with either CDC ASE or Sepsis-3 clinical criteria. Additionally, 21% of patients with explicit ICD coding for septic shock did not have evidence for continuous vasopressor administration, which is a component of the consensus septic shock clinical criteria. Further, sepsis and septic shock onset dates cannot be reliably discerned with ICD coding alone. Researchers applying explicit ICD codes alone for sepsis case identification must understand the selection bias and limitations created with this decision [5]. Strengths and limitations Our study has several strengths. We utilized an open-source framework to enhance methodological soundness and replicability, including annotated rationale for decision-making for ambiguous criteria components, enabling collaboration across the field to promote and facilitate best standards for sepsis and septic shock identification. We assessed emergency department visits in addition to hospital admissions and considered care transfers within our health system as one continuous encounter, providing data throughout acute care episodes for case adjudication and patient characterization. We restricted cohort inclusion to the first sepsis encounter on the patient level to reduce bias and align with sepsis pathobiology. Indeed, of all sepsis encounters identified, 35% of CDC ASE, 65% of Sepsis-3, and 20% of ICD encounters were recurrences of sepsis in survivors. Finally, our sepsis and septic shock cohort is large and diverse, inclusive of both medical and surgical patients and patients from two different US regions, increasing the robustness of our findings. To our knowledge, this is the largest cohort of patients with sepsis and septic shock identified with Sepsis-3 clinical criteria described in the literature, and the second-largest cohort identified with CDC ASE criteria. Several limitations warrant discussion. Although our workflow incorporated data-driven and manual checks for case identification, we did not compare performance with another tool such as OpenSep. However, our workflow was meticulously created and reviewed collaboratively by clinicians and data scientists, enhancing validity. Additionally, while we sought to adhere to published guidance for sepsis case identification, some components required decision-making. Although we transparently describe our approach, we did not explore alternative decisions (e.g., omission of hyperlactatemia as an organ dysfunction criterion for CDC ASE, or calculation of norepinephrine-equivalent doses for cardiovascular SOFA sub-score determination for Sepsis-3 criteria), which could have altered case adjudication and cohort comparisons. Further, our observed mortality for sepsis and septic shock cases is lower than in previous descriptions. While we cannot rule out misspecification of our case finding workflow, we hypothesize the lower observed mortality is reflective of process of care improvements over time in our analysis of a contemporary period inclusive of patient admissions after implementation of the US Centers for Medicare & Medicaid Services Severe Sepsis and Septic Shock: Management Bundle (SEP-1) quality measure. Lastly, despite our substantial cohort size and assessment of 25 locations, we only included patients from one integrated healthcare system, which may limit generalizability. Conclusions We describe an open-source data workflow for retrospective sepsis case identification, which will be useful for individuals seeking to build and analyze sepsis cohorts. Our cohort of sepsis and septic shock patients is large and diverse. There was inadequate agreement between cases identified with CDC ASE, Sepsis-3, and ICD criteria. Clinicians, researchers, and quality improvement practitioners retrospectively identifying patient cohorts with sepsis and septic shock should consider applying multiple criteria. Supplementary Information Supplementary Material 1. Supplementary Material 1.