Our client (a large pharmaceutical company) works with partners who analyze tissue samples in clinical studies. Typically, more than one study would be active at a time, with multiple test panels. Raw analysis results are generated in machine-readable files with embedded metadata. Files are uploaded to the Client's systems in a nested file hierarchy that already contains thousands of files. Due to the fast pace of work, up to several dozen result files may be uploaded in any 24-hr period.
In the past, Client scientists used to manually track newly arrived files and to annotate them with tissue sample IDs and metadata before forwarding work requests to partners. With a somewhat worrying regularity, incorrect sample IDs were assigned to new files or errors in the embedded metadata were overlooked. Incorrectly tagged results were then sent to downstream reviewers who were unaware of the mis-annotations. Quality of the review work was at risk.
Automated tracking and validation of lab results improves data quality in clinical studies.
Our Client realized that a degree of automation had to be introduced to track new submissions. In addition to tracking new files, they decided that metadata embedded in these files also had to be validated. This in retrospect turned out to be a very smart decision.
Saber Informatics was asked to set up an automated book-of-work file tracker which would also parse out embedded metadata from newly submitted files. The metadata would then be auto-validated against internal company records. Typos would be flagged and either corrected automatically or sent in log files to the review team. In each run of the tracker a log file would be generated and sent to data reviewers who could then make corrections downstream or go back to the labs and request corrections. Looking back, establishing this feedback loop proved to be extremely valuable.
Partner scientists can now focus on their primary review work. Over 99% of submissions are now processed automatically without the need for human intervention. Every few days there are metadata corrections that need to be made based on auto-generated alerts. Such occasional corrective steps no longer interfere with the main workstream.
Using the receive-validate-alert automated approach, our Client has been able to control the quality of raw results that are submitted by clinical laboratories. This immediately led to an improvement in the quality of data reviews conducted by partners.
