Wong, Florence

INTRODUCTION: Readmission and death in cirrhosis are common, expensive, and difficult to predict. Our aim was to evaluate the abilities of multiple artificial intelligence (AI) techniques to predict clinical outcomes based on variables collected at admission, during hospitalization, and at discharge. METHODS: We used the multicenter North American Consortium for the Study of End-Stage Liver Disease (NACSELD) cohort of cirrhotic inpatients who are followed up through 90-days postdischarge for readmission and death. We used statistical methods to select variables that are significant for readmission and death and trained 3 AI models, including logistic regression (LR), kernel support vector machine (SVM), and random forest classifiers (RFC), to predict readmission and death. We used the area under the receiver operating characteristic curve (AUC) from 10-fold crossvalidation for evaluation to compare sexes. Data were compared with model for end-stage liver disease (MELD) at discharge. RESULTS: We included 2,170 patients (57 ± 11 years, MELD 18 ± 7, 61% men, 79% White, and 8% Hispanic). The 30-day and 90-day readmission rates were 28% and 47%, respectively, and 13% died at 90 days. Prediction for 30-day readmission resulted in 0.60 AUC for all patients with RFC, 0.57 AUC with LR for women-only subpopulation, and 0.61 AUC with LR for men-only subpopulation. For 90-day readmission, the highest AUC was achieved with kernel SVM and RFC (AUC = 0.62). We observed higher predictive value when training models with only women (AUC = 0.68 LR) vs men (AUC = 0.62 kernel SVM). Prediction for death resulted in 0.67 AUC for all patients, 0.72 for women-only subpopulation, and 0.69 for men-only subpopulation, all with LR. MELD-Na model AUC was similar to those from the AI models. DISCUSSION: Despite using multiple AI techniques, it is difficult to predict 30- and 90-day readmissions and death in cirrhosis. AI model accuracies were equivalent to models generated using only MELD-Na scores. Additional biomarkers are needed to improve our predictive capability (See also the visual abstract at http://links.lww.com/AJG/B710).

Background and Aims Compared to other chronic diseases, patients with chronic liver disease (CLD) have significantly higher inpatient mortality; accurate models to predict inpatient mortality are lacking. Serum lactate (LA) may be elevated in patients with CLD due to both tissue hypoperfusion as well as decreased LA clearance. We hypothesized that a parsimonious model consisting of Model for End‐Stage Liver Disease (MELD) and LA at admission may predict inpatient mortality in patients with CLD. Approach and Results We examined all patients with CLD in two large and diverse health care systems in Texas (North Texas [NTX] and Central Texas [CTX]) between 2010 and 2015. We developed (n = 3,588) and validated (n = 1,804) a model containing MELD and LA measured at the time of hospitalization. We further validated the model in a second cohort of 14 tertiary care hepatology centers that prospectively enrolled nonelective hospitalized patients with cirrhosis (n = 726). MELD‐LA was an excellent predictor of inpatient mortality in development (concordance statistic [C‐statistic] = 0.81, 95% confidence interval [CI] 0.79‐0.82) and both validation cohorts (CTX cohort, C‐statistic = 0.85, 95% CI 0.78‐0.87; multicenter cohort C‐statistic = 0.82, 95% CI 0.74‐0.88). MELD‐LA performed especially well in patients with specific cirrhosis diagnoses (C‐statistic = 0.84, 95% CI 0.81‐0.86) or sepsis (C‐statistic = 0.80, 95% CI 0.78‐0.82). For MELD score 25, inpatient mortality rates were 11.2% (LA = 1 mmol/L), 19.4% (LA = 3 mmol/L), 34.3% (LA = 5 mmol/L), and >50% (LA > 8 mmol/L). A linear increase (P < 0.01) was seen in MELD‐LA and increasing number of organ failures. Overall, use of MELD‐LA improved the risk prediction in 23.5% of patients compared to MELD alone. Conclusions MELD‐LA (bswh.md/meldla) is an early and objective predictor of inpatient mortality and may serve as a model for risk assessment and guide therapeutic options.

The automated low‐flow ascites pump (alfapump) is an implantable device that drains ascites directly into the urinary bladder. We studied its safety (absence of serious complications) and efficacy (decreased large‐volume paracentesis [LVP] requirement and improved quality of life [QoL]) in the management of ascites in a cohort of North American patients with cirrhosis and recurrent ascites ineligible for transjugular intrahepatic portosystemic shunt (TIPS). QoL was measured by the Chronic Liver Disease Questionnaire (CLDQ) and Ascites Questionnaire (Ascites Q). Following alfapump implantation, patients were monitored for ascites control, laboratory abnormalities, QoL, adverse events, and survival at 12 months. A total of 30 patients (60.0 ± 9.9 years; 57% male; Model for End‐Stage Liver Disease score, 11.4 ± 2.7) received an alfapump, mostly by an interventional radiology approach (97%), followed by longterm prophylactic antibiotics. The alfapump removed a mean ascites volume of 230.6 ± 148.9 L/patient at 12 months, dramatically reducing the mean LVP frequency from 2.4 ± 1.4/patient/month before pump implantation to 0.2 ± 0.4/patient/month after pump implantation. All surviving patients had improved QoL (baseline versus 3 months; CLDQ, 3.9 ± 1.21 versus 5.0 ± 1.0; Ascites Q, 51.7 ± 21.9 versus 26.7 ± 18.6; P < 0.001 for both) and a better biochemical index of nutritional status (prealbumin 87.8 ± 37.5 versus 102.9 ± 45.3 mg/L at 3 months; P = 0.04). Bacterial infections (15 events in 13 patients), electrolyte abnormalities (11 events in 6 patients), and renal complications (11 events in 9 patients) were the most common severe adverse events. By 12 months, 4 patients died from complications of cirrhosis. Alfapump insertion may be a definitive treatment for refractory ascites in cirrhosis, especially in patients who are not TIPS candidates.

OBJECTIVES: Nosocomial infections (NIs) can be a major cause of morbidity and mortality in cirrhosis. This study aims to define the determinants of NI development and its impact on 30-day outcomes among hospitalized patients with cirrhosis. METHODS: North American Consortium for the Study of End-Stage Liver Disease enrolled patients with cirrhosis who were admitted nonelectively. Admission variables and 30-day outcomes were compared between patients with and without NI. These were also compared based on whether there was an isolated admission infection, NI, or both. Models were created for NI development using admission variables and for 30-day mortality. RESULTS: The study included 2,864 patients; of which, 15% (n = 436) developed NI. When comparing NI vs no NI, 1,866 patients were found to be infection free, whereas 562 had admission infections only, 228 had only NI, and 208 had both infections. At admission, patients with NI were more likely to be infected and have advanced cirrhosis. NIs were associated with higher rates of acute-on-chronic liver failure, death, and transplant regardless of admission infections. Patients with NI had higher respiratory infection, urinary tract infection, Clostridium difficile infection, fungal infections, and infection with vancomycin-resistant enterococci compared with patients without NI. Risk factors for NIs were admission infections, model for end-stage liver disease (MELD) > 20, systemic inflammatory response syndrome criteria, proton pump inhibitor, rifaximin, and lactulose use, but the regression model (sensitivity, 0.67; specificity, 0.63) was not robust. Age, alcohol etiology, admission MELD score, lactulose use, acute-on-chronic liver failure, acute kidney injury, intensive care unit, and NI increased the risk of death, whereas rifaximin decreased the risk of death. DISCUSSION: NIs are prevalent in hospitalized patients with cirrhosis and are associated with poor outcomes. Although higher MELD scores and systemic inflammatory response syndrome are associated with NI, all hospitalized patients with cirrhosis require vigilance and preventive strategies.

We hypothesize that the prevalence of chronic kidney disease (CKD) among patients with cirrhosis has increased due to the increased prevalence of CKD‐associated comorbidities, such as diabetes. We aimed to assess the characteristics of hospitalized patients with cirrhosis with CKD and its impact on renal and patient outcomes. The North American Consortium for the Study of End‐Stage Liver Disease (NACSELD) prospectively enrolled nonelectively admitted patients with cirrhosis and collected data on demographics, laboratory results, in‐hospital clinical course, and postdischarge 3‐month outcomes. CKD positive (CKD+) patients, defined as having an estimated glomerular filtration rate (eGFR; Modification of Diet in Renal Disease–4 variable formula) of ≤60 mL/minute for >3 months, were compared with chronic kidney disease negative (CKD–) patients for development of organ failures, hospital course, and survival. There were 1099 CKD+ patients (46.8% of 2346 enrolled patients) who had significantly higher serum creatinine (2.21 ± 1.33 versus 0.83 ± 0.21 mg/dL in the CKD– group) on admission, higher prevalence of nonalcoholic steatohepatitis cirrhosis etiology, diabetes, refractory ascites, and hospital admissions in the previous 6 months compared with the CKD– group (all P < 0.001). Propensity matching (n = 922 in each group) by Child‐Pugh scores (9.78 ± 2.05 versus 9.74 ± 2.04, P = 0.70) showed that CKD+ patients had significantly higher rates of superimposed acute kidney injury (AKI; 68% versus 21%; P < 0.001) and eventual need for dialysis (11% versus 2%; P < 0.001) than CKD– patients. CKD+ patients also had more cases of acute‐on‐chronic liver failure as defined by the NACSELD group, which was associated with reduced 30‐ and 90‐day overall survival ( P < 0.001 for both). A 10 mL/minute drop in eGFR was associated with a 13.1% increase in the risk of 30‐day mortality. In conclusion, patients with CKD should be treated as a high‐risk group among hospitalized patients with cirrhosis due to their poor survival, and they should be monitored carefully for the development of superimposed AKI.

Acute‐on‐chronic liver failure (ACLF) characterized with ≥2 extrahepatic organ failures in cirrhosis carries a high mortality. Outcomes of patients listed for liver transplantation (LT) after ACLF and after LT are largely unknown. The North American Consortium for the Study of End‐Stage Liver Disease prospectively enrolled 2793 nonelectively hospitalized patients with cirrhosis; 768 were listed for LT. Within 3 months, 265 (35%) received a LT, 395 remained alive without LT, and 108 died/delisted. Compared with nonlisted patients, those listed were younger and more often had ACLF, acute kidney injury, and a higher admission Model for End‐Stage Liver Disease (MELD) score. ACLF was most common in patients who died/delisted, followed by those alive with and without LT respectively, (30%, 22%, and 7%, respectively; P < 0.001). At LT, median MELD was 27.9% and 70% were inpatients; median time from hospitalization to LT was 26 days. Post‐LT survival at 6 months was unchanged between those with and without ACLF (93% each at 6 months). There was no difference in 3‐ and 6‐month mean post‐LT creatinine in those with and without ACLF, despite those with ACLF having a higher mean pre‐LT creatinine and a higher rate of perioperative dialysis (61%). In conclusion, patients with and without ACLF had similar survival after transplant with excellent renal recovery in both groups.