1
|
Kornblith LZ, Robles AJ, Conroy AS, Redick BJ, Howard BM, Hendrickson CM, Moore S, Nelson MF, Moazed F, Callcut RA, Calfee CS, Jay Cohen M. Predictors of postinjury acute respiratory distress syndrome: Lung injury persists in the era of hemostatic resuscitation. J Trauma Acute Care Surg 2019; 87:371-378. [PMID: 31033882 PMCID: PMC6660388 DOI: 10.1097/ta.0000000000002331] [Citation(s) in RCA: 12] [Impact Index Per Article: 2.4] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 01/02/2023]
Abstract
BACKGROUND Acute respiratory distress syndrome (ARDS) following trauma is historically associated with crystalloid and blood product exposure. Advances in resuscitation have occurred over the last decade, but their impact on ARDS is unknown. We sought to investigate predictors of postinjury ARDS in the era of hemostatic resuscitation. METHODS Data were prospectively collected from arrival to 28 days for 914 highest-level trauma activations who required intubation and survived more than 6 hours from 2005 to 2016 at a Level I trauma center. Patients with ratio of partial pressure of oxygen to fraction of inspired oxygen of 300 mmHg or less during the first 8 days were identified. Two blinded expert clinicians adjudicated all chest radiographs for bilateral infiltrates in the first 8 days. Those with left-sided heart failure detected were excluded. Multivariate logistic regression was used to define predictors of ARDS. RESULTS Of the 914 intubated patients, 63% had a ratio of partial pressure of oxygen to fraction of inspired oxygen of 300 or less, and 22% developed ARDS; among the ARDS cases, 57% were diagnosed early (in the first 24 hours), and 43% later. Patients with ARDS diagnosed later were more severely injured (ISS 32 vs. 20, p = 0.001), with higher rates of blunt injury (84% vs. 72%, p = 0.008), chest injury (58% vs. 36%, p < 0.001), and traumatic brain injury (72% vs. 48%, p < 0.001) compared with the no ARDS group. In multivariate analysis, head/chest Abbreviated Injury Score scores, crystalloid from 0 to 6 hours, and platelet transfusion from 0 to 6 hours and 7 to 24 hours were independent predictors of ARDS developing after 24 hours. CONCLUSIONS Blood and plasma transfusion were not independently associated with ARDS. However, platelet transfusion was a significant independent risk factor. The role of platelets warrants further investigation but may be mechanistically explained by lung injury models of pulmonary platelet sequestration with peripheral thrombocytopenia. LEVEL OF EVIDENCE Prognostic study, level IV.
Collapse
Affiliation(s)
- Lucy Z Kornblith
- Department of Surgery, Zuckerberg San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| | - Anamaria J Robles
- Department of Surgery, Zuckerberg San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| | - Amanda S Conroy
- Department of Surgery, Zuckerberg San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| | - Brittney J Redick
- Department of Surgery, Zuckerberg San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| | - Benjamin M Howard
- Department of Surgery, Zuckerberg San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| | - Carolyn M Hendrickson
- Department of Medicine, University of California, San Francisco; San Francisco, California
| | - Sara Moore
- Department of Biostatistics, University of California, Berkeley; Berkeley, California
| | - Mary F Nelson
- Department of Surgery, Zuckerberg San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| | - Farzad Moazed
- Department of Medicine, University of California, San Francisco; San Francisco, California
| | - Rachael A Callcut
- Department of Surgery, Zuckerberg San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| | - Carolyn S Calfee
- Department of Medicine, University of California, San Francisco; San Francisco, California
| | - Mitchell Jay Cohen
- Department of Surgery, Denver Health Medical Center and the University of Colorado; Denver, Colorado
| |
Collapse
|
2
|
Christie SA, Conroy AS, Callcut RA, Hubbard AE, Cohen MJ. Dynamic multi-outcome prediction after injury: Applying adaptive machine learning for precision medicine in trauma. PLoS One 2019; 14:e0213836. [PMID: 30970030 PMCID: PMC6457612 DOI: 10.1371/journal.pone.0213836] [Citation(s) in RCA: 26] [Impact Index Per Article: 5.2] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [MESH Headings] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 07/09/2018] [Accepted: 03/03/2019] [Indexed: 11/25/2022] Open
Abstract
OBJECTIVE Machine learning techniques have demonstrated superior discrimination compared to conventional statistical approaches in predicting trauma death. The objective of this study is to evaluate whether machine learning algorithms can be used to assess risk and dynamically identify patient-specific modifiable factors critical to patient trajectory for multiple key outcomes after severe injury. METHODS SuperLearner, an ensemble machine-learning algorithm, was applied to prospective observational cohort data from 1494 critically-injured patients. Over 1000 agnostic predictors were used to generate prediction models from multiple candidate learners for outcomes of interest at serial time points post-injury. Model accuracy was estimated using cross-validation and area under the curve was compared to select among predictors. Clinical variables responsible for driving outcomes were estimated at each time point. RESULTS SuperLearner fits demonstrated excellent cross-validated prediction of death (overall AUC 0.94-0.97), multi-organ failure (overall AUC 0.84-0.90), and transfusion (overall AUC 0.87-0.9) across multiple post-injury time points, and good prediction of Acute Respiratory Distress Syndrome (overall AUC 0.84-0.89) and venous thromboembolism (overall AUC 0.73-0.83). Outcomes with inferior data quality included coagulopathic trajectory (AUC 0.48-0.88). Key clinical predictors evolved over the post-injury timecourse and included both anticipated and unexpected variables. Non-random missingness of data was identified as a predictor of multiple outcomes over time. CONCLUSIONS Machine learning algorithms can be used to generate dynamic prediction after injury while avoiding the risk of over- and under-fitting inherent in ad hoc statistical approaches. SuperLearner prediction after injury demonstrates promise as an adaptable means of helping clinicians integrate voluminous, evolving data on severely-injured patients into real-time, dynamic decision-making support.
Collapse
Affiliation(s)
- S. Ariane Christie
- Department of Surgery, Zuckerberg San Francisco General Hospital and Trauma Center and the University of California, San Francisco; San Francisco, California, United States of America
| | - Amanda S. Conroy
- Department of Surgery, Zuckerberg San Francisco General Hospital and Trauma Center and the University of California, San Francisco; San Francisco, California, United States of America
| | - Rachael A. Callcut
- Department of Surgery, Zuckerberg San Francisco General Hospital and Trauma Center and the University of California, San Francisco; San Francisco, California, United States of America
| | - Alan E. Hubbard
- Department of Biostatistics, University of California, Berkeley School of Public Health; Berkeley, California, United States of America
| | - Mitchell J. Cohen
- Denver Health Medical Center and the University of Colorado; Denver, Colorado, United States of America
| |
Collapse
|
3
|
Robles AJ, Kornblith LZ, Hendrickson CM, Howard BM, Conroy AS, Moazed F, Calfee CS, Cohen MJ, Callcut RA. Health care utilization and the cost of posttraumatic acute respiratory distress syndrome care. J Trauma Acute Care Surg 2018; 85:148-154. [PMID: 29958249 PMCID: PMC6029709 DOI: 10.1097/ta.0000000000001926] [Citation(s) in RCA: 15] [Impact Index Per Article: 2.5] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/25/2022]
Abstract
BACKGROUND Posttraumatic acute respiratory distress syndrome (ARDS) is associated with prolonged mechanical ventilation and longer hospitalizations. The relationship between posttraumatic ARDS severity and financial burden has not been previously studied. We hypothesized that increasing ARDS severity is associated with incrementally higher health care costs. METHODS Adults arriving as the highest level of trauma activation were enrolled in an ongoing prospective cohort study. Patients who survived 6 hours or longer are included in the analysis. Blinded review of chest radiographs was performed by two independent physicians for any intubated patient with PaO2:FIO2 ratio of 300 mmHg or lower during the first 8 days of admission. The severity of ARDS was classified by the Berlin criteria. Hospital charge data were used to perform standard costing analysis. RESULTS Acute respiratory distress syndrome occurred in 13% (203 of 1,586). The distribution of disease severity was 33% mild, 42% moderate, and 25% severe. Patients with ARDS were older (41 years vs. 35 years, p < 0.01), had higher median Injury Severity Score (30 vs. 10, p < 0.01), more chest injury (Abbreviated Injury Scale score, ≥ 3: 51% vs. 21%, p < 0.01), and blunt mechanisms (85% vs. 53%, p < 0.01). By ARDS severity, there was no significant difference in age, mechanism, or rate of traumatic brain injury. Increasing ARDS severity was associated with higher Injury Severity Score and higher mortality rates. Standardized total hospital charges were fourfold higher for patients who developed ARDS compared with those who did not develop ARDS (US $434,000 vs. US $96,000; p < 0.01). Furthermore, the daily hospital charges significantly increased across categories of worsening ARDS severity (mild, US $20,451; moderate, US $23,994; severe, US $33,316; p < 0.01). CONCLUSION The development of posttraumatic ARDS is associated with higher health care costs. Among trauma patients who develop ARDS, total hospital charges per day increase with worsening severity of disease. Prevention, early recognition, and treatment of ARDS after trauma are potentially important objectives for efforts to control health care costs in this population. LEVEL OF EVIDENCE Economic and value-based evaluations, level IV.
Collapse
Affiliation(s)
- Anamaria J Robles
- Department of Surgery, San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| | - Lucy Z Kornblith
- Department of Surgery, San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| | - Carolyn M Hendrickson
- Department of Medicine, San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| | - Benjamin M Howard
- Department of Surgery, San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| | - Amanda S Conroy
- Department of Surgery, San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| | - Farzad Moazed
- Department of Medicine, San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| | - Carolyn S Calfee
- Department of Medicine, San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| | - Mitchell J Cohen
- Department of Surgery, Denver Health Medical Center and the University of Colorado; Denver, Colorado
| | - Rachael A Callcut
- Department of Surgery, San Francisco General Hospital and the University of California, San Francisco; San Francisco, California
| |
Collapse
|
4
|
Hendrickson CM, Gibb SL, Miyazawa BY, Keating SM, Ross E, Conroy AS, Calfee CS, Pati S, Cohen MJ. Elevated plasma levels of TIMP-3 are associated with a higher risk of acute respiratory distress syndrome and death following severe isolated traumatic brain injury. Trauma Surg Acute Care Open 2018; 3:e000171. [PMID: 30023434 PMCID: PMC6045722 DOI: 10.1136/tsaco-2018-000171] [Citation(s) in RCA: 5] [Impact Index Per Article: 0.8] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/07/2018] [Accepted: 05/10/2018] [Indexed: 01/15/2023] Open
Abstract
Background: Complications after injury, such as acute respiratory distress syndrome (ARDS), are common after traumatic brain injury (TBI) and associated with poor clinical outcomes. The mechanisms driving non-neurologic organ dysfunction after TBI are not well understood. Tissue inhibitor of matrix metalloproteinase-3 (TIMP-3) is a regulator of matrix metalloproteinase activity, inflammation, and vascular permeability, and hence has plausibility as a biomarker for the systemic response to TBI. Methods: In a retrospective study of 182 patients with severe isolated TBI, we measured TIMP-3 in plasma obtained on emergency department arrival. We used non-parametric tests and logistic regression analyses to test the association of TIMP-3 with the incidence of ARDS within 8 days of admission and in-hospital mortality. Results: TIMP-3 was significantly higher among subjects who developed ARDS compared with those who did not (median 2810 pg/mL vs. 2260 pg/mL, p=0.008), and significantly higher among subjects who died than among those who survived to discharge (median 2960 pg/mL vs. 2080 pg/mL, p<0.001). In an unadjusted logistic regression model, for each SD increase in plasma TIMP-3, the odds of ARDS increased significantly, OR 1.5 (95% CI 1.1 to 2.1). This association was only attenuated in multivariate models, OR 1.4 (95% CI 1.0 to 2.0). In an unadjusted logistic regression model, for each SD increase in plasma TIMP-3, the odds of death increased significantly, OR 1.7 (95% CI 1.2 to 2.3). The magnitude of this association was greater in a multivariate model adjusted for markers of injury severity, OR 1.9 (95% CI 1.2 to 2.8). Discussion: TIMP-3 may play an important role in the biology of the systemic response to brain injury in humans. Along with clinical and demographic data, early measurements of plasma biomarkers such as TIMP-3 may help identify patients at higher risk of ARDS and death after severe isolated TBI. Level of evidence III.
Collapse
Affiliation(s)
- Carolyn M Hendrickson
- Division of Pulmonary and Critical Care Medicine, Department of Medicine, University of California San Francisco, San Francisco, California, USA
| | - Stuart L Gibb
- Department of Laboratory Medicine, University of California San Francisco, San Francisco, California, USA.,Blood Systems Research Institute, San Francisco, California, USA
| | - Byron Y Miyazawa
- Department of Laboratory Medicine, University of California San Francisco, San Francisco, California, USA.,Blood Systems Research Institute, San Francisco, California, USA.,Department of Surgery, University of California San Francisco, Zuckerberg San Francisco General Hospital, San Francisco, California, USA
| | - Sheila M Keating
- Blood Systems Research Institute, San Francisco, California, USA
| | - Erin Ross
- Division of Pulmonary and Critical Care Medicine, Department of Medicine, University of California San Francisco, San Francisco, California, USA
| | - Amanda S Conroy
- Department of Surgery, University of California San Francisco, Zuckerberg San Francisco General Hospital, San Francisco, California, USA
| | - Carolyn S Calfee
- Division of Pulmonary and Critical Care Medicine, Department of Medicine, University of California San Francisco, San Francisco, California, USA
| | - Shibani Pati
- Department of Laboratory Medicine, University of California San Francisco, San Francisco, California, USA.,Blood Systems Research Institute, San Francisco, California, USA
| | - Mitchell J Cohen
- Department of Surgery, University of California San Francisco, Zuckerberg San Francisco General Hospital, San Francisco, California, USA.,Department of Surgery, University of Colorado, Denver, Colorado, USA
| |
Collapse
|
5
|
Howard BM, Kornblith LZ, Christie SA, Conroy AS, Nelson MF, Campion EM, Callcut RA, Calfee CS, Lamere BJ, Fadrosh DW, Lynch S, Cohen MJ. Characterizing the gut microbiome in trauma: significant changes in microbial diversity occur early after severe injury. Trauma Surg Acute Care Open 2017; 2:e000108. [PMID: 29766103 PMCID: PMC5877916 DOI: 10.1136/tsaco-2017-000108] [Citation(s) in RCA: 58] [Impact Index Per Article: 8.3] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/03/2017] [Revised: 06/15/2017] [Accepted: 06/26/2017] [Indexed: 01/25/2023] Open
Abstract
Background Recent studies have demonstrated the vital influence of commensal microbial communities on human health. The central role of the gut in the response to injury is well described; however, no prior studies have used culture-independent profiling techniques to characterize the gut microbiome after severe trauma. We hypothesized that in critically injured patients, the gut microbiome would undergo significant compositional changes in the first 72 hours after injury. Methods Trauma stool samples were prospectively collected via digital rectal examination at the time of presentation (0 hour). Patients admitted to the intensive care unit (n=12) had additional stool samples collected at 24 hours and/or 72 hours. Uninjured patients served as controls (n=10). DNA was extracted from stool samples and 16S rRNA-targeted PCR amplification was performed; amplicons were sequenced and binned into operational taxonomic units (OTUs; 97% sequence similarity). Diversity was analyzed using principle coordinates analyses, and negative binomial regression was used to determine significantly enriched OTUs. Results Critically injured patients had a median Injury Severity Score of 27 and suffered polytrauma. At baseline (0 hour), there were no detectable differences in gut microbial community diversity between injured and uninjured patients. Injured patients developed changes in gut microbiome composition within 72 hours, characterized by significant alterations in phylogenetic composition and taxon relative abundance. Members of the bacterial orders Bacteroidales, Fusobacteriales and Verrucomicrobiales were depleted during 72 hours, whereas Clostridiales and Enterococcus members enriched significantly. Discussion In this initial study of the gut microbiome after trauma, we demonstrate that significant changes in phylogenetic composition and relative abundance occur in the first 72 hours after injury. This rapid change in intestinal microbiota represents a critical phenomenon that may influence outcomes after severe trauma. A better understanding of the nature of these postinjury changes may lead to the ability to intervene in otherwise pathological clinical trajectories. Level of evidence III Study type Prognostic/epidemiological
Collapse
Affiliation(s)
- Benjamin M Howard
- Department of Surgery, San Francisco General Hospital, University of California San Francisco, California, USA
| | - Lucy Z Kornblith
- Department of Surgery, San Francisco General Hospital, University of California San Francisco, California, USA
| | - Sabrinah A Christie
- Department of Surgery, San Francisco General Hospital, University of California San Francisco, California, USA
| | - Amanda S Conroy
- Department of Surgery, San Francisco General Hospital, University of California San Francisco, California, USA
| | - Mary F Nelson
- Department of Surgery, San Francisco General Hospital, University of California San Francisco, California, USA
| | - Eric M Campion
- Department of Surgery, Denver Health and Hospital Authority, University of Colorado, Denver, Colorado, USA
| | - Rachael A Callcut
- Department of Surgery, San Francisco General Hospital, University of California San Francisco, California, USA
| | - Carolyn S Calfee
- Division of Pulmonary and Critical Care Medicine, Departments of Medicine and Anesthesia, University of California San Francisco, California, USA
| | - Brandon J Lamere
- Division of Gastroenterology, Department of Medicine, University of California San Francisco, California, USA
| | - Douglas W Fadrosh
- Division of Gastroenterology, Department of Medicine, University of California San Francisco, California, USA
| | - Susan Lynch
- Division of Gastroenterology, Department of Medicine, University of California San Francisco, California, USA
| | - Mitchell Jay Cohen
- Department of Surgery, Denver Health and Hospital Authority, University of Colorado, Denver, Colorado, USA
| |
Collapse
|
6
|
Kornblith LZ, Robles AJ, Conroy AS, Miyazawa BY, Callcut RA, Cohen MJ. Tired Platelet: Functional Anergy after Injury. J Am Coll Surg 2017. [DOI: 10.1016/j.jamcollsurg.2017.07.131] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [What about the content of this article? (0)] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/26/2022]
|
7
|
Callcut RA, Wakam G, Conroy AS, Kornblith LZ, Howard BM, Campion EM, Nelson MF, Mell MW, Cohen MJ. Discovering the truth about life after discharge: Long-term trauma-related mortality. J Trauma Acute Care Surg 2016; 80:210-7. [PMID: 26606176 PMCID: PMC4731245 DOI: 10.1097/ta.0000000000000930] [Citation(s) in RCA: 19] [Impact Index Per Article: 2.4] [Reference Citation Analysis] [What about the content of this article? (0)] [Affiliation(s)] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 12/19/2022]
Abstract
BACKGROUND Outcome after traumatic injury has typically been limited to the determination at time of discharge or brief follow-up. This study investigates the natural history of long-term survival after trauma. METHODS All highest-level activation patients prospectively enrolled in an ongoing cohort study from 2005 to 2012 were selected. To allow for long-term follow-up, patients had to be enrolled at least 1 year before the latest available data from the National Death Index (NDI, 2013). Time and cause of mortality was determined based on death certificates. Survival status was determined by the latest date of either care in our institution or NDI query. Kaplan-Meier curves were created stratified for Injury Severity Score (ISS). Survival was compared with estimated actuarial survival based on age, sex, and race. RESULTS A total of 908 highest-level activation patients (median ISS, 18) were followed up for a median 1.7 years (interquartile range 1.0-2.9; maximum, 9.8 years). Survival data were available on 99.8%. Overall survival was 73% (663 of 908). For those with at least 2-year follow-up, survival was only 62% (317 of 509). Severity of injury predicted long-term survival (p < 0.0001) with those having ISS of 25 or greater with the poorest outcome (57% survival at 5 years). For all ISS groups, survival was worse than predicted actuarial survival (p < 0.001). When excluding early deaths (≤30 days), observed survival was still significantly lower than estimated actuarial survival (p < 0.002). Eighteen percent (44 of 245 deaths) of all deaths occurred after 30 days. Among late deaths, 53% occurred between 31 days and 1 year after trauma. Trauma-related mortality was the leading cause of postdischarge death, accounting for 43% of the late deaths. CONCLUSION Postdischarge deaths represent a significant percentage of total trauma-related mortality. Despite having "survived" to leave the hospital, long-term survival was worse than predicted actuarial survival, suggesting that the mortality from injury does not end at "successful" hospital discharge. LEVEL OF EVIDENCE Prognostic study, level III.
Collapse
Affiliation(s)
- Rachael A. Callcut
- Department of Surgery, San Francisco General Hospital, University of California, San Francisco
| | - Glenn Wakam
- Department of Surgery, San Francisco General Hospital, University of California, San Francisco
| | - Amanda S. Conroy
- Department of Surgery, San Francisco General Hospital, University of California, San Francisco
| | - Lucy Z. Kornblith
- Department of Surgery, San Francisco General Hospital, University of California, San Francisco
| | - Benjamin M. Howard
- Department of Surgery, San Francisco General Hospital, University of California, San Francisco
| | - Eric M. Campion
- Department of Surgery, Denver Health Medical Center and the University of Colorado School of Medicine, Denver, CO
| | - Mary F. Nelson
- Department of Surgery, San Francisco General Hospital, University of California, San Francisco
| | - Matthew W. Mell
- Department of Surgery, Stanford University, Stanford, California
| | - Mitchell J. Cohen
- Department of Surgery, San Francisco General Hospital, University of California, San Francisco
| |
Collapse
|