Adverse Event Monitoring

Adverse Event (AE) Monitoring is the RoseRx Platform’s built-in pharmacovigilance feature. When enabled, the system uses AI to automatically detect potential adverse drug reactions mentioned in conversations, alert your team, and provide an appropriate response to the user.

Access Adverse Event Monitoring by selecting an agent from the Agents list and clicking the Adverse Events tab.

How it works

When a user describes a symptom, side effect, or negative experience related to a medicine or treatment, the AE monitoring system:

Detect

The AI analyzes each message in real time, looking for language that indicates a potential adverse event — such as descriptions of symptoms, side effects, or negative reactions.

Alert

When a potential adverse event is detected, the system immediately sends notifications to your configured pharmacovigilance contacts via email.

Respond

The agent displays your configured AE response template to the user, providing appropriate guidance and contact information.

Example scenario: A patient mentions “I’ve been experiencing a rash on my arms since I started taking the medication last week.” The system detects this as a potential adverse event, notifies your pharmacovigilance team, and shows the patient your configured response message with reporting contact details.

Enabling monitoring

To activate AE monitoring for your agent:

Navigate to the Adverse Events tab
Toggle the Enable Adverse Event Monitoring switch to on

Once enabled, the remaining configuration options become available.

AE monitoring only applies to conversations that occur after it is enabled. It does not retroactively scan previous conversations.

Detection sensitivity

The detection sensitivity slider controls how aggressively the system identifies potential adverse events. Choose the level that best balances thoroughness with practical workload for your pharmacovigilance team.

Sensitivity	Approximate detection rate	Description
Low	~70%	Catches clearly described adverse events. Minimal false positives. Best when your team has limited capacity for review.
Medium	~85%	Balances detection coverage with manageable alert volume. Suitable for most use cases.
High	~95%	Catches most potential adverse events, including less obviously described ones. May generate some false positives.
Very High	~98%	Maximum coverage. Flags anything that could remotely indicate an adverse event. Expect a higher rate of false positives.

Start with Medium sensitivity and adjust based on your team’s experience. Higher sensitivity catches more genuine events but also increases the number of alerts your team needs to review.

Detection policies

Two policy settings control when a mention becomes an adverse event signal:

Detection Policy

Option	Behavior
Product or report context	A reportable reaction only creates an AE signal when the message also includes product, exposure, or report context. Recommended for patient-facing agents.
Keyword only	Configured reportable terms create AE signals on their own, without extra context.

Educational Inquiry Policy

Controls how the system treats educational or hypothetical side-effect questions (e.g. “What are the common side effects of this medication?”):

Option	Behavior
Answer without AE	Educational questions are answered normally and don’t create AE signals unless the message also reports a real symptom.
Treat as AE signal	Educational side-effect questions can still be flagged for review.

Keyword lists

Tune detection with three keyword lists specific to your product and therapy area:

List	Purpose
Reportable Reactions	Symptoms or reactions that should be considered reportable — include both clinical terms (“hepatotoxicity”) and everyday language patients actually use (“skin turned yellow”)
Product Terms	Product names, aliases, active ingredients, and brand terms that link a reported reaction to your product (up to 50 terms)
Routine Presenting Concerns	Expected intake or cosmetic concerns that should not create AE signals unless the message also reports product-linked harm

You can also define side-effect severity rules — mapping specific symptoms (and their aliases) to a severity level — so detected events arrive pre-classified for your team.

Include both clinical terms and everyday language in Reportable Reactions. Patients rarely use formal medical terminology.

Testing your configuration

The Adverse Events tab includes a built-in preview tool: paste a sample patient message (and optionally prior conversation turns) to see exactly how the detector scores it, which triggers fired, and whether an AE signal would be created. Use it to validate your keyword lists and policies before publishing.

Advanced detection tuning

Thresholds, context weights, and the gated evaluator

For teams that need finer control, the Advanced section exposes:

Detection thresholds — Adjust the score at which the evaluator is triggered and the score at which an AE signal fires.
Context weights — Tune how much product mentions, prior product links, and report language contribute to the detection score.
Gated evaluator — An optional, bounded AI second-pass that reviews borderline cases, with a configurable model tier and maximum score adjustment.

Most teams won’t need to change these — the sensitivity levels above adjust them automatically.

AE response template

Configure the message your agent displays when a potential adverse event is detected. This message replaces the agent’s normal response and provides the user with clear guidance on what to do next.

Maximum length: 2,000 characters

A good AE response template should:

Acknowledge the user’s concern
Explain that the information has been noted for review
Provide clear instructions for reporting the event
Include relevant contact information
Advise the user to seek medical attention if needed

Example template:

Thank you for sharing this information. Your safety is our priority. We have noted your experience and our pharmacovigilance team has been notified. If you are experiencing a medical emergency, please contact your doctor or call emergency services immediately. You can also report this experience directly by contacting us at the details below.

Reporting contact information

Provide the contact details that appear in AE notifications and may be referenced in the response template.

Field	Description
Email	The pharmacovigilance team’s email address
Phone	A phone number for adverse event reporting
Website	A URL for online adverse event reporting (e.g., your company’s AE reporting portal)

Notification recipients

Add the email addresses of team members who should receive alerts when a potential adverse event is detected.

Maximum recipients: 20

All listed recipients receive an email notification for every detected adverse event. Include your pharmacovigilance team, medical affairs contacts, and any other stakeholders required by your organization’s pharmacovigilance procedures.

Adverse event monitoring is an assistive tool — it does not replace your organization’s formal pharmacovigilance processes. Always ensure detected events are reviewed and reported through your established regulatory reporting pathways.

Best practices

Enable monitoring before publishing — Ensure AE monitoring is configured and active before your agent goes live
Craft a thorough response template — The AE response is often the first thing a concerned user sees. Make it reassuring, clear, and actionable
Include multiple contact methods — Provide email, phone, and website options so users can report through their preferred channel
Review alerts promptly — Establish an internal process for triaging and responding to AE notifications within your required timeframes
Refine sensitivity over time — Monitor the volume and quality of alerts and adjust the sensitivity level as you gather real-world data

Getting Started

Agents

Deploy

Testing

Monitor

Sales Engagement

Organization

Adverse Event Monitoring

How it works

Enabling monitoring

Detection sensitivity

Detection policies

Detection Policy

Educational Inquiry Policy

Keyword lists

Testing your configuration

Advanced detection tuning

AE response template

Reporting contact information

Notification recipients

Best practices

​How it works

​Enabling monitoring

​Detection sensitivity

​Detection policies

​Detection Policy

​Educational Inquiry Policy

​Keyword lists

​Testing your configuration

​Advanced detection tuning

​AE response template

​Reporting contact information

​Notification recipients

​Best practices

How it works

Enabling monitoring

Detection sensitivity

Detection policies

Detection Policy

Educational Inquiry Policy

Keyword lists

Testing your configuration

Advanced detection tuning

AE response template

Reporting contact information

Notification recipients

Best practices