Cognitive Impairment Predicts Mortality And Longer Admissions For Index Presentations Of Alcohol-Related Liver Disease

Introduction Alcohol-related liver disease (ARLD) can present with neuropsychiatric complications. Epidemiological studies have not investigated the impact of confusional states on patient journeys. As part of the Connected Health Cities programme, we retrospectively investigated coding data from a regional administrative dataset to identify the burden of recognised cognitive impairment (CI) in a pre-identified cohort of patients during their index admission of ARLD (fiscal years 2014–18). Methodology Inpatient spells for a cohort of 3,887 index ARLD admissions were screened for ICD-10 codes indicative of acute and chronic CI. Descriptive analytics and stepwise multivariate logistic regression models were generated using Stata 15 (StataCorp, 2017) for predefined outcomes: inpatient mortality, length of stay, all-cause outpatient attendance and all-cause A&E attendance. These were casemix-adjusted for age, sex, co-morbidity, deprivation and variables associated with severity of liver decompensation. Results 20 codes corresponded to acute or chronic CI, most frequently encoding encephalopathy and alcohol-related amnesiac syndrome respectively. Codes for intoxication and withdrawal were excluded and adjusted for. 277 spells (7.1%) were coded with > 1 ICD code for acute CI, and 78 (2.0%) for chronic CI, with minimal overlap (0.6%). Comparisons were made with patients without relevant codes. Overall, these patients were older (mean age 57.5 and 64.1 respectively) with higher levels of co-morbidity (mean Charlson index 16.4 and 12.6 respectively) with median bed-days of 13 and 14. Multivariate logistic regression models demonstrated patients with acute CI had higher odds of inpatient mortality (OR 2.13) and long admission >21 days (OR 2.30). Patients coded with chronic CI had higher odds of long admission (OR 3.09) and A&E attendance within 90 days (OR 2.01), and were less likely to attend an outpatient clinic at 14-day (OR 0.28) and 30-day (OR 0.4) intervals post-discharge. CI did not predict risk of readmission. Conclusion In terms of mortality, acute CI is likely to reflect higher disease severity, particularly as encephalopathy is a poor prognosticator in this group of patients. The evidence suggests that chronic CI in patients with ARLD is under-detected in clinical practice. Our analysis demonstrates patients with recognised chronic CI experience extended admissions, are more likely to experience unplanned care and less likely to engage with outpatient care up to 90 days post-discharge. This suggests that current care models may not be appropriate for this subgroup, and alternative pathways integrating prompt identification and supported discharge mechanisms should be developed.