A recent press release about the collaboration between GenoSpace and Caris Life Sciences got me thinking about the future of the management of cancer genomics information (see: Caris Life Sciences collaborates with GenoSpace to advance cancer care through research). Below is an excerpt from it:
Caris Life Sciences...announced a collaboration with GenoSpace, a Massachusetts-based technology company that develops robust software solutions for genomic and health data. Caris will apply GenoSpace's population analytics offerings to uncover and better utilize key insights from the company's... tumor profiling service....The partnership also leverages GenoSpace's analytics architecture expertise to unlock key treatment insights from the Caris Registry, a database of clinicopathologic and outcome variables from consenting patients whose tumors underwent multi-technology profiling by Caris.... Caris' tumor profiling databases are comprised of multiple assay technologies performed in combination to achieve [a]...molecular profile of a tumor. The resulting analysis and insights, enabled in part by this collaboration, will create a HIPAA-compliant informatics platform that provides Caris' institutional partners the ability to investigate institution-specific profiling data....It also allows Caris to more efficiently and expediently isolate novel and/or critical drug-biomarker associations, ultimately providing better treatment information to oncologists and their patients.
We see here a collaboration between GenoSpace with its cancer genomic support and profiling tools and Caris with its cancer profiling technology. This relationship suggests to me that a new computing model is evolving for cancer genomics that is different from the preexisting LIS computing model. Here are short descriptions of the two models for comparison purposes:
- Older classic LIS model: LIS software installed in hospital => LIS receives and stores patient-specific data generated in the hospital labs => LIS reports data to hospital clinicians for analysis, usually these days via the EHR.
- New cancer genomics model: for-profit companies develop cancer genomic profiling databases and analytic software that runs in the cloud => hospital labs upload individual patient tumor genomic data to this diagnostic cloud node => individualized patient interpretive reports are generated in the diagnostic cloud node => PDF reports are viewed by hospital clinicians via EHR web readers or hand-held tablets or smart phones.
Most of the processing in this new model takes place in the cloud. Reports are generated under this new model using sophisticated databases, analytic software, and algorithms that may not be available to hospital labs. This data and software non-availability is related, in part, to its status as the intellectual property of the genomic reference labs. Moreover, the software and data is also in a state of flux as the pool of knowledge about cancer genomic changes. In the old model, patient databases were created at the hospital level by analyzing patient serum samples that were then interpreted by pathologists and the test-ordering clinicians. In the new model, some of the information necessary to interpret the patient test results belongs to the vendor.
Painting with broad brush strokes, here is the IT architecture I see evolving to manage cancer genomic data, a process separate from hospital EHRs and LISs. The job of the EHR is largely to replicate the paper medical record, to document work flow pertaining to patients, and also to aggregate relevant clinical data generated at the hospital level. The IT nodes that provide cancer genomic analysis exist only in the cloud. In most cases, the hospital clinicians will access the cancer genomic reports for their patients via web readers supported by the EHR or via hand-held tablets and smart phones. Similar and parallel sophisticated cloud diagnostic nodes will also evolve for other diagnostic services like cardiology and radiology.
It may be relevant to refer to these specialized, cloud-based, diagnostic, computing nodes as SMAC diagnostic nodes (see: Time to SMAC the Healthcare Consumer). The term SMAC was described by John who blogs over at Chilmark Research in the following way:
SMAC – Social [i.e., Personalized], Mobile, Analytics, Cloud – is a popular framework for optimizing business performance through IT. The basic idea is that these four elements all play key roles in generating value from data through capture, storage, and application. Social [i.e., Personalized] refers to the consumer or end-user level, where data is created and collected. Mobile describes the shift to smartphone and tablet-driven computing. Analytics speaks to the growing ability and need to interpret and understand data big and small. Cloud refers to the advent of virtual computing through untethered storage and access to data, applications, services, and more.