Table 3. Lessons learned after... | American Society of Hematology

Table 3.

Lessons learned after developing an electronic SCD cohort from a data warehouse

Area of consideration	Lesson learned
Personnel of a multidisciplinary team required	Context expertise: a hematologist knowledgeable about the clinical care of SCD who can provide content expertise about SCD and medical complications. Informatics expertise: an informatics expert knowledgeable about clinical information systems (electronic health records, laboratory systems, and clinical data warehouses), data standards, terminologies, and research informatics. Programmer expertise: a computer programmer knowledgeable about extracting the data from the data warehouse and electronic health record. Research coordinators: research personnel knowledgeable about SCD-related medical complications with the ability to identify distinct complications in the electronic health records. Preferably at least 2 individuals extract data from the electronic health records, so at least 10% of the data extracted can be double-checked.
Effort for the team needed	Initial creation and validation of cohort: 200 h of context expertise 2500 h of informatics expertise 4000 h of programmer time 5350 h of validation time (this can vary depending on cohort size and selected SCD phenotypes) An ongoing effort is needed to assess and update the cohort to double check data structures and integrity. This effort could be substantially more if moving to a new electronic health record or data warehouse system. The minimal amount of time required annually: 40 h of programmer time 40 h of validation time
Practical lessons learned	A recurring feedback loop is needed between members of the team and ongoing assessment of data quality and validation work. The feedback loop included the following: Weekly meetings between all members of the team to resolve and troubleshoot issues. Weekly programmer-specific meetings to develop algorithms and complete data extraction. Anticipate data fragmentation requiring vigilance to ensure complete data capture.
Exportability	Every electronic health record data warehouse has its unique attributes, including data structures, terminologies, access, and security that the team needs to familiarize themselves with and use. Changes in electronic health record data entry and clinical information systems may require modifications to code and extract data or additional validation to ensure data is correct and accurate. Harmonizing data systems across electronic health records is essential. There are certain areas in which harmonization can occur (eg, laboratory values) and those in which harmonization can be difficult (eg, reports and clinical notes).

Area of consideration	Lesson learned
Personnel of a multidisciplinary team required	Context expertise: a hematologist knowledgeable about the clinical care of SCD who can provide content expertise about SCD and medical complications. Informatics expertise: an informatics expert knowledgeable about clinical information systems (electronic health records, laboratory systems, and clinical data warehouses), data standards, terminologies, and research informatics. Programmer expertise: a computer programmer knowledgeable about extracting the data from the data warehouse and electronic health record. Research coordinators: research personnel knowledgeable about SCD-related medical complications with the ability to identify distinct complications in the electronic health records. Preferably at least 2 individuals extract data from the electronic health records, so at least 10% of the data extracted can be double-checked.
Effort for the team needed	Initial creation and validation of cohort: 200 h of context expertise 2500 h of informatics expertise 4000 h of programmer time 5350 h of validation time (this can vary depending on cohort size and selected SCD phenotypes) An ongoing effort is needed to assess and update the cohort to double check data structures and integrity. This effort could be substantially more if moving to a new electronic health record or data warehouse system. The minimal amount of time required annually: 40 h of programmer time 40 h of validation time
Practical lessons learned	A recurring feedback loop is needed between members of the team and ongoing assessment of data quality and validation work. The feedback loop included the following: Weekly meetings between all members of the team to resolve and troubleshoot issues. Weekly programmer-specific meetings to develop algorithms and complete data extraction. Anticipate data fragmentation requiring vigilance to ensure complete data capture.
Exportability	Every electronic health record data warehouse has its unique attributes, including data structures, terminologies, access, and security that the team needs to familiarize themselves with and use. Changes in electronic health record data entry and clinical information systems may require modifications to code and extract data or additional validation to ensure data is correct and accurate. Harmonizing data systems across electronic health records is essential. There are certain areas in which harmonization can occur (eg, laboratory values) and those in which harmonization can be difficult (eg, reports and clinical notes).

We have identified the most salient strategies for creating a contemporaneous cohort of children and adults with SCD from a data warehouse.