Algorithm-enabled Patients Activated in Cancer care through Teams (A-PACT) toimprove goals of care communication for people with cancer - Project Summary/Abstract For patients with cancer, goals of care (GoC) conversations regarding prognosis, values, and advance care planning are a critical but greatly underutilized component of oncology practice. Early GoC communication among patients with cancer reduces unwanted care utilization (hospitalization), improves patient mood (anxiety and depression), and improves communication regarding care preferences. However, reliance on oncology clinicians to identify appropriate patients who need urgent conversations and to initiate these conversations in clinic are major barriers to GoC communication. To that end, we developed Patients Activated in Cancer care through Teams (PACT), a six-month telephonic lay health worker-(LHW)-led intervention to deliver structured education and engage patients in GoC conversations between oncology clinic visits. In single-institution efficacy studies, PACT doubled GoC conversations and halved end-of-life hospitalizations. Reliance on research or clinical staff to manually abstract eligible patients for PACT limits scalability in busy community oncology settings, where most patients with cancer receive their oncology care. To address this critical gap, we developed and demonstrated feasibility of Algorithm-Enabled PACT (A-PACT), which uses machine learning algorithms to automatically identify high-risk patients to expedite referral to PACT. We have demonstrated both prospective and external validation of this machine learning algorithm in community oncology. Building on our studies showing efficacy, we propose to test A-PACT’s effectiveness on healthcare utilization and patient- reported outcomes and explore factors shaping effectiveness and implementation across community oncology sites, using a hybrid type I effectiveness-implementation study. Our randomized trial is conducted through NCI’s SWOG Cancer Research Network and implemented in NCI’s National Community Oncology Research Program, a network of >1000 cancer practices nationwide. We test the effectiveness of A-PACT on reducing hospitalizations and intensive end-of-life care (aim 1) and on improving patient-reported anxiety, depression, and metrics of communication and care preferences (aim 2). To guide scale and reach in community oncology, we will assess how patient, caregiver, clinician, and organizational factors shape A-PACT effectiveness using mixed-methods guided by implementation science frameworks (aim 3). The A-PACT intervention is innovative for its multi-level approach and integration of machine learning to scale access to lay health worker GoC engagement. The proposal is innovative in using principles of implementation science and by producing dissemination toolkits for algorithm and lay health worker workflows to guide future scale in diverse oncology settings. Responding to PAR-24-072 (Cancer Prevention and Control Clinical Trials Grant Program) and NCI’s NOSI on use of telehealth in cancer-related care (NOT-CA-21-043), this proposal, if successful, will provide a scalable framework to engage patients with cancer in earlier GoC communication in order to reduce unwanted care utilization and improve patient-reported outcomes, particularly near the end-of-life.
