Search Engines Using Your Personal Health Information: Creepy or Cutting Edge?

When using a search engine, should results be customized based on your personal health information (PHI)?  Should your search engine of choice take into account your previous history of medical searches, or even provide results tailored from data about your personal medical history?

Two companies — Aetna and Microsoft — have come up with 180 degree different answers.

In this post, I will:

• Describe how Aetna’s and Microsoft’s approaches differ
• Speculate on why their approaches make strategic sense for the respective companies
• Explore how technology and expectations about using PHI for search are likely to change over time

Aetna’s SmartSource and Microsoft HealthVault 

Aetna has adopted a more aggressive strategy of using your PHI to improve search results. Microsoft HealthVault is employing a much more cautious strategy and (for now) has pledged NOT to use personalized health information to target health search.

Aetna is a health insurance plan with an existing member base of 17 million members.  Aetna’s recent press release announced a partnership with Healthline introducing

Aetna SmartSource, the first-ever personalized search engine that mines information from Aetna’s vast data resources to deliver meaningful, personalized health and health benefits information to Aetna members. 

Based on existing medical claims data that Aetna already has about members, your search term will be matched against your own medical history to provide the most relevant results.

Microsoft HealthVault search is different.  From the Microsoft HealthVault Fact Sheet:

Microsoft does not link users’ personal information with HealthVault search queries.

Why the Difference?

These companies likely see vastly different upside rewards and downside risks.  They will view the pros and cons of using PHI to target searches very differently.

First, let’s look at it from Aetna’s vantage point. How might it see the pros of a more aggressive strategy using PHI to target searches?

• Aetna has a strong need to differentiate itself in the marketplace, both directly to members and to employer customers.  As I’ve written previously, the strategy employed by health plans over the past decade has mostly run its course. Health plans must become better managers of  medical risks and costs.  They do not want to be viewed simply as commoditized claims processors.
• Savings will accrue directly to Aetna (or its customers).  Aetna is assuming that targeted search results will provide you with better information for you to manage your health. If you’re healthier, your medical costs should be lower.
• Aetna already has substantial information about it’s members.

However, Aetna will need to confront cons to using PHI to target searches:

• Will people use and trust a personalized search delivered by a health plan? For example, see Jane Sarasohn-Kahn’s insightful commentary and data.
• Will Aetna become a lightning rod for criticisms by privacy advocates? 
• Will claims based data be sufficient to improve search capabilities? Claims data about you captured by health plans is not the same as clinical data about you.  Their search engine will have to make inferences from the claims data.

What about Microsoft HealthVault?  What are the pros of a more cautious strategy for Microsoft, i.e, NOT using PHI to target searches?

• Microsoft’s short-term goals must focus on building awareness and usage for HealthVault. Personalized search is just one feature, and it is one that brings possibilities for controversy.
• Microsoft HealthVault has already become a lightning rod for privacy advocates.  Microsoft is already challenged to gain the trust of the public.  
• Microsoft’s business model does not allow it to benefit directly from reduced health costs.  These savings will accrue to patients, health plans and employers.
• As a brand new offering, HealthVault does not have much information about its members…it will take time to acquire information.
• The policy can evolve and change over time.

What are the cons of a more cautious strategy for Microsoft?

• Foregone revenue.  Microsoft HealthVault’s current business model is based on advertising revenues.  By not matching your PHI to searches, Microsoft presumably will be able to deliver less precise information about you to advertisers.

My observation: Aetna’s more aggressive approach makes sense for Aetna, while Microsoft’s more cautious approach makes sense for Microsoft.

Another Scenario — How Would You React to Your Doctor Using Your PHI to Personalize Care?

My guess is that most of us would be far more comfortable with this than with the prospect of Aetna or Microsoft using our PHI.

Think about the implicit understandings with which we develop trusted relationships with our physicians:

“Doctor, let me tell you about my medical history and symptoms. I know that you have a lot of medical training and experience.  Please use your knowledge to provide personalized medical care for my situation. If you don’t have the expertise to deal with my situation — that’s OK — but please refer me to one of your colleagues that does.”

The use of searches and clinical decision support systems using PHI by physicians strikes me as a logical extension of the existing relationship between doctor and patient:

“Doctor, while I’ve tried to explain my past history and current symptoms to you, I might not have told you everything that you need to know about me.  Thus, you certainly SHOULD use information about me that is already stored in your EHR or my PHR to provide context about my situation.

I also understand that the rate of medical knowledge is expanding rapidly. While I  appreciate your knowledge and experience, I know that you don’t know everything and can’t keep everything in your head. Thus, please do use computer based decision support tools to apply a wider base of knowledge to my situation. Not only do I want to allow you to do so, but I EXPECT you to do so.”

Implications? What are some of the possible lessons here?

First, health search technology is very immature.  Search algorithms will improve and over time search engines will have access to richer data, including clinical data.

Second, let’s recognize that this issue is driven by evolving cultural norms as well as technology.  For example, while there were initial concerns about sharing credit card information over the Internet, this concern has moderated over time.

Third, recognize that search based on PHI is highly context dependent.  People will react differently based on who provides the search tools and who uses them on your behalf.

Finally, recognize that there is a lot of middle ground here. The Aetna/Microsoft examples represent the two extremes — the black/white scenarios of using or not using PHI to target search.  There is also a lot of gray territory — for example:

• Searches using your PHI could be conditioned on your permission (in general)
• Search using your PHI could be customized to use or not use specific information:.  For example, it’s OK to search based on PHI describing my headaches, but don’t search based on PHI describing my psychiatric conditions.
• Searches using your PHI could be customized for individual searches.  For example, it’s OK to use my PHI when I search on the term “migraine”, but keep my search on “HIV” private.

Of course there are complicating factors here. Will people trust that their withholding permission is respected? …and is it good clinical practice to include or exclude specific PHI?

Creepy or cutting edge? 

Personally, I’m excited about the opportunities to improved self care and medical care. I think Internet guru Esther Dyson summed up well where this trend is headed:

Doing a search without your health record attached to it 10 years from now will be called crazy.