|Home | About | Journals | Submit | Contact Us | Français|
At its core, the practice of medicine is an information intensive endeavor. Most of what physicians do involves the collection, review and management of information. Examples of such activities include obtaining and recording patient information, consulting colleagues, reading the scientific literature, planning diagnostic procedures, devising strategies for patient care, interpreting tests, and conducting research. The ever-increasing biomedical knowledge-base that must be considered in order to deliver optimal patient care only adds to the challenges facing medicine today.
Successfully addressing these challenges in order to deliver the best healthcare possible requires not only the existence of valid and generalizable datasets derived from systematic basic, clinical and epidemiologic research efforts, but also the ability to apply the knowledge derived from these research efforts at the point-of-care. It is easy to understand, therefore, why the field of biomedical informatics, a field that is the concerned with collecting, managing and optimally using information in healthcare and biomedicine, is critical to the current and future practice of medicine and the study of healthcare outcomes that result from such practice 1, 2.
Biomedical informatics approaches and related health information technology (Health IT) platforms are key to enabling knowledge-driven healthcare and practice improvement initiatives based on a solid research foundation. Similar biomedical informatics approaches and resources are also critical to advancing outcomes research. Indeed, the emergence of such technologies such as electronic health records (EHRs), clinical data repositories, and research-specific data management systems are already transforming the way we practice medicine and conduct research. This transformation is being further advanced by federally directed funding and research infrastructure development efforts 3, 4.
In the sections that follow, we provide an overview of how Biomedical Informatics and Health IT processes and tools can impact the conduct of research and the delivery of evidence-based healthcare from the authors’ perspective. Given the current state of development in this area, we introduce a conceptual framework by which clinicians, researchers, and the healthcare community at-large can optimally leverage biomedical informatics processes and tools to facilitate outcomes research and drive improvements in healthcare outcomes. We conclude by proposing a course of action intended to support greater collaboration between the clinical, research, and biomedical informatics communities to achieve the goals set forth.
Outcomes research, “seeks to understand the end results of particular health care practices and interventions,” on individual patients and populations 5. This often includes the evaluation of economic impacts linked to health outcomes, such as cost effectiveness, cost utility, and comparative effectiveness. Such research involves a range of data collection and aggregation methods including drawing upon primary studies as well as collecting data de novo for such research 6. The conduct of outcomes research therefore can be thought of as involving an information-cycle that includes collecting data about healthcare practices and patient/population level outcomes, analyzing those data, and the reporting on the findings.
When considering the types of information relevant to the conduct of outcomes research and the practice of evidence-based medicine, two major types are relevant: (1) patient-specific information – or information generated during the care of patients such as numerical values, free text, imaging information etc. as might be found in an EHR; and (2) knowledge-based information – or information derived from the scientific medical literature that is based on biomedical and healthcare research. As the collection, storage, retrieval, and optimal use of these types of information constitute the material focus of biomedical informatics, it becomes clear why the intersection of biomedical informatics and outcomes research is critical to the advancement of medical science and practice.
Indeed, the efficient conduct of outcomes research requires access to robust clinical and population-level datasets as well as existing research and knowledge datasets in order to evaluate relevant metrics such as procedural complications, days of hospitalization, health status, and mortality. Moreover, the results of research can be used to develop system-level interventions designed to facilitate the improvement and/or optimization of healthcare policies and practice, often by leveraging clinical informatics platforms such as EHRs and clinical decision support systems.
Biomedical informatics can be seen as key a key enabler of the outcomes research cycle. One way to explore Biomedical Informatics’ role is in terms of the resources and platforms developed and used for contemporary clinical practice and research practice. Examples of such systems including EHR platforms, clinical data warehouses, and research information management platforms, all of which are becoming increasingly available in the healthcare environment.
At each stage in the research process, general purpose or research-specific IT systems may be of utility. Payne et al. provided a model for clinical research in general, and as a subset of clinical research, the same can be said of the applicability of this model for outcomes research 3. Examples of general and clinical information systems that are able to support the conduct of clinical research include:
In addition to the preceding general-purpose and clinical IT systems, research-specific IT systems have been developed that include:
Numerous reports have concluded that the use of such tools and platforms can lead to increased data and research quality 7, 11, 32, 35. Furthermore, the use of informatics platforms across multi-site studies has been shown to increase the efficiency and efficacy of such team-science endeavors 32, 34, 36–38. The ability to use IT in support of clinical research relies on the ability to collect, store and analyze data. Examples of information systems that are becoming increasingly common include Electronic Data Capture (EDC) systems as well as integrated clinical trials management systems (CTMS) that target multiple aspects of the clinical trials process, including EDC, financial management, data quality assurance and research participant tracking 39, 40. Indeed, the use of such IT for clinical research is growing rapidly, and it is projected that nearly 50% of all studies will soon utilize information management technologies 41.
A common problem for the outcomes researcher is the aggregation of data from varied and disparate information resources. Resources that bring together clinical data from otherwise inaccessible and non-integrated medical record, laboratory, hospital-based and other healthcare enterprise systems throughout a given region in order to facilitate more efficient and effective care processes are therefore becoming an important part of the research IT solution set.
The growing availability of such integrative datasets via the development of regional health information exchange (HIE) organizations, is one area of particular relevance to this discussion. While HIEs are in various stages of development across the country and some have struggled to establish themselves, there are early examples of successful HIE models that have enabled the use of population level data for public benefit. Examples include the use of HIEs to significantly improve reporting of common diseases over standard means, allow surveillance for disease outbreaks, and facilitate population level epidemiology studies that would otherwise have been difficult if not impossible 42.
Fundamental to the success and utility of such efforts, whether at the regional or even at the individual institutional level where multiple data sources are common, are the informatics underpinnings that enable the aggregation and integration of data from disparate sources into data repositories. One key element of such Informatics methods underpinning the exchange of data is transactional standards. A commonly employed example is Health Level 7 (HL7) version 2, a health data messaging interchange standard employed to transfer information between systems and which resources like HIEs often employ to exchange information between source and destination systems 43. It is worth noting however, that due to the fact that such standards traditionally define the mechanism of exchanging data but do not define the semantic annotations necessary to ensure a shared understanding of the meaning of such data across platforms or organizations, their adoption alone is not sufficient to enable the seamless exchange of complex data sets. Work is ongoing to address this issue in new versions of such standards such as HL7 version 3 44.
Another key element is the methodologies and approaches used to maintain integrative data repositories of data in resources such as data warehouses. A data warehouse (DW) is a type of database or data repository that is designed to have certain characteristics that enable its use for purposes such as research including the following definitional factors 45–47:
There are many potential uses for a DW 48, 49, but it is the ability to perform longitudinal or episodic queries based upon one or more criteria of interest in support of research activities (e.g., study planning, retrospective data analyses) 45, and the delivery of task or role-specific datamarts to support context-specific access to data sets by other applications or direct query and analysis by authorized end-users that make these resources so useful in the biomedical research domain.
Indeed, within the biomedical domain, numerous reports provide contexts in which data warehouses have been utilized, including:
Along with all of these benefits, there are also challenges associated with the use of DWs, including 45, 54: 1) overcoming regulatory and data privacy and confidentiality concerns; 2) ensuring the provenance and quality of data being included in the DW; 3) implementing and supporting sufficiently robust and timely interfaces between production systems and DWspecific extraction, transformation, and load (ETL) processes; and 4) providing timely access to limited or de-identified data sets for retrospective research or research planning purposes. However, despite such potential limitations and challenges, the use of data warehouse platforms in biomedical settings has been and continues to be associated within major increases in productivity and efficiency surrounding many application scenarios.
Another factor motivating the development of resources that enable connectivity and information exchange within and between major research centers is the increased NIH funding in recent years for research informatics infrastructure. Examples include the Cancer Biomedical Informatics Grid (caBIG), Informatics for Integrating Biology and the Bedside (i2b2), and Clinical and Translational Science Award (CTSA) initiatives 40, 55, 56. These efforts have led to the development of additional resources and approaches to advancing information integration and accessibility to advance all kinds of research.
As previously mentioned, patient or population level data is one key type of information needed to advance evidence-based medicine. The other is knowledge-based information such as that contained in the scientific literature. Here, too, informatics resources such as Medline/Pubmed, the Cochrane database, and other data repositories of medical knowledge including guideline repositories have become critical components of the evidence-driven medicine solution. In addition to relying upon manual review of such resources on an ad hoc basis in the course of hypothesis generation or to answer clinical questions, these repositories have the potential to influence care directly through biomedical informatics approaches.
One such example involves the combination of knowledge-based information with patient-level data in order to drive healthcare support via rules-engines, often as part of Electronic Health Records. Such Clinical Decision Support Systems (CDSS) assesses individual rules and determines their applicability in a specific case or situation 57. At the most basic level, such a rules engine is comprised of two components, a knowledge-base (consisting of rules represented in a computational format), and an inference or execution engine which can reason about incoming data based upon the contents of the knowledge-base in order to generate some form of output such as an instruction set or alert. An additional critical component of a rules engine is the knowledge engineering facility, which provides the capability to curate the contents of the knowledge-base on an automated, semi-automated, or manual basis.
Biomedical Informaticians have contributed various models for representing clinical decision support rules throughout the years. Examples such as the Arden syntax which can be used to write what are called Medical Logic Modules (MLM’s), form the basis of clinical alerting systems 58. Work has also been done on frameworks and systems to represent the knowledge of clinical guidelines. Examples such as the Guideline Interchange Format (GLIF), EON, and EON’s successor, SAGE, allow for computation on guidelines for the purposes of point-of-care decision support, and have been utilized in multiple studies 59, 60. Indeed, products of the EON project have been deployed in a production-level advisory system for the management of hypertensive patients in VA hospitals, named ATHENA 61.
As the preceding descriptions make evident, the biomedical informatics community has generated a broad variety of techniques, platforms, and theoretical models capable of supporting the myriad of information management and analysis activities essential to the conduct of outcomes research and ultimately the practice of evidence-based medicine. However, such physical and methodological resources are often underutilized, usually due to a combination of human-factors, socio-technical issues and technology-based barriers 3, 4. In analogous research domains such as clinical research and bio-molecular translational science, the development of conceptual frameworks that illustrate the interrelationships of critical components of the socio-technical system has been shown to help in advancing the development of informatics solutions to similar issues 19, 62. Therefore we propose the basis for such a model to advance the integration of biomedical informatics and outcomes research. The proposed model builds upon the AHRQ description of outcomes research, which states that, “Outcomes research seeks to understand the end results of particular health care practices and interventions… For clinicians and patients, outcomes research provides evidence about benefits, risks, and results of treatments so they can make more informed decisions” 5.
Building upon this description, we first acknowledge the relevant methodological approaches and operational or research products, as follows:
Spanning the preceding four methodological and operational/research products are a set of enabling biomedical informatics practice areas, which are concerned with the design, application, and evaluation of tools and techniques concerned with:
Taken together, the combination of the four methodological and research/operational products introduced earlier, and the corresponding and enabling informatics practice areas described above, create a framework to facilitate and support the improvement or optimization of healthcare practice and policies, based upon the systems-level understanding of outcomes, risks, and benefits that they afford researchers, practitioners, and policy-makers (Figure 1). It is important to note that this model includes an iterative feedback cycle to ensure that experiences with interventions inform the creation and optimization of healthcare practices and policies including those related to the development of clinical and research informatics solutions.
As reviewed in the preceding sections, the conduct of outcomes research and the corresponding provision of evidence-based medicine is both a complex and information-intensive endeavor. Central to such activities is a systems-level approach to information and knowledge generation, collection, analysis, and dissemination. We believe that the development of a community-accepted conceptual model that advances a collective understanding of the complex interplay between the constituent operational, research, and informatics components is both necessary and advantageous to those working in these domains. In the preceding section, we outlined the basis for such a model in the hopes of catalyzing the further discussion and research required to formalize such a construct. However, to realize the benefit of such a model, and the processes by which it will be fully conceived and validated, a number of important factors must be addressed by our clinical and research communities, namely:
The conduct of outcomes research is an information intensive endeavor and therefore benefits from the application of biomedical informatics approaches, resources, and platforms. It is our contention that a tighter integration of biomedical informatics, clinical care, healthcare policy, and outcomes research can serve to advance improvements in healthcare research and practice. As the preceding overview of the current state of knowledge in these domains illustrates, such integration will require at the most basic level the development of team-science approaches to outcomes research that include clinicians, researchers, and biomedical informaticians. However, the formation and support of such teams will not be possible without addressing some of the barriers and requirements we have summarized in our discussion. By building upon the recent and ongoing advances in biomedical informatics and addressing the issues raised above, there exists a tremendous opportunity to link research, healthcare delivery and policy in such a way as to have direct and demonstrable impact on the health and quality of life of the public.
Dr. Embi’s contributions were supported in part by grants from the NIH/NLM (R01-LM009533) and NIH/NCRR (UL1-RR026314). Dr. Payne’s contributions were supported in part by grants from the NIH/NCI (P01-CA081534, R01-CA134232) and NIH/NCRR (U54-RR024384).
The authors have no disclosures to report in relation to this manuscript.