A confidential client engaged me to explore an AI agent that predicts thyroid surgery complications using LLMs, patient-specific data, surgical trends, and historical outcomes.
A confidential client engaged me to explore an AI agent that predicts thyroid surgery complications using LLMs, patient-specific data, surgical trends, and historical outcomes.
A confidential client engaged me to explore an AI agent that predicts thyroid surgery complications using LLMs, patient-specific data, surgical trends, and historical outcomes.
Overview
Objectives
The app supports preoperative planning by predicting surgical risks and complications, helping teams improve patient safety, decision-making, and surgical outcomes.
PROCESS
✅ I sent an initial questionnaire to the client & stakeholders to gather basic requirements.
✅ Facilitated a design sprint brainstorming with stakeholders to identify objectives, goals, and workflows, especially for sensitive health areas outside of my SME.
✅ Crafted SWOT analysis of competitors.
✅ Flesh out user (JBTD) for surgeons, scrub nurses, and admins.
✅ Rapid wireframing to aid in client discussions. Validating assumptions with clients rapidly via Slack.
☐ Identify required engineering resources and set client expectations on cost and engineering time. (TBD)
☐ Refine wireframes in collaboration with AI engineers. (TD
☐ Test refined wireframes with the beta user group.
Audience
Surgeons
Medical Researchers
Scrub nurses
Surgical Admins
Key skills
Complex data visualization.
AI interaction design.
Product ideation.
User experience (UX) strategies for high-volume AI systems.
Designing information-dense dashboards with a clear hierarchy.
Using visual storytelling to make complex AI outputs actionable.
Implementing trend lines and drill-down capabilities.
Overview
Objectives
The app supports preoperative planning by predicting surgical risks and complications, helping teams improve patient safety, decision-making, and surgical outcomes.
PROCESS
Initial questionnaire to the client & stakeholders to gather basic requirements.
Facilitated a design sprint brainstorming with stakeholders to identify objectives, goals, and workflows, especially for sensitive health areas outside my SME.
Crafted SWOT analysis of competitors.
Rapid wireframing to aid in client discussions. Validating assumptions with clients rapidly via Slack.
Identify required engineering resources and set client expectations on cost and engineering time. (TBD)
Refine wireframes in collaboration with AI engineers. (TBD)
Test refined wireframes with the beta user group.
Audience
Surgeons
Medical Researchers
Scrub nurses
Surgical Admins
Key skills
Complex data visualization.
AI interaction design.
Product ideation.
User experience (UX) strategies for high-volume AI systems.
Designing information-dense dashboards with a clear hierarchy.
Using visual storytelling to make complex AI outputs actionable.
Implementing trend lines and drill-down capabilities.
Overview
Objectives
The app supports preoperative planning by predicting surgical risks and complications, helping teams improve patient safety, decision-making, and surgical outcomes.
Audience
Surgeons
Medical Researchers
Scrub nurses
Surgical Admins
Key skills
Complex data visualization.
AI interaction design.
Product ideation.
User experience (UX) strategies for high-volume AI systems.
Designing information-dense dashboards with a clear hierarchy.
Using visual storytelling to make complex AI outputs actionable.
Implementing trend lines and drill-down capabilities.
Jobs to be done (WIP)
For Surgeons (primary audience Beta)
I want to adjust the surgical approach based on predicted complications.
I want to be able to plan mitigation strategies with the medical team.
I want to evaluate patient recovery trends.
I want to use AI-assisted checklists for precision.
I want to compare past surgeries for continuous improvement.
For Scrub nurses ( secondary audience v.1)
I need to ensure risk-specific precautions are in place.
I want to generate AI-assisted procedural checklists.
I want to see a trend analysis of surgical checklists.
I need to log intraoperative events for AI-driven analysis.
For Surgical Admins (secondary audience v.2)
I need to predict and mitigate supply shortages based on historical data.
I need to ensure adherence to AI-recommended regulations.
I want to automate reports on surgical outcomes.
I need to predict the length of stay of surgical patients.
For Surgeons (primary audience Beta)
I want to adjust the surgical approach based on predicted complications.
I want to be able to plan mitigation strategies with the medical team.
I want to evaluate patient recovery trends.
I want to use AI-assisted checklists for precision.
I want to compare past surgeries for continuous improvement.
For Scrub nurses ( secondary audience v.1)
I need to ensure risk-specific precautions are in place.
I want to generate AI-assisted procedural checklists.
I want to see a trend analysis of surgical checklists.
I need to log intraoperative events for AI-driven analysis.
For Surgical Admins (secondary audience v.2)
I need to predict and mitigate supply shortages based on historical data.
I need to ensure adherence to AI-recommended regulations.
I want to automate reports on surgical outcomes.
I need to predict the length of stay of surgical patients.
AI Technical mapping
This is a rough draft based on my understanding as an AI product designer, it is not complete, 100% correct, it's intended to aid discussions with engineering and as a learning tool for me!
Natural Language Processing (NLP) for Data Extraction
The SCP will process structured (EHRs, medical databases) and unstructured (doctor’s notes, research papers) data.
It will extract key risk factors from clinical notes using NLP techniques.
It will extract insights from surgeon case notes.
It will generate human-readable risk reports for decision support
Predictive Modeling and Risk Assessment
The model will correlate past surgical data with patient vitals, profile, and digital twin technology to estimate the probabilities of complications (e.g., bleeding, voice changes, and hypocalcemia).
It will generate a confidence score (e.g., 92%) to indicate
prediction reliability.The AI will continuously learn from new patient outcomes and feedback loops from surgeons and nurses to refine predictions over time.
Personalized Recommendations (TDB)
The SCP will suggest preoperative optimizations, such as alternative approaches for high-risk patients.
It can generate adaptive surgical plans based on risk assessment.
It will offer contextual explanations of predictions to aid decision-making.
AI-Assisted Trend Analysis
The model will analyze global surgical trends (e.g., average procedure times and complication trends) to offer benchmarking insights.
It can identify shifts in surgical best practices.
The AI can analyze trends in specific surgeon's cases.
The trend analysis can be trained by manual input of data and continual updating of trend indicators.
Jobs to be done
For Surgeons (primary audience Beta)
I want to adjust the surgical approach based on predicted complications.
I want to be able to plan mitigation strategies with the medical team.
I want to evaluate patient recovery trends.
I want to use AI-assisted checklists for precision.
I want to compare past surgeries for continuous improvement.
For Scrub nurses ( secondary audience v.1)
I need to ensure risk-specific precautions are in place.
I want to generate AI-assisted procedural checklists.
I want to see a trend analysis of surgical checklists.
I need to log intraoperative events for AI-driven analysis.
For Surgical Admins (secondary audience v.2)
I need to predict and mitigate supply shortages based on historical data.
I need to ensure adherence to AI-recommended regulations.
I want to automate reports on surgical outcomes.
I need to predict the length of stay of surgical patients.
AI Technical mapping
NLP: Extract themes, detect sentiment, and categorize feedback with refinement options.
Sentiment Analysis: Gauge user sentiment to calculate frustration scores.
Topic Modeling: Identify recurring feedback themes and pain points.
Predictive Analytics: Forecast trends like churn and prioritize UX risks.
Machine Learning: Automate behavior analysis with manual refinement.
Recommendation Systems: Deliver personalized UX suggestions.
Anomaly Detection: Spot deviations to flag potential issues.
Clustering & Segmentation: Group users/feedback for targeted improvements.
Technical flow map (coming soon)
Define data flow and decision points, integrating AI workflow automation with manual user intervention to manage, prioritize, and address customer feedback or issues. (coming soon)
Rough wireframe
This wireframe was created as a result of the initial design sprint brainstorm. My goals were to:
Ensure the patient does not get lost in the data.
Ensure the surgeon feels confident in the data, trust the AI predictions.
Make AI driven content transparent and allow for human training of the model.
Ensure predictions are simple and scannable.
Potential challenges & mitigations
Overview
Objectives
The primary goal of this app is to enhance preoperative planning by predicting potential complications associated with surgeries. The model aims to improve patient safety, streamline decision-making, and optimize surgical success rates by analyzing data from vast surgical outcomes and risk factors using multiple data sources.
PROCESS
✅ I sent an initial questionnaire to the client & stakeholders to gather basic requirements.
✅ Facilitated a design sprint brainstorming with stakeholders to identify objectives, goals, and workflows, especially for sensitive health areas outside of my SME.
✅ Crafted SWOT analysis of competitors.
✅ Flesh out user (JBTD) for surgeons, scrub nurses, and admins.
✅ Rapid wireframing to aid in client discussions. Validating assumptions with clients rapidly via Slack.
☐ Identify required engineering resources and set client expectations on cost and engineering time. (TBD)
☐ Refine wireframes in collaboration with AI engineers. (TD
☐ Test refined wireframes with the beta user group.
Audience
Surgeons
Medical Researchers
Scrub nurses
Surgical Admins
Key skills
Complex data visualization.
AI interaction design.
Product ideation.
User experience (UX) strategies for high-volume AI systems.
Designing information-dense dashboards with a clear hierarchy.
Using visual storytelling to make complex AI outputs actionable.
Implementing trend lines and drill-down capabilities.