Hybrid ARIMA-Prophet-LSTM with SHAP Interpretation for Forecasting Outpatient Visits by Medical Specialty

Vlandimir Welem L; Sani Muhamad  Isa

doi:10.59261/jequi.v8i4.427

Authors

Vlandimir Welem L Universitas Bina Nusantara
Sani Muhamad Isa Universitas Bina Nusantara

DOI:

https://doi.org/10.59261/jequi.v8i4.427

Keywords:

ARIMA, Hybrid Forecasting, LSTM, Outpatient Visit Forecasting, SHAP

Abstract

Background: Accurate forecasting of outpatient visits is important for hospital staffing, scheduling, facility utilization, and resource allocation. Outpatient demand in the study data shows strong weekly seasonality and holiday-related variation, creating challenges for both forecasting accuracy and model interpretation.

Objective: This study evaluates ARIMA, Prophet, LSTM, and four hybrid configurations for daily outpatient-visit forecasting across five medical specialties and interprets the best-performing hybrid model using SHAP.

Method: Historical daily outpatient-visit data from Internal Medicine, Orthopedic and Traumatology Surgery, Cardiology, Pediatrics, and Neurology were modeled using ARIMA, Prophet, LSTM, ARIMA-Prophet, ARIMA-LSTM, Prophet-LSTM, and ARIMA-Prophet-LSTM. Time-series cross-validation with an expanding window was used, with MAE, RMSE, and MAPE as performance metrics. Friedman and post hoc Nemenyi tests were applied for model comparison, and SHAP's GradientExplainer was used for the LSTM component of Prophet-LSTM.

Results: Prophet achieved the best performance in four specialties, whereas Prophet-LSTM was best in Neurology. The triple hybrid was significantly worse than Prophet (p = 0.017) and did not differ significantly from ARIMA-Prophet (p = 0.9996). SHAP identified day-of-week cyclical features as the most influential predictors across specialties, while national holidays showed low global importance but were associated with large errors in specific cases.

Conclusion: Greater model complexity did not improve forecasting accuracy in these data. Prophet provided the strongest overall performance, while Prophet-LSTM remained a competitive, interpretable hybrid alternative.

Downloads

Download data is not yet available.

References

Band, S. S., Yarahmadi, A., Hsu, C.-C., Biyari, M., Sookhak, M., Ameri, R., Dehzangi, I., Chronopoulos, A. T., & Liang, H.-W. (2023). Application of explainable artificial intelligence in medical health: A systematic review of interpretability methods. Informatics in Medicine Unlocked, 40, 101286. https://doi.org/10.1016/j.imu.2023.101286

Chandekar, A., Akande, O., Makar, C., Grace, K., Abduselam, S., Balogun, O., Nwoke, M., Okeke, C., Gebreyes, M., & Ricks-Oddie, J. (2026). Impact of primary care visit frequency on non-urgent emergency department visits in a large urban medical center. Journal of Primary Care & Community Health, 17, 21501319261420570.

Cousineau, D. (2025). There are no alternative hypotheses in tests of null hypotheses. Frontiers in Psychology, 16. https://doi.org/10.3389/fpsyg.2025.1708313

Di Martino, F., & Delmastro, F. (2023). Explainable AI for clinical and remote health applications: A survey on tabular and time series data. Artificial Intelligence Review, 56(6), 5261–5315. https://doi.org/10.1007/s10462-022-10304-3

Fattouh, M., Lyssenko, L., Heilmeyer, F., Ball, T., & Haverkamp, C. (2026). Forecasting hospital bed occupancy: A time series approach with Prophet. BMC Medical Informatics and Decision Making, 26(1), 162. https://doi.org/10.1186/s12911-026-03542-w

Hariyanto, S. (2022). Forecasting the number of outpatient patient visits using ARIMA at Mitra Sehat Clinic: Case study in Sukoharjo, Central Java Indonesia. Universal Journal of Public Health, 10(5), 482–491. https://doi.org/10.13189/ujph.2022.100506

Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9(8), 1735–1780. https://doi.org/10.1162/neco.1997.9.8.1735

Hyndman, R. J., & Athanasopoulos, G. (2021). Forecasting: Principles and practice (3rd ed.). OTexts.

Jung, J., Lee, H., Jung, H., & Kim, H. (2023). Essential properties and explanation effectiveness of explainable artificial intelligence in healthcare: A systematic review. Heliyon, 9(5), e16110. https://doi.org/10.1016/j.heliyon.2023.e16110

Krichen, M., & Mihoub, A. (2025). Long Short-Term Memory Networks: A Comprehensive Survey. In AI (Switzerland) (Vol. 6, Number 9). Multidisciplinary Digital Publishing Institute (MDPI). https://doi.org/10.3390/ai6090215

Kwiatkowski, D., Phillips, P. C. B., Schmidt, P., & Shin, Y. (1992). Testing the null hypothesis of stationarity against the alternative of a unit root: How sure are we that economic time series have a unit root? Journal of Econometrics, 54(1–3), 159–178. https://doi.org/10.1016/0304-4076(92)90104-Y

Liu, J., Li, C., Ouyang, P., Liu, J., & Wu, C. (2023). Interpreting the prediction results of the tree-based gradient boosting models for financial distress prediction with an explainable machine learning approach. Journal of Forecasting, 42(5). https://doi.org/10.1002/for.2931

Lundberg, S. M., & Lee, S.-I. (2017). A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems, 30, 4765–4774.

Mahmud, A., Noor, S. H. N. S. H., Musa, K. I., Hamzah, F. M., Yudin, Z. M., Kamaruddin, N., Madawana, A. M., & Nawi, M. A. A. (2025). Hybrid ARIMA-LSTM for COVID-19 forecasting: A comparative AI modeling study. PeerJ Computer Science, 11, e3195. https://doi.org/10.7717/peerj-cs.3195

Makridakis, S., Spiliotis, E., & Assimakopoulos, V. (2018). Statistical and machine learning forecasting methods: Concerns and ways forward. PLOS ONE, 13(3), e0194889. https://doi.org/10.1371/journal.pone.0194889

Makridakis, S., Spiliotis, E., Assimakopoulos, V., Semenoglou, A. A., Mulder, G., & Nikolopoulos, K. (2023). Statistical, machine learning and deep learning forecasting methods: Comparisons and ways forward. Journal of the Operational Research Society, 74(3). https://doi.org/10.1080/01605682.2022.2118629

Saini, K. P., & Sharma, A. (2024). Impact of data preprocessing on Prophet-LSTM hybrid models for time series forecasting. Educational Administration: Theory and Practice, 30(5), 14647–14650.

Shaikh, A. S., Samant, R. M., Patil, K. S., Patil, N. R., & Mirkale, A. R. (2023). Review on Explainable AI by using LIME and SHAP models for healthcare domain. International Journal of Computer Applications, 185(45), 18–23. https://doi.org/10.5120/ijca2023923263

Sprangers, O., Wadman, W., Schelter, S., & de Rijke, M. (2024). Hierarchical forecasting at scale. International Journal of Forecasting, 40(4). https://doi.org/10.1016/j.ijforecast.2024.02.006

Statistics, C. for D. C. N. C. for H. (2017). Fast stats. Ambulatory Care Use and Physician Office Visits.

Taylor, S. J., & Letham, B. (2018). Forecasting at scale. The American Statistician, 72(1), 37–45. https://doi.org/10.1080/00031305.2017.1380080

Wang, Y., Yi, X., Luo, M., Wang, Z., Qin, L., Hu, X., & Wang, K. (2023). Prediction of outpatients with conjunctivitis in Xinjiang based on LSTM and GRU models. PLOS ONE, 18(9), e0290541. https://doi.org/10.1371/journal.pone.0290541

Wei, Y., Li, W., Tan, J., Yuan, J., Wu, Z., Li, Y., Mao, Y., & Huang, D. (2022). A method for improving the prediction of outpatient visits for hospital management: Bayesian autoregressive analysis. Computational and Mathematical Methods in Medicine, 2022, 4718157. https://doi.org/10.1155/2022/4718157

Zhang, Y., Wang, L., Li, H., & Chen, X. (2023). Time-series forecasting of healthcare utilization incorporating meteorological factors: A machine learning approach. BMC Health Services Research, 23, 1189.