This collection doesn't say how many patients can't finish an AI-run medical interview because of devices, internet or digital skills.

This explores how many patients can't finish an AI-run clinical interview because of technology barriers. The short answer is that this collection doesn't contain that number. None of the retrieved notes reports a dropout or non-completion rate tied to devices, connectivity or digital skills. If you need a real figure, look for feasibility or implementation studies of AI intake tools that report completion rates broken down by age or digital literacy. Those studies aren't in this set.

The closest evidence comes from the opposite direction. In a study where a conversational AI took histories from 100 real urgent-care patients, none of the interviews needed a safety stop, and patients' attitudes toward AI improved afterward Can conversational AI safely take patient histories without supervision?. That sounds like strong completion, but treat it carefully. It was a single-arm study, and the summary doesn't say who was screened out before enrolling. That is usually where technology barriers hide: people who can't or won't use the system are filtered out before the study starts, so they never show up as dropouts.

What the corpus does suggest is that the bigger barriers may be psychological rather than technical. Patients resist medical AI for three separate reasons. They feel it can't address their particular situation, they believe it performs worse than a human, and they see it as harder to hold accountable. These beliefs persist regardless of how capable the system actually is Why do patients distrust medical AI systems?. So a patient who 'doesn't complete' an AI interview may be opting out rather than locked out, and a count of technology failures would miss them.

The less obvious lesson is that the medium and the workflow shape whether people engage at all. In a therapy study, a robot and a paper worksheet reduced distress, while a chatbot running the same language model did not. The format and the sense of social presence did the work, not the model Why do robots outperform chatbots in therapy despite identical language models?. Likewise, the Nairobi clinic deployment credits its error reductions to interface design and active rollout, not to the model alone Can AI safety nets reduce errors in live clinical practice?. This suggests that a completion rate is partly a design outcome rather than a fixed property of patients. The same population might finish an AI interview at very different rates depending on whether it runs as a chat window, a staffed kiosk or a voice call.