Modelling and forecasting Positive and Negative Syndrome Scale scores to achieve remission using time series analysis.

Objectives: Schizophrenia is a chronic mental condition. The objective of this study is to apply time series modelling to Positive and Negative Syndrome Scale scores of outpatients with schizophrenia, observed at regular intervals of time, and hence forecast the number of visits required to reach remission.

Methods: A retrospective data of outpatients diagnosed with chronic paranoid-type schizophrenia were extracted from the records of outpatient department of a tertiary hospital in New Delhi, India. Autoregressive integrated moving average (ARIMA) and ARIMAX models (ARIMA with explanatory variable as Clinical Global Impression Severity scale) are fitted to the data. The best fit models are employed to forecast the number of visits required to reach remission for the outpatients who did not achieve remission by the end of study. Prediction accuracy of the two models is compared using mean absolute percentage error and mean absolute deviation.

Results: The ARIMA (1, 2, 1) and ARIMAX (1, 2, 1) models are identified to be suitable models after a series of statistical tests.

Conclusions: ARIMA and ARIMAX models are suitable to predict number of visits required to reach remission. Further, ARIMAX model performed better than ARIMA model.