Abstract 10384: Validation of a Simple Score to Determine Risk of Hospital Mortality After the Norwood Procedure

Circulation ◽  
2015 ◽  
Vol 132 (suppl_3) ◽  
Author(s):  
Shahryar M Chowdhury ◽  
Eric M Graham ◽  
Andrew M Atz ◽  
Scott M Bradley ◽  
Minoo M Kavarana ◽  
...  
Keyword(s):  
Hospital Mortality ◽  
Risk Score ◽  
Validation Cohort ◽  
Univariate Analysis ◽  
Risk Groups ◽  
Norwood Procedure ◽  
Analysis Model ◽  
Multivariable Logistic Regression Model ◽  
Risk Of Mortality ◽  
Weighted Regression Analysis

Background: The NIH/NHLBI Pediatric Heart Network Single Ventricle Reconstruction (SVR) trial identified risk factors for hospital mortality after the Norwood procedure. However, the ability to quantify pre-operative risk remains elusive. This study aimed to develop an accurate and clinically feasible score to assess the risk of hospital mortality in neonates undergoing the Norwood procedure. Methods: All patients (n = 549) in the publically available SVR database were included in the analysis. Patients were randomly divided into a derivation (75%) and validation (25%) cohort. Pre-operative patient, center, and surgeon-related covariates found to be associated with mortality upon univariate analysis (p < 0.2) were included in the initial multivariable logistic regression model. The final model was derived by including only variables independently associated with mortality (p < 0.05). A risk score was then developed using relative magnitudes of the covariates’ odds ratio. The score was then tested in the validation cohort. Results: A 20-point risk score using 6 variables (see Table) was developed. The derivation and validation cohorts did not differ in age, sex, mortality, and the score covariates. Mean score in derivation and validation cohort were 5.2 ± 3.2 and 5.6 ± 3.5, p = 0.35, respectively. In weighted regression analysis, model predicted risk of mortality correlated closely with actual rates of mortality in the derivation (R 2 = 0.87, p < 0.01) and validation cohorts (R 2 = 0.82, p 10). Patients were classified as low (score 0-5), medium (6-10), or high risk (>10). Mortality differed significantly between risk groups in the derivation (6% vs. 22% vs. 77%, p < 0.01) and validation (4% vs. 30% vs. 53%, p < 0.01) cohorts. Conclusion: This mortality score is accurate in determining risk of hospital mortality in neonates undergoing the Norwood operation. The score has the potential to be used in clinical practice to aid in risk assessment prior to surgery.

A New Risk Score to Predict In-Hospital Mortality in Elderly Patients With Acute Heart Failure: On Behalf of the Journey HF-TR Study Investigators

Angiology ◽  
2020 ◽  
Vol 71 (10) ◽  
pp. 948-954
Author(s):  
Gülay Gök ◽  
Mehmet Karadağ ◽  
Ümit Yaşar Sinan ◽  
Mehdi Zoghi
Keyword(s):  
Heart Failure ◽  
Elderly Patients ◽  
Hospital Mortality ◽  
Risk Score ◽  
Validation Cohort ◽  
Characteristic Curve ◽  
Univariate Analysis ◽  
Previous History ◽  
Turkish Population ◽  
Patient Journey

We aimed to predict in-hospital mortality of elderly patients with heart failure (HF) by using a risk score model which could be easily applied in routine clinical practice without using an electronic calculator. The study population (n = 1034) recruited from the Journey HF-TR (Patient Journey in Hospital with Heart Failure in Turkish Population) study was divided into a derivation and a validation cohort. The parameters related to in-hospital mortality were first analyzed by univariate analysis, then the variables found to be significant in that analysis were entered into a stepwise multivariate logistic regression (LR) analysis. Patients were classified as low, intermediate, and high risk. A risk score obtained by taking into account the regression coefficients of the significant variables as a result of the LR analysis was tested in the validation cohort using receiver operating characteristic curve analysis. In total, 6 independent variables (age, blood urea nitrogen, previous history of hemodialysis/hemofiltration, inotropic agent use, and length of intensive care stay) associated with in-hospital mortality were included in the analysis. The risk score had a good discrimination in both the derivation and validation cohorts. A new validated risk score to determine the risk of in-hospital mortality of elderly hospitalized patients with HF was developed by including 6 independent predictors.

scholarly journals Development of a new sternal dehiscence prediction scale for decision making in sternal closure techniques after cardiac surgery

2021 ◽  
Vol 16 (1) ◽  
Author(s):  
Ehab Nooh ◽  
Colin Griesbach ◽  
Johannes Rösch ◽  
Michael Weyand ◽  
Frank Harig
Keyword(s):  
Logistic Regression ◽  
Risk Score ◽  
Characteristic Curve ◽  
Multivariable Analysis ◽  
Individual Risk ◽  
Dependent Diabetes Mellitus ◽  
Analysis Model ◽  
Multivariable Logistic Regression Model ◽  
Sternal Dehiscence ◽  
Multivariable Logistic Regression

Abstract Background After sternotomy, the spectrum for sternal osteosynthesis comprises standard wiring and more complex techniques, like titanium plating. The aim of this study is to develop a predictive risk score that evaluates the risk of sternum instability individually. The surgeon may then choose an appropriate sternal osteosynthesis technique that is risk- adjusted as well as cost-effective. Methods Data from 7.173 patients operated via sternotomy for all cardiovascular indications from 2008 until 2017 were retrospectively analyzed. Sternal dehiscence occurred in 2.5% of patients (n = 176). A multivariable analysis model examined pre- and intraoperative factors. A multivariable logistic regression model and a backward elimination based on the Akaike Information Criterion (AIC) a logistic model were selected. Results The model showed good sensitivity and specificity (area under the receiver-operating characteristic curve, AUC: 0.76) and several predictors of sternal instability could be evaluated. Multivariable logistic regression showed the highest Odds Ratios (OR) for reexploration (OR 6.6, confidence interval, CI [4.5–9.5], p < 0.001), obesity (body mass index, BMI > 35 kg/m2) (OR 4.23, [CI 2.4–7.3], p < 0.001), insulin-dependent diabetes mellitus (IDDM) (OR 2.2, CI [1.5–3.2], p = 0.01), smoking (OR 2.03, [CI 1.3–3.08], p = 0.001). After weighting the probability of sternum dehiscence with each factor, a risk score model was proposed scaling from − 1 to 5 points. This resulted in a risk score ranging up to 18 points, with an estimated risk for sternum complication up to 74%. Conclusions A weighted scoring system based on individual risk factors was specifically created to predict sternal dehiscence. High-scoring patients should receive additive closure techniques.

Blood mRNA signature to predict survival in patients with metastatic melanoma treated with tremelimumab.

2013 ◽  
Vol 31 (15_suppl) ◽  
pp. 9080-9080 ◽  
Author(s):  
Yvonne M. Saenger ◽  
Jay Magidson ◽  
Bobby Chi-Hung Liaw ◽  
Karl Wassmann ◽  
William Barker ◽  
...  
Keyword(s):  
Metastatic Melanoma ◽  
Risk Score ◽  
Peripheral Blood ◽  
Validation Cohort ◽  
Risk Groups ◽  
Phase Iii ◽  
Gene Model ◽  
Training Cohort ◽  
Melanoma Patients

9080 Background: Tremelimumab (Ticilimumumab, Pfizer), a monoclonal antibody targeting CTLA-4, a T cell inhibitory molecule, has shown activity in metastatic melanoma. Ipilimumab (Yervoy, BMS), another antibody targettingCTLA-4, improves survival relative to a peptide vaccine and is now FDA approved. A minority of patients will achieve durable tumor control with CTLA-4 blockade and biomarkers are urgently needed to identify those patients. Methods: 170 inflammatory, melanoma-specific and CTLA4-pathway related mRNA transcripts were measured using RT-PCR in pre-treatment peripheral blood samples from 218 patients with refractory melanoma receiving tremelimumab in a multi-center phase II study. A 2-class latent model yielded a risk score based on 4-genes that was highly predictive of survival (p<0.001), and was used to categorize patients into low, medium and high-risk groups. An independent cohort of 260 treatment naïve melanoma patients receiving tremelimumab as part of a multi-center phase III study was then used to validate the risk score as well as the 3 risk groups defined using the pre-specified cut-points. Results: There was no significant difference between the two cohorts in terms of age, gender, stage of disease or ECOG status. Median time of follow up was 297 days for the training cohort and 386 days for the validation cohort. 67% of patients in the training cohort and 70% of patients in the validation died during time of follow-up. Collectively, the ability of the 170 genes to predict survival exhibited a high degree of consistency across the cohorts (p < 0.001). A 4-gene model including cathepsin D (CTSD), Phopholipase A2 group VII (PLA2G7), Thioredoxin reductase 1 (TXNRD-1) and Interleukin 1 receptor associated kinase 3 (IRAK3) predicted survival in the validation cohort (p=0.001 by log rank test). Multivariable cox analysis showed that the 4-gene model added to the predictive value of clinical predictors (p<0.0001). Conclusions: Expression levels of CTSD, PLA2G7, TXNRD1, and IRAK3 in peripheral blood are predictive of survival in melanoma patients treated with ticilimumab (αCTLA-4). Blood mRNA signatures should be further explored to define patient subsets likely to benefit from immunotherapy.

scholarly journals Prognosis Score System to Predict Survival for COVID-19 Cases: a Korean Nationwide Cohort Study (Preprint)

2020 ◽  
Author(s):  
Sung-Yeon Cho ◽  
Sung-Soo Park ◽  
Min-Kyu Song ◽  
Young Yi Bae ◽  
Dong-Gun Lee ◽  
...  
Keyword(s):  
Overall Survival ◽  
Survival Rate ◽  
High Risk ◽  
Hospital Stay ◽  
Risk Score ◽  
Validation Cohort ◽  
Length Of Hospital Stay ◽  
Risk Groups ◽  
Prognosis Score ◽  
Very High

BACKGROUND As the COVID-19 pandemic continues, an initial risk-adapted allocation is crucial for managing medical resources and providing intensive care. OBJECTIVE In this study, we aimed to identify factors that predict the overall survival rate for COVID-19 cases and develop a COVID-19 prognosis score (COPS) system based on these factors. In addition, disease severity and the length of hospital stay for patients with COVID-19 were analyzed. METHODS We retrospectively analyzed a nationwide cohort of laboratory-confirmed COVID-19 cases between January and April 2020 in Korea. The cohort was split randomly into a development cohort and a validation cohort with a 2:1 ratio. In the development cohort (n=3729), we tried to identify factors associated with overall survival and develop a scoring system to predict the overall survival rate by using parameters identified by the Cox proportional hazard regression model with bootstrapping methods. In the validation cohort (n=1865), we evaluated the prediction accuracy using the area under the receiver operating characteristic curve. The score of each variable in the COPS system was rounded off following the log-scaled conversion of the adjusted hazard ratio. RESULTS Among the 5594 patients included in this analysis, 234 (4.2%) died after receiving a COVID-19 diagnosis. In the development cohort, six parameters were significantly related to poor overall survival: older age, dementia, chronic renal failure, dyspnea, mental disturbance, and absolute lymphocyte count &lt;1000/mm<sup>3</sup>. The following risk groups were formed: low-risk (score 0-2), intermediate-risk (score 3), high-risk (score 4), and very high-risk (score 5-7) groups. The COPS system yielded an area under the curve value of 0.918 for predicting the 14-day survival rate and 0.896 for predicting the 28-day survival rate in the validation cohort. Using the COPS system, 28-day survival rates were discriminatively estimated at 99.8%, 95.4%, 82.3%, and 55.1% in the low-risk, intermediate-risk, high-risk, and very high-risk groups, respectively, of the total cohort (<i>P</i>&lt;.001). The length of hospital stay and disease severity were directly associated with overall survival (<i>P</i>&lt;.001), and the hospital stay duration was significantly longer among survivors (mean 26.1, SD 10.7 days) than among nonsurvivors (mean 15.6, SD 13.3 days). CONCLUSIONS The newly developed predictive COPS system may assist in making risk-adapted decisions for the allocation of medical resources, including intensive care, during the COVID-19 pandemic.

scholarly journals In-Hospital Mortality of COVID-19 Patients Treated with High-Flow Nasal Oxygen: Evaluation of Biomarkers and Development of the Novel Risk Score Model CROW-65

Life ◽  
2021 ◽  
Vol 11 (8) ◽  
pp. 735
Author(s):  
Toni Kljakovic Gaspic ◽  
Mirela Pavicic Ivelja ◽  
Marko Kumric ◽  
Andrija Matetic ◽  
Nikola Delic ◽  
...  
Keyword(s):  
Mechanical Ventilation ◽  
Hospital Mortality ◽  
Risk Score ◽  
Univariate Analysis ◽  
High Flow ◽  
The Novel ◽  
Single Center ◽  
Therapeutic Possibility ◽  
Male Patients ◽  
Score Model

To replace mechanical ventilation (MV), which represents the cornerstone therapy in severe COVID-19 cases, high-flow nasal oxygen (HFNO) therapy has recently emerged as a less-invasive therapeutic possibility for those patients. Respecting the risk of MV delay as a result of HFNO use, we aimed to evaluate which parameters could determine the risk of in-hospital mortality in HFNO-treated COVID-19 patients. This single-center cohort study included 102 COVID-19-positive patients treated with HFNO. Standard therapeutic methods and up-to-date protocols were used. Patients who underwent a fatal event (41.2%) were significantly older, mostly male patients, and had higher comorbidity burdens measured by CCI. In a univariate analysis, older age, shorter HFNO duration, ventilator initiation, higher CCI and lower ROX index all emerged as significant predictors of adverse events (p < 0.05). Variables were dichotomized and included in the multivariate analysis to define their relative weights in the computed risk score model. Based on this, a risk score model for the prediction of in-hospital mortality in COVID-19 patients treated with HFNO consisting of four variables was defined: CCI > 4, ROX index ≤ 4.11, LDH-to-WBC ratio, age > 65 years (CROW-65). The main purpose of CROW-65 is to address whether HFNO should be initiated in the subgroup of patients with a high risk of in-hospital mortality.

scholarly journals Development of a New Sternal Dehiscence Prediction Scale for Decision Making in Sternal Closure Techniques After Cardiac Surgery

2020 ◽  
Author(s):  
Frank Harig ◽  
Ehab Nooh ◽  
Colin Griesbach ◽  
Michael Weyand ◽  
Johannes Rösch
Keyword(s):  
Logistic Regression ◽  
Risk Score ◽  
Characteristic Curve ◽  
Multivariable Analysis ◽  
Individual Risk ◽  
Dependent Diabetes Mellitus ◽  
Analysis Model ◽  
Multivariable Logistic Regression Model ◽  
Sternal Dehiscence ◽  
Multivariable Logistic Regression

Abstract BackgroundAfter sternotomy, the spectrum for sternal osteosynthesis comprises standard wiring and more complex techniques, like titanium plating. The aim of this study is to develop a predictive risk score that evaluates the risk of sternum instability individually. The surgeon may then choose an appropriate sternal osteosynthesis technique that is risk- adjusted as well as cost-effective.MethodsData from 7.173 patients operated via sternotomy for all cardiovascular indications from 2008 until 2017 were retrospectively analyzed. Sternal dehiscence occurred in 2.5% of patients (n=176). A multivariable analysis model examined pre- and intraoperative factors. A multivariable logistic regression model and a backward elimination based on the Akaike Information Criterion (AIC) a logistic model were selected.ResultsThe model showed good sensitivity and specificity (area under the receiver-operating characteristic curve, AUC: 0.76) and several predictors of sternal instability could be evaluated. Multivariable logistic regression showed the highest Odds Ratios (OR) for reexploration (OR 6.6, confidence interval, CI [4.5-9.5], p <0.001), obesity (body mass index, BMI >35kg/m²) (OR 4.23, [CI 2.4-7.3], p<0.001), insulin-dependent diabetes mellitus (IDDM) (OR 2.2, CI [1.5-3.2], p=0.01), smoking (OR 2.03, [CI 1.3-3.08], p=0.001). After weighting the probability of sternum dehiscence with each factor, a risk score model was proposed scaling from -1 to 5 points. This resulted in a risk score ranging up to 18 points, with an estimated risk for sternum complication up to 74%.ConclusionsA weighted scoring system based on individual risk factors was specifically created to predict sternal dehiscence. High-scoring patients should receive additive closure techniques.

scholarly journals The NIV Outcomes (NIVO) Score: prediction of in-hospital mortality in exacerbations of COPD requiring assisted ventilation

2021 ◽  
pp. 2004042
Author(s):  
Tom Hartley ◽  
Nicholas D. Lane ◽  
John Steer ◽  
Mark W. Elliott ◽  
Milind P. Sovani ◽  
...  
Keyword(s):  
Hospital Mortality ◽  
Validation Cohort ◽  
A Priori ◽  
External Validation ◽  
Risk Groups ◽  
Assisted Ventilation ◽  
Hospital Death ◽  
Prediction Tool ◽  
Trial Methodology ◽  
Derivation Cohort

IntroductionAcute exacerbations of COPD (AECOPD) complicated by acute (acidaemic) hypercapnic respiratory failure (AHRF) requiring ventilation are common. When applied appropriately, ventilation substantially reduces mortality. Despite this, there is evidence of poor practice and prognostic pessimism. A clinical prediction tool could improve decision making regarding ventilation, but none is routinely used.MethodsConsecutive patients admitted with AECOPD and AHRF treated with assisted ventilation (principally non-invasive ventilation) were identified in two hospitals serving differing populations. Known and potential prognostic indices were identified a priori. A prediction tool for in-hospital death was derived using multivariable regression analysis. Prospective, external validation was performed in a temporally separate, geographically diverse 10-centre study. The trial methodology adhered to TRIPOD recommendations.ResultsDerivation cohort, n=489, in-hospital mortality 25.4%; validation cohort, n=733, in-hospital mortality 20.1%. Using 6 simple categorised variables; extended Medical Research Council Dyspnoea score (eMRCD)1–4/5a/5b, time from admission to acidaemia >12 h, pH<7.25, presence of atrial fibrillation, Glasgow coma scale ≤14 and chest radiograph consolidation a simple scoring system with strong prediction of in-hospital mortality is achieved. The resultant NIVO score had area under the receiver operated curve of 0.79 and offers good calibration and discrimination across stratified risk groups in its validation cohort.DiscussionThe NIVO score outperformed pre-specified comparator scores. It is validated in a generalisable cohort and works despite the heterogeneity inherent to both this patient group and this intervention. Potential applications include informing discussions with patients and their families, aiding treatment escalation decisions, challenging pessimism, and comparing risk-adjusted outcomes across centres.

scholarly journals COVID-19-related outcomes in immunocompromised patients: A nationwide study in Korea

PLoS ONE ◽  
2021 ◽  
Vol 16 (10) ◽  
pp. e0257641
Author(s):  
Moon Seong Baek ◽  
Min-Taek Lee ◽  
Won-Young Kim ◽  
Jae Chol Choi ◽  
Sun-Young Jung
Keyword(s):  
Mortality Rate ◽  
Hospital Mortality ◽  
Multivariable Analysis ◽  
Immunocompromised Patients ◽  
Multivariable Logistic Regression Model ◽  
Inverse Probability ◽  
Risk Of Mortality ◽  
Vulnerable Patients ◽  
Increased Risk ◽  
Immunocompromised Status

Background Given the rapid increased in confirmed coronavirus disease 2019 (COVID-19) and related mortality, it is important to identify vulnerable patients. Immunocompromised status is considered a risk factor for developing severe COVID-19. We aimed to determine whether immunocompromised patients with COVID-19 have an increased risk of mortality. Method The groups’ baseline characteristics were balanced using a propensity score-based inverse probability of treatment weighting approach. Odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated for the risks of in-hospital mortality and other outcomes according to immunocompromised status using a multivariable logistic regression model. We identified immunocompromised status based on a diagnosis of malignancy or HIV/AIDS, having undergone organ transplantation within 3 years, prescriptions for corticosteroids or oral immunosuppressants for ≥30 days, and at least one prescription for non-oral immunosuppressants during the last year. Results The 6,435 COVID-19 patients (≥18 years) included 871 immunocompromised (13.5%) and 5,564 non-immunocompromised (86.5%). Immunocompromised COVID-19 patients were older (60.1±16.4 years vs. 47.1±18.7 years, absolute standardized mean difference: 0.738). The immunocompromised group had more comorbidities, a higher Charlson comorbidity index, and a higher in-hospital mortality rate (9.6% vs. 2.3%; p < .001). The immunocompromised group still had a significantly higher in-hospital mortality rate after inverse probability of treatment weighting (6.4% vs. 2.0%, p < .001). Multivariable analysis adjusted for baseline imbalances revealed that immunocompromised status was independently associated with a higher risk of mortality among COVID-19 patients (adjusted odds ratio [aOR]: 2.09, 95% CI: 1.62–2.68, p < .001). Conclusions Immunocompromised status among COVID-19 patients was associated with a significantly increased risk of mortality.

P1763Derivation and validation of a risk score to predict incomplete ST segment resolution in STEMI patients undergoing primary PCI: association with the benefit of Glycoprotein IIb-IIIa inhibitors Use

2019 ◽  
Vol 40 (Supplement_1) ◽  
Author(s):  
S Cornara ◽  
A Somaschini ◽  
M Ferlini ◽  
A Demarchi ◽  
A Mandurino Mirizzi ◽  
...  
Keyword(s):  
Risk Score ◽  
Survival Benefit ◽  
Validation Cohort ◽  
Univariate Analysis ◽  
University Hospital ◽  
Discrimination Power ◽  
Killip Class ◽  
Derivation Cohort ◽  
C Statistic ◽  
St Segment

Abstract Incomplete ST segment resolution (iSTR) is a well-known marker of poor outcome in patients undergoing primary percutaneous coronary intervention (pPCI) for ST elevation myocardial infarction (STEMI). The use of glycoprotein IIbIIIa inhibitors (GPIs) was suggested to be associated with a survival benefit in high-risk patients. A simple score to predict the risk for developing iSTR could help early identification of these patients and could allow a tailored use of pharmacological tools, such as GPIs. The aim of this study was to create and validate a numerical score to predict iSTR occurrence in STEMI patients undergoing pPCI and to assess its association with the potential benefit of GPIs use. We prospectively enrolled all STEMI patients undergoing pPCI in our University Hospital (2005–2017). iSTR was defined as a <70% resolution of initial ST segment shift in the lead with maximal ST deviation 60 min after reperfusion. Our population was randomly divided in two group: a derivation cohort (60%) and a validation cohort (40%). Potential predictors of iSTR were selected at univariate analysis and were then inserted in a multivariate binary stepwise-backward logistic regression. To create a risk score, numerical values were obtained considering the odds ratio of each independent predictor rounding to the nearest unit or half. A ROC curve with its c-statistic was then used to test the discrimination power of the score both in the derivation and in the validation cohort. Out of a total of 2959 patients, 1774 were included in the derivation: 480 (27%) of them presented iSTR. All-cause mortality at 30 days was significantly higher in patients with iSTR (OR 3.2, 95% CI 2.1–4.9, p<0.001). Anterior MI (OR 2.46, 95% CI 1.90–3.14, p<0.001, score 2.5), anemia at admission (OR 1.76, 95% CI 1.29–2.4, p<0.001, score 2), blood glucose >198 mg/dl at admission (OR 1.77, 95% CI 1.29–2.49, p<0.001, score 2), age >75 years (OR 1.54, 95% CI 1.15–2.10, p=0.004, score 1.5), female sex (OR 1.41, 95% CI 1.06–1.88, p=0.02, score 1.5) and Killip class >2 (OR 1.44, 95% CI 1.05–1.98, p=0.024, score 1.5) were identified as independent predictors of iSTR, creating a ISTR-score that ranged from 0 to 11. The validation cohort consisted in 1185 patients, with 31% showing iSTR. The c-statistic was 0.67 and 0.66 in the derivation and validation cohorts. Patients with score ≥4 versus <4 showed present a worst prognosys but a similar GPI use. Notably, GPIs were associated with a significant survival benefit among patients≥4 but not among patients <4 (Figure). The use of GPIs was not associated to any clinically relevant difference, the increase in bleeding risk appeared similar. A simple pre-procedural risk score may predict iSTR following pPPCI, allowing a rapid risk stratification and the identification of patients who show a favorable risk/benefit ratio for the use of more aggressive strategies such as GPIs. These findings deserve a prospective, randomized evaluation.

scholarly journals An Individualized Algorithm to Predict Mortality in COVID-19 Pneumonia: a Machine Learning Based Study

2022 ◽  
Author(s):  
Maria Elena Laino ◽  
Elena Generali ◽  
Tobia Tommasini ◽  
Giovanni Angelotti ◽  
Alessio Aghemo ◽  
...  
Keyword(s):  
Machine Learning ◽  
Hospital Mortality ◽  
Risk Score ◽  
Machine Learning Techniques ◽  
Clinical Settings ◽  
Risk Of Mortality ◽  
Learning Techniques ◽  
Machine Learning Approach ◽  
Importance Analysis ◽  
High Level

IntroductionIdentifying SARS-CoV-2 patients at higher risk of mortality is crucial in the management of a pandemic. Artificial intelligence techniques allow to analyze big amount of data to find hidden patterns. We aimed to develop and validate a mortality score at admission for COVID-19 based on high-level machine learning.Material and methodsWe conducted a retrospective cohort study on hospitalized adults COVID-19 patients between March and December 2020. The primary outcome was in-hospital mortality. A machine learning approach on vital parameters, laboratory values, and demographic features was applied to develop different models. Then, a feature importance analysis was performed to reduce the number of variables included in the model, to develop a risk score with good overall performance, that was finally evaluated in terms of discrimination and calibration capabilities. All results underwent cross-validation.Results1,135 consecutive patients (median age 70 years, 64% males) were enrolled, 48 patients were excluded, the cohort was randomly divided in training (760) and test (327). During hospitalization, 251 (22%) patients died. After feature selection, the best performing classifier was random forest (AUC 0.88±0.03). Based on the relative importance of each variable, a pragmatic score was developed, showing good performances (AUC 0.85, ±0.025), and three levels were defined that correlated well with in-hospital mortality.ConclusionsMachine learning techniques were applied in order to develop an accurate in-hospital mortality risk score for COVID-19 based on ten variables. The application of the proposed score has utility in clinical settings to guide the management and prognostication of COVID-19 patients.

Sign in / Sign up

Export Citation Format

Share Document