<![CDATA[Background: Falls are a major health problem in older adults, with serious physical and psychological consequences. Identifying fall risk factors is crucial to support prevention efforts, particularly for older adults in nursing homes. Purpose: To analyze fall risk scores for older adults based on daily activities and machine learning. Methods: The study used a mixed methods design with 31 elderly respondents in nursing homes. Quantitative analysis was conducted using logistic regression, ROC, and Spearman Rank correlation. Qualitative data were obtained through in-depth interviews and analyzed thematically. Results: Logistic regression analysis showed that the Timed Up and Go Test (TUGT) was a significant predictor of fall risk (p = 0.012; Exp(B) = 0.202), with an AUC of 0.75 and an accuracy of 73%. Qualitative findings identified six main themes: fall experience, feelings after a fall, daily activities, use of walking aids, fear of activities, and ways to prevent falls. Integration of the results demonstrated consistency between objective mobility limitations (TUGT) and the subjective experiences of older adults, who tend to limit activities due to fear of falling. Based on these findings, a simple risk score prototype was developed using the TUGT (Triggered Daily Activity Score), daily step count, activity calories, and cognitive status (MMSE) as indicators. A total score of 4 or greater is categorized as high risk for falls. Conclusion: The TUGT has been shown to be an important indicator of fall risk in older adults, but subjective experience suggests that psychological factors also play a significant role in limiting activity. The developed daily activity-based risk score has the potential to be a rapid screening tool in nursing homes, although external validation in a broader population is still needed. Suggestion: Healthcare workers and nursing home managers are advised to use this risk score as an initial screening tool and to provide preventive interventions for high-risk seniors. Future researchers are encouraged to conduct prospective validation with a larger sample size and consider integrating the risk score into wearable technology or digital applications for rapid screening and ongoing monitoring. Keywords: Daily Activity; Fall Risk; Machine Learning; Older Adults. Pendahuluan: Jatuh merupakan salah satu masalah kesehatan utama pada lansia yang dapat menimbulkan dampak serius, baik fisik maupun psikologis. Identifikasi faktor risiko jatuh menjadi penting untuk mendukung upaya pencegahan, terutama pada lansia di panti werdha. Tujuan: Untuk menganalisis skor risiko jatuh lansia berbasis aktivitas harian dan machine learning. Metode: Penelitian menggunakan desain mixed methods dengan 31 responden lansia di Panti Werdha. Analisis kuantitatif dilakukan menggunakan uji regresi logistik, ROC, serta korelasi Spearman Rank. Data kualitatif diperoleh melalui wawancara mendalam dan dianalisis tematik. Hasil: Analisis regresi logistik menunjukkan bahwa Timed Up and Go Test (TUGT) merupakan prediktor signifikan risiko jatuh (p = 0.012; Exp(B) = 0.202), dengan AUC sebesar 0.75 dan akurasi 73%. Temuan kualitatif mengidentifikasi enam tema utama, yaitu pengalaman jatuh, perasaan setelah jatuh, aktivitas sehari-hari, penggunaan alat bantu jalan, rasa takut beraktivitas, dan cara mencegah jatuh. Integrasi hasil menunjukkan konsistensi antara keterbatasan mobilitas objektif (TUGT) dengan pengalaman subjektif lansia yang cenderung membatasi aktivitas karena takut jatuh. Berdasarkan temuan ini, dikembangkan prototipe skor risiko sederhana dengan indikator TUGT, jumlah langkah harian, kalori aktivitas, dan status kognitif (MMSE). Skor total ≥ 4 dikategorikan risiko tinggi jatuh. Simpulan: TUGT terbukti sebagai indikator penting risiko jatuh pada lansia, namun pengalaman subjektif memperlihatkan bahwa faktor psikologis juga berperan besar dalam pembatasan aktivitas. Skor risiko berbasis aktivitas harian yang dikembangkan berpotensi menjadi alat skrining cepat di panti werdha, meskipun masih memerlukan validasi eksternal pada populasi yang lebih luas. Saran: Bagi tenaga kesehatan dan pengelola panti, disarankan untuk menggunakan skor risiko ini sebagai alat skrining awal serta memberikan intervensi preventif bagi lansia berisiko tinggi. Bagi peneliti selanjutnya, dianjurkan melakukan validasi prospektif dengan sampel lebih besar dan mempertimbangkan integrasi skor risiko ke dalam teknologi wearable atau aplikasi digital untuk skrining cepat dan pemantauan berkelanjutan. Kata Kunci: Aktivitas Harian; Lansia; Machine Learning; Risiko Jatuh.]]>
