Through new telehealth technologies and increased data analysis physicians are gaining insights into patients like never before, allowing them to facilitate early interventions, improve adherence, and reduce readmission rates -- not to mention at a price more affordable than ever. The companies you’ll hear from in this session are using a healthy and innovative mix of data, educational tools, sensors, and more to improve patient outcomes.
4. Data is driving the future…
From the University College London
Big Data Institute
5. …of improved health outcomes
Decoding the Top 10 Buzzwords of Healthcare Big Data Analytics
Jennifer Bresnick, Health IT Analytics, Nov 16, 2015
Learning
Health
System
Healthcare
Informatics
Precision
Medicine
Population
Health
Mgmt
Clinical
Decision
Support
9. 8
Wireless and comfortable
Device form factor and comfort allows longer wear
time and improved compliance
Barrett P., Komatireddy R., Haaser S. et al. Comparison of 24 hour Holter monitoring versus 14 day novel
adhesive patch electrocardiographic monitoring. Am J Med. 2014;127: 95.e11-95.e17.
36. VitalConnect-PhysIQ System detected change 10
days in advance of hospitalization.
Multivariate Change Index – Daily Average
Hospitalized
Sept.17
Initial
Discharge
Improving
Declining
Discharge
Sept. 22
10 days
advance
warning
37. Meaningful IoT solutions for remote patient
monitoring depend on:
Comfortable, easy to use, reliable solutions
Accurate, comprehensive and timely data
Editor's Notes
I’m going to focus today on the role of data generated from IOT devices in the future of remote care and mobile solutions.
I’m with VitalConnect and we are developing innovative models of care with our partners.
The VitalConnect Platform includes a patch like the one I’m wearing. This device collects patient data and transmits it by BLE to a phone or other relay and then to the cloud where data can be received by the care team.
To generate remote and mobile solutions in healthcare, we need good inputs. I’m going to focus most of my talk today on the patient and the starting point of the system…Here
of high quality data that is reliably collected from patients as the foundation of the digital health ecosystem and the big data revolution.
Everywhere you look, big data will take us to new places and solve many complex problems.
And in healthcare, big data is linked to opportunity for improved health outcomes. It is also implicit in the buzzwords around digital healthcare
However, while Google may have done a great job of taking unstructured data in tracking the progression of the flu. Google beat the CDC with rapid data, lots of data, lots of people participating. A lot of the data may have been junk, but they had more to balance it out and were just establishing trends
Individual patient data and treatment responses look different. You might know the flu is happening in your area, but google is not telling you whether or not you have the flu, need to take tylenol now or need to go to a hospital because you’re dehydrated. These treatment paths are based on timely and accurate data collected from the individual patient.
Today, I won’t be talking about trending and epidemiology, but about individual patient care. This is our patient. Not only are we going to have an individual patient, but we’re going to have one data point. A heart rate value
I like HR because we all know what it means. It seems straightforward
But what do we need to know about this number when we see it? What is important?
I’m going to start by saying it’s important and non-trivial that we as the care team are actually looking at this data. That means the data was actually collected and got to you the medical professional to view.
How would our patient like this?
Better?
Seems like an obvious conclusion…Your patient is more likely to wear something comfortable and unobtrusive
That has been a big part of our design philosophy at VitalConnect. This is our HealthPatch MD product. I have one of these on down here and wear it frequently.
Looking at a recent report from Argus Insights, you can see that even within a class of similar devices, there’s a range of consumer delight.
Of course the meaningful metrics are a bit different for medical devices. We have two key design guideposts – comfortable and easy to use
”…logistical problems with syncing bands to apps and data across devices lead to widespread frustration”
”…apps crash and fail, leading to infuriating load times and even lost data."
A consumer may want to have an interesting life experience with their fitbit, but a nurse wants a new device to not add significantly to her overburdened work schedule and a patient doesn’t want much impact on their life.
Easy and fast.
Get the patch on the patient and data flowing as soon as possible
For that reason, we just launched a fully disposable version of our patch. Sound like a minor enhancement, it overcomes significant objections that could be show stoppers in implementation.
But let’s not stop at the hardware aspects of ease of use
Single use, disposable
No cross-contamination risk
No need to disinfect between uses
No management of reusable components
No risk of losing reusable components
Fewer steps, improved work flow
Peel and stick
Efficient for nurses at a hospital or office
Easy for patients at home
Small, discrete, comfortable factor
The app has a large impact on product acceptance. You cans ee for this consumer example, there was a lot more delight with the hardware than with the apps.
”…logistical problems with syncing bands to apps and data across devices lead to widespread frustration”
”…apps crash and fail, leading to infuriating load times and even lost data.”
Our goal is fast and reliable. Get the patch on the patient, get data flowing and be done. The nurse or patient doesn’t want to restart the app overnight. BLE autoreconnect is key.
We had been working with partners to develop apps specific to their clinical need, but we realized that this focus on fast and reliable was independent of the clinical specifics and this year have been designing an app this year which will be used in its first inpatient clinical study in April.
Here’s a view of our data screen, but this is not the most important aspect of the system. Easy and fast to this screen.
Now my patient is connected, but where does the data go…
I used to work for an ENT company and our sales teams supported cases for balloon sinuplasty because rotating technicians weren’t comfortable using the inflator.
I won’t spend a lot of time on interoperability, but obviously it’s key to this equation. I talked to a community hospital in the midwest at HIMSS. Their solution to connecting EHRs at certain specialty clinics into the main hospital EHR was a pdf file import. That is the state of technology today.
API is essential. Our partners have access to our API for iOS, Android, and secure server. We’re investing to make the API easier for a new user to understand and start using.
BLE is great, but we can’t use standard protocols with our data.
HL7 is great and we’ll use for importing subsets of data into EHRs, but not the best for continuous upload, analysis and response to data.
So back to our data point. We’ve collected the data from the patient and connected patient to the system.
But what else is relevant about this data point?
Was the raw ECG or PPG taken with good equipment, was the algorithm for going from ECG to HR high quality, was this an erroneous point possibly motion artifact, how long was HR at this value, is this an average over a long period, when did this happen and what happened before and after, what other data exists for the patient
A data point may seem like a non-negotiable fact, but it’s a bit more nuanced.
In the 60s Walter Cronkite presented the news as a set of facts. We know today that from Fox News to John Stewart, the apparent facts are affected by the sampling, selection and presentation.
Digital health data may be quantifiable and come out in numbers, but you can easily add apparent significant digits that are within the accuracy and precision of the device. You can easily lack data that is relevant to the patient’s condition, and you can easily receive data too late or not in sync with other data sources.
Accurate, comprehensive, timely data is key
Comprehensive data is step 1. Putting a single number in context.
In addition to comprehensive data… timely data is key. Continuous data allows fast response. In comparison to current standard of care outpatient that might involve data collection only during a monthly doctor’s visit.
Sepsis… every hour of delayed diagnosis changes a patient’s odds of survival significantly.
Inpatient checks may be every 6-12 hrs as the current standard of care and outpatient might be a week to a month
And at the center of it all is accurate data. Garbage in, garbage out definitely applies to advanced analytics. Your foundation is your data. It must be collected from the patient, comprehensive and timely. But all of this is for naught, if it is not sufficiently accurate.
We have spent significant time validating our measurements that has been part of regulatory submissions.
And for our last dive into this simple HR measure, let’s look at the type of data. Is this instantaneous point in time, averaged, or an alarm based on a threshold. How are these different types used in remote patient monitoring
All of these are refinements that go from raw data to data that is meaningful and can be acted on
We are collecting hundreds of MB per patient per day on a single VitalConenct sensor. How is all of this data used? We need to go from raw patient data to usable, actionable data
What’s the physician response to this? Please make it stop. I don’t have time for this.
There’s my 120 bpm. Does is mean anything?
How do we go from data to information?
This is starting to look better, but there’s still a lot to wade through and no obvious area to review. And of course, this is just 50 min.
And there’s my 120 bpm. Any more meaning to it now?
It still takes a long time to look at this?
Can I see all that I need to in a 24 hr view or I need to further zoom in?
What’s noise like motion artifacts versus a real area of concern?
Now this is starting to be more actionable. I have an alarm off a number of interest and I know what happened.
This seems simple, but of course we haven’t figured out how to do alarms well yet. Alarm fatigue?
So we need Reduce sensitivity, change the thresholds, add multiple parameters… But we have to be careful as we get more complex
Models are our future, but it is hard to make them perfect. There is a balance between minimizing false negatives and overalarming. In most cases, we are not smart enough to avoid physician review even on simple tasks. But we are moving from single parameter to multiparameter alarming and from alarming to diagnosing
The most that can be expected from any model is that it can supply a useful approximation to reality, also from Box in 2005
Sensitivity and specificity
Need for physician confirmation (regulatory implications)
I’ll wrap up by saying that …
We partner with other companies on remote and mobile solutions. A few of our partners…
Care innovations - remote and cloud-based healthcare delivery at home
Philips - tailored ambulatory care at home managed through a telehealth center
Validic - cloud-based technology platform that connects patient-recorded data from multiple sources
MediBioSense has helped us design our app. They also have a multidevice app that is being piloted by the Scottish RedCross that includes our HealthPatch and other devices for emergency response in that red backpack
We are also integrated with the BePatient patient portal that connects patients to educational material, to other patients at similar states of disease, and to their care team and their data. It’s been used in multiple sites in France post bariatric surgery and we have two upcoming CHF studies on the West Coast this year.
PhysIQ has combined our HealthPatch MD signals with their advanced FDA cleared personal baselining algorithm. It’s currently in studies at 4 VA sites and 2 other hospitals for postdischarge care. This is an impressive result from the VA CHF study that is currently enrolling that shows deterioration 10 days prior to rehospitalization and shows in hospital recovery.
FDA-cleared proprietary algorithm relying on multiple VitalConnect data streams to create a personal baseline for each patient.
Deterioration may be caught earlier in comparison to a patient’s personal baseline rather than compared to population averaged signs of deterioration.
CHF study underway for post discharge monitoring at 4 VA sites with the goal of preventing readmissions.