Use of Epidemiological Databases for health research Arindam Basu [email protected] September 28, 2016 

Roadmap of this talk: Start with Two examples Breast cancer incidence before and after 2003 Construction of a polygenic risk score for heart disease based on GWAS Use of Databases for Health Services Research Opportunities and Caveats Questions? 

Use case 1: Trend of Breast cancer before & After HRT Peak in 2003 Could only be possible with IARC datasets http://ci5.iarc.fr Download the entire data set Filter the data to show only needed years (1996-2007, Breast Cancer, Women 45-85 years) Eight registries (US, Canada, Australia, UK, New Zealand, Japan, France, Netherlands) Analyse peak Breast Cancer before & after peak in HRT usage 

Results //Show the figures 

Use case 2: Construction of a Polygenic Risk Score For Cardiovascular Illness’ Quality of Life Score Genetic Epidemiology study Download entire database from ebi.ac.uk https://www.ebi.ac.uk/gwas/docs/downloads 

Limitations of RCTs & Primary Studies Small sample size that often under-represents Focus on short-term efficacy and safety in a controlled environment Far from routine clinical Fails to answer the more relevant questions that face doctors and patients 

Data in Administrative Databases Made up of the automated electronic recording of Filled prescriptions, Professional services, and Hospitalizations Collected routinely for the payment and administration of health services 

Advantages of Electronic medical records Contain detailed clinical information, patients’ reports of symptoms, the findings of physical examinations, and the results of diagnostic tests 

Advantages of Large Databases Allows the study of rare events Study real-world effectiveness Utilization patterns Relatively low cost and without long delays 

Uses in Health Services Research Studies of the adherence to chronic therapy in routine care. Research using Medicare and Medicaid database studies showed that Adherence to statin therapy among poor elderly patients was below 60% days covered) after 6 months Continued to decline afterward 

Value of Health Care Utilization databases Evaluating the clinical and economic effects of drug reimbursement policy changes They measure actual utilization and economic outcomes accurately They are broadly representative They are large enough to detect small changes in major clinical outcomes 

Value proposition of administrative databases Study the safety of drugs They include large numbers of patients for long periods of time Useful attribute for the study of rare events Very little risk of recall bias when used in epidemiological studies 

Access to Databases are becoming increasingly difficult All jurisdictions or database owners require careful attention to data privacy Minimum requirement for drug utilization research is the availability of a pharmacy dispensing database, which allows descriptive analyses of drug use Keep the linking process with the original patient identifiers physically separated from the data analysis files that instead use study-specific coded ID numbers very small cell sizes in highly cross-classified data may allow the identification of individuals 

Privacy regulations threaten access & this is sad Health Insurance Portability and Accountability Act (HIPAA) has constrained the availability of health care data for uses other than the direct care of patients Europe and Canada have imposed similar problems for researchers modern methods of data anonymization, coupled with oversight by institutional review boards, can make such research possible while adequately protecting the privacy of patients Societal benefits of such database research in pharmacoepidemiology, ranging from quicker ascertainment of drug risks to a more accurate depiction of the quality of use and cost-effectiveness of comparable medications, are considerable, whereas the risks of properly authorized use of such data are negligible 

What to do when we have large enough sample size Use Confidence limits provide more information than simple P values CI more useful in describing the precision of statistical estimation without suggesting any notion of meaningful difference Pay attention to whether that difference is clinically meaningful 

Large Health Care Databases Enormous growth in the use of large health care databases Large data sets Large sample size Allows conducting longitudinal studies Objective measures are possible Minimize Bias 

Summary Large databases can be gainfully used in Epidemiological Research While using large databases, think of the purpose for which you are using them When analysing large databases for Epidemiological research think in terms of confidence intervals and relevance of the findings