safety and privacy. Ross Anderson. Cambridge University. London, April 24 2013. Synopsis. Health privacy is everywhere under threat with . tussles in one country after . another. Everyone from drug companies to insurers want access to masses of personal data. ID: 308444
DownloadNote - The PPT/PDF document "Online patient records –" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Online patient records –safety and privacy
Ross AndersonCambridge University
London, April 24 2013Slide2
Health privacy is everywhere under threat with tussles in one country after anotherEveryone from drug companies to insurers want access to masses of personal dataNonconsensual access to health data is currently against European lawBut: the medical lobby wants to carve a huge loophole in the Data Protection RegulationIn Britain, the NHS Information Centre wants to hoover everything up and build an empire
London, April 24 2013Slide3
We’ve been here before!
Big row in 1998 when a startup (DeCODE) offered the health service free IT systems in return for access to records for researchFunding was from Swiss drug company RocheRecords to be ‘de-identified’ by encrypting the social security number, but would be linked to genetic, family dataIcelandic Medical Association got 11% of citizens to opt outEventually the supreme court ruled the system should be opt-in, and the scheme collapsed
London, April 24 2013Slide4
We’ve been here before (2)
European law based on s8 ECHR right to privacy, clarified in the I v Finland caseMs I was a nurse in Helsinki, and was HIV+Her hospital’s systems let all clinicians see all patients’ recordsSo her colleagues noticed her status – and hounded her out of her jobFinnish courts wouldn’t give her compensation but Strasbourg overruled themNow: we have the right to restrict our personal health information to the clinicians caring for us
London, April 24 2013Slide5
We’ve been here before (3)
Tony Blair ordered a “National Programme for IT” in the NHS in 2002Idea: replace all IT systems with standard ones, giving “a single electronic health record” with access for everyone with a “need to know”This became the biggest public-sector IT disaster in British historyBillions wasted, suppliers dropped out, huge lawsuits, and the flagship software didn’t work
London, April 24 2013Slide6
Assorted Things Going Wrong
Some stuff did get fielded though – over half of family doctors’ systems are now ‘hosted’ Some hospital systems that let receptionists read all patients’ psychiatric casenotesThere’s the PDS “address book” which gets abused – lawsuit pending from a woman who was traced by her ex-husband who broke her arm (No-one knew they could opt her out, or how)An emergency care record system in Scotland let curious people browse celebrities’ records
London, April 24 2013Slide7
We’ve had big tussles over ‘shared care’E.g.: giving social workers access to GP records in some areas has made young mums reluctant to discuss post-natal depressionBig win: after the 2010 election, we killed the ‘childrens’ databases’ designed to share data between health, school, probation and social work (‘Database State’, Munro review)The NHS Information Centre now wants to revive the idea, but under its control
London, April 24 2013Slide8
London, April 24 2013
2,231 adults asked October 2006 on central records database with no opt
strong support 12%
tend to support 15%
tend to oppose 17%
Several surveys since say the same: most don’t want wide sharing, or research use without consent
And there’s the Catholic Bishops’ ConferenceSlide9
Cameron policy announced last January: make ‘anonymised’ data available to researchers, both academic and commercial, but with opt-outWe’ve already had a laptop stolen in London with 8.63m people’s anonymised records on itIn September 2012, CPRD went live – a gateway for making anonymised data available from both primary and secondary careFrom July, GPES hoovering stuff up to the ICSo: how easy is it to anonymise health records?
London, April 24 2013Slide10
London, April 24 2013Slide11
Also known as ‘statistical security’ or ‘statistical disclosure control’Started about 1980 with US census Before then only totals & samples had been published, e.g. population and income per ward, plus one record out of 1000 with identifiers removed manuallyMove to online database system changed the game Dorothy Denning bet her boss at the US census that she could work out his salary – and won!
London, April 24 2013Slide12
Inference Control (2)
Query set size controls are very common. E.g. in New Zealand a medical-records query must be answered from at least six recordsProblem: tracker attacks. Find a set of queries that reveal the target. E.g for our female prof’s salary‘Average salary professors’‘Average salary male professors’Or even these figures for all ‘non-professors’!On reasonable assumptions, trackers exist for almost all sensitive statistics
London, April 24 2013Slide13
Inference Control (3)
Contextual knowledge is really hard to deal with! For example in the key UK law case, Source Informatics (sanitised prescribing data):
Week 1Week 2Week 3Week 4Doctor 1 17 21 15 19Doctor 2 20 14 3 25Doctor 3 18 17 26 17
London, April 24 2013Slide14
Inference Control (4)
Perturbation – add random noise (e.g. to mask small values)Trimming – to remove outliers (the one HIV positive patient in Chichester in 1995)We can also use different scales: practice figures for coronary artery disease, national figures for liver transplantsRandom sampling – answer each query with respect to a subset of records, maybe chosen by hashing the query with a secret key
London, April 24 2013Slide15
Inference Control (5)
Modern theory: differential privacy (pessimistic)Practical problem in medical databases: context‘Show me all 42-yo women with 9-yo daughters where both have psoriasis’If you link episodes into longitudinal records, most patients can be re-identifiedAdd demographic, family data: worse stillActive attacks: worse still (Iceland example)Social-network stuff: worse stillPaul Ohm’s paper has alerted lawyers at last!
London, April 24 2013Slide16
Next problem – rogue officials
Cameron promised our records would be anonymised, and we’d have an opt outThe opt-out is like Facebook: the defaults are wrong, the privacy mechanisms are obscure, and they get changed whenever a lot of people learn to use themShould not Kelsey follow Cameron’s stated policy of allowing an opt-out?
London, April 24 2013Slide17
London, April 24 2013Slide18
The Coming Policy Tussle
UK Data Protection Act 1998 failed to incorporate recital 26 of the DirectiveDefinition of ‘personal data’ was too narrowCreated loophole for UK firms, government departments to use ‘pseudo anonymised’ dataWe hoped the new DP Regulation would fix thisBut medical researchers, drug companies put down amendments to sections 81, 83 in draft DP regulationWill Europe move to the UK free-for-all?
London, April 24 2013Slide19
Problems building in Europe
Data Protection Regulation currently making its way through the EuroparlAttempt to exempt medical data (art 81, 83)You’ll be deemed to consent to secondary use and forbidden to opt out retrospectively, or even claim that consent was coercedAmendments came from NHS confederation, COCIR, Wellcome TrustIntroduced by the Baroness Sarah Ludford MEP (Vice-President of LGBT+ Lib Dems)
London, April 24 2013Slide20
Think safety and privacy togetherScale matters! A national system with 50,000,000 records is too big a targetIt will also be cumbersome, fragile and unsafePrivacy failure will have real costs in safety and access especially for those most at riskOfficials are ignoring Cameron’s promisesEventually a scandal will lead to public revolt
London, April 24 2013Slide21Slide22Slide23