Schrödinger's Better Patients FATE
UNCERTAINTY
MEDICINE
COMPLIANCE
Relevance to VistA, Population Analytics, and the OSEHRA Roadmap REPORT ON THE Q-UEL PROJECT with special reference to EHR structure, Automated Systematic Review, and Clinical and Public Health Decision Support Professor Barry Robson PhD DSc 1
Section 1 FOREWORD. QUANTUM-UNIVERSAL EXCHANGE LANGUAGE Preamble on Declarations, relation to VistA vision, older origins in Big Data mining., and synopsis and overview.
Modern form of Q-UEL - origins in the PCAST 2010 Report Mathematical Basis in Dirac Notation, algebra, and in semantic theory
2
Preamble 1. Declarations Consortium. The Q-UEL language grew from research tools developed for the QEXL Consortium for a Future THINKING WEB, an informal consortium of workers from ingine.com, LEXIKL.org, diracfoundation.org/com, quantal semantics group, University of North Carolina Chapel Hill, University of Michigan, University of Wisconsin-Stout, St. Matthews University School of Medicine, and recently the University of York (UK).
Technology. The Q-UEL language and related technologies are based largely on ideas and mathematics due to physicist and Nobel Laureate Paul A. M Dirac, combined with linguistic and semantic research. The Dirac Foundation, Provenance, and Charter. This mathematics first began to be adapted by The Dirac Foundation, Oxfordshire, UK (Company Limited By Guarantee) founded in 1994 with the written support of Ms. Margit Dirac, widow of Professor Paul Dirac, with the charter of promoting Professor Dirac’s ideas and Dirac family interests in human and animal medicine. Publication. The preference is to promote these largely by publication in reputable peer reviewed scientific journals, rather than by media that are not peer reviewed. Commercialization. Whereas in the interest of wide adoption the policy is of dissemination by open source, and currently more particularly by publication in sufficient detail so that ideas regarding interoperability can be readily and freely implemented, commercial work on specific applications and architectures to promote 3 the adoption of the ideas in industry is not discouraged.
Preamble 2. The True Interoperability Vision
VistA VistA 4 Vision • An EHR that is interoperable with the DoD EHR, and which has achieved ONC Compliance. • A modernized EHR to enhance patient-centered, team-based, and evidencebased quality- driven care by giving health care providers a complete picture of a patient’s care and treatment history • Sharable Clinical Decision Support (CDS) to promote best clinical practices tailored to the patient's clinical condition and health-related goals.
THINKING WEB for MEDICINE (PROBABILISTIC SEMANTIC WEB)
Q-UEL
MINING, PUBLIC HEALTH ANALYSIS, CLINICAL DECISION SUPPORT (BioIngine prototype)
Continuity of Care Overseas
HL7 V2 pipe-hat HL7 CDA, CCD Other EHRs, Entity Attribute Value docs, Intermediate Data Systems 4
Preamble 3. The Dragon on the Gold. • Even Before Q-UEL, its Zeta-Theory-of-Data was Being Developed for Medical “Big Data” and Natural Language Text Mining, and Inference from it. •
Origins in widely used GOR method in bioinformatics. B. Robson (1974) , "Analysis of the Code Relating Sequence to Conformation in Globular Proteins: Theory and Application of Expected Information", Biochem. J. 141, 853-867
•
“A particular dragon protects the gold therein: the dragon is the curse of dimensionality, and its formidable fire weapon, which is burning researchers, is the combinatorial explosion.” B. Robson, “The Dragon on the Gold: Myths and Realities for Data Mining in Biomedicine and Biotechnology Using Digital and Molecular Libraries” J. Proteome Res., 2004, 3 (6), pp 1113–1119) Continued in Robson, B. and Vaithiligam, A. (2010) “Drug Gold and Data Dragons: Myths and Realities of Data Mining in the Pharmaceutical Industry” pp25-85 in Pharmaceutical Data Mining, Ed Balakin, K. V. , John Wiley & Sons
•
The zeta theory/number theory techniques were used in data mining 0.67m to 6.7m records in… Mullins, I. M., Siadaty, M. S., Lyman, J., Scully, K., Garrett, C. T., Miller, W. G., Robson, B., Apte, C., Weiss, S., Rigoutsos, Platt, D., Cohen, S., Knaus, W. A. (2006) , “Data mining and clinical data repositories: Insights from a 667,000 patient data set” Computers in Biology and Medicine, 36(12):1351-77 Robson, B., Li, J., Dettinger, R., Peters, A., and Boyer, S.K. (2011), “Drug discovery using very large numbers of patents. General strategy with extensive use of match and edit operations..” Journal of Computer-Aided Molecular Design 25(5): 427-441 (2011) – included mining subsets of chemical formulae as graph structures.
B. Robson
5
Preamble 4. Synopsis and Key Points 1. 2. 3.
4. 5. 6.
7.
8.
The QEXL Consortium researches a THINKING Semantic Web for mining, linking, managing, and inferencing from, all data and knowledge. Q-UEL is being designed as an alternative to IBM’s Watson for medicine by distributing computation across the Internet rather than on centralized high performance machines. Q-UEL started with data miners but in modern form was in response to the 2010 Report of the President’s Council of Advisors in Science and Technology, requesting an “XMLlike” Universal Exchange Language for healthcare. Q-UEL has a formal unifying basis: it closely follows the Dirac notation and probabilistic algebra that has been a standard in physics since the 1940s. Q-UEL positions itself as a Universal Second Language as Pivot/Hub, meaning 2N interconversions between N standards and implementations, not N2-N!!! Q-UEL is capable of encoding, storing and communicating electronic health records, biomedical data and summary statistics from archive and population studies, all as statements about observations, information, and knowledge; it also has metastatements (rules) about manipulating statements. Q-UEL tags can autosurf and spawn on the web to extract knowledge for automated reasoning on a “Thinking Web”. Here a recent development has been application to automated Systematic Reviews and meta-analysis. Q-UEL is described in some 12 publications in major “hardcopy” peer-reviewed journals, 6 IEEE etc., and there are some 100 related and/or older publications.
The Main Beginning: The PCAST Report 2010
REPORT TO THE PRESIDENT REALIZING THE FULL POTENTIAL OF HEALTH INFORMATION TECHNOLOGY TO IMPROVE HEALTHCARE FOR AMERICANS: THE PATH FORWARD
President’s Council of Advisors on Science and Technology (December 2010) “In other sectors, universal exchange standards have resulted in new products that knit together fragmented systems into a unified infrastructure.” “The resulting ‘ network effect’ then increases the value of the infrastructure for all, and spurs rapid adoption.” “By contrast, health IT has not made this transition.” They call for an XML-like … UNIVERSAL EXCHANGE LANGAGE
The Tower of Babel
UEL
7
What Should a Universal Exchange Language Look Like? “….a good notation can be of great value in helping the development of a theory, by making it easy to write down those quantities or combinations of quantities that are important, and difficult or impossible to write down those that are unimportant….”. P. A. M. Dirac (1939) “The methods of theoretical physics should be applicable to all those branches of thought in which the essential features are expressible with numbers.” Paul A. M. Dirac, Nobel Prize Banquet Speech (1933)
< subject expression | relationship expression | object expression> • The above is a bi-directed edge in a general graph of probabilistic knowledge representation, a probability dual quantified by a special form of imaginary algebra. It is the element in a general graph or net of knowledge representation. • Dirac’s bracket or braket notation often looks conveniently like XML and the kind of XML extension we want for probabilistic semantics. The above is just one type of Dirac “tag” . We also have Dirac’s braket , and bra vector , ketbra matrix |…> . 8
But Why Keep the Quantum Q in Q-UEL? Quantum mechanics models the world not as a static “Directed Acyclic Graph” like the “Bayes Net” popular in inference , but an evolving general graph or diffuse field of interactions and cyclic processes, and a combinatorial explosion of possibilities!
Re-enter the Dragon Refers to the large number of considerations and operations required when relationships, functions, or models are deduced from, fitted to, or optimized for, items associated with N parameters (graph dimensions, spread-sheet columns, metadata, or types of data item in a collection), when N is large. When quantitative data is to be fitted by a statistic or model with an error ε or n = 1/ ε distinguished data values, then nN evaluations are required (That is 10100, with 10% error or 10 distinguished values, when N=100!! ).
Quantum Chaos as Statistics: Zeta Theory Combined with the Quantum Theory of Paul Dirac Allows for Integrating several approaches resulting into Hyperbolic Complex Algebraic Model
Algebra • Mathematical meaning, relationships
Semantics • Human Meaning, relationships
Vector Space • Probability of States
Zeta Theory • Combinatorial Explosion, Sparse Data, Combining Belief and Objective Data
Bayesian Net and HDN • Knowledge Representation
9
Data Science Driven Large Data Sets Transformation
•Patient Longitudinal Records •Multivariate / High Dimensional •Handle complexity in establishing conditional and joint probabilities •Handle uncertainty about the state of the world •Is probabilistic as opposed to deterministic
Millions of Patient Records
Quantum Transformation* •Q-UEL – Dirac Notation •Hyperbolic Dirac Net – Dual Probabilities •Bidirectional Bayes •Riemann Zeta Summation •Unsupervised and Supervised Machine Learning driven Data Mining
•Public Health Statistics •Epidemiology •Public Health Surveillance •Evidence Based Medicine •Continuity of Care •Preventive •Personalization
Knowledge By Inference
VistA/ BigData Platform / Health Information Exchange / IoT * Literally, as a Lorentz rotation i → h of Schrödinger's wave mechanics. Exponentials of i-complex expressions relate to periodic functions, but with h-complex they relate to normal distributions. 10
Q-UEL’s Hyperbolic-Complex Algebra in a Nutshell • hh = +1, Dirac’s , time etc., is the hyperbolic imaginary number (relates to physicists’ 5 with same property) • Q-UEL is a Lorentz Rotation i → h of Schrödinger's Wave Mechanics; unlike with i, exponentials of h-complex expressions relate to normal distributions, not waves (deep relation to Wick rotation, i-complex ground state of harmonic oscillator, and to collapse of wave function).
• ≡ ≡ ≡ = * = ½ [P(A|B)+P(B|A)] + h½ [P(A|B)-P(B|A)] (Hermitian commutator notation) = P (A|B) + *P(B|A) (iota notation) = [ P(A) + *P(B) ] eI(A; B) = [ P(A) + *P(B) ] K(A; B) (association constant notation)
•
= ½ (1+h), * = ½ (1-h) are physicists’ spinor projectors, easier to manipulate than h. Note = , * * = *, * = * = 0, + * = 1.
• = [< 1|B>, < 2|B>, < 3|C>, …. ]T (ket vector) • i< i|B> = = (better than assuming A B independence) • Universal reference quantum
is usually gender with age distribution11
Q-UEL’s Probabilistic Semantics All statements about the world have a degree of truth (default 1) • Dirac braket notation maps to natural S-V-O languages.
– The relation expression is in our system (as typically in QM) is always a real or complex Hermitian operator/matrix expression (if real, it is “trivially Hermitian”). – The bra-operator-ket responds to complex conjugation *, changing the sign of the imaginary part.
= = * (dogs chase cats) = (cats are chased by dogs) = (cats chase dogs)*
• Example extract from medical record allowed by fine-grained consent language (Confidence Interval implies probability). •
• Statistical summary statement from data-mining such. •
Attributes on Q-UEL Tags Use Attribute Metadata Language (AML) and themselves handle ontology, links to MUMPS (M) •
A Q-UEL XML-like attribute is minimally a nominal-categorical value , a word such as male … but is typically a metadata:=orthodata construct such as ‘Systolic BP(mmHg)’:= ‘143(+/-13CI)(consented)(Sat Jan 19 16:36:54 2013)’. Attributes are separated by whitespace, which in absence of any other operator implies logical AND. If there is whitespace within metadata or orthodata, these must be placed in single quotes ‘…’
•
Generally, an attribute is optionally a branched structure implying ontology, e.g. Cardiovascular:=((pulse:= 75, units:=/sec, consent:=‘consented to nearest 5 and month’, time:=gmt:=‘Jan 2013’), ‘blood pressure’:=(systolic:=140, diastolic:=80, units:=mmHg, consent:= ‘consented to nearest 5 and month’, time:=gmt:=‘Jan 2013’)))
•
Consistent with a basic rule that the text right of the rightmost:= in a branch can be encrypted, using = instead of := can be used to encrypt metadata. metadata:=‘systolic BP(mmHg)’=140 (GMT Sat Jan 19 16:36:54 2013)’ in private data becomes metadata:=Tn43yBoG681LtUvWzNpS69(encrypted).
• Q-UEL AML maps to MUMPS (M) traditionally used for VistA… set ^home(^captain(^ship("1ST FLEET", "BOSTON","FLAG")))="PORTSMOUTH“
13
Attributes on Q-UEL Tags often Hold Distributions and HbA1c(%):='6.60+/-(3.26P95, 1.66SD, 1.76SDs, .14SE, .28CI, .47Skew, -0.74Kurtosis)' :='(( 'probabilities':=comment:='Breakdown distribution (probabilities all summing to 1)':= (M:='P(male,AVERAGE FOR HbA1c(%)=6.72)=0.31':=( 'P(male,age:=30-39,average HbA1c(%):=7)=0.0652', This gender and age group distribution is 'P(male,age:=40-49,average HbA1c(%):=7)=0.0435', (at least in healthcare) also Q-UEL’s most 'P(male,age:=50-59,average HbA1c(%):=7)=0.1159', common choice as the counterpart of the 'P(male,age:=60-69,average HbA1c(%):=6)=0.0580', 'P(male,age:=70-79,average HbA1c(%):=5)=0.0145', of quantum mechanics, the universal 'P(male,age:=80-89,average HbA1c(%):=6)=0.0072', wave function, universal quantum state, 'P(male,age:=90-99,average HbA1c(%):=8)=0.0072',), F:=('P(female,AVERAGE FOR HbA1c(%)=6.55)=0.69':=( or universal reference state! If you don’t 'P(female,age:=10-19,average HbA1c(%):=6)=0.0072', have , a good estimate is…. 'P(female,age:=20-29,average HbA1c(%):=5)=0.0290', 'P(female,age:=30-39,average HbA1c(%):=5)=0.0362', 'P(female,age:=40-49,average HbA1c(%):=7)=0.1812', i< i|B> = 'P(female,age:=50-59,average HbA1c(%):=6)=0.2681', 'P(female,age:=60-69,average HbA1c(%):=7)=0.1014', 'P(female,age:=70-79,average HbA1c(%):=7)=0.0580', It is better than assuming independence. 'P(female,age:=80-89,average HbA1c(%):=7)=0.0072',)), 'distributed by age as':= (M:=(30-39:=7,40-49:=7,50-59:=7,60-69:=6,70-79:=5,80-89:=6,90-99:=8,), F:=(10-19:=6,20-29:=5,30-39:=5,40-49:=7,50-59:=6,60-69:=7,70-79:=7,80-89:=7,)), 'distributed by year of study as':= (M:=(2011:=6,2012:=6,), F:=(2011:=7,2012:=6,)) 14 )
Special Tag Value Attributes Carry the values of a tag when it resides on the Web •
Pfwd:=0.8 – Example probabilistic value forward attribute – Corresponds to P(A|B) in case the = = – Placed in subject expression (bra) part of bra-relationship-ket tag
•
Pbwd:=0.5 – Example probabilistic value backward attribute – Corresponds to P(B|A) in case the = = – Placed in object expression (ket) part of bra-relationship-ket tag
•
assoc:=2.4 – – – – –
•
Example association constant K Usually appended via := to relator in relationship expression Corresponds to eigenvalue of e of quantum mechanics If Pfwd = P(A|B) and Pbwd=P(B|A), then K(A; B) = P(A, B) / P(A)P(B) All probabilities like P(A), P(B | not A), et.c can be calculated from these three values, and hence many measures of EBM and epidemiology.
{0.8, 0.5} would be the dual probability value of the tag. – Simple probabilistic categorical semantic and algebraic behavior follow from h-complex • ½ [0.8 + 0.5] = 0.65 is said to be the real, existential or “some’ component • ½ [0.8 – 0.5] = 0.3 is said to be the imaginary, universal, or “all” component 15 • {a, b} x {c, d} = {ac, bd}, {a, b}+{c + d} = {a+c, b+d}
Simple Q-UEL Tag examples (1) All Q-UEL can be expressed in XML, but Q-UEL also represents an algebraic system
EBM TAG (usually obtained from data mining Consented Incidence Tags) DEFFINITTIONAL TAG Adverse Drug Reaction/Allergy Tag (RDF-USE - example SNOMED Specification Usually processed from EBM-RULES and X-TRACTS ) 16
Simple Q-UEL Tag examples (2) Use of an EHR ket tag to carry GENOMIC DATA Data ket tag from an attribute from an EHR decrypted for system development purposes. Note that following the EHR disaggregation/shredding algorithm described later, you cannot see who this data belongs to, or what other data it relates to, even if decrypted. |metadata:=‘DNA for epitope':=CODE:=GMS:=‘version 2’:=‘filedata;[*(by $PW unlock data)*];number;[87 base pairs/];squeeze DNA; ATGGTGTGTC TGAAGCTCCC TGGAGGCTCC TGCATGACAG CGCTGACAGT GACACTGATG GTGCTGAGCT CCCCACTGGC TTTGGCT/; name;[experimental start of mature peptide/]; GGGGAC;name;[snp/];G;comment;[allotype is g in dbr1*0101, r in db1*0105/] (decrypted data)' club:=1 Q-UELshred110Tue> GMS is a stream of DNA ATGGTGTGTC…. that can however embed instructions, comment, code etc. The above is an actual example from Section “GMS data content 3.2.3 Advanced Considerations for Legacy Input” in Robson, B., and Mushlin,
•
R. (2004) “Genomic Messaging System for Information-Based Personalized Medicine with Clinical and Proteome Research Applications”, J. Proteome Res. (Am. Chem. Soc.) 3(5); 930948 [Award winning paper with Am. Chem. Soc. Press Release] For bit codes for DNA , amino acids and commands etc. see Robson and R. Mushlin J2005) “The Genomic Messaging System Language Including Command Extensions for Clinical Data 17 Categories” J. Proteome Res. (Am. Chem. Soc.) 4 (2), 275 -299
Simple Q-UEL Tag examples (3) Fine grained visible data consent and backtrack/alert notification systems
CONSENTED INCIDENCE TAG (can be generated by Q-UEL’s fine-grained consent language in the record). •
18
Simple Q-UEL Tag examples (4) Probabilistic Semantic Web
X-TRACT TAG (an X-TRACT is a canonicalized Internet source text extract – auto-surf, spawn, and point to, other X-TRACTS)
•
•
One issue relates to the IBM Watson computer, which beat human champions at Jeopardy but thought O’Hare airport was in Toronto [35]. A Q-UEL metastatement did already know that if one travels from A to B, then A is not B. A key rule in that process was that below reached by an automatically generated Google query:-
•
Simple Q-UEL Tag Examples (5) Example of POPPER code (uses tags stripped of web management detail).
= #Test case establishing "TB causes harm". Robson, B. "POPPER, a Simple = 95, 2 Programming Language for Probabilistic = 99, 1 Semantic Inference in Medicine. = 70,2 Computers in Biology and Medicine " Computers in biology and Medicine", 56,
= 95,3 107 (2015). ################################ #1.Syllogism Barbara. Verb form 2. #E.g.All animals eat vegetables. All vegetables are plants. All animals eat plants. = The tags containing binding variables #Test case establishing "TB causes a pathological state". $...come from Q-UEL metastatement tags. These tags search out and edit = 95,0.001 one or more tags to one or more = 100,5 different tags, so enabling syllogisms, ################################## definitions, grammar rules etc. #2. Syllogism Celarent. #E.g. All humans are mammals. No mammals eat rocks. No humans eat rocks. = 20
Simple Q-UEL Tag examples (6) Probabilistic Semantic Web: Meaning of Words.
•
THESAURUS PRIOR PROBABILITY STATEMENT. This kind of tag can be used to provide prior probabilities amongst other input in order to help resolve ambiguities in natural language source text. It represents a probability density distribution of interpretations. By default Pfwd:=1 in this case. In addition, by text analytics of natural language text, text analytics of thesauruses, and access to the Semantic Web, it is also used build thesaurus of meaning, and define the QUELIC language (shown later). 21
Simple Q-UEL Tag examples (7) Start of Automated Systematic Review Tag
23
Simple Q-UEL Tag examples (9) Patient data usually exists on the web in a disaggregated, triple-shredded form •
SHREDDED “BRA” < Q-UELshred110Tue metadata:='History of diabetes':=‘ lnCLAFXgw4.s3D NTSy7p2 2YPKyjpR/UlQ2fBt6fnLJ8Qq9BX Hegd0gMrE5Am UlcZu YDU2hVIaOu1Ul8sw GdjT18YKTgC7.3GPpUCi5JLgNht xgRfpDVTbG VngBZlT/ CM nhAsserL.1mPT2UWBmF87LWl1QMnCuTw8xcAbBP7ToyzF7wd41jhp8WmOkChebkWpXLDkQhwmqfIxhjYMTOt H9p NexMO9/5F5LVsFQ0I xKGyrKMrUHckjNBl0H3k (encrypted data)' |
THAT QUERIES FOR THE NEXT…
SHREDDDED ENCRYPTED DATA “KET” |…> TAG. We can think of forming each one of these by cutting out each attribute from a tag. A Q-UEL EHR on the web is thus comprised of encrypted disaggregated and scattered , mixed-up, elements data belonging to specified metadata, or each an arbitrary small shred of e.g. a DICOM image. By an algorithmic trick, it can carry only encrypted data, but can mixed with everyone else’s data on the web. Each ket requires a bra and a ketbra tag to join it back into the original record. |metadata:='Total cholesterol':= '21b41023b23110191a10981a1d999c1d181a1019189819414b237321032b2b139bc383ab2321032b03ab0928 4374716274702465636279707475646024616471692137393831333935333334383d3132323028267(encrypt 24 ed data)' club:=1 Q-UELshred110Tue> THAT QUERIES FOR THE NEXT “BRA” TAG.
Section 2 Q-UEL Relevance to VistA / OSEHRA Q-UEL Examples modeled for VistA BioIngine Platform and public health analysis DiracMiner – Tags generated DiracBuilder – HDN driven Inference 25
Q-UEL in Relation to Vista ACCOMPLISHED AT PROOF-OF-CONCEPT AND WORKING RESEARCH TOOL LEVEL • Transformation from information to knowledge – large numbers of records can be transformed by Machine Learning and HYPERBOLIC DIRAC NETS for inference – “Quantum transformation” • Enabled tacit or “buried in Big Data” knowledge to better facilitate both population health and patient medicine by EBM / biostatistics etc. in real world public health project. • Employed Dirac algebra and Big Data mining to progress the architecture from lower maturity to medium maturity ; facilitated Deep Big Data mining (in-memory processing) • Created second order semantics by employing probabilistic ontology to achieve semantic interoperability. VistA is now better disposed towards the interoperability challenge - including with ALHTA (DoD) ( http://www.osehra.org/content/joint-dod-va-virtual-patient-record-vpr-iehr-enterprise-information-architecture-eia-0)
• Q-UEL has captured patient information from VistA and HL7 CDA sources (amongst others). • EHR security, privacy, disaggregation, authority, grained consent and alert back-track mechanisms. • Production and automatic use of Risk Factors. Means of extending Framingham scores etc. • LEXIKL as a kind of “SNOMED FOR ALL NATURAL LANGUAGE” - A Global Representation Framework
IN PROGRESS • Smarter CONVERTERS for more fully automated Q-UEL conversion to/from VistA ,HL7 V2 pipehat, HL7 CDA, JSON (Q-UEL’s semantic/ontological capabilities are important here) • Intuitively introduce hypothesis models (multi- node transformation - BigData Hadoop and MapReduce capability this may require our Perl to python conversion) • Payor / Provider analysis, Provider / Pharma analysis, Pharmacovigillence. • NOT LEAST - Reduce medical errors and improve quality through automated translational research use of Outcomes Analysis/Comparative Effectiveness Research/ EBM. • QUELIC – Like LEXIKL but based on a quantum mechanical operator interpretation of natural language (longer term 26 project).
DOMAIN
REQUIREMENTS SUPPORTED BY Q-UEL EXTENSIONS TO PARADIGMS TO CREATE NEW I NTEGRATED PARADIGM
Development • As in OSEHRA roadmap, based on Extreme Programming and Agile philosophy (of using define/build/test Management component team, two-level planning/tracking, iteration , frequent releases initially, concurrent testing,
continuous integration, and regular reflection and adaptation.) • The Q-UEL communication language is also a programming language and a development language. Tags represent an XML-like, probabilistic Semantic Web language but are also algebraic variables (scalars, vectors, matrices) with values, or diadic functions, or operators). They can invoke and download code by name or RDF. It maps to XML: there is an XML-compatible form. Maps to Semantic Web. • Initial phases developed without constant voting systems as e.g. in HL7. • Q-UEL Tags are manually designed in the implementation environment, aided by:1. Tags must conform to the GENERAL SPECIFICATION (which does evolve slowly and must be backward- compatible) and within that many SPECIAL SPECIFICATIONS (which more rapidly evolves). 2. Division into STATEMENT TAGS and METASTATEMENT (or RULE) TAGS. The latter edit one or more statement tags to one or more statement tags (e.g. a syllogism, or definition). 3. Usually hierarchic control with QCOMMAND tags at apex that override SETUP file for a Q-UEL application. 4. Attribute Metadata Language (AML) extends XML attributes to ontological structure or graph. 5. Tag names as special case of attribute with status description as values. 6. Q-UEL binding variables $A… and action $a… variables 7. Report/comment/status/ tracking and logging alert system inside brackets {!...!} 8. Standard use of QUELLOG log report and debug aid file.
Technology
As in OSEHRA roadmap uses Service Oriented Architecture (SOA) approach but enables it by… 1. Use of stakeholder attribute to control flow. 2. RDF pointers NOT ONLY to definitions BUT ALSO to downloadable, actionable codes/applications.. 3. Use of metadata as attribute features that point to data and services that are described by metadata in a form that (i) software systems can use to configure dynamically by discovery and incorporation of defined services, and (ii) maintain coherence and integrity, and such that system designers can understand and manage easily.
Informatics Design
As in OSEHRA roadmap, but in a fundamental way , Q-UEL supports Medical Decision Support (CDS) • Knowledge comes from structured data mining and Q-UEL tags that surf and autospawn the web. 27 • User Experience (UX) provides feedback to progressively replace users by automated protocols.
OSEHRA Priorities
DOMAIN
ACTUAL or PLANNED TERMINOLOGY
1
Allergies/ADRs
RxNorm, IUPAC, SMILES, InChI
2
Immunizations
CVX, epitopes in IUPAC-aa, LEXIKL
3
Problems List
SNOMED CT, LEXIKL
4
Vital Signs
LOINC VS subset, LEXIKL
5
Documents
LOINC (documents), Transcripts in LEXIKL
6-10
Text-based Lab results (bacteriological, chemical, hematology, pathology).
LOINC, LEXIKL
10B
Imaging
DICOM (may be pointer or embedded byte stream)
11-12
Workflow. Appointments, admissions
Encount. DXs – SNOMED, LEXIKL, Q-UEL Stakeholder codes
13
Procedures
CPT4/HCPCS
14
Demographics
OMB/CDC, ISO 6392, alpha 3 codes, LEXIKL
15
Social History/Lifestyle
SNOMED CT, LEXIKL
16
Family History, genetics
SNOMED CT, Q-UEL codes incl. LEXIKL, Resident self, parentgrandparent count (0-7), IUPAC DNA/aa, GMS,
17
Imported scanned legacy records, images
LOINC (documents)
18
Workflow. Med Orders.
RxNorm, IUPAC, SMILES, InChI
18
Workflow. Lab Orders.
LOINC, Q-UEL Stakeholder codes
18
Workflow. Radiology Orders:
LOINC, CPT4, Q-UEL Stakeholder codes, LEXIKL
18
Workflow. Plan-of-care, Continuity of Care
Q-UEL CODE, Q-UEL Stakeholder codes, LEXIKL
LEXIKL because
patients and an doctors don't speak
28 could. in SNOMED, LOINC etc., and that isn't all they need to say even if they
Start of European HL7-CDA-Derived Q-UEL Patient Tag 
