2019 CYIR Finalized

5 3. No prospective evaluation was performed, and thus, the “real world” role of the AI Clinician is unknown. 4. Multiple modeling and statistical assumptions could be varied in future evaluations, such as the handling of missing data or choice of states for which treatments were optimized (concur- rent vs. future). 5. It is hypothesis generating, and consistent with growing body of observational data, that the optimal resuscitation policy from the AI Clinician favored higher doses of vasopressors with less use of IV fluids compared to clinician choices. BUNDLED CARE Levy MM, Gesten FC, Phillips GS, Terry KM, Seymour CW, Prescott HC, Friedrich M, Iwashyna TJ, Osborn T, Lemeshow S. Mortality Changes Associated with Mandated Public Reporting for Sepsis. The Results of the New York State Initiative. Am J Respir Crit Care Med. 2018; 198(11): 1406-1412. Summary Public reporting and performance measurement of health- care programs are intended to improve patient outcomes and reduce cost. In sepsis, there is limited experience and uncertainty surrounding these policy measures and their ability to drive clinician behavior. In 2013, New York State mandated bundled care for all patients with sepsis and septic shock, including a 3-hr bundle of blood cultures antibiotics, and serum lactate measurement, and a 6-hour bundle including a 30 ml/kg fluid bolus, administration of vasopressors, and re-measurement of serum lactate. In this study, Levy et al. reports on the rates of bundle initiation, completion, and risk-adjusted association with outcomes during the first 2 years after implementation. Among 91,557 patients, more than 8 of 10 had a sepsis protocol initiated, and the 3-hour bundle compliance increased from approxi- mately 53 to 65% during the study period. Compliance with the 6-hour bundle was lower, increasing from 24 to 31%. Using an extensive set of risk-adjustment variables, the authors modeled the association between bundle compli- ance and in-hospital mortality, finding that over 2 years, the mortality rate decreased from 29 to 24% (p<0.001). Meanwhile, septic patients in whom the protocol was not initiated had a stable risk-adjusted mortality. Although not causal, these data suggest a strong relationship between increasing compliance with the state mandated sepsis bundles and improved risk-adjusted outcomes. Comments 1. This study evaluated a first-in-the-nation healthcare policy aimed at improving sepsis outcomes through statewide man- dated protocols and public reporting. 2. Audits of case ascertainment and individual data elements sug- gested high accuracy for sepsis recognition. 3. This report included all patients in the New York State Clinical Sepsis database, not just those recognized in the emergency department. 4. No data was available about the pre-policy period or from states in the nearby geographic area, which could provide con- text to the temporal trends. 5. The study was not randomized, but rather a retrospective analysis of an existing cohort, and therefore we cannot make causal inferences about the relationship between sepsis bundle completion and patient outcomes. SEPSIS SUBCLASSES Antcliffe DB, Burnham KL, Al-Beidh F, Santhakumaran S, Brett SJ, Hinds CJ, Ashby D, Knight JC, Gordon AC. Transcriptomic Signatures in Sepsis and a Differential Response to Steroids: From the VANISH Randomized Trial. Am J Respir Crit Care Med , 2018; Oct, EPUB ahead of Print Summary Not all sepsis is the same. Recently, sepsis endotypes (e.g. subclasses with similar molecular profiles) were derived using genome wide gene expression profiling, identify- ing one type (SRS1) that is relatively immunosuppressed and one type (SRS2) that is relatively immunocompetent. Previously, these endotypes are reported to be prognostic but not found to be predictive of treatment response. In this study, Antcliffe et al. performed a secondary analy- sis of a double blind, factorial design randomized clinical trial of vasopressors and corticosteroids to determine if patients with SRS transcriptomic profiles had differential treatment responses. Among a 43% (N=176) subset of RCT patients with samples, the authors found that SRS types were distributed evenly. There was no significant interaction between SRS types and vasopressor treatment (Norepinephrine vs. Vasopressin, p=0.50), but there was a significant interaction for corticosteroid therapy (p=0.02). In other words, hydrocortisone treatment was only associated with increased mortality in those with the immunocompe- tent SRS2 phenotype (OR=7.9, 95%CI=1.6-39.9) but not SRS1. These data require validation but are a proof of concept that future trials should consider incorporation of sepsis subclasses in design and measurement of treatment effects. Comments 1. Sepsis endotypes are prognostic in other studies, but are not uniformly associated with treatment response. 2. These data are one of the first to report a treatment response linked to SRS transcriptomic profiles in sepsis. 3. More than half of patients in the VANISH trial did not have samples and were not included, which raises generalizability as a limitation. SEPSIS AND SEPTIC SHOCK

RkJQdWJsaXNoZXIy MTM1ODMw