# Identifying patients at high risk of decompensated liver disease through unscheduled care attendance data: a retrospective cohort study

## TL;DR

This study identifies patients at high risk of severe liver disease using data from unscheduled healthcare visits, aiming to detect undiagnosed liver issues early.

## Contribution

A novel predictive model using simple lab and demographic data to identify high-risk patients with undiagnosed liver disease.

## Key findings

- 1,609 out of 173,486 patients had liver-related admissions within five years.
- A predictive model with a Harrell’s C statistic of 0.78 was developed using Fib4 score, deprivation, and sex.

## Abstract

Liver cirrhosis is one of the leading causes of mortality and morbidity in those of working age. Mortality from liver disease in the UK has continued to rise over the past decade. A significant proportion of patients presenting with decompensated liver disease have no prior diagnosis of liver disease despite multiple acute healthcare interactions providing opportunities for detection.

We aimed to characterise patients presenting to unscheduled care with no known liver disease who subsequently had a liver related admission (DLD), and determine if a simple predictive score could identify those at high risk.

All patients attending unscheduled care in our health board between the beginning of 2018 and the end of 2020 were included with clinical follow up until end 2022. Exclusion criteria were known liver disease, early (< 6 months) presentation with DLD or missing key laboratory data. A predictive model was developed based on demographic and laboratory parameters.

Following exclusions, a group of 173,486 patients were included in our analysis, of whom 1,609 (0.9%) went on to have a DLD-related admission in the 5 year-follow up period. A model to predict future admission was developed based on Fib4 score (using the common blood tests Aspartate aminotransferase (AST), Alanine Transaminase (ALT) and platelet count), geographical deprivation decile, and sex. This model had a Harrell’s C statistic of 0.78.

Unscheduled care presentations provide an opportunity to identify those at high risk of advanced liver disease and decompensation. It is likely these patients have undiagnosed liver disease at the time of presentation, and a model using simple laboratory and demographic data may aid detection in this setting of those at risk of future liver-related admission. External validation of this model is required.

The online version contains supplementary material available at 10.1186/s12876-025-04534-2.