The impulse provided by public administrations in the early 2000s has resulted in the widespread implementation of health information systems in healthcare organizations. These systems have the capacity to collect, store, manage, and transmit electronic medical records of patients. Additionally, there is an ongoing academic effort to utilize these systems as a source of data for clinical research, as they offer advantages in terms of cost and speed compared to traditional systems.
As a result of this effort, the concept of computable phenotype has arisen. A computable phenotype refers to a search process that is solely supported by computer data, allowing for the identification of a cohort - a group of individuals with shared characteristics. Although this term originated in genomic research to associate genetic markers with phenotypic traits, its utility extends to various types of epidemiological, biomedical, and administrative research.
Searching through records is a common procedure in any database. Healthcare organizations have developed their own search processes since implementing their health information systems. However, creating a computable phenotype requires an assessment to determine the quality and comprehensiveness of the search within the target population. In the absence of standardization, seemingly similar definitions can yield different results when applied to the same dataset, or the effectiveness of the algorithm may vary when applied to a different population or environment.
Consequently, public repositories with computable phenotypes have emerged to promote sharing, reuse, and iterative improvement of the definitions. Standardizing and disseminating definitions are believed to facilitate analytical transparency, promote the use of common data models, enhance quality and consistency, and reduce duplication of efforts.
Reusing a phenotype is more cost-effective than creating a new one, but it necessitates a process of adaptation to the characteristics of the information system and the new target population, as well as subsequent verification and validation to ensure its satisfactory performance.
The validation of a computable phenotype is akin to a diagnostic test, in which its ability to accurately identify the required patients from the health information system is evaluated. Accuracy is measured in terms of predictive values, sensitivity, and specificity. The reference standard can be a more complex computable phenotype, or alternatively, a diagnostic panel in which a group of experts determines the definitive characteristics of each patient using additional data.
To summarize, there is an ongoing collaborative effort to develop computable phenotypes, with the aim of establishing a methodology and a common language that facilitates the adaptation of created computable phenotypes to different healthcare organizations. This process faces challenges due to the complexity and fragmentation inherent to health information systems. However, the benefits are significant, as it will streamline the identification of patient cohorts in various locations, thus enhancing the quality and quantity of local research.
References:
Rethinking Clinical Trials: A Living Textbook of Pragmatic Clinical Trials. https://rethinkingclinicaltrials.org
A users's guide to computable phenotypes by C. Blake Cameron, M.D., M.B.I. https://dcricollab.dcri.duke.edu/sites/NIHKR/KR/Blake_Users_Guide_to_Computable_Phenotypes.pdf
Assessed and Endorsed by the MedReport Medical Review Board
Comments