Time-Series Forecasting of Radiotherapy Utilization in Older Adults in Southern Thailand's Largest Quaternary Hospital: A Retrospective Study.

OBJECTIVE

This study aimed to forecast the radiotherapy demand among geriatric patients in Southern Thailand's largest quaternary hospital by 2030 using an Autoregressive Integrated Moving Average (ARIMA) model.

METHODS

This retrospective analysis was conducted using data from January 2004 and December 2022 and comprised patients aged ≥65 years who received radiotherapy. Monthly time-series data were analyzed in two phases. First, descriptive statistics were used to summarize patient demographics, cancer types, and treatment intent over time. Time-series decomposition and automatic machine learning were used to explore these patterns. Stationarity was assessed using the augmented Dickey-Fuller test. The model parameters were selected based on autocorrelation and partial autocorrelation plots and refined through optimization. Model selection was performed based on the Akaike Information Criterion, and forecasting accuracy was measured using the Mean Absolute Percentage Error (MAPE). Residual diagnostics included the Ljung-Box and Jarque-Bera tests as well as the assessment of heteroskedasticity.

RESULTS

Of the 39,653 patients who underwent radiotherapy, 10,717 (27%) were aged ≥65 years (mean age 71.8; 60% male). The most common cancers were head and neck, lung, colorectal, and breast. Most patients received curative treatment, with increasing trends in radiotherapy utilization, particularly for lung, colorectal, breast, and prostate cancers. The optimal model, ARIMA(3,1,0)(0,0,1,4), incorporating exogenous variables related to the older adult population in Southern Thailand, achieved a MAPE of 0.17 and successfully passed all residual diagnostics. By 2030, the model forecasted approximately 74.7 new monthly cases of geriatric radiotherapy, with a 95% confidence interval of 53.8-95.7.

CONCLUSION

The demand for radiotherapy among older adults is projected to increase, underscoring the need for capacity planning. Future studies should explore sophisticated prediction techniques and include more clinical variables to enhance the accuracy of forecasts and aid thorough oncology planning.