Hospitalizations due to spontaneous intracerebral hemorrhage in the region of Nis (Serbia): 11-year time-series analysis

Abstract

Background: The study of seasonal variability of intracerebral hemorrhage (ICH) occurrence may contribute to a better understanding of the nature of this disease and open up new perspectives in its prevention. The aim of this study was to test seasonal patterns in the number of admissions of ICH patients and determine which months have maximal and minimal number of admissions.

Methods: The main data source for this study was a hospital-based registry at the Clinic of Neurology in Nis, Serbia. During the studied period (1997-2007) a total of 1569 ICH patients were registered. Time series, consisting of the monthly number of hospitalized patients, for the 128 months of the study duration, has been successfully modeled using the multiplicative Auto Regressive Integrated Moving Average (ARIMA) model.

Results: Using the maximum likelihood method, utilizing Melrad's algorithm, the parameters of this ARIMA model have been calculated: constant (estimate 12.068, p<0.001), auto regressive-AR(1) (estimate 0.866, p<0.001), moving average-MA(1) (estimate 0.775, p<0.001), seasonal moving average-SMA(12) (estimate -0.198, p=0.036). ARIMA modeling has been successful and showed that there is a clear seasonal pattern in the data analyzed.

Conclusion: Based on the seasonal multiplicative ARIMA model and the seasonal time series decomposition, we showed that, in the period covered by the study, the peak of admissions occurred in March, and the trough of admissions was found in August.