However, ITS also can suffer from common time series problems, such as non-linear trends, autocorrelation and seasonality, which make the interpretation of intervention effects harder. The application of ARIMA model in our paper can better address these issues than the commonly used segmented regression models, by allowing more flexibility in modelling the impacts from other unrelated trends in the time series data.
We found the total number of outpatient visits stayed at the same level, but with a shift of modality from in-person to virtual visits. On the other hand, visits with psychopharmaceutical prescribing significantly declined after the pandemic. Further analysis stratify those visits by patient demographics (i.e. race, sex, age) did not reveal any uneven decline. However, when stratifying by the type of mental health diagnosis, significant decline was seen in visits related to ADHD and mood disorders.
We learned how COVID-19 impacts our health service utilization at Kaiser, and we can quantify the change, both immediately and long-term, to make intelligent business decision and to improve the quality of care. For example, we can add more availability for mental health virtual visits to meet the greater demands, and further investigate the prescribing pattern in the context of a virtual setting to locate the cause of greater decline for ADHD and mood disorders visits.
Read more about the study: "COVID-19 and adolescent outpatient mental health service utilization: an interrupted time series analysis"
Click here if you want to try out my analysis code: ITS python code.