Sushant Govindan

Is Statin Use Associated with Reduced Mortality After Pneumonia? A Systematic Review and Meta-analysis

OBJECTIVE The objective of this study was to perform a systematic review and meta-analysis of the effects of statins on mortality following pneumonia. METHODS We searched MEDLINE, EMBASE, BIOSIS, Cochrane CENTRAL Register of Controlled Trials, Cambridge Scientific Abstracts, BIOSIS, and Scopus. Studies were included if they involved: participants ≥18 years of age; patients with community-acquired pneumonia; current statin users; and reported overall or adjusted mortality after pneumonia. RESULTS Of 491 citations identified, 13 studies involving 254,950 patients met eligibility criteria. Pooled unadjusted data showed that statin use was associated with lower mortality after pneumonia (odds ratio [OR] 0.62, 95% confidence interval [CI], 0.54-0.71). Pooling of adjusted data also showed reduced mortality after pneumonia (OR 0.66, 95% CI, 0.55-0.79). However, this effect was attenuated in subgroup analysis by confounders and in prospective studies. CONCLUSIONS Although statin use is associated with decreased mortality after pneumonia, this effect weakens in important subgroups. Only a randomized controlled study can fully explore the link between statins and pneumonia mortality.

Death certificates underestimate infections as proximal causes of death in the U.S.

Background Death certificates are a primary data source for assessing the population burden of diseases; however, there are concerns regarding their accuracy. Diagnosis-Related Group (DRG) coding of a terminal hospitalization may provide an alternative view. We analyzed the rate and patterns of disagreement between death certificate data and hospital claims for patients who died during an inpatient hospitalization. Methods We studied respondents from the Health and Retirement Study (a nationally representative sample of older Americans who had an inpatient death documented in the linked Medicare claims from 1993–2007). Causes of death abstracted from death certificates were aggregated to the standard National Center for Health Statistics List of 50 Rankable Causes of Death. Centers for Medicare and Medicaid Services (CMS)-DRGs were manually aggregated into a parallel classification. We then compared the two systems via 2×2, focusing on concordance. Our primary analysis was agreement between the two data sources, assessed with percentages and Cohens kappa statistic. Results 2074 inpatient deaths were included in our analysis. 36.6% of death certificate cause-of-death codes agreed with the reason for the terminal hospitalization in the Medicare claims at the broad category level; when re-classifying DRGs without clear alignment as agreements, the concordance only increased to 61%. Overall Kappa was 0.21, or “fair.” Death certificates in this cohort redemonstrated the conventional top 3 causes of death as diseases of the heart, malignancy, and cerebrovascular disease. However, hospitalization claims data showed infections, diseases of the heart, and cerebrovascular disease as the most common diagnoses for the same terminal hospitalizations. Conclusion There are significant differences between Medicare claims and death certificate data in assigning cause of death for inpatients. The importance of infections as proximal causes of death is underestimated by current death certificate-based strategies.

Do Clinicians Know Which of Their Patients Have Central Venous Catheters?: A Multicenter Observational Study

Context Central venous catheters (CVCs) are commonly used to care for hospitalized patients; however, their continued presence creates substantial risks, including infection and thrombosis. Although indwelling CVCs should be removed as soon as they are no longer needed, clinicians may not be aware of their presence. Contribution This study found that many interns, residents, and attending physicians were unaware that their patients had indwelling catheters, including triple-lumen and peripherally inserted central catheters. Implication Increased efforts are needed to ensure that clinicians are mindful of the presence of their patients CVCs. The Editors Central venous catheters (CVCs) are instrumental for the safe and comprehensive care of many hospitalized patients. Often inserted in intensive care unit (ICU) and non-ICU settings, CVCs provide reliable venous access for tasks ranging from hemodynamic monitoring to delivery of irritants, vesicants, and intravenous antibiotics. In adults, 2 devices are most used in this context: nontunneled triple-lumen catheters placed in the subclavian, jugular, or femoral veins and peripherally inserted central catheters (PICCs) inserted into upper-extremity veins (1, 2). Despite their many advantages, triple-lumen catheters and PICCs are associated with important risks, including central lineassociated bloodstream infections (CLABSIs) and venous thromboembolism (3, 4). In particular, PICC-related CLABSI and thromboembolism have recently garnered attention owing to their frequency, attributable cost, and potential for prevention (5, 6). Because the risk for these adverse outcomes increases with time, early removal of CVCs that are no longer clinically warranted is a key strategy for prevention (7, 8). However, accumulating evidence suggests that clinicians often do not remove unnecessary CVCs. For example, a study done at a large academic medical center found many patients with PICCs that were idle and not clinically justifiable (3). In another study, 6.6% of CVCs in non-ICU settings were found to be inappropriate and clinically unnecessary at the time of review (4). These findings are not unique to the United Statesconcerns about inappropriately prolonged use of vascular access devices are well-documented worldwide (912). In survey-based studies of inpatient providers, nearly half of all hospitalists stated that they had, at least once, forgotten that their patient had a PICC in situ (13, 14). These findings mirror trends noted with indwelling urinary catheters, in which 1 in 3 physicians was unaware that these devices were present (15). Given this background, we sought to determine how often interns, residents, general medicine attendings, hospitalists, and subspecialists know which of their hospitalized patients have a PICC or triple-lumen catheter. We hypothesized that clinicians who write orders for or those who are most proximal to patients (for example, interns and hospitalists) would be most likely to correctly identify which of their patients have CVCs. Further, we postulated that clinicians who insert CVCs (such as critical care specialists) or consciously deliberate on the choice of a vascular access device (such as hematologists or oncologists) would be more likely to be aware that a device was present. Methods Patients and Study Population Between April 2012 and September 2013, we conducted face-to-face interviews with hospitalized patients and their responsible clinicians at 3 academic medical centers in the United States. A responsible clinician was defined as an intern, resident, physician extender (for example, nurse practitioner or physician assistant), or attending physician who had provided care to a patient for at least 24 hours. Housestaff were defined as interns (year 1 of training), residents (beyond year 1 of training), or physician extenders who cared for patients under the supervision of an attending physician. At each site, patients and providers were randomly selected from general medicine teaching, hospitalist-only, and subspecialty services that often use CVCs (such as cardiology, gastroenterology and hepatology, hematology and oncology, and critical care) and were primarily responsible for patient care (for example, not consultants). Thus, any provider on duty or patient receiving care in a specialty or discipline of interest was eligible for study inclusion. Eligible patients were identified using electronic patient lists for provider teams at each site. At 2 sites, patients were approached for participation and interviewed for the presence of a CVC. At 1 site, CVCs were identified through use of a validated electronic tool (98% accuracy at correctly identifying CVC presence); these patients were included but not directly examined for device presence. After patients were interviewed or electronically identified, providers were approached to ascertain their awareness of device presence. Clinicians were blinded to which patients were participating in the study and were queried for all patients on their roster. Survey Methods Before morning team rounds, we approached patients to seek written informed consent for participation at 2 of our 3 sites. If patients could not provide consent and a family member was available, consent was obtained from next of kin. Requirement for informed consent was waived at our third site. After consent was obtained, patients were interviewed and a focused examination was done to determine the presence of a PICC or triple-lumen catheter in the jugular, subclavian, or femoral veins at 2 study sites. Central venous catheters were defined as PICCs inserted in any upper-extremity vein or triple-lumen catheters placed in the neck, chest, or groin. Patients with specialty catheters, including hemodialysis catheters, small-bore catheters (such as Pro-Line CTs [Medcomp]), tunneled lines, and midlines were excluded. We surveyed patients in ICU and non-ICU settings; those on surgical services were excluded. Patients were surveyed only once during hospitalization; surveys were done weekly at all sites. After team rounds, we interviewed medical providers for each patient and asked, As of this morning, does your patient have a PICC or a triple-lumen catheter in the neck, chest, or groin? All clinicians were interviewed after morning rounds to ensure that they had seen the patient the day that the survey was administered. Clinicians were interviewed separately and were not notified of our visit beforehand. We allowed clinicians to use such materials as written notes or sign-outs during the interview but they were not allowed access to electronic health records. Because teaching attendings were responsible for all patients on the team, they were queried about CVC presence for all patients on their list. Subspecialists were similarly queried for all patients on their inpatient specialty teams. Interns, residents, and physician extenders were questioned only about patients for whom they were primarily responsible, regardless of assignment to a general medicine or specialist team. All clinicians were surveyed the day that patients were examined. Statistical Analysis Descriptive statistics for patient, provider, device, and site characteristics were used to define the study samples. The primary outcome of interest was unawareness of PICC or triple-lumen catheter presence. Given the categorical nature of the data, differences among provider types and training levels were compared using chi-square tests, where appropriate. Stata MP, version 13.0 (StataCorp), was used for all statistical analyses, and Pvalues less than 0.050 were considered statistically significant. Institutional review boards at each site provided ethical and regulatory approval for the study. Role of the Funding Source The study was not funded by any agency. Results Of the 1082 patients approached, 990 (91.5%) consented to participate in the study. For these 990 patients, we did 1881 clinician assessments across the 3 study sites (Table 1). Clinician responses from interns (454), residents and physician extenders (513), general medicine teaching attendings (245), subspecialty attendings (176), intensivists (95), and hospitalists (398) were included. An average of 1.9 clinician assessments were associated with each patient. Table 1. Patient, Provider, and Site Characteristics The overall prevalence of CVCs (triple-lumen catheter or PICC) was 21.1% (209 of 990). More than one half of the 209 devices were PICCs (n= 126 [60.3%]); the remaining 83 devices were triple-lumen catheters inserted in the neck (n= 41 [19.6%]), chest (n= 24 [11.5%]), or groin (n= 18 [8.6%]). A total of 47.0% (39 of 83) of triple-lumen catheters were found in patients in ICU settings; conversely, 92.9% (117 of 126) of PICCs were found in non-ICU patients. For the 209 patients with CVCs, 21.2% (90 of 425) of responsible clinicians were unaware of the presence of a triple-lumen catheter or PICC (Table 2). Unawareness of PICCs was greatestmore than 1 in 4 responsible clinicians (25.1% [60 of 239]) were not aware of their presence. Lack of awareness of a triple-lumen catheter or PICC varied from 16.3% to 31.1% across sites (P= 0.038) and was most pronounced in non-ICU settings, where PICCs are most common (24.8% vs. 12.6% in non-ICU and ICU settings; P= 0.005). Of note, a small but substantial number of clinicians (5.6% [82 of 1456]) stated that their patients had a CVC when no device was found on examination. Table 2. Awareness of CVC Presence We tested whether proximity to patients was associated with awareness of device presence. Although interns were the clinicians most likely to write orders, almost 1 of every 5 surveyed was not aware that his or her patient had a triple-lumen catheter or PICC (19.1% [22 of 115]). Although medical residents were more frequently aware of device presence than interns, the difference did not reach statistical significance (13.8% vs. 19.1%; P= 0.27). However, teaching

Issues of Survivorship Are Rarely Addressed during Intensive Care Unit Stays. Baseline Results from a Statewide Quality Improvement Collaborative

UNLABELLED RATIONALE/OBJECTIVE: In the context of increasing survivorship from critical illness, many studies have documented persistent sequelae among survivors. However, few evidence-based therapies exist for these problems. Support groups have proven efficacy in other populations, but little is known about their use after an intensive care unit (ICU) stay. Therefore, we surveyed critical care practitioners regarding their hospitals practice regarding discussing post-ICU problems for survivors with patients and their loved ones, communicating with primary care physicians, and providing support groups for current or former patients and families. METHODS A written survey was administered to 263 representatives of 73 hospitals attending the January 2013 annual meeting of the Michigan Health and Hospitals Association Keystone ICU initiative, a quality improvement collaborative focused on enhancing outcomes across Michigan ICUs. RESULTS There were 174 completed surveys, a 66% response rate. Representatives included staff nurses, nursing leadership, physicians, hospital administrators, respiratory therapists, and pharmacists. Sixty-nine percent of respondents identified at least one issue facing ICU survivors after discharge. The concerns most commonly identified by these ICU practitioners were weakness, psychiatric pathologies, cognitive dysfunction, and transitions of care. However, most respondents did not routinely discuss post-ICU problems with patients and families, and only 20% had a mechanism to formally communicate discharge information to primary care providers. Five percent reported having or being in the process of creating a support group for ICU survivors after discharge. CONCLUSIONS Despite growing awareness of the problems faced by ICU survivors, in this statewide quality improvement collaborative, hospital-based support groups are rarely available, and deficiencies in transitions of care exist. Practice innovations and formal research are needed to provide ways to translate awareness of the problems of survivorship into improved outcomes for patients.

Identifying Barriers to Delivering the Awakening and Breathing Coordination, Delirium, and Early Exercise/Mobility Bundle to Minimize Adverse Outcomes for Mechanically Ventilated Patients: A Systematic Review

Background Improved outcomes are associated with the Awakening and Breathing Coordination, Delirium, and Early exercise/mobility bundle (ABCDE); however, implementation issues are common. As yet, no study has integrated the barriers to ABCDE to provide an overview of reasons for less successful efforts. The purpose of this review was to identify and catalog the barriers to ABCDE delivery based on a widely used implementation framework, and to provide a resource to guide clinicians in overcoming barriers to implementation. Methods We searched MEDLINE via PubMed, CINAHL, and Scopus for original research articles from January 1, 2007, to August 31, 2016, that identified barriers to ABCDE implementation for adult patients in the ICU. Two reviewers independently reviewed studies, extracted barriers, and conducted thematic content analysis of the barriers, guided by the Consolidated Framework for Implementation Research. Discrepancies were discussed, and consensus was achieved. Results Our electronic search yielded 1,908 articles. After applying our inclusion/exclusion criteria, we included 49 studies. We conducted thematic content analysis of the 107 barriers and identified four classes of ABCDE barriers: (1) patient‐related (ie, patient instability and safety concerns); (2) clinician‐related (ie, lack of knowledge, staff safety concerns); (3) protocol‐related (ie, unclear protocol criteria, cumbersome protocols to use); and, not previously identified in past reviews, (4) ICU contextual barriers (ie, interprofessional team care coordination). Conclusions We provide the first, to our knowledge, systematic differential diagnosis of barriers to ABCDE delivery, moving beyond the conventional focus on patient‐level factors. Our analysis offers a differential diagnosis checklist for clinicians planning ABCDE implementation to improve patient care and outcomes.

Did They Just Prove That a Diagnosis of “Septic Shock” Is Meaningless?

Annane and colleagues’ oft-cited 2005 review of septic shock in The Lancet features a lovely set of graphics (1). In their Figure 1, there is a small oval at the top labeled “Bacteria.” This generic infection triggers at least eight separately identified effector pathways, which ramify out to show the multiple systems that lead “from bacteria to disease.” In their Figure 2, dozens of intracellular interactions are laid out, but the only vestige of the bacteria is an extracellular lipopolysaccharide. By 2013, Figure 1 of Angus and van der Poll’s NEJM review is entirely about the “Host Response in Severe Sepsis”—the pathogens are nearly invisible (2). In this understanding of severe sepsis, the story is about the host response, particularly the dysregulated inflammatory and coagulopathic cascades. Pathogens enter only to the extent that they create physiologically interesting molecular patterns that trigger this host response. In this issue of the Journal (pp. 1204–1213), the Co-operative Antimicrobial Therapy of Septic Shock (CATSS) Database Research Group, led by Dr. Leligdowicz, seeks to inject a note of discord into this perspective (3). This is a group well known to intensive care unit practitioners and researchers alike—the group whence our best evidence for the critical importance of time to antibiotics in septic shock came (4). In an expanded database, the authors now ask, are all infections really the same once they produce septic shock? Leligdowicz and colleagues suggest there are important differences within septic shock. The group examined a cohort of nearly 8,000 patients diagnosed with septic shock. They found that there was clinically meaningful and statistically significant variation in hospital mortality as a function of the source of infection. Adjusted mortality varied among sites from about one-third (diverticulitis and obstruction-related urinary tract infection) to nearly three-fourths (several abdominal infections). This variation persisted after adjusting for a multitude of predisposing and downstream factors, including year of admission, demographics, 12 comorbidities, and even Acute Physiology and Chronic Health Evaluation II (APACHE II) score. The authors suggest that we should take into account sources of infection so that patients are appropriately risk stratified and all potential factors impacting mortality are evaluated for interventions. The authors have shown that there are crucial differences in short-term outcomes by source of infection in patients with septic shock. Have they thereby proven the “host response” consensus to be wrong? More generally, have they shown that our current understanding of “sepsis” as a meaningful diagnosis is too severe an oversimplification? These questions hinge on what exactly we want from a diagnosis. The conflict over the Berlin definition of acute respiratory distress syndrome may be interpreted in a similar light (5, 6). It may be, we would like to suggest, that we want too many things from a single diagnosis—even a disease diagnosis, let alone an admittedly “syndromic” diagnosis (see Table 1). Table 1: The Uses of Diagnosis For some situations, particularly those of research, a diagnosis should be straightforward: it is a clinical representation of a unique pathological disturbance. What we want from a diagnosis is to define a sufficiently homogenous clinical entity for which we can work to identify the specific mechanism that produces said diagnosis. Such an understanding of diagnosis is particularly useful for driving forward animal-based research and ensuring close correlations between animal and human models of a condition. In contrast, at other times, the point of making a diagnosis is that it allows one to bring treatment to bear. Distinctions in diagnosis that do not lead to differences in treatment may be intellectually intriguing, but are thought to be of no clinical consequence. The art of diagnosis is then the art of matching efficacious treatments to the patients who will benefit from those interventions. In still other situations, we want a diagnosis to imply a coherent natural history and predictable course. A diagnosis that implies too wide an array of outcomes can lead to little informed patient decision making. In the long term, such a diagnosis implies the need to simply wait and see. In the short term, such variability precludes the ability to detect when things are going off track, or even when a second look is necessary to reconsider an alternative diagnosis. From this prognostic standpoint, consistency of future course is the key requirement for a workable diagnosis. Finally, sometimes the point of a diagnosis is communication—with ourselves and with other clinicians. A diagnosis is a cognitive shorthand that lets us reduce the complexity of an individual’s story into an archetype so we can remember what is going on with that patient and share that understanding with others to coordinate care. In an ideal world, these different desires would all point in one direction. In the real world, there is at best a loose coupling between any of these. Dr. Leligdowicz and colleagues’ work certainly suggests that septic shock, as currently implemented, fails to meet the coherent natural history standard. This is true at least for short-term mortality, and we can only speculate about longer-term patient-centered outcomes. Such heterogeneity implies concerns about a host of dependent issues, including unmeasured heterogeneity in clinical trials and insufficient risk adjustment in current severity of illness scores that lump everything together as “septic shock.” These findings also imply that there is important work to be done in understanding the variation in host response between sources of infection, and in finding practical ways to get all abdominal infections to behave more likely diverticulitis. This article may suggest that insufficient attention is being devoted to source-specific treatment investigations. In the meantime, septic shock remains a pragmatically useful organizing concept—although perhaps more like “cancer” than “HER2/neu–overexpressing stage IIA breast cancer.” Our resuscitations are still usefully guided by a notion of septic shock while we complete our efforts at source control (7, 8). Long-term outcome studies have not yet shown meaningful differences in outcomes across sites of infection, although existing efforts were underpowered to rule out such a possibility (9). The urgent challenge remains to integrate epidemiologic insights such as those from the current article into clinically relevant animal models and treatment trials. Dr. Leligdowicz and colleagues suggest there may be great benefits from bringing the source of the infection back into the forefront of septic shock research.

Do Clinicians Understand Quality Metric Data? An Evaluation in a Twitter-Derived Sample.

OBJECTIVE: Despite significant efforts and cost, quality metrics do not consistently influence practice. While research has focused on improving data through statistical risk‐adjustment, whether clinicians understand these data is unknown. Therefore, we assessed clinician comprehension of central line‐associated blood stream infection (CLABSI) quality metric data. DESIGN: Cross‐sectional survey with an 11‐item test of CLABSI data comprehension. Each question assessed 1 of 3 concepts concerning CLABSI understanding: basic numeracy, risk‐adjustment numeracy, and risk‐adjustment interpretation. Hypothetical data were used and presented in a validated format. PARTICIPANTS: Clinicians were recruited from 6 nations via Twitter to take an online survey. Clinician eligibility was confirmed by assessing responses to a question regarding CLABSI. MAIN MEASURES: The primary outcome was percent correct of attempted questions pertaining to the presented CLABSI data. RESULTS: Ninety‐seven clinicians answered at least 1 item, providing 939 responses; 72 answered all 11 items. The mean percentage of correct answers was 61% (95% confidence interval [CI], 57%‐65%). Overall, doctor performance was better than performance by nurses and other respondents (68% [95% CI, 63%‐73%] vs. 57% [95% CI, 52%‐62%], P = 0.003). In basic numeracy, mean percent correct was 82% (95% CI, 77%‐87%). For risk‐adjustment numeracy, the mean percent correct was 70% (95% CI, 64%‐76%). Risk‐adjustment interpretation had the lowest average percent correct, 43% (95% CI, 37%‐49%). All pairwise differences between concepts were statistically significant at P <0.05. CONCLUSIONS: CLABSI quality metric comprehension appears low and varies substantially among clinicians. These findings may contribute to the limited impact of quality metric reporting programs, and further research is needed.

Quick Sequential Organ Failure Assessment: Illness Severity Indicator, Clinical Decision Support Tool, or Both?*

Although the devastating consequences of sepsis are well-known (1), the diagnosis of this clinical syndrome has proven to be highly challenging (2). Until recently, sepsis was defined by two or more systemic inflammatory response syndrome (SIRS) criteria in the presence of infection (3). However, af

Inpatient notes: Sepsis-3 for hospitalists-sepsis without SIRS

Spurred by inconsistent clinical recognition of sepsis, significant mortality, and lack of reliability of the current definition, the Sepsis-3 Task Force recently published a new approach to sepsis (1). Sepsis was once believed to be the direct toxic effect of pathogens. The Sepsis-1 (1992) and -2 (2001) Task Forces argued, instead, that infection often kills through the bodys inflammatory response. Coupled to this reframing was the hope that systemic inflammatory response syndrome (SIRS) criteria would identify patients with such a response. For 20 years, we have taught that the diagnosis of sepsis is based on having 2 or more SIRS criteria with an infectious source. Alas, research on the SIRS hypothesis did not validate this hope. Among patients with infection and organ dysfunction in the intensive care unit (ICU), data suggest that 12% do not meet SIRS criteria (2). Further, nearly all hospitalized patients eventually meet SIRS criteria, regardless of sepsis status, if they stay long enough (3). Because SIRS lacks discriminant and convergent validity, the Sepsis-3 Task Force abandoned the use of SIRS and changed the recommended clinical approach to sepsis. The Sepsis-3 Task Force thus sought to identify characteristics that could raise clinical suspicion of sepsis among patients who were suspected or confirmed to be infected. Several scores, signs, and symptoms were evaluated in a cohort of 1.3 million medical records. Confirmatory analysis was done in 4 data sets of 706 399 encounters. Among infected patients in the ICU, the Sequential Organ Failure Assessment (SOFA) score most effectively identified patients who would die. Therefore, the Sepsis-3 Task Force recommended that the SOFA score be used to assist in identifying sepsis in critically ill patents, with an increase of 2 SOFA points (above baseline) raising probability of the diagnosis among infected patients. For infected patients outside the ICU, the novel and simpler quick SOFA (qSOFA) score was developed. This score assigns 1 point for each of the following characteristics: systolic blood pressure under 100 mm Hg, altered mental status (formally, Glasgow Coma Scale score of 15 or less, but practically any degree of acute alteration), or respiratory rate of 22 breaths/min or greater. For patients outside the ICU, qSOFA is superior to SIRS at predicting mortality (area under the receiver-operating characteristic curve, 0.81 vs. 0.76; P < 0.001) and noninferior to the more complex SOFA score (4). To be clear: qSOFA does not define sepsis outside the ICUrather, it is a warning of increased risk for death among those who are infected. Likewise, the absence of qSOFA points does not exclude sepsis, but it lowers the likelihood of a poor outcome. Progress aside, the response to the updated sepsis definition has not been uniformly positive (5). Some have called into question whether change was really necessary: SIRS positivity identifies nearly 88% of patients with sepsis, and sepsis mortality is decreasing worldwide with the current model. Concerns have also been raised about the interaction between the new definition and prior clinical experience and studies, as well as its generalizability given the relative lack of gender and geographic diversity in both the Sepsis-3 Task Force and data used in the development of these scores. Whether the Sepsis-3 definition will improve patient outcomes is an empirical question that needs further investigation. However, hospitalists should focus on the conceptual shift in transitioning from Sepsis-2 to Sepsis-3. The latter approach reinforces the primacy of a clinical definition of sepsis as life-threatening organ dysfunction caused by a dysregulated host response to infection. Using SOFA and qSOFA, clinicians are able to monitor organ dysfunction among infected patients that then informs the suspicion of sepsis and consequent risk for mortality. This differs from the Sepsis-2 paradigm in which sepsis is ruled in or out simply by counting SIRS criteria. It must be emphasized that our current understanding of sepsis precludes a simple check-box approach and perhaps even a dichotomous categorization of a patient as either septic or not septic. Instead, patients exhibit varying degrees of maladaptive dysregulation. The reality remains that all hospitalized patients are at risk for sepsis. Without a gold-standard definition, hospitalists must remain vigilant for the clinical features of the sepsis syndrome and make treatment decisions on the basis of the level and trajectory of suspicion. Sepsis-3 emphasizes clinical judgment in this diagnosis but also calls for additional research on sepsis to guide more reliable clinical definitions. We hope that these efforts will ultimately reduce sepsis mortality as well as the long-term morbidity that plagues its survivors.

Sample size implications of mortality definitions in sepsis: A retrospective cohort study

BackgroundMany randomized controlled trials (RCTs) employ mortality at a given time as a primary outcome. There are at least three common ways to measure 90-day mortality: first, all-location mortality, that is, all-cause mortality within 90 days of randomization at any location. Second, ARDSnet mortality is death in a healthcare facility of greater intensity than the patient was in prior to the hospitalization during which they were randomized. Finally, in-hospital mortality is death prior to discharge from the primary hospitalization of randomization. Data comparing the impact of these different measurements on sample size are lacking. We evaluated the extent to which event rates vary by mortality definition.MethodsThis was a retrospective cohort study of 30,691 patients hospitalized at Veterans Affairs (VA) hospitals for sepsis during 2009. 12,727 (41.5%) received care in an ICU setting. For each patient, we measured event rates for three different 90-day mortality outcomes: all-location mortality, ARDSnet mortality, and in-hospital mortality. We also calculated sample sizes necessary to power an example RCT given those event rates.ResultsAt 90 days, all-location mortality was 26.4% (95% CI 25.9–26.9%), ARDSnet mortality was 19.2% (95% CI 18.8–19.7%), and in-hospital mortality was 13.4% (95% CI 13.0–13.8%) (p < 0.01 all comparisons). These respective event rates result in different required sample sizes to achieve a 20% relative reduction in mortality with 80% power and a 5% false positive rate. Such a trial of VA sepsis patients would require 2080 patients for all-location mortality, 3080 for ARDSnet mortality, and 4796 for in-hospital mortality. Among sepsis patients mechanically ventilated in an ICU, 2438 experienced all-location mortality (46.2% [95% CI 44.8–47.5%]), 2181 experienced ARDSnet mortality (41.3% [95% CI 40.0–42.6%]), and 1894 experienced in-hospital mortality (36.0% [95% CI 34.7–37.3%]).ConclusionsEvent rates vary substantially in sepsis patients based on the chosen 90-day mortality definition. This could have important implications for RCT design trade-offs.