1
|
Bashiri FS, Caskey JR, Mayampurath A, Dussault N, Dumanian J, Bhavani SV, Carey KA, Gilbert ER, Winslow CJ, Shah NS, Edelson DP, Afshar M, Churpek MM. Identifying infected patients using semi-supervised and transfer learning. J Am Med Inform Assoc 2022; 29:1696-1704. [PMID: 35869954 PMCID: PMC9471712 DOI: 10.1093/jamia/ocac109] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.5] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/28/2022] [Revised: 06/13/2022] [Accepted: 07/01/2022] [Indexed: 11/12/2022] Open
Abstract
Abstract
Objectives
Early identification of infection improves outcomes, but developing models for early identification requires determining infection status with manual chart review, limiting sample size. Therefore, we aimed to compare semi-supervised and transfer learning algorithms with algorithms based solely on manual chart review for identifying infection in hospitalized patients.
Materials and Methods
This multicenter retrospective study of admissions to 6 hospitals included “gold-standard” labels of infection from manual chart review and “silver-standard” labels from nonchart-reviewed patients using the Sepsis-3 infection criteria based on antibiotic and culture orders. “Gold-standard” labeled admissions were randomly allocated to training (70%) and testing (30%) datasets. Using patient characteristics, vital signs, and laboratory data from the first 24 hours of admission, we derived deep learning and non-deep learning models using transfer learning and semi-supervised methods. Performance was compared in the gold-standard test set using discrimination and calibration metrics.
Results
The study comprised 432 965 admissions, of which 2724 underwent chart review. In the test set, deep learning and non-deep learning approaches had similar discrimination (area under the receiver operating characteristic curve of 0.82). Semi-supervised and transfer learning approaches did not improve discrimination over models fit using only silver- or gold-standard data. Transfer learning had the best calibration (unreliability index P value: .997, Brier score: 0.173), followed by self-learning gradient boosted machine (P value: .67, Brier score: 0.170).
Discussion
Deep learning and non-deep learning models performed similarly for identifying infection, as did models developed using Sepsis-3 and manual chart review labels.
Conclusion
In a multicenter study of almost 3000 chart-reviewed patients, semi-supervised and transfer learning models showed similar performance for model discrimination as baseline XGBoost, while transfer learning improved calibration.
Collapse
Affiliation(s)
- Fereshteh S Bashiri
- Department of Medicine, University of Wisconsin-Madison , Madison, Wisconsin, USA
| | - John R Caskey
- Department of Medicine, University of Wisconsin-Madison , Madison, Wisconsin, USA
| | - Anoop Mayampurath
- Department of Biostatistics and Medical Informatics, University of Wisconsin-Madison , Madison, Wisconsin, USA
| | - Nicole Dussault
- Pritzker School of Medicine, University of Chicago , Chicago, Illinois, USA
| | - Jay Dumanian
- Pritzker School of Medicine, University of Chicago , Chicago, Illinois, USA
| | | | - Kyle A Carey
- Department of Medicine, University of Chicago , Chicago, Illinois, USA
| | - Emily R Gilbert
- Department of Medicine, Loyola University , Chicago, Illinois, USA
| | - Christopher J Winslow
- Department of Medicine, NorthShore University HealthSystem , Evanston, Illinois, USA
| | - Nirav S Shah
- Department of Medicine, University of Chicago , Chicago, Illinois, USA
- Department of Medicine, NorthShore University HealthSystem , Evanston, Illinois, USA
| | - Dana P Edelson
- Department of Medicine, University of Chicago , Chicago, Illinois, USA
| | - Majid Afshar
- Department of Medicine, University of Wisconsin-Madison , Madison, Wisconsin, USA
- Department of Biostatistics and Medical Informatics, University of Wisconsin-Madison , Madison, Wisconsin, USA
| | - Matthew M Churpek
- Department of Medicine, University of Wisconsin-Madison , Madison, Wisconsin, USA
- Department of Biostatistics and Medical Informatics, University of Wisconsin-Madison , Madison, Wisconsin, USA
| |
Collapse
|
2
|
Jarczak D, Kluge S, Nierhaus A. Sepsis-Pathophysiology and Therapeutic Concepts. Front Med (Lausanne) 2021; 8:628302. [PMID: 34055825 PMCID: PMC8160230 DOI: 10.3389/fmed.2021.628302] [Citation(s) in RCA: 118] [Impact Index Per Article: 39.3] [Reference Citation Analysis] [Abstract] [Key Words] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 11/11/2020] [Accepted: 04/09/2021] [Indexed: 12/12/2022] Open
Abstract
Sepsis is a life-threatening condition and a global disease burden. Today, the heterogeneous syndrome is defined as severe organ dysfunction caused by a dysregulated host response to infection, with renewed emphasis on immune pathophysiology. Despite all efforts of experimental and clinical research during the last three decades, the ability to positively influence course and outcome of the syndrome remains limited. Evidence-based therapy still consists of basic causal and supportive measures, while adjuvant interventions such as blood purification or targeted immunotherapy largely remain without proof of effectiveness so far. With this review, we aim to provide an overview of sepsis immune pathophysiology, to update the choice of therapeutic approaches targeting different immunological mechanisms in the course of sepsis and septic shock, and to call for a paradigm shift from the pathogen to the host response as a potentially more promising angle.
Collapse
Affiliation(s)
- Dominik Jarczak
- Department of Intensive Care Medicine, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
| | - Stefan Kluge
- Department of Intensive Care Medicine, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
| | - Axel Nierhaus
- Department of Intensive Care Medicine, University Medical Center Hamburg-Eppendorf, Hamburg, Germany
| |
Collapse
|
3
|
Nierhaus A, Berlot G, Kindgen-Milles D, Müller E, Girardis M. Best-practice IgM- and IgA-enriched immunoglobulin use in patients with sepsis. Ann Intensive Care 2020; 10:132. [PMID: 33026597 PMCID: PMC7538847 DOI: 10.1186/s13613-020-00740-1] [Citation(s) in RCA: 29] [Impact Index Per Article: 7.3] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/21/2020] [Accepted: 09/06/2020] [Indexed: 12/20/2022] Open
Abstract
Background Sepsis is a life-threatening organ dysfunction caused by a dysregulated host response to infection. Despite treatment being in line with current guidelines, mortality remains high in those with septic shock. Intravenous immunoglobulins represent a promising therapy to modulate both the pro- and anti-inflammatory processes and can contribute to the elimination of pathogens. In this context, there is evidence of the benefits of immunoglobulin M (IgM)- and immunoglobulin A (IgA)-enriched immunoglobulin therapy for sepsis. This manuscript aims to summarize current relevant data to provide expert opinions on best practice for the use of an IgM- and IgA-enriched immunoglobulin (Pentaglobin) in adult patients with sepsis. Main text Sepsis patients with hyperinflammation and patients with immunosuppression may benefit most from treatment with IgM- and IgA-enriched immunoglobulin (Pentaglobin). Patients with hyperinflammation present with phenotypes that manifest throughout the body, whilst the clinical characteristics of immunosuppression are less clear. Potential biomarkers for hyperinflammation include elevated procalcitonin, interleukin-6, endotoxin activity and C-reactive protein, although thresholds for these are not well-defined. Convenient biomarkers for identifying patients in a stage of immune-paralysis are still matter of debate, though human leukocyte antigen–antigen D related expression on monocytes, lymphocyte count and viral reactivation have been proposed. The timing of treatment is potentially more critical for treatment efficacy in patients with hyperinflammation compared with patients who are in an immunosuppressed stage. Due to the lack of evidence, definitive dosage recommendations for either population cannot be made, though we suggest that patients with hyperinflammation should receive an initial bolus at a rate of up to 0.6 mL (30 mg)/kg/h for 6 h followed by a continuous maintenance rate of 0.2 mL (10 mg)/kg/hour for ≥ 72 h (total dose ≥ 0.9 g/kg). For immunosuppressed patients, dosage is more conservative (0.2 mL [10 mg]/kg/h) for ≥ 72 h, without an initial bolus (total dose ≥ 0.72 g/kg). Conclusions Two distinct populations that may benefit most from Pentaglobin therapy are described in this review. However, further clinical evidence is required to strengthen support for the recommendations given here regarding timing, duration and dosage of treatment.
Collapse
Affiliation(s)
- Axel Nierhaus
- University Medical Center Hamburg, Hamburg, Germany. .,Dep. of Intensive Care Medicine, University Medical Center Hamburg-Eppendorf, Martinistr. 52, 20246, Hamburg, Germany.
| | | | | | | | | |
Collapse
|
4
|
Jarczak D, Kluge S, Nierhaus A. Use of Intravenous Immunoglobulins in Sepsis Therapy-A Clinical View. Int J Mol Sci 2020; 21:E5543. [PMID: 32756325 PMCID: PMC7432410 DOI: 10.3390/ijms21155543] [Citation(s) in RCA: 14] [Impact Index Per Article: 3.5] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 07/03/2020] [Revised: 07/29/2020] [Accepted: 07/30/2020] [Indexed: 12/29/2022] Open
Abstract
Sepsis is a life-threatening organ dysfunction, defined by a dysregulated host immune response to infection. During sepsis, the finely tuned system of immunity, inflammation and anti-inflammation is disturbed in a variety of ways. Both pro-inflammatory and anti-inflammatory pathways are upregulated, activation of the coagulation cascade and complement and sepsis-induced lymphopenia occur. Due to the manifold interactions in this network, the use of IgM-enriched intravenous immunoglobulins seems to be a promising therapeutic approach. Unfortunately, there is still a lack of evidence-based data to answer the important questions of appropriate patient populations, optimal timing and dosage of intravenous immunoglobulins. With this review, we aim to provide an overview of the role of immunoglobulins, with emphasis on IgM-enriched formulations, in the therapy of adult patients with sepsis and septic shock.
Collapse
Affiliation(s)
| | | | - Axel Nierhaus
- Department of Intensive Care Medicine, University Medical Center Hamburg-Eppendorf, 20246 Hamburg, Germany; (D.J.); (S.K.)
| |
Collapse
|
5
|
C-C Chemokine Receptor Type 2 Expression on Monocytes Before Sepsis Onset Is Higher Than That of Postsepsis in Septic Burned Patients: A New Predictor for Sepsis in Burned Injury. Ann Surg 2017; 264:392-8. [PMID: 26727083 DOI: 10.1097/sla.0000000000001531] [Citation(s) in RCA: 6] [Impact Index Per Article: 0.9] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 12/20/2022]
Abstract
OBJECTIVE The present study aims to investigate the alterations in monocytes (Mo) and dendritic cells (DCs) in septic burned patients with a special focus on C-C chemokine receptor type 2 (CCR2) expressions on classical Mo. BACKGROUND The phenotypes of Mo and DCs, particularly CCR2 expression on Mo, are not fully explored in severely burned patients with sepsis. METHODS The prospective cohort study was conducted in Ross Tilley Burn Centre and Sunnybrook Research Institute (Toronto, Canada). We enrolled 8 healthy patients and 89 burned patients with various burned sizes, of those burned patients, 12 were with sepsis. Blood was collected upon admission to the hospital and throughout their course in hospital. The expression of human leukocyte antigen-DR was determined on all DCs and Mo, along with CCR2 on CD14/CD16 Mo. RESULTS We found a profound decrease in human leukocyte antigen-DR on Mo and DCs in burned patients with sepsis compared with healthy controls and nonseptic burned patients. In addition, septic burned patients presented an increased CCR2 expression on classical Mo (CD14/CD16), which was paralleled by greater chemokine (C-C motif) ligand 2 concentrations in the plasma when compared with controls and nonseptic burned patients. Furthermore, burned patients with sepsis had a more profound expansion of CD14/CD16 Mo when compared with nonseptic burned patients. CONCLUSION Our results demonstrate that burned patients with sepsis have more profound impairment of monocytes and dendritic cells than burned patients without sepsis. With CCR2 level on Mo before sepsis onset being higher than postsepsis, CCR2 expression could be a new predictor of sepsis onset in severe burn injury.
Collapse
|
6
|
Bermejo-Martin JF, Andaluz-Ojeda D, Almansa R, Gandía F, Gómez-Herreras JI, Gomez-Sanchez E, Heredia-Rodríguez M, Eiros JM, Kelvin DJ, Tamayo E. Defining immunological dysfunction in sepsis: A requisite tool for precision medicine. J Infect 2016; 72:525-36. [PMID: 26850357 DOI: 10.1016/j.jinf.2016.01.010] [Citation(s) in RCA: 58] [Impact Index Per Article: 7.3] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 12/15/2015] [Revised: 01/24/2016] [Accepted: 01/26/2016] [Indexed: 02/05/2023]
Abstract
OBJECTIVES Immunological dysregulation is now recognised as a major pathogenic event in sepsis. Stimulation of immune response and immuno-modulation are emerging approaches for the treatment of this disease. Defining the underlying immunological alterations in sepsis is important for the design of future therapies with immuno-modulatory drugs. METHODS Clinical studies evaluating the immunological response in adult patients with Sepsis and published in PubMed were reviewed to identify features of immunological dysfunction. For this study we used key words related with innate and adaptive immunity. RESULTS Ten major features of immunological dysfunction (FID) were identified involving quantitative and qualitative alterations of [antigen presentation](FID1), [T and B lymphocytes] (FID2), [natural killer cells] (FID3), [relative increase in T regulatory cells] (FID4), [increased expression of PD-1 and PD-ligand1](FID5), [low levels of immunoglobulins](FID6), [low circulating counts of neutrophils and/or increased immature forms in non survivors](FID7), [hyper-cytokinemia] (FID8), [complement consumption] (FID9), [defective bacterial killing by neutrophil extracellular traps](FID10). CONCLUSIONS This review article identified ten major features associated with immunosuppression and immunological dysregulation in sepsis. Assessment of these features could help in utilizing precision medicine for the treatment of sepsis with immuno-modulatory drugs.
Collapse
Affiliation(s)
- Jesús F Bermejo-Martin
- Infection and Immunity Medical Investigation Unit (IMI), Hospital Clínico Universitario de Valladolid, SACYL/IECSCYL, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain; Grupo de Investigación Biomédica en Cuidados Críticos (BioCritic), Hospital Clínico Universitario de Valladolid, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain.
| | - David Andaluz-Ojeda
- Grupo de Investigación Biomédica en Cuidados Críticos (BioCritic), Hospital Clínico Universitario de Valladolid, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain; Servicio de Medicina Intensiva, Hospital Clínico Universitario de Valladolid, SACYL, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain.
| | - Raquel Almansa
- Infection and Immunity Medical Investigation Unit (IMI), Hospital Clínico Universitario de Valladolid, SACYL/IECSCYL, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain; Grupo de Investigación Biomédica en Cuidados Críticos (BioCritic), Hospital Clínico Universitario de Valladolid, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain.
| | - Francisco Gandía
- Grupo de Investigación Biomédica en Cuidados Críticos (BioCritic), Hospital Clínico Universitario de Valladolid, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain; Servicio de Medicina Intensiva, Hospital Clínico Universitario de Valladolid, SACYL, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain.
| | - Jose Ignacio Gómez-Herreras
- Grupo de Investigación Biomédica en Cuidados Críticos (BioCritic), Hospital Clínico Universitario de Valladolid, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain; Servicio de Anestesiología y Reanimación, Hospital Clínico Universitario de Valladolid, SACYL, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain.
| | - Esther Gomez-Sanchez
- Grupo de Investigación Biomédica en Cuidados Críticos (BioCritic), Hospital Clínico Universitario de Valladolid, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain; Servicio de Anestesiología y Reanimación, Hospital Clínico Universitario de Valladolid, SACYL, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain.
| | - María Heredia-Rodríguez
- Grupo de Investigación Biomédica en Cuidados Críticos (BioCritic), Hospital Clínico Universitario de Valladolid, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain; Servicio de Anestesiología y Reanimación, Hospital Clínico Universitario de Valladolid, SACYL, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain.
| | - Jose Maria Eiros
- Grupo de Investigación Biomédica en Cuidados Críticos (BioCritic), Hospital Clínico Universitario de Valladolid, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain.
| | - David J Kelvin
- Division of Experimental Therapeutics, Toronto General Hospital Research Institute, University Health Network, 200 Elizabeth Street, Toronto, ON M5G 2C4, Canada; Sezione di Microbiologia Sperimentale e Clinica, Dipartimento di Scienze Biomediche, Universita' degli Studi di Sassari, Piazza Università, 21, 07100 Sassari SS, Italy; International Institute of Infection and Immunity, Shantou University Medical College, 22 Xinling Road, Shantou, 515041 Guangdong Province, PR China.
| | - Eduardo Tamayo
- Grupo de Investigación Biomédica en Cuidados Críticos (BioCritic), Hospital Clínico Universitario de Valladolid, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain; Servicio de Anestesiología y Reanimación, Hospital Clínico Universitario de Valladolid, SACYL, Avenida Ramón y Cajal, 3, 47005 Valladolid, Spain.
| |
Collapse
|