Talk to the Veterans Crisis Line now
U.S. flag
An official website of the United States government

VA Health Systems Research

Go to the VA ORD website
Go to the QUERI website

PPO 15-155 – HSR Study

PPO 15-155
Electronic Medical Record (EMR) data-mining for identifying clinical risk modifiers of drug-induced liver injury
Ayako Suzuki, MD PhD MSc
Durham VA Medical Center, Durham, NC
Durham, NC
Funding Period: July 2016 - June 2017
Drug-related adverse events are increasing as critical public health problem. The reported number of serious or fatal adverse drug events in the US increased nearly three-fold from 1998 to 2005. The VA Adverse Drug Event Reporting System recorded over 16,000 events per quarter in 2010, which was a significant increase from prior years. This may be partly explained by increased polypharmacy among the elderly. In hospital settings, the annual incidence of reported DILI ranges from 5 to 271 per 100,000 patients depending on the studied populations and drugs. Most patients with DILI present with self-limited liver enzyme elevation. Yet, DILI often results in serious clinical outcomes; about 10% of patients who developed clinically significant DILI exhibit acute liver failure, requiring urgent liver transplant or resulting in death. Eliminating the potential "harm" caused by a therapeutic intent must be our priority in patient care. Due to challenges in clinical diagnosis of DILI and its complex disease pathobiology, we are still a long way from actualizing the identification of susceptible individuals for a specific drug. How patients' clinical factors impact DILI occurrence and severity remains largely unknown. The lack of proper data source and methodology to cope with the complexity of DILI significantly limits our ability to systematically explore such clinical risk factors. Thus, currently DILI is not predictable. Knowledge required for drug-specific risk assessment is currently lacking.

To develop a database for future systematic investigations of the clinical risk factors of DILI.

To use a topic modeling procedure to identify associations between amoxicillin/clavulanate (AMX/CLA)-induced liver injury and co-medications taken at the time of the adverse liver event.

At the VA Informatics and Computing Infrastructure, a relational database was developed using clinical and administrative data available from the Corporate Data Warehouse. We extracted medical records of patients who received any of the study drugs (N=124) between 1999 and 2015, which are known to cause hepatotoxicity in humans and possess diverse drug properties. Refining an existing computer algorithm of prescription records, laboratory data, and ICD-9 codes, DILI cases and corresponding medication-matched controls were identified. DILI was defined as significantly elevated liver enzymes during a high-risk period (HRP: first 90 days after drug initiation or within 30 days following drug discontinuation, whichever was shorter), after excluding other causes of acute liver injury. Comprehensive clinical/administrative information of the identified cases and controls were retrieved and accumulated in the database (VHA DILI database). We created a pilot cohort of AMX/CLA exposure (1.8 million exposures between 2004 and 2014 from 1.1 million unique patients) and characterized frequency, phenotypes, and fatal outcomes of AMX/CLA-related DILI. We also identified age-, gender-, and race/ethnicity-disparities in the AMX/CLA-related DILI risk while assessing potential biases caused by missing information. In the cohort, 32.6% of patients received AMX/CLA on more than one occasion. The descriptive analysis was performed using the first exposure of 'liver healthy' population (847,749 exposures/patients) after excluding patients with underlying liver diseases, liver enzyme elevations during pre-exposure 12 months, hepatitis B viral infection, hepatitis C viral infection or HIV infection.
Topic modeling was applied to the VA pilot cohort (1.1 million patients) and an existing clinical data from a DILI registry (888 DILI cases). Mutual information with 95% confidence interval was computed using the topic model outputs for 2, 3, 4, ..6 word combinations to explore multifactorial risk determinants and drug-host interactions in DILI. Different data-mining approaches were also applied to complement our investigation.

In the liver healthy population, frequency of AMX/CLA-related DILI was 0.14%, 3-4 times higher than previous reports (vs. 0.043% and 0.033%). Multiple logistic regression models showed that the likelihood of AMX/CLA-related DILI was increased in men (OR and 95%CI=1.7[1.3, 2.3] vs. women, p=0.0001), White Hispanics (OR and 95%CI=1.4[1.1, 1.8] vs. White non-Hispanics, p=0.015) and Unknown race (OR and 95%CI=1.3[1.1, 1.5] vs. White non-Hispanics, p=0.022). The likelihood of AMX/CLA-related DILI was also increased with advanced age (1.23- to 2.61-fold increase in those over 76 years vs. other age groups). We also assessed lab test availability within the HRP in different age, gender, and race/ethnicity groups. The lab availability varied up to 38% among different age/gender groups, but no apparent difference was observed among race/ethnicity groups. Overall, 28% had hepatocellular injury, 61% cholestatic injury, and 11% exhibited mixed injury. We demonstrated first time that rates of AMX/CLA-related cholestatic injury increased with aging while AMX/CLA-related hepatocellular injury remained stable across different age groups. Serious hepatocellular injury accompanied by total bilirubin elevation affected 0.007%. Among the DILI cases, 4.85% had liver-related death within 6 months and 5.99% within 2 years. These death rates exceed previously reported rates of 2.9% to 3.4% from existing DILI registries.
Data-mining analyses also revealed several drug-host interactions related to chronic DILI and multiple age, gender, and race/ethnicity interactions in AMX/CLA-related DILI and generated hypotheses to be tested in future projects.

Compared with the general public, veterans are burdened with a higher rate of various diseases. Consequently, polypharmacy is a major drug safety concern for the VHA. This pilot study provided unique data resource for future data-mining analyses on drug-specific DILI and an investigational tool to explore multifactorial DILI risks, including drug-drug/drug-host interactions, which enable knowledge discovery that is directly applicable to clinical implementation of drug safety strategies.

External Links for this Project

NIH Reporter

Grant Number: I21HX001865-01A1

Dimensions for VA

Dimensions for VA is a web-based tool available to VA staff that enables detailed searches of published research and research projects.

Learn more about Dimensions for VA.

VA staff not currently on the VA network can access Dimensions by registering for an account using their VA email address.
    Search Dimensions for this project


Journal Articles

  1. Suzuki A, Tillmann H, Williams J, Hauser RG, Frund J, Suzuki M, Prior F, Aithal GP, Lucena MI, Andrade RJ, Tong W, Hunt CM. Assessment of the Frequency, Phenotypes, and Outcomes of Acute Liver Injury Associated with Amoxicillin/Clavulanate in 1.4 Million Patients in the Veterans Health Administration. Drug Safety. 2023 Feb 1; 46(2):129-143. [view]
  2. Tillmann HL, Suzuki A, Barnhart HX, Serrano J, Rockey DC. Tools for causality assessment in drug-induced liver disease. Current Opinion in Gastroenterology. 2019 May 1; 35(3):183-190. [view]
Journal Other

  1. Suzuki A, Austen M, Williams J, Tillmann HL, Hunt CM. Higher rates and significant disparities in the risk and phenotypes of Amoxicillin:clavulanate-related liver injury were observed in the Veterans Health Administration (VHA). [Abstract]. Hepatology. 2017 Jan 1; 66(1):(Suppl): 452A. [view]
  2. Lonardo A, Suzuki A. Nonalcoholic fatty liver disease: does sex matter? [Editorial]. Hepatobiliary surgery and nutrition. 2019 Apr 1; 8(2):164-166. [view]

DRA: Health Systems, Other Conditions
DRE: Epidemiology, Research Infrastructure, TRL - Applied/Translational
Keywords: none
MeSH Terms: none

Questions about the HSR website? Email the Web Team

Any health information on this website is strictly for informational purposes and is not intended as medical advice. It should not be used to diagnose or treat any condition.