S1224 Computable Phenotypes for NAFLD and NASH Identify Patients with Significant Comorbidities yet Most Remain Undiagnosed

Abstract

Introduction: Non-alcoholic fatty liver disease (NAFLD) is the most common cause of chronic liver disease, yet remains underdiagnosed in real-world practice. Several clinical tools exist to identify possible NAFLD and nonalcoholic steatohepatitis (NASH). Incorporating these into computable phenotypes may help identify patients using electronic health records (EHR), improve clinical care and long-term outcomes. Our objective was to create and evaluate computable phenotypes for NAFLD and NASH using previously validated scores. Methods: We conducted a retrospective study of adults using the Duke University Health System (DUHS) between 2014 and 2019. Patients (pts) with a body mass index < 25 kg/m2 and liver disease diagnosis other than NAFLD or NASH were excluded. The first cohort (NAFLD1) included pts with alanine aminotransferase (ALT) > 40 for men and > 31 for women on 2 occasions six months apart within 2 years. The second NAFLD cohort (NAFLD2) included patients with an ICD9/10 code for NAFLD (K76.0). The first cohort of NASH (NASH1) pts included those that met NAFLD1 criteria plus a FIBROSIS-4 score > 2.67. The second NASH (NASH2) cohort included pts with a NASH diagnosis (K75.81). Chi squared and post hoc Z tests were used to compare cohorts. Results: Out of 359,643 pts who met inclusion criteria, 33,971 pts met criteria for NAFLD1, and 12,544 had a NAFLD diagnosis (NAFLD2). 7,888 met criteria for NASH1 and 2,648 had a NASH diagnosis (NASH2). For both NAFLD and NASH, clinical comorbidities (Table) were significantly more prevalent in NAFLD1 and NASH1 cohorts. White pts were more likely to be diagnosed with NASH than those identified by FIBROSIS-4 score alone (82% vs 69%, P < 0.005). Fewer black pts were diagnosed with NASH (7.8% vs. 24.8%, P < 0.005). Fewer NAFLD2 or NASH2 pts were treated with statins. All cause hospitalizations were high regardless of diagnosis (Figure 1). Conclusion: Very few pts were diagnosed with NAFLD (3.5%) or NASH (0.7%) in this 5 years period. The number of pts meeting criteria by computable phenotype was higher suggesting significant underdiagnosis, particularly among black pts. Our results suggest that pts with NAFLD and NASH phenotypes but no diagnosis are more likely to have downstream comorbidities. Metabolic syndrome features like hypertension and hyperlipidemia should trigger suspicion for NAFLD and NASH in overweight pts. Further studies are needed to understand the reasons behind the underdiagnosis and the increased downstream comorbidities.Figure 1.: Best Practice Advisory (BPA) alerting provider of the need for screening ultrasoundTable 1.: Comparison of Demographics, Comorbidities, and Medications of Different Phenotypes of NAFLD and NASH. T2DM = Type 2 Diabetes Mellitus. CAD = Coronary artery disease. PAD = peripheral artery disease. HFrEF = Heart failure with reduced ejection fraciton. HFpEF = Heart failure with preserved ejection fraction. ACEi = ACE inhibitor. ARB = angiotensin receptor blocker. BB = beta blocker.