1 / 36

Understanding the Challenges with Medical Data Segmentation

Understanding the Challenges with Medical Data Segmentation. Ellick M. Chan, Peifung E. Lam, and John C. Mitchell Stanford University. Health Information Exchange (HIE). Health Information Exchange Cloud. Federal HIPAA HITECH State laws on Mental Health Substance Abuse STDs

nodin
Download Presentation

Understanding the Challenges with Medical Data Segmentation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Understanding the Challenges with Medical Data Segmentation Ellick M. Chan, Peifung E. Lam, and John C. Mitchell Stanford University

  2. Health Information Exchange (HIE) Health Information Exchange Cloud • Federal • HIPAA • HITECH • State laws on • Mental Health • Substance Abuse • STDs • Genetic testing • Organizational

  3. Compliance approaches compliantWithALaw( A ) Data segmentation AND permittedBySomeClause( A ) notForbiddenBy AnyClause( A ) Health Record AND • Medications • Previous diagnoses • Labs permittedBy Clause1( A ) notForbidden ByClauseM( A ) notForbidden ByClause1( A ) … AND clause1 Applicable( A ) permittedBySome RefOfClause1( A ) clauseM NotApplicable( A ) meetReq Clause1( A ) Sensitive conditions permittedByClause Ref_I,J( A ) According to research by the California HealthCare Foundation, 15 percent of patients who know their information will be shared would hide information from their doctor, and another 33 percent would consider hiding information[1]. IHI 2012 Automated Policy HIPAA Law § 164.502 Uses and disclosures of protected health information: general rules.(a) Standard. A covered entity may not use or disclose protected health information, except as permitted or required by this subpart or by subpart C of part 160 of this subchapter. (1) Permitted uses and disclosures. A covered entity is permitted to use or disclose protected health information as follows: (i) To the individual; (ii) For treatment, payment, or health care operations, as permitted by and in compliance with §164.506; (iii) Incident to a use or disclosure otherwise permitted or required by this subpart, provided that the covered entity has complied with the applicable requirements of §164.502(b), §164.514(d), and §164.530(c) with respect to such otherwise permitted >| %%Standard rules for "uses and disclosures" permitted_by_164_502_a(A) :- is_from_coveredEntity(A), is_phi(A), (permitted_by_160_C(A); permitted_by_164_502_a_1(A); required_by_164_502_a_2(A)). permitted_by_164_502_a_1(A):- permitted_by_164_502_a_1_i(A); permitted_by_164_502_a_1_ii(A); permitted_by_164_502_a_1_iii(A); permitted_by_164_502_a_1_iv(A); permitted_by_164_502_a_1_v(A); permitted_by_164_502_a_1_vi(A). …010110...

  4. Example Letter I hope you and your partner had a great weekend in Provincetown and that the thrush has improved with the mouthwash sample I gave you. Adapted from J. Halamka, 2012 Medications 1.  Tylenol 2.  Sudafed 3.  AZT 4.  Bactrim Problem List 1.  Headache 2.  Sinus Infection 3.  HIV positive 4.  UTI

  5. Example Letter I hope you and your partner had a great weekend in Provincetown and that the thrush has improved with the mouthwash sample I gave you. Hide HIV/AIDS ICD-9 code 042 Adapted from J. Halamka, 2012 Medications 1.  Tylenol 2.  Sudafed 3.  AZT 4.  Bactrim Problem List 1.  Headache 2.  Sinus Infection 3.  HIV positive 4.  UTI

  6. Example Letter I hope you and your partner had a great weekend in Provincetown and that the thrush has improved with the mouthwash sample I gave you. +side effects Zidovudine (INN) or azidothymidine (AZT) is a type of antiretroviral drug used for the treatment of HIV/AIDS. Side effects: anemia, neutropenia, hepatotoxicity, cardiomyopathy, and myopathy Adapted from J. Halamka, 2012 Medications 1.  Tylenol 2.  Sudafed 3.  AZT 4.  Bactrim Problem List 1.  Headache 2.  Sinus Infection 3.  HIV positive 4.  UTI

  7. Example Letter I hope you and your partner had a great weekend in Provincetown and that the thrush has improved with the mouthwash sample I gave you. +side effects Trimethoprim/sulfamethoxazole or co-trimoxazole (abbreviated SXT, TMP-SMX, TMP-SMZ or TMP-sulfa) is a sulfonamideantibiotic combination of trimethoprim and sulfamethoxazole, in the ratio of 1 to 5, used in the treatment of a variety of bacterial infections. ? Adapted from J. Halamka, 2012 Medications 1.  Tylenol 2.  Sudafed 3.  AZT 4.  Bactrim Problem List 1.  Headache 2.  Sinus Infection 3.  HIV positive 4.  UTI

  8. Example Letter I hope you and your partner had a great weekend in Provincetown and that the thrush has improved with the mouthwash sample I gave you. +side effects Prophylaxis (preventative med) for immunocompromised patient? Patient has urinary tract infection (UTI), plausibly deniable case. Adapted from J. Halamka, 2012 Medications 1.  Tylenol 2.  Sudafed 3.  AZT 4.  Bactrim Problem List 1.  Headache 2.  Sinus Infection 3.  HIV positive 4.  UTI

  9. Example Letter I hope you and your partner had a great weekend in Provincetown and that the thrush has improved with the mouthwash sample I gave you. +side effects Highest rate of same-sec couples in Provincetown, MA. Karen ChristelKrahulik, Provincetown: From Pilgrim Landing to Gay Resort, NYU Press, 2007, p. 51. Adapted from J. Halamka, 2012 Medications 1.  Tylenol 2.  Sudafed 3.  AZT 4.  Bactrim Problem List 1.  Headache 2.  Sinus Infection 3.  HIV positive 4.  UTI

  10. Example Letter I hope you and your partner had a great weekend in Provincetown and that the thrush has improved with the mouthwash sample I gave you. +side effects Candidiasis (thrush) - Candidiasis or thrush is a fungal infection (mycosis). Commonly causes mouth yeast infections, which manifest as white patches in the mouth. 15% of immuno-compromised patients may develop this. Adapted from J. Halamka, 2012 Medications 1.  Tylenol 2.  Sudafed 3.  AZT 4.  Bactrim Problem List 1.  Headache 2.  Sinus Infection 3.  HIV positive 4.  UTI

  11. Example Letter I hope you and your partner had a great weekend in Provincetown and that the thrush has improved with the mouthwash sample I gave you. +side effects Headaches & HIV: 24/535 patients – 4.5% CDC NHDS 2010 dataset. 115,000 patients. Mononucleosis-like symptoms Adapted from J. Halamka, 2012 Medications 1.  Tylenol 2.  Sudafed 3.  AZT 4.  Bactrim Problem List 1.  Headache 2.  Sinus Infection 3.  HIV positive 4.  UTI

  12. Effects

  13. Threat Model • Attacker has direct access to redacted health record, medical literature • Attacker does not have the computational capability to circumvent security mechanisms that protect the primary sensitive codes

  14. Treatments

  15. Treatments

  16. Formal model d2 d3 d1 D C M m1 m2 m3 m4 • Reggia’s set cover model • Plausibility – set cover • Likelihood – Occam’s razor and fitness

  17. Formal model d4 d2 d3 d1 D C M m1 m2 m3 m4 m5 • Reggia’s set cover model • Plausibility – set cover • Likelihood – Occam’s razor and fitness

  18. Explanation of manifestations • Best explanation E of manifestations: • Covers all observed manifestations M+ • Is the simplest (parsimonious) definition • Heuristics for “best cover” • Minimality- |E| is minimal • Criticism: minimal cardinality covers can be too restrictive • Occam’s razor vs Hickam’s dictum • Irredundancy – removing any disorder results in a non-cover of M+ • Relevancy – Every din D must be causally associated with some min M+

  19. Medical concepts • Diseases • Manifestations • Kaposi’s Sarcoma • Rotavirus • Cervical Cancer • AIDS • Alcohol Abuse • Delusions • Schizophrenia • Psychosis Sensitive Concepts • Stroke • Memory Loss

  20. Manifestations Delusion Memory loss Alcoholism Stroke Hallucination HIV+ Psychosis Alzheimer’s Disease Diseases Source: PubMed, NIH.gov

  21. Predicate-Reducer definition A – Medical algorithm/Action π – Policy determines sensitive code s M – Medical record Predicate P(M, π) – Determines if s M Reducer R(M, π) – Removes s from M Ideal reducer Rhinovirus Mental health STDs Substance abuse X-ray [Free text] Rhinovirus Mental health STDs Substance abuse X-ray [Free text] R

  22. Inference approach • Input: Reduce(Diseases U Manifestations U Treatments) • Output: Inferred Diseases • 1. For each input, evoke hypotheses • 2. Evaluate hypotheses • 3. Rank hypotheses according to fitness • Hypothesis fitness • Competing hypotheses, e.g. d1 or d2

  23. Algorithm overview EHR: 042 (HIV), 112.0 (Thrush), 136.3 (Pneumocystosis) R(EHR) Extract Concepts Salient Concepts Retrieve Documents Docs Extract and rank Hypo Hypotheses

  24. Algorithm overview EHR: 042 (HIV), 112.0 (Thrush), 136.3 (Pneumocystosis) R(EHR) Extract Concepts EHR: 112.0 (Thrush), 136.3 (Pneumocystosis) Salient Concepts Retrieve Documents Docs Extract and rank Hypo Hypotheses

  25. Algorithm overview EHR: 042 (HIV), 112.0 (Thrush), 136.3 (Pneumocystosis) R(EHR) Extract Concepts EHR: 112.0 (Thrush), 136.3 (Pneumocystosis) Salient Concepts Retrieve Documents EHR1, EHR2, …, EHR n Docs Extract and rank Hypo Hypotheses

  26. Algorithm overview EHR: 042 (HIV), 112.0 (Thrush), 136.3 (Pneumocystosis) R(EHR) Extract Concepts EHR: 112.0 (Thrush), 136.3 (Pneumocystosis) Salient Concepts Retrieve Documents EHR1, EHR2, …, EHR n Docs Extract and rank Hypo Hypotheses

  27. Algorithm overview

  28. Concept Support Index d1 {d1, d2}

  29. Hypothesis Fitness Index {d1, d2}

  30. Results

  31. Empirical results Title

  32. Possible defenses • Deniability through relative strengths of hypotheses • Hide non-sensitive EHR as well • Enhance competing hypothesis, e.g. Citalopram or immunosuppression • Association rule hiding d4 d2 d3 d1 D C M m1 m2 m3 m4 m5

  33. Future work • Epidemiology CA Hidalgo, N Blumm, A-L Barabasi, NA ChristakisPLoS Computational Biology (2009). Towards Precision Medicine. National Research Council, 2011. Title

  34. A message from our sponsors… • We thank: • Carl Gunter and Mike Berry - predicate-reducer model • James Reggia - formalization of the hypothetico-deductive model • Brad Malin - helpful resources • Ivan Handler - health information exchange level • FisayoOsitelu - medical insight.

  35. Questions?

  36. Ask your physician!

More Related