AR: Comparative effectiveness raging for diagnostic evidence
Comparative effectiveness research (CER) will likely continue to be an important effort in generating evidence regarding the most effective diagnostic and therapeutic procedures, according to an article published in the September edition of Academic Radiology.
“Both outcomes researchers and informaticians are concerned with information and data,” wrote Nabile M. Safdar, MD, from the Sheikh Zayhed Institute of Pediatric Surgical Innovation from the department of diagnostic imaging at the Children’s National Medical Center in Washington, D.C., and colleagues. “As such, some of the central challenges to conducting successful comparative effectiveness research can be addressed with informatics solutions.”
The authors reviewed informatics tools which address how data in CER are enriched, stored, shared and analyzed. Retrospective studies, or those that require the use of clinical records do not always have the ability to define the method of data collection from research subjects. To alleviate these problems, there have been efforts to standardize and enrich the content of data which serve as input for clinical and research uses, particularly in imaging and its reporting.
“Although lexicons and structured reporting efforts create additional constraints at the time of report creation to maximize the ease with which information in such reports can later be accessed, the alternative is to utilize automated techniques which create structure and extract information from otherwise unstructured text, such as with natural language processing (NLP),” Safdar and colleagues wrote. “The utilization of NLP techniques are a recognition that traditional prose clinical reports are likely to persist for some time and remain the preferred method of communicating results. NLP techniques have been used to determine whether findings and recommendations are present in radiology reports, categorize reports as normal or abnormal and to identify specific imaging features in reports.
“Such tools can be leveraged to empower studies of comparative effectiveness when retrospective reviews of large amounts of uncoded text data are desired,” the authors wrote.
In the context of CER, one of the most fruitful developments has been the advent of data warehouses, according to the authors. “Although production clinical systems are typically not well-suited for large, complicated queries that may burden a database, a data warehouse created for nonclinical purposes can allow CER researchers with appropriate permissions to access information that is not easily accessible in the EMR, radiology information system, pathology information system and other clinical and business systems.”
In context of sharing, file transfer protocol sites or sending physical hard drives through traditional mail may not have sufficient security or scale well. “Fortunately, there has been increasing activism in the research community to provide informatics infrastructure for data sharing and several national organizations have made their data registries available to researchers online. Increasingly, organizations that fund research are requiring recipients of funding to provide a data sharing plan,” the authors wrote adding that data regarding outcomes, claims, clinical information, or in some cases, actual imaging data from large research studies can be also be shared through data registries that are often accessible online.
According to the authors, informatics support for CER will likely be driven by the needs of clinical communities. “Those in clinical and research communities are hopeful that standards and efforts will translate in a more widespread fashion through both clinical and research realms to best inform our decisions regarding the most effective diagnostic and therapeutic interventions.
“In the same way that outcomes researchers use diverse methods to assign meaning to store, share and analyze their data, there is a diverse arsenal of informatics tools designed to meet their needs in these respective areas,” the authors concluded. “Because there is further collaboration between outcomes researchers and informaticians, the arsenal of tools developed with the goal of enabling improvement in patient outcomes is likely to increase in size and sophistication.”
“Both outcomes researchers and informaticians are concerned with information and data,” wrote Nabile M. Safdar, MD, from the Sheikh Zayhed Institute of Pediatric Surgical Innovation from the department of diagnostic imaging at the Children’s National Medical Center in Washington, D.C., and colleagues. “As such, some of the central challenges to conducting successful comparative effectiveness research can be addressed with informatics solutions.”
The authors reviewed informatics tools which address how data in CER are enriched, stored, shared and analyzed. Retrospective studies, or those that require the use of clinical records do not always have the ability to define the method of data collection from research subjects. To alleviate these problems, there have been efforts to standardize and enrich the content of data which serve as input for clinical and research uses, particularly in imaging and its reporting.
“Although lexicons and structured reporting efforts create additional constraints at the time of report creation to maximize the ease with which information in such reports can later be accessed, the alternative is to utilize automated techniques which create structure and extract information from otherwise unstructured text, such as with natural language processing (NLP),” Safdar and colleagues wrote. “The utilization of NLP techniques are a recognition that traditional prose clinical reports are likely to persist for some time and remain the preferred method of communicating results. NLP techniques have been used to determine whether findings and recommendations are present in radiology reports, categorize reports as normal or abnormal and to identify specific imaging features in reports.
“Such tools can be leveraged to empower studies of comparative effectiveness when retrospective reviews of large amounts of uncoded text data are desired,” the authors wrote.
In the context of CER, one of the most fruitful developments has been the advent of data warehouses, according to the authors. “Although production clinical systems are typically not well-suited for large, complicated queries that may burden a database, a data warehouse created for nonclinical purposes can allow CER researchers with appropriate permissions to access information that is not easily accessible in the EMR, radiology information system, pathology information system and other clinical and business systems.”
In context of sharing, file transfer protocol sites or sending physical hard drives through traditional mail may not have sufficient security or scale well. “Fortunately, there has been increasing activism in the research community to provide informatics infrastructure for data sharing and several national organizations have made their data registries available to researchers online. Increasingly, organizations that fund research are requiring recipients of funding to provide a data sharing plan,” the authors wrote adding that data regarding outcomes, claims, clinical information, or in some cases, actual imaging data from large research studies can be also be shared through data registries that are often accessible online.
According to the authors, informatics support for CER will likely be driven by the needs of clinical communities. “Those in clinical and research communities are hopeful that standards and efforts will translate in a more widespread fashion through both clinical and research realms to best inform our decisions regarding the most effective diagnostic and therapeutic interventions.
“In the same way that outcomes researchers use diverse methods to assign meaning to store, share and analyze their data, there is a diverse arsenal of informatics tools designed to meet their needs in these respective areas,” the authors concluded. “Because there is further collaboration between outcomes researchers and informaticians, the arsenal of tools developed with the goal of enabling improvement in patient outcomes is likely to increase in size and sophistication.”