Smart Analysis and Decision-Making for Emerging Infectious Diseases

Official title

AI-enabled Emergency Clinical Research for Emerging and Re-emerging Infectious Diseases: Protocol for an International Consensus

Overview

Emerging infectious diseases, such as COVID-19, mpox, and dengue fever, are characterized by rapid transmission, wide impact, and high uncertainty, posing ongoing threats to global public health. While China achieved significant success in COVID-19 control, the response also revealed key challenges, including fragmented information, delayed risk perception, experience-dependent assessment, and inefficiencies in complex decision-making. This study aims to establish a smart technology system covering the full chain of "risk perception-situational assessment-intelligent decision-making-comprehensive evaluation." Specific objectives include: Constructing a global disease burden database and knowledge graph for emerging infectious diseases; Developing early risk assessment models covering the full transmission spectrum (cross-species, imported, and local outbreak); Building an AI-driven collective intelligence decision-support tool for epidemic control; Developing precise intervention frameworks and comprehensive evaluation indicators for key populations (e.g., elderly, students); Integrating the above technologies into a multi-agent toolkit and evaluating its effectiveness through a cluster randomized controlled trial across 52 CDC sites in five provinces (Guangdong, Zhejiang, Hubei, Sichuan, and Shanghai). The study population includes public health professionals and managers responsible for epidemic surveillance, risk assessment, decision-making, and emergency response at the city/district/county CDC levels across the five provinces. Approximately 780 participants will be enrolled. The intervention group will use the smart toolkit alongside routine practices, while the control group will follow routine practices only. The primary outcome is response time for epidemic assessment and decision-making (hours from risk perception to decision completion). Secondary outcomes include epidemic control effectiveness, user satisfaction, and socioeconomic benefits. The intervention period is 3 months, starting around July 2026 and ending in December 2027. This study has been approved by the Peking University Biomedical Ethics Committee. The study does not involve individual patient data; all data are aggregated at the district/county level from CDC sources or publicly available data. Anonymous questionnaires do not collect any personal identifiable information.

Detailed description

This is a multicenter, cluster-randomized controlled trial (cRCT) with a single-blind design (blinding of statisticians). The study will be conducted across five provinces/municipalities: Guangdong, Zhejiang, Hubei, Sichuan, and Shanghai. A total of 52 district/county/city-level Centers for Disease Control and Prevention (CDCs) will be selected as study clusters and randomized 1:1 to either the intervention group (26 clusters) or the control group (26 clusters).

Randomization Procedure: For the four provinces (Zhejiang, Guangdong, Hubei, Sichuan), CDC clusters will be stratified by socioeconomic level (high, medium, low), with 2 prefecture-level CDCs randomly selected from each stratum and allocated to intervention or control. For Shanghai municipality, CDCs will be stratified by urban functional zone (central urban vs. new/suburban districts), with 2 district-level CDCs selected from each stratum and randomly allocated.

Intervention: The intervention group will use a multi-agent integrated toolkit (including data-knowledge agent, assessment agent, decision agent, and evaluation agent) to assist with epidemic risk perception, situational assessment, and emergency decision-making, in addition to routine practices. The control group will follow routine practices only.

Follow-up Plan: The intervention period is 3 months, timed to coincide with peak seasons for specific infectious diseases (winter/spring for respiratory infections; summer/autumn for vector-borne diseases like dengue). Follow-up assessments will occur every 3 months, with the endpoint defined as the conclusion of an emerging infectious disease event.

Sample Size: Using PASS software (α=0.05, Power=80%, ICC=0.05, CV=0.5, average cluster size m=15), assuming an 80% improvement in decision-making efficiency in the intervention group (response time reduced from 24 to approximately 19 hours), with a 10% attrition rate, a minimum of 28 clusters is required. This study will enroll 52 clusters (approximately 780 participants), exceeding the minimum requirement.

Data Management: Dual independent data entry will be performed. Data will be stored on Peking University's encrypted servers, with backups on the university cloud platform and offline encrypted hard drives (AES-256 encryption). All data will be physically destroyed after the retention period.

Missing Data: Analysis will follow the intention-to-treat (ITT) principle. Missing primary outcome data will be handled using the last observation carried forward (LOCF) method.

Safety Evaluation: Adverse events include headache and absenteeism, classified using a five-level attribution scale (definitely, probably, possibly, probably not, definitely not related), with the first three categories counted as adverse reaction rates. Any serious adverse event must be reported immediately to the sponsor and/or ethics committee.

Early Termination: The study may be terminated early under the following conditions: (1) identification of serious safety issues; (2) the toolkit proves ineffective or futile; (3) major protocol flaws or implementation deviations; (4) request by the applicant or administrative authority.

Interventions

• Other Multi-Agent Integrated Smart Toolkit for Emerging Infectious Diseases
  The multi-agent integrated smart toolkit consists of four integrated agents: (1) Data-Knowledge Agent - for early risk perception based on historical event experience; (2) Assessment Agent - for risk assessment and situational analysis; (3) Decision Agent - for emergency decision support; and (4) Evaluation Agent - for effect simulation and comprehensive evaluation. The toolkit is designed to assist CDC staff with epidemic risk perception, situational assessment, and emergency decision-making. I
• Other Routine Practices
  Routine infectious disease prevention and control practices currently implemented at the CDC, including standard epidemic surveillance, information collection, risk assessment, and emergency response procedures.

Primary outcome measures

• Response Time for Risk Assessment Report Generation and Submission [Time frame: Measured at baseline (enrollment) and at the end of the 3-month intervention period]

Secondary outcome measures (9)

• Consistency of Risk Assessment Results between Multi-Agent Toolkit and Expert Panel [Time frame: Assessed at the end of the 3-month intervention period]
• Epidemic Control Effectiveness [Time frame: Assessed continuously throughout the 3-month intervention period and summarized at the end of the intervention]
• User Experience and Satisfaction with the Smart Toolkit [Time frame: Measured at the end of the 3-month intervention period]
• Healthcare Resource Consumption [Time frame: Assessed at the end of the 3-month intervention period]
• Prevention and Control Resource Inputs [Time frame: Assessed at the end of the 3-month intervention period]
• Reduction in Hospitalization Burden [Time frame: Assessed at the end of the 3-month intervention period]
• Reduction in Severe Disease Burden [Time frame: Assessed at the end of the 3-month intervention period]
• Cost-Effectiveness Ratio [Time frame: Assessed at the end of the 3-month intervention period]
• Macroeconomic Impact [Time frame: Assessed at the end of the 3-month intervention period]

Eligibility criteria

Inclusion criteria

• Working at a district/county/city-level CDC in one of the five participating provinces/municipalities (Zhejiang, Guangdong, Hubei, Sichuan, or Shanghai) where at least one emerging infectious disease (COVID-19, mpox, influenza, dengue, chikungunya, or avian influenza) has occurred.
• Currently responsible for or involved in infectious disease epidemic prevention and control work, including information collection, risk perception, risk assessment, decision-making, risk management, and emergency response at the CDC.
• Willing to voluntarily participate in this study and provide written informed consent.

Exclusion criteria

• Under 18 years of age.
• Diagnosed with severe mental illness or other conditions that impede normal communication.
• Employed in the current CDC position for less than 1 year.

Criteria are shown verbatim from the registry (in English). Final eligibility is always assessed by the study center.

Healthy volunteers: No

Study design

Allocation

Randomized

Model

Parallel assignment

Masking

Single blind

Primary purpose

Prevention

Study locations

China · 1 center

• Jue Liu — Beijing

Identifiers

NCT: NCT07678619 · PKUINF-2026

Primary sources (government registries)

• ClinicalTrials.gov (US NIH)

View this study on ClinicalTrials.gov ↗