アプリケーションガイド

AI-Drafted Replies to Patient Portal Messages

AI-drafted replies are responses that a language model writes inside a clinician's EHR inbox for a patient's portal message.

4 分で読めます
最終更新日 2026年9月15日

読み始めるクイズに答えてください

このページでは4 分で読めます

概要
ディープダイブ
戦略的影響
The Future of AI-Drafted Replies to Patient Portal Messages
現実世界の実装
リスクとガードレール
実装ロードマップ
探検を続けましょう
よくある質問

概要

A clinician must review, edit and send each one. Early studies suggest the drafts can reduce the mental burden of inbox work and are often seen as empathetic, but they have not reliably saved time, and their safety depends on careful human review.

ディープダイブ

Large-scale use started in 2023, when Epic and Microsoft began offering GPT-4-based draft replies through Azure OpenAI. Early adopters included UC San Diego Health, UW Health and Stanford Health Care. In a typical setup, the system reads a patient message and some chart context, then shows a suggested reply. The clinician can start from the draft or start blank. Nothing goes to the patient automatically. The research is mixed. A 2023 study by Ayers and colleagues in JAMA Internal Medicine compared ChatGPT answers with physician answers to public health questions posted on Reddit. Evaluators preferred the chatbot answers in most comparisons and rated them higher for quality and empathy. That setting was not a real clinical inbox. Studies inside health systems give a more modest picture. A Stanford study by Garcia and colleagues in JAMA Network Open in 2024 found clinicians used drafts for about a fifth of messages, with reductions in task load and burnout scores but no significant time savings. A 2024 UC San Diego study by Tai-Seale and colleagues found no drop in reply time, longer time spent reading messages, and longer replies. Safety research shows why review matters. A 2024 simulation study at Mass General Brigham, published in Lancet Digital Health, found that a small share of unedited GPT-4 drafts could have caused severe harm if sent. Physicians who used drafts sometimes left the model's content in place. That is automation bias: people tend to accept a fluent draft. The main misconception is that drafting is mainly a time-saver. The evidence so far points more to reduced mental burden than to saved minutes. Disclosure is becoming a legal issue too. California's AB 3030, effective in 2025, requires health providers to disclose AI-generated patient communications unless a licensed provider reviewed them.

技術的な洞察

A typical pipeline starts by classifying the incoming message, for example as a medication question, symptom report, scheduling request or test result question. Some categories may be excluded from drafting. The system then builds a prompt from the message, limited chart context such as medication list, allergies, problem list and recent visits, and instructions on tone, reading level and what the model must not do, such as diagnose or change doses. The reply appears in a draft pane, and a clinician decides whether to use it. Risk is concentrated in missing context: the model sees only what the prompt includes, so an unfiled lab result or an outside prescription can lead to a confident wrong answer. Useful safeguards include excluding high-risk categories, keeping the start-blank option, flagging messages with emergency language for triage, and length limits. Log how often drafts are used and how much clinicians edit them, and review a sample of sent messages in which the draft was barely changed.

戦略的影響

ビルドの選択

AI が実際の成果を向上させるかどうかは、アプリケーションレベルの設計によって決まります。

チームとワークフロー

ワークフローを適切に統合すると、ユーザーが信頼できる生産性が向上します。

リスクと安全性

適切な範囲のユースケースにより、変更の疲労と実装のリスクが軽減されます。

The Future of AI-Drafted Replies to Patient Portal Messages

Inbox drafting is likely to spread because message volume keeps rising, but the reasons to adopt it are shifting from saved minutes to reduced strain on clinicians. That claim needs longer follow-up to confirm. Better use of chart context and more careful triage of which messages get drafts may improve accuracy. The bigger risk remains automation bias, so training, audits and interface designs that make editing easy will matter as much as the model. Expect more attention to patient disclosure and consent as states and health systems set rules. The open research question is whether drafts change outcomes for patients, not only for clinicians.

現実世界の実装

A nurse opens a patient's message about a cough that has lasted three weeks. She finds a pre-written draft asking about fever and shortness of breath, edits it to offer a same-week visit, and sends it.

A physician discards a draft that told a patient to double a blood pressure medication, because the model did not see yesterday's lab results showing high potassium.

A health system adds a line to replies saying the message was drafted with AI help and reviewed by a clinician, so patients know how their care team uses the tool.

An organization turns off drafting for test result messages and possible emergencies, sending those to nurse triage with no AI draft.

リスクとガードレール

壊れたプロセスを自動化すると、既存の問題がさらに拡大する可能性があります。
チームが過剰に自動化し、必要な人間の判断を排除してしまう可能性があります。
出力が継続的に評価されないと、品質が変動する可能性があります。

実装ロードマップ

現在のワークフローをマッピングし、最も摩擦が大きいステップを特定します。
完全自動化の前に人間によるチェックポイントを定義します。
プロンプト、エスカレーションパス、品質基準についてユーザーをトレーニングします。
タスクレベルの結果を追跡して、持続的な価値を確認します。

探検を続けましょう

AIアシスタント

有用で信頼できるデザインアシスタントのワークフロー。ガイドを読む

AIコーディング

応用された AI がソフトウェア配信をどのように改善するかをご覧ください。ガイドを読む

Free newsletter

Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.

One email each weekday. Unsubscribe in one click. We never sell or share your address.

Test yourself

Take the AI-Drafted Replies to Patient Portal Messages quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

クイズを開始する

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

よくある質問

What is AI-Drafted Replies to Patient Portal Messages?

AI-drafted replies are responses that a language model writes inside a clinician's EHR inbox for a patient's portal message. A clinician must review, edit and send each one. Early studies suggest the drafts can reduce the mental burden of inbox work and are often seen as empathetic, but they have not reliably saved time, and their safety depends on careful human review.

What did the 2024 Stanford study by Garcia and colleagues find about AI-drafted inbox replies?

Clinicians used drafts for about a fifth of messages. Burden measures improved, but time did not change significantly.

Why does the Ayers 2023 study not prove AI drafts beat physicians in real clinical inboxes?

The study rated chatbot and physician answers to public forum questions, which lack real chart context and patient relationships.

The UC San Diego study by Tai-Seale and colleagues found which pattern?

Reply time did not fall, reading time rose, and replies got longer. That undercuts the idea that drafting mainly saves time.

A physician leaves a fluent draft mostly unchanged even though it contains an error. What is this tendency called?

Automation bias is the tendency to accept machine output, especially fluent output, without enough scrutiny.

A draft tells a patient to double a blood pressure medication, missing yesterday's high potassium result. What risk does this show?

If key data such as a new lab result is not in the prompt, the model can give a confident but unsafe answer.

学び続ける