Effects of Compliance With the Early Management Bundle (SEP-1) on Mortality Changes Among Medicare Beneficiaries With Sepsis
2021; Elsevier BV; Volume: 161; Issue: 2 Linguagem: Inglês
10.1016/j.chest.2021.07.2167
ISSN1931-3543
AutoresSean R. Townsend, Gary Phillips, Reena Duseja, Lemeneh Tefera, Derek Cruikshank, Robert B. Dickerson, H. Bryant Nguyen, Christa Schorr, Mitchell M. Levy, R. Phillip Dellinger, William A. Conway, Warren S. Browner, Emanuel P. Rivers,
Tópico(s)Health Systems, Economic Evaluations, Quality of Life
ResumoBackgroundUS hospitals have reported compliance with the SEP-1 quality measure to Medicare since 2015. Finding an association between compliance and outcomes is essential to gauge measure effectiveness.Research QuestionWhat is the association between compliance with SEP-1 and 30-day mortality among Medicare beneficiaries?Study Design And MethodsStudying patient-level data reported to Medicare by 3,241 hospitals from October 1, 2015, to March 31, 2017, we used propensity score matching and a hierarchical general linear model (HGLM) to estimate the treatment effects associated with compliance with SEP-1. Compliance was defined as completion of all qualifying SEP-1 elements including lactate measurements, blood culture collection, broad-spectrum antibiotic administration, 30 mL/kg crystalloid fluid administration, application of vasopressors, and patient reassessment. The primary outcome was a change in 30-day mortality. Secondary outcomes included changes in length of stay.ResultsWe completed two matches to evaluate population-level treatment effects. In standard match, 122,870 patients whose care was compliant were matched with the same number whose care was noncompliant. Compliance was associated with a reduction in 30-day mortality (21.81% vs 27.48%, respectively), yielding an absolute risk reduction (ARR) of 5.67% (95% CI, 5.33-6.00; P < .001). In stringent match, 107,016 patients whose care was compliant were matched with the same number whose care was noncompliant. Compliance was associated with a reduction in 30-day mortality (22.22% vs 26.28%, respectively), yielding an ARR of 4.06% (95% CI, 3.70-4.41; P < .001). At the subject level, our HGLM found compliance associated with lower 30-day risk-adjusted mortality (adjusted conditional OR, 0.829; 95% CI, 0.812-0.846; P < .001). Multiple elements correlated with lower mortality. Median length of stay was shorter among cases whose care was compliant (5 vs 6 days; interquartile range, 3-9 vs 4-10, respectively; P < .001).InterpretationCompliance with SEP-1 was associated with lower 30-day mortality. Rendering SEP-1 compliant care may reduce the incidence of avoidable deaths. US hospitals have reported compliance with the SEP-1 quality measure to Medicare since 2015. Finding an association between compliance and outcomes is essential to gauge measure effectiveness. What is the association between compliance with SEP-1 and 30-day mortality among Medicare beneficiaries? Studying patient-level data reported to Medicare by 3,241 hospitals from October 1, 2015, to March 31, 2017, we used propensity score matching and a hierarchical general linear model (HGLM) to estimate the treatment effects associated with compliance with SEP-1. Compliance was defined as completion of all qualifying SEP-1 elements including lactate measurements, blood culture collection, broad-spectrum antibiotic administration, 30 mL/kg crystalloid fluid administration, application of vasopressors, and patient reassessment. The primary outcome was a change in 30-day mortality. Secondary outcomes included changes in length of stay. We completed two matches to evaluate population-level treatment effects. In standard match, 122,870 patients whose care was compliant were matched with the same number whose care was noncompliant. Compliance was associated with a reduction in 30-day mortality (21.81% vs 27.48%, respectively), yielding an absolute risk reduction (ARR) of 5.67% (95% CI, 5.33-6.00; P < .001). In stringent match, 107,016 patients whose care was compliant were matched with the same number whose care was noncompliant. Compliance was associated with a reduction in 30-day mortality (22.22% vs 26.28%, respectively), yielding an ARR of 4.06% (95% CI, 3.70-4.41; P < .001). At the subject level, our HGLM found compliance associated with lower 30-day risk-adjusted mortality (adjusted conditional OR, 0.829; 95% CI, 0.812-0.846; P < .001). Multiple elements correlated with lower mortality. Median length of stay was shorter among cases whose care was compliant (5 vs 6 days; interquartile range, 3-9 vs 4-10, respectively; P < .001). Compliance with SEP-1 was associated with lower 30-day mortality. Rendering SEP-1 compliant care may reduce the incidence of avoidable deaths. Take-home PointsStudy Question: What is the association between compliance with the Early Management Bundle, Severe Sepsis/Septic Shock (SEP-1) and 30-day mortality among Medicare beneficiaries?Results: In this post hoc analysis of data submitted by US hospitals to Medicare, compliance with SEP-1 was associated with a significant reduction in 30-day mortality. Among patients matched by propensity score using standard matching criteria, there was a 5.67% absolute risk reduction (ARR) in 30-day mortality, and when applying the most stringent literature-based matching criteria, there was a 4.06% ARR.Interpretation: If the relationship between SEP-1 compliance and 30-day mortality is causal, rendering SEP-1 compliant care likely reduces the incidence of avoidable deaths.FOR EDITORIAL COMMENT, SEE PAGE 303 Study Question: What is the association between compliance with the Early Management Bundle, Severe Sepsis/Septic Shock (SEP-1) and 30-day mortality among Medicare beneficiaries? Results: In this post hoc analysis of data submitted by US hospitals to Medicare, compliance with SEP-1 was associated with a significant reduction in 30-day mortality. Among patients matched by propensity score using standard matching criteria, there was a 5.67% absolute risk reduction (ARR) in 30-day mortality, and when applying the most stringent literature-based matching criteria, there was a 4.06% ARR. Interpretation: If the relationship between SEP-1 compliance and 30-day mortality is causal, rendering SEP-1 compliant care likely reduces the incidence of avoidable deaths. FOR EDITORIAL COMMENT, SEE PAGE 303 In the prelude to a national sepsis measure, hospitalization rates for sepsis more than doubled from 2000 to 2008.1Hall M.J. Williams S.N. DeFrances C.J. Golosinskiy A. Inpatient care for septicemia or sepsis: a challenge for patients and hospitals.NCHS Data Brief. 2011; : 1-8Google Scholar,2Liu V. Escobar G.J. Greene J.D. et al.Hospital deaths in patients with sepsis from 2 independent cohorts.JAMA. 2014; 312: 90-92Google Scholar During this time, sepsis was present in > 50% of US hospital deaths and was the costliest disease at $24 billion annually.3Gaieski D.F. Edwards J.M. Kallan M.J. Carr B.G. Benchmarking the incidence and mortality of severe sepsis in the United States.Crit Care Med. 2013; 41: 1167-1174Google Scholar,4Torio C (AHRQ), Moore B (Truven Health Analytics)National inpatient hospital costs: the most expensive conditions by payer, 2013. HCUP Statistical Brief #204. Agency for Healthcare Research and Quality, 2016Google Scholar Congress encouraged the Centers for Medicare and Medicaid Services (CMS) to improve sepsis care after gripping testimony from Ciaran Staunton, father of Rory Staunton, a boy who died of sepsis.5CSPANSenate Health Committee: father of Rory Staunton, who died of sepsis.https://www.c-span.org/video/?c4465991/senate-health-committee-father-rory-staunton-died-sepsisGoogle Scholar CMS identified sepsis as a priority and in 2013 asked the Measure Applications Partnership, a multistakeholder group convened by the National Quality Forum and required by Congress to review quality measures for CMS, to review measures for possible inclusion in Medicare's Inpatient Quality Reporting Program.6National Quality ForumMAP 2015 considerations for implementing measures in federal programs. Draft for public comment.https://www.qualityforum.org/Setting_Priorities/Partnership/Draft_Programmatic_Deliverable_Report.aspxGoogle Scholar In 2014, the National Quality Forum reendorsed a measure developed by Henry Ford Hospital (Detroit, MI), which CMS incorporated as SEP-1, into the Inpatient Quality Reporting Program on October 1, 2015. Controversy remains regarding the efficacy of SEP-1, which may have limited compliance with the measure.7Moreira H. Sinert R. How effective is the early management bundle for severe sepsis/septic shock?.JAMA Intern Med. 2020; 180: 716-717Google Scholar Using data from CMS, we examined the association between compliance with SEP-1 and 30-day mortality. We used chart-abstracted data from 1,312,024 patients obtained from October 1, 2015, to March 31, 2017, at 3,241 US hospitals. Hospitals submitted data from all or a predefined sample of patients with severe sepsis or septic shock. SEP-1 specifies an algorithm to ascertain compliance including a start time (time zero) set by suspicion of infection, two systemic inflammatory response syndrome criteria, and an organ dysfunction, or a provider's designation of time zero. Elements included measurement of lactate, obtaining blood cultures, broad-spectrum antibiotic administration, 30 mL/kg of crystalloid fluids for hypotension or lactate ≥ 4.0 mM, vasopressors for persistent hypotension, lactate remeasurement if elevated, and patient reassessment (see e-Table 1 for description of the SEP-1 measure). CMS' Clinical Data Abstraction Center randomly audits 600 hospitals annually for accuracy of abstracted data.8Department of Health and Human ServicesChart-abstracted data validation.https://qualitynet.org/inpatient/data-management/chart-abstracted-data-validationGoogle Scholar Patients ≥ 18 years of age with a qualifying International Classification of Diseases, Tenth Revision, Clinical Modification code were eligible for SEP-1 if hospitalized < 120 days (see e-Table 2 for qualifying codes). Hospital personnel audited the sample to verify patients met clinical criteria and excluded those that did not (see e-Table 3 for SEP-1 clinical criteria). The SEP-1 dataset included demographics, timestamps, diagnosis codes, and limited clinical data. We obtained Medicare-verified dates of death (Fig 1, box 2). Compliance was defined as passing all eligible SEP-1 elements. The algorithm excluded 327,687 patients leaving 333,770 in the SEP-1 eligible cohort (Fig 1, boxes 3 and 4). The Third International Consensus Definitions for Sepsis and Septic Shock were not applied because SEP-1 antedated their release.9Seymour C.W. Liu V.X. Iwashyna T.J. et al.Assessment of clinical criteria for sepsis: for the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3) [published correction appears in JAMA. 2016;315(20):2237].JAMA. 2016; 315: 762-774Google Scholar In overview of our methods, we analyzed the association between compliance with SEP-1 and mortality. We used propensity score matching to estimate the marginal treatment effect, the average effect at the population level of moving a cohort of patients from untreated to treated,10Greenland S. Interpretation and choice of effect measures in epidemiologic analyses.Am J Epidemiol. 1987; 125: 761-768Google Scholar specifically estimating the average treatment effect among the treated (ATT).11Imbens G. Nonparametric estimation of average treatment effects under exogeneity: a review.Review of Economics and Statistics. 2004; 86: 4-29Google Scholar We also estimated the conditional treatment effect, the average effect at the subject level of moving a subject from untreated to treated,10Greenland S. Interpretation and choice of effect measures in epidemiologic analyses.Am J Epidemiol. 1987; 125: 761-768Google Scholar using a hierarchical generalized linear model (HGLM). Mortality hereinafter refers to 30-day mortality unless otherwise stated. This study received a waiver of consent from the institutional review board for human research at Henry Ford Hospital, Detroit, Michigan (reference No. 12252). We sought to identify which variables were associated with both the exposure and the outcome, potentially confounding results. To inform our selection of variables, we relied on sepsis bundle studies (finding increasing compliance) and previously validated administrative sepsis mortality models (studying the model proposed by Darby et al12Darby J.L. Davis B.S. Barbash I.J. Kahn J.M. An administrative model for benchmarking hospitals on their 30-day sepsis mortality.BMC Health Serv Res. 2019; 19: 221Google Scholar most closely). To ensure that these relationships were properly mapped, we organized the collected variables as a directed acyclic graph13Lederer D.J. Bell S.C. Branson R.D. et al.Control of confounding and reporting of results in causal inference studies. Guidance for authors from editors of respiratory, sleep, and critical care journals [published correction appears in Ann Am Thorac Soc. 2019;16(2):283].Ann Am Thorac Soc. 2019; 16: 22-28Google Scholar (e-Fig 1) using a publicly available, web-based environment (http://dagitty.net). Studying the resultant graph, we determined that many of the variables that confounded this relationship were related to the severity of illness. Variables included comorbidities, infection sources, acute organ failures, demographics, among others. Following contemporary practice, we selected true confounders known prior to, or just after, any SEP-1 algorithm failure point for inclusion in our models.14Biondi-Zoccai G. Romagnoli E. Agostoni P. et al.Are propensity scores really superior to standard multivariable analysis?.Contemp Clin Trials. 2011; 32: 731-740Google Scholar, 15Austin P.C. Grootendorst P. Anderson G.M. A comparison of the ability of different propensity score models to balance measured variables between treated and untreated subjects: a Monte Carlo study.Stat Med. 2007; 26: 734-753Google Scholar, 16Brookhart M.A. Schneeweiss S. Rothman K.J. Glynn R.J. Avorn J. Stürmer T. Variable selection for propensity score models.Am J Epidemiol. 2006; 163: 1149-1156Google Scholar Variables were modeled as indicator covariates unless otherwise noted (see e-Table 4 for variable descriptions). Demographics included age (continuous), sex, race, and ethnicity. Hospital characteristics included square root of bed count (continuous), accreditation status, and urban, rural, or critical access designation. Comorbidities were defined as per Elixhauser et al17Elixhauser A. Steiner C. Harris D.R. Coffey R.M. Comorbidity measures for use with administrative data.Med Care. 1998; 36: 8-27Google Scholar and Quan et al.18Quan H. Sundararajan V. Halfon P. et al.Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data.Med Care. 2005; 43: 1130-1139Google Scholar Acute organ failures were defined as per Elias et al19Elias K.M. Moromizato T. Gibbons F.K. Christopher K.B. Derivation and validation of the acute organ failure score to predict outcome in critically ill patients: a cohort study.Crit Care Med. 2015; 43: 856-864Google Scholar (see e-Table 5 for ICD-9 to IDC-10 crosswalk). Infection sources (bacteremia, septicemia, fungal, peritoneal, heart, upper respiratory, lung, CNS, GI, skin, and other infections) were modeled as hierarchical infection categories by strength of the infection's association with unadjusted mortality as per Darby et al12Darby J.L. Davis B.S. Barbash I.J. Kahn J.M. An administrative model for benchmarking hospitals on their 30-day sepsis mortality.BMC Health Serv Res. 2019; 19: 221Google Scholar and Iwashyna et al20Iwashyna T.J. Odden A. Rohde J. et al.Identifying patients with severe sepsis using administrative claims: patient-level validation of the angus implementation of the international consensus conference definition of severe sepsis.Med Care. 2014; 52: e39-e43Google Scholar (see e-Appendix 1 for additional methods, e-Table 6 for International Classification of Diseases, Ninth Revision, Clinical Modification to International Classification of Diseases, Tenth Revision, Clinical Modification crosswalk, and e-Table 7 for model parameter estimates). Severity of illness covariates included lactate range (≤ 2.0, > 2.0 and < 4.0, and ≥ 4.0 mM), persistent hypotension, and septic shock. We estimated the propensity score for compliance using an HGLM using a binomial family and a logit link function accounting for patients clustered within hospitals.21Arpino B. Mealli F. The specification of the propensity score in multilevel observational studies.Computational Statistics & Data Analysis. 2011; 55: 1770-1780Google Scholar,22Arpino B. Cannas M. Propensity score matching with clustered data. An application to the estimation of the impact of caesarean section on the Apgar score.Stat Med. 2016; 35: 2074-2091Google Scholar We calculated the intraclass correlation coefficient then added patient-level fixed effects (infection source, acute organ failures, comorbidities, age, race, ethnicity, sex, initial lactate range, persistent hypotension, septic shock, discharge quarter) and hospital-level fixed effects (bed count; accreditation status; urban, rural, or critical access designation), iteratively checking fit.23Ene M. Leighton E.A. Blue G.L. Bell B.A. Multilevel models for categorical data using SAS® PROC GLIMMIX: the basics.https://support.sas.com/resources/papers/proceedings15/3430-2015.pdfGoogle Scholar The variable defining each hospital was included in both levels as a random effect. A similar HGLM predictive mortality model estimated conditional treatment effects. Calibration was assessed by plotting observed vs predicted outcome.12Darby J.L. Davis B.S. Barbash I.J. Kahn J.M. An administrative model for benchmarking hospitals on their 30-day sepsis mortality.BMC Health Serv Res. 2019; 19: 221Google Scholar Discrimination was assessed by exporting HGLM predictive probabilities then applying logistic regression.24Kiernan K. Insights into using the GLIMMIX procedure to model categorical outcomes with random effects.https://www.sas.com/content/dam/SAS/support/en/sas-global-forum-proceedings/2018/2179-2018.pdfGoogle Scholar Once the propensity score was generated, compliant cases were matched random seed, 1:1, without replacement, nearest neighbor using the Mahalanobis distance (constituted by logit of the propensity score [LPS] and binary lactate variables) within propensity score calipers.25Rubin D.B. Thomas N. Combining propensity score matching with additional adjustments for prognostic covariates.J Am Stat Assoc. 2000; 95: 573-585Google Scholar Given competing literature-based standards to assess variable balance, standard match targeted covariate absolute standardized mean differences (SMDs) of ≤ 0.25,26Rubin D.B. Using propensity scores to help design observational studies: application to the tobacco litigation.Health Services & Outcomes Research Methodology. 2001; 2: 169-188Google Scholar,27Stuart E.A. Matching methods for causal inference: a review and a look forward.Stat Sci. 2010; 25: 1-21Google Scholar and stringent match SMDs of ≤ 0.10.28Austin P.C. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples.Stat Med. 2009; 28: 3083-3107Google Scholar,29Austin P.C. Grootendorst P. Normand S.L. Anderson G.M. Conditioning on the propensity score can result in biased estimation of common measures of treatment effect: a Monte Carlo study.Stat Med. 2007; 26: 754-768Google Scholar We calculated the treated-to-control variance ratio.26Rubin D.B. Using propensity scores to help design observational studies: application to the tobacco litigation.Health Services & Outcomes Research Methodology. 2001; 2: 169-188Google Scholar,27Stuart E.A. Matching methods for causal inference: a review and a look forward.Stat Sci. 2010; 25: 1-21Google Scholar ATT was estimated as the mean mortality difference between groups after matching. We partitioned matched cohorts into deciles by propensity score. We used the general estimating equation with independent working correlation accounting for matched pairs and robust SEs. Estimands included the absolute risk reduction (ARR), relative risk (RR), number needed to treat (NNT), and marginal OR.29Austin P.C. Grootendorst P. Normand S.L. Anderson G.M. Conditioning on the propensity score can result in biased estimation of common measures of treatment effect: a Monte Carlo study.Stat Med. 2007; 26: 754-768Google Scholar, 30Austin P.C. An introduction to propensity score methods for reducing the effects of confounding in observational studies.Multivariate Behav Res. 2011; 46: 399-424Google Scholar, 31Austin P.C. The performance of different propensity score methods for estimating marginal odds ratios.Stat Med. 2007; 26: 3078-3094Google Scholar We used adaptive Gaussian quadrature to approximate maximum likelihood estimation.32Capanu M. Gönen M. Begg C.B. An assessment of estimation methods for generalized linear mixed models with binary outcomes.Stat Med. 2013; 32: 4550-4566Google Scholar To estimate conditional treatment effects, the compliance parameter estimate was exponentiated as a conditional adjusted OR (AORc). We studied quarter 4, 2015, to quarter 2, 2016, for element-level analysis because data were fully abstracted. We assessed length of stay as medians with interquartile ranges (IQRs) using the Wilcoxon rank-sum test. Regarding sensitivity analyses, we assessed whether clustering of patients within hospitals was accounted for by our HGLM repeating matches using preferential within-cluster matching, which matches within the same then other clusters.21Arpino B. Mealli F. The specification of the propensity score in multilevel observational studies.Computational Statistics & Data Analysis. 2011; 55: 1770-1780Google Scholar,22Arpino B. Cannas M. Propensity score matching with clustered data. An application to the estimation of the impact of caesarean section on the Apgar score.Stat Med. 2016; 35: 2074-2091Google Scholar We also assessed whether Mahalanobis distance improved on propensity score matching, repeating matches using the propensity score alone. We performed quantitative bias analysis to derive an unmeasured confounder capable of nullifying our estimates. Analyses were performed using SAS 9.3 (SAS Institute Inc). We selected true confounders from our causal directed acyclic graph for our models (e-Fig 1), and Table 1 depicts their distribution. The average age was 73.5 years. The commonest comorbidity was hypertension (70.9%), then fluid and electrolyte disorders (65.4%), and then diabetes (40%). The commonest acute organ failure was renal failure (47.9%) then respiratory failure (31.3%). The commonest infections were lung (34.3%), septicemia alone (24.5%), and genitourinary (24.3%) (Table 1). Unadjusted mortality was 26.7%.Table 1Case Characteristics of the SEP-1 Eligible Cohort Across Compliance and Noncompliance With the SEP-1 MeasureVariableAll Cases (N = 333,770)Compliant (n = 140,504; 42.1%)Noncompliant (n = 193,266; 57.9%)Hospital certification bed count Mean ± SD308.9 ± 273.4293.2 ± 259.5320.3 ± 282.6 Median240231249 Interquartile range138-390130-375144-405Hospital accreditation AOA/HFAP13,580 (4.1)5,598 (4.0)7,982 (4.1) CIHQ2,026 (0.6)1,052 (0.7)974 (0.5) DNV31,305 (9.4)12,620 (9.0)18,685 (9.7) NONE9,661 (2.9)3,554 (2.5)6,107 (3.2) TJC277,198 (83.0)117,680 (83.8)159,518 (82.5)Critical access hospital5,152 (1.5)2,255 (1.6)2,897 (1.5)Urban hospital location253,458 (75.9)106,039 (75.5)147,419 (76.3)Hospital region New England11,807 (3.5)4,987 (3.6)6,820 (3.5) Northeast27,725 (8.3)11,250 (8.0)16,475 (8.5) Mideast47,391 (14.2)20,151 (14.3)27,240 (14.1) Southeast75,951 (22.8)31,902 (22.7)44,049 (22.8) Great Lakes57,063 (17.1)22,697 (16.1)34,366 (17.8) South38,893 (11.7)16,260 (11.6)22,633 (11.7) Midwest15,950 (4.8)6,753 (4.8)9,197 (4.8) Mountain West9,096 (2.7)4,002 (2.9)5,094 (2.6) Southwest38,842 (11.6)17,973 (12.8)20,869 (10.8) Northwest11,052 (3.3)4,529 (3.2)6,523 (3.4)Discharge quarter Quarter 4, 201551,785 (15.5)18,158 (12.9)33,627 (17.4) Quarter 1, 201654,858 (16.4)22,154 (15.8)32,704 (16.9) Quarter 2, 201653,003 (15.9)23,754 (16.9)29,249 (15.1) Quarter 3, 201653,936 (16.2)21,764 (15.5)32,172 (16.7) Quarter 4, 201657,953 (17.4)25,402 (18.1)32,551 (16.8) Quarter 1, 201762,235 (18.6)29,272 (20.8)32,963 (17.1)Patient Hispanic ethnicity (yes)18,756 (5.6)7,527 (5.4)11,229 (5.8)Patient sex (male)169,804 (50.9)72,224 (51.4)97,580 (50.5)Patient race American Indian/Alaska Native2,210 (0.6)864 (0.6)1,346 (0.7) Asian6,983 (2.1)3,397 (2.4)3,586 (1.9) Black38,313 (11.5)14,529 (10.4)23,784 (12.3) Native Hawaiian/Pacific Islander941 (0.3)452 (0.3)489 (0.2) Unable to determine11,706 (3.5)4,920 (3.5)6,786 (3.5) White273,617 (82.0)116,342 (82.8)157,275 (81.4)Patient age at admission, y Mean ± SD73.5 ± 13.073.6 ± 13.073.4 ± 13.0 Median747574 Interquartile range66-8366-8366-83Length of stay, d Mean7.8 ± 7.47.1 ± 6.48.3 ± 8.1 Median656 Interquartile range3-103-94-10Persistent hypotension (yes)16,666 (5.0)5,008 (3.6)11,658 (6.0)Initial lactate level Not collected44,838 (13.4)0 (0.0)44,838 (23.2) ≤ 2.0 mM99,121 (29.7)57,833 (41.2)41,288 (21.4) > 2.0 and < 4.0 mM133,659 (40.1)70,426 (50.1)63,233 (32.7) ≥ 4.0 mM56,152 (16.8)12,245 (8.7)43,907 (22.7)Severe sepsis263,238 (78.9)120,479 (85.7)142,759 (73.9)Septic shock70,532 (21.1)20,037 (14.3)50,495 (26.1)Mortality88,998 (26.7)30,444 (21.7)58,554 (30.3) Severe sepsis63,510 (24.1)23,791 (19.8)39,719 (27.8) Septic shock25,488 (36.1)6,653 (33.2)18,835 (37.3)Site of infection Bacteremia160 (0.1)55 (0.1)105 (0.0) CNS849 (0.2)331 (0.2)518 (0.3) Fungal4,814 (1.4)1,887 (1.3)2,927 (1.5) GI3,712 (1.1)1,302 (0.9)2,410 (1.3) Genitourinary81,089 (24.3)35,425 (25.2)45,664 (23.6) Heart2,539 (0.8)980 (0.7)1,559 (0.8) Lung114,391 (34.3)50,878 (36.2)63,513 (32.9) Unknown1,238 (0.4)535 (0.4)703 (0.4) Other29,596 (8.9)12,600 (9.0)16,996 (8.8) Peritoneal6,778 (2.0)1,726 (1.2)5,052 (2.6) Septicemia only81,682 (24.5)31,871 (22.7)49,811 (25.8) Soft tissue6,131 (1.8)2,547 (1.8)3,584 (1.8) Upper respiratory tract791 (0.2)367 (0.3)424 (0.2)Organ failures Cardiovascular33,297 (10.0)11,580 (8.2)21,717 (11.2) Hematologic12,410 (3.7)3,471 (2.5)8,939 (4.6) Metabolic91,398 (27.4)34,173 (24.3)57,225 (29.6) Neurologic32,501 (9.7)12,457 (8.9)20,044 (10.4) Renal159,946 (47.9)62,215 (44.3)97,731 (50.6) Respiratory104,296 (31.3)39,230 (27.9)65,066 (33.7)Comorbidities AIDS995 (0.3)426 (0.3)569 (0.3) Alcohol abuse10,440 (3.1)3,901 (2.8)6,539 (3.4) Chronic blood loss anemia3,851 (1.2)1,410 (1.0)2,441 (1.3) Chronic pulmonary disease113,087 (33.9)49,357 (35.1)63,730 (33.0) Coagulopathy49,797 (14.9)18,726 (13.3)31,071 (16.1) Congestive heart failure102,126 (30.6)39,890 (28.4)62,236 (32.2) Deficiency anemias108,137 (32.4)45,025 (32.1)63,112 (32.7) Depression45,665 (13.7)19,894 (14.2)25,771 (13.3) Diabetes with chronic complications72,176 (21.6)29,548 (21.0)42,628 (22.1) Diabetes without chronic complications61,547 (18.4)25,896 (18.4)35,651 (18.5) Drug abuse7,223 (2.2)3,060 (2.2)4,163 (2.2) Fluid and electrolyte disorders218,179 (65.4)88,210 (62.8)129,969 (67.3) Hypertension236,558 (70.9)100,108 (71.3)136,450 (70.6) Hypothyroidism61,692 (18.5)26,225 (18.7)35,467 (18.4) Liver disease19,099 (5.7)6,901 (4.9)12,198 (6.3) Lymphoma5,912 (1.8)2,539 (1.8)3,373 (1.8) Metastatic cancer14,775 (4.4)5,822 (4.1)8,953 (4.6) Obesity51,505 (15.4)20,646 (14.7)30,859 (16.0) Other neurologic disorders67,194 (20.1)29,319 (20.9)37,875 (19.6) Paralysis31,207 (9.4)13,430 (9.6)17,777 (9.2) Peptic ulcer disease excluding bleeding4,914 (1.5)1,750 (1.3)3,164 (1.6) Peripheral vascular disease35,393 (10.6)14,066 (10.0)21,327 (11.0) Psychoses15,916 (4.8)7,207 (5.1)8,709 (4.5) Pulmonary circulation disease6,361 (1.9)2,388 (1.7)3,973 (2.1) Renal failure102,247 (30.6)40,503 (28.8)61,744 (32.0) Rheumatoid arthritis/collagen vascular diseases15,456 (4.6)6,637 (4.7)8,819 (4.6) Solid tumor without metastasis13,853 (4.2)5,757 (4.1)8,096 (4.2) Valvular disease30,745 (9.2)12,451 (8.9)18,294 (9.5) Weight loss55,562 (16.7)21,558 (15.3)34,004 (17.6)Values are No. (%) or as otherwise indicated. AOA = American Osteopathic Association; cert = certification; CIQH = Center for Improvement in Healthcare Quality; DNV = Det Norske Veritas; HFAP = Healthcare Facilities Accreditation Program; Q = quarter; TJC = The Joint Commission. Open table in a new tab Values are No. (%) or as otherwise indicated. AOA = American Osteopathic Association; cert = certification; CIQH = Center for Improvement in Healthcare Quality; DNV = Det Norske Veritas; HFAP = Healthcare Facilities Accreditation Program; Q = quarter; TJC = The Joint Commission. The intraclass correlation coefficients between hospital and compliance and hospital and mortality were 0.145 and 0.29, respectively. Factors most strongly associated with compliance were initial lactate value, discharge quarter, persistent hypotension, septic shock, critical access hospital status, among others (see e-Table 8 for compliance model parameter estimates). The mortality model associations were similar excepting discharge quarter and critical access hospital status (see e-Table 9 for mortality model parameter estimates). The compliance and mortality models showed good discrimination with C-statistics of 0.835 and 0.788, respectively. For the compliance model, the calibration plot's slope and intercept were 1.02 and 0.008, respectively, and the mortality model's slope and intercept were 1.02 and 0.004, respectively (e-Fig 2). Regarding match results, the caliper was set at 0.40 and 0.22 times the pooled estimate of the SD of the LPS for standard match and stringent match, respectively. Not all patients could be matched, leaving 245,740 (73.6%) and 214,032 (64.1%) in the standard match and stringent match analysis cohorts, respectively (Fig 1, boxes 6a and 6b). Observed characteristics among the patients whose care was compliant with SEP-1 were similar to their matched counterparts (see e-Tables 10 and 13 for standard match and stringent match distributions, respectively). In standard match and stringent match, all variables achieved absolute SMDs ≤ 0.25 and 0.10 and variance ratios between 0.5 and 1.3 and 0.7 a
Referência(s)