Aggregating the response in time series lag models using quad weight kernel: impact of environmental factors on schizophrenia admissions

There has been a growing apprehension in the past few years concerning the issue of pollution and climate change. Several articles have shown the impact of air pollutants and atmosphere factors like temperature and relative humidity on health. This encompasses a deterioration in cognitive function and a heightened susceptibility to neurological diseases like schizophrenia. This work constructed a new dataset for hospital admissions of schizophrenia patients and daily environmental values from various locations in Bangalore City, India. The number of admissions to hospitals is used as a proxy for the incidence of schizophrenia emergence. In this kind of time series data where a response has a delayed impact on dependent variables, distributed lag models (DLM) are applicable. The newly created data is used to analyze the impact of pollution and climate factors on hospital admissions in Bengaluru, India. We have compared the applicability of the aggregated response technique using the distributed lag non-linear model (ARDLNM) in comparison with existing techniques, the distributed lag non-linear model and the non-linear autoregressive distributed lag model. ARDLNM using the Epanechnikov kernel showed improved performance over DLNM by 25%, 15%, 17% and 72% for performance metrics MSE, MAE, MAPE and R² respectively. Within aggregation methods, we have introduced a quad-weight kernel that further improved performance over the Epanechnikov kernel by 17% in terms of MSE.