Efficacy of time series gray model and exponential smoothing model in predictions of daily outpatient visits, numbers of SARS-CoV-2 nucleic acid and antigen positive among the visits during COVID-19 epidemic: an empirical analysis in Yunnan province

Objective To examine the efficacy of time series gray model and exponential smoothing mode in predictions of daily fever outpatient visits, number of severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) nucleic acid and antigen positive among the visits during conronavirus disease 2019 (COVID-19) epidemic for developing short-term prediction model of COVID-19 incidence.

Methods The data on 54 836 individuals visiting fever clinics during a COVID-19 epidemic period (December 12 – 31, 2022) were collected from medical facilities in a prefecture and two districts of Yunnan province and analyzed statistically. The series gray model GM(1,1) and exponential smoothing model were fitted to the data to construct models for predicting daily number of fever clinic visits, SARS-CoV-2 nucleic acid and antigen positivity in the study region. Average relative errors between the actual and the predicted values were used to evaluate the efficacy of the prediction models.

Results The posterior ratio C values of the constructed time series gray model GM(1,1) were 0.50, 0.40, and 0.43 and the average relative errors were103%, 72%, and 84% in predicting the daily number of fever clinic visits, SARS-CoV-2 nucleic acid positivity, and antigen positivity. In predicting the three numbers mentioned above, the R² values were 0.36, 0.56, and 0.53 and the average relative errors were 103%, 66% and 51%, respectively, for the constructed Holt linear trend non-seasonal exponential smoothing model.

Conclusion The efficacy of gray model is comparable to that of exponential smoothing model in predicting the number of fever outpatient visits but better than that of the latter model in predicting the number of SARS-CoV-2 nucleic acid and antigen positivity.