Perancangan Platform Chatbot Berbasis Large Language Model Untuk Rekomendasi Obat Spesialisasi Telinga, Hidung, dan Tenggorokan Berdasarkan Panduan Praktik Klinis PERHATI-KL

Langit, Hening Mestika (2026) Perancangan Platform Chatbot Berbasis Large Language Model Untuk Rekomendasi Obat Spesialisasi Telinga, Hidung, dan Tenggorokan Berdasarkan Panduan Praktik Klinis PERHATI-KL. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5049221004-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy

Abstract

Perkembangan Large Language Model (LLM) mendorong pemanfaatan chatbot sebagai media penyediaan informasi kesehatan yang cepat dan interaktif. Namun, penggunaan LLM secara langsung masih menghadapi kendala terkait akurasi, konsistensi jawaban, serta kesesuaian dengan panduan praktik klinis nasional. Permasalahan ini menjadi krusial pada bidang Telinga, Hidung, dan Tenggorokan (THT), mengingat tingginya praktik swamedikasi dan keterbatasan masyarakat dalam mengakses informasi obat yang valid dan terstandarisasi di Indonesia. Menanggapi kondisi tersebut, penelitian ini merancang sebuah platform chatbot berbasis LLM yang berfokus pada rekomendasi obat spesialisasi THT dengan mengacu pada Panduan Praktik Klinis (PPK) PERHATI-KL. Pendekatan Retrieval-Augmented Generation (RAG) digunakan untuk mengintegrasikan kemampuan generatif LLM dengan proses document chunking, embedding, penyimpanan pada vector database, serta strategi retrieval berbasis kemiripan semantik. Sistem dirancang dengan pembatasan keluaran agar setiap respon tetap terikat pada konteks panduan klinis, sehingga resiko terjadinya hallucination dapat diminimalkan. Evaluasi sistem dilakukan menggunakan metrik evaluasi RAG menunjukkan kinerja yang baik, dengan nilai faithfulness sebesar ±91,67%, answer relevancy sebesar ±89,87%, context precision sebesar ±90,90%, dan context recall sebesar 100%. Selain itu, pengujian usability menunjukkan tingkat penerimaan yang sangat baik, dengan skor SUS sbesar 80.6 pada user dokter, dan 84 pada user masyarakat umum. Secara keseluruhan, sistem ini berpotensi meningkatkan literasi obat masyarakat serta berfungsi menjadi alat bantu informasi yang aman dan dapat dipertanggungjawabkan.
===========================================================================================================================
The advancement of Large Language Models (LLMs) has encouraged the adoption of chatbots as a medium for delivering fast and interactive health information. However, the direct application of LLMs still faces challenges related to accuracy, respone consistency, and alignment with national clinical practice guidelines. These issues are particularly critical in the field of Otorhinolaryngology (Ear, Nose, and Throat/ENT), given the high prevalence of self-medication practices and the limited public access to valid and standardized drug information in Indonesia. In respone to these conditions, this study proposes an LLM-based chatbot platform focused on ENT drug recommendations, grounded in the Clinical Practice Guidelines issued by PERHATI-KL. A Retrieval-Augmented Generation (RAG) approach is employed to integrate the generative capabilities of LLMs with document chunking, embedding, vector database storage, and semantic similarity–based retrieval strategies. The system incorporates output constraints to ensure that respones remain anchored to clinical guideline contexts, thereby minimizing the risk of hallucination. System evaluation using RAG metrics demonstrates strong performance, with faithfulness of approximately 91.67%, answer relevancy of 89.87%, context precision of 90.90%, and context recall of 100%. Furthermore, usability testing indicates a high level of user acceptance, achieving SUS scores of 80.6 among physician users and 84 among general users. Overall, the proposed system shows potential to enhance public drug literacy and serve as a reliable and accountable health information support tool.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Chatbot, Large Language Model (LLM), Retrieval-Augmented Generation (RAG), Rekomendasi Obat Telinga, Hidung, dan Tenggorokan =============================================================================================================================================== Chatbot, Large Language Model (LLM), Retrieval-Augmented Generation (RAG), ENT Drug recommendations
Subjects:	R Medicine > RF Otorhinolaryngology T Technology > T Technology (General) > T58.6 Management information systems
Divisions:	Faculty of medicine and health (MEDICS) > Medical Technology > 11503-(S1) Undergraduate Thesis
Depositing User:	Hening Mestika Langit
Date Deposited:	03 Aug 2026 04:37
Last Modified:	03 Aug 2026 04:38
URI:	http://repository.its.ac.id/id/eprint/142007

Actions (login required)

View Item