Effect of Automated Methods on Forensic DNA Workflow Management – ISHI News

Jan 29 2018

Effect of Automated Methods on Forensic DNA Workflow Management

TipsForensic

In 2013, the Texas Department of Public Safety (DPS) Houston Crime Laboratory had two robots for automated purification and one robot for automated assay set-up. All were under-utilized and none were incorporated into a cohesive automated workflow; rather, made available for independent use at the analysts’ discretion. We were working under a case-ownership workflow where analysts were responsible for all parts of a case- from screening to DNA report writing.

While highly satisfying for analysts, the case-ownership approach was not allowing us to keep up with the demands of our customers. We soon began reviewing our processes, identifying pinch-points, and exploring options that would achieve our goals to improve traceability, decrease quality incidents, increase productivity, standardize processes, and ultimately reduce the backlog and turnaround time.

By 2016 we were batching samples using a Hamilton AutoLys StarPLUS capable of fully automated lysis, substrate removal, purification, quantification assay set-up, normalization, and amplification set-up of up to 96 samples at a time. To support the Hamilton instrument, we designed a batching workflow where case processing responsibilities were divided into discrete functional groups (screeners, robotics team, report writers and reviewers). In 2017 we reviewed the goals we established at the beginning of the project to evaluate how the new automated workflow impacted our processes.

 

written by: jennifer duncan, texas department of public safety

 

 

Increase in sample traceability

Prior to 2014, the extent of the DNA extract chain-of-custody was a record of its final storage location written on a worksheet in the case file. Now, this chain-of-custody is much more detailed and available on demand within our Laboratory Information Management System (LIMS), JusticeTrax. Samples selected for extraction are placed in AutoLys® A tubes and registered to the DNA extract in our LIMS using the barcode on the bottom of the tube. Once the sample is purified, it is eluted into a microcentrifuge tube that also has a barcode registered to the extract in LIMS. The extract’s chain-of-custody is a combination of the inherited chain-of-custody from the parent item of evidence and record of the extract’s independent movements from pending extraction in the AutoLys tube thru final permanent freezer storage in the microcentrifuge tube.

We take a similar approach in tracking reagent blanks. Analysts, technical leaders, and supervisors can quickly check the status of the case by determining the location of the DNA extracts. Are there samples pending extraction? Are the samples currently on the liquid handling workstation? Is processing complete and extracts stored in the freezer?

In addition to the improved basic chain-of-custody, we increased traceability by integrating our LIMS into our automated methods and using output files generated by the liquid handling system to create the standard forms we use as case records. The Hamilton instrument scans the barcodes on the tubes then sends them to LIMS. LIMS returns sample information including case number, item number, item description, and origin (questioned, known, or reagent blank) to the instrument which then generates a worklist based on that sample information. Then the instrument uses the worklist to validate the sample information and run the method (lysis/purification, quantification assay set-up, or amplification assay set-up).

When the instrument completes the method, it creates an output file containing sample information, sample positions, sample barcodes, and pipetting volumes among other run-specific details. The output file is imported into a validated intermediary Microsoft Excel workbook that runs a macro and transforms the raw data into the familiar forms we use as our case records. Since the liquid handling system’s worklist is generated from LIMS information, and the instrument uses said worklist to carry out the method and create the output file, and the output file is then used by a validated macro to generate the case record, we can ensure that our records are traceable and accurate.

 

Decrease in quality incidents

In 2013, under the case-ownership workflow, all paperwork and tube labels were manually transcribed, and the majority of DNA processes were manually performed. The level of human intervention required by these processes left the system vulnerable to error. There were 2 sample switches, 2 transcription errors, 17 contamination events, 1 accidental sample discard, 8 plate misloads, and 6 other quality issues such as loading plates into the genetic analyzer backwards, using improper pipette tips, using expired reagents, etc. Thirty-five of those 36 quality events were caused by human error. Only 1 incident was attributed to an automated process where a purification robot failed to elute a sample.

By re-vamping our workflow with Hamilton robotics, and eliminating human intervention where possible, we reduced quality incidents by 58% in 2016 compared to 2013. In fact, the average number of quality incidents per analyst fell from 2.25 to 1.36. The majority (87%) of quality events that occurred in 2016 were attributed to non-automated parts of our workflow that still require human intervention such as differential extractions, loading plates on genetic analyzers, etc. Two quality issues were related to the introduction of robotics – one being a software update pushed out by our IT department that inadvertently caused our customized settings to return to default values, and the other being a hardware failure in the middle of an extraction. Neither incident, however, resulted in sample loss or re-work (re-extraction, re-quantification, re-amplification, or re-typing). Between 2013 and 2016 there was a 63% decrease in the amount of re-work required by quality events, which represents a substantial saving of time, money, and manpower.

The types of errors that were completely eliminated through the implementation of a LIMS-connected automated batching workflow included transcription errors, sample switches, plate misloading, accidental sample discard, and non-differential extraction sample contamination.

 

Increase in productivity

In 2013, the Houston lab had 16 fully-trained DNA analysts working in a case-ownership system who processed approximately 7,392 DNA extracts amounting to 462 samples per analyst. By 2016, DNA analyst ranks had slimmed to only 11. When working within a batching system, however, these 11 individuals were able to process 5,863 extracts for an average of 533 samples per analyst. While the total number of samples processed in 2016 was less than the total in 2013, the average number of samples per analyst increased by 15% indicating that the batching workflow is measurably more efficient than the case-ownership workflow. The lab is currently on pace to process over 8,500 samples in 2017 with 12 DNA analysts.

  

Increase in standardization

Optional wash steps, incubation duration ranges, personal preferences, and other processes open to “analyst discretion” are virtually eliminated in an automated batching workflow. Every sample is processed in the same way, all paperwork is generated the same way, and all documents are stored in the same way from batch to batch. This standardization can extend to other processes and improve efficiency throughout the workflow. While report writing and review are not automated, they benefit from the predictable nature of the protocols and documentation of the automated batching workflow such that parts of the review process can be standardized as well.  Controlling documents related to all cases in a batch by storing them in read-only electronic files means they only need to be thoroughly reviewed once per batch, not once per case per batch which saves significant time during an individual case’s technical review. Additionally, using output files generated by the Hamilton AutoLys STARplus with a validated macro to create the documents means the reviewer does not have to spend time looking for transcription errors. When analysts know what information is already validated versus what needs to be reviewed, the result is a faster, more efficient reporting and review process.

 

Reduction of backlog and turnaround time

Success in reducing backlog and turnaround time depends upon how you measure it. DPS measures backlog reduction by the number of DNA reports being released. There should be more reports released than incoming requests in order to achieve successful backlog reduction. Predictably, the improvements we made upstream in the workflow created a bottleneck downstream at report writing. We have the data, but not enough analysts to write and review the number of cases required to be “successful” in backlog and turnaround time reduction if success is defined in terms of the number of reports released. The loss of 31% of DNA report writers, multiple validations, new amplification chemistries, and major changes to interpretation protocols reduced the total number of reports released and masked any gains achieved from automation in 2016. One benefit that was realized, however, was improved turnaround time for rush DNA requests. Generally the samples in these cases had already been processed, the data was already in the folder, and the case just needed someone to write the report.

The bottleneck at report writing illustrates the importance of taking a holistic approach to workflow design and resource allocation. When improving an upstream process, one must anticipate the impact on downstream processes and adjust operations accordingly. With high-throughput DNA processing comes an avalanche of data to be analyzed and reviewed which requires additional resources beyond that which is already necessary to maintain an acceptable level of case output. Simply having robotics in the laboratory is not sufficient, one must also construct a supportive workflow around the instrumentation in order to maximize their capacity and leverage their advantages.

 

So how did we do?

The effect of automated methods on our forensic DNA workflow management was net positive. With Hamilton as a partner, we made measurable improvements in traceability, quality, productivity, and process standardization. We are very pleased with how the combination of automation and supportive workflow design has streamlined sample processing in our lab. Four out of five of our original goals were achieved, but the fifth and most visible goal – backlog and turnaround time reduction – remains unaccomplished at this time.

 

Future projects

At the Texas Department of Public Safety Houston Crime Laboratory, we are constantly re-evaluating our processes and looking for ways to improve. We recently implemented a DNA-side Y-screening process for sexual assault cases with great success, which I hope to present at ISHI 2018. We are also looking to develop a fully automated, differential extraction at some point in the near future. We currently have several analysts in training that will soon write reports and help us meet our goal for backlog and turnaround time reduction!

 

Collaborators

Dr. Kevin Miller, from Hamilton Robotics, and I collaborated on both the ISHI poster and this blog post.

 

WOULD YOU LIKE TO SEE MORE ARTICLES LIKE THIS? SUBSCRIBE TO THE ISHI BLOG BELOW!

 

SUBSCRIBE NOW!