Overview
Name
Sichuan Provincial Third People's Hospital
Industry
Healthcare & Social Services
Sichuan Provincial Third People's Hospital faced challenges in ensuring consistent quality of internship reports and efficient supervision by teaching physicians, with limited data to support objective evaluation. By partnering with a HAP implementation partner, the hospital leveraged HAP to integrate large language models into its report review process. This AI-driven solution enabled automated risk screening, issue alerts, and structured analytics. As a result, report accuracy rates rose to 96%, review efficiency increased by 70%, and medical risks were significantly reduced, accelerating the development of intern competencies.
Challenge
Interns write numerous diagnostic reports daily—highly subjective, experience-dependent documents where inconsistent descriptions, inaccurate conclusions, or omitted critical details risk patient safety. Supervising physicians manually review each report, facing fatigue and scalability limits. Training lacks structured analytics on error rates, patterns, and competency progression—hindering targeted feedback.
Difficulty Quantifying Teaching Quality
Training outcomes for resident physicians largely rely on instructors’ subjective judgments, with little structured, visualized data to back them up. This makes it hard to spot common weaknesses and individual gaps, hinders the development of targeted improvement plans, and slows the creation of a closed-loop teaching and continuous improvement mechanism.
Rising medical malpractice risk
Interns must draft numerous reports daily, and inexperience plus fatigue often lead to data entry errors or misjudged diagnoses. Such flawed reports can trigger inappropriate treatments or surgeries, heighten the risk of disputes and patient safety incidents, and undermine hospital reputation and compliance.
Lack of standardization and process controllability
Lack of unified standards and key quality control checkpoints in report drafting and review leads to inconsistent interpretations among interns and supervising physicians. Limited process traceability and consistency hinder issue retrospectives, accountability, and implementation of quality improvements, increasing management complexity.
Low information utilization efficiency
Most reports are unstructured text, making rapid search, comparison, and data analysis difficult. Teaching and quality control cannot efficiently extract key lessons and error patterns from historical cases, so knowledge fails to accumulate. The same issues recur, slowing the learning curve and hindering improvements in care quality.
Excessive mentoring review burden
Supervising physicians must review large volumes of interns’ text reports, which are dense and repetitive, leading to fatigue and reduced attention. Under heavy workloads, the risk of missed findings and misjudgments rises, slowing clinical workflows and undermining department efficiency and teaching quality.
Compliance and cost pressures coexisting
Compounded medical malpractice risks and inefficient reviews can trigger compliance issues, inflate payouts, and raise communication costs. At the same time, repetitive manual reviews and training consume significant staff hours, drive up operating expenses, and limit sustained investment in teaching and care quality.
Solution
The company leverages large language models integrated with existing hospital reporting rules—built on its AI platform—to create an intelligent report review agent. This agent automatically screens intern-submitted diagnostic reports, highlights high-risk cases for attending physicians, and structures error data to inform teaching evaluations and targeted training.
AI Report Review Agent Deployment
Built on Nocoly HAP and hospital report review rules integrated with large language models, the hospital created an AI middle platform with a report-audit agent that automatically delivers concrete edit suggestions and risk alerts. It flags nonstandard wording, contradictory conclusions, and missing key points, helping reduce medical incident risk.
Compliance and traceability management
End-to-end operation logs capture review records, rule versions, and change recommendations, enabling audit traceability and a clear chain of accountability. Role-based access control and data segregation ensure compliant use of physician and patient information.
Report Quality Visualization and Quantitative Evaluation
Powered by a structured data model, HAP tracks doctors’ report error rates in real time, analyzes error type distribution and trends, and aggregates teaching evaluation scores—delivered as dual-layer quality dashboards at individual and department levels. Daily/weekly/monthly reports and leaderboards enable management to drive quality improvement and training reviews.
Automated report review process
Using HAP scheduled jobs to integrate with the hospital report repository, the system automatically retrieves pending cases and invokes the review agent to complete checks. Reports with suggested edits are flagged in orange (default: green) and pushed directly to physicians. Supervising doctors focus on orange reports in a priority queue, significantly improving review efficiency and pass rates.
Rule–Model Co-iteration Mechanism
Configure a visual rules library in HAP to let hospitals maintain discipline-specific review rules and thresholds. Use closed-loop feedback from review outcomes to capture high-frequency error patterns, driving continuous optimization of agent prompts and rule upgrades, ensuring steady gains in accuracy and coverage.
Workstation Optimization for Physicians
Integrate review results and revision recommendations into the physician client to reduce back-and-forth, including issue localization, suggested text, and links to rule references. Use an orange status with a checklist view plus filtering and sorting to quickly surface problematic reports by department, report type, or time batch, shortening average review time.
Core Value
Getting Intern Doctors Up to Speed:Standardized mentoring workflows and templates accelerate intern radiologists’ mastery of report writing and review—cutting training time.
Report Review Accuracy 96%:Established a multi-dimensional proofreading mechanism and quality control rules, ensuring stable report review accuracy at 96%.
Reducing medical malpractice risk:Standardized reporting review and process control significantly reduce diagnostic and treatment errors, continuously lowering medical incident risk.
Visualizing Doctor Report-Writing Statistics:Provides a multi-dimensional visualization dashboard that displays core metrics, including report volume, duration, and accuracy, in real time to enable fine-grained departmental management.
Improving physician productivity:Streamlined report authoring and review processes, cutting report turnaround time and significantly boosting overall efficiency.
Precepting Physician Review Efficiency Up 70%:Streamlined mentoring review with AI-powered auto-prompts boosts mentor physicians’ review efficiency by 70%, enabling sharper academic focus.
