Nov 4, 2026 · 27 min · 10 segments
The patient in front of you today could help the patient someone else sees tomorrow. Dr. Amy Abernethy has spent her career making that possible, and she’s worked at nearly every stop that data passes…
Amy AbernethyGuest
Scott SigmanHost
That sounds like a crazy number, but I've got the whole system in play to sort of keep people moving.

What happens to all of that information? How do we maintain the confidentiality of HIPAA, but still be able to get that data and use it?

There was a time when we were writing the legislation, for example, around high tech that we sort of had what I used to call push button thinking.

Like if we could just have everything electronic, then there would be some magic button you push.

And bing, bang, boom, that information was now going to be ready to be analyzed so we can learn from every patient.

And along the way, we learned the importance of, first of all, making sure information that is stored is now transported and modified into a common framework, essentially a lingua franca.

through standardization and harmonization and other technical activities that now makes it able to be combined together with data that's coming in from not only other patients in your orthopedic clinic or other patients in your health system, but also other patients in Oregon and Maine, right? And so that the data can now actually be aggregated together so that in order to learn for the care of the next patient who's looking similar and is in your clinic, you actually have a cohort a population of individuals that have similar features and now can provide the kind of information ready for analysis.

It's also important that data are now prepared in a way that combines together not just the clinical notes, but also other data like the laboratory data and the radiology data and maybe the genomic information or other kinds of data sets, and that that data is now stored in a way that it's ready to be analyzed.

So correct me if I'm wrong, but it sounds to me like what we can do is collate all of this information.

We're going to have to filter it because my mother, Judy, doesn't want her data just out there for the world to be identified.

So it has to be de-identified in a manner that it could be collated so that we can use that information.

So share that process so our listeners can feel good that their data is going to help humanity, but not necessarily give up any personal details that they don't want to have exposed.

As information comes together and the data sets are prepared in a way that is on the path towards analysis, before you get to a place where you hand it over to the people who are doing the analysis, you want to actually strip out any of the details that might identify your mother as herself or another patient who might call her Patrice as Patrice.

or make it possible to trace back to this as Patrice in any way within the context of the data.

That process called de-identification is particularly important and is a technical process.

We're also going to have an expert cross-check, take a look at how the data set has been prepared, and say, you know what? In addition to the standard technical de-identification processes, we'd like you to add in these extra steps in order to absolutely reduce the risk of any re-identification.

How does this real-world data, we've got it now, we've got this information, how is that going to help to change what we do?

As data sets now come together, so let's, for example, use a data set around people with non-small cell lung cancer, some of whom have received immunotherapy and some of whom received standard chemotherapy, and you want to understand what is the potential incremental benefit of immunotherapy.

So as these data sets first are de-identified and prepared for analysis, It allows us then to analyze and ask a question like immunotherapy as compared to standard chemotherapy, but also drill in further.

That sounds like a crazy number, but I've got the whole system in play to sort of keep people moving.

What happens to all of that information? How do we maintain the confidentiality of HIPAA, but still be able to get that data and use it?

There was a time when we were writing the legislation, for example, around high tech that we sort of had what I used to call push button thinking.

Like if we could just have everything electronic, then there would be some magic button you push.

And bing, bang, boom, that information was now going to be ready to be analyzed so we can learn from every patient.

And along the way, we learned the importance of, first of all, making sure information that is stored is now transported and modified into a common framework, essentially a lingua franca.

through standardization and harmonization and other technical activities that now makes it able to be combined together with data that's coming in from not only other patients in your orthopedic clinic or other patients in your health system, but also other patients in Oregon and Maine, right? And so that the data can now actually be aggregated together so that in order to learn for the care of the next patient who's looking similar and is in your clinic, you actually have a cohort a population of individuals that have similar features and now can provide the kind of information ready for analysis.

It's also important that data are now prepared in a way that combines together not just the clinical notes, but also other data like the laboratory data and the radiology data and maybe the genomic information or other kinds of data sets, and that that data is now stored in a way that it's ready to be analyzed.

So correct me if I'm wrong, but it sounds to me like what we can do is collate all of this information.

We're going to have to filter it because my mother, Judy, doesn't want her data just out there for the world to be identified.

So it has to be de-identified in a manner that it could be collated so that we can use that information.

So share that process so our listeners can feel good that their data is going to help humanity, but not necessarily give up any personal details that they don't want to have exposed.

As information comes together and the data sets are prepared in a way that is on the path towards analysis, before you get to a place where you hand it over to the people who are doing the analysis, you want to actually strip out any of the details that might identify your mother as herself or another patient who might call her Patrice as Patrice.

or make it possible to trace back to this as Patrice in any way within the context of the data.

That process called de-identification is particularly important and is a technical process.

We're also going to have an expert cross-check, take a look at how the data set has been prepared, and say, you know what? In addition to the standard technical de-identification processes, we'd like you to add in these extra steps in order to absolutely reduce the risk of any re-identification.

How does this real-world data, we've got it now, we've got this information, how is that going to help to change what we do?

As data sets now come together, so let's, for example, use a data set around people with non-small cell lung cancer, some of whom have received immunotherapy and some of whom received standard chemotherapy, and you want to understand what is the potential incremental benefit of immunotherapy.

So as these data sets first are de-identified and prepared for analysis, It allows us then to analyze and ask a question like immunotherapy as compared to standard chemotherapy, but also drill in further.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.