What's unique about Comprehend compared to all other BI tools? Tableau was mentioned in the article, but there are plenty of other applications like QlikView or MS SSAS, etc.
Comprehend isn't a BI tool. If all of a company's data is well-structured in a single database, with a non-changing data structure and without missing or null data, then it's easy to use BI tools like Tableau, Spotfire, Excel, etc.
But this is not the case in the modern day enterprise. Instead, there are dozens of different data collection systems, with different and changing data structures. This is what Comprehend's core technology was built to handle, in real-time.
In order to get similar functionality, companies are typically relying on teams of programmers to manually write and run scripts over and over again. Or, they're trying to put in place data warehouses, which rarely contain everything required and are often out of date.
What you described is just standard ETLs in any Data Warehouse system - again in my opinion a typical functionality in a BI stack. QlikView tries to do it in-memory and live but it puts strain on OLTP systems for example freezing cash machines (seen it!). Every system requires maintenance, question is where's Comprehend's maintenance effort placed?
This is an interesting side to to push on, but I haven't heard many people claiming this is where pharma has data woes. The real data woes are from high-throughput biological data such as gene expression and now sequencing, and that's a field that's crowded with tiny startups (and though my startup has great tools for this, surviving on pharma does not seem like a sustainable business model).
This other side of information management, clinical trials, is typically handled by Oracle's offerings in the area. However, I haven't heard many complaints about the data side of clinical trials, it's the actual patient accrual and patient handling that's typically expensive on clinical trials. That's not to say that Oracle has solved all the issues. As there are more trials with a bent towards personalized medicine, accruing the patients with the desired traits is going to get more difficult, and will require the participation and coordination of many many more different trial sites. Having additional trial sites is the real cost, as all the sites have to use exactly the same treatment and data collection protocols, and these protocols almost certainly differ from the hospital's or office's standard protocols. And trials also require the doctors to attend training sessions fro their particular data collection routines, even if they have previously learned similar techniques, which is extremely wasteful, but necessary without proper certification schemes. These are the areas that a forward thinking startup can really get ahead of Oracle, so even though I don't see much indication in this PR about it, hopefully Comprehend is thinking in that direction.
That said, typical Silicon Valley startups are woefully naive about healthcare (not necessarily a bad thing!!), and typical startups here underestimate the difficulty of working their way into long-standing relationships, and of changing practices in a field that is change-averse without proper evidence. Additionally it can be extremely difficult to see the areas that can use improvement unless you've studied the terrain carefully, and by the time you've found pain points that you can address, you may have lost the helpful parts of naivete that let you look past all the myriad hurdles in the way of making progress.
I'm under the impression that these guys are in the pharma space because that is where the money is for a service provider. Pharma underwrites a lot of this stuff. Once you have a toehold in the larger health care space you can branch out and expand to other opportunities. If I were to approach healthcare on the whole I could think of worse places in the pipeline to start than big pharma.
There is a difference between bioinformatics such as genetic sequencing and its better half, analysis, and medical /clinical/research informatics. Imho, the current ability to sequence vast amounts of dna far outstrips our ability to analyze it. Better, quicker analysis will come in the form of better algorithms and through machine learning/data mining.
Medical informatics, the practice of managing patient information is a separate problem. Here the problem is connecting different systems that were not meant to be connected when designed, and great tools for parsing and presenting that data by non-programmers. I'm working with tools like Amalga and RedCap right now which each play in their own space but do provide a service. They could both be... better. The field is wide open for disruption.
Programmer time is the limiting resource factor in making better things happen for a lot of different institutions. The other major limiting factor is simply institutional politics. No amount of silicon valley magic will make that go away.
This article is surprisingly information-free with respect to what they are actually doing. It mostly says that clinical trials are lengthy and costly...
6 comments
[ 3.8 ms ] story [ 28.5 ms ] threadComprehend isn't a BI tool. If all of a company's data is well-structured in a single database, with a non-changing data structure and without missing or null data, then it's easy to use BI tools like Tableau, Spotfire, Excel, etc.
But this is not the case in the modern day enterprise. Instead, there are dozens of different data collection systems, with different and changing data structures. This is what Comprehend's core technology was built to handle, in real-time.
In order to get similar functionality, companies are typically relying on teams of programmers to manually write and run scripts over and over again. Or, they're trying to put in place data warehouses, which rarely contain everything required and are often out of date.
This other side of information management, clinical trials, is typically handled by Oracle's offerings in the area. However, I haven't heard many complaints about the data side of clinical trials, it's the actual patient accrual and patient handling that's typically expensive on clinical trials. That's not to say that Oracle has solved all the issues. As there are more trials with a bent towards personalized medicine, accruing the patients with the desired traits is going to get more difficult, and will require the participation and coordination of many many more different trial sites. Having additional trial sites is the real cost, as all the sites have to use exactly the same treatment and data collection protocols, and these protocols almost certainly differ from the hospital's or office's standard protocols. And trials also require the doctors to attend training sessions fro their particular data collection routines, even if they have previously learned similar techniques, which is extremely wasteful, but necessary without proper certification schemes. These are the areas that a forward thinking startup can really get ahead of Oracle, so even though I don't see much indication in this PR about it, hopefully Comprehend is thinking in that direction.
That said, typical Silicon Valley startups are woefully naive about healthcare (not necessarily a bad thing!!), and typical startups here underestimate the difficulty of working their way into long-standing relationships, and of changing practices in a field that is change-averse without proper evidence. Additionally it can be extremely difficult to see the areas that can use improvement unless you've studied the terrain carefully, and by the time you've found pain points that you can address, you may have lost the helpful parts of naivete that let you look past all the myriad hurdles in the way of making progress.
There is a difference between bioinformatics such as genetic sequencing and its better half, analysis, and medical /clinical/research informatics. Imho, the current ability to sequence vast amounts of dna far outstrips our ability to analyze it. Better, quicker analysis will come in the form of better algorithms and through machine learning/data mining.
Medical informatics, the practice of managing patient information is a separate problem. Here the problem is connecting different systems that were not meant to be connected when designed, and great tools for parsing and presenting that data by non-programmers. I'm working with tools like Amalga and RedCap right now which each play in their own space but do provide a service. They could both be... better. The field is wide open for disruption.
Programmer time is the limiting resource factor in making better things happen for a lot of different institutions. The other major limiting factor is simply institutional politics. No amount of silicon valley magic will make that go away.