BACKGROUND: In the present study we aimed to create a model able to predict the short-term need of hospital beds for COVID-19 patients, during SARS-CoV-2 outbreak. METHODS: We retrospectively revised data about all COVID-19 patients hospitalized at a University Hospital in Northern Italy, between March 1 and April 29, 2020. Several polynomial models (from first to fourth order) were fitted to estimate the relationship between the time and the number of occupied hospital beds during the entire period and after the local peak of the outbreak and to provide the prediction of short-term hospital beds demand. Model selection was based on the adjusted R2 (aR2) Index and likelihood ratio test (LRT). RESULTS: We included 836 hospitalizations (800 COVID-19 patients). The median length of hospital in-stay was 12 days. According to the aR2, the fourth order models best fitted the data considering the entire time period. When only the data after the peak was selected, no statistical improvement was found adding terms of order 3 and 4 and lower order polynomial models were considered for the forecasting of the hospital beds demand. Both approaches had a decreasing trend in the number of occupied beds along with time; however, the quadratic one showed a faster reduction in the predicted number of beds required by patients affected by COVID-19. CONCLUSIONS: We propose a model to predict the hospital bed requirement during the descending phase of COVID-19 outbreak, the validation of which might contribute to decision makers policy in the next weeks of pandemic.

Modelling hospital bed necessity for COVID-19 patients during the decline phase of the epidemic trajectory

AIROLDI C.;SCOTTI L.;BELLAN M.
;
SAINAGHI P. P.;PIRISI M.
2021-01-01

Abstract

BACKGROUND: In the present study we aimed to create a model able to predict the short-term need of hospital beds for COVID-19 patients, during SARS-CoV-2 outbreak. METHODS: We retrospectively revised data about all COVID-19 patients hospitalized at a University Hospital in Northern Italy, between March 1 and April 29, 2020. Several polynomial models (from first to fourth order) were fitted to estimate the relationship between the time and the number of occupied hospital beds during the entire period and after the local peak of the outbreak and to provide the prediction of short-term hospital beds demand. Model selection was based on the adjusted R2 (aR2) Index and likelihood ratio test (LRT). RESULTS: We included 836 hospitalizations (800 COVID-19 patients). The median length of hospital in-stay was 12 days. According to the aR2, the fourth order models best fitted the data considering the entire time period. When only the data after the peak was selected, no statistical improvement was found adding terms of order 3 and 4 and lower order polynomial models were considered for the forecasting of the hospital beds demand. Both approaches had a decreasing trend in the number of occupied beds along with time; however, the quadratic one showed a faster reduction in the predicted number of beds required by patients affected by COVID-19. CONCLUSIONS: We propose a model to predict the hospital bed requirement during the descending phase of COVID-19 outbreak, the validation of which might contribute to decision makers policy in the next weeks of pandemic.
File in questo prodotto:
File Dimensione Formato  
Minerva Respir Med-1925_Figures3_V1_2021-02-10.pdf

Open Access dal 25/05/2022

Licenza: Copyright dell'editore
Dimensione 439.63 kB
Formato Adobe PDF
439.63 kB Adobe PDF Visualizza/Apri
Minerva Respir Med-1925_Figures2_V1_2021-02-10.pdf

Open Access dal 25/05/2022

Licenza: Copyright dell'editore
Dimensione 463.49 kB
Formato Adobe PDF
463.49 kB Adobe PDF Visualizza/Apri
Minerva Respir Med-1925_Figures1_V1_2021-02-10.pdf

Open Access dal 25/05/2022

Licenza: Copyright dell'editore
Dimensione 450.44 kB
Formato Adobe PDF
450.44 kB Adobe PDF Visualizza/Apri
Minerva Respir Med-1925_Manuscript_V1_2021-02-10.doc

Open Access dal 25/05/2022

Licenza: Copyright dell'editore
Dimensione 113.5 kB
Formato Microsoft Word
113.5 kB Microsoft Word Visualizza/Apri

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: https://hdl.handle.net/11579/127966
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 5
  • ???jsp.display-item.citation.isi??? 5
social impact