Artificial Intelligence-Assisted Care After Hospital Discharge for People With Decompensated Cirrhosis

Participants will receive standard care plus 90 days of AI agent-assisted proactive management after discharge. Participants or authorized caregivers will submit brief daily health reports and may report symptoms as they occur. The service combines AI-supported information structuring, rules-based risk alerts, nurse verification, physician-led clinical decisions, and documentation of actions and follow-up.

A 90-day postdischarge management service involving daily symptom and health-status reporting, AI-supported information structuring and task coordination, rules-based risk alerts, nurse verification, physician-led clinical decisions, and documented follow-up of actions. The service covers volume status and renal function-related symptoms, infection, hepatic encephalopathy, gastrointestinal bleeding, medication adherence, and follow-up visits or referrals. Each routine daily report is expected to take approximately 3-5 minutes. The system does not autonomously diagnose, prescribe, or change medication. Critical symptoms prompt immediate emergency care without waiting for platform review.

Usual care for decompensated cirrhosis, including etiology-specific treatment, discharge instructions, written education on warning symptoms, routine follow-up arrangements, existing outpatient or telephone follow-up at each center, usual clinical contact channels, and emergency care. Existing healthcare services will be maintained in both study groups.

Participants will receive standard care without access to the study AI agent-assisted proactive management service. Existing outpatient care, telephone follow-up, clinical contact channels, and access to emergency care will be maintained. Scheduled study outcome ascertainment and safety follow-up will be performed at the same frequency as in the experimental group.

Usual care for decompensated cirrhosis, including etiology-specific treatment, discharge instructions, written education on warning symptoms, routine follow-up arrangements, existing outpatient or telephone follow-up at each center, usual clinical contact channels, and emergency care. Existing healthcare services will be maintained in both study groups.

Proportion of participants with at least one unplanned emergency department visit, hospital observation stay, acute day-case treatment, or readmission due to acute deterioration of liver disease within 90 days after index-hospital discharge. Contacts arising from the same episode of deterioration count as one composite event. Planned care without acute deterioration and routine consultations do not qualify. Events will be verified from hospital records, authorized external records, and identical follow-up schedules in both groups. Two independent adjudicators blinded to group assignment will review events; a third will resolve disagreements. Death and liver transplantation before a qualifying event will be recorded as competing events. Uncertain event status will not be classified as no event.

Proportion of randomized participants with at least one unplanned inpatient readmission due to acute deterioration of liver disease within 90 days after index-hospital discharge. Readmissions will be verified using local hospital records, authorized external records, and identical follow-up procedures in both groups, with blinded endpoint adjudication. Planned admissions without acute deterioration are excluded. Hospitalizations after liver transplantation are not counted as liver-related target events under this protocol.

Proportion of participants with at least one unplanned emergency department visit, hospital observation stay, acute day-case treatment, or readmission due to acute deterioration of liver disease within 30 days after index-hospital discharge. Contacts arising from the same episode of deterioration count as one composite event. Planned care without acute deterioration and routine consultations do not qualify. Events will be verified from hospital records, authorized external records, and identical follow-up schedules in both groups. Two independent adjudicators blinded to group assignment will review events; a third will resolve disagreements. Death and liver transplantation before a qualifying event will be recorded as competing events. Uncertain event status will not be classified as no event.

Proportion of randomized participants with at least one unplanned emergency department visit attributable to acute deterioration of liver disease. Emergency department records, authorized external records, and follow-up information will be reviewed using the blinded adjudication procedures for the emergency department component of the primary composite outcome. Routine consultations and planned care without acute deterioration do not qualify. Events after liver transplantation are not counted as liver-related target events.

Proportion of randomized participants with at least one qualifying unplanned hospital observation stay attributable to acute deterioration of liver disease. Observation records and related clinical documentation will be reviewed to confirm occurrence, clinical indication, and date, followed by blinded endpoint adjudication. Planned observation without acute deterioration does not qualify. Events after liver transplantation are not counted as liver-related target events.

Proportion of randomized participants receiving at least one episode of unplanned acute day-case treatment required for acute deterioration of liver disease. A documented acute medical indication and treatment record are required, with blinded endpoint adjudication. Routine consultations and scheduled day-case treatment without acute deterioration do not qualify. Events after liver transplantation are not counted as liver-related target events.

Time in days from actual discharge to the first unplanned emergency department visit, hospital observation stay, acute day-case treatment, or readmission due to acute deterioration of liver disease, as confirmed by blinded adjudication. Contacts arising from the same episode of deterioration count as one event. Planned care without acute deterioration and routine consultations do not qualify. Death and liver transplantation before a qualifying event are competing events. Participants without a target or competing event are followed through Day 90; participants lost to follow-up are censored at the last time their outcome status was confirmed.

Proportion of randomized participants who die from any cause during the 90-day postdischarge observation period. Vital status and date of death will be verified using medical records, available death documentation, and authorized follow-up information. Deaths after liver transplantation remain included in this all-cause outcome.

Proportion of randomized participants who undergo liver transplantation during the 90-day postdischarge observation period. Transplantation and its date will be verified from operative or hospitalization records, including authorized records from other hospitals.

Number of days alive and out of hospital during the 90-day postdischarge observation period, calculated as 90 minus inpatient days and days after death within that period. Overlapping days are deducted only once. All-cause inpatient days are counted, including hospitalization for liver transplantation. Days alive and out of hospital after transplantation can contribute to this outcome. Emergency department visits and day-case treatment without inpatient admission are not deducted as inpatient days. The possible range is 0 to 90 days; higher values indicate more days alive and out of hospital.

Proportion of randomized participants with at least one adverse event of any cause during follow-up. Events will be collected in both randomized groups through study contacts, medical records, and safety reports. Seriousness, expectedness, and relatedness to the study system will be assessed separately. Participants remain in the safety summary regardless of whether they used the study system.

Proportion of randomized participants with at least one serious adverse event of any cause during follow-up. Events will be verified through study contacts, medical records, and safety reports using the protocol's seriousness criteria. Liver disease deterioration meeting serious adverse event criteria will be recorded even if it also qualifies as a clinical outcome event. Planned hospitalization for routine treatment is not automatically classified as a serious adverse event. Relatedness to the study system will be assessed separately.

For participants assigned to proactive management, the percentage of scheduled eligible home-reporting days on which a report containing all required fields is completed. The numerator is the number of valid reporting days and the denominator is the number of scheduled eligible home-reporting days. Hospitalization days, days after death or liver transplantation, formally suspended reporting days, and days after withdrawal are excluded from the denominator. Reporting status will be determined from the reporting records. Actual use and retention among all randomized participants will be described separately.

Proportion of all randomized participants for whom the primary clinical outcome status during the 90-day postdischarge observation period and survival and liver transplantation status can be established. A confirmed death or liver transplantation is not classified as loss to follow-up. Target event status before death or transplantation must also be established; knowing the competing event alone does not resolve missing preceding event information. Ascertainment will use scheduled follow-up contacts and available local and authorized external records. Attendance at a follow-up visit alone does not establish complete outcome ascertainment.