Reka Dwi Syaputra, Sari Widya Ningsih, Deno Harmanto

SISTEM APLIKASI DETEKSI TINGKAT RISIKO KEHAMILAN PADA AKI DI PUSKESMAS TELAGA DEWA

Introduction

Sistem aplikasi deteksi tingkat risiko kehamilan pada aki di puskesmas telaga dewa. Sistem aplikasi deteksi risiko kehamilan pada AKI di Puskesmas Telaga Dewa. Menggunakan Random Forest, aplikasi web ini bantu bidan skrining dini & intervensi efektif.

Abstract

Angka Kematian Ibu (AKI) yang persisten tinggi di Indonesia, khususnya yang disebabkan oleh komplikasi selama masa kehamilan, menuntut adanya inovasi dalam metode skrining dini di fasilitas pelayanan kesehatan primer. Penelitian ini bertujuan untuk mengembangkan, memvalidasi, dan membangun aplikasi untuk prediksi tingkat risiko kehamilan di Puskesmas Telaga Dewa, Bengkulu. Dengan menggunakan kerangka kerja Cross-Industry Standard Process for Data Mining (CRISP-DM), sebuah dataset yang terdiri dari 488 rekam medis pasien ibu hamil dengan 13 variabel klinis objektif dianalisis. Dua algoritma klasifikasi, yaitu Decision Tree dan Random Forest, dievaluasi menggunakan strategi validasi silang 10-fold (10-fold cross-validation) untuk memastikan estimasi kinerja yang robust. Hasil evaluasi menunjukkan keunggulan signifikan dari model Random Forest, yang mencapai nilai rata-rata akurasi 0.97, presisi 0.94, recall 0.99, dan F1-score 0.96. Kinerja ini secara konsisten melampaui model Decision Tree (akurasi 0.85, F1-score 0.85). Analisis feature importance mengidentifikasi tekanan darah, usia, dan riwayat abortus sebagai prediktor paling berpengaruh. Temuan ini menggarisbawahi potensi besar model Random Forest sebagai alat bantu keputusan klinis (Clinical Decision Support) yang akurat dan andal bagi bidan di tingkat Puskesmas. Implementasi model ini dalam bentuk aplikasi berbasis web dapat memfasilitasi stratifikasi risiko pasien secara efisien, memungkinkan alokasi sumber daya yang lebih terfokus, dan mendukung intervensi dini untuk menekan AKI, sejalan dengan agenda transformasi digital kesehatan nasional.

Review

The submitted work presents a timely and highly relevant initiative aimed at tackling the persistently high Maternal Mortality Rate (MMR) in Indonesia through innovative early screening methods. The study's objective to develop, validate, and build a risk prediction application for pregnant women at Puskesmas Telaga Dewa is commendable, addressing a critical need within primary healthcare settings. Utilizing the well-established CRISP-DM framework, the authors systematically analyzed a dataset of 488 medical records with 13 objective clinical variables, employing robust machine learning algorithms—Decision Tree and Random Forest—evaluated through 10-fold cross-validation, which lends credibility to their methodological approach. The findings from the comparative analysis of the machine learning models are particularly compelling. The Random Forest model demonstrated significantly superior performance, achieving impressive metrics including an accuracy of 0.97, precision of 0.94, recall of 0.99, and an F1-score of 0.96. This substantially outperforms the Decision Tree model, which had an accuracy and F1-score of 0.85. The identification of blood pressure, age, and history of abortion as the most influential predictors aligns well with existing clinical understanding of pregnancy risks, reinforcing the model's clinical validity. These strong results position the Random Forest model as a highly accurate and reliable candidate for a Clinical Decision Support (CDS) tool. The practical implications of this research are substantial. The proposed implementation of the validated Random Forest model within a web-based application offers a powerful solution for efficient patient risk stratification by midwives at the Puskesmas level. This digital tool has the potential to streamline clinical workflows, enable more focused allocation of scarce resources, and facilitate timely interventions, ultimately contributing to a reduction in MMR. Furthermore, this initiative aligns perfectly with Indonesia's national digital health transformation agenda, showcasing a tangible application of data-driven approaches to improve maternal health outcomes and demonstrating a significant step forward in leveraging technology for public health.

Full Text

Comments

You need to be logged in to post a comment.