This study will collect prospective longitudinal data to characterize rates and identify correlates of a) physical harms due to follow-up tests, b) financial harms, and c) inappropriate screening using electronic medical record data and manual chart review. The study will also use surveys and semi-structured interviews to characterize rates and identify correlates of screening-related psychological harms, e.g. cancer specific worry, situational anxiety, mood disturbances, and decisional regret. Lastly, investigators will create and disseminate a balance sheet of benefits and harms to inform patients, providers, healthcare organizations, payers, and policymakers about the role of hepatocellular carcinoma screening in patients with cirrhosis.

We will prospectively follow the cohort using electronic medical record to document the hepatocellular carcinoma screening process and characterize physical and financial harms related to positive or indeterminate screening results and burden of inappropriate screening. Patients are anticipated to undergo hepatocellular carcinoma screening every 6-12 months, so each patient will have ~4-8 screening encounters over the study duration. We will use manual chart review to determine intent of ultrasound exams (screening vs. diagnostic) and test results. Receipt of follow-up tests after positive or indeterminate screening results will be identified through electronic medical record extraction using Current Procedural Terminology (CPT) codes for CT, MRI, and biopsy.

Physical harms (contrast injury, radiation exposure, and biopsy complications) can result from screening or follow-up testing and extends beyond medical complications to include discomfort. A binary outcome (harm vs. no harm) will be defined for each person and each type of physical harm (contrast injury, radiation exposure, biopsy, and any physical harm). We will report the point estimate and 95% confidence interval for the proportion of patients with each type of harm in each arm, stratified by health system. Using an intention-to-treat principle, we will use Chi squared test to compare the proportion of patients with physical harms between the screening and usual care arms, with a secondary analysis stratified by health system. We will also perform a sensitivity analysis based on test intent, in which we will only include tests (and corresponding harms) performed as a direct result of hepatocellular carcinoma screening.

Financial harms may include anticipated or real costs of hepatocellular carcinoma screening and diagnostic evaluation including, indirect costs such as missed work, and opportunity costs such as distraction from other health-related activities. Financial harms will be summarized for each arm using descriptive analyses as average and range of costs per person. Degree of financial harms will be compared between the hepatocellular carcinoma screening and usual care arms using Student T test, with a secondary analysis stratified by health system. In a secondary analysis, a mixed-effect model approach will be employed to identify patient-, provider- and system-level factors associated with financial harm.

Defined as hepatocellular carcinoma diagnoses that are unlikely to have an effect on mortality, specifically among patients with: 1) significant comorbid conditions or 2) severe liver dysfunction, i.e. Child Pugh C cirrhosis, who are not candidates for liver transplantation, at hepatocellular carcinoma diagnosis. For the primary analysis, Chi squared test will be used to compare the proportion of patients with overdiagnosis between the screening arm and usual care arm, stratified by health system.

As a secondary analysis, we will construct mixed-effect logistic regression models to identify patient-, provider-, and system-level factors associated with physical harms (contrast injury, radiation exposure, biopsy, and any physical harm). Models will include random effects for providers and health systems to account for potential correlation at different levels. Final models will include covariates identified by stepwise variable selection procedure and those considered clinically important a priori (BMI, race, liver dysfunction, and gastroenterology care). Statistical significance will be declared for p<0.05.