Background and aim: There is a need to determine which clinical variables predict the severity of COVID-19. We analyzed a series of critically ill COVID-19 patients to see if any of our dataset’s clinical variables were associated with patient outcomes. Methods: We retrospectively analyzed the data of COVID-19 patients admitted to the ICU of the Hospital in Pordenone from March 11, 2020, to April 17, 2020. Patients’ characteristics of survivors and deceased groups were compared. The variables with a different distribution between the two groups were implemented in a generalized linear regression model (LM) and in an Artificial Neural Network (NN) model to verify the “robustness” of the association with mortality. Results: In the considered period, we reviewed the data of 22 consecutive patients: 8 died. The causes of death were a severe respiratory failure (3), multi-organ failure (1), septic shock (1), pulmonary thromboembolism (2), severe hemorrhage (1). Lymphocyte and the platelet count were significantly lower in the group of deceased patients (p-value 0.043 and 0.020, respectively; cut-off values: 660/mm3; 280,000/mm3, respectively). Prothrombin time showed a statistically significant trend (p-value= 0.065; cut-off point: 16.8/sec). The LM model (AIC= 19.032), compared to the NN model (Mean Absolute Error, MAE = 0.02), was substantially alike (MSE 0.159 vs. 0.136). Conclusions: In the context of critically ill COVID-19 patients admitted to ICU, lymphocytopenia, thrombocytopenia, and lengthening of prothrombin time were strictly correlated with higher mortality. Additional clinical data are needed to be able to validate this prognostic score. (www.actabiomedica.it).

Artificial neural network model from a case series of covid-19 patients: A prognostic analysis

Orso D.;D'andrea N.;Vetrugno L.;Bove T.
2021

Abstract

Background and aim: There is a need to determine which clinical variables predict the severity of COVID-19. We analyzed a series of critically ill COVID-19 patients to see if any of our dataset’s clinical variables were associated with patient outcomes. Methods: We retrospectively analyzed the data of COVID-19 patients admitted to the ICU of the Hospital in Pordenone from March 11, 2020, to April 17, 2020. Patients’ characteristics of survivors and deceased groups were compared. The variables with a different distribution between the two groups were implemented in a generalized linear regression model (LM) and in an Artificial Neural Network (NN) model to verify the “robustness” of the association with mortality. Results: In the considered period, we reviewed the data of 22 consecutive patients: 8 died. The causes of death were a severe respiratory failure (3), multi-organ failure (1), septic shock (1), pulmonary thromboembolism (2), severe hemorrhage (1). Lymphocyte and the platelet count were significantly lower in the group of deceased patients (p-value 0.043 and 0.020, respectively; cut-off values: 660/mm3; 280,000/mm3, respectively). Prothrombin time showed a statistically significant trend (p-value= 0.065; cut-off point: 16.8/sec). The LM model (AIC= 19.032), compared to the NN model (Mean Absolute Error, MAE = 0.02), was substantially alike (MSE 0.159 vs. 0.136). Conclusions: In the context of critically ill COVID-19 patients admitted to ICU, lymphocytopenia, thrombocytopenia, and lengthening of prothrombin time were strictly correlated with higher mortality. Additional clinical data are needed to be able to validate this prognostic score. (www.actabiomedica.it).
File in questo prodotto:
File Dimensione Formato  
Artificial neural network model from a case series of covid-19 patients.pdf

non disponibili

Licenza: Non pubblico
Dimensione 229.37 kB
Formato Adobe PDF
229.37 kB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11390/1207174
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? ND
social impact