Using a chatbot to reduce emergency department visits and unscheduled hospitalizations among patients with gynecologic malignancies during chemotherapy: A retrospective cohort study

Ming-Yuan Huang; Chia-Sui Weng; Hsiao-Li Kuo; Yung-Cheng Su

doi:10.1016/j.heliyon.2023.e15798

Using a chatbot to reduce emergency department visits and unscheduled hospitalizations among patients with gynecologic malignancies during chemotherapy: A retrospective cohort study

Heliyon. 2023 May 1;9(5):e15798. doi: 10.1016/j.heliyon.2023.e15798. eCollection 2023 May.

Authors

Ming-Yuan Huang^{1

2}, Chia-Sui Weng³, Hsiao-Li Kuo^{3

4}, Yung-Cheng Su⁵

Affiliations

¹ Department of Emergency, Mackay Memorial Hospital, Taipei, Taiwan.
² Department of Medicine, MacKay Medical College, New Taipei City, Taiwan.
³ Department of Obstetrics and Gynecology, MacKay Memorial Hospital, Taipei, Taiwan.
⁴ School of Nursing, Chang Gung University, New Taipei City, Taiwan.
⁵ Department of Emergency Medicine, Ditmanson Medical Foundation Chia-Yi Christian Hospital, Chiayi City, Taiwan.

Abstract

Background: A chatbot is an automatic text-messaging tool that creates a dynamic interaction and simulates a human conversation through text or voice via smartphones or computers. A chatbot could be an effective solution for cancer patients' follow-up during treatment, and could save time for healthcare providers.

Objective: We conducted a retrospective cohort study to evaluate whether a chatbot-based collection of patient-reported symptoms during chemotherapy, with automated alerts to clinicians, could decrease emergency department (ED) visits and hospitalizations. A control group received usual care.

Methods: Self-reporting symptoms were communicated via the chatbot, a Facebook Messenger-based interface for patients with gynecologic malignancies. The chatbot included questions about common symptoms experienced during chemotherapy. Patients could also use the text-messaging feature to speak directly to the chatbot, and all reported outcomes were monitored by a cancer manager. The primary and secondary outcomes of the study were emergency department visits and unscheduled hospitalizations after initiation of chemotherapy after diagnosis of gynecologic malignancies. Multivariate Poisson regression models were applied to assess the adjusted incidence rate ratios (aIRRs) for chatbot use for ED visits and unscheduled hospitalizations after controlling for age, cancer stage, type of malignancy, diabetes, hypertension, chronic renal insufficiency, and coronary heart disease.

Result: Twenty patients were included in the chatbot group, and 43 in the usual-care group. Significantly lower aIRRs for chatbot use for ED visits (0.27; 95% CI 0.11-0.65; p = 0.003) and unscheduled hospitalizations (0.31; 95% CI 0.11-0.88; p = 0.028) were noted. Patients using the chatbot approach had lower aIRRs of ED visits and unscheduled hospitalizations compared to usual-care patients.

Conclusions: The chatbot was helpful for reducing ED visits and unscheduled hospitalizations in patients with gynecologic malignancies who were receiving chemotherapy. These findings are valuable for inspiring the future design of digital health interventions for cancer patients.

Keywords: Chatbot; Gynecologic malignancies; Patient-reported symptoms.