Why Healthcare Innovation is Broken and How to Fix It
As ThinkSono has matured, I have been thinking a lot about what the key issues are that makes healthcare innovation so slow and expensive compared to other industries.
This topic gets brought up to the forefront fairly frequently, and in particular when I am asked to give advice to healthcare startup founders who are earlier in their journey.
The difficulties in bringing in new technology or services within a healthcare system is really brought to light, when we consider that it took a worldwide pandemic to force hospitals to use the most basic technologies such as video conferencing. Startups have been trying to convince hospitals to use video conferencing for a decade, and it’s only now that some hospitals are adapting the use of such ubiquitous technology.
For many founders, investors and the general public, the outdated technology used by healthcare systems worldwide seems rather perplexing. Sometimes it feels like the consumer world of technology is light years ahead of where the healthcare system should be.
It took me a while to fully digest the core reasons for this, and having been in direct contact with hospitals, regulators, administrators and clinicians for 6 years now, I have come to see that the common denominator is a mindset issue.
The result of this mindset generates a range of internal and external processes which in turn slow the entire engine of innovation down.
Just in case this might seem too abstract, when the rate of innovation within healthcare is slow, healthcare providers and patients suffer.
A real-life example was given during the COVID-19 pandemic, when every day that the vaccine was delayed, we had thousands of people dying due to not being vaccinated and entire economies were getting destroyed.
Obviously in that case there was so much pressure to speed up the development, that we essentially bypassed a huge number of structural issues. It truly was a “all hands on deck” moment. And what we achieved was remarkable.
Clinicians and scientists had to keep reminding the public that no shortcuts were taken, but understandably, people were still sceptical. If the average time a vaccine takes to develop is 10–15 years, how could they have possibly done it safely within a year?
Well, this is an extreme case, but the gist is that things were done in parallel instead of sequentially, the funding was given immediately, scientists collaborated and shared data worldwide and the research that was used had been built over 50 years. An article explaining more of the details in this case is found here.
The point is the “real” science part of the development took 10 days from its discovery to the sequencing being completed. The rest is in some sense just the logistics. Obviously I’m exaggerating a little here since you still have to develop the actual vaccine, run the clinical trials and then get the approval etc. But the point I am making is that the infrastructure and the fundamental science had not changed. What had changed was the mindset that we all had.
The crux of this mindset is our perception of risk had changed, and we added time as a huge cost.
We knew that the risk of the vaccine being harmful was very, very low due to 50 years of research in the field. However, we also took into account that anything that takes time, which is not fundamentally related to safety was considered a massive cost.
Simple things like the bureaucracy of getting funding approved or speaking to regulators which usually takes months if not years was basically set to hours and days. No one wanted to be the person that slows down the development of the vaccine and everyone wanted to help.
The actual science took 10 days, the safety check (via clinical trials) to around a year and the rest of the overhead was set as close to zero as possible.
The impact the vaccine had worldwide is so large it’s hard to even quantify. From human lives saved to economies starting to recover and to the future emphasis on preventing infections disease from spreading.
Now, this was clearly a unique situation, and we can’t treat every other treatment or healthcare innovation the same way. However, we should take a look at all the “overhead” that stops innovations from being introduced.
OK, so what can be done to fix these issues — Regulatory improvements
The first thing that is required is for regulators such as the FDA in the US and Notified Bodies in the EU to significantly improve their communication with companies. These are the gatekeepers of all medical devices and therapies to be introduced. Hence, their view is vital for anything to be done.
The FDA on this end is actually ahead of the EU, as they have a “Q-submission” process, whereby a company can send them information to review and set up an hour long meeting to discuss the planned intended use, clinical claims, planned clinical study etc.
However, there is still much to be improved in this process. It can take 3 months to simply speak to them for 1 hour. In many cases, this is not enough time and a company has to potentially request several meetings in order to get enough feedback to understand what the FDA expects and how to meet those expectations.
Many innovations in healthcare are complex and require a fair amount of discussion. Furthermore, questions that are raised by the FDA might require a company to gather more clinical evidence and this can be hugely time consuming and expensive. Hence, getting it right can make or break a startup company.
Furthermore, the discussion itself is in many cases very limited. The FDA as a regulator tries to be as neutral as possible so as not to be seen to advise the company and hence be biased. This sounds great in theory, but in practice it just means that discussion is very difficult to have, as everyone is defensive. Put simply, due to the limited communication, the company presents their case like a lawyer would do it, and become extremely strategic with their wording.
They try to avoid as many potential questions by the FDA as possible in order to avoid a line of questioning that veers off course and forces them to wait another 3 months to have yet another conversation.
The FDA on the other hand might have some good solutions or suggestions to the company. However, they don’t want to “advise” the company too much, so they don’t say anything or try to limit their input to the point that some of the suggestions given are far too vague.
The FDA is also generally very risk averse, so anything that appears different from what is standard practice will have a much, much harder time getting through. This means several meetings will be required and causes a delay by months to several years.
The result is that the so called “innovations” that go through the FDA tend mostly to be “me too” innovations where the company simply claims they are equivalent to an already approved device. Hence, most devices approved by the FDA are not in fact innovative at all.
Another problem is that the initial team that a company speaks with at the FDA might have changed by the next meeting. This means that they may have to cover the same points that were raised previously or have different issues raised by the new team at the FDA.
This uncertainty causes many startup companies to think twice about submitting anything at all to the FDA, or if they do, they submit a claim that is so similar and low-risk that arguably is not much of an innovation at all.
To be fair to the FDA, they are actively working on this. They do have a breakthrough device program that tries to solve this problem, at least for innovations that might have a big impact. But the point still stands that the FDA’s regular timeline and communication with companies must be improved so as not to overwhelm the startups when they have not even begun their development journey.
If it will take a company a year of discussions with the FDA and $50,000 in consulting fees with regulatory experts just to get clarity on what they need to do, then the whole enterprise might simply not be worth it.
The above discussion on the FDA’s limitation might seem quite dire. However, unfortunately in the industry this is actually considered fairly efficient. In the EU, Notified Bodies don’t even have a process to discuss with companies about getting clearance.
Basically, the company simply has to gather any evidence that they believe is sufficient and send all the data over to a Notified Body for review. There are a handful number of Notified Bodies in the EU and they may all differ somewhat on what they consider is good clinical evidence.
At this point, it’s even worse than normal. Due to a change of regulations by the EU, there is a 13–18 months delay to even consider new device applications. Some Notified bodies won’t even add anyone on their waiting list since they have no idea when they will have enough capacity to review the data.
This all of course assumes that the startup has read the mind of the reviewer and the evidence generated is perfectly suitable. The reality is more likely that they need to go back and gather further evidence/make changes to their claims or innovation and hence get even further delayed prior to getting their device cleared.
How does this impact patients and the healthcare system?
The key point here is that the costs are hidden. Since new devices don’t actually get approved or startups don’t get funding, no one will know what would have been the case if the innovations were approved. The regulators incentive is to reduce risk, so for them an “unsafe” or “unproven” technology was simply not approved. So they don’t feel the impact.
The hidden cost of course is that a new device or a treatment could have improved patient lives or reduced costs etc but since it’s a potential and not an actuality, we don’t see that as readily as a loss. This is even harder to sense when healthcare startups simply don’t get built in the first place.
Notice that the above issues are structural. They are not scientific. I.e If we applied this mode of thinking during the COVID-19 pandemic, then there would be an outcry over the number of unnecessary deaths that occurred as a result. Furthermore, we don’t even know if the pandemic could have been avoided if we had better processes and a different mindset when it comes to innovations in healthcare.
What if we had an innovation to speed up clinical trials, or a better overall vaccine that protects against many more variations of corona viruses. This may be scientifically impossible — I don’t know, but that’s the point. The costs are hidden, so we don’t know what we’re missing.
What we do know is that the regulatory hurdles are becoming so large, that most innovations are either not invented or gets shut down before they have a chance to get through to the other side for us to see their benefits.
All this is supposed to be done in the name of patient safety. However, we also heavily affect patient safety when we make the introduction of new healthcare innovations so hard that they end up never being used.
Healthcare Systems — Streamline clinical admin
OK, so we discussed the problems from the regulators perspective. However, a lot of the issues with regulators are questions of evidence. It makes everyones life easier if the evidence is robust, because that lowers the risk of having the regulators reject the innovation and helps ease their decision making.
In healthcare, the discussion is always around the evidence of the proposed innovation. It is the job of the startup to provide such evidence and hospitals are the main place where this evidence is generated.
Unfortunately, even here we see serious process and mindset issues. Again, the actual science part can be done fairly quickly. Clinicians and healthcare entrepreneurs are generally aware of what evidence they need to show that their technology works and are very motivated in generating this evidence.
However, they are usually severely slowed down by the giant bureaucratic machine of hospitals and health systems. The number of staff required to sign off on a simple low-risk anonymous data collection study can be immense. Any time “patient data” is mentioned, it feels like talking about Voldemort from Harry Potter.
In many cases, all data is anonymised anyway and the startup is not interested in any patient information whatsoever. Usually, they just want to find out if their technology works, and they need to gather text, or images or some blood work to test against.
However, people generally feel so sensitive about this, that it takes a lot of effort to get it done. Just the contracting between a company and the hospital can take months to sort out. Since the number of people who have to get involved to sign off is so large, the probability that someone is on annual leave or simply not available is very high.
This means that even a simple task like a minor contract amendment can end up taking 3–6 months to complete. There is no physical or fundamental scientific reason for this of course. It’s just that there is no urgency or time cost on the side of the hospital to get it done.
We saw during the pandemic that things can indeed get done very very quickly, with zero compromise. It’s just that the priority had changed. In normal practice though, depending on the mood of the person dealing with a particular administrative case, a full trial can get severely delayed.
This isn't just a startup problem. Clinicians frequently complain about these types of issues and are usually just as frustrated by it as the startup founders themselves.
This is a hard problem to solve because it is related to the management of hospitals and a streamlined approach to completing any necessary paperwork should be created.
Not only that, hospital administrators need to have a strict timeline when things should get done. Otherwise, the incentive to simply wait is too strong for them to avoid. This is not about hospital administrators being particularly slow, it’s a general issue with any human. We all need strict timelines to get things done and require external pressure to make sure we stick to those timelines.
Startups need to understand upfront how long the admin processes will likely take so they can plan accordingly. A very thorough assessment of the real admin requirements is very important to carry out. Based on my experience, most of the requirements from an admin perspective has nothing to do with patient safety or even likely legal issues. They are just historical processes that has been set up for reasons that just don’t exist anymore.
Reimbursement & clinical evidence
The core component of any technology or service is the need for reimbursement. This is a huge topic for hospitals and startups alike as anything they do or develop needs to get paid for. Hence, payers play a pivotal role in the healthcare innovation ecosystem.
However, they are very rarely seen or engaged in the process at all. Whether the payers are insurance companies or government agencies, they don’t get involved until all the work has already been done in terms of clinical trials and regulatory clearance.
This however, is a huge issue for startups who want to actually build innovative technologies. By definition, innovative technologies don’t sit neatly on top of existing technologies. This means that a payment for the product may not be there yet and hence even if the technology works perfectly, hospitals will not use them for purely economic reasons.
They need to “tick the box” that is required by payers to get reimbursed for their service and thus the incentive to use a new technology, even if shown to be beneficial to patients will not be utilised.
For example, if a clinic or a hospital won’t get the same reimbursement for seeing a patient via a video call vs in person, then they will never want to adopt that technology, even if it means the patient will have to travel a few hours just to be seen and get the exact same clinical result.
This topic is somewhat of a taboo subject, and sometimes is not even explicitly stated by the hospitals as they wish to be seen as always wanting to improve their service and care for the patients. Their reputation for doing whatever is in the patients best interest must not be damaged.
Hence, they might find other reasons for why they justify not adopting a particular technology. For example, they might ask the company to gather evidence via a study to show that their technology does in fact improve patient outcomes. That way they can get potentially get reimbursement via the payers.
This may be the case even if it’s very obvious that the technology will work. Not everything needs a full scale Randomized Controlled Trial to show that the technology will have a benefit. However, if the mindset is that there is any risk even residual risk, then it needs to be mitigated. The payer will likely also want to see health economic evidence to create a reimbursement for the technology.
In many cases, the payer and the healthcare provider (e.g hospital) are at odds. The payer wants to see lower costs while the hospital wants to increase revenue. This sounds very cynical, but it’s nuanced. The hospital needs revenue to support other functions as they might be losing a lot of money and are not running sustainably.
Using a video conferencing system might indeed make healthy and young patients life easier and avoid them having to attend the hospital. However, those patients might be the very cohort that the hospital needs to charge the payers in order to support the older, more complex patients who need additional services.
Adopting a more streamlined system is great for the admin and those young patients, but it does increase the problem in terms of economics for the hospital.
The payer might be funded by the government and since we all know that our spend on healthcare is completely unsustainable, so their wish to reduce costs is also not necessarily coming from a negative place.
In this video conferencing example, the hospital might decide not to use a system that is better for patients and lowers costs mainly because it might result in less funding. Hence, they will unlikely want to run a study that involves health economic data.
Of course, the payer won’t want to even talk to the startup unless they have clear data showing this improvement in patient outcomes and the lowering of costs.
With the above being said, it is possible to create innovations that please both stakeholders (e.g improve patient outcomes but at the same cost). However, the main issue is to prove this, startup companies will have to run potentially such large studies that take too long and are too expensive to be worth it.
And even then, since payers are not actively involved in any of these early discussions, there is a risk that they might not provide reimbursement anyway.
Payers and providers alongside startups should be in much, much closer contact when it comes to making plans on running studies and gathering economic evidence in a way that makes sense for all sides.
And by this, I really mean something as simple as a process where they meet and can discuss the innovation proposed in a similar way to the FDA “Q-submission” process. But hopefully in a much more frequent and open way.
Technology iteration cycles
The key feature with technology is that rapid iteration cycle enables amazing improvements over time. That is what makes it so powerful. This is done in hardware products to an extent, but in the software world this is probably one of the most important features of the technology.
The number of times that a software gets updated in the consumer market before it truly solves a customers problem is huge. The number of small iterations in the code may go into the thousands.
Now that software is entering the healthcare space, this becomes very, very tricky indeed. Clinical trials are not usually designed to take this into account. They are usually designed based on traditional drug development iteration cycles. And in this sense they are averse to any change to the product.
In practice, it means that once a study has started, no change to the software can occur. And since it is near impossible to know how well a software product might do upfront, it is very unlikely to succeed in its first version. As the cost of each iteration cycle is simply too high and a company can’t run a clinical trial for every version of the software that is developed, they might just decide to avoid building a medical product in the first place.
For this reason, healthcare software tends to be so outdated. Very few companies want to commit to only developing one version of their software product as it loses one of the biggest advantages, which is its ability to be changed and improved rapidly.
Well, what’s the solution?
Well, one way to solve this problem is to design studies that take advantage of this feature. The study might be designed to be adaptive, and not static. If there is no major risk to patients, then the changes should happen as frequently as required.
It is also possible to “baby” step the approvals of software, whereby it is only used in a limited way to reduce the risk and as the changes happen, then have a process set up that can test the change without affecting the patient.
The details of this process is perhaps too nuanced to discuss via a post on Medium, but suffice to say that it should be doable. However, it requires very good communication and open mindedness from both hospitals and regulators.
Regulators might see risks in this approach, but they should not ignore the potential benefits either. “A limited use” approval might give hospitals and the startup companies an ability to use iterate on their technology, and do a simple submission once they wish to upgrade their approval and extend the intended use.
This sort of happens already with drugs being used “off-label”. In a similar way, software can be used relatively safely even when they are still in development.
What’s the Conclusion?
Healthcare innovation is broken because of structural issues due to administration of the healthcare system and a misguided mindset of perceived risk. Very little cost is placed on missed opportunities of new technology and almost no cost is associated with the time it takes to develop new technologies.
These issues don’t enhance patient safety or improve patient outcomes, although they are perceived to do exactly that.
The result is that even relatively mature technologies struggle to get adopted by the healthcare system and misaligned incentives between payers and healthcare providers exacerbate this issue.
The speed of iteration is not well understood by the healthcare system in general, even though it’s what makes the technology so powerful in the first place.
Finally, there are solutions to these problems. Ironically, these solutions themselves don’t need a new technology. They simply need a mindset change and simple processes set up related to how startups and other stakeholders communicate that can help alleviate a huge number of problems for all parties.
