Advertisement

Secondary analysis of large databases for hepatology research

  • Philip N. Okafor
    Correspondence
    Corresponding author. Address: Division of Gastroenterology and Hepatology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, United States.
    Affiliations
    Division of Gastroenterology and Hepatology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, United States
    Search for articles by this author
  • Maria Chiejina
    Affiliations
    Department of Internal Medicine, Good Shepard Medical Center, Longview, TX 75601, United States
    Search for articles by this author
  • Nicolo de Pretis
    Affiliations
    Division of Gastroenterology and Gastrointestinal Endoscopy, Department of Medicine, University of Verona, Piazzale L.A. Scuro, 10, 37134 Verona, Italy
    Search for articles by this author
  • Jayant A. Talwalkar
    Affiliations
    Division of Gastroenterology and Hepatology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, United States

    Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery, Mayo Clinic, 200 First Street SW, Rochester, MN 55905, United States
    Search for articles by this author
Published:December 28, 2015DOI:https://doi.org/10.1016/j.jhep.2015.12.019

      Summary

      Secondary analysis of large datasets involves the utilization of existing data that has typically been collected for other purposes to advance scientific knowledge. This is an established methodology applied in health services research with the unique advantage of efficiently identifying relationships between predictor and outcome variables but which has been underutilized for hepatology research. Our review of 1431 abstracts published in the 2013 European Association for the Study of Liver (EASL) abstract book showed that less than 0.5% of published abstracts utilized secondary analysis of large database methodologies.
      This review paper describes existing large datasets that can be exploited for secondary analyses in liver disease research. It also suggests potential questions that could be addressed using these data warehouses and highlights the strengths and limitations of each dataset as described by authors that have previously used them. The overall goal is to bring these datasets to the attention of readers and ultimately encourage the consideration of secondary analysis of large database methodologies for the advancement of hepatology.

      Keywords

      Introduction

      Secondary analysis of large datasets involves the utilization of existing data that has typically been collected for other purposes to advance scientific knowledge. This is an established methodology applied in health services research (HSR) with the unique advantage of efficiently identifying relationships between predictor and outcome variables that could potentially serve as preliminary data for larger studies or provide hypotheses for testing using prospective study designs. Over the last three decades, the “computer revolution” has led to the generation of an increasing amount of healthcare related data mostly in the form of electronic medical records or administrative claims-based records that are stored in large databases or registries. In addition, it is now easier to survey large groups of individuals and store large quantities of data effectively. With the evolving climate in research funding, we believe that large database analyses have a greater role to play in answering pertinent research questions relating to acute and chronic liver diseases (CLD). Apart from United Network for Organ Sharing (UNOS), we posit that existing large databases are underutilized by researchers in hepatology. This is supported with evidence from our review of 1431 abstracts published in the 2013 European Association for the Study of Liver (EASL) abstract book, which showed that less than 0.5% of published abstracts utilized secondary analysis of large database methodologies. We also reviewed 2276 abstracts from the 2013 American Association for the Study of Liver Diseases (AASLD) abstract book, and found that less than 3% of published abstracts employed large database analyses methodologies. For both reviews of EASL and AASLD abstracts, we excluded UNOS abstracts.
      In most cases, these large databases exist in the form of: 1) administrative and claims-based datasets; 2) clinical registries; and 3) surveys. Importantly, because observations in these datasets are occasionally generated from International Classification of Diseases (ICD-9) or Current Procedural Terminology codes, there is room for misclassification, which could significantly bias results, occasionally away from the null, leading to significant errors. To minimize misclassification with claims-based data, one strategy that could be applied involves the use of strict validation strategies in cohort identification, for instance, requiring an ICD-9 code to appear on multiple visits before inclusion. This can be combined with codes of drugs or treatments that are specific to the disease in question making the selection process more rigorous. In addition, when using hospital level datasets, the use of diagnosis related group codes may be more accurate than ICD-9 codes for identifying medical conditions.
      Missing data are another limitation of large datasets. They can be missing completely at random suggesting that a missing data point is unrelated to observed and unobserved data; missing at random, suggesting that a missing data point can be explained by the observed data; or missing not at random, suggesting that the missing data is dependent on unobserved values [
      • Newgard C.D.
      • Lewis R.J.
      Missing data: how to best account for what is not known.
      ]. Approaches to dealing with missing data include case wise deletion where only complete cases are analyzed, use of missing indicator variables, and imputation during the process of data analysis [
      • Greenland S.
      • Finkle W.D.
      A critical look at methods for handling missing covariates in epidemiologic regression analyses.
      ]. Single value imputation methods estimate what each missing value might have been and replace it with a single value in the data set. These methods include mean imputation, last observation carried forward, and random imputation [
      • Newgard C.D.
      • Lewis R.J.
      Missing data: how to best account for what is not known.
      ]. Multiple imputation methods handle missing data better by estimating and replacing missing values many times while case wise deletion or complete case analyses may introduce bias.
      Another major limitation is the difficulty in establishing temporality and causality. Appropriate time must be allocated to assess the database for appropriateness to answer the research question. In most cases, the external validity of results obtained from secondary analysis of large datasets depends on the type of dataset studied. Datasets from the U.S. may have variables that are not directly applicable to patients in other countries, so results from these analyses must be critically appraised before generalization. As with any rigorous scientific study, the need for an a priori hypothesis before statistical analysis of these large datasets cannot be overstated. Some authors have reported that the ease of analyzing these datasets without a hypothesis often leads to their abuse, leading to results that have significant p values because of random effects due to the sheer large number of observations in the datasets [
      • Kaplan B.
      • Schold J.
      • Meier-Kriesche H.U.
      Overview of large database analysis in renal transplantation.
      ], especially as some of these results may have limited clinical significance or applicability.
      However, despite these limitations, compared to prospective methodologies, secondary analysis of large datasets can be performed in relatively short time periods and at relatively lower costs, usually costs associated with obtaining datasets and statistical analyses. Secondary data analysis has been employed in the description of time trends and prevalence rates of various liver diseases, and also health resource utilization. In addition, this methodology has not only been employed in the identification of variations in practice and treatment patterns of liver diseases, but also in the exploration of disparities in access to healthcare resources among patients with liver diseases, as described below. These questions would otherwise be challenging to answer using other study designs. Observations from these studies can also be used as inputs for cost-effectiveness modeling and in comparative effectiveness studies. The aim of this paper is to describe existing large databases that can be used for hepatology research. In the process, we hope to suggest potential questions that could be addressed by prospective investigators using these data warehouses. In addition, we will highlight the strengths and limitations of these datasets as described by authors that have previously used them for research. Our foremost objective is to bring these datasets to the attention of readers and ultimately encourage the consideration of secondary database methodologies in the advancement of hepatology and HSR.

      Why are chronic liver diseases suitable for health services research?

      CLD cause significant morbidity and mortality globally [
      • Byass P.
      The global burden of liver disease: a challenge for methods and for public health.
      ]. Multiple studies have demonstrated that patients with CLD (including chronic hepatitides, alcoholic and non-alcoholic fatty liver disease (NAFLD), primary biliary cirrhosis, primary sclerosing cholangitis and hepatocellular cancer (HCC)) have increased utilization of healthcare resources [
      • Rogal S.S.
      • Winger D.
      • Bielefeldt K.
      • Rollman B.L.
      • Szigethy E.
      Healthcare utilization in chronic liver disease: the importance of pain and prescription opioid use.
      ,
      • Ganesh S.
      • Rogal S.S.
      • Yadav D.
      • Humar A.
      • Behari J.
      Risk factors for frequent readmissions and barriers to transplantation in patients with cirrhosis.
      ]. Chronic hepatitis C virus infection (HCV) remains the leading indication for liver transplant in the U.S. [
      • Kabiri M.
      • Jazwinski A.B.
      • Roberts M.S.
      • Schaefer A.J.
      • Chhatwal J.
      The changing burden of hepatitis C virus infection in the United States: model-based predictions.
      ], soon to be overtaken by NAFLD. While the availability of direct-acting antivirals and implementation of HCV screening is expected to reduce the burden associated with HCV, alcoholic and NAFLD still remain significant public health problems and are increasing in prevalence [
      • Lazo M.
      • Hernaez R.
      • Eberhardt M.S.
      • Bonekamp S.
      • Kamel I.
      • Guallar E.
      • et al.
      Prevalence of nonalcoholic fatty liver disease in the United States: the Third National Health and Nutrition Examination Survey, 1988–1994.
      ]. With the improvement in treatment modalities for CLD, we anticipate increasing healthcare resource utilization (HCRU) by patients, which could further drive healthcare costs. However, studies to identify patient-level predictors, health system level and regional variations in HCRU, and disparities in access to healthcare among patients with liver diseases are few and far between. For this review, we will not discuss the various commercial health insurance databases which exist; instead, we will focus on some of the national databases, registries, and surveys in Europe and the U.S.

      General hepatology outcomes

      General Practice Research Databases (Europe)

      Large general practice research databases are useful for the study of risk factors, treatment patterns, incidence rates, and health resource utilization of various medical conditions [
      • Chen L.
      • Reed C.
      • Happich M.
      • Nyhuis A.
      • Lenox-Smith A.
      Health care resource utilisation in primary care prior to and after a diagnosis of Alzheimer’s disease: a retrospective, matched case-control study in the United Kingdom.
      ,
      • Ratib S.
      • West J.
      • Crooks C.J.
      • Fleming K.M.
      Diagnosis of liver cirrhosis in England, a cohort study, 1998–2009: a comparison with cancer.
      ]. Among the most utilized are the General Practice Research Database (GPRD), The Health Improvement Network (THIN), and the French Longitudinal Patient Database (FLPD). The GPRD was a prospectively maintained primary care database developed in the United Kingdom (UK). It contained records of diagnoses, laboratory tests, consultations and medications from primary care and also information communicated by hospitals in discharge summaries or outpatient letters. It had previously been shown to be a representative 4–6% sample of the UK population and its accuracy and completeness have been validated in several studies. Researchers have used it for pharmacoepidemiological, pharmacoeconomics, and outcomes studies in patients with liver diseases [
      • Seaman H.E.
      • de Vries C.S.
      • Farmer R.D.
      The risk of liver disorders in women prescribed cyproterone acetate in combination with ethinyloestradiol (Dianette): a nested case-control study using the GPRD.
      ,
      • Myers M.W.
      • Jick H.
      Terfenadine and risk of acute liver disease.
      ,
      • Clifford G.M.
      • Logie J.
      • Farmer R.D.
      No risk of drug-associated liver injury with alpha1-adrenoreceptor blocking agents in men with BPH: results from an observational study using the GPRD.
      ]. In 2012, the GPRD was replaced by the Clinical Practice Research Datalink (CPRD) http://www.cprd.com/intro.asp, a new observational data and interventional research service, jointly funded by the NHS, National Institute for Health Research and the Medicines and Healthcare products Regulatory Agency [

      Clinical Practice Research Datalink. Available at: <http://www.cprd.com/intro.asp>. Accessed 7/24/2015.

      ]. The major strengths of the CPRD lie in its coverage of different patient demographics, and also in the prospective nature of data collection that minimizes recall bias. The large size affords the ability to study rare diseases. It has also been used to study a wide variety of liver diseases including drug-induced liver diseases, cirrhosis, and HCC [
      • Ratib S.
      • West J.
      • Crooks C.J.
      • Fleming K.M.
      Diagnosis of liver cirrhosis in England, a cohort study, 1998–2009: a comparison with cancer.
      ,
      • McGlynn K.A.
      • Hagberg K.
      • Chen J.
      • Graubard B.I.
      • London W.T.
      • Jick S.
      • et al.
      Statin use and risk of primary liver cancer in the Clinical Practice Research Datalink.
      ]. Access to the CPRD comes with a price, and data costs are charged at a fixed rate depending on data sources and linkage. Data can be tailored to specific interests of the investigator. The CPRD is comparable to the FLPD, which also provides representative data from patients in France who have been treated in general practitioner facilities [

      IMS Health Strategic Data. Available at: <https://www.sd.imshealth.com/fr/Pages/default.aspx>. Accessed 7/24/2015.

      ,
      • Sabouret P.
      • Bricard M.
      • Hermann M.A.
      • Cotte F.E.
      • Deret-Bixio L.
      • Rushton-Smith S.
      Discrepancy between guidelines for stroke prevention in atrial fibrillation and practice patterns in primary care. The nationwide French AFIGP survey.
      ].
      THIN is a de-identified database of patient information from over 450 general practices in the UK using the INPS vision software (http://www.thin-uk.net). It has data on over 11 million electronic patient records and has been determined to be representative of the UK population in terms of age, gender, medical conditions (including alcohol use) [
      • Keane M.G.
      • Horsfall L.
      • Rait G.
      • Pereira S.P.
      A case-control study comparing the incidence of early symptoms in pancreatic and biliary tract cancer.
      ]. Like the CPRD, it contains data on diagnoses, symptoms, prescriptions, test and results. Secondary data on hospitalizations, outpatient consultations and tests are entered retrospectively. Investigators have used THIN to study the incidence of early symptoms in pancreatic and biliary tract cancers [
      • Keane M.G.
      • Horsfall L.
      • Rait G.
      • Pereira S.P.
      A case-control study comparing the incidence of early symptoms in pancreatic and biliary tract cancer.
      ].

      Limitations

      GPRDs may be limited in the amount of data collected for medical conditions outside of primary care. This was highlighted by Sabouret et al. as a major limitation of the FLPD [
      • Sabouret P.
      • Bricard M.
      • Hermann M.A.
      • Cotte F.E.
      • Deret-Bixio L.
      • Rushton-Smith S.
      Discrepancy between guidelines for stroke prevention in atrial fibrillation and practice patterns in primary care. The nationwide French AFIGP survey.
      ]. However, the GPRD has been used to study outcomes and resource utilization among patients with various gastrointestinal diseases including Barrett’s esophagus, celiac sprue, inflammatory bowel diseases etc. [
      • Alexandropoulou K.
      • van Vlymen J.
      • Reid F.
      • Poullis A.
      • Kang J.Y.
      Temporal trends of Barrett’s oesophagus and gastro-oesophageal reflux and related oesophageal cancer over a 10-year period in England and Wales and associated proton pump inhibitor and H2RA prescriptions: a GPRD study.
      ,
      • Suissa D.
      • Delaney J.A.
      • Dial S.
      • Brassard P.
      Non-steroidal anti-inflammatory drugs and the risk of Clostridium difficile-associated disease.
      ,
      • Wright E.
      • Schofield P.T.
      • Seed P.
      • Molokhia M.
      Bisphosphonates and risk of upper gastrointestinal cancer–a case control study using the General Practice Research Database (GPRD).
      ], while both THIN and CPRD have been used to study hepatobiliary malignancies and cirrhosis [
      • Ratib S.
      • West J.
      • Crooks C.J.
      • Fleming K.M.
      Diagnosis of liver cirrhosis in England, a cohort study, 1998–2009: a comparison with cancer.
      ,
      • McGlynn K.A.
      • Hagberg K.
      • Chen J.
      • Graubard B.I.
      • London W.T.
      • Jick S.
      • et al.
      Statin use and risk of primary liver cancer in the Clinical Practice Research Datalink.
      ]. Further linkage of these general practice repositories to other databases (such as the Hospital Episode Statistics) may provide more opportunities for the study of liver diseases. As with any large database, general practice databases are susceptible to coding errors. This is especially important as inaccuracies in coding may lead to ascertainment and misclassification biases, as highlighted by Ratib et al. [
      • Ratib S.
      • West J.
      • Crooks C.J.
      • Fleming K.M.
      Diagnosis of liver cirrhosis in England, a cohort study, 1998–2009: a comparison with cancer.
      ].

      National Hospital Ambulatory Medical Care Survey Database (United States)

      The National Hospital Ambulatory Medical Care Survey (NHAMCS) is a federally funded survey of the utilization of a nationally representative sample of ambulatory care services (NHAMCS-OPD) and hospital emergency departments (NHAMCS-ED) in the U.S. This is done via a national probability sampling method of geographic areas, hospitals, emergency service areas within the hospitals, and patient visits to the emergency room. It includes data from nonfederal, general and short-stay hospitals (including children’s hospitals) but excludes data from federal, military, and veteran’s administration hospitals. The National Ambulatory Medical Survey was conducted annually from 1973 to 1981, in 1985, and annually since 1989. These surveys are conducted by the National Center for Health Statistics and Centers for Disease Control and Prevention (CDC). NHAMCS is publically available for free via the CDC website http://www.cdc.gov/nchs/ahcd/ahcd_questionnaires.htm [

      National Hospital Ambulatory Medical Care Survey Database. Available at: <http://www.cdc.gov/nchs/ahcd/ahcd_questionnaires.htm>.

      ]. In addition, there is a restricted dataset that contains more data including emergency service area, annual emergency room visit volume, teaching hospital/medical school affiliation, trauma level rating and other geographic variables that would be essential in identifying variations in practice patterns. It also includes data suitable for analyzing trends in emergency room utilization by patients with gastrointestinal diseases including CLD (Table 1).
      Table 1Summary table of large databases for hepatology research.
      NHAMCS has been used to study time trends, disparities in length of stay and wait times for acute gastrointestinal illnesses [
      • Wu B.U.
      • Banks P.A.
      • Conwell D.L.
      Disparities in emergency department wait times for acute gastrointestinal illnesses: results from the National Hospital Ambulatory Medical Care Survey, 1997–2006.
      ], and healthcare utilization among CLD patients including chronic HCV patients [
      • Tsui J.I.
      • Maselli J.
      • Gonzales R.
      Sociodemographic trends in national ambulatory care visits for hepatitis C virus infection.
      ,
      • Cheung R.
      • Mannalithara A.
      • Singh G.
      Utilization and antiviral therapy in patients with chronic hepatitis C: analysis of ambulatory care visits in the US.
      ]. Cheung et al. retrospectively analyzed national trends of ambulatory visits in patients with HCV using NHAMCS between 2000 and 2006 [
      • Cheung R.
      • Mannalithara A.
      • Singh G.
      Utilization and antiviral therapy in patients with chronic hepatitis C: analysis of ambulatory care visits in the US.
      ]. In their study, they showed that 71% of chronic HCV patients utilizing emergency room and ambulatory resources were Caucasian, while 22% were African-Americans. They went on to investigate gender, race and health coverage among patients with HCV infection and showed an increasing trend in HCRU between 2000 and 2003, which eventually leveled off. Utilizing this database, they were also able to elucidate other comorbidities apart from CLD in these patients that accounted for increased HCRU. In a similar study, Tsui et al. also showed an increase in ambulatory care visits in the U.S., with an increasing percentage of visits among non-white patients and Medicaid patients [
      • Tsui J.I.
      • Maselli J.
      • Gonzales R.
      Sociodemographic trends in national ambulatory care visits for hepatitis C virus infection.
      ]. They postulated reasons for this disparity including differential outcomes by race. They also showed that almost half of HCV related visits occurred with a primary care provider supporting the shortage of hepatologists providing care to HCV patients.
      Roberts et al. used the NHAMCS on a broader scale capturing data on various CLD including hepatitis B virus (HBV), HCV, alcoholic and NAFLD, cirrhosis, HCC and other CLD [
      • Roberts H.W.
      • Utuama O.A.
      • Klevens M.
      • Teshale E.
      • Hughes E.
      • Jiles R.
      The contribution of viral hepatitis to the burden of chronic liver disease in the United States.
      ]. They described racial, gender, and insurance coverage trends for these CLD between 2006 and 2010 and showed that about 50% of all CLD ambulatory visits were attributed solely to HBV and HCV infections.

      Limitations

      One major limitation of the NHAMCS survey is that data is collected on a visit basis; as such, estimates of individual persons cannot be obtained [
      • McCaig L.F.
      • Burt C.W.
      Understanding and interpreting the National Hospital Ambulatory Medical Care Survey: key questions and answers.
      ]. Patients may have multiple visits during the reporting period for the same disease and still be captured. As a result, incidence and prevalence rates of diseases cannot be examined. For the same reason, it would not be possible to study rehospitalization rates among patient with CLD, a major driver of healthcare costs. Cheung et al. were unable to determine prescription medication and comorbidity trends over time because of the sampling and weighting methods used to establish national estimates [
      • Cheung R.
      • Mannalithara A.
      • Singh G.
      Utilization and antiviral therapy in patients with chronic hepatitis C: analysis of ambulatory care visits in the US.
      ]. They were also unable to investigate possible covariates such as socioeconomic status and educational level, which potentially could influence HCRU among these patients. In addition, they were unable to study progression of liver disease as it was unlikely that the same patient would be entered at different time points due to the survey design. They concluded that the exclusion of ambulatory visits to federal medical centers from the NAMCS and NHAMCS was likely to underestimate HCRU among patients with HCV infection. As with any other database, another limitation is the potential for misclassification, as visits relating to HCV care might have been improperly captured because the survey only allowed for three diagnosis codes [
      • Tsui J.I.
      • Maselli J.
      • Gonzales R.
      Sociodemographic trends in national ambulatory care visits for hepatitis C virus infection.
      ].

      National Health and Nutrition Examination Survey (United States)

      The National Health and Nutrition Examination Survey (NHANES) is a program of studies designed to assess the health and nutritional status of adults and children in the U.S. related with a public-linked mortality file [

      NHANES. Available at: <http://www.cdc.gov/nchs/nhanes.htm>. Accessed 10/01/2014.

      ]. It was started in the early 1960s and became a continuous program in 1999 http://www.cdc.gov/nchs/nhanes/nhanes_questionnaires.htm. It examines a nationally representative sample of about 5000 persons each year and includes demographic, socioeconomic, dietary, and health-related questions. NHANES has been utilized in the estimation of prevalence rates and determination of risk factors for different diseases.
      NHANES collects data on variables such as alcohol and tobacco use, and laboratory data including serum aminotransferase level, albumin, complete blood count, creatinine, iron, total iron binding capacity and ferritin, hepatitis A, B, C, D, and E [

      NHANES Survey Data. Available at: <http://www.cdc.gov/nchs/data/nhanes/survey_content_99_14.pdf>. Accessed 9/30/2014.

      ]. It has provided data for many landmark epidemiologic studies exploring the prevalence and incidence rates of the viral hepatitides. For instance, Coleman et al. described the incidence of HBV in the U.S. population [
      • Coleman P.J.
      • McQuillan G.M.
      • Moyer L.A.
      • Lambert S.B.
      • Margolis H.S.
      Incidence of hepatitis B virus infection in the United States, 1976–1994: estimates from the National Health and Nutrition Examination Surveys.
      ]. Alter et al. subsequently described HCV prevalence [
      • Alter M.J.
      • Kruszon-Moran D.
      • Nainan O.V.
      • McQuillan G.M.
      • Gao F.
      • Moyer L.A.
      • et al.
      The prevalence of hepatitis C virus infection in the United States, 1988 through 1994.
      ], while Kuniholm et al. calculated estimates of hepatitis E (HEV) [
      • Kuniholm M.H.
      • Purcell R.H.
      • McQuillan G.M.
      • Engle R.E.
      • Wasley A.
      • Nelson K.E.
      Epidemiology of hepatitis E virus in the United States: results from the Third National Health and Nutrition Examination Survey, 1988–1994.
      ]. Investigators have also described time trends in vaccination rates using NHANES [
      • Younossi Z.M.
      • Stepanova M.
      Changes in hepatitis A and B vaccination rates in adult patients with chronic liver diseases and diabetes in the U.S. population.
      ]. Studies such as these afforded identification of the birth cohort effect of HCV among baby boomers that eventually led to cost-effectiveness studies informing changes in health policy [
      • Rein D.B.
      • Smith B.D.
      • Wittenborn J.S.
      • Lesesne S.B.
      • Wagner L.D.
      • Roblin D.W.
      • et al.
      The cost-effectiveness of birth-cohort screening for hepatitis C antibody in U.S. primary care settings.
      ]. NHANES has also provided data used in exploring racial and ethnic differences in CLD [
      • Ioannou G.N.
      • Dominitz J.A.
      • Weiss N.S.
      • Heagerty P.J.
      • Kowdley K.V.
      Racial differences in the relationship between hepatitis C infection and iron stores.
      ,
      • Flores Y.N.
      • Yee Jr., H.F.
      • Leng M.
      • Escarce J.J.
      • Bastani R.
      • Salmeron J.
      • et al.
      Risk factors for chronic liver disease in Blacks, Mexican Americans, and Whites in the United States: results from NHANES IV, 1999–2004.
      ], and has been to establish risk factors and predictors of CLD [
      • Ioannou G.N.
      • Weiss N.S.
      • Boyko E.J.
      • Mozaffarian D.
      • Lee S.P.
      Elevated serum alanine aminotransferase activity and calculated risk of coronary heart disease in the United States.
      ,
      • Ruhl C.E.
      • Menke A.
      • Cowie C.C.
      • Everhart J.E.
      Relationship of hepatitis C virus infection with diabetes in the U.S. population.
      ].

      Limitations

      The cross-sectional nature of NHANES is a major limitation of the database. Even though it is a representative sample of the U.S. population, there may be underrepresentation of certain groups limiting generalizability of results. The cross-sectional design also prevents establishment of causality or temporality. Another limitation is that of selection bias as healthier persons are more likely to participate in the survey, which would lead to an underestimation of true prevalence or incidence rates in many of these studies.

      National Health Interview Survey (United States)

      The National Health Interview Survey (NHIS) is comparable to the NHANES as it is one of the data collection programs of the National Center for Health Statistics (NCHS), which is also part of the Centers for Disease Control and Prevention (CDC) [

      National Health Interview Survey: Available at: <http://www.cdc.gov/nchs/nhis/about_nhis.htm>. Accessed 10/01/2014.

      ]. It has been conducted since 1957 with the main objective of monitoring the health of the U.S. population on a broad range of health topics http://www.cdc.gov/nchs/nhis/nhis_questionnaires.htm. The sample design is a cross-sectional household survey following a multistage area probability design permitting oversampling of certain demographics e.g. African-American, Hispanic, and Asian persons. The rigorous sampling methods and quality controls employed by the NCHS and the complex sample weights allow generalization of data to the U.S population. Most often, the NHIS has been used to estimate vaccine coverage, particularly HBV, across the U.S. [
      Hepatitis B vaccination coverage among adults–United States, 2004.
      ]. Authors have also used this database to identify factors associated that influence vaccination rates. For instance, Jain et al. showed that young age, unmarried status, past blood donation and prior human immunodeficiency virus testing were independently associated with HBV vaccination [
      • Jain N.
      • Yusuf H.
      • Wortley P.M.
      • Euler G.L.
      • Walton S.
      • Stokley S.
      Factors associated with receiving hepatitis B vaccination among high-risk adults in the United States: an analysis of the National Health Interview Survey, 2000.
      ].

      Limitations

      The major limitation of the NHIS is that it is essentially self-reported. As such, findings are not independently verified or confirmed by chart review or laboratory testing, which significantly differentiates it from NHANES. In the study by Jain et al., an emphasis was made on the fact that no prior studies have validated the self-reporting of HBV vaccination [
      • Jain N.
      • Yusuf H.
      • Wortley P.M.
      • Euler G.L.
      • Walton S.
      • Stokley S.
      Factors associated with receiving hepatitis B vaccination among high-risk adults in the United States: an analysis of the National Health Interview Survey, 2000.
      ]. They also suggest that the nature of the survey limits determination of differences among sub-groups of high-risk individuals. In addition, self-reported healthcare resource utilization may not be as accurate as claims-based assessments. However, some authors believe that this limitation is offset by the ability to assess out-of-pocket costs and thus the individual burden of disease [
      • Silverberg J.I.
      Healthcare utilization, patient costs, and access to care in US adults with eczema: a population-based study.
      ]. These limitations may explain the underutilization of the NHIS for the study of liver diseases.

      Medical Expenditure Panel Survey (United States)

      The Medical Expenditure Panel Survey (MEPS) is a nationally representative survey of noninstitutionalized individuals, families and their healthcare providers and employers across the U.S. http://meps.ahrq.gov/mepsweb/about_meps/survey_back.jsp [

      Medical Expenditure Panel Survey. Available at: <http://meps.ahrq.gov/mepsweb/about_meps/survey_back.jsp>. Accessed 3/23/2015.

      ]. It began in 1996 and is conducted annually by the Agency for Healthcare Research and Quality (AHRQ) allowing exploration of changes in time. In summary, MEPS collects data on the costs and utilization of healthcare services by Americans. Available data includes age, gender, race, marital status, employment status, poverty category, education, health status and urban status [
      • Bernard D.S.
      • Farr S.L.
      • Fang Z.
      National estimates of out-of-pocket health care expenditure burdens among nonelderly adults with cancer: 2001 to 2008.
      ]. Its major components include: a) household component which collects data from a sample of families and individuals across the U.S.; and b) insurance component: collects data from a sample of private and public sector employers on the health insurance plans they offer their employees via a survey known as Health Insurance Cost Study. Various authors have used this database to study differences in the prevalence of chronic diseases and explore disparities in access to healthcare [
      • Wang Y.R.
      • Loftus Jr., E.V.
      • Cangemi J.R.
      • Picco M.F.
      Racial/Ethnic and regional differences in the prevalence of inflammatory bowel disease in the United States.
      ,
      • Shi L.
      • Chen C.C.
      • Nie X.
      • Zhu J.
      • Hu R.
      Racial and socioeconomic disparities in access to primary care among people with chronic conditions.
      ]. In addition, MEPS has been used to estimate healthcare expenditures including out-of-pocket burdens for adults with multiple chronic conditions because it includes data on all payers and components of expenditures, unlike claims data which includes only covered services [
      • Bernard D.S.
      • Farr S.L.
      • Fang Z.
      National estimates of out-of-pocket health care expenditure burdens among nonelderly adults with cancer: 2001 to 2008.
      ,
      • Machlin S.R.
      • Soni A.
      Health care expenditures for adults with multiple treated chronic conditions: estimates from the Medical Expenditure Panel Survey, 2009.
      ]. Some of the other major MEPS topic areas include access to care, disabilities, insurance coverage, elderly healthcare etc. Its strengths include the ability to study HCRU across different racial groups on a wide variety of topics such as prescription drug use and preventative services [
      • Holden C.D.
      • Chen J.
      • Dagher R.K.
      Preventive care utilization among the uninsured by race/ethnicity and income.
      ].

      Limitations

      As with any survey, some authors have stated recall and social desirability bias as limitations of the MEPS [
      • Li R.
      • Barker L.E.
      • Shrestha S.
      • Zhang P.
      • Duru O.K.
      • Pearson-Clarke T.
      • et al.
      Changes over time in high out-of-pocket health care burden in U.S. adults with diabetes, 2001–2011.
      ]. Wang et al. report that among Asian-Americans, subgroup analysis is not possible and that some data regarding census region, and specific data such as state, city, and zip code may not be available [
      • Wang J.
      • Mullins C.D.
      • Zuckerman I.H.
      • Walker G.D.
      • Suda K.J.
      • Yang Y.
      • et al.
      Medical Expenditure Panel Survey: a valuable database for studying racial and ethnic disparities in prescription drug use.
      ].

      Behavioral Risk Factor Surveillance System (United States)

      The Behavioral Risk Factor Surveillance System (BRFSS) is the world’s largest telephone based health survey http://www.cdc.gov/brfss/ [

      Behavioral Risk Factor Surveillance System: Accessed on 10/01/2014. Available at: <http://www.cdc.gov/brfss/>.

      ]. This survey provides prevalence data on a state level for various disease conditions. It was established in 1984 by the Centers for Disease Control and Prevention. For some states, the samples have been stratified in such a way to allow for prevalence estimation. It became a national system in 1993 and an estimated 100,000 interviews were performed in 1993, which was expanded to more than 500,000 interviews by 2011. The BRFSS also collects demographic data including age, gender, ethnicity, and marital status etc. Its cross-sectional design is carried out by annual phone based surveys and captures data on certain behaviors including smoking, alcohol consumption, exercise, and obesity.
      Researchers using the BRFSS have combined it with other databases to assess the effect of the prevalence of certain risk factors and their relationship to other conditions. Davila et al. combined BRFSS results with HCC data from Surveillance Epidemiology and End Results (SEER) [
      • Davila J.A.
      • Petersena N.J.
      • Nelson H.A.
      • El-Serag H.B.
      Geographic variation within the United States in the incidence of hepatocellular carcinoma.
      ]. In their hierarchical model, HCC data from SEER was added to population-level prevalence estimates of potential HCC risk factors obtained from the BRFSS system in an attempt to explain geographic variations in HCC incidence rates. The behaviors studied did not reduce the unexplained variation in the incidence of HCC among geographic regions.

      Limitations

      Davila et al. reported population-level prevalence estimates of HCC from 1990 and concluded that since HCC estimates vary over time, they might not really reflect the prevalence at the time of acquisition of these factors by HCC patients in their study. This is especially the case as respondent behaviors in BRFSS are not linked to individual patient clinical data. As with other survey based databases, a major limitation of the BRFSS is the inability to verify respondent data, which may lead to under- or overestimation of true prevalence rates.

      Healthcare Cost and Utilization Project Database (United States)

      The Healthcare Cost and Utilization Project Database (HCUP) was initiated in 1988 and consists of a set of databases sponsored by the AHRQ and includes administrative data from patient encounters [

      Healthcare Cost and Utilization Project: Available at: <http://www.ahrq.gov/research/data/hcup/index.html>.

      ].
      • 1)
        Nationwide Inpatient Sample (http://www.hcup-us.ahrq.gov/nisoverview.jsp).
      • 2)
        Kids’ Inpatient Database.
      • 3)
        Nationwide Emergency Department Sample.
      • 4)
        State Inpatient Databases.
      • 5)
        State Ambulatory Surgery and Services Database.
      • 6)
        State Emergency Department Databases.
      For this review, we will focus on the Nationwide Inpatient Sample (NIS), which is the largest publicly available all-payer inpatient healthcare database in the United States. It contains a 20% sample of all discharges from hospitals participating in HCUP allowing for more precise estimates and reduced sampling errors. The following data elements are included in NIS.
      • Primary and secondary diagnoses and procedures.
      • Demographic data including gender, age, race, median household income for ZIP code.
      • Hospital characteristics (e.g. teaching status, rural vs. urban location, bed size).
      • Expected payment source (e.g. private, Medicare/medicaid, self-pay).
      • Charges.
      • Discharge status.
      • Length of stay.
      • Severity and comorbidity measures.
      The HCUP contains an extensive set of variables that have been utilized in HSR including healthcare resource utilization, hospital stay and healthcare costs, variations in practice patterns, patient safety and quality of care etc. The availability of hospital characteristics allows comparison of outcomes across a diverse group of hepatobiliary conditions. In addition, since the NIS samples across a wide variety of hospital types, biases associated with single institution studies are obviated. Bichoupan et al. employed data points from HCUP to estimate the costs associated with triple therapy in patients with HCV [
      • Bichoupan K.
      • Martel-Laferriere V.
      • Sachs D.
      • Ng M.
      • Schonfeld E.A.
      • Pappas A.
      • et al.
      Costs of telaprevir-based triple therapy for hepatitis C: $189,000 per sustained virological response.
      ], while other investigators have used the NIS to investigate regional variations in transplant wait time among patients with HCC [
      • Anderson J.E.
      • Hemming A.W.
      • Chang D.C.
      • Talamini M.A.
      • Mekeel K.L.
      Are hepatocellular carcinoma patients more likely to receive liver resection in regions with longer transplant wait times?.
      ]. The NIS has also been used to explore disparities related to the surgical treatment of liver metastases and outcomes following ablation of hepatic tumors [
      • Munene G.
      • Parker R.D.
      • Shaheen A.A.
      • Myers R.P.
      • Quan M.L.
      • Ball C.G.
      • et al.
      Disparities in the surgical treatment of colorectal liver metastases.
      ,
      • Fox J.P.
      • Gustafson J.
      • Desai M.M.
      • Hellan M.
      • Thambi-Pillai T.
      • Ouellette J.
      Short-term outcomes of ablation therapy for hepatic tumors: evidence from the 2006–2009 nationwide inpatient sample.
      ]. Nathan et al. described national trends in surgical procedures for HCC [
      • Nathan H.
      • Segev D.L.
      • Mayo S.C.
      • Choti M.A.
      • Cameron A.M.
      • Wolfgang C.L.
      • et al.
      National trends in surgical procedures for hepatocellular carcinoma: 1998–2008.
      ], while Kim et al. described HCC mortality rates and hospital utilization in the U.S. [
      • Kim W.R.
      • Gores G.J.
      • Benson J.T.
      • Therneau T.M.
      • Melton 3rd., L.J.
      Mortality and hospital utilization for hepatocellular carcinoma in the United States.
      ]. Ananthakrishnan et al. used NIS data to describe the effect of hospital volume and teaching status on acute liver failure outcomes [
      • Ananthakrishnan A.N.
      • McGinley E.L.
      • Saeian K.
      Effect of hospital volume and teaching status on outcomes of acute liver failure.
      ].

      Limitations

      Kim et al. support the fact that the NIS captures data on individual hospitalizations and not individual patients. As a result, it does not afford investigators the ability to link records preventing longitudinal tracking of patients, for instance, to determine how often a patient is hospitalized. Some authors have minimized the effect of these limitations by restricting analysis to one-time events [
      • Kim W.R.
      • Gores G.J.
      • Benson J.T.
      • Therneau T.M.
      • Melton 3rd., L.J.
      Mortality and hospital utilization for hepatocellular carcinoma in the United States.
      ]. In addition, with the administrative nature of the NIS database, specific tumor characteristic such as size, location and number are not available limiting the ability to adjust for their effect as potential confounders. The NIS database also does not provide data on outpatient visits or procedures limiting its use to inpatient hospitalization and procedures.

      Liver transplant outcomes

      United Kingdom Transplant Registry (Europe)

      The UK Transplant Registry (UKTR) collects data from various organ transplants across the UK and Ireland, including liver transplants http://www.odt.nhs.uk/uk-transplant-registry/data/. These data are collected via forms submitted by hospital staff and includes information on donor and recipient variables including donor type, age, gender, ethnicity, country of residence, primary diagnosis, and cause of death. Specific transplant data including year of transplant, cold ischemia time, human leukocyte antigen mismatch, urgency of transplant, graft and patient survival are also collected. These datasets can be requested by emailing [email protected] .
      The longitudinal nature of the database allows analysis of transplant related trends and outcomes. Edeghere et al. analyzed the UKTR to assess the impact of chronic HCV infection on liver transplantation in the UK [
      • Edeghere O.
      • Verlander N.Q.
      • Aboulhab J.
      • Costella A.
      • Harris H.E.
      • Balogun M.A.
      • et al.
      Retrospective cohort study of liver transplantation in the United Kingdom between 1994 and 2010: the impact of hepatitis C infection.
      ]. They reported that alcohol and hepatitis C related cirrhosis were the most common indications for liver transplantation during their study period. In addition, they described demographics of patients undergoing liver transplantation and outcomes following these procedures. In a separate study, Rowe et al. utilized the registry to study outcomes following retransplantation for graft failure among patients with chronic hepatitis C [
      • Rowe I.A.
      • Barber K.M.
      • Birch R.
      • Curnow E.
      • Neuberger J.M.
      Retransplantation for graft failure in chronic hepatitis C infection: a good use of a scarce resource?.
      ].

      Limitations

      Edeghere et al. report that the potential of bias from missing data in the registry [
      • Edeghere O.
      • Verlander N.Q.
      • Aboulhab J.
      • Costella A.
      • Harris H.E.
      • Balogun M.A.
      • et al.
      Retrospective cohort study of liver transplantation in the United Kingdom between 1994 and 2010: the impact of hepatitis C infection.
      ]. However, they report that misclassification is minimal because of the well maintained nature of the registry. Another limitation was their inability to perform a survival analysis because of the limitation of certain variables in the registry. However, other authors have generated survival analysis outcomes following liver transplantation using the same dataset [
      • Rowe I.A.
      • Barber K.M.
      • Birch R.
      • Curnow E.
      • Neuberger J.M.
      Retransplantation for graft failure in chronic hepatitis C infection: a good use of a scarce resource?.
      ].

      European Liver Transplant Registry (Europe)

      The European Liver Transplant Registry (ELTR) was created in 1985 as a registry of liver transplant procedures (pediatric and adult) in Europe, linking data from European liver transplant centers for scientific use (http://www.eltr.org). Between 1968 and 2013 the ELTR had collected data on 118,441 liver transplant performed in 155 centers from 28 countries. The registry collects data on the indication for liver transplant, transplant type, donor and recipient blood group, outcomes including complications, and immunosuppression data. Participation is completely voluntary with data collected using a standardized questionnaire. This standardized data collection tool minimizes the impact of variations in terminology and classification that could otherwise influence results of the ELTR. Adam et al. utilized the ELTR to study survival in liver transplant patients receiving prolonged release tacrolimus. It has also been utilized to study trends in indications and patient selection for liver transplantation [
      • Le Treut Y.P.
      • Gregoire E.
      • Klempnauer J.
      • Belghiti J.
      • Jouve E.
      • Lerut J.
      • et al.
      Liver transplantation for neuroendocrine tumors in Europe-results and trends in patient selection: a 213-case European liver transplant registry study.
      ,
      • Adam R.
      • Karam V.
      • Delvart V.
      • O’Grady J.
      • Mirza D.
      • Klempnauer J.
      • et al.
      Evolution of indications and results of liver transplantation in Europe. A report from the European Liver Transplant Registry (ELTR).
      ]. Adam et al. also audited the ELTR to identify risk factors for increased mortality associated with liver transplantation [
      • Adam R.
      • Cailliez V.
      • Majno P.
      • Karam V.
      • McMaster P.
      • Caine R.Y.
      • et al.
      Normalised intrinsic mortality risk in liver transplantation: European Liver Transplant Registry study.
      ]. Individual European countries also have organ procurement organizations and prospectively maintained transplant databases (e.g. the French Organ Procurement Organization [Agence de la Biomédicine] database) that have been used to describe liver transplant outcomes [
      • Ichai P.
      • Legeai C.
      • Francoz C.
      • Boudjema K.
      • Boillot O.
      • Ducerf C.
      • et al.
      Patients with acute liver failure listed for superurgent liver transplantation in France: reevaluation of the Clichy-Villejuif criteria.
      ].

      Limitations

      Adam et al. highlighted the fact that the ELTR questionnaire does not include information on ethnicity, which would limit the ability to explore racial disparities in outcomes and treatment patterns [
      • Adam R.
      • Karam V.
      • Delvart V.
      • Trunecka P.
      • Samuel D.
      • Bechstein W.O.
      • et al.
      Improved survival in liver transplant recipients receiving prolonged-release tacrolimus in the European Liver Transplant Registry.
      ]. They also reported their inability to explore socioeconomic differences in outcomes using the ELTR. Van Keimpema et al. in their study of survival after liver transplantation for polycystic kidney disease reported that pooling of results from multiple centers with different surgical and treatment protocols may influence interpretation and applicability of results [
      • van Keimpema L.
      • Nevens F.
      • Adam R.
      • Porte R.J.
      • Fikatas P.
      • Becker T.
      • et al.
      Excellent survival after liver transplantation for isolated polycystic liver disease: an European Liver Transplant Registry study.
      ]. They also had to exclude patients because they were unable to retrieve additional clinical information from participating centers and were concerned that this might have introduced selection bias. In addition, they reported the possibility of bias since the listing indication for liver transplant was based on the opinion of the individual transplant team.

      Organ Procurement and Transplant Network, United Network for Organ Sharing & Scientific Registry for Transplant Recipients (United States)

      The Organ Procurement and Transplant Network (OPTN) collects data on liver transplant recipients including adult and pediatric patients using data collection forms http://optn.transplant.hrsa.gov/converge/data/. These data are collected from before surgery at the time of listing to the post-transplant period. Collected data include demographic and clinical information, transplant type, and outcomes following surgery. Massie et al. highlighted that the OPTN also collects waitlist information including MELD scores, waitlist removals, organ acceptance, and rejections [
      • Massie A.B.
      • Kucirka L.M.
      • Segev D.L.
      Big data in organ transplantation: registries and administrative claims.
      ]. The aforementioned data collected by the OPTN is linked by the UNOS to social security death master files to augment ascertainment of mortality among donors and recipients [
      • Massie A.B.
      • Kucirka L.M.
      • Segev D.L.
      Big data in organ transplantation: registries and administrative claims.
      ].
      The Scientific Registry for Transplant Recipients (SRTR) was founded in 1987 and maintains a database on all solid organ transplants in the U.S. http://www.srtr.org/who.aspx. The data are from different sources but mostly from the OPTN, which allows further ascertainment of death via the National Death Index and graft failure via the Centers for Medicare and Medicaid Services. SRTR uses these data to create reports for OPTN committees that make organ allocation policy, and files for researchers interested in secondary analysis. In addition, SRTR and OPTN produce an annual data report and provide information upon request for the general public [
      • Leppke S.
      • Leighton T.
      • Zaun D.
      • Chen S.C.
      • Skeans M.
      • Israni A.K.
      • et al.
      Scientific Registry of Transplant Recipients: collecting, analyzing, and reporting data on transplantation in the United States.
      ].
      These databases have been extensively utilized for research in liver transplantation. For instance, Wong et al. utilized OPTN/UNOS data to demonstrate a significant increase in patients with non-alcoholic steatohepatitis awaiting liver transplant [
      • Wong R.J.
      • Aguilar M.
      • Cheung R.
      • Perumpail R.B.
      • Harrison S.A.
      • Younossi Z.M.
      • et al.
      Nonalcoholic steatohepatitis is the second leading etiology of liver disease among adults awaiting liver transplantation in the United States.
      ]. Other investigators have used OPTN/UNOS data to describe the impact of comorbidities such as hepatitis C and diabetes on liver transplant outcomes [
      • Perumpail R.B.
      • Wong R.J.
      • Scandling J.D.
      • Ha L.D.
      • Todo T.
      • Bonham C.A.
      • et al.
      HCV Infection is Associated with Lower Survival in Simultaneous Liver Kidney Transplant Recipients in the U.S.
      ,
      • Uemura T.
      • Ramprasad V.
      • Hollenbeak C.S.
      • Bezinover D.
      • Kadry Z.
      Liver transplantation for hepatitis C from donation after cardiac death donors: an analysis of OPTN/UNOS data.
      ,
      • Hoehn R.S.
      • Singhal A.
      • Wima K.
      • Sutton J.M.
      • Paterno F.
      • Steve Woodle E.
      • et al.
      Effect of pretransplant diabetes on short-term outcomes after liver transplantation: a National cohort study.
      ]. The linked nature of these databases allows efficient study of post-transplant outcomes. Goldberg et al. used the SRTR to describe the risk of end-stage renal disease among living donor liver transplant recipients [
      • Goldberg D.S.
      • Ruebner R.L.
      • Abt P.L.
      The Risk of End-Stage Renal Disease Among Living Donor Liver Transplant Recipients in the United States.
      ], while Wong et al. described the impact of pretransplant hepatic encephalopathy on post-liver transplant survival [
      • Wong R.J.
      • Aguilar M.
      • Gish R.G.
      • Cheung R.
      • Ahmed A.
      The impact of pretransplant hepatic encephalopathy on survival following liver transplantation.
      ].

      Limitations

      Goldberg et al. highlight the fact that while the SRTR accurately identifies patients with graft failure, it does not identify patients with early graft dysfunction [
      • Goldberg D.S.
      • Ruebner R.L.
      • Abt P.L.
      The Risk of End-Stage Renal Disease Among Living Donor Liver Transplant Recipients in the United States.
      ]. Other investigators have reported that data on HBV DNA levels, genotype, and HBV treatment regimens are lacking in the UNOS database [
      • Dugum M.
      • Hanouneh I.
      • Lopez R.
      • Aucejo F.
      • Eghtesad B.
      • Zein N.
      Hepatocellular Carcinoma in the Setting of Chronic Hepatitis B Virus Infection: Tumor Recurrence and Survival Rates After Liver Transplantation.
      ]. In this study of HCC recurrence among patients with HBV after liver transplant, the authors report that their findings may not be generalizable to populations outside the U.S. because of differences in HBV biology, treatment and post-liver transplant monitoring. Bambha et al. acknowledge a significant selection bias at the level of the transplant centers regarding their choice to transplant patients, a limitation that may not be adequately controlled for during secondary analysis of these large databases [
      • Bambha K.M.
      • Dodge J.L.
      • Gralla J.
      • Sprague D.
      • Biggins S.W.
      Low, rather than high, body mass index confers increased risk for post-liver transplant death and graft loss: risk modulated by MELD.
      ].

      Pharmacovigilance databases

      Pharmacovigilance databases in Spain, France and Italy have been used for research [
      • Lapeyre-Mestre M.
      • Grolleau S.
      • Montastruc J.L.
      Adverse drug reactions associated with the use of NSAIDs: a case/noncase analysis of spontaneous reports from the French pharmacovigilance database 2002–2006.
      ,
      • Lapeyre-Mestre M.
      • de Castro A.M.
      • Bareille M.P.
      • Del Pozo J.G.
      • Requejo A.A.
      • Arias L.M.
      • et al.
      Non-steroidal anti-inflammatory drug-related hepatic damage in France and Spain: analysis from national spontaneous reporting systems.
      ]. These are national databases that collect information on adverse drug reactions as reported by healthcare providers. Conforti et al. utilized the Italian pharmacovigilance database to study reporting patterns among nurses [
      • Conforti A.
      • Opri S.
      • D’Incau P.
      • Sottosanti L.
      • Moretti U.
      • Ferrazin F.
      • et al.
      Adverse drug reaction reporting by nurses: analysis of Italian pharmacovigilance database.
      ]. Lapeyre-Mestre et al. using information from the Spanish and French pharmacovigilance databases reported liver-related adverse drug reactions associated with non-steroidal anti-inflammatory drug usage and described rates of serious gastrointestinal and hepatic side effects [
      • Lapeyre-Mestre M.
      • Grolleau S.
      • Montastruc J.L.
      Adverse drug reactions associated with the use of NSAIDs: a case/noncase analysis of spontaneous reports from the French pharmacovigilance database 2002–2006.
      ,
      • Lapeyre-Mestre M.
      • de Castro A.M.
      • Bareille M.P.
      • Del Pozo J.G.
      • Requejo A.A.
      • Arias L.M.
      • et al.
      Non-steroidal anti-inflammatory drug-related hepatic damage in France and Spain: analysis from national spontaneous reporting systems.
      ].

      Limitations

      The major limitation of pharmacovigilance databases is that most data are spontaneously reported, which may lead to gross underestimation of the true incidence of adverse drug events.

      Cancer outcomes

      NORDCAN database (Europe)

      The NORDCAN database collects epidemiological data including incidence, mortality, prevalence, and survival statistics of about 50 major cancers (including liver malignancies) from the Nordic countries comprising Denmark, Finland, Faroe Islands, Greenland, Iceland, Norway and Sweden http://www-dep.iarc.fr/NORDCAN/english/frame.asp [
      • Engholm G.
      • Ferlay J.
      • Christensen N.
      • Bray F.
      • Gjerstorff M.L.
      • Klint A.
      • et al.
      NORDCAN – A Nordic tool for cancer information, planning, quality control and research.
      ]. Participating countries except Sweden have a central cancer notification system that allows excellent collection of data on incident cancer cases [
      • Engholm G.
      • Ferlay J.
      • Christensen N.
      • Bray F.
      • Gjerstorff M.L.
      • Klint A.
      • et al.
      NORDCAN – A Nordic tool for cancer information, planning, quality control and research.
      ]. NORDCAN has been utilized in describing time trends in cancer rates and survival, and also for HSR [
      • Kalseth J.
      • Halvorsen T.
      • Kalseth B.
      • Sarheim Anthun K.
      • Peltola M.
      • Kautiainen K.
      • et al.
      Cross-country comparisons of health-care costs: the case of cancer treatment in the Nordic countries.
      ]. In addition, it has been used in the development of prediction modules for various cancers [
      • Engholm G.
      • Ferlay J.
      • Christensen N.
      • Bray F.
      • Gjerstorff M.L.
      • Klint A.
      • et al.
      NORDCAN – A Nordic tool for cancer information, planning, quality control and research.
      ].

      Limitations

      Even though cross-country comparison of healthcare costs among the Nordic countries can be performed with the NORDCAN database [
      • Kalseth J.
      • Halvorsen T.
      • Kalseth B.
      • Sarheim Anthun K.
      • Peltola M.
      • Kautiainen K.
      • et al.
      Cross-country comparisons of health-care costs: the case of cancer treatment in the Nordic countries.
      ], the homogeneity of the Nordic population may limit the ability to investigate for racial differences in outcomes, and also the generalizability of findings to more diverse parts of Europe [
      • Annertz K.
      • Anderson H.
      • Palmer K.
      • Wennerberg J.
      The increase in incidence of cancer of the tongue in the Nordic countries continues into the twenty-first century.
      ]. Kalseth et al. also reported the possibility of introduction of bias because of differences in data accessibility, patient classification systems, and healthcare organization among participating countries [
      • Kalseth J.
      • Halvorsen T.
      • Kalseth B.
      • Sarheim Anthun K.
      • Peltola M.
      • Kautiainen K.
      • et al.
      Cross-country comparisons of health-care costs: the case of cancer treatment in the Nordic countries.
      ].

      Surveillance Epidemiology and End Results – Medicare linked database (United States)

      The SEER program is sponsored by the National Cancer Institute and reports cancer statistics in the U.S. [

      Surveillance, Epidemiology, and End Results Program. Available at: <http://seer.cancer.gov/>.

      ]. SEER contains data on demographics (age and gender), tumor morphology (including size), stage at diagnosis (localized, regional, distant, and unstaged), treatment and survival for patients with HCC which has been linked to Medicare claims data http://healthcaredelivery.cancer.gov/seermedicare/. It has been utilized in the reporting of HCC and intrahepatic bile duct cancer related incidence and mortality trends, especially in relation to various treatment patterns [
      • Eggert T.
      • McGlynn K.A.
      • Duffy A.
      • Manns M.P.
      • Greten T.F.
      • Altekruse S.F.
      Fibrolamellar hepatocellular carcinoma in the USA, 2000–2010: a detailed report on frequency, treatment and outcome based on the Surveillance, Epidemiology, and End Results database.
      ,
      • Amini N.
      • Ejaz A.
      • Spolverato G.
      • Kim Y.
      • Herman J.M.
      • Pawlik T.M.
      Temporal trends in liver-directed therapy of patients with intrahepatic cholangiocarcinoma in the United States: a population-based analysis.
      ,
      • Altekruse S.F.
      • Henley S.J.
      • Cucinelli J.E.
      • McGlynn K.A.
      Changing hepatocellular carcinoma incidence and liver cancer mortality rates in the United States.
      ].
      The availability of incidence data on a state level allows investigation of geographical differences in HCC incidence and mortality, which could highlight differences in care systems. For instance, Altecruse et al. reported an increase in HCC mortality rates especially among the baby boomers while decreasing rates were noted in the 35–49 year old demographics. SEER data has also been used to report disparities in HCC related mortality showing significantly higher mortality rates in African-Americans compared to other ethnic/racial groups [
      • Altekruse S.F.
      • McGlynn K.A.
      • Reichman M.E.
      Hepatocellular carcinoma incidence, mortality, and survival trends in the United States from 1975 to 2005.
      ,
      • Ahmed F.
      • Perz J.F.
      • Kwong S.
      • Jamison P.M.
      • Friedman C.
      • Bell B.P.
      National trends and disparities in the incidence of hepatocellular carcinoma, 1998–2003.
      ]. In addition, HCC data in the SEER database has been used in comparative and cost-effectiveness studies of HCC treatment strategies [
      • Shaya F.T.
      • Breunig I.M.
      • Seal B.
      • Mullins C.D.
      • Chirikov V.V.
      • Hanna N.
      Comparative and cost effectiveness of treatment modalities for hepatocellular carcinoma in SEER-Medicare.
      ]. Hyder et al. utilized the SEER database linked to Medicare data to explore referral patterns and treatment choices for patients with HCC. They identified regional variations in referral and receipt of treatment, which eventually did not translate to survival differences. Yang et al. investigated the impact of sex on HCC survival and showed that after multivariate analysis, women had a significantly greater median overall survival than men, especially among younger cohorts [
      • Yang D.
      • Hanna D.L.
      • Usher J.
      • LoCoco J.
      • Chaudhari P.
      • Lenz H.J.
      • et al.
      Impact of sex on the survival of patients with hepatocellular carcinoma: a surveillance, epidemiology, and end results analysis.
      ].

      Limitations

      The major limitation of the SEER-Medicare linked database is the restriction of data to patients 65 years and older. Hyder et al. reported that SEER-linked Medicare data did not capture variables such as performance status, alpha-fetoprotein level, and severity of cirrhosis that are essential in clinical decision making [
      • Hyder O.
      • Dodson R.M.
      • Nathan H.
      • Herman J.M.
      • Cosgrove D.
      • Kamel I.
      • et al.
      Referral patterns and treatment choices for patients with hepatocellular carcinoma: a United States population-based study.
      ]. In addition, the SEER registry also does not contain specific primary tumor data including number of liver lesions and margin status [
      • Parikh A.A.
      • Ni S.
      • Koyama T.
      • Pawlik T.M.
      • Penson D.
      Trends in the multimodality treatment of resectable colorectal liver metastases: an underutilized strategy.
      ]. For investigators interested in studying treatment patterns, they reported that SEER is also limited by the capture of extent of surgical resection, type of ablation performed, and type of TACE agent used. Yang et al. report that the SEER database lacks data on HCC predisposing risk factors e.g. HBV, HCV, and NAFLD. In addition, there is limited data on comorbidities and socioeconomic status of HCC patients limiting generalizability of findings. Altecruse et al. reported an increase in U.S. liver cancer related mortality but also showed that the increase in HCC incidence rates was no longer statistically significant. They attributed this difference to a potential limitation of the SEER registry – that the SEER registry covers only 28% of the U.S. while mortality statistics are obtained from 98% of the U.S. As a result, they suggested that since mortality statistics from Texas, Mississippi and District of Columbia, which have the highest mortality rates, are not in the SEER catchment area, SEER results may not be reflective of the actual mortality data of the U.S. as a whole.

      Hepatitis C outcomes

      National Veterans Administration HCV Clinical Case Registry (United States)

      The U.S. Department of Veterans Affairs (VA) has a comprehensive HCV registry developed from its electronic medical record system known as Veterans Health Information Systems and Technology Architecture and its graphical interphase termed Computer Patient Record System. Backus et al. extensively described the VA HCV and HIV clinical case registries (CCR) in their 2009 review paper [
      • Backus L.I.
      • Gavrilov S.
      • Loomis T.P.
      • Halloran J.P.
      • Phillips B.R.
      • Belperio P.S.
      • et al.
      Clinical Case Registries: simultaneous local and national disease registries for population quality management.
      ]. They describe the methodology of verifying HCV cases in this registry utilizing HCV antibody, recombinant immunoblot assay, genotype and viral load testing. They summarized that from 1998 to 2007, 292,205 patients in the registry had laboratory evidence of HCV. Key data extracted by the CCR include inpatient hospitalizations and procedures, medications, laboratory tests, pathology, problem list, outpatient visits, radiology and vital signs.
      The strengths of this registry include the aggregation of data on a national level from all VA medical facilities. The authors propose that these registries can be used to study quality of care and safety in U.S. veterans with the goal of improving healthcare delivery. Data points from this registry can also be used in cost modeling and healthcare planning. In addition, the registries are robust enough to provide the unique opportunity to study HCV patients co-infected with HIV. The VA HCV registry has been used to study predictors of HCV treatment and postmarketing surveillance of drugs in veterans.

      Limitations

      One limitation of this database is the fact that some veterans may receive testing and treatment outside of the VA care system and as such these data may not be captured in the CCR. In addition, since this registry collects information only in veterans, it limits generalizability to the general population. External validity is also impacted since the VA patient population is mostly male, as a result, studies investigating gender differences may not be feasible with the HCV CCR. Another limitation reported in one study is that certain variables such as the degree of cirrhosis and cause of mortality are not captured as data points in the registry [
      • Backus L.I.
      • Boothroyd D.B.
      • Phillips B.R.
      • Belperio P.
      • Halloran J.
      • Mole L.A.
      A sustained virologic response reduces risk of all-cause mortality in patients with hepatitis C.
      ]. As a result, these authors concluded that it was not possible to ascertain if in all-cause mortality was due to a reduction in liver-related mortality.

      Surgical outcomes

      National Surgical Quality Improvement Program (United States)

      The National Surgical Quality Improvement Program (NSQIP) is an initiative of the American College of Surgeons (ACS) that collects data from patient’s charts in participating hospitals, which is used to assess the quality of care provided by surgeons across various surgical specialties https://www.facs.org/quality-programs/acs-nsqip. Data is collected by on-site clinical reviewers; risk adjusted using validated models, and used to assess patient outcomes such as 30-day postoperative complication rates. This allows the ACS to provide blinded, risk adjusted information to participating hospitals allowing them to benchmark their complication rates and surgical outcomes. The major strengths of this program include the collection of actual patient data by on-site surgical clinical reviewers, which has been established to be more accurate than claims-based data. In addition, the ACS provides regular training of clinical reviewers and audits to ensure reliable collection of data. The NSQIP has been used to report perioperative hepatectomy outcomes in patients with HCC.

      Limitations

      Some authors have highlighted that the NSQIP does not collect hospital costs, patient charges, or reimbursement figures [
      • Kim S.
      • Maynard E.C.
      • Shah M.B.
      • Daily M.F.
      • Tzeng C.W.
      • Davenport D.L.
      • et al.
      Risk factors for 30-day readmissions after hepatectomy: analysis of 2444 patients from the ACS-NSQIP database.
      ]. The absence of these data prevents direct cost analysis on the impact of hospital readmissions and reoperations after surgery. In addition, for cancer related procedures, Bartlett et al. state that specific pathological data may not be available [
      • Bartlett E.K.
      • Meise C.
      • Roses R.E.
      • Fraker D.L.
      • Kelz R.R.
      • Karakousis G.C.
      Morbidity and mortality of cytoreduction with intraperitoneal chemotherapy: outcomes from the ACS NSQIP database.
      ]. Furthermore, there is no data on liver biochemical tests, and the 30-day reported outcomes may be insufficient in assessing for intermediate and long-term outcomes such as biliary stenosis that may be seen after liver surgeries.

      Conclusion

      In summary, large databases are an underutilized resource in hepatology research. In a previous article, we (J.A.T.) described available mechanisms for healthcare delivery research training in gastroenterology and hepatology [
      • Talwalkar J.A.
      • Oxentenko A.S.
      • Katzka D.A.
      Health care-delivery research-training opportunities in gastroenterology and hepatology.
      ]. We emphasized the need to train emerging healthcare providers in the various HSR methodologies. We strongly advise future investigators to thoroughly review the strengths and limitations of individual databases prior to the study design phase to ensure feasibility and internal validity. As we continue to generate more patient-related data, and refine the ability to link existing databases, novel opportunities will be created that can be successfully exploited to close prevailing knowledge and practice gaps in the healthcare delivery process with the overarching goal of improving patient outcomes and containing costs.

      Conflict of interest

      The authors declared that they do not have anything to disclose regarding funding or conflict of interest with respect to this manuscript.

      References

      Author names in bold designate shared co-first authorship

        • Newgard C.D.
        • Lewis R.J.
        Missing data: how to best account for what is not known.
        JAMA. 2015; 314: 940-941
        • Greenland S.
        • Finkle W.D.
        A critical look at methods for handling missing covariates in epidemiologic regression analyses.
        Am J Epidemiol. 1995; 142: 1255-1264
        • Kaplan B.
        • Schold J.
        • Meier-Kriesche H.U.
        Overview of large database analysis in renal transplantation.
        Am J Transplant. 2003; 3: 1052-1056
        • Byass P.
        The global burden of liver disease: a challenge for methods and for public health.
        BMC Med. 2014; 12: 159
        • Rogal S.S.
        • Winger D.
        • Bielefeldt K.
        • Rollman B.L.
        • Szigethy E.
        Healthcare utilization in chronic liver disease: the importance of pain and prescription opioid use.
        Liver Int. 2013; 33: 1497-1503
        • Ganesh S.
        • Rogal S.S.
        • Yadav D.
        • Humar A.
        • Behari J.
        Risk factors for frequent readmissions and barriers to transplantation in patients with cirrhosis.
        PLoS One. 2013; 8: e55140
        • Kabiri M.
        • Jazwinski A.B.
        • Roberts M.S.
        • Schaefer A.J.
        • Chhatwal J.
        The changing burden of hepatitis C virus infection in the United States: model-based predictions.
        Ann Intern Med. 2014; 161: 170-180
        • Lazo M.
        • Hernaez R.
        • Eberhardt M.S.
        • Bonekamp S.
        • Kamel I.
        • Guallar E.
        • et al.
        Prevalence of nonalcoholic fatty liver disease in the United States: the Third National Health and Nutrition Examination Survey, 1988–1994.
        Am J Epidemiol. 2013; 178: 38-45
        • Chen L.
        • Reed C.
        • Happich M.
        • Nyhuis A.
        • Lenox-Smith A.
        Health care resource utilisation in primary care prior to and after a diagnosis of Alzheimer’s disease: a retrospective, matched case-control study in the United Kingdom.
        BMC Geriatr. 2014; 14: 76
        • Ratib S.
        • West J.
        • Crooks C.J.
        • Fleming K.M.
        Diagnosis of liver cirrhosis in England, a cohort study, 1998–2009: a comparison with cancer.
        Am J Gastroenterol. 2014; 109: 190-198
        • Seaman H.E.
        • de Vries C.S.
        • Farmer R.D.
        The risk of liver disorders in women prescribed cyproterone acetate in combination with ethinyloestradiol (Dianette): a nested case-control study using the GPRD.
        Pharmacoepidemiol Drug Saf. 2003; 12: 541-550
        • Myers M.W.
        • Jick H.
        Terfenadine and risk of acute liver disease.
        Br J Clin Pharmacol. 1998; 46: 251-253
        • Clifford G.M.
        • Logie J.
        • Farmer R.D.
        No risk of drug-associated liver injury with alpha1-adrenoreceptor blocking agents in men with BPH: results from an observational study using the GPRD.
        Pharmacoepidemiol Drug Saf. 2005; 14: 75-80
      1. Clinical Practice Research Datalink. Available at: <http://www.cprd.com/intro.asp>. Accessed 7/24/2015.

        • McGlynn K.A.
        • Hagberg K.
        • Chen J.
        • Graubard B.I.
        • London W.T.
        • Jick S.
        • et al.
        Statin use and risk of primary liver cancer in the Clinical Practice Research Datalink.
        J Natl Cancer Inst. 2015; 107 (pii: djv009)
      2. IMS Health Strategic Data. Available at: <https://www.sd.imshealth.com/fr/Pages/default.aspx>. Accessed 7/24/2015.

        • Sabouret P.
        • Bricard M.
        • Hermann M.A.
        • Cotte F.E.
        • Deret-Bixio L.
        • Rushton-Smith S.
        Discrepancy between guidelines for stroke prevention in atrial fibrillation and practice patterns in primary care. The nationwide French AFIGP survey.
        Arch Cardiovasc Dis. 2015; 108: 544-553
        • Keane M.G.
        • Horsfall L.
        • Rait G.
        • Pereira S.P.
        A case-control study comparing the incidence of early symptoms in pancreatic and biliary tract cancer.
        BMJ Open. 2014; 4: e005720
        • Alexandropoulou K.
        • van Vlymen J.
        • Reid F.
        • Poullis A.
        • Kang J.Y.
        Temporal trends of Barrett’s oesophagus and gastro-oesophageal reflux and related oesophageal cancer over a 10-year period in England and Wales and associated proton pump inhibitor and H2RA prescriptions: a GPRD study.
        Eur J Gastroenterol Hepatol. 2013; 25: 15-21
        • Suissa D.
        • Delaney J.A.
        • Dial S.
        • Brassard P.
        Non-steroidal anti-inflammatory drugs and the risk of Clostridium difficile-associated disease.
        Br J Clin Pharmacol. 2012; 74: 370-375
        • Wright E.
        • Schofield P.T.
        • Seed P.
        • Molokhia M.
        Bisphosphonates and risk of upper gastrointestinal cancer–a case control study using the General Practice Research Database (GPRD).
        PLoS One. 2012; 7: e47616
      3. National Hospital Ambulatory Medical Care Survey Database. Available at: <http://www.cdc.gov/nchs/ahcd/ahcd_questionnaires.htm>.

        • Wu B.U.
        • Banks P.A.
        • Conwell D.L.
        Disparities in emergency department wait times for acute gastrointestinal illnesses: results from the National Hospital Ambulatory Medical Care Survey, 1997–2006.
        Am J Gastroenterol. 2009; 104: 1668-1673
        • Tsui J.I.
        • Maselli J.
        • Gonzales R.
        Sociodemographic trends in national ambulatory care visits for hepatitis C virus infection.
        Dig Dis Sci. 2009; 54: 2694-2698
        • Cheung R.
        • Mannalithara A.
        • Singh G.
        Utilization and antiviral therapy in patients with chronic hepatitis C: analysis of ambulatory care visits in the US.
        Dig Dis Sci. 2010; 55: 1744-1751
        • Roberts H.W.
        • Utuama O.A.
        • Klevens M.
        • Teshale E.
        • Hughes E.
        • Jiles R.
        The contribution of viral hepatitis to the burden of chronic liver disease in the United States.
        Am J Gastroenterol. 2014; 109 (quiz 6, 94): 387-393
        • McCaig L.F.
        • Burt C.W.
        Understanding and interpreting the National Hospital Ambulatory Medical Care Survey: key questions and answers.
        Ann Emerg Med. 2012; 60 (e1): 716-721
      4. NHANES. Available at: <http://www.cdc.gov/nchs/nhanes.htm>. Accessed 10/01/2014.

      5. NHANES Survey Data. Available at: <http://www.cdc.gov/nchs/data/nhanes/survey_content_99_14.pdf>. Accessed 9/30/2014.

        • Coleman P.J.
        • McQuillan G.M.
        • Moyer L.A.
        • Lambert S.B.
        • Margolis H.S.
        Incidence of hepatitis B virus infection in the United States, 1976–1994: estimates from the National Health and Nutrition Examination Surveys.
        J Infect Dis. 1998; 178: 954-959
        • Alter M.J.
        • Kruszon-Moran D.
        • Nainan O.V.
        • McQuillan G.M.
        • Gao F.
        • Moyer L.A.
        • et al.
        The prevalence of hepatitis C virus infection in the United States, 1988 through 1994.
        N Engl J Med. 1999; 341: 556-562
        • Kuniholm M.H.
        • Purcell R.H.
        • McQuillan G.M.
        • Engle R.E.
        • Wasley A.
        • Nelson K.E.
        Epidemiology of hepatitis E virus in the United States: results from the Third National Health and Nutrition Examination Survey, 1988–1994.
        J Infect Dis. 2009; 200: 48-56
        • Younossi Z.M.
        • Stepanova M.
        Changes in hepatitis A and B vaccination rates in adult patients with chronic liver diseases and diabetes in the U.S. population.
        Hepatology. 2011; 54: 1167-1178
        • Rein D.B.
        • Smith B.D.
        • Wittenborn J.S.
        • Lesesne S.B.
        • Wagner L.D.
        • Roblin D.W.
        • et al.
        The cost-effectiveness of birth-cohort screening for hepatitis C antibody in U.S. primary care settings.
        Ann Intern Med. 2012; 156: 263-270
        • Ioannou G.N.
        • Dominitz J.A.
        • Weiss N.S.
        • Heagerty P.J.
        • Kowdley K.V.
        Racial differences in the relationship between hepatitis C infection and iron stores.
        Hepatology. 2003; 37: 795-801
        • Flores Y.N.
        • Yee Jr., H.F.
        • Leng M.
        • Escarce J.J.
        • Bastani R.
        • Salmeron J.
        • et al.
        Risk factors for chronic liver disease in Blacks, Mexican Americans, and Whites in the United States: results from NHANES IV, 1999–2004.
        Am J Gastroenterol. 2008; 103: 2231-2238
        • Ioannou G.N.
        • Weiss N.S.
        • Boyko E.J.
        • Mozaffarian D.
        • Lee S.P.
        Elevated serum alanine aminotransferase activity and calculated risk of coronary heart disease in the United States.
        Hepatology. 2006; 43: 1145-1151
        • Ruhl C.E.
        • Menke A.
        • Cowie C.C.
        • Everhart J.E.
        Relationship of hepatitis C virus infection with diabetes in the U.S. population.
        Hepatology. 2014; 60: 1139-1149
      6. National Health Interview Survey: Available at: <http://www.cdc.gov/nchs/nhis/about_nhis.htm>. Accessed 10/01/2014.

      7. Hepatitis B vaccination coverage among adults–United States, 2004.
        MMWR Morb Mortal Wkly Rep. 2006; 55: 509-511
        • Jain N.
        • Yusuf H.
        • Wortley P.M.
        • Euler G.L.
        • Walton S.
        • Stokley S.
        Factors associated with receiving hepatitis B vaccination among high-risk adults in the United States: an analysis of the National Health Interview Survey, 2000.
        Fam Med. 2004; 36: 480-486
        • Silverberg J.I.
        Healthcare utilization, patient costs, and access to care in US adults with eczema: a population-based study.
        JAMA Dermatol. 2015; 151: 743-752
      8. Medical Expenditure Panel Survey. Available at: <http://meps.ahrq.gov/mepsweb/about_meps/survey_back.jsp>. Accessed 3/23/2015.

        • Bernard D.S.
        • Farr S.L.
        • Fang Z.
        National estimates of out-of-pocket health care expenditure burdens among nonelderly adults with cancer: 2001 to 2008.
        J Clin Oncol. 2011; 29: 2821-2826
        • Wang Y.R.
        • Loftus Jr., E.V.
        • Cangemi J.R.
        • Picco M.F.
        Racial/Ethnic and regional differences in the prevalence of inflammatory bowel disease in the United States.
        Digestion. 2013; 88: 20-25
        • Shi L.
        • Chen C.C.
        • Nie X.
        • Zhu J.
        • Hu R.
        Racial and socioeconomic disparities in access to primary care among people with chronic conditions.
        J Am Board Fam Med. 2014; 27: 189-198
        • Machlin S.R.
        • Soni A.
        Health care expenditures for adults with multiple treated chronic conditions: estimates from the Medical Expenditure Panel Survey, 2009.
        Prev Chronic Dis. 2013; 10: E63
        • Holden C.D.
        • Chen J.
        • Dagher R.K.
        Preventive care utilization among the uninsured by race/ethnicity and income.
        Am J Prev Med. 2015; 48: 13-21
        • Li R.
        • Barker L.E.
        • Shrestha S.
        • Zhang P.
        • Duru O.K.
        • Pearson-Clarke T.
        • et al.
        Changes over time in high out-of-pocket health care burden in U.S. adults with diabetes, 2001–2011.
        Diabetes Care. 2014; 37: 1629-1635
        • Wang J.
        • Mullins C.D.
        • Zuckerman I.H.
        • Walker G.D.
        • Suda K.J.
        • Yang Y.
        • et al.
        Medical Expenditure Panel Survey: a valuable database for studying racial and ethnic disparities in prescription drug use.
        Res Social Adm Pharm. 2008; 4: 206-217
      9. Behavioral Risk Factor Surveillance System: Accessed on 10/01/2014. Available at: <http://www.cdc.gov/brfss/>.

        • Davila J.A.
        • Petersena N.J.
        • Nelson H.A.
        • El-Serag H.B.
        Geographic variation within the United States in the incidence of hepatocellular carcinoma.
        J Clin Epidemiol. 2003; 56: 487-493
      10. Healthcare Cost and Utilization Project: Available at: <http://www.ahrq.gov/research/data/hcup/index.html>.

        • Bichoupan K.
        • Martel-Laferriere V.
        • Sachs D.
        • Ng M.
        • Schonfeld E.A.
        • Pappas A.
        • et al.
        Costs of telaprevir-based triple therapy for hepatitis C: $189,000 per sustained virological response.
        Hepatology. 2014; 60: 1187-1195
        • Anderson J.E.
        • Hemming A.W.
        • Chang D.C.
        • Talamini M.A.
        • Mekeel K.L.
        Are hepatocellular carcinoma patients more likely to receive liver resection in regions with longer transplant wait times?.
        Transplant Proc. 2014; 46: 199-201
        • Munene G.
        • Parker R.D.
        • Shaheen A.A.
        • Myers R.P.
        • Quan M.L.
        • Ball C.G.
        • et al.
        Disparities in the surgical treatment of colorectal liver metastases.
        J Natl Med Assoc. 2013; 105: 128-137
        • Fox J.P.
        • Gustafson J.
        • Desai M.M.
        • Hellan M.
        • Thambi-Pillai T.
        • Ouellette J.
        Short-term outcomes of ablation therapy for hepatic tumors: evidence from the 2006–2009 nationwide inpatient sample.
        Ann Surg Oncol. 2012; 19: 3677-3686
        • Nathan H.
        • Segev D.L.
        • Mayo S.C.
        • Choti M.A.
        • Cameron A.M.
        • Wolfgang C.L.
        • et al.
        National trends in surgical procedures for hepatocellular carcinoma: 1998–2008.
        Cancer. 2012; 118: 1838-1844
        • Kim W.R.
        • Gores G.J.
        • Benson J.T.
        • Therneau T.M.
        • Melton 3rd., L.J.
        Mortality and hospital utilization for hepatocellular carcinoma in the United States.
        Gastroenterology. 2005; 129: 486-493
        • Ananthakrishnan A.N.
        • McGinley E.L.
        • Saeian K.
        Effect of hospital volume and teaching status on outcomes of acute liver failure.
        Liver Transpl. 2008; 14: 1347-1356
        • Edeghere O.
        • Verlander N.Q.
        • Aboulhab J.
        • Costella A.
        • Harris H.E.
        • Balogun M.A.
        • et al.
        Retrospective cohort study of liver transplantation in the United Kingdom between 1994 and 2010: the impact of hepatitis C infection.
        Public Health. 2015; 129: 509-516
        • Rowe I.A.
        • Barber K.M.
        • Birch R.
        • Curnow E.
        • Neuberger J.M.
        Retransplantation for graft failure in chronic hepatitis C infection: a good use of a scarce resource?.
        World J Gastroenterol. 2010; 16: 5070-5076
        • Le Treut Y.P.
        • Gregoire E.
        • Klempnauer J.
        • Belghiti J.
        • Jouve E.
        • Lerut J.
        • et al.
        Liver transplantation for neuroendocrine tumors in Europe-results and trends in patient selection: a 213-case European liver transplant registry study.
        Ann Surg. 2013; 257: 807-815
        • Adam R.
        • Karam V.
        • Delvart V.
        • O’Grady J.
        • Mirza D.
        • Klempnauer J.
        • et al.
        Evolution of indications and results of liver transplantation in Europe. A report from the European Liver Transplant Registry (ELTR).
        J Hepatol. 2012; 57: 675-688
        • Adam R.
        • Cailliez V.
        • Majno P.
        • Karam V.
        • McMaster P.
        • Caine R.Y.
        • et al.
        Normalised intrinsic mortality risk in liver transplantation: European Liver Transplant Registry study.
        Lancet. 2000; 356: 621-627
        • Ichai P.
        • Legeai C.
        • Francoz C.
        • Boudjema K.
        • Boillot O.
        • Ducerf C.
        • et al.
        Patients with acute liver failure listed for superurgent liver transplantation in France: reevaluation of the Clichy-Villejuif criteria.
        Liver Transpl. 2015; 21: 512-523
        • Adam R.
        • Karam V.
        • Delvart V.
        • Trunecka P.
        • Samuel D.
        • Bechstein W.O.
        • et al.
        Improved survival in liver transplant recipients receiving prolonged-release tacrolimus in the European Liver Transplant Registry.
        Am J Transplant. 2015; 15: 1267-1282
        • van Keimpema L.
        • Nevens F.
        • Adam R.
        • Porte R.J.
        • Fikatas P.
        • Becker T.
        • et al.
        Excellent survival after liver transplantation for isolated polycystic liver disease: an European Liver Transplant Registry study.
        Transpl Int. 2011; 24: 1239-1245
        • Massie A.B.
        • Kucirka L.M.
        • Segev D.L.
        Big data in organ transplantation: registries and administrative claims.
        Am J Transplant. 2014; 14: 1723-1730
        • Leppke S.
        • Leighton T.
        • Zaun D.
        • Chen S.C.
        • Skeans M.
        • Israni A.K.
        • et al.
        Scientific Registry of Transplant Recipients: collecting, analyzing, and reporting data on transplantation in the United States.
        Transplant Rev (Orlando). 2013; 27: 50-56
        • Wong R.J.
        • Aguilar M.
        • Cheung R.
        • Perumpail R.B.
        • Harrison S.A.
        • Younossi Z.M.
        • et al.
        Nonalcoholic steatohepatitis is the second leading etiology of liver disease among adults awaiting liver transplantation in the United States.
        Gastroenterology. 2015; 148: 547-555
        • Perumpail R.B.
        • Wong R.J.
        • Scandling J.D.
        • Ha L.D.
        • Todo T.
        • Bonham C.A.
        • et al.
        HCV Infection is Associated with Lower Survival in Simultaneous Liver Kidney Transplant Recipients in the U.S.
        Clin Transplant. 2015; 29: 920-926
        • Uemura T.
        • Ramprasad V.
        • Hollenbeak C.S.
        • Bezinover D.
        • Kadry Z.
        Liver transplantation for hepatitis C from donation after cardiac death donors: an analysis of OPTN/UNOS data.
        Am J Transplant. 2012; 12: 984-991
        • Hoehn R.S.
        • Singhal A.
        • Wima K.
        • Sutton J.M.
        • Paterno F.
        • Steve Woodle E.
        • et al.
        Effect of pretransplant diabetes on short-term outcomes after liver transplantation: a National cohort study.
        Liver Transpl. 2015; 35: 1902-1909
        • Goldberg D.S.
        • Ruebner R.L.
        • Abt P.L.
        The Risk of End-Stage Renal Disease Among Living Donor Liver Transplant Recipients in the United States.
        Am J Transplant. 2015; 15: 2732-2738
        • Wong R.J.
        • Aguilar M.
        • Gish R.G.
        • Cheung R.
        • Ahmed A.
        The impact of pretransplant hepatic encephalopathy on survival following liver transplantation.
        Liver Transpl. 2015; 21: 873-880
        • Dugum M.
        • Hanouneh I.
        • Lopez R.
        • Aucejo F.
        • Eghtesad B.
        • Zein N.
        Hepatocellular Carcinoma in the Setting of Chronic Hepatitis B Virus Infection: Tumor Recurrence and Survival Rates After Liver Transplantation.
        Transplant Proc. 2015; 47: 1939-1944
        • Bambha K.M.
        • Dodge J.L.
        • Gralla J.
        • Sprague D.
        • Biggins S.W.
        Low, rather than high, body mass index confers increased risk for post-liver transplant death and graft loss: risk modulated by MELD.
        Liver Transpl. 2015; 21: 1286-1294
        • Lapeyre-Mestre M.
        • Grolleau S.
        • Montastruc J.L.
        Adverse drug reactions associated with the use of NSAIDs: a case/noncase analysis of spontaneous reports from the French pharmacovigilance database 2002–2006.
        Fundam Clin Pharmacol. 2013; 27: 223-230
        • Lapeyre-Mestre M.
        • de Castro A.M.
        • Bareille M.P.
        • Del Pozo J.G.
        • Requejo A.A.
        • Arias L.M.
        • et al.
        Non-steroidal anti-inflammatory drug-related hepatic damage in France and Spain: analysis from national spontaneous reporting systems.
        Fundam Clin Pharmacol. 2006; 20: 391-395
        • Conforti A.
        • Opri S.
        • D’Incau P.
        • Sottosanti L.
        • Moretti U.
        • Ferrazin F.
        • et al.
        Adverse drug reaction reporting by nurses: analysis of Italian pharmacovigilance database.
        Pharmacoepidemiol Drug Saf. 2012; 21: 597-602
        • Engholm G.
        • Ferlay J.
        • Christensen N.
        • Bray F.
        • Gjerstorff M.L.
        • Klint A.
        • et al.
        NORDCAN – A Nordic tool for cancer information, planning, quality control and research.
        Acta Oncol. 2010; 49: 725-736
        • Kalseth J.
        • Halvorsen T.
        • Kalseth B.
        • Sarheim Anthun K.
        • Peltola M.
        • Kautiainen K.
        • et al.
        Cross-country comparisons of health-care costs: the case of cancer treatment in the Nordic countries.
        Health Policy. 2014; 115: 172-179
        • Annertz K.
        • Anderson H.
        • Palmer K.
        • Wennerberg J.
        The increase in incidence of cancer of the tongue in the Nordic countries continues into the twenty-first century.
        Acta Otolaryngol. 2012; 132: 552-557
      11. Surveillance, Epidemiology, and End Results Program. Available at: <http://seer.cancer.gov/>.

        • Eggert T.
        • McGlynn K.A.
        • Duffy A.
        • Manns M.P.
        • Greten T.F.
        • Altekruse S.F.
        Fibrolamellar hepatocellular carcinoma in the USA, 2000–2010: a detailed report on frequency, treatment and outcome based on the Surveillance, Epidemiology, and End Results database.
        United European Gastroenterol J. 2013; 1: 351-357
        • Amini N.
        • Ejaz A.
        • Spolverato G.
        • Kim Y.
        • Herman J.M.
        • Pawlik T.M.
        Temporal trends in liver-directed therapy of patients with intrahepatic cholangiocarcinoma in the United States: a population-based analysis.
        J Surg Oncol. 2014; 110: 163-170
        • Altekruse S.F.
        • Henley S.J.
        • Cucinelli J.E.
        • McGlynn K.A.
        Changing hepatocellular carcinoma incidence and liver cancer mortality rates in the United States.
        Am J Gastroenterol. 2014; 109: 542-553
        • Altekruse S.F.
        • McGlynn K.A.
        • Reichman M.E.
        Hepatocellular carcinoma incidence, mortality, and survival trends in the United States from 1975 to 2005.
        J Clin Oncol. 2009; 27: 1485-1491
        • Ahmed F.
        • Perz J.F.
        • Kwong S.
        • Jamison P.M.
        • Friedman C.
        • Bell B.P.
        National trends and disparities in the incidence of hepatocellular carcinoma, 1998–2003.
        Prev Chronic Dis. 2008; 5: A74
        • Shaya F.T.
        • Breunig I.M.
        • Seal B.
        • Mullins C.D.
        • Chirikov V.V.
        • Hanna N.
        Comparative and cost effectiveness of treatment modalities for hepatocellular carcinoma in SEER-Medicare.
        Pharmacoeconomics. 2014; 32: 63-74
        • Yang D.
        • Hanna D.L.
        • Usher J.
        • LoCoco J.
        • Chaudhari P.
        • Lenz H.J.
        • et al.
        Impact of sex on the survival of patients with hepatocellular carcinoma: a surveillance, epidemiology, and end results analysis.
        Cancer. 2014; 120: 3707-3716
        • Hyder O.
        • Dodson R.M.
        • Nathan H.
        • Herman J.M.
        • Cosgrove D.
        • Kamel I.
        • et al.
        Referral patterns and treatment choices for patients with hepatocellular carcinoma: a United States population-based study.
        J Am Coll Surg. 2013; 217: 896-906
        • Parikh A.A.
        • Ni S.
        • Koyama T.
        • Pawlik T.M.
        • Penson D.
        Trends in the multimodality treatment of resectable colorectal liver metastases: an underutilized strategy.
        J Gastrointest Surg. 2013; 17: 1938-1946
        • Backus L.I.
        • Gavrilov S.
        • Loomis T.P.
        • Halloran J.P.
        • Phillips B.R.
        • Belperio P.S.
        • et al.
        Clinical Case Registries: simultaneous local and national disease registries for population quality management.
        J Am Med Inform Assoc. 2009; 16: 775-783
        • Backus L.I.
        • Boothroyd D.B.
        • Phillips B.R.
        • Belperio P.
        • Halloran J.
        • Mole L.A.
        A sustained virologic response reduces risk of all-cause mortality in patients with hepatitis C.
        Clin Gastroenterol Hepatol. 2011; 9 (e1): 509-516
        • Kim S.
        • Maynard E.C.
        • Shah M.B.
        • Daily M.F.
        • Tzeng C.W.
        • Davenport D.L.
        • et al.
        Risk factors for 30-day readmissions after hepatectomy: analysis of 2444 patients from the ACS-NSQIP database.
        J Gastrointest Surg. 2015; 19: 266-271
        • Bartlett E.K.
        • Meise C.
        • Roses R.E.
        • Fraker D.L.
        • Kelz R.R.
        • Karakousis G.C.
        Morbidity and mortality of cytoreduction with intraperitoneal chemotherapy: outcomes from the ACS NSQIP database.
        Ann Surg Oncol. 2014; 21: 1494-1500
        • Talwalkar J.A.
        • Oxentenko A.S.
        • Katzka D.A.
        Health care-delivery research-training opportunities in gastroenterology and hepatology.
        Gastroenterology. 2014; 146: 878-883