Development of a repository of computable phenotype definitions using the clinical quality language

Abstract

Objective The objective of this study is to create a repository of computable, technology-agnostic phenotype definitions for the purposes of analysis and automatic cohort identification. Materials and Methods We selected phenotype definitions from PheKB and excluded definitions that did not use structured data or were not used in published research. We translated these definitions into the Clinical Quality Language (CQL) and Fast Healthcare Interoperability Resources (FHIR) and validated them using code review and automated tests. Results A total of 33 phenotype definitions met our inclusion criteria. We developed 40 CQL libraries, 231 value sets, and 347 test cases. To support these test cases, a total of 1624 FHIR resources were created as test data. Discussion and Conclusion Although a number of challenges were encountered while translating the phenotypes into structured form, such as requiring specialized knowledge, or imprecise, ambiguous, and conflicting language, we have created a repository and a development environment that can be used for future research on computable phenotypes.

Fast Healthcare Interoperability Resources (FHIR) for Interoperability in Health Research: Systematic Review

“…Of the 49 studies, most (73%, n=36) studies covered the research domain of clinical research, of which 10 (20%) studies were clinical trials [ 22 , 29 - 31 , 36 , 39 , 43 , 56 , 65 , 66 ]; 3 (6%) studies focused on solutions in public health and epidemiology [ 38 , 40 , 64 ], and the remaining studies did not specify their research domain (20%, n=10; Figure 4 ) [ 24 , 32 , 41 , 42 , 45 - 47 , 50 , 63 , 69 ]. The included studies used FHIR for the standardization of data (41%, n=20) [ 23 , 26 , 30 , 34 , 41 , 45 - 48 , 51 - 53 , 57 - 60 , 63 , 66 , 67 , 70 ], data capture (29%, n=14) [ 1 , 12 , 22 , 24 , 27 , 35 - 37 , 43 , 44 , 55 , 61 , 64 , 65 ], recruitment (14%, n=7) [ 28 , 29 , 31 , 32 , 49 , 56 , 62 ], analysis (12%, n=6) [ 25 , 38 , 42 , 50 , 68 , ...…”

“…Of the 49 studies, the majority were conducted in Germany (47%, n=23) [12,26,[28][29][30][31]34,35,[40][41][42][45][46][47]52,53,[56][57][58]60,62,63,69], the United States (27%, n=13) [22,25,36,44,[48][49][50]61,[64][65][66]68,70], and Australia (6%, n=3) [1,43,67]. The remaining studies were performed in Austria (2%, n=1) [32], Canada (2%, n=1) [24], France (2%, n=1) [51], Greece (2%, n=1) [59], Japan (2%, n=1) [27], Pakistan (2%, n=1) [38], Spain (2%, n=1) [55], Switzerland (2%, n=1) [39], Taiwan (2%, n=1) [23], and the United Kingdom (2%, n=1)…”

“…In terms of medical specialty, most (55%, 27/49) of the studies [24,[27][28][29][30][31][32]34,[36][37][38][39][40][41][42]44,46,48,49,53,56,[60][61][62][63]67,70] were using a generic approach-implementable in any kind of specialty (Table 2). Of the remaining studies, 16% (8/49) use cases focused on infectious disease [1,22,23,26,45,47,50,59], whereas 12% (6/49) focused on oncology [25,55,57,58,65,66] and 8% (4/49) on genomics [43,52,68,69].…”

“…Among the 49 studies, 37% (n=18) did not report on or use additional standards or terminologies [12,22-24,27,28,30-32, 38,39,48,50,55,57,64,66,69]. SNOMED CT [1,25,26,35,37,43,[45][46][47]51,55,56,65,70] and LOINC [25,26,35,37,42-45, 47,49,51,55,56,58,61,65,68,70] were reported to be used by 29% (n=14) and 37% (n=18) of the studies, respectively; 18% (n=9) of the studies used International Classification of Diseases 10th Revision [25,26,35,37,49,51,58,65,68] and 12% (n=6) used OMOP CDM [26,29,35,47,52,60]; and 43% (n=21) of the studies used additional standards which were categorized under "Other" (Table 1) [26,34-37,40,42,43,45-47,49,51,56,58-60, 62,63,67,70]. The implemented FHIR resources by each study are listed in Table 2; 5 (10%) studies did not precisely list their FHIR resources used [36,38,56,64,69].…”

Characterizing Variability of EHR-Driven Phenotype Definitions

“…We used a set of 33 phenotype definitions that were represented using FHIR and CQL. Full details about the creation of the phenotype definitions are explained elsewhere,[Brandt2021] but briefly, the data set is comprised of phenotype algorithms that utilized structured data, were marked as “Final” in PheKB, and were used in a published research study. These phenotype algorithms were then translated into FHIR (selected for its growth and use in healthcare and research contexts) and CQL (selected for its use within eCQMs and ability to represent phenotype algorithms), and validated using manual review and automated testing.…”