Electronic case report forms and electronic data capture within clinical trials and pharmacoepidemiology

Aims Researchers in clinical and pharmacoepidemiology fields have adopted information technology (IT) and electronic data capture, but these remain underused despite the benefits. This review discusses electronic case report forms and electronic data capture, specifically within pharmacoepidemiology and clinical research. Methods The review used PubMed and the Institute of Electrical and Electronic Engineers library. Search terms used were agreed by the authors and documented. PubMed is medical and health based, whereas Institute of Electrical and Electronic Engineers is technology based. The review focuses on electronic case report forms and electronic data capture, but briefly considers other relevant topics; consent, ethics and security. Results There were 1126 papers found using the search terms. Manual filtering and reviewing of abstracts further condensed this number to 136 relevant manuscripts. The papers were further categorized: 17 contained study data; 40 observational data; 27 anecdotal data; 47 covering methodology or design of systems; one case study; one literature review; two feasibility studies; and one cost analysis. Conclusion Electronic case report forms, electronic data capture and IT in general are viewed with enthusiasm and are seen as a cost‐effective means of improving research efficiency, educating participants and improving trial recruitment, provided concerns about how data will be protected from misuse can be addressed. Clear operational guidelines and best practises are key for healthcare providers, and researchers adopting IT, and further work is needed on improving integration of new technologies with current systems. A robust method of evaluation for technical innovation is required.


Introduction
Information technology (IT) provides a fast and efficient way to collect scientific and clinical data and has become the most effective way to collaboratively share data. The benefits have underpinned the incremental introduction of electronic patient records in healthcare organizations which has been suggested as the principal reason for the increasing allocation of healthcare industry funding to IT; from 2% of total revenue, in the 1990s, to 5-7% in recent years [1]. This in turn has contributed to investment in the use of IT and electronic case report forms (eCRFs) in clinical research. Whilst these systems are designed and used differently, they share a common goal of storing, and communicating in a safe and confidential way private clinical data in a structured format [2]. Pharmacoepidemiology and clinical research have undoubtedly benefitted from IT; however, developments in these areas have continued to lag behind the healthcare sector, with investment limited due to various concerns. Reasons cited for not further using IT in research include: technical issues in setting up infrastructure, financing and maintaining the newest technology, and ethical fears [3]. Additionally, different funding streams and personnel involved in development of electronic patient records used for healthcare purposes, and those used for data capture for research, make it difficult to integrate solutions that would satisfy both aims. The objectives of both types of system are often different, which can also lead to conflicts. Different regulatory processes govern systems used in routine healthcare and research. However, clinical research relying on IT and electronic data capture (EDC) often depends on interfacing with healthcare IT systems, which generally comprise numerous dissimilar software systems and storage formats for storing patient data. Clinical research also often operates over large geographical areas, incorporating several different healthcare providers, further compounding challenges when interfacing with diverse local systems. Although there is a drive towards IT unification in the National Health Service primary care practises and hospital trusts in the UK are under no obligation to use collaborative IT systems or storage formats, nor are they required to make these data available for research purposes. While the need to exploit healthcare data for research to cost effectively drive healthcare improvements has never been greater, it is largely for these reasons that the task of collecting, storing and amalgamating health service data is likely to become increasingly difficult in the future.

Objective
The objective of this review is to assess the advantages and disadvantages of eCRF and EDC technologies in pharmacoepidemiology and clinical research, and to explore where further research should be best directed. For the purpose of this paper the term eCRF will refer to a system used to capture clinical data for research and EDC will refer to the generic process of data capture.

Methods
A literature review was conducted to identify articles pertaining to pharmacoepidemiology (drug epidemiology) and clinical research, and their use of eCRFs and EDC. Whilst the use of IT in routine healthcare is increasingly commonplace, the emphasis of this review was on the use of EDC and eCRFs in the conduct of clinical research. Common themes relating to these topics emerged covering a broad range of issues including technical and practical matters, consent, ethics, and security. PubMed and the Institute of Electrical and Electronic Engineers (IEEE) libraries were searched using to cast a wide net over the subject area; electronic case report form, eCRF, electronic data capture, and electronic data collection. Filters were applied to search terms to condense results to relevant articles (see Appendix). The search was conducted between 2014 and 2015 with a final analysis of the literature completed in August 2016. PubMed is a clinical library while IEEE is technology based.
All returned abstracts were read and articles deemed irrelevant to eCRFs and EDC, or articles that did not involve pharmacoepidemiology or clinical research, were excluded. Unlike clinical studies, IT has no universally accepted quality scoring system for academic papers. Therefore, it was decided that any published and peer reviewed article that was returned from the IEEE or PubMed search would be included. Exceptions to this were where there was an overt conflict of interest or the journal was not available in English. Figure 1 depicts a flow diagram of the review process. The authors endeavoured to adhere fully to the PRISMA checklist [4] in structuring this review; however, the nonstandard output of technical papers made this impractical. The included papers were sorted by relevance, and categorized according to whether they contained opinion or data. Papers reporting data included anecdotal data, observational data from selected data sources, observational data in population-based studies, prospective observational data and experimental data such as clinical trials. Papers were analysed to identify reported positive and negative aspects of the IT tools being discussed.

Results
A total of 1126 papers were returned from all search topics. After review and consideration, 136 manuscripts were deemed relevant to the review. Each topic was further separated into manuscript types. There were 17 papers documenting a study or clinical trial that used EDC where the system was the primary focus of the manuscript; 40 papers discussed observational studies comparing or evaluating EDC; 27 papers contained anecdotal evidence or opinion regarding EDC; 47 papers detailed EDC models or designs. There was one literature review, one cost benefit analysis, two feasibility studies, and one case study comparing the use of EDC in five studies (Table 1). During this review, papers were further discarded that were found to be of poor overall quality or adding little to the topic. For a list of all included publications see Table 2.
Research has been conducted into ways to maximize data accuracy and efficiency using IT. Trials have taken data from patient's electronic medical records (EMRs) and transferred these directly into eCRFs. The cost savings, quality improvements, and reduction of data entry errors, were significant [5][6][7]. Whilst not all required data ARE available from the patient EMR, studies have found varying results with as much as 69% of data required being found and used to prepopulate trial eCRFs [8]. Discussions around the design and theoretical modelling of EMRs, eCRFs and ECD were prevalent within the included papers [9][10][11][12][13][14][15][16][17]. The electronic systems reported vary in quality, with some being used in mock environments and others being purely theoretical. Commercial software packages are available, but are generally not cost effective and in some circumstances it is unclear who owns the data entered into them [18][19][20]. Observational studies have compared paper based systems against EDC or canvassed opinion on the use of EDC systems [21][22][23][24]. These papers were overwhelmingly in favour of EDC as long as security could be maintained.
Obtaining patient consent is an ethical necessity, and up until recently, has almost always required a physical signature. Varnhagen et al. [25] considered obtaining informed consent online and questioned whether it is ethical to obtain consent electronically. Recently, electronic consent has been accepted by the National Health Service as a viable alternative to a written signature [26]. This review found one trial where consent had successfully been captured online [27]. Collecting participant consent electronically is a novel, and largely unexplored, method that invites further innovation. There are ethical implications of conducting research entirely online. IT is advancing faster than ethical review panels can address and there is a need for greater ethical consideration of conducting research online and how we share data between IT systems and within organizations [28][29][30][31][32][33][34][35]. Government attempts to legislatethe Health Insurance and Portability Act [36] in the USA, and the Data Protection Act [37] in the UK, have had little impact on alleviating public scepticism [38]. Patient privacy is critical and despite the well-intentioned zeal for the mass adoption of IT within healthcare, serious security concerns remain [39,40]. However, patients are open to technology being used to store their medical data if trust and privacy concerns can be addressed [41].   Feasibility studies 2 1 . 5 Literature review 1 0 . 7 Cost-benefit analysis 1 0 . 7 Case study 1 0 . 7     Clinical research and pharmacoepidemiology often involve interdisciplinary research. This not only means that various researchers deal with different data sources and formats but also that they have different workflows and organizational structures. As a consequence, there are no off-the-shelf solutions to facilitate this. This often results in individual solutions being developed that, over time, evolve and are ultimately difficult to maintain [42]. Unfortunately, it is often much easier to change time points, interventions, and assessment tools on paper than it is to suddenly change the programming of a computerized system. The reality demands future IT systems be flexible and adaptable with more automation [43].

Advantages and disadvantages
There are distinct advantages to EDC in research and pharmacoepidemiology. However, there are pragmatic concerns that need to be addressed. The role of clinical research and pharmacoepidemiology is to improve healthcare by generating and providing access to high quality data. Due to the limitations of paper based records this is not possible with the status quo [44]. The objectives of ECD are to reduce medical errors, improve communication between healthcare providers, collect information for educational and research purposes and to gather complete and accurate data whilst avoiding duplication.
EDC's distinct advantage over paper-based systems of research is that it is able to detect protocol violations and data outside the normal range at the time of entry and not days, weeks or months after. EDC systems have been shown to improve the quality of clinical trials, halt the development of ineffective or unsafe drugs earlier, reduce unnecessary work, reduce cost, and accelerate time to market of new drugs [45][46][47][48][49][50][51]. There are also benefits in relation to data quality, performance, productivity and costs in clinical trial management [52][53][54][55][56]. Observational data suggest that it is now considered a preferred method of data capture in clinical research [57][58][59]. It is well accepted by users and has been shown to contribute to patient empowerment, allowing them to be more engaged in research and to take direct control of their own data [60][61][62]. By contrast, paper-based questionnaires can suffer from incomplete forms, questions being answered twice or skipped questions. Paper forms are considered time consuming, require dual checking, and data cleansing [63], whereas EDC can alert people to missing answers before any attempt to proceed, and can be easily incorporated into electronic health records. Remote data collection offers the additional advantage of convenience to patients, particularly those who are incapacitated or live far from the nearest clinic [47], and may provide a safer environment for questionnaires than paper-based methods eliciting the answers to potentially sensitive questions [64,65].
Despite the advantages, EDC has not been universally accepted. Perceived disadvantages and concerns regarding EDC include: a lack of available technical support, a lack of investigator motivation, complexity of installation, maintenance of software, high initial investment cost, and complexity of use [66,67]. Reliable data handling methods, effective project management, and expert technical architecture and infrastructure are all key factors for successful implementation, and should not be underestimated [68]. There are concerns over patient privacy and the need for computer literacy, which may affect generalisability of any research findings [60]. Study retention is considered to be higher where there is direct patient interaction because of the explicit alignment of patient incentives; the patient learns about the study directly, understands what is required, self-consents to participate, and then self-reports study information [69]. Jamison et al. [70] found better rates of compliance with electronic patient reported outcomes (PROs) than paper based PROs. Despite data suggesting benefits of EDC use, Alexander [71] reports that physicians lack motivation and will only use structured electronic records if the system reduces overhead while at the same time minimizing their work load. In the UK, it has been suggested that development of these technologies suffers from the lack of a clear national direction towards unifying clinical and medical data, with no common format for all data systems. Not only would EDC benefit clinical research, but pharmacovigilance and drug safety regulation could also be improved [72].

Discussion
IT and how it is used in pharmacoepidemiology and clinical research is a relatively new field with no substantial guidelines in place and few recommendations. There is a consensus that EDC has clear benefits for use in research but there are fears over security and data protection which must be addressed. IT offers an opportunity to improve pharmacoepidemiology and clinical research and to facilitate the continual improvement of healthcare. If the use of IT in research is to succeed fully, change is required: specifically, investment in infrastructure and the provision of support for integration of interoperable systems. Further efforts will be essential to alleviate healthcare providers and users legitimate concerns regarding IT. Policy makers will need to find ways to supply adequate financial resources to IT to counter a historical lack of investment within the public sector.
Healthcare providers and researchers require a governance-led support network of technology experts to assist in integrating ever more complex systems and providing guidance on compliance and security. IT security is a challenging and fast moving field and requires careful consideration. There is a need for clearer and more consistent policies and more trained data managers, software architects and semantic web specialists working in medical research groups. These architectures will need ongoing support from a robust legal system protecting patient privacy.
Furthermore, the exchange of information between systems is essential. Data format differences need to be resolved, and a solution found for interoperability between healthcare systems. Motivating software vendors, healthcare providers and researchers to agree on a common path will be difficult but worthwhile endeavour. Future technical development needs to focus on creating adaptable and generic software that can be tailored to specific trial needs without major re-development.

Limitations
This work has several limitations. Firstly, a publication bias is very likely as less successful IT projects are unlikely to be reported in published literature. This review only searched the IEEE library and the PubMed database and we did not include papers in non-English languages. In addition, researchers from low-income countries are known to have lower publication rates. The relative novelty of the field means that evaluation studies, in particular, are missing and rapid developments in the field may not yet have been published at the time of conducting the literature search. There are currently no widely accepted methods to evaluate technical publications in the same way as has been developed for reports of clinical trials, for example. Therefore, subjective interpretation had to be used to decide if a journal was of sufficient quality to be referenced. The authors took steps to avoid selection and objectivity bias by including all peer reviewed and published articles. The only exceptions authors made were where there was an overt conflict of interest, or the journal was not available in English. This review aimed to capture the full range of reported advantages and benefits of IT use. It did not measure the relative frequency or impact of individual factors of the utility of EDC and eCRFs. Despite the limitations detailed above, the authors believe this review to be an unbiased appraisal of publications on EDC and eCRFs in pharmacoepidemiology and clinical research.

Conclusion
It is apparent from the results of this literature review that the following areas would benefit from further development: • Clearer legislation and operational frameworks governing electronic health records. • Guidelines and best practises for researchers to follow in the use of IT and EDC. • Standard methods of reporting and evaluating technical innovation to facilitate comparison.
Regardless of the challenges, it is the imperative that healthcare organizations ensure that patients receive safe medications.
Effective clinical research and pharmacoepidemiology are essential to this process.