In this systematic review, studies leveraging everything from wearables to free text data tout solid AUCs, but almost all crumble under the weight of bias and poor validation. PGHD-AI may promise streamlined pre-care triage, yet for now the evidence is mostly smoke and mirrors. Any clinician asked to rely on these models for pre-visit risk screening is left holding considerable uncertainty as little work measures impact on outcomes or equity.
Are AI Models Using Patient-Generated Data Ready for the Pre-Care Front Desk?
Can AI models that use patient-generated health data reliably improve triage, history-taking, or symptom assessment before elective care?