Time Series Analysis to Assess Non Linear Lag Impact of Environmental Factors on Schizophrenia Admissions

Abstract

Recent studies have increasingly focused on understanding the impact of climatic factors on public health, particularly regarding the adverse effects of air quality on human well-being. This research investigates the complex relationship between environmental elements and psychiatric health, with a specific emphasis on schizophrenia. Schizophrenia is a psychiatric illness that significantly affects perception, cognition, and social identity. While existing statistical evidence supports the influence of atmospheric factors such as air pollutants and climate on schizophrenia, the non-linear dynamics of this association have yet to be explored using advanced time series and deep learning techniques. This study aims to clarify the significant impacts of atmospheric factors on schizophrenia in India. This study emphasizes the importance of considering cumulative effects arising from the interplay between air pollutants, temperature, and humidity. This study introduces an analysis of the association between atmospheric factors and schizophrenia in the specific context of India, with a particular focus on the city of Bangalore. We analyzed the connection between two different time series, i.e. daily environmental variables and hospital admission rates of Schizophrenia in Bangalore City. The time delay between pollution exposure and the manifestation of symptoms of schizophrenia is a key consideration. In our primary data analysis, we explored conventional time series algorithms such as ARIMA (SARIMAX) and VAR (VECM) and evaluated the significance of air quality exposure and schizophrenia response using Bangalore data. By utilizing the time series regression algorithms, we aim to elucidate the substantial impacts of various environmental factors on the rate of hospital admissions. ARIMA models have shown better performance than VAR in all efficiency measures. For instance, the R 2 value is higher for ARIMA in all lag models than VAR models, which shows better ARIMA performance.