Validasi Internal Model Skor ICU RSUP Dr. Sardjito Yogyakarta Untuk Kematian di ICU RSUP Dr. Sardjito Yogyakarta

Pramuditya Haryo Sarjono, Dr. Bowo Adiyanto, Sp.AN.; KIC.; dr. Yunita Widyastuti, MKes.; Sp.An.; KAP., PhD.

2023 | Tesis-Spesialis | S2 Anestesiologi

Latar Belakang: Sistem penilaian yang paling umum digunakan pada pasien dewasa di ICU adalah APACHE II dan III, GCS, SAPS, MPM, SOFA, MODS, dan LODS. Namun, pada banyak kasus, terjadi kegagalan akurasi dan kalibrasi apabila sistem-sistem skoring tersebut digunakan pada populasi baru. Sistem skoring yang dapat memprediksi mortalitas pasien ICU di Indonesia belum ditemukan. Penelitian sebelumnya di RSUP Dr. Sardjito telah mengidentifikasi faktor risiko yang kemudian akan dikembangkan sebagai bahan pembuatan alat skoring prediksi risiko kematian. Namun, sistem skoring tersebut belum divalidasi dan dikalibrasi untuk digunakan dalam populasi besar di Indonesia. Tujuan: Melakukan validasi internal model skoring prediksi kematian yang sedang dikembangkan UGM dengan pasien ICU RSUP Dr. Sardjito. Metode: Penelitian ini menggunakan rancangan penelitian kohort retrospektif observasional. Jumlah sampel yang diambil merupakan pasien ICU dari 1 Januari – 31 Desember 2020. Didapatkan jumlah sampel 698 pasien dengan jumlah sampel yang dieksklusi 27 pasien, sehingga total 671 sample yang selanjutnya dapat dilakukan pengolahan data. Setelah itu, dilakukan penilaian kekuatan diskriminasi dan kalibrasi menggunakan metode Area Under The Curve (AUC). Uji kalibrasi dengan Hosmer-Lameshow dilakukan untuk mendapatkan nilai model Skor Mortalitas ICU RSUP Dr. Sardjito. Hasil : Kekuatan diskriminasi AUC ROC sebesar 0.765 (p < 0 xss=removed> 0,05) pada Hosmer–Lemeshow test memprediksi predicted mortalitas terhadap mortalitas aktual. Kesimpulan : Skor Mortalitas ICU RSUP Dr. Sardjito Yogyakarta mempunyai hasil vaildasi nilai diskriminasi cukup baik untuk memprediksi mortalitas namun dengan kalibrasi yang kurang baik.

Background: The most common scoring systems used in adult patients in the ICU setting are APACHE II and III, GCS, SAPS, MPM, SOFA, MODS, and LODS. However, in most cases, calibration failures occur when these scoring systems are applied to new populations. A scoring system that can predict mortality of ICU patients in Indonesia has not been found. Previous research at RSUP Dr. Sardjito has identified risk factors which will then be developed as materials for making a mortality risk prediction scoring tool. However, the scoring system has not yet been validated and calibrated for use in Indonesias large population. Objective: Conducting internal validation of the mortality prediction scoring model that is being developed by UGM with ICU patients at RSUP Dr. Sardjito.  Method: This study will use an observational retrospective cohort study design. The number of samples taken are ICU patients from January 1 to December 31, 2020. Total of 698 patient were found but 27 of them were excluded, so 671 sample were taken to be analyzed. After that, an assessment of the power of discrimination and calibration was carried out using the Area Under The Curve (AUC) method. Calibration test with Hosmer-Lameshow was carried out to get the value of the ICU Score model at RSUP Dr. Sardjito. Result : The strength of discrimination is acceptable with sensitivity of 77% and specificity of 69.4% shown by AUC ROC of 0.765 (p<0 xss=removed> 0,05) in the Hosmer–Lemeshow test of predicted mortality to the actual mortality.  Conclusion : The ICU score of RSUP Dr. Sardjito Yogyakarta has an acceptable discrimination strenght for predicting mortality but poor calibration result.

Kata Kunci : Skor ICU, Mortalitas, Validasi, ICU score, Mortality, Validation