Background: Determining treatment options for patients with cancer can be difficult for providers, especially when it comes to deciding the appropriate time to stop therapy. Many factors come into play when making important treatment and quality of life decisions including cancer responsiveness to treatment, patient preferences, labs, and diagnostic studies. Providers use their experience, instinct, and patient input in making these decisions, and can still have a difficult time knowing the best option for the patient. To support providers in identifying patients at high risk for poor outcomes, we developed a 90-day mortality risk machine-learning model (MRM) for patients with metastatic cancer. The tool was designed to facilitate earlier advance care planning (ACP) discussions and prognostic awareness leading to alignment of treatment with patient goals and values. The US Oncology Network (The Network) implemented the MRM beginning in early 2021 through late 2022. We assessed the impact of the MRM using performance on 5 EOL Quality Measures included in Medicare’s Merit-based Incentive Payment System (MIPS) Program. Methods: MRM predictions were made every two weeks and provided to practices through a secure web-based dashboard. Practices customized their use of the MRM based on their unique workflow. MRM utilization was collected July – December 2023 and determined how frequently practices accessed the dashboard. Practices were separated into two groups based on MRM utilization: High Utilizers (HU) used the tool more than once a week and Low Utilizers (LU) did not use the tool or used it less than or equal to once a week. EOL MIPS measure performance was collected and evaluated from the 2023 calendar year. For each of the EOL MIPS measures, we compared the average score between the two practice groups. A 2-sample unequal variance t-test with 1-tailed distribution was used to determine if there was a significant difference between the two practice groups for each EOL MIPS measure. Results: A total of 25 practices had available EOL MIPS scores. Five practices were classified as a HU group and 20 were in the LU group. EOL MIPS measure scores were higher among HU. All results trended towards improvement with a statistically significant difference observed for PIMSH 1 and MIPS 457 for HU compared to LU. The table shows the average MIPS scores across the two practice groups. Conclusions: Practices with the highest reported utilization of the MRM also reported improved MIPS EOL quality scores for the 2023 calendar year. While each practice implemented MRM independently, those using the tool more than once per week had improved performance in their MIPS EOL quality scores.