1
|
Luo Y, Wang Z, Wang C. Improvement of APACHE II score system for disease severity based on XGBoost algorithm. BMC Med Inform Decis Mak 2021; 21:237. [PMID: 34362354 PMCID: PMC8344327 DOI: 10.1186/s12911-021-01591-x] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.7] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 12/02/2020] [Accepted: 07/21/2021] [Indexed: 11/30/2022] Open
Abstract
BACKGROUND Prognostication is an essential tool for risk adjustment and decision making in the intensive care units (ICUs). In order to improve patient outcomes, we have been trying to develop a more effective model than Acute Physiology and Chronic Health Evaluation (APACHE) II to measure the severity of the patients in ICUs. The aim of the present study was to provide a mortality prediction model for ICUs patients, and to assess its performance relative to prediction based on the APACHE II scoring system. METHODS We used the Medical Information Mart for Intensive Care version III (MIMIC-III) database to build our model. After comparing the APACHE II with 6 typical machine learning (ML) methods, the best performing model was screened for external validation on anther independent dataset. Performance measures were calculated using cross-validation to avoid making biased assessments. The primary outcome was hospital mortality. Finally, we used TreeSHAP algorithm to explain the variable relationships in the extreme gradient boosting algorithm (XGBoost) model. RESULTS We picked out 14 variables with 24,777 cases to form our basic data set. When the variables were the same as those contained in the APACHE II, the accuracy of XGBoost (accuracy: 0.858) was higher than that of APACHE II (accuracy: 0.742) and other algorithms. In addition, it exhibited better calibration properties than other methods, the result in the area under the ROC curve (AUC: 0.76). we then expand the variable set by adding five new variables to improve the performance of our model. The accuracy, precision, recall, F1, and AUC of the XGBoost model increased, and were still higher than other models (0.866, 0.853, 0.870, 0.845, and 0.81, respectively). On the external validation dataset, the AUC was 0.79 and calibration properties were good. CONCLUSIONS As compared to conventional severity scores APACHE II, our XGBoost proposal offers improved performance for predicting hospital mortality in ICUs patients. Furthermore, the TreeSHAP can help to enhance the understanding of our model by providing detailed insights into the impact of different features on the disease risk. In sum, our model could help clinicians determine prognosis and improve patient outcomes.
Collapse
Affiliation(s)
- Yan Luo
- Present Address: School of Computer Science (National Pilot Software Engineering School)
, Beijing University of Posts and Telecommunications, Beijing, 100876 China
- Key Laboratory of Trustworthy Distributed Computing and Service (BUPT), Ministry of Education, Beijing, 100876 China
| | - Zhiyu Wang
- Present Address: School of Computer Science (National Pilot Software Engineering School)
, Beijing University of Posts and Telecommunications, Beijing, 100876 China
- Key Laboratory of Trustworthy Distributed Computing and Service (BUPT), Ministry of Education, Beijing, 100876 China
| | - Cong Wang
- Present Address: School of Computer Science (National Pilot Software Engineering School)
, Beijing University of Posts and Telecommunications, Beijing, 100876 China
- Key Laboratory of Trustworthy Distributed Computing and Service (BUPT), Ministry of Education, Beijing, 100876 China
| |
Collapse
|
2
|
Mnatzaganian G, Hiller JE, Fletcher J, Putland M, Knott C, Braitberg G, Begg S, Bish M. Socioeconomic gradients in admission to coronary or intensive care units among Australians presenting with non-traumatic chest pain in emergency departments. BMC Emerg Med 2018; 18:32. [PMID: 30268098 PMCID: PMC6162924 DOI: 10.1186/s12873-018-0185-2] [Citation(s) in RCA: 4] [Impact Index Per Article: 0.7] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 12/18/2017] [Accepted: 09/21/2018] [Indexed: 11/10/2022] Open
Abstract
BACKGROUND Socioeconomic inequalities in cardiovascular morbidity have been previously reported showing direct associations between socioeconomic disadvantage and worse health outcomes. However, disagreement remains regarding the strength of the direct associations. The main objective of this panel design was to inspect socioeconomic gradients in admission to a coronary care unit (CCU) or an intensive care unit (ICU) among adult patients presenting with non-traumatic chest pain in three acute-care public hospitals in Victoria, Australia, during 2009-2013. METHODS Consecutive adults aged 18 or over presenting with chest pain in three emergency departments (ED) in Victoria, Australia during the five-year study period were eligible to participate. A relative index of inequality of socioeconomic status (SES) was estimated based on residential postcode socioeconomic index for areas (SEIFA) disadvantage scores. Admission to specialised care units over repeated presentations was modelled using a multivariable Generalized Estimating Equations approach that accounted for various socio-demographic and clinical variables. RESULTS Non-traumatic chest pain accounted for 10% of all presentations in the emergency departments (ED). A total of 53,177 individuals presented during the study period, with 22.5% presenting more than once. Of all patients, 17,579 (33.1%) were hospitalised over time, of whom 8584 (48.8%) were treated in a specialised care unit. Female sex was independently associated with fewer admissions to CCU / ICU, whereas, a dose-response effect of socioeconomic disadvantage and admission to CCU / ICU was found, with risk of admission increasing incrementally as SES declined. Patients coming from the lowest SES locations were 27% more likely to be admitted to these units compared with those coming from the least disadvantaged locations, p < 0.001. Men were significantly more likely to be admitted to such units than similarly affected and aged women among those diagnosed with angina pectoris, arrhythmia, myocardial infarction, heart failure, chest pain, and general signs and symptoms. CONCLUSIONS This study is the first to report socioeconomic gradients in admission to CCU / ICU in patients presenting with chest pain showing a dose-response effect. Our findings suggest increased cardiovascular morbidity as socioeconomic disadvantage increases.
Collapse
Affiliation(s)
- George Mnatzaganian
- La Trobe Rural Health School, College of Science, Health and Engineering, La Trobe University, PO Box 199, Bendigo, VIC, 3552, Australia.
| | - Janet E Hiller
- School of Health Sciences, Faculty of Health, Arts and Design, Swinburne University of Technology, John Street, Hawthorn, VIC, Australia.,School of Public Health, The University of Adelaide, North Terrace, Adelaide, SA, Australia
| | - Jason Fletcher
- Intensive Care Unit, Bendigo Health, Barnard Street, Bendigo, VIC, Australia
| | - Mark Putland
- Department of Emergency Medicine, Royal Melbourne Hospital, Parkville, VIC, Australia
| | - Cameron Knott
- Intensive Care Unit, Bendigo Health, Barnard Street, Bendigo, VIC, Australia.,Monash Rural Health Bendigo, Monash University, Bendigo, VIC, Australia.,Department of Intensive Care, Austin Health, Heidelberg, VIC, Australia
| | - George Braitberg
- Department of Emergency Medicine, Royal Melbourne Hospital, Parkville, VIC, Australia.,Centre for Integrated Critical Care Medicine, Department of Medicine and Radiology, The University of Melbourne, Parkville, VIC, Australia
| | - Steve Begg
- La Trobe Rural Health School, College of Science, Health and Engineering, La Trobe University, PO Box 199, Bendigo, VIC, 3552, Australia
| | - Melanie Bish
- La Trobe Rural Health School, College of Science, Health and Engineering, La Trobe University, PO Box 199, Bendigo, VIC, 3552, Australia
| |
Collapse
|
3
|
Mnatzaganian G, Braitberg G, Hiller JE, Kuhn L, Chapman R. Sex differences in in-hospital mortality following a first acute myocardial infarction: symptomatology, delayed presentation, and hospital setting. BMC Cardiovasc Disord 2016; 16:109. [PMID: 27389522 PMCID: PMC4937590 DOI: 10.1186/s12872-016-0276-5] [Citation(s) in RCA: 24] [Impact Index Per Article: 3.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 12/14/2015] [Accepted: 05/13/2016] [Indexed: 11/10/2022] Open
Abstract
BACKGROUND Women generally wait longer than men prior to seeking treatment for acute myocardial infarction (AMI). They are more likely to present with atypical symptoms, and are less likely to be admitted to coronary or intensive care units (CCU or ICU) compared to similarly-aged males. Women are more likely to die during hospital admission. Sex differences in the associations of delayed arrival, admitting ward, and mortality have not been thoroughly investigated. METHODS Focusing on presenting symptoms and time of presentation since symptom onset, we evaluated sex differences in in-hospital mortality following a first AMI in 4859 men and women presenting to three emergency departments (ED) from December 2008 to February 2014. Sex-specific risk of mortality associated with admission to either CCU/ICU or medical wards was calculated after adjusting for age, socioeconomic status, triage-assigned urgency of presentation, blood pressure, heart rate, presenting symptoms, timing of presentation since symptom onset, and treatment in the ED. Sex-specific age-adjusted attributable risks were calculated. RESULTS Compared to males, females waited longer before seeking treatment, presented more often with atypical symptoms, and were less likely to be admitted to CCU or ICU. Age-adjusted mortality in CCU/ICU or medical wards was higher among females (3.1 and 4.9 % respectively in CCU/ICU and medical wards in females compared to 2.6 and 3.2 % in males). However, after adjusting for variation in presenting symptoms, delayed arrival and other risk factors, risk of death was similar between males and females if they were admitted to CCU or ICU. This was in contrast to those admitted to medical wards. Females admitted to medical wards were 89 % more likely to die than their male counterparts. Arriving in the ED within 60 min of onset of symptoms was not associated with in-hospital mortality. Among males, 2.2 % of in-hospital mortality was attributed to being admitted to medical wards rather than CCU or ICU, while for females this age-adjusted attributable risk was 4.1 %. CONCLUSIONS Our study stresses the need to reappraise decision making in patient selection for admission to specialised care units, whilst raising awareness of possible sex-related bias in management of patients diagnosed with an AMI.
Collapse
Affiliation(s)
- George Mnatzaganian
- School of Allied Health, Faculty of Health Sciences, Australian Catholic University, Fitzroy, Victoria, 3065, Australia.
| | - George Braitberg
- Department of Medicine, The University of Melbourne, Parkville, Victoria, 3010, Australia.,Department of Emergency Medicine, Royal Melbourne Hospital, Parkville, Victoria, 3010, Australia
| | - Janet E Hiller
- School of Health Sciences, Faculty of Health, Arts and Design, Swinburne University of Technology, Hawthorn, Victoria, 3122, Australia.,Discipline of Public Health, School of Population Health, The University of Adelaide, Adelaide, South Australia, 5000, Australia
| | - Lisa Kuhn
- School of Nursing and Midwifery, Faculty of Health, Deakin University, Geelong, Victoria, 3220, Australia
| | - Rose Chapman
- School of Physiotherapy and Exercise Science, Faculty of Health Sciences, Curtin University, Bentley, Western Australia, 6102, Australia
| |
Collapse
|
4
|
Di Martino V, Weil D, Cervoni JP, Thevenot T. New prognostic markers in liver cirrhosis. World J Hepatol 2015; 7:1244-50. [PMID: 26019739 PMCID: PMC4438498 DOI: 10.4254/wjh.v7.i9.1244] [Citation(s) in RCA: 18] [Impact Index Per Article: 2.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Received: 12/01/2014] [Revised: 02/02/2015] [Accepted: 02/10/2015] [Indexed: 02/06/2023] Open
Abstract
Determining the prognosis of cirrhotic patients is not an easy task. Prognostic scores, like Child-Pugh and Model of End-stage Liver Disease scores, are commonly used by hepatologists, but do not always reflect superimposed events that may strongly influence the prognosis. Among them, bacterial intestinal translocation is a key phenomenon for the development of cirrhosis-related complications. Several biological variables (C-reactive protein, serum free cortisol, copeptin, von Willebrand factor antigen) are surrogates of "inflammatory stress" and have recently been identified as potential prognostic markers in cirrhotic patients. Most of these above mentioned markers were investigated in pilot studies with sometimes a modest sample size but allow us to catch a glimpse of the pathophysiological mechanisms leading to the worsening of cirrhosis. These new data should generate further well-designed studies to better assess the benefit for liver function of preventing intestinal bacterial translocation and microvascular thrombosis. The control of infection is vital and among all actors of immunity, vitamin D also appears to act as an anti-infective agent and therefore has probably a prognostic value.
Collapse
Affiliation(s)
- Vincent Di Martino
- Vincent Di Martino, Delphine Weil, Jean-Paul Cervoni, Thierry Thevenot, Service D'hépatologie Et De Soins Intensifs Digestifs, CHU Jean Minjoz, 25000 Besançon, France
| | - Delphine Weil
- Vincent Di Martino, Delphine Weil, Jean-Paul Cervoni, Thierry Thevenot, Service D'hépatologie Et De Soins Intensifs Digestifs, CHU Jean Minjoz, 25000 Besançon, France
| | - Jean-Paul Cervoni
- Vincent Di Martino, Delphine Weil, Jean-Paul Cervoni, Thierry Thevenot, Service D'hépatologie Et De Soins Intensifs Digestifs, CHU Jean Minjoz, 25000 Besançon, France
| | - Thierry Thevenot
- Vincent Di Martino, Delphine Weil, Jean-Paul Cervoni, Thierry Thevenot, Service D'hépatologie Et De Soins Intensifs Digestifs, CHU Jean Minjoz, 25000 Besançon, France
| |
Collapse
|
5
|
Hui C, Lin MC, Jao MS, Liu TC, Wu RG. Previous antibiotic exposure and evolution of antibiotic resistance in mechanically ventilated patients with nosocomial infections. J Crit Care 2013; 28:728-34. [DOI: 10.1016/j.jcrc.2013.04.008] [Citation(s) in RCA: 14] [Impact Index Per Article: 1.3] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/09/2013] [Revised: 04/07/2013] [Accepted: 04/13/2013] [Indexed: 11/28/2022]
|
6
|
Lengliné E, Raffoux E, Lemiale V, Darmon M, Canet E, Boissel N, Schlemmer B, Dombret H, Azoulay E. Intensive care unit management of patients with newly diagnosed acute myeloid leukemia with no organ failure. Leuk Lymphoma 2012; 53:1352-9. [PMID: 22233111 DOI: 10.3109/10428194.2011.649752] [Citation(s) in RCA: 81] [Impact Index Per Article: 6.8] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 12/13/2022]
Abstract
Patients with acute myeloid leukemia (AML) may present with early complications from sepsis or leukemic infiltration. Benefits from early in-intensive care unit (ICU) hematological management was evaluated in 42 adults with newly diagnosed AML with hematological risk of early death (age 46 years, French-American-British [FAB] M4/5 58%, leukocytes 103 × 10(9)/L) first admitted to the ICU without immediate life support (early-ICU). Controls were 42 patients primarily admitted to hematology wards, matched for age, leukocytes and FAB subtype. Twenty (47.6%) control patients were subsequently admitted to the ICU (late-ICU). Late-ICU patients presented with increased respiratory and cardiac rates, decreased oxygen saturation (SpO(2)) and blood pressure, at hospital admission. Late-ICU admission resulted in increased use of mechanical ventilation (60% vs. 33%) and vasopressors (60% vs. 16%), longer ICU stay (9 [6-25] vs. 5 [2-9] days) and decreased ICU survival (65% vs. 79%). Direct admission to the ICU of patients with high-risk AML with physiological disturbances but no organ dysfunction is associated with improved outcomes.
Collapse
Affiliation(s)
- Etienne Lengliné
- Medical ICU, Hôpital Saint-Louis, AP-HP and UFR de Médecine, University Paris-7 Paris-Diderot, Paris, France
| | | | | | | | | | | | | | | | | |
Collapse
|
7
|
Impact of delayed admission to intensive care units on mortality of critically ill patients: a cohort study. CRITICAL CARE : THE OFFICIAL JOURNAL OF THE CRITICAL CARE FORUM 2011; 15:R28. [PMID: 21244671 PMCID: PMC3222064 DOI: 10.1186/cc9975] [Citation(s) in RCA: 261] [Impact Index Per Article: 20.1] [Reference Citation Analysis] [Abstract] [Track Full Text] [Download PDF] [Figures] [Subscribe] [Scholar Register] [Received: 08/26/2010] [Revised: 11/11/2010] [Accepted: 01/18/2011] [Indexed: 01/20/2023]
Abstract
Introduction When the number of patients who require intensive care is greater than the number of beds available, intensive care unit (ICU) entry flow is obstructed. This phenomenon has been associated with higher mortality rates in patients that are not admitted despite their need, and in patients that are admitted but are waiting for a bed. The purpose of this study is to evaluate if a delay in ICU admission affects mortality for critically ill patients. Methods A prospective cohort of adult patients admitted to the ICU of our institution between January and December 2005 were analyzed. Patients for whom a bed was available were immediately admitted; when no bed was available, patients waited for ICU admission. ICU admission was classified as either delayed or immediate. Confounding variables examined were: age, sex, originating hospital ward, ICU diagnosis, co-morbidity, Acute Physiology and Chronic Health Evaluation (APACHE) II score, therapeutic intervention, and Sequential Organ Failure Assessment (SOFA) score. All patients were followed until hospital discharge. Results A total of 401 patients were evaluated; 125 (31.2%) patients were immediately admitted and 276 (68.8%) patients had delayed admission. There was a significant increase in ICU mortality rates with a delay in ICU admission (P = 0.002). The fraction of mortality risk attributable to ICU delay was 30% (95% confidence interval (CI): 11.2% to 44.8%). Each hour of waiting was independently associated with a 1.5% increased risk of ICU death (hazard ratio (HR): 1.015; 95% CI 1.006 to 1.023; P = 0.001). Conclusions There is a significant association between time to admission and survival rates. Early admission to the ICU is more likely to produce positive outcomes.
Collapse
|
8
|
Diminution de l’expression monocytaire de HLA-DR et risque d’infection hospitalière. ACTA ACUST UNITED AC 2010; 29:368-76. [DOI: 10.1016/j.annfar.2010.02.015] [Citation(s) in RCA: 12] [Impact Index Per Article: 0.9] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/16/2009] [Accepted: 02/05/2010] [Indexed: 01/25/2023]
|
9
|
Mandel M. The competing risks illness–death model under cross-sectional sampling. Biostatistics 2009; 11:290-303. [DOI: 10.1093/biostatistics/kxp048] [Citation(s) in RCA: 11] [Impact Index Per Article: 0.7] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/14/2022] Open
|