Sole Source Justification to post.pdf

Synthetic Data Generation for COVID VA-20-00090770

JUSTIFICATION FOR AN EXCEPTION TO FAIR OPPORTUNITY

1. Contracting Activity: Department of Veterans Affairs (VA)

Office of Procurement, Acquisition, and Logistics Technology Acquisition Center 23 Christopher Way

Eatontown, New Jersey 07724

2. Description of Action: The proposed action is for a firm-fixed price delivery order to be awarded under the National Aeronautics and Space Administration (NASA) Solutions for an Enterprise-Wide Procurement (SEWP) V to Thundercat Technology (Thundercat) for brand name MD Clone Platform which consists of MD Clone Platform XL POC, the integrated MD Clone User Portal and MD Clone Associated Business Services software licenses. In addition, the proposed action includes platform set up and training services.

3. Description of the Supplies or Services: VA is interested in unlocking data access to engage the widest possible audience in innovation, research, and quality improvement efforts across a wide range of opportunity areas to include COVID response. VA requires a method of data sanitization which is mathematically proven to preserve privacy while simultaneously maintaining data sets and data lakes which are complex and diverse to sustain meaningful and relevant data relationships. This can be accomplished through synthetic data analysis. Synthetic data is statistically identical to data from real patients, but is not permitted to be associated with individual patients which protects the patient from being associated with their data. Synthetic data generation, which drastically speeds up the rate of data analysis, is not readily available in VA which creates a choke point for rapid data analysis on fast moving clinical and social issues, to include the COVID pandemic.

The brand name MD Clone platform enables synthetic data generation and consists of MD Clone Platform XL POC, the integrated MD Clone User Portal and MD Clone

Associated Business Services software licenses. VA requires this platform to be used with VA’s existing data warehouses to create synthetic data which maintains information privacy while simultaneously having computational clinical validity and maintaining meaningful clinical relationships throughout the created data sets. VA will utilize this platform to provide synthetic data as a means to innovate, research, and improve quality and speed of efforts across a wide range of opportunity areas without compromising personally identifiable information. The MD Clone Platform will provide the following: a self-service user interface that can be understood and utilized by a non-clinical, non-researcher (non-data scientist) VA staff while maintaining full VA security compliance, masking the complexity of the data available; a data lake that can incorporate and meaningfully organize diverse large data sets; synthetic data generation with metrics on validity and privacy of data outputted; clearly defined data and rules/logic; and synthetic data set while including measures of sensitivity and specificity. In addition, the MD Clone Platform will be able to operate on VA based cloud or VA infrastructure and provide data support for five use cases where VA has identified a critical area of need. Furthermore, access to the user interface for quality assurance and clinical improvement purposes will not be limited among VA staff which will enable collaboration with its users to create large, privacy preserving healthcare data sets to unlock insights and provide improvements to care. This will expand VA’s current capabilities by allowing collaboration and sharing of information across VISNs within VA and will allow for potential collaboration with external Agencies, if needed. As part of this proposed action, training will be provided to VHA training coordinators who can assist MD Clone Portal end users in utilizing the MD Clone platform. All delivery, installation, configuration, testing, go-live, and training shall be completed within 95 days from contract award. MD Clone Platform XL POC software subscription shall be for 12-month date of go-live.

4. Statutory Authority: The statutory authority permitting an exception to fair opportunity is Section 41 U.S.C. 4106(c)(2) as implemented by the Federal Acquisition Regulation (FAR) Subpart 16.505(b)(2)(i)(B), entitled “Only one awardee is capable of providing the supplies or services required at the level of quality required because the supplies or services ordered are unique or highly specialized.”

5. Rationale Supporting Use of Authority Cited Above: The proposed source for this action is Thundercat Technology, 1925 Isaac Newton Square E, Suite 180, Reston, VA 20190. Based on extensive market research as described in Section 8 of this document, it was determined that there is no competition among vendors. Thundercat Technology is the sole reseller of the MD Clone Platform.

The MD Clone Platform generates synthetic data sets and provides curated, mathematically detectable, computational validity automatically. The MD Clone Platform either makes corrections autonomously or suggests corrections to the data scientists managing the data set generation. This is a unique capability that only the MD Clone Platform can provide. This unique capability, to provide automated synthetic data generation, makes for faster and easier inclusion of the most recent data ensuring relevancy to potential solutions. Synthetic data, through use of the MD Clone Platform, can reduce cycles of discovery from months and years to hours and days through the user portal and access to synthetic data without the need for institutional review board

(IRB) approval. The MD Clone Platform user portal and software is used to extract data from MD Clone’s synthetic data through queries. These queries aid research on social and clinically exigent problems; both critical and pressing problems. The MD Clone Platform user portal software interface masks the complexity of the data making it easier to be understood by non-clinical, non-researcher and non-data scientists. MD Clone’s synthetic data and the integrated MD Clone user portal software will ultimately provide VA a single process for the rapid generation of multiple synthetic data sets and the democratization of access for a larger diverse pool of research, business and clinical staff.

There is currently no other software platform that is capable of providing the required automated data set generation with fully curated and mathematically proven computational validity and meaningful clinical relationships accessible through a self-service user query portal. Currently, each time data sets need to be curated, VA analysts need to seek approval from the IRB. This requires, on average, a 4-6 week application and review process. Once a board review can be scheduled, assuming no outstanding issues are found in the review, approval is granted from the IRB. Following approval, a 1-2 week process ensues to gain access to the data that needs to be curated. Then, after access to data is granted, each data point is required to be traced back to its original data source and regression testing completed to prove the data set retains its clinical relevance and nearest neighbor analysis to confirm that the data set maintains privacy. This effort is very time consuming and takes multiple months to complete once data access is approved following IRB or appropriate quality improvement project. As a result, the data sets become historical and unable to embrace changes in current data.

The data sets created with available free open source software (i.e. CorGAN, MedGAN and Synthea) do not meet VA’s requirements as they have to be manually curated. All other potentially comparable approaches to synthetic data generation require significant manual effort of a trained data science expert to curate a data set, generate synthetic data and then evaluate the data for privacy and quality. Use of any other sources platform to provide synthetic data will result in a multiple month delay each time a new synthetic data set is needed which will not enable VA with the ability to analyze current trends and provide optimal COVID response.

VA’s only avenue to avoiding these delays each time a data set needs curation is through use of the MD Clone Platform. VA requires the MD Clone Platform, specifically the MD Clone User Portal, to provide VA a single process for the rapid generation of multiple synthetic data sets and the democratization of access for a larger diverse pool of research and clinical staff. The MD Clone Platform is the only platform that can bypass the need for an extended process each time data sets need to be generated by providing VA with automatically curated synthetic data sets ready for query within minutes or hours. No other platform can provide VA this functionality and meet the need to rapidly increase the rate of response and analysis on fast moving clinical and social issues, to include the COVID pandemic. Additionally, no other platform is capable of providing software capable of integrating both a user query portal, rapid synthetic data generation, and automatic reports on the privacy and validity of data sets.

Synthetic data democratizes the data to non-clinical, non-researcher, non-data scientists, while maintaining the privacy and clinical utility of the data sets for the research and innovation community. The MD Clone Platform automated synthetic dataset generation and the user query portal allow for more users to unlock insights about emerging data concerns like the COVID pandemic and how to respond. No other platform can meet this requirement.

Though many research organizations provide data, the provided data is proprietary and secure, principally for data scientists, and access is limited. Other free open source software platforms are fully manual and inefficient for synthetic data set generation which in turn causes delays and does not allow for rapid analysis and response. The ability to provide this response is critical for VA to provide adequate care to Veterans.

The automated data set generation available via MD Clone’s Platform is critical to VA’s COVID response to provide the very best in-patient care based on evidence the data provides. Failure to procure this platform will degrade the ability for VA to provide quick and efficient response to fast moving clinical and social issues, most notably the COVID pandemic.

6. Efforts to Obtain Competition: Market research was conducted, details of which are in section 8 of this justification. This effort did not yield any additional sources that can meet the Government's requirements. There is no competition anticipated for this acquisition. In accordance with FAR 16.505(a)(4)(iii)(A)(2), this justification will be provided with the solicitation to all appropriate NASA SEWP V contract holders.

Furthermore, in accordance with FAR 5.301, 16.505(b)(2)(ii)(D), and 16.505(a)(4)(iii), the award notice for this action will be synopsized on the Contracting Opportunities website and this justification will be made publicly available within 14 days of award.

7. Actions to Increase Competition: The Government will continue to conduct market research to ascertain if there are changes in the marketplace that would enable future actions to be fully competed. Currently, Thundercat has exclusive rights to sell and support the brand name MD Clone Platform which consists of MD Clone Platform XL POC, the integrated MD Clone User Portal and MD Clone Associated Business Services software licenses.

8. Market Research: The Government’s technical experts conducted market research by reviewing other similar platforms. This market research is an ongoing process and continuously being conducted through July 2020. The other similar platforms reviewed were CorGAN, MedGAN and Synthea (GAN- Generated Adversarial Networks) free open source synthetic data generation software. These products do not meet the requirements as outlined in Section 5 as they require manual mathematical regression and curation of data to be performed by a data scientists. Manually curated data sets take multi-months to develop making them historical and unable to incorporate current data which will not enable VA with the ability to analyze current trends and provide timely COVID response.

Based on reviews of these products, the Government’s technical experts determined that no other platforms can meet the Government’s requirements as outlined in this Justification. Additionally, none of these other products can offer the single integration automated data set generation with mathematically proven privacy while simultaneously maintaining computational validity and meaningful clinical relationships. Based on all of these market research efforts, the Government’s technical experts have determined that only brand name MD Clone Platform which consists of MD Clone Platform XL POC, the integrated MD Clone User Portal and MD Clone Associated Business Services software licenses can meet all of VA’s needs.

In addition, the Government representatives received confirmation from MD Clone on August 26, 2020, through a signed letter confirming that Thundercat Technology is the sole reseller of the MD Clone platform. Furthermore, MD Clone confirmed the only way to purchase its products is via NASA SEWP V contract vehicle to Thundercat.

Therefore, there is no competition anticipated for this acquisition.

9. Other Facts: None.