Pursue economical, secure, resilient and nimble information technologies that support knowledge-sharing across diverse groups of collaborators; enable evidence-based clinical decision models to be built, evaluated, and continually evolved through knowledge feedback loops between clinicians, researchers, Informaticists and patients; and select data sets, algorithms, presentation outputs, transport protocols, and data stores that met the requirements of particular use cases. Abandon the notion of a silver bullet; there is no single right way to share and learn.

Open dialogue like this is invaluable in creating a culture that recognizes that data sharing and analysis in a responsible, evidence-based manner that applies rigor and excellence will advance science and medicine. I’ve shared some information about our experience with the YODA project in which clinical trial data are shared. Another excellent example of scientific collaboration is OHDSI, the Observational Health Data Sciences and Informatics program, which is an international network of researchers and observational health databases coordinated through Columbia University. I think efforts in both of these areas – clinical trial and observational data – will lead to new insights and better care of patients and I really appreciated all of the insights from this discussion.

Most patients are not interested in viewing, downloading and transmitting their records unless they have a serious illness. I think Gary's list is terrific!

There is a debate raging in the academic community now regarding traditional models of publishing and open access that relates to this post... Peer-reviewed quality academic journals are part of solution, but not sufficient on their own ... there are characteristics of published manuscripts that add value and are more often being used, such as supplementary content that includes the data, methods allowing for reuse and reproducibility , which not all publishers do ... this is important to facilitate a learning health system ... There is also the option of publishing in peer-reviewed conferences, which is frequently used by computer scientists to speed the bench to public cycle and can be attractive... around all this, having great indexers and curators of data is important so that the right data, credible data can be found. The ability to include social interaction around the data can be useful as well, (such as this platform). An area for development may be a rethinking of how we value tools/content value for data sharing.Potential measures for inclusion on evaluation of data sharing models, tools suggest as start: Availability and Access; Findability; Usability; Comprehensibility;Quality (completeness/cleanliness); Linking and Combining; Data Provider Support; and Timeliness. (have a working paper with more details on some of these....)

In Massachusetts, one of only seven states that require an opt-in consent process, opting-in is a major problem. It requires resources from the place the patient is visiting and also the central HIE itself. It increases cost and makes busy work for intelligent people who have other needs and concerns.

Most HIE vendors charge way too much for the platform software. Our regional HIE (Wellport) could only have happened because of the generosity of our platform vendor. Our first five years were very affordable for our 200 physician IPA, and produced one of the most robust HIEs I have ever seen. With a change on leadership, or a sale of the vendor, or a bait and switch, it is possible for someone to see a marked increase in price that is unanticipated at the end of the first contract. That can mean the demise of the HIE.

Persistent technical barriers to interoperability are the failure to (across exchange instances):
• Ensure persistent evidence of who did (took) what (action), when, where and why.
• Ensure persistent evidence of who documented (authored) what (content) and when.
• Ensure persistent evidence of health data/record attestation/attribution, non-alteration of content, and preservation of context, provenance and meaning.
• Maintain fidelity to source (i.e., the source of truth at the point of origination), whether or not health data/record content is transformed in the course of exchange:
> From source representation to the representation required in an exchange artifact, whether a message (e.g., HL7 v2, NCPDP), document (e.g., HL7 CDA/CCDA) or resource (e.g., HL7 FHIR);
> From the exchange artifact to the representation required by the receiver.
(Record/data transformation typically results in alterations, omissions or errors.)
• Continuously and consistently yield gold nuggets from an avalanche of often irrelevant exchanged data fragments. The true “gold nuggets” in health data/records must (as the result of interoperability):
> Be readily accessed and discoverable, 

> Show provenance from their source, 

> Bear evidence of truth,
> Be shown in full context without loss of meaning, 

> Be fully relevant (to the condition/task at hand), and 

> Be immediately actionable.
• Ensure emphasis is on qualitative rather than on quantitative measurement of interoperability achievement. Enumerating transaction volumes without measuring actual fitness for use (of exchanged health data/record content) is of little value.
• Banish the current intransigency that the collection of so-called “interoperability standards” offer a proper path toward achieving anything close to true interoperability – in fact interoperability that actually supports:
> Primary use;
> Integrity of the clinical process; and
> Most importantly, patient safety.

I wonder if Joseph’s concern about HIEs could be resolved by Adrian Gropper’s HIE-of-one model presented in his earlier comment: “Patient-controlled independent health records are real.” As I previously discussed, centralized data stores have their use cases (for research and other purposes), they have downsides that could be addressed with distributed data stores and point-to-point data exchange.

I concur with much of what Gary stated. I think it would be productive to examine each item in his post for possible solutions. For example, could app-based algorithms on sender and receiver ends provide the necessary interoperability intelligence? Could those apps evaluate the data payload before being sent and upon receipt, do any necessary transformations and filtering, perform requisite tracking, and provide feedback and alerts regarding the adequacy of the payload’s contents.

The beauty of our HIE is that all clinical records are instantaneously available at all times. It meets almost every requirement Gary listed, but depends on the EMR to be able to send all the clinical data. Most emrs that are certified can send all the data reliably and immediately as patients leave the office or hospital (when and if the note is created and signed). Unfortunately, because we are in an opt-in state, we have 415,000 individuals patients’records but can display only the 25,000 of those who have opted in.

If only the HIE could solve the problems I previously identified. Unfortunately by the time it hits the HIE it's too late (in the data/record flow) and the HIE can't fix it. In my experience: 1) the source system does not capture, retain or convey key information; 2) transformation to/from the exchange artifact is lossy (to content, to bindings of authorship/attestation/attribution and provenance, in translation to/from various code/value sets). Many disjunctions (errors, alterations and omissions) occur before/during/after almost every transaction. What is finally displayed to the receiving clinician is often at least two orders of derivation from the actual source data/record content. This is an ongoing risk to clinical practice integrity and crucially, to patient safety. The EHR/EMR system testing/certification process is mostly superficial and reflects ideally configured/populated exchange artifacts (orders of magnitude removed from the real-world).

Having records quickly accessible is important, no doubt, whether the data are centrally stored, locally stored, or access from a distributed network. The opt-in issue might be resolved if patients were in control of their data, but that's another issue. It would be interesting to compare the relative benefits and advantages of different data sharing models in different real life use case scenarios, including their ability to address Gary's issues, handle security risks, exchange data across boundaries (e.g., between HIEs), provide rapid data access, be resilient in disaster situations and when internet connectivity is disrupted, be low cost to build and deploy, avoid opt-in problems, etc. I believe the various models can be implemented in a complementary manner that would covers all bases.

Gary is right that some clinical data may not be complete or accurate, and that could be compounded by an HIE. But that is true of medical records themselves. Perfection is the enemy of amazing. Our HIE shows future appointments in our community for most emrs as an example. Our case managers love that feature and have learned which practices don't show them.

Well, Gary addressed one of my questions: Something needs to be done BEFORE the records reach the HIE. This gets into the need for technologies smart enough to know, a priori, how to make sure the data to be sent will be of the best quality and usefulness for the particular recipient(s). As such, it seems to go beyond interoperability, per se, and focuses on how to share and build clinical knowledge.

I've always considered that true interoperability starts at the point of health data/record origination and ends at each ultimate point of data/record access/use. To this end, I've acted as author/project lead on ISO 21089, Trusted End-to-End Information Flows, initially approved/published as a 2004 ISO Technical Report (informative), recently updated/approved (but not yet published) as an ISO Technical Specification (normative).

Colleagues, I'm joining the conversation late but agree with many/all of your comments. As an emergency physician & researcher who works closely with our (also opt-in) state HIE, I echo @Joseph Heyman's frustrations!

Other thoughts:
* I like the idea of the personal HIE - but this may not solve the issue that I face in the emergency department, where I don't have a pre-existing relationship with the patient, and where the patient may not be in a position to upload her/his data to my EHR via API (I would presume that they need to give consent in order to share their data).
* I am also concerned about the very real limitations that a personal HIE would put on our ability to use data to inform the learning healthcare system. We need large quantities of aggregated data to find patterns & to measure improvements.
* I want to echo others' comments about the importance of standards for provenance and transformation of data between different systems. As I work with my HIE's data for both clinical care improvement and for research purposes, we are repeatedly challenged by a lack of a data dictionary / lack of clear info on what got transfomed by whom. One of my students presented a paper on this at AMIA last month - even the data on patient disposition is full of errors!
* No one has discussed the issues of (a) cost [of implementing standards for data, like Gary's comment above] and (b) navigating the challenges with clinicians [each of whom wants to handle their clinical data in their own way]. These are both major barriers to retrofitting EHR/HIE communication to new (or existing, but not-implemented) standards.

Megan makes good points and they call for a hybrid solution, imo.

• A patient’s personal HIE may not be available in an emergency, but his/her providers would have their own records that could be sent to the emergency personnel. Also, inexpensive medical records on a USB drive have been around for a while (tinyurl.com/yabdv4oa). Using all these options may be ideal.

• In addition to HIEs, there are various ways to aggregate patient data for research. Some systems query and harmonize data from multiple disparate sources using common data models (e.g., signetaccel.com/). The Structured Data Capture project uses electronic forms with a standardized schema and transport methods that populate research registries (tinyurl.com/ydcv5ej9). EHR companion apps (“bolt-ons”) can be programmed with rules that automatically export de-identified data to central repositories. Again, multiple options exist.

• Whether HIE or not, it would be great to have a universal standards dictionary that maps to all existing clinical concepts, tests, codes and value sets. A daunting task, if possible.

• I don’t know about an HIE, but if a universal standards data dictionary is available to an app, it need not be expensive at all to implement requisite data transformations, and if different dictionaries are needed for different recipients, that, too, need not be difficult or costly if the recipient’s needs are known.

• I’m not sure what is meant by “navigating the challenges with clinicians [each of whom wants to handle their clinical data in their own way].” If it’s a workflow issue, then a clinician-facing app can use a configurable rules engine to navigate/handle their data the way they want.

The single most technical hurdle to data sharing - and rapidly the most consequential one -- is Semantic Interoperability -- as measured by [1] the scope of diverse semantic representation of data elements and concepts in healthcare [2] the volume of that inbound data [3] the velocity of that inbound data streams [4] the convergence of healthcare, wellness, social (determinants of health), life sciences and transnational medicine data (omics) [5] optimizing value and economics of the scope of reference terminology and code sets to map these to (e.g. ICD, CPT, RxNorm, LOINC, SNOMED, etc etc etc) [6] the unsynchronized rate of change of each of the selected reference terminology and code sets -- translating to version synchronization strategy / policy / economics challenges [7] Configuration management of 1-6 [8] talent to figure out and manage 1-7 [9] funding to make 1-8 a reality

Ron's comments are spot on! One thing I can offer technologically is a familiar data structuring/modelling tool that can provide a useful way to address the semantic interop issue as described. It's a Spreadsheet-based Software Framework (SSF). I've been working on a diagram that describes an SSF through a deep understanding of a spreadsheet cell that comes from 35 years’ experience; the diagram is attached.

An SSF is used to build apps that provide new ways to take advantage of a spreadsheet’s power, flexibility, and simplicity. At the core of an SSF app are grids composed of millions of spreadsheet cells. Each cell is a content container with a built-in computation engine that provides a rich set of capabilities for inputting, storing, modifying, calculating and displaying data (described below). SSF apps organize cells into different data models that: (a) harmonize disparate data for aggregation and integration using maps, (b) perform analytics and create charts using data arrays, and (c) enable data inputs and report outputs using lists and tables. These cells are embedded in worksheets, and the worksheets reside in workbooks. The workbooks also provide macros for process automation that includes data access, transformation, organization, storage, transmission, analysis, and rendering, as well as utilization of user forms and integration with third-party tools.

#Semantics is the next big thing in Health Information #Interoperability per @Gartner_inc requires new #Framework twitter.com/RonRibitzkyMD/stat...

More about the TRANSLATIONAL MEDICINE STACK in Chapter 2: "Domain frameworks and key concepts" of the report to the National Science Foundation (NSF) "Knowledge Mining & Bio-informatics Techniques to Advance Personalized Medicine: The Case for White Space R&D" at cake.fiu.edu/Publications/Ribi...

Let's imagine that the following 'forces' are formidable, clear and present disruptors: Medical Devices connected in IoT, are AI enabled, and Blockchained.

Now, let's imagine the data streams I'v discussed previously going through that disruptive environment.

Need to be more concrete? So let's examine the confluence of aging population and smart medical homes.

Looking for a specific example? check this 'Aging-in-Place innovation day in Singapore' at linkedin.com/pulse/aging-in-pl...

Just to make it a bit more exciting, let's add the 'reduction' of AI to an AI-purposed 'Neuromorphic' CPU such as Intel's LOIHI youtube.com/watch?v=EgCRwZw4p8... (or GPU, such as Nvidianvidia.com/object/machine-lear...) embedded in some or all of these IoT'd Blockchained medical device.

Then, let's imagine that environment operating w/o semantic integrity of inbound, in-process, and outbound data streams.

Can we?

Hence the thought that Semantic Interoperability is the the single most technical hurdle to data sharing - and rapidly the most consequential one to progress -- clinical, scientific, and economic.

Worldwide.

Here we go...

In Stephen Beller's world (thanks for the kind note!), SSF means "Spreadsheet-based Software Framework"

In Ron's world, SSF means "Semantic Services Framework", designed to eliminate data ambiguity and enable extraction of semantically vetted facts from unstructured and partially structured text (the latter is 'believed' to constitute over 80% of medical content)

Go figure!

I've started reading Ron's book online (see the link in his Semantic Interoperability - Reference Points post) and have thus far found it a great resource focused on translational research and personalized medicine. There appears to be a deep convergence in many aspects of our understanding about issues related to the creation, evolution and use of knowledge in healthcare. I just want to say that the AMA/IHMI--by bringing people together from diverse backgrounds who openly share and discuss what they know, perceive, and envision--provides great potential benefits to our professions and patients.

With our YODA experience, Johnson & Johnson shares both summary (clinical study reports) as well as de-identified individual participant level data with researchers. We have had over 70 requests from around the world - the vast majority have already been approved by the Yale/YODA group and the number of publications submitted or accepted (7 thus far) exceeded our expectations. For those looking for more details, the policies and experience with requests and approvals are transparently listed on the YODA website (yoda.yale.edu/).

This is a great example of where I think we all hope to move towards. I was wondering if you or anyone on here has experience requesting and receiving the data, and can share how that experience was? Some major concerns that have been voiced throughout this discussion have been maintaining the privacy of patients/participants. I am curious if some of the steps taken in the YODA group process can be applied towards patient data sharing.

We took confidentiality and privacy very seriously in our design of the YODA/JNJ data sharing initiative. In addition to de-identifying the data itself, researchers must sign a data use agreement and agree to keep the data confidential and not try to re-identity participants. The data are on a secure website for analysis, and not available for download.

An initiative named "Synch 4 Science" is underway to allow patients to grant some researchers/institutions/studies get access to their Electronic Health Records. If successful, this will allow patients to play a more active role in sharing their own data.

Thank you for sharing, Lucila! This is new to me, and I am excited to see it grow. Dr. Zwerling has also mentioned another platform, "Health Wizz", where patients can sell their data. These are great initiatives. I wonder what will help propel them to success? Financial reward never hurts, but what can gain patient awareness and trust?

Why has there been such an emphasis on patient engagement? Why can't ONC and so many others realize that patients are generally not interested in becoming engaged to their medical care? The last thing most people want to think about, unless they have a chronic illness, is going to the doctor (or the "healthcare team"). We seem to have every institution trying to force feed people portals, devices, educational materials, and contacts that most do not want, i.e. human foie gras. If we have to make this much of an effort to sell it, we should realize that there is little demand. In my practice, I used an EMR for the last thirteen years, and not a single patient asked me for an electronic version of her record. I used to print out a copy of the visit which I automatically sent to the patient's primary care physician, and another for the patient. I had a waste paper basket immediately outside my office door where invariably there would be HIPAA violations by my patients.
Let's stop kidding ourselves!

Forty three states allow patients to opt out of an HIE or data sharing. Seven don't. To me, that says it all. It is much easier and cheaper to share data when it does not require specific consent. As long as a patient has information about where and how the data is to be used, and has a relatively simple way to opt out, then I think that is all we need.

Minnesota weather becomes dangerous for many patients 4 months per year. We help Veterans sign up for VA patient portal MyHealtheVet. "When it's -30 below, this is your best bet. Might save you a trip to the VA." I agree with Dr. Heyman's statements, most older veterans just want their doctors to have all the information. How the doctors receive it is far less important. St Cloud VA lead the nation in VHIE adoption the last couple years. Weather and the convince of HIE and patient portals played a role.

Hello Leonard, thank you for joining our conversation on Data Acquisition, Curation, and Use for a Continuously Learning Health System! Please feel free to contribute to our ongoing discussion. Under Resources, you will see the short paper we read that is driving the conversation. I will leave it to the experts if they have thoughts on your product. Thank you!

The text is not clear about what kinds of data are implicated and what "cloud based security" means.

I appreciate that you took interest to query for clarification. There is nothing like the product we designed.
Data: The data [collected] can be anything. BLYNK just monitors light being on or off (let's relate it to binary or Morse code) so if something [like a medical device] has an indicator light it can remotely {see below: "Security} track when that light is on or off. To continue this part of the discussion BLYNK can also determine the frequency of flashing plus if light flashing is in long short pulses {see below area added: Customization}. BLYNK can also determine the color (this is important because some LED change color to indicate different things on a medical device) of the light.
Security: This was a question for the experts. What platform are you most comfortable. Right now BLYNK transmits through the hospital wifi system. It is cloud based with high encryption. Is there a medical standard? Later development of Blynk will include cell transmission. This way the medical device can remotely (out of hospital range) alert the clinician. So I was asking, is there a standard for moving data 'securely' which you are more comfortable?
Customization: Stating that BLYNK is just a transmitter of information, BLYNK will have a rudimentary receiving software that can be installed on a caregiver's phone, laptop, ipad, computer, etc. This will advise the BLYNK that has a light illuminating and what it means (for example the light could be that a feeding pump is sending the right amount of food) and this would suggest one or more caregivers does not need to visit the patients room. Because data is binary, this could also be customized to forward to a program that stores (manipulates) information to automatically insert into a patient record that the patient was fed for X minutes. We are leaving this programming to personnel with appropriate skill and facility to know the need to store (may change by area acuity). It can also assist research. [ran out of room]

HIPAA currently grants individuals a right to access their health
information; however, it does not specify what form that access should take, so many health services organizations do not provide electronically. Under Meaningful Use certification language, there was criteria measure "to provide patients the ability to view online, download, and transmit their health information within 4 business days of the information being available to the eligible provider (EP)". The bar for how many patients varied by provider type and measure goal sought by EP, from at least 1 (yes, one patient) views, downloads or transmits to a third party his or her health information to > 50% of patients ... in 21st Century Cures Act that passed, there is more language on patient access electronically (that is technically law, although I am not sure enforcement teeth) in the certification criteria support-- ``(i) patient access to their electronic health information, including in a single longitudinal format that is easy to understand, secure, and may be updated automatically" ... how easy to understand and use may be interesting challenge ... This CHIME crosswalk chimecentral.org/wp-content/up... and Bill text (for wonks) has more detail on what in 21 Century Cures patient access, congress.gov/bill/114th-congre... .. last sidenote, the law states that patient should have access to their personal electronic health information for research at the option of the patient... long story short, patients given access by statute, but enforcement strength and timeline unclear...

Thank you for sharing this information, Kenyon. It is interesting to consider what was mentioned in the question posted by Dr. Zwerling- Dr. Chen asks how realistic it is to expect a patient to remember portal access. I wonder how realistic it is to assume that patients are aware of any of these rights to data access? What seems to be a repetitive barrier to me in all this is the limited awareness.

Indeed, I don't think vast majority of patients have any clue what their health data rights are, and defer to the staff of providers, many of whom on front line also probably don't know the nuances, but default to organizational policy that traditionally is overly restrictive and burdensome. On the patient side and consistent with an increasing focus on health IT usability, patients simply want an easy to use mechanism to get / use the information they need... I think we will see better and better interfaces, and also with the rise of the AI chatbot, perhaps more responsive patient information systems that let people interact with natural language-ish ways. Siri: send my bloodwork to Dr. Jane... Many people are accustomed to portals for their shopping, banking, and will become more accustomed to using for health data transactions, but I am concerned about the digital health equity divide this may promote without careful policy and systems development. I am curious, how the folks with practices reading this thread handle provisioning of patient data and barriers faced or lesson learned?

I’d like to add to Kenyon’s excellent response. The bottom line is that all patient rights relating to their information are rooted in laws. HIPAA, HITECH, and 21st Century Cures Act are just a few of the federal laws that grant patients certain rights to their health information. Other federal laws can further define patient rights to their information based on the specific context. For example, the VA Claims Confidentiality Statute and regulations (38 CFR §§ 1.500 et seq.) defines certain rights of veterans to their health records, and the regulations at 42 C.F.R. Part 2 define patient rights to access substance abuse treatment records.
These, of course, are all federal laws. States also have laws granting certain rights to patients. A 2016 study of state Health Information Exchange (HIE) laws (described here: ncbi.nlm.nih.gov/pubmed/291065...), found 10 states that address the issue of ownership or property rights (AK,AR,DE,IA,MS,NC,NH,PA,VT,WV) concerning health information in the context of HIE, and 18 states promote public transparency or patient engagement in HIE . Moreover, these laws are not always easy for laypersons to understand. Arkansas for example, grants a property right to individuals who provide data to the HIE, but grants a “nonexclusive license to retrieve and use that information” to other HIE users.
If a learning health system is to exist and function at the national level, then meaningful clarification of patient’s rights will remain elusive. There is simply too much variation and nuance in the dozens of federal privacy laws and thousands of state laws relating to health information.
healthit.gov/sites/default/fil...
ncbi.nlm.nih.gov/pubmed/288136...
A federal effort to define and clarify patient rights at a national level is one option to address the existing confusion and uncertainty surrounding patients’ rights to their health information

There is a critique of the dynamics of the model of care proposed by the National Health Agency (ANS) here in Brazil by Health Lawyers who see risks to patients due to the sharing of medical information.
In common, all programs do not have the patient as the primary purpose of the project benefit. In fact, these measures are aimed primarily at reducing the costs of health care providers.
In fact, the costs need to be reduced. The problem is the order of importance. Instead of being a program designed to benefit health care providers, with possible good reflexes for the patient, it should be a program designed to benefit the patient, with possible good reflexes for health care providers.
They warn that this model can be used by the operators as a risk control tool, by canceling the plans of elderly users.
On the other hand, the National Health Agency (ANS) focuses on improving the care provided to the elderly and pregnant women - two priority care lines for health systems in Brazil - and have the patient as the center of care.
The projects also aim to reorient the pay model of service providers towards a patient-centered model rather than the number of procedures performed. The search strategy lead, above all, the quality of care, without losing sight of the sector's sustainability.

Thank you everyone for your responses! I hope we continue to see improved interfaces, Kenyon. Cason, thank you for sharing what is legally connected to patient's rights. Based on this information, it is clear we need to see this more fully addressed and generalized across the map. What steps can we take to do this? Maria - this is a great point that the purpose should be focused on the patient. Too often in the data world, we will see models driven by reducing the frequency of "high cost events" rather than focusing on the patient outcome and trying to reduce the frequency of a "high impact state" in disease progression. While these often overlap, sometimes one sees that had we looked at the patient outcome instead of the costs, we could have intervened long before the high cost event occurred.

I believe patients should be able to obtain their data whenever they want, and should be able to see whatever they want. They should be able to control where it goes and who else sees it. They are not entitled to the original copy of either their electronic or paper medical record. We clinicians need to preserve their records as stewards, on the one hand, and for medico-legal protection on the other. We also need to stop trying to force them to take data they don't want in the interest of "patient engagement".

This is a good point, Dr. Heyman. We should be careful in our conversations not to conflate the two, as it changes the message and goal. Do you see any issues associated with the patient sharing their data?

Hi Stephen, great question! I think that having data as broadly available as possible would face many challenges, such as standardized acquisition and curation, and appropriate measures taken for privacy. What do you see are the main challenges to implementing such a network? Where would the first step need to be taken to move towards LCNs? And are there any reasons not to pursue it?

I wouldn't call it a "downside" per se, Antonia, but if a central authority wants to keep tight control of who can communicate whom, then it is not a reasonable use case for an LCN. On the other hand, if two or more clinicians or researchers (etc.) want to exchange information or have conversations among themselves without external constraints--just like making a phone call, texting, email, faxing, etc.--then an LCN makes sense. The trade-off is control and relative homogeneity versus the freedom and diversity LCNs provide. Neither is right in all situations. This IHMI forum has some LCN components: It enables people outside the APA, across disciplines, and with different roles to participate; it encourages collaboration among diverse groups of people; and it stimulates a level of creativity and insights less likely in tightly controlled, restrictive groups. Another example is an independent PCP and his/her community of referrals (possibly a PCMH) who freely collaborate in different care teams. One challenge implementing an LCN is that people must know about it, which can be accomplished in different way (e.g., advertising, word of mouth, sending a referral, using a provider directory).

I find that a lot of the issues we are discussing, and their potential solutions, all face the roadblock to some extent of awareness within the community where the solution would be implemented. For the case of LCNs, teams must become aware of it as an option; for patient portals, patients need to be informed of this access and given a feasible way to maintain access to all them (per Dr. Chen's point in a different post); Dr. Guyer mentions a solution in effect for the VA, but this also requires spreading the word. What can we do to overcome this issue not just within our data problem, but across healthcare? Is there an effective mode of communication for clinicians, hospitals, research groups, and patients?

I see two related issues here: 1. Spreading the word about an LCN and 2. Searching for participants in an LCN. Spreading the word is done by advertising the LCN, which can be through emails, posts, snail-mail, word of mouth, F2F, publications, etc. For development of an LCN care team beyond (or within) organizational boundaries, software can tracks where referrals are sent and automatically associates and tracks team members around a patient’s encounters. To join an LCN, directories could used. There are authoritative directories at the national, state, and organization levels with highly structured standardized schemas that are searchable by different categories. They may be in the cloud or on paper, free or purchased. In addition, many organization published their own directories that can be searched. For example, IHMI, Facebook, LinkedIn, provider organizations, and insurers have directories. These directories may have curated constraints by its owner to avoid trolling/spamming. In addition, one can use a general search engine (such as Google) to try to find an LCN.

The answer depends on what you consider "ethical requirements" that must first be met. For example, we know that even so-called "de-identified" data can be re-identified if you have some information about the targeted individual. Therefore, sharing "de-identified" data publicly may be legal, but can be considered unethical because it puts the privacy of individuals at risk.
Sharing clinical data always poses such dilemmas, as there is a trade-off between (a) increasing benefits to society in general from what is learnt from the data, and (b) the risk of privacy breaches for individuals.
Letting individuals decide what they consider thresholds for the risk of re-identification (which may depend on who is the recipient of the data) is a potential way to address this problem. This way institutions would not have to make unilateral decisions that may not align with the person whose data are going to be shared.

In terms of the strength of weak networks, the discussion relates to the understanding of who qualifies as a member of these networks and whether there is control about who gets the data and whether the recipient can further disseminate these data (somewhat similar to the concept of licenses for data).

Good points re-identification and individuals’ control of sharing their PHI. Regarding the strength of weak ties in loosely-coupled social networks (LCNs), the rules for membership should focus on the purpose/goals of the network. In addition, the LCN should welcome potential members who have relevant knowledge, experiences, and expertise that differ widely since this diversity is its core strength. And yes, if PHI is exchanged, there should be rules about sharing it based on ethical requirements.

Even if the people with access to the data have only the best intentions, the dissemination of this information to an increased audience means increased device access and modes of mistakes, and greater opportunity for people of ill intent to gain access. While I think a majority of people here are not security experts, I believe we have all seen instances of poor data security within our own experience. Has anyone seen good solutions to this?

I believe that no data security solution is foolproof. Nevertheless, it's wise to consider the risks in different scenarios. For example, it is more likely that malicious hackers will attack places where large amounts of sensitive are stored rather than attacking a person's standalone device with small amounts of data (although both are at risk). In either case, attacks will be more damaging if the data are not encrypted. As such, arguably the most secure way to manage data is to store them in a widely distributed manner rather than in a large centralized warehouse where hackers’ potential payoff is greater. Furthermore, the data should be encrypted while at rest and in transit, which can be done with digital certificates, encrypted drives (e.g., Bit locker), and "hardened" devices (i.e., ones that have been test for vulnerabilities and protected using secure policy settings, as well as software and hardware solutions). And since data must be decrypted for use, they should remain exposed for as little time as possible. To avoid spoofing, there should also be identity certificates. So, ideally, no e-mail should be opened unless the sender's identity is confirmed, no attachments should be opened unless they are encrypted, and no links should be clicked unless the sender's identity is validated and the linked web sites have been examined by up-to-date security software and are https.

Hello Stephen, these are great points to learn about. I wonder if anyone has seen excellent security systems throughout their work, and if individuals who had access to the data were compliant with the structure?

Travel any direction out of St Cloud MN and you'll pass a large solar power project. State, federal and utility company incentives gave solar power a positive business case.

Google decided going to the Moon should be cheaper. So, they created the X-Prize. "The $30M Google Lunar XPRIZE is an unprecedented competition to challenge and inspire engineers, entrepreneurs and innovators from around the world to develop low-cost methods of robotic space exploration." 5 teams are scheduled to reach the Moon in 2017.

Meaningful Use Stage 2 provides similar incentives for Electronic Health Records and interoperability.

Government Incentives. If big systems make money helping little systems connect, many of these other problems will work themselves out. This will happen faster if incentives are increased in proportion to national interoperability.

Like the cost of solar power or a trip to the Moon, a little push in the right direction and the technology becomes affordable for all.

Call me crazy but I think auctions are the way to go. Most of the current initiatives on value based care and quality measures are cumbersome to say the least(I recently attended a webinar on Meaningful Measures and the Q&A session was a "delight"). There has to be some form of monetary compulsion make the industry move, while ensuring that the right people are incentivized. Instead of data exchanges, if we create a data selling model that allows patients to benefit from selling their data with providers as an intermediary, it might actually renew interest from all parties. I am taking cue from online ads industry which has created a robust ecosystem around leveraging data. I believe their model deserves a look.

Thank you both for sharing your different approaches to solving this problem! Jonathon, where do you see that we could get the ball rolling? It seems to me we need to put this idea in front of a large enough entity (i.e. Google or the government) who can influence this. For the healthcare domain, what companies can we look to for the next challenge? Where do we share these ideas to see government incentives form? Bodhisattya, I think you are right that engagement and benefit from multiple sides would stimulate engagement. Previous posts have mentioned this model as well, mostly with start-up companies. Do you think this can succeed with a small group or does this model need to be absorbed by a healthcare data giant? Could this be implemented through insurance companies?

Google, Amazon, Microsoft, General Electric and many other big names are moving into healthcare. This is very similar to our current commercial space race. The government offered contracts for supply missions to the International Space Station and commercial companies made it happen.

Increasing Meaningful Use funding will do the trick. The government sets the bar and companies rise to the challenge.

We're past the "Too big to fail" point. Like computers and the internet, data sharing and interoperability is happening. How long it will take is the only variable.