Shaw, Jawaid

INTRODUCTION: Nosocomial infections (NIs) in cirrhosis are associated with high mortality but could be preventable. Logistic regression (LR) models have failed to identify high-risk patients. We aimed to develop machine learning (ML) models to predict NI. METHODS: The CLEARED consortium consists of prospectively enrolled cirrhosis inpatients from >120 centers. Using day-of-admission clinical data, 3 ML approaches (random forest [RF], extreme gradient boosting, and neural networks [NNs]) were used to predict NI. Data were split 80:20 for training and testing stratified by the outcome. Models were compared using area under the receiver operating characteristic curve (AUC). RESULTS: In total, 8,263 patients (55.90 ± 13.34 years; 64.1% men) from 127 centers in 37 countries were included. NI developed in 869 (10.5%), a median of 6 (4–11) days of postadmission. Major NIs were respiratory (29.6%) and urinary tract infection (15.7%), and spontaneous bacterial peritonitis (13.5%). NIs occurred more frequently in patients from low/low-middle income countries and those with severe liver disease, alcohol etiology, and admission infections. NIs were associated with inpatient mortality (31.9% vs 8.1%, P < 0.001) and liver transplantation (4.8% vs 1.9%, P < 0.001). Although the RF model (AUC 0.69) showed good calibration (Brier score 0.09), outperforming extreme gradient boosting, neural network, and LR models (AUC 0.66 for all; LR comparison P = 0.043), no model achieved AUC ≥0.80 for clinical utility. At 10% predicted probability threshold, the RF model demonstrated only 75.4% sensitivity, 52.9% specificity, and 15.9% positive predictive value (PPV). DISCUSSION: NIs cannot be accurately predicted from day-of-admission data using ML models, even in a large, prospective, global cirrhosis cohort. Every hospitalized patient with cirrhosis should receive protocolized infection control measures.

Background and Aims: Prognosticating outcomes such as hospitalizations in outpatients with cirrhosis is challenging, especially with changing etiologies and demographics. This study aims to determine the impact of multi-omic strategies on outcome prediction. Approach and Results: NACSELD3 enrolls outpatients with cirrhosis with controlled/eradicated etiologies from 10 centers and follows them systematically. At baseline, clinical/demographic and cirrhosis details were recorded and saliva and serum samples were collected for microbiome and metabolome analysis, respectively. Multi-omic bioinformatic studies to determine the interaction of microbiota and metabolites with the clinical prediction of 6-month hospitalizations were performed. Five hundred sixty-five patients (60.2 y, 68% men, 35% alcohol, 33% metabolic dysfunction-associated steatohepatitis, 21%, eradicated HCV with MELD 3.0 12) were enrolled. One hundred sixty-three (29%) required 6-month hospitalizations; most (75%) were liver-related. Those hospitalized had worse cirrhosis severity and comorbidity indices but similar demographics and oral health variables. Salivary microbiome alpha-diversity was lower (1.96±0.48 vs. 2.09±0.45, p =0.018) with greater pathobionts ( Streptococcus , Treponema, Enterococcaceae) and lower commensal genhospitalized/noera ( Veillonella, Prevotella, Haemophilus , Lachnospiraceae spp) at baseline. Serum metabolomics showed significant separation at baseline between hospitalized/non-hospitalized patients using supervised analyses with microbial-origin (phenyllactate, secondary bile acids, indoles), choline moieties, and polyamine/GABA (3-ureidopropionate/spermidine) metabolites being most prominent. Area under the curve using random forest for clinical, microbial, and metabolomic variables was higher than that of these individually. Latent factor analysis showed clinical variables (MELD 3.0, hemoglobin, and albumin) with the greatest impact, followed by salivary microbiota and then serum microbiome for hospitalization prediction. Conclusions: In a multicenter North American outpatient cirrhosis cohort with controlled etiologies, serum metabolomics and salivary microbiome add to clinical variables to prognosticate 6-month hospitalization.

Background and Aims: Reducing avoidable readmissions in cirrhosis is challenging. Enhanced engagement using health information technology (HIT) interventions and caregivers lowered readmissions in an open-label study of the Patient Buddy App (PBA). Aim : Multicenter trial of PBA versus standard of care (SOC) to reduce avoidable readmissions. Approach and Results: An open-label, randomized clinical trial was performed at 3 sites to study the effect of PBA (HIT) versus SOC in cirrhosis inpatients with adult caregivers (dyads). Initial randomization was 1:1:1 between SOC, HIT only, and HIT+ visits. However, due to COVID-19, an unplanned study redesign required a combined HIT versus SOC. Primary outcome : Avoidable readmissions (decided by a blinded monitoring board). Secondary outcomes were all-cause readmission and stakeholder input. PBA focused on medication adherence, cognitive testing, and symptoms, and was remotely monitored by study staff. In all, 464 subjects (232 dyads) were enrolled [Virginia Commonwealth University (VCU): 120, Mayo: 40, Department of Veterans Affairs (VA): 72; 116 dyads/group]. Avoidable readmissions were significantly higher in SOC versus HIT (19.8% vs. 10.3%, p =0.04) with OR of 2.14 (95% CI 1.01–4.54) and remained significant even after removing pre-COVID HIT+ visits patients (19.8% vs. 9.3%, p =0.040) with OR of 2.41 (95% CI 1.02–5.69). All-cause readmissions were higher in SOC versus HIT (48% vs. 30%, p =0.005). App evaluation/engagemen t : 1660 alerts were sent; mostly related to HE. Most dyads were satisfied with the app. Conclusions: In a multicenter randomized clinical trial of 464 cirrhosis inpatients and their CGs across several practice settings, the PBA was associated with lower avoidable readmissions at 30 days post-discharge compared to SOC.

INTRODUCTION: Hospitalized patients with cirrhosis can develop respiratory failure (RF), which is associated with a poor prognosis, but predisposing factors are unclear. METHODS: We prospectively enrolled a multicenter North American cirrhosis inpatient cohort and collected admission and in-hospital data (grading per European Association for the Study of Liver-Chronic Liver Failure scoring system, acute kidney injury [AKI], infections [admission/nosocomial], and albumin use) in an era when terlipressin was not available in North America. Multivariable regression to predict RF was performed using only admission day and in-hospital events occurring before RF. RESULTS: A total of 511 patients from 14 sites (median age 57 years, admission model for end-stage liver disease [MELD]-Na 23) were enrolled: RF developed in 15%; AKI occurred in 24%; and 11% developed nosocomial infections (NI). At admission, patients who developed RF had higher MELD-Na, gastrointestinal (GI) bleeding/AKI-related admission, and prior infections/ascites. During hospitalization, RF developers had higher NI (especially respiratory), albumin use, and other organ failures. RF was higher in patients receiving albumin (83% vs 59%, P < 0.0001) with increasing doses (269.5 ± 210.5 vs 208.6 ± 186.1 g, P = 0.01) regardless of indication. Admission for AKI, GI bleeding, and high MELD-Na predicted RF. Using all variables, NI (odds ratio [OR] = 4.02, P = 0.0004), GI bleeding (OR = 3.1, P = 0.002), albumin use (OR = 2.93, P = 0.01), AKI (OR = 3.26, P = 0.008), and circulatory failure (OR = 3.73, P = 0.002) were associated with RF risk. DISCUSSION: In a multicenter inpatient cirrhosis study of patients not exposed to terlipressin, 15% of patients developed RF. RF risk was highest in those admitted with AKI, those who had GI bleeding on admission, and those who developed NI and other organ failures or received albumin during their hospital course. Careful volume monitoring and preventing nosocomial respiratory infections and renal or circulatory failures could reduce this risk.

Background: App-based technologies could enhance patient and caregiver communication and provide alerts that potentially reducing readmissions. However, the burden of App alerts needs to be optimized to reduce provider burnout. Aim: The purpose of this study was to determine subjective and objective burden of using the Patient Buddy App, a health information technology (HIT) on providers in a randomized multicenter trial, who completed a semi-quantitative Likert scale survey regarding training procedures, data and privacy concerns, follow-up details, and technical support. This randomized multicenter trial recruits cirrhosis inpatients and their caregivers, and randomizes them into standard-of-care, HIT (communication only via App) and HIT+visits (App+phone calls/visits) for 30 days after discharge. The alerts are monitored by providers through a central iPad. The reason(s) and number of alerts were recorded as the objective burden. A total of 1442 messages were sent as alerts from the 103 dyads (patient + caregiver) (n=206) randomized to HIT arms. The most common messages related to Hepatic Encephalopathy (HE) (high or low bowel movement=50% or orientation tests=37%). Twelve providers completed the surveys reflecting the following themes—92% and 100%, felt adequately trained and confident about educating the patients and caregivers before roll out of App and had no concerns related to data and privacy; 70%, felt that appropriate time was spent on pursuing reason for data not being logged; 60% each, had issues with availability of adequate technical support and connectivity. Conclusion: The Patient Buddy App randomized multicenter trial till date shows an overall favorable rating regarding training procedures/education, privacy concerns, and ease of message follow-up, from providers. However, it is important to gauge and address subjective and objective burdens of monitoring human resources in current and future HIT studies to avoid burnout and to ensure successful study completion.