Dataset Information

Medical Specialty Recommendations by an Artificial Intelligence Chatbot on a Smartphone: Development and Deployment.

ABSTRACT:

Background

The COVID-19 pandemic has limited daily activities and even contact between patients and primary care providers. This makes it more difficult to provide adequate primary care services, which include connecting patients to an appropriate medical specialist. A smartphone-compatible artificial intelligence (AI) chatbot that classifies patients' symptoms and recommends the appropriate medical specialty could provide a valuable solution.

Objective

In order to establish a contactless method of recommending the appropriate medical specialty, this study aimed to construct a deep learning-based natural language processing (NLP) pipeline and to develop an AI chatbot that can be used on a smartphone.

Methods

We collected 118,008 sentences containing information on symptoms with labels (medical specialty), conducted data cleansing, and finally constructed a pipeline of 51,134 sentences for this study. Several deep learning models, including 4 different long short-term memory (LSTM) models with or without attention and with or without a pretrained FastText embedding layer, as well as bidirectional encoder representations from transformers for NLP, were trained and validated using a randomly selected test data set. The performance of the models was evaluated on the basis of the precision, recall, F₁-score, and area under the receiver operating characteristic curve (AUC). An AI chatbot was also designed to make it easy for patients to use this specialty recommendation system. We used an open-source framework called "Alpha" to develop our AI chatbot. This takes the form of a web-based app with a frontend chat interface capable of conversing in text and a backend cloud-based server application to handle data collection, process the data with a deep learning model, and offer the medical specialty recommendation in a responsive web that is compatible with both desktops and smartphones.

Results

The bidirectional encoder representations from transformers model yielded the best performance, with an AUC of 0.964 and F₁-score of 0.768, followed by LSTM model with embedding vectors, with an AUC of 0.965 and F₁-score of 0.739. Considering the limitations of computing resources and the wide availability of smartphones, the LSTM model with embedding vectors trained on our data set was adopted for our AI chatbot service. We also deployed an Alpha version of the AI chatbot to be executed on both desktops and smartphones.

Conclusions

With the increasing need for telemedicine during the current COVID-19 pandemic, an AI chatbot with a deep learning-based NLP model that can recommend a medical specialty to patients through their smartphones would be exceedingly useful. This chatbot allows patients to identify the proper medical specialist in a rapid and contactless manner, based on their symptoms, thus potentially supporting both patients and primary care providers.

SUBMITTER: Lee H

PROVIDER: S-EPMC8104000 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BackgroundChatbots empowered by artificial intelligence (AI) can increasingly engage in natural conversations and build relationships with users. Applying AI chatbots to lifestyle modification programs is one of the promising areas to develop cost-effective and feasible behavior interventions to promote physical activity and a healthy diet.ObjectiveThe purposes of this perspective paper are to present a brief literature review of chatbot use in promoting physical activity and a healthy diet, describe the AI chatbot behavior change model our research team developed based on extensive interdisciplinary research, and discuss ethical principles and considerations.MethodsWe conducted a preliminary search of studies reporting chatbots for improving physical activity and/or diet in four databases in July 2020. We summarized the characteristics of the chatbot studies and reviewed recent developments in human-AI communication research and innovations in natural language processing. Based on the identified gaps and opportunities, as well as our own clinical and research experience and findings, we propose an AI chatbot behavior change model.ResultsOur review found a lack of understanding around theoretical guidance and practical recommendations on designing AI chatbots for lifestyle modification programs. The proposed AI chatbot behavior change model consists of the following four components to provide such guidance: (1) designing chatbot characteristics and understanding user background; (2) building relational capacity; (3) building persuasive conversational capacity; and (4) evaluating mechanisms and outcomes. The rationale and evidence supporting the design and evaluation choices for this model are presented in this paper.ConclusionsAs AI chatbots become increasingly integrated into various digital communications, our proposed theoretical framework is the first step to conceptualize the scope of utilization in health behavior change domains and to synthesize all possible dimensions of chatbot features to inform intervention design and evaluation. There is a need for more interdisciplinary work to continue developing AI techniques to improve a chatbot's relational and persuasive capacities to change physical activity and diet behaviors with strong ethical principles.

Project description:ImportanceConsumers are increasingly using artificial intelligence (AI) chatbots as a source of information. However, the quality of the cancer information generated by these chatbots has not yet been evaluated using validated instruments.ObjectiveTo characterize the quality of information and presence of misinformation about skin, lung, breast, colorectal, and prostate cancers generated by 4 AI chatbots.Design, setting, and participantsThis cross-sectional study assessed AI chatbots' text responses to the 5 most commonly searched queries related to the 5 most common cancers using validated instruments. Search data were extracted from the publicly available Google Trends platform and identical prompts were used to generate responses from 4 AI chatbots: ChatGPT version 3.5 (OpenAI), Perplexity (Perplexity.AI), Chatsonic (Writesonic), and Bing AI (Microsoft).ExposuresGoogle Trends' top 5 search queries related to skin, lung, breast, colorectal, and prostate cancer from January 1, 2021, to January 1, 2023, were input into 4 AI chatbots.Main outcomes and measuresThe primary outcomes were the quality of consumer health information based on the validated DISCERN instrument (scores from 1 [low] to 5 [high] for quality of information) and the understandability and actionability of this information based on the understandability and actionability domains of the Patient Education Materials Assessment Tool (PEMAT) (scores of 0%-100%, with higher scores indicating a higher level of understandability and actionability). Secondary outcomes included misinformation scored using a 5-item Likert scale (scores from 1 [no misinformation] to 5 [high misinformation]) and readability assessed using the Flesch-Kincaid Grade Level readability score.ResultsThe analysis included 100 responses from 4 chatbots about the 5 most common search queries for skin, lung, breast, colorectal, and prostate cancer. The quality of text responses generated by the 4 AI chatbots was good (median [range] DISCERN score, 5 [2-5]) and no misinformation was identified. Understandability was moderate (median [range] PEMAT Understandability score, 66.7% [33.3%-90.1%]), and actionability was poor (median [range] PEMAT Actionability score, 20.0% [0%-40.0%]). The responses were written at the college level based on the Flesch-Kincaid Grade Level score.Conclusions and relevanceFindings of this cross-sectional study suggest that AI chatbots generally produce accurate information for the top cancer-related search queries, but the responses are not readily actionable and are written at a college reading level. These limitations suggest that AI chatbots should be used supplementarily and not as a primary source for medical information.

Project description:BackgroundAdvanced analytics, such as artificial intelligence (AI), increasingly gain relevance in medicine. However, patients' responses to the involvement of AI in the care process remains largely unclear. The study aims to explore whether individuals were more likely to follow a recommendation when a physician used AI in the diagnostic process considering a highly (vs. less) severe disease compared to when the physician did not use AI or when AI fully replaced the physician.MethodsParticipants from the USA (n = 452) were randomly assigned to a hypothetical scenario where they imagined that they received a treatment recommendation after a skin cancer diagnosis (high vs. low severity) from a physician, a physician using AI, or an automated AI tool. They then indicated their intention to follow the recommendation. Regression analyses were used to test hypotheses. Beta coefficients (ß) describe the nature and strength of relationships between predictors and outcome variables; confidence intervals [CI] excluding zero indicate significant mediation effects.ResultsThe total effects reveal the inferiority of automated AI (ß = .47, p = .001 vs. physician; ß = .49, p = .001 vs. physician using AI). Two pathways increase intention to follow the recommendation. When a physician performs the assessment (vs. automated AI), the perception that the physician is real and present (a concept called social presence) is high, which increases intention to follow the recommendation (ß = .22, 95% CI [.09; 0.39]). When AI performs the assessment (vs. physician only), perceived innovativeness of the method is high, which increases intention to follow the recommendation (ß = .15, 95% CI [- .28; - .04]). When physicians use AI, social presence does not decrease and perceived innovativeness increases.ConclusionPairing AI with a physician in medical diagnosis and treatment in a hypothetical scenario using topical therapy and oral medication as treatment recommendations leads to a higher intention to follow the recommendation than AI on its own. The findings might help develop practice guidelines for cases where AI involvement benefits outweigh risks, such as using AI in pathology and radiology, to enable augmented human intelligence and inform physicians about diagnoses and treatments.

Project description:BackgroundIt is unlikely that applications of artificial intelligence (AI) will completely replace physicians. However, it is very likely that AI applications will acquire many of their roles and generate new tasks in medical care. To be ready for new roles and tasks, medical students and physicians will need to understand the fundamentals of AI and data science, mathematical concepts, and related ethical and medico-legal issues in addition with the standard medical principles. Nevertheless, there is no valid and reliable instrument available in the literature to measure medical AI readiness. In this study, we have described the development of a valid and reliable psychometric measurement tool for the assessment of the perceived readiness of medical students on AI technologies and its applications in medicine.MethodsTo define medical students' required competencies on AI, a diverse set of experts' opinions were obtained by a qualitative method and were used as a theoretical framework, while creating the item pool of the scale. Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA) were applied.ResultsA total of 568 medical students during the EFA phase and 329 medical students during the CFA phase, enrolled in two different public universities in Turkey participated in this study. The initial 27-items finalized with a 22-items scale in a four-factor structure (cognition, ability, vision, and ethics), which explains 50.9% cumulative variance that resulted from the EFA. Cronbach's alpha reliability coefficient was 0.87. CFA indicated appropriate fit of the four-factor model (?2/df?=?3.81, RMSEA?=?0.094, SRMR?=?0.057, CFI?=?0.938, and NNFI (TLI)?=?0.928). These values showed that the four-factor model has construct validity.ConclusionsThe newly developed Medical Artificial Intelligence Readiness Scale for Medical Students (MAIRS-MS) was found to be valid and reliable tool for evaluation and monitoring of perceived readiness levels of medical students on AI technologies and applications. Medical schools may follow 'a physician training perspective that is compatible with AI in medicine' to their curricula by using MAIRS-MS. This scale could be benefitted by medical and health science education institutions as a valuable curriculum development tool with its learner needs assessment and participants' end-course perceived readiness opportunities.

Project description:BackgroundArtificial intelligence (AI) chatbots have recently gained use in medical practice by health care practitioners. Interestingly, the output of these AI chatbots was found to have varying degrees of hallucination in content and references. Such hallucinations generate doubts about their output and their implementation.ObjectiveThe aim of our study was to propose a reference hallucination score (RHS) to evaluate the authenticity of AI chatbots' citations.MethodsSix AI chatbots were challenged with the same 10 medical prompts, requesting 10 references per prompt. The RHS is composed of 6 bibliographic items and the reference's relevance to prompts' keywords. RHS was calculated for each reference, prompt, and type of prompt (basic vs complex). The average RHS was calculated for each AI chatbot and compared across the different types of prompts and AI chatbots.ResultsBard failed to generate any references. ChatGPT 3.5 and Bing generated the highest RHS (score=11), while Elicit and SciSpace generated the lowest RHS (score=1), and Perplexity generated a middle RHS (score=7). The highest degree of hallucination was observed for reference relevancy to the prompt keywords (308/500, 61.6%), while the lowest was for reference titles (169/500, 33.8%). ChatGPT and Bing had comparable RHS (β coefficient=-0.069; P=.32), while Perplexity had significantly lower RHS than ChatGPT (β coefficient=-0.345; P<.001). AI chatbots generally had significantly higher RHS when prompted with scenarios or complex format prompts (β coefficient=0.486; P<.001).ConclusionsThe variation in RHS underscores the necessity for a robust reference evaluation tool to improve the authenticity of AI chatbots. Further, the variations highlight the importance of verifying their output and citations. Elicit and SciSpace had negligible hallucination, while ChatGPT and Bing had critical hallucination levels. The proposed AI chatbots' RHS could contribute to ongoing efforts to enhance AI's general reliability in medical research.

Project description:BackgroundArtificial intelligence-based chatbots are emerging as instruments of psychological intervention; however, no relevant studies have been reported in Latin America.ObjectiveThe objective of the present study was to evaluate the viability, acceptability, and potential impact of using Tess, a chatbot, for examining symptoms of depression and anxiety in university students.MethodsThis was a pilot randomized controlled trial. The experimental condition used Tess for 8 weeks, and the control condition was assigned to a psychoeducation book on depression. Comparisons were conducted using Mann-Whitney U and Wilcoxon tests for depressive symptoms, and independent and paired sample t tests to analyze anxiety symptoms.ResultsThe initial sample consisted of 181 Argentinian college students (158, 87.2% female) aged 18 to 33. Data at week 8 were provided by 39 out of the 99 (39%) participants in the experimental condition and 34 out of the 82 (41%) in the control group. On an average, 472 (SD 249.52) messages were exchanged, with 116 (SD 73.87) of the messages sent from the users in response to Tess. A higher number of messages exchanged with Tess was associated with positive feedback (F2,36=4.37; P=.02). No significant differences between the experimental and control groups were found from the baseline to week 8 for depressive and anxiety symptoms. However, significant intragroup differences demonstrated that the experimental group showed a significant decrease in anxiety symptoms; no such differences were observed for the control group. Further, no significant intragroup differences were found for depressive symptoms.ConclusionsThe students spent a considerable amount of time exchanging messages with Tess and positive feedback was associated with a higher number of messages exchanged. The initial results show promising evidence for the usability and acceptability of Tess in the Argentinian population. Research on chatbots is still in its initial stages and further research is needed.