Presentation at the Workshop on Data Management held at King Saud University Riyad Saudi Arabia on April 30, 2013 by Professor Omar Hasan Kasule Sr. MB ChB (MUK), MPH (Harvard), DrPH (Harvard) Chairman Institutional Review Board and Department of Bioethics King Fahad Medical City Riyadh EM: omarkasule@yahoo.com

INTRODUCTION

· Data includes words, numbers, images, and voice most often in an electronic form.

· Modern information technology handling large and multiple datasets has spawned new ethical issues that researchers dealing with a single research data base in one institution did not face. These issues are: data costs, data ownership, data confidentiality, and patient safety based on data validity.

· These ethical issues arise at the stages of data sourcing / collection, data editing, and data storage and retrieval.

· They also arise in the processes of data sharing and data integration.

USES OF EXISTING DATA SOURCES

· Data from existing data sources for clinical epidemiology[1]^,[2]

· Use of existing data bases for morbidity information: injury[3]^,[4], chronic conditions[5],

· Use of existing data to analyze policy[6]

· Use of existing data for research on specific conditions: heart failure,[7] diabetes[8] [9] [10], HIV[11], asthma[12], knee arthroplasty[13]

· Genome data interpretation using internet and public data bases[14]

· Others: herbal medicine[15], drug safety[16]

QUALITY ISSUES OF EXISTING DATA SOURCES

· Assessing quality of existing data sources[17]

· Assessing coverage and completeness of existing data[18]

· Assessing data sources on occupational fatality[19]

· Comprehensiveness of data sources[20]

· Under registration of death in Thailand[21]

ETHICS OF USING EXISTING DATA SOURCES

· Operational data generated by hospitals, health insurance companies, and administrative units is not collected with due care to ensure research-quality accuracy (accuracy, coverage). The defects of these data bases can be overcome by instituting quality control programs and using multiple sources for cross validation. Is it ethical to have low quality routine operational data that cannot be used for research or valid policy analysis?

· Operational data lies unused in data banks while researchers apply for and get grants to collect new data for their research purposes. Is it ethical to collect new data when we have a lot of routinely collected data?

· Ethical issues in genetic data collection from critically ill patients[22]

· Using human samples for environmental bio-monitoring[23]

· Data ownership and responsible research[24]. Which researcher has a right to use data in the charts.

· Trade in data: can we buy data from subjects or from collectors? Can we sell data for research? Sell/trade/purchase?

· Data ownership?: data from public publications, data in patient charts (epidemiological & clinical analysis), case reports

· The data can be used for research by permission of the legal owner; laws and regulations are not yet clear on this issue because potential owners include the patients, the physicians, and the institutions.

· The issue of ownership leads to another question whether routinely collected data can be sold to researchers. A corollary to this is whether researchers can engage in selling or buying data with other researchers or commercial marketing and advertising agencies.

ACCESS TO MORE DATA BY DATA INTEGRATION AND DATA SHARING 1

· Data integration and data sharing, facilitated by modern information technology, enable access to more data in other institutions for analysis and standard setting.

· Information technology provides the algorithms for fast integration and sharing.

· Both integration and sharing are an ethical imperative to advance knowledge that benefits patients.

· Integration and sharing have been used mostly in genomic sequencing, nuclear mapping, imaging, clinical trials, and organ transplantation research.

DATA INTEGRATION

· Three levels of integration (within species and interspecies) in plant science to gain different understanding[25]

· Comparing and integrating data from various sources[26]

· Algorithm for fast integration of data from various sources[27]

· Integration of distributed data sources[28]

· Integration of data from various sources on adverse effects[29]

· Interoperability and data integration[30]

· Methodology[31]

DATA SHARING 1

· Benefits of data sharing: making data sharing count[32], making data sharing work[33], sharing online data among people with epilepsy[34], medical research data sharing: the public good[35]

· Processes of data sharing: data sharing not simple[36], different practices of data sharing among scientists[37], bottle necks in data sharing[38], data standardization[39]

· Technology of data sharing: electronic data sharing[40], platform for data sharing[41], cryptographic approach to data sharing[42], informatics to enable data sharing[43], Secure data sharing portals[44]

· Policies and regulations: Need for data sharing policy[45], Institute wide data sharing policy[46]

· Examples of data shared: genetic[47], nuclear mapping[48], mass spectrometry[49], neuroimaging[50], neuroimaging data[51], Health care[52], prevention of alcohol related violence[53], Data sharing on lung injury[54], Sharing of clinical trials data[55], Data sharing with e-health[56], Analysis of organ sharing data[57], Sharing autism research data[58]

· Analysis: Post liver transplant obesity based on analysis of shared data[59], Analysis based on data sharing[60]

ETHICAL ISSUES IN DATA SHARING 1

· Data sharing: ethical and practical issues[61]

· Intellectual property, informed consent, privacy, and confidentiality. Codes, standards, policies and mapping at local and international levels are being developed to address these issues.

· Owners of data collected at great expense are reluctant to share with others without proper acknowledgement of intellectual property.

· Informed consent from patients is needed for data sharing unless fully anonymised.

· Data privacy and confidentiality are assured by use of secure data portals and cryptography.

ETHICAL ISSUES IN DATA SHARING 2

· Sharing HIV data: confidentiality and acceptability[62]

· Genetic data: benefits vs risks[63], code for sharing international genetic data[64], Consent for data sharing in genomic research[65], risks in sharing aggregate genetic data[66]

· Intellectual property in data sharing[67]

· Data sharing vs protection of privacy[68], Privacy compliance in data sharing[69], Preserving privacy with data sharing[70]

· neuroimaging data[71], Standardization and risks[72], patient control of neuroimaging data[73]

· Misuse of Data use for dual use[74]

DATA PROCESSING

· Data processing within one research project has its own ethical issues. The data manager could introduce biases, random or non-random, in the processes of adjusting for missing data, data transformation, and creation of derived variables.

· Data processing mistakes underlie several types of bias: misclassification, selection bias, and sampling bias. Any mistakes introduced at this stage will impact the final research results and eventually affect patient safety due to clinical interventions based on false research.

· The usual procedures for data privacy and data confidentiality must be respected. As far as possible personal identifiers should not be accessible except to a few selected members of the research team. The data should be kept locked up or in password protected computers. If the computers are connected online the institution should have policies and software to assure data security.

DATA EDITING AND VALIDATION

· Data editing can lead to biases that will eventually impact on patient safety through wrong research data and conclusions.

· Potential biases in dealing with missing data

· Sampling biases

· Biases in data transformation: qualitative vs quantitative, ordinal vs nominal, continuous vs discrete, creating new variables by combining variables or by mathematical transformations

DATA VALIDATION

· Cross validation using multiple sources[75]^{,[76],[77],[78],[79],[80],[81]}