How to handle eligibility pre screening chatbot questions…

Before you build the bot, gather the documents that define exactly who can join each trial. Put them in a folder or a shareable PDF so you can upload them in one go.

• The full inclusion and exclusion criteria for the trial, ideally in the exact wording used in the protocol.
• Any IRB-approved patient-facing language (recruitment flyers, consent form summaries, FAQ sheets).
• A short list of common pre-screening questions your coordinators already answer by phone or email – these will help you test the agent later.
• A Chatref account with at least one workspace. (If you work across multiple trials, a dedicated workspace per trial keeps the knowledge bases clean.) For a broader view of how Chatref fits a research organization, see the Clinical Trial Sites & CROs guide.

1. Upload your trial criteria as a knowledge base

In your Chatref workspace, go to Knowledge Base and add a source. Upload the inclusion/exclusion document you gathered earlier, or point the importer at a page on your site that lists the criteria. Chatref will read and learn the content immediately. If a trial has separate documents for different phases (e.g., a screening script vs. the official protocol), add them all – the agent can cross-reference them.

2. Configure the agent to answer from the criteria

Create a new AI Agent and attach the knowledge base you just built. Write a short, direct system prompt:

You are a pre-screening assistant for the [Trial Name]. Use only the uploaded criteria to determine if a person may qualify. Never guess. If the criteria don’t cover a question, say you’ll connect the person with a coordinator.

This keeps the agent grounded in your documents. You can also set the agent’s tone here – for research sites a calm, professional voice usually works best.

3. Build custom actions to collect structured details

After a visitor describes their condition, the agent should gather the exact data your coordinators need: age range, diagnosis, prior treatments, key lab values, contact details. Use Custom Actions to build small forms right in the chat. For example:

• Action: Capture age – ask the visitor’s date of birth or age range, store the response.
• Action: Confirm diagnosis – ask which condition the visitor was diagnosed with and when.

Each action can store the answer for your review or trigger an external step (like writing to a Google Sheet). Start with 3–4 actions that cover the most frequent criteria – too many steps will frustrate people before they finish.

4. Test before you publish

Open the agent’s Playground and run through a handful of real scenarios: a perfect match, a borderline case, a clear mismatch, and a few “what if I had X instead?” questions. Check that the agent stays within the criteria and doesn’t make claims beyond what’s in your uploaded documents. When a case genuinely falls between the lines, the agent should offer to hand off to a person.

5. Embed the widget on your recruitment pages

Copy the widget snippet from the agent’s settings and add it to your trial landing page or the “Join a Study” section of your site. The widget loads immediately; once it’s live, visitors can start the pre-screening conversation without leaving the page.

Once the widget is live, the three pieces run together without your team pushing anything:

• The knowledge base answers every eligibility question straight from your uploaded criteria. There’s no internet search and no generic health advice – every answer cites the protocol you trained it on.
• The AI agent handles the conversation. It asks follow-ups, explains criteria when a visitor is unsure, and stays in the brand voice you set. When a question goes beyond the protocol (for example, a very rare comorbidity not listed in the exclusion), the agent recognizes the gap and offers to connect the visitor to a coordinator.
• Custom actions capture the structured information coordinators need without a phone call: age, diagnosis, prior treatments, contact details. That data can be forwarded to a trial management spreadsheet or CRM, so your staff begin the next steps with the key details already in hand.

The result: routine eligibility checks resolve right on your website, and coordinators only step in for the nuanced cases that really need a person – no more chasing people by phone to ask the same ten questions.

• Update the knowledge base the moment a protocol amendment posts. An outdated inclusion criterion is the single biggest cause of false positives. Schedule a monthly re-check, and always re-upload after an IRB revision.
• Design actions for the path, not the exception. Ask for age and primary diagnosis first. Deeper lab values or genetic markers can wait until later in the conversation, after the visitor has had a chance to self-identify as a potential match.
• Use the agent playground as a training lab. Every time a coordinator explains a nuanced eligibility call, turn that scenario into a test conversation. If the agent gives the wrong answer, expand the knowledge base with a short clarifying note – the agent will learn from it.
• Let the agent say “I don’t know.” A pre-screening bot that guesses can create serious compliance risk. In the agent prompt, explicitly instruct it to hand off when it can’t find a clear match in the criteria.
• Route handoffs to the right person. If you run several studies, label each agent with the trial name. When a conversation escalates, coordinators see immediately which study it’s about and can pick it up without digging.

What causes eligibility pre screening chatbot problems for Clinical Trial Sites & CROs?

The most common failures come from outdated or vague criteria. If the uploaded document is a year-old protocol that hasn’t been updated for an amendment, the agent will use the wrong rules. Ambiguous phrasing (like “history of significant cardiac disease” without defining “significant”) leaves too much room for the bot to guess. Finally, many teams skip testing with borderline cases – a bot that handles perfect matches but stumbles on a common gray-area scenario will frustrate visitors and coordinators alike.

How do I improve eligibility pre screening chatbot for Clinical Trial Sites & CROs?

Start by auditing a week’s worth of conversations. Look for patterns – which criteria does the bot get wrong, and which questions lead to an unhelpful handoff? Add those exact edge cases to your knowledge base as short explanatory notes. Tighten your custom actions so coordinators receive structured data (age, diagnosis, contact info) instead of a long transcript. Finally, set a recurring 15-minute review where a coordinator runs fresh examples through the playground and tweaks the agent prompt based on what they see. Small, regular tuning keeps a pre-screening bot accurate as your trials evolve.