Automated conversational agents for post-intervention follow-up: a systematic review

Area of Science:

• Natural Language Processing
• Machine Learning
• Healthcare Technology

Background:

• Automated agents, or chatbots, leverage advances in machine learning and natural language processing to simulate human conversation.
• Current applications are primarily commercial, with medical uses including symptom checking and psychotherapy.
• This review focuses on chatbots for post-physical healthcare intervention patient follow-up.

Purpose of the Study:

• To systematically review the acceptability and implementation success of chatbots in patient follow-up after physical healthcare interventions.

Main Methods:

• A PRISMA-compliant systematic review of multiple databases (MEDLINE, EMBASE, PsychINFO, CINAHL, CENTRAL) and grey literature up to September 2020.
• Duplicate abstract screening and data extraction were performed, with risk of bias and quality assessments for included studies.
• Included studies comprised three randomized controlled trials, one non-randomized clinical trial, and six cohort studies.

Main Results:

• Ten studies met inclusion criteria, utilizing chatbots for monitoring after interventions for cancer, hypertension, asthma, orthopaedic procedures, ureteroscopy, and varicose veins.
• All chatbots were deployed on mobile devices, showing engagement rates from 31% to 97% response rates for system-generated questions.
• No included study assessed patient safety outcomes.

Conclusions:

• A variety of chatbot designs and applications were identified for patient monitoring.
• Further investigation into the acceptability, efficacy, and mechanisms of chatbots in outpatient settings is warranted.
• Evidence may support the integration of chatbots into routine clinical care pathways.