Kevin S. Hughes, MD, FACS, is the co-director of the Avon Foundation Comprehensive Breast Evaluation Center at Massachusetts General Hospital in Boston. He is also the co-founder and medical advisor for Hughes RiskApps, a software company focused on cancer risk assessment. Learn more about Hughes RiskApps at www.hughesriskapps.com.
Q: Why do you feel it is important for organizations to incorporate screening and preventive medicine measures into their cancer and breast programs?
Dr. Kevin Hughes: Screening and prevention measures allow us to find cancer at an earlier, more treatable stage and sometimes to prevent cancers altogether. When we can find cancer earlier, it is easier to treat, there’s less deforming surgery and there’s less need for chemotherapy. In addition, there are some women whose risk is so high that preventative medication — what we call “chemoprevention” — is extremely useful, and in a smaller subset, even removing the breast in advance — prophylactic mastectomy — can have significant value.
It all comes down to the ability to perform a risk assessment and determine which patients are at the highest risk and which are at more average risk.
Q: Why is performing a risk assessment a particularly valuable measure to integrate?
KH: There are ways of determining how high a woman’s risk of breast cancer actually is. That is going to vary by her age. But you can and should be much more specific than just saying younger equals less risk, older equals higher risk as there are certainly instances when some younger women can have an extremely high risk, and require very intensive screening and prevention measures.
This is why you need the means to segment your population — which women who are younger than 40 might need mammography; which women at any age might need MRI, a more intensive screening approach; which women will need chemoprevention; and who needs genetic testing to try to identify a mutation in a cancer-causing gene. By segmenting the population — stratifying by risk — you can determine the right management for the right patient.
Q: What role do you believe technology and care coordination can play in improving screening and care coordination efforts?
KH: If you look at the guidelines for how we manage patients, they are difficult for a human to deal with. For example, one guideline talks about when you should do an MRI, suggesting screening breast MRI if the patient’s lifetime risk is over 20% according to one of three models: Tyrer-Cuzick, BRCAPRO or Claus.
On the surface, it doesn’t sound that hard. But then you must ask yourself how you run Tyrer-Cuzick, BRCAPRO and Claus? How do you have the time to collect all of the data needed for each model and put it into a computer to run each model? Then how do you get that patient booked for an MRI? Then how do you make sure she gets one every year? And when she comes back every year, how do you make sure she’s still over a 20% lifetime risk and that her risk hasn’t dropped below that threshold? The idea that humans are going to be able to do all of this in their heads — Not likely! This is why I think this is a perfect area for technology. We know 99% of the women who need MRIs don’t get them; 95% of patients who need genetic testing for hereditary breast cancer don’t have it done. Just look at any screening test and it’s grossly underutilized. Note: To view a video of Dr. Hughes discussing the benefits of genetic testing, click here.
You have to identify the patients, segment them into risk categories, identify for each risk category how you would manage them — what screening or other intervention you would do — and then make sure it gets done again on a regular basis. It is critical to confirm compliance with screening and medications.
All of that goes into the care coordination component. These patients are not just one shot and done. We need to get them onto this trail and make sure they keep going on it for years into the future.
From a risk assessment perspective, the problem we have right now in terms of getting it done for all patients is that risk assessment requires a large amount of data collection. You need to identify from the patient her family history, history of pregnancies, her hormonal use, birth control use, etc. Collecting all of that information is time consuming, and, unfortunately, physicians have less and less time to do it.
With that in mind, we developed Hughes RiskApps.
Q: What is Hughes RiskApps?
KH: Hughes RiskApps is software available as a web service that allows patients to enter their own data either online through a website or via a tablet in the physician office so that the data collection piece is all done on the patient’s time without taking up staff or physician time.
Once that data is in the computer, even if you were to present it to the physician for a fast review, most physicians lack the expertise to stratify risk just by eyeballing the information the patient has delivered. So a physician would be better off having risk models that can help tell them what to do.
What RiskApps does is run a variety of risk models on this data. These are algorithms that have been developed over time at various academic institutions that help to categorize risk more quantitatively. Then, using that risk level, guidelines exist saying what should be done for different patients at different levels of risk. This helps a physician make the best care decision for the patient.
We work very closely with Dana-Farber Cancer Institute in Boston where the web service exists. I work with Giovanni Parmigiani, who is the originator of BRCAPRO, one of the major risk models we use. Dr. Parmigiani and I try to be sure the web service always has the latest and most up to date models.
Q: How many users do you have to date?
KH: We now have about 60-70 centers around the country and a few around the world that are using our software. Each software package uses the web service to run the risk calculations. We’re running about 15,000 to17,000 risk calculations a month, which means that patients are benefiting from our service.
I’m a surgeon who is used to treating maybe 100-200 breast cancer patients a year, so numbers like 15,000-17,000 a month are staggering to me, and I think we’re just scratching the surface.
Q: What clinical need were you hoping to meet by developing this software?
KH: We’re doing this for breast cancer risk, but we’re also starting to do this for colon cancer risk, where the need is as great or greater. The basic approach of the patient entering data, the computer running risk algorithms and the computer then helping the physician know what to do next really can apply to almost all of clinical medicine.
While we started out with breast cancer risk, moving out into other cancers and their risks is on our timeline, as is moving out into benign diseases, such as cardiac disease, diabetes, etc., where the same need exists.
It’s been a great collaboration between Dr. Parmigiani and myself in looking at these models and developing ways to make them more useful on the day-to-day level.
Q: What role do you hope technology will play in the future for determining treatment and care coordination?
KH: In the next five years, I would like to see more systems using our approach. It’s the basic idea of patients entering their data, algorithms being run and guidelines being followed without humans needing to remember every last bit of information. That is where I think medicine is going in the next few years.
I think EHRs will start to become more modular. EHRs have to learn to work with other software entities. Hughes RiskApps is now embedded in Epic. Epic didn’t have to develop all the risk calculations on its own; it just used ours.
An external module for risk assessment for breast cancer, an external module for management of osteoporosis or an external module like Cordata, which would trace a patient’s care from their diagnosis through their management, could be built as modules that latch onto any EHR rather than each EHR building their own from scratch. It doesn’t make sense for each EHR to start from scratch when so many useful software packages exist.
I’ve put a lot of work into trying to make systems interoperable. Our system and web service is HL7 compliant. HL7 is the major language for interoperating across EHRs. We not only use the web service to run risk calculation for Hughes RiskApps; we also use it to run risk calculations for Epic and other hospitals and software packages. If you make these things with a basic interoperable message, they now become usable by multiple systems.
The age of the physician doing everything from memory is a thing of the past. We cannot keep up with the massive increase in the amount of information and guideline-based approaches. This has to be done at a much more systematic level, and that means developing and using good software systems.
Note: To view a video of Dr. Hughes discussing breast density, click here.