Domain

Clinical Informatics

Type

Protocol

Theme

operations

Start Date

7-6-2014 1:15 PM

End Date

7-6-2014 2:45 PM

Structured Abstract

1) Objective

The objective of this study was to develop an automated system for rapidly identifying potential candidates for enrollment in the Oncology Nurse Navigator Trial.

2) Background/Context

A major barrier to conducting pragmatic clinical trials is the efficient enrollment of eligible participants. Automated surveillance of electronic medical record (EMR) data to identify study participants has not been used widely.

Methods

We used Group Health Cooperative EMR, administrative claims data, and pathology reports to identify candidates for enrollment in the Oncology Nurse Navigator trial on a daily basis. Clinical progress notes in the EMR were extracted from the operational system daily and evaluated for caseness using natural language processing. A list of candidates meeting inclusion criteria was emailed automatically to study staff each morning. Potential candidates were validated using automated filters and expert review.

3) Findings

The algorithms used for rapid case ascertainment are relatively simple to program and use. Moreover, the algorithms can be used adaptively and will “learn” on the basis of feedback from expert review. Importantly, the algorithms developed were robust – allowing us to reserve human attention for the tough cases.

4) Lessons Learned

We successfully implemented an automated rapid case ascertainment algorithm in Group Health Cooperative. A major strength of the approach is that it is relatively simple to program and use and requires minimal human input. At the same time, the approach can be used adaptively and will learn from expert input. There are several limitations. First, the algorithm must be trained on pre-categorized documents. Generally, the more training, the better the classification of caseness. Second, classifications apply to the document as a whole. For example, an investigator can feed the algorithm text from the “Diagnosis” section only, but there is no way to focus on text applying to say, the left lung only. Third, our approach was not very amenable to post-hoc evaluation. That is, there was No easy way of quickly determining the particular characteristics about a document that led the algorithm to assign the category (caseness) it did.

5) Conclusion

Large-scale effectiveness trials and quality improvement projects must become more efficient in identifying candidates for intervention or case management. Approaches that routinely access multiple data streams and apply adaptive “smart” algorithms to identify candidates with minimal human input will be an important tool for advancing effectiveness research and program evaluation. Our experience in the Oncology Nurse Navigator Trial demonstrates that automated approaches to rapid case ascertainment are both feasible and efficient.

Acknowledgements

n/a

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License.

Share

COinS
 
Jun 7th, 1:15 PM Jun 7th, 2:45 PM

Rapid Automated Case Ascertainment for Clinical Trial Enrollment

1) Objective

The objective of this study was to develop an automated system for rapidly identifying potential candidates for enrollment in the Oncology Nurse Navigator Trial.

2) Background/Context

A major barrier to conducting pragmatic clinical trials is the efficient enrollment of eligible participants. Automated surveillance of electronic medical record (EMR) data to identify study participants has not been used widely.

Methods

We used Group Health Cooperative EMR, administrative claims data, and pathology reports to identify candidates for enrollment in the Oncology Nurse Navigator trial on a daily basis. Clinical progress notes in the EMR were extracted from the operational system daily and evaluated for caseness using natural language processing. A list of candidates meeting inclusion criteria was emailed automatically to study staff each morning. Potential candidates were validated using automated filters and expert review.

3) Findings

The algorithms used for rapid case ascertainment are relatively simple to program and use. Moreover, the algorithms can be used adaptively and will “learn” on the basis of feedback from expert review. Importantly, the algorithms developed were robust – allowing us to reserve human attention for the tough cases.

4) Lessons Learned

We successfully implemented an automated rapid case ascertainment algorithm in Group Health Cooperative. A major strength of the approach is that it is relatively simple to program and use and requires minimal human input. At the same time, the approach can be used adaptively and will learn from expert input. There are several limitations. First, the algorithm must be trained on pre-categorized documents. Generally, the more training, the better the classification of caseness. Second, classifications apply to the document as a whole. For example, an investigator can feed the algorithm text from the “Diagnosis” section only, but there is no way to focus on text applying to say, the left lung only. Third, our approach was not very amenable to post-hoc evaluation. That is, there was No easy way of quickly determining the particular characteristics about a document that led the algorithm to assign the category (caseness) it did.

5) Conclusion

Large-scale effectiveness trials and quality improvement projects must become more efficient in identifying candidates for intervention or case management. Approaches that routinely access multiple data streams and apply adaptive “smart” algorithms to identify candidates with minimal human input will be an important tool for advancing effectiveness research and program evaluation. Our experience in the Oncology Nurse Navigator Trial demonstrates that automated approaches to rapid case ascertainment are both feasible and efficient.