One of the first lessons learned was that researchers are naive of medical coding standards used by EMRs.  For instance, they ask for “Lyme Disease” which could be an ICD10 diagnosis code or a positive lab result with multiple test options.  Since we frequently found ourselves translating as part of the sales process, we built a cohort search tool to make it easier to see the different coding variants associated with lay nomenclature, and see prevalence in the Real-World-Data so that users were guided to pick the most relevant EMR codes to find their patients.  The cohort tool also gave real-time feasibility feedback showing the total matching population and a run-rate based on data.  We had to dial in our database optimization for very fast response times here.  This gave incredibly useful real-time feasibility feedback on cohorts for specimen requests and level-set expectations for time to collect specimens, and became a very powerful sales and project setup tool, removing a lot of the manual back-and-forth with a researcher.

The lab workflow and logistics were complicated. Lab staff had to pull a specimen just before it was discarded, pipette into a new container without PHI, then print a shipping label and manifest before packaging and taking the box to a FedEx pickup area.  All labs do “send out”; sending specimens to reference labs for tests that aren’t run in-house.  We piggy-backed on this process to minimize impact on the lab staff.  In addition, we designed our lab UI/UX to flexibly accommodate any workflow variant for shipment preparation.  This included specimen handling, specimen preparation, batching, and special shipping instructions.  To further reduce burden on lab staff, we integrated with FedEx to dynamically produce the shipping label and the manifest (required by IATA regs).  We conducted time-and-motion studies with Applied Management Systems (lab consultants) to demonstrate a 10 minute average time to prepare a shipment for Novaseek.  All of our supplier labs found the lab tools and workflow for Novaseek to be easy to understand and use.

This program used a wide array of structured and unstructured data.  We created generic EMR specific reports (NPR reports for Meditech, CCL reports for Cerner) that included 95% of the work needed at each site.  We also supported HL7, but most sites opted to use the reports with a few last-mile modifications to tailor to the site.  This allowed us to very quickly deploy EMR integrations at a low cost with our fastest being 6 weeks from start to finish.  As FHIR adoption increased we started using FHIR as the canonical data format in our own back-end but did not find wide enough utilization at hospitals to ever make the complete switch.

This product had to comport with multiple regulatory regimes including HIPAA, NIH Common Rule, and IATA regulations, with a nod to FDA part 23 data chain of custody for regulated diagnostic validation studies.  We operated under an umbrella IRB protocol and all technology and operations had to align to the protocol.  This included administering ICFs where applicable and working with data under waiver of consent.  Additionally, we had to become conversant in research regulations because most hospitals required some education as part of the sales process.  We never had any breaches or reportable adverse events, and we never lost a sale to a hospital due to security/privacy concerns.

We initially attempted to obtain a consent from every patient whose remnant specimens we collected.   While this added some additional value to the specimens, we discovered that not all projects required consent and this greatly limited our access to specimen supply.  I led a revision of our IRB and technical program that allowed us to access essentially every patient at the hospital under waiver of consent, increasing the number of specimens we could access by ~20x-30x, and allowing us to supply larger orders faster.

Selecting good hospital partners was essential to our success.  Unfortunately several of our hospitals were financially unstable and either sold or filed for bankruptcy protection within the first year of our supplier contract with them effectively ending our relationship.  This led to a lot of wasted effort implementing and training the sites and ultimately significantly hindered our ability to serve specimens to researchers reliably.  We created a supplier qualification form that covered technical and lab operational issues, but failed to adequately qualify organizational stability.  This is an area I would approach differently based on this experience.