Commercial Solutions for OMOP/OHDSI Research Acceleration

General
, , , ,
1

Dear Community,

I would like to understand why commercial solutions (ie. software that is developed by for-profit companies, eg IQVIA, et. al.) have utterly failed to gain any traction in this community of researchers.

I have spent 3.5 years developing in this space, and have contributed to Broadsea as part and parcel of my engagement with the International Rett Syndrome Foundation.

We were able to convert their 15 year NIH funded Natural History data into the CDM 5.4, which enabled them to co-operate with 3 Pharma developers to bring to market novel therapies through the FDA/NDA process.

We launched the Parthenon platform back in March, and received support and interest in how we have engineered a common, singular, comprehensive “protocol-to-evidence-to-publication” pipeline. Since then we have proven that this AI-native platform can serve researchers from Korea, to Chile, to 20+ institutions in the US, to the OHDSI community in India.

I have reached out to @Chris_Knoll, @Paul_Nagy, @Christian_Reich, @Ajit_Londhe @clairblacketer and others who have contributed massively to this community, and I have received little or no response.

Here is the deal: www.acumenus.io

Please let me know whether our Parthenon platform is worth investing in, as beta testers, as code evaluators, as institutions that you work for, across the gamut of AMCs, Pharma, etc. I genuinely need your informed opinions.

To others who are long time users/contributors to the fragmented set of solutions from ATLAS to HADES and all other sub-solutions that you need to tie together… Please let me know of your level of satisfaction of where we are as an FOSS dependent community. My intention is to ALWAYS provide a FOSS solution, but also for those who can AFFORD it, a vibrant and evolving platform paid for by those who can support accelerated advancement to the rest of the world of outcomes researchers.

As an example of what our fine-tuned LLM based on the free and open source MedGemma:27b model which we call Abby, also a harness for the larger frontier models, is capable of… I am attaching our latest test study on Hypertension and our earlier work on Rett Syndrome.

I genuinely hope for a sincere response and guidance.

Yours,
-Sanjay M. Udoshi MD.

Under-Diagnosis_of_Hypertension_JAMA-style.pdf (718.7 KB)
IRSF_Vocabulary_Mapping_Appendix.pdf (258.3 KB)
IRSF_Genotype_Severity_Paper.pdf (2.4 MB)

1 Like
2

Hi Sanjay, thanks for your post. From my perspective, the engagement problem you’re experiencing is not because you have a commercial solution. It may be in part because for many teams (like mine), a monolithic, all-in-one platform is not the right solution. As a global analytics team at a large pharma company we have a complex set of workflow and IT requirements, so modularity and flexibility are important to us. I think this also applies to the OHDSI community as a whole. OHDSI caters to such a wide user base that keeping things modular and low-level is important in driving adoption.

Relatedly, as an open-source maintainer in OHDSI, I prefer to spend the little spare time I have contributing to core OHDSI software. While our open-source community may progress much slower than vendor-led projects, I believe strongly that the value of OHDSI lies in its community-led, distributed nature.

That said, there is absolutely space for vendor solutions in OHDSI. Some institutions likely are seeking an all in one platform solution, and many definitely want/need to offload their OMOP/OHDSI implementation to a vendor. It’s commendable that you’ve made your solution open-source, but if you’re seeking contributors I think you’ll need to hire them, or source them from within your user base.

If you want contributions from OHDSI maintainers and contributors you’ll need to get buy in that your solution is a priority for the community, and that it’s worth it for those individuals to volunteer their time on. Doing so can be incredibly challenging and isn’t guaranteed to pan out. So it’s definitely understandable (and expected, and a good thing!) that vendors develop independent solutions. You’ll just need to resource those differently than a project developed by & for the community.

These are just my opinions, and I hope I understood your questions correctly. OHDSI does currently lack a structured pathway for new open-source solution investment/prioritization. One aim of the Technical Advisory Board has been to create such a pathway, but our progress has been limited. I will admit that my contributions in this area have slipped recently, but this thread has gotten me thinking about it again - so thank you for that :slight_smile:

3

Thank you @katy-sadowski. Commercial vendors continue to have success in OHDSI, but most are boutique services like OMOP transformation, phenotyping or study design and execution. There are proprietary platform/product companies in OHDSI. The other problem is, with vibe coding, there is a lot of platforms that looks very similar and there is a lot of me too replicas. There is a lot of AI experimentation and that may be a factor too.