A machine-learning decision support system for demand-driven nurse staffing in outpatient clinics: Design and validation at an Indonesian public hospital

Muhammad Fadhillah Hadi; Asmuliardi Muluk

doi:10.65112/tcmis.10092

Authors

Muhammad Fadhillah Hadi Department of Industrial Engineering, Faculty of Engineering, Universitas Andalas, Padang, Indonesia https://orcid.org/0009-0001-1126-2853
Asmuliardi Muluk Department of Industrial Engineering, Faculty of Engineering, Universitas Andalas, Padang, Indonesia https://orcid.org/0009-0008-9564-4867

DOI:

https://doi.org/10.65112/tcmis.10092

Keywords:

decision support systems, demand forecasting, ARIMA, nurse staffing, healthcare operations, time-series analysis

Abstract

Public hospitals in developing countries often plan their health workforce reactively despite highly variable daily patient demand, leading to uneven workloads and longer waiting times. This study designed, verified, and retrospectively evaluated a web-based decision support system (DSS) that couples daily demand forecasting with a transparent workload-based nurse-staffing rule and automated cross-unit reallocation, using five years of outpatient records (2021–2025; 364,849 visits) from a class-B public hospital in Batam, Indonesia. ARIMA, long short-term memory (LSTM), and Prophet models were compared for the five highest-volume clinics under a single consistent protocol across three horizons—one-step-ahead, rolling-origin seven-day-ahead, and long-horizon multi-step—using MAE, RMSE, and the scale-free MASE against a seasonal-naive benchmark, with Diebold–Mariano tests for significance. ARIMA was the most reliable model: at the operational seven-day horizon it achieved a mean MASE of 0.79 and significantly outperformed the naive benchmark in four of five clinics, whereas at long horizons all models converged toward naive-level accuracy. A retrospective counterfactual simulation, in which implied nurse requirements are derived from the same workload rule used by the DSS, indicates that forecast-driven recommendations would have reduced total staffing mismatch by 55.9% under realistic seven-working-day planning (60.3% in the one-step upper bound), versus 43.2% for seasonal-naive staffing recommendations, and reveals structural misallocation across clinics, 56% of which could be covered by same-day reallocation. Sensitivity analyses over preprocessing choices, workload parameters, and weekly-lag specifications support the robustness of these findings. System computations were verified exactly, and a single-expert review by the hospital’s Head of Services established face validity of the tool. The results demonstrate that an auditable forecasting-to-staffing pipeline is feasible on routine hospital data, and that evaluation protocol and forecast horizon critically shape conclusions about model superiority.

Downloads

Download data is not yet available.

References

[1] W. D. Savira and W. Subadi, “Kualitas pelayanan kesehatan dilihat dari aspek tangible dalam rangka meningkatkan pelayanan kesehatan pada Puskesmas Hikun,” Jurnal Administrasi Publik dan Administrasi Bisnis, vol. 6, no. 2, pp. 749–762, 2023.

[2] Badan Pusat Statistik, Profil Statistik Kesehatan 2025. Jakarta, Indonesia: Badan Pusat Statistik, 2025.

[3] R. Amran, A. Apriyani, and N. P. Dewi, “Peran penting kelengkapan rekam medik di rumah sakit,” Baiturrahmah Medical Journal, vol. 1, no. 1, pp. 69–76, 2021.

[4] A. Komashie, S. Hinrichs-Krapels, and P. J. Clarkson, “Systems approaches to healthcare systems design and care delivery: an overview of the literature,” Proceedings of the Design Society, vol. 1, pp. 2941–2950, 2021, doi: 10.1017/pds.2021.555. DOI: https://doi.org/10.1017/pds.2021.555

[5] Peraturan Wali Kota Batam No. 140 Tahun 2024 tentang Rumah Sakit Umum Daerah Embung Fatimah, 2024.

[6] A. Sebaa, A. Nouicer, A. Tari, R. Tarik, and O. Abdellah, “Decision support system for health care resources allocation,” Electronic Physician, vol. 9, no. 6, pp. 4661–4668, 2017, doi: 10.19082/4661. DOI: https://doi.org/10.19082/4661

[7] D. J. Power, Decision Support Systems: Concepts and Resources for Managers. Westport, CT, USA: Quorum Books, 2002.

[8] E. Turban, J. E. Aronson, and T. P. Liang, Decision Support Systems and Intelligent Systems, 7th ed. New Delhi, India: Prentice-Hall, 2007.

[9] A. Aditya et al., Machine Learning. Padang, Indonesia: PT Global Eksekutif Teknologi, 2023.

[10] B. M. Porto and F. S. Fogliatto, “Enhanced forecasting of emergency department patient arrivals using feature engineering approach and machine learning,” BMC Medical Informatics and Decision Making, vol. 24, no. 1, Art. no. 377, 2024, doi: 10.1186/s12911-024-02788-6. DOI: https://doi.org/10.1186/s12911-024-02788-6

[11] B. Rostami-Tabar, J. Browell, and I. Svetunkov, “Probabilistic forecasting of hourly emergency department arrivals,” Health Systems, vol. 13, no. 2, pp. 133–149, 2024, doi: 10.1080/20476965.2023.2200526. DOI: https://doi.org/10.1080/20476965.2023.2200526

[12] I. Basri K., “Peramalan jumlah kunjungan pasien rawat jalan menggunakan metode ARIMA, SES dan Holt-Winters di Balai Kesehatan Masyarakat XYZ,” Jurnal Tata Kelola dan Kerangka Kerja Teknologi Informasi, vol. 7, no. 1, pp. 29–39, 2021, doi: 10.34010/jtk3ti.v7i1.5548. DOI: https://doi.org/10.34010/jtk3ti.v7i1.5548

[13] N. Manabung, C. Mongi, and Y. Langi, “Peramalan banyaknya pasien yang berobat di Puskesmas Bengkol dengan menggunakan metode ARIMA,” Jurnal LPPM Bidang Sains dan Teknologi, vol. 9, no. 1, pp. 11–19, 2024. DOI: https://doi.org/10.35801/jlppmsains.9.1.2024.49768

[14] H. Bui, S. D. Ekşioğlu, A. A. Villafranca, J. A. Sanford, and K. W. Sexton, “A long short-term memory model for forecasting the surgical case volumes at a hospital,” IISE Transactions on Healthcare Systems Engineering, vol. 13, no. 3, pp. 226–236, 2023. DOI: https://doi.org/10.1080/24725579.2023.2205180

[15] V. I. Kontopoulou, A. D. Panagopoulos, I. Kakkos, and G. K. Matsopoulos, “A review of ARIMA vs. machine learning approaches for time series forecasting in data driven networks,” Future Internet, vol. 15, no. 8, Art. no. 255, 2023, doi: 10.3390/fi15080255. DOI: https://doi.org/10.3390/fi15080255

[16] I. G. T. Suryawan, I. K. N. Putra, P. M. Meliana, and I. G. I. Sudipa, “Performance comparison of ARIMA, LSTM, and Prophet methods in sales forecasting,” Sinkron: Jurnal dan Penelitian Teknik Informatika, vol. 8, no. 4, pp. 2410–2421, 2024, doi: 10.33395/sinkron.v8i4.14057. DOI: https://doi.org/10.33395/sinkron.v8i4.14057

[17] S. Albahli, “LSTM vs. Prophet: achieving superior accuracy in dynamic electricity demand forecasting,” Energies, vol. 18, no. 2, Art. no. 278, 2025, doi: 10.3390/en18020278. DOI: https://doi.org/10.3390/en18020278

[18] R. Anwar and L. Rassiyanti, “Analisis komparasi model peramalan Prophet dan ARIMA dalam memprediksi harga saham penutupan PT ANTM,” Lattice Journal: Journal of Mathematics Education and Applied, vol. 5, no. 1, pp. 57–74, 2025. DOI: https://doi.org/10.30983/lattice.v5i1.9478

[19] S. J. Taylor and B. Letham, “Forecasting at scale,” The American Statistician, vol. 72, no. 1, pp. 37–45, 2018, doi: 10.1080/00031305.2017.1380080. DOI: https://doi.org/10.1080/00031305.2017.1380080

[20] T. Fischer and C. Krauss, “Deep learning with long short-term memory networks for financial market predictions,” European Journal of Operational Research, vol. 270, no. 2, pp. 654–669, 2018, doi: 10.1016/j.ejor.2017.11.054. DOI: https://doi.org/10.1016/j.ejor.2017.11.054

[21] I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning. Cambridge, MA, USA: MIT Press, 2016.

[22] A. Graves, Supervised Sequence Labelling with Recurrent Neural Networks. Berlin, Germany: Springer, 2012. DOI: https://doi.org/10.1007/978-3-642-24797-2

[23] R. J. Hyndman and G. Athanasopoulos, Forecasting: Principles and Practice, 2nd ed. Melbourne, Australia: OTexts, 2018. DOI: https://doi.org/10.32614/CRAN.package.fpp2

[24] S. Kim and H. Kim, “A new metric of absolute percentage error for intermittent demand forecasts,” International Journal of Forecasting, vol. 32, no. 3, pp. 669–679, 2016, doi: 10.1016/j.ijforecast.2015.12.003. DOI: https://doi.org/10.1016/j.ijforecast.2015.12.003

[25] F. X. Diebold and R. S. Mariano, “Comparing predictive accuracy,” Journal of Business & Economic Statistics, vol. 13, no. 3, pp. 253–263, 1995, doi: 10.1080/07350015.1995.10524599. DOI: https://doi.org/10.1080/07350015.1995.10524599

[26] A. D. Wahuningtyas, R. B. M. T. Pangestu, R. P. Hariono, F. Falah, O. M. Prawiryanto, and D. W. Sutha, “Analisis kebutuhan tenaga kerja berdasarkan beban kerja dengan metode Workload Indicator of Staffing Need (WISN) unit rekam medis di Puskesmas ‘X’,” MEJORA: Medical Journal Awatara, vol. 2, no. 1, pp. 25–31, 2024, doi: 10.61434/mejora.v2i1.142. DOI: https://doi.org/10.61434/mejora.v2i1.142

[27] J. Natassa, N. Maimun, H. M. Ulfa, S. Wardani, and R. Isnani, “Analisis kebutuhan tenaga rekam medis menggunakan metode WISN,” Jurnal Kesehatan Komunitas, vol. 10, no. 3, pp. 546–554, 2025, doi: 10.25311/keskom.Vol10.Iss3.1734. DOI: https://doi.org/10.25311/keskom.Vol10.Iss3.1734

[28] A. Epizitone, S. P. Moyane, and I. E. Agbehadji, “A systematic literature review of health information systems for healthcare,” Healthcare, vol. 11, no. 7, Art. no. 959, 2023, doi: 10.3390/healthcare11070959. DOI: https://doi.org/10.3390/healthcare11070959

[29] K. F. Nisa, T. Dompak, and K. T. Lodan, “Perbandingan sistem pelayanan kesehatan di Indonesia dan di Singapura,” Prosiding Seminar Nasional Ilmu Sosial dan Teknologi (SNISTEK), vol. 6, pp. 72–77, 2024.

[30] R. B. Cleveland, W. S. Cleveland, J. E. McRae, and I. Terpenning, “STL: a seasonal-trend decomposition procedure based on loess,” Journal of Official Statistics, vol. 6, no. 1, pp. 3–73, 1990.