This site is 100% ad supported. Please add an exception to adblock for this site.

IntroHI final


undefined, object
copy deck
what is the MIB?
medical information bus - the standard for interconnectivity of patient monitoring devices, developed by the IEEE.
what is closed-loop therapy
when a patient monitoring system can administer the appropriate treatment indicated by its readout.
virtual memory
a scheme by which users can access information stored in auxiliary memory as though it were main memory.
What is ANSI
American National Standards Institute, nonprofit organization responsible for approving standards in the US, government authorized.
What does CPT stand for?
Current Procedural Terminology - developed by the AMA , 4th version most widely used.
define pre-coordinated and post-coordinated
in standard vocabularies, a pre-coordinated vocabulary attempts to define every possible condition ahead of time. A Post-coordinated vocabulary can combine terms to express any idea, not just those anticipated by the designers. Some have sanctions that limit what you can combine, which decreases expressiveness somewhat
Name two hierarchical standard vocabularies
READ codes
three characteristics of good ontologies
1. inheritance - attributes carried from inheritance. e.g. in defining kinds of transportation, and you divide them into wheels vs. hull, all the items in the "hull" branch will have hulls.
2. relationships are shared across a level.
3. items can have multiple inheritance.
define sensitivity
probability of a positive test given the presence of the condition
what is RISC?
reduced instruction set computing - attempts to reduce the number of instructions for an operation. Thought to speed up processing
what features of computers make them useful for medical record keeping?
1. more accurate at counting and math
2. perfect memory and searchable memory
3. not subject to cognitive bias
define artificial intelligence
the branch of computer science that tries to get computers to do the things humans can do
what components of medical record-keeping are humans better at?
1. recognizing patterns
2. interpreting "fuzzy" information such as emotional content
3. tasks requiring physical dexterity
4. taking responsibility for events
reasons why doctors are not using CPR
- workflow (bedside rounds)
- politics or organizational reasons
- fear
- startup costs
- lack of proven outcomes
- security concerns
- data entry problems
- conversion of old records
problems with paper records
- only in one place at a time, may have lots of different versions, copies; transportation issues
- storage, ensuring safety, backing up are difficult
- hard to read - illegible, abbreviated
- difficult to search or do chart review
- only one view of information offered
- no ability to abstract
- cannot perform automatic computations for quality control, correlations, alerts, etc.
ways to evaluate a CPR's usefulness
1. comprehensiveness of information
2. duration of use; years of information stored
3. degree of data structure
4. ubiquity of access
A network that connects computers owned by independent institutions and distributed over long distances
wide-area network
Key issue in vocabulary development
expressivity vs. consistency
What is GALEN?
a standard vocabulary, designed by Alan Rector, which is postcoordinated and uses a "structured medical knowledge"
define specificity
probability of a negative test given the absence of the condition.
5 assumptions of decision analysis
1. probability rule = everything can be assigned a probability
2. order rule - payoffs can be ordered according t preference
3. equivalence rule - given prospects A>B>C, we can assign p such the certainty of A vs C is prob B
4. substitution rule - equivelent rule can be embedded in a more complicated rule
5. choice rule - people behave logically; if A>B>C, people choose A over B.
define computationally intractable
problems where the best solution you can come up with is O(c^n)
which is more efficient, symmetric or assymetric keys?
Uses for imaging in medicine
1. diagnosis
2. Assessment of known problem
3. guidance of treatment (i.e. surgery)
4. communication (telemedicine)
5. education and training
6. research
4 tasks we can do with an image
1. generation
2. management
3. manipulation
4. integration
Plain films, Ultrasound are what sort of image?
CT, MRI are what sort of images?
define tomographic
an imaging technique that takes slices of the body
define pixel
a single element in a picture
define voxel
a volume element in a 3-d imaging technique
define CCD
charge-coupled device: arrays of sensors that capture image in a digital camera
name and define 3 types of image resolution
spatial: sharpness, how well you can distinguish points that are close together
Contrast: how well you can distinguish points that are similar in color/intensity
temporal: time needed to create an image.
how many frames per second are needed to have real-time imaging?
thirty frames/second
define PACS
picture archiving and communication systems - used for digital radiology systems
name and define (with numbers) two kinds of compression algorithm
Lossy - loses some data but can get compression on the order of 20:1
Lossless - keeps all information, can compress on the order of 2-3:1
Comment on the definitions, differences in JPEG and MPEG
both are lossy compression algorithms. JPEG is picture, MPEG is music.
what is a computer virus?
a software program that is written for malicious purposes to spread from one machine to another and do some kind of damage. Many are self-replicating
without compression, how much storage would be required for a digital radiology system?
assuming 250 exams/day, 10 MB/exam, this is approx 1 terabyte per year.
define DICOM: what is it, who does it, what does it do?
digital imaging and communication in medicine - is the ACR-NEMA vendor-neutral standard for medical digital images, patient demographic and clinical data, and exam-specific data
relationship between DICOM and PACS
PACS systems are vendor-specific systems but if they use the DICOM standard then information can be shared between systems
enabling technologies needed for PACS systems
1. software
2. storage hardware
3. transmission hardware
4. query/display hardware
How is image data different from lab results?
not easy to scan and read; difficult to set up alert system for abnormal exams.
name two methods of retrieving images
1. index images (tag with text, retrieve tag)
2. retrieve by content (get images that look similar to a query image). This is very difficult, requires the computer to "understand" the image
what are 3-D and 4-D stacking of images?
3-D stacks tomographic images to make a 3D representation of the body. 4-D compares stacked images over time.
name and define 4 types of image manipulation; give examples
1. global processing - computation on entire image, i.e. grey-scale windowing on CT scans to highlight bone vs. lung windows
2. segmentation - extract region of interest from overall image. Computationally difficult. example - identify an organ based on its border
3. feature detection - extract useful information about the image eg volume of heart or size of fetus
4. classification - determine type of object found - eg identify cxrs with pneumonia
define BI-RADS
standardized vocabulary for reporting mammogram results - ACR standard
define Ultra-Star
Image reporting vocabulary by Bell and Greenes at Harvard, now incorporated into UMLS
problem with developing software for reporting image data?
developing a standard vocabulary. Currently reports limited to areas that have a controlled vocabulary, limited applications
What is the visible human project? Who does it
NLM contract with the U. of Colorado at Denver. attempt to link visual images to medical/anatomical knowledge. Uses transverse sections, CT, and MRI from male and female cadavers.
a storage medium that is suitable for reuse but cannot be erased or rewritten
WORM (write once, read many)
visual-analog scale
a method for valuing health outcomes, wherein a person simply rates the quality of life with a health outcome on a scale from 0 to 100.
What is ISO
International Standards Organization
What is ICD?
International classification of Diseases - currently available thru 10th edition - designed to track mortality data. Has clinical modification (ICD-9 CM) which is used for diagnosis coding in many clinical settings.
What is the DSM?
Diagnostic and Statistical Manual - developed by APA, coordinates with ICD 10.
define abstraction as it relates to standard vocabularies
examine the data and label it
what is a reference information model?
an ontology used in HL7 version 3.
Bayes theorem
p[D/F] = p[F/D]p[D]/p[F]

= (sens)(prev)/[(sens)(prev)+(1-spec)(1-prev)]
What is an assembly language?
the instrucction set architecture. It's the interface between higher-level languages and machine language.
define information retrieval
part of computer science which studies the retrieval of information (not data) from a collection of written documents. The retrieved documents aim at satisfying a user information need usually expressed in natural language
reasons why information retrieval may be imperfect
-unintentional biases in database structure
-intentional biases (marketing manipulation)
-study design
3 types of text content you may want to do information retrieval on - give examples:
1. original content (primary literature) - medical journals that report original research
2. bibliographic or synoptic (secondary literature - i.e. medline
3. synthesis (tertiary literature) - systematic reviews
define synoptic
indexed literature (from synopsis - a summary of relevant information of a source)
what is primary, secondary, and tertiary literature?
primary - original text, like a research article
secondary - a listing of primary literature, indexed based on concepts (synoptic)
tertiary - a synthesis of primary literature, such as a systematic review
define query
an expression of an information need for input into a data retrieval system.
how can a query be expressed?
- natural language
- constrained language such as boolean queries, fill-in-the-blank, or pick-lists.
two general computational approaches to information retrieval
1. semantic - trying to understand the text
2. statistical - no attempt to understand, just use statistical properties to answer questions
define NLP
Natural language processing - the study of how we hear and interpret natural language.
define indexing
the process of assigning tags from an ontology to a set of content - improves retrieval performance.
define MeSH.
a medical indexing system, used in medline, that uses concept extraction to assign tags to text.
define concept extraction. why is it difficult?
defining the important concepts in a work to be indexed. "a shorthand representation of content". difficult because there's no gold standard.
how good is inter-observer agreement in manual indexing?
define recall in the context if information retrieval
what percentage of all relevant articles are retrieved.
define precision in information retrieval
what percentage of articles retrieved are relevant
which is easier to measure, recall or precision? Why?
precision is easier to measure because you're looking at a known set. With recall, you don't have any way of knowing how many things you missed.
What kind of recall and precision percentages can expert searchers (medical librarians) get? how about novice clinicians?
recall - experts get about 50%, novice get about 27%.
precision - experts get 58%, novice get 38%
what is a web search engine?
service that maintains large database of web pages and allows users to search that database. examples - google, altavista, yahoo, DirectHit.
4 important characteristics of web search engines
1. how do they interact with the user?
2. how do they build, structure, and maintain the data
3. how do they retrieve results?
4. how do they order results?
advantages and disadvantages of automatic web crawlers for building a search engine database
advantages - frequent updating, can maintain very large database
disadvantages - no quality control, difficult to search
advantages/disadvantages of manual curation of a web search engine database
advantages: quality control, can be categorized to facilitate retrieval, can have an ontology (eg yahoo)
disadvantages: slow, expensive, difficult to keep updates
4 ways web search engines may handle queries to select "hits"
2. keyword only
3. stemming (run=ran=running)
4. stop words (like "and" - usually not included)
4 ways web search engines may rank results
1. relevance ranking e.g. excite
2. connectivity analysis - measure of link popularity - e.g. google, most search engines. very effective
3. date e.g. pubmed
4. result type (PDF vs. HTML vs. Citation)
define bioinformatics
the study of how information is represented and transmitted in biological systems
name the two major types of biological sequence
1. Nucleic acids - DNA, RNA
2. Amino acids - proteins
what are five things you might do with sequence information?
1. alignment
2. similarity assessment
3. search
4. motifs
5. prediction of structure, function
3 ways to study biological function
1. mathematical models
2. simulations
3. sequence --> structure --> function
what is a biological network? give examples
interconnected biological elements leading to a functional whole. examples include gene regulatory networks, interacting biochemical pathways, interacting physiologies, and interacting populations.
what is Gleevec?
a cancer drug that was a product of rational drug design.
what is a microarray (AKA array)?
high-throughput technique for generating biological data, up to a thousand reactions take place in microscopic spots on a glass slide. results have to be evaluated statistically.
define In Silico experiments
experiments that are run on a computer using data correlated from multiple databases (rather than in vivo = in the body or in vitro = in the lab)
four possible future goals for bioinformatics
1. computational model of physiology for in silica experiments
2. design of new compounds for medical and scientific use
3. engineering new biological pathways
4. data mining for new knowledge
5 reasons to evaluate a health informatics product
1. promotional
2. scholarly
3. pragmatic - to improve the system
4. ethical - make sure system is safe
5. medico-legal
define NP hard
NP = non-polynomial
if a problem cannot be solved in polynomial time, it is called NP-hard
define NP-complete
a problem that can be reduced to a small set of problems that have no known polynomial time solution
what is O-notation?
a way of evaluating algorithms by how they scale with increasing input sizes.
what is the objectivist approach to evaluation?
AKA logical-positivist, this approach uses properties of the resource being studied that can be measures such that rational observers agree regarding what to measure and the measurement result. Prefers numbers over words.
what is the subjectivist approach to evaluation?
AKA intuitionist-pluralist, this approach evaluates resources in ways that may legitimately be interpreted differently by different observers. uses verbal descriptions, and describes situations and contexts rather than defining "better" and "worse".
Name some objectivist study designs
- comparison-based
- objectives-based
- decision facilitation
- goal free
name some subjectivist study designs
- quasi-legal
- art criticism
- professional review
- responsive-illuminative
describe an objectives-based study
asks the question of whether a resource meets the designer's objectives. Measures specific benchmarks. Example - usability studies.
describe a decision facilitation study
seeks to help people make decisions about the future of a resource. Data collection methods are designed to answer the questions posed. Tend to be conducted at early states of development, to guide resource development.
describe a goal-free study
evaluators are blinded to the designer's goals and attempt to understand all the effects of the resource rather than limiting to specific questions.
describe a quasi-legal study
set up as a mock trial, where the resource gets judged.
describe an art criticism study
a critic uses the information resource and then writes a review of it.
describe a professional review study
a panel of experienced peers spend several days in the environment where the resource is installed and draft a report after their visit
describe a responsive/illuminating study
investigators immerse themselves in the environment where the resource is operational, take an ethnographic approach to understand the impact of the resource. goal is understanding, not judgement. all viewpoints are represented.
name the three stages of technology studies
1. technical characteristics
2. efficacy
3. outcomes
define independent variable (in the context of experimental design)
is included in a study to explain the measured values of dependent variables.
define dependent variable (in the context of experimental design)
the measured outcomes of interest
define internal validity
conclusions drawn from the specific circumstances of an experiment are justified.
define external validity
conclusions drawn from a study are generalizable to a non-study population.
explain the assessment bias and how it can be avoided
in a study, measurements are skewed by the observers expectations or feelings. This is avoided by blinding.
explain the allocation bias and how it can be avoided
experimental groups are different in some important way. This is avoided by randomization.
Explain the Hawthorne effect and how it can be avoided
performance improves when people know they are being studied. avoid with a placebo group or control group.
explain the checklist effect
if people are given a form, data collection is better.
explain the placebo effect
people get some therapeutic effect if they expect to.
define QALY
quality-adjusted life year.
what is the typical cutoff for cost-effectiveness in medicine?
$50,000 per quality-adjusted life year. arbitrary figure derived from hemodialysis.
definition of NP-hard
according to the national institute of standards and technology, "The complexity class of decision problems that are intrinsically harder than those that can be solved by a nondeterministic Turing machine in polynomial time." example is the travelling salesman problem "is there a tour with length less than k" is NP-complete: it is easy to determine if a proposed certificate has length less than k. The optimization problem, "what is the shortest tour?", is NP-hard, since there is no easy way to determine if a certificate is the shortest.
definition of NP-complete
a problem that can be demonstrated to be both NP and NP-hard; solution to such a problem would lead to solution of all NP-hard problems. an example is minesweeper game!
what is a problem-oriented medical record?
data is organized around problems that a patient has.
X-ray: average bytes per study, number of images per study, pixels per image
20 million bytes per study
2 images per study
2048 x 2650 pixels per image
MRI : pixels per image, images per study, total bytes per study
256 x 256 pixels
100 images per study
12 million bytes per study
CT: pixels per image, images per study, bytes per study
512 x 512 pixels per image, 60 image per study, 30 million bytes per study
Ultrasound: pixels per image, images per study, bytes per study
512 x 512 pixels
30 images per study
7.5 million bytes per study
Nuclear medicine: pixels per image, images per study, bytes per study
128 x 128 pixels per image, 30 image per study, 0.5 million bytes.
what is MYCIN?
an early clinical decision support system that de-emphasized diagnosis to concentrate on appropriate management of patients who have infections. rules are conditional statements that indicated what conclusions can be reached or actions taken if a specified set of conditions is found to be true.
arden's syntax
is a shareable formalism so guidelines can be shared once encoded; is an HL7/ANSI standard
define HTTP
hypertext transfer protocol, used to make multimedia information resources accessible on the world wide web
volume performance standard
a system authorized by congress for paying for medicare physicians services, intended to control volume.
what are the stages of developing a standard?
early implementation
What is a DRG?
diagnosis related group - developed at Yale for classifying hospitalizations for the purposes of payments
Define ontology
a set of concepts and the relationships between them
What is UMLS?
Unified Medical Language System - encompases the metathesaurus, the semantic network, and a specialist lexicon.
What is a declarative language?
involved goals and operators, you give it a function, a starting point, and a desired goal, and it figures out how to get there. Used in AI; not very fast or efficient
What is a procedural language?
a language that follows a list of instructions, like Pascal or Perl.
what is run-length encoding?
a compression technique for graphics
define cryptography
the science of using mathematical tools to protect data.
what does NP stand for
name three important early EMR's
Regenstreif (IUPUI)
what are the only universally reportable conditions?
births, deaths, fetal deaths
what is consumer informatics
concerned with issues related to biomedical information as it relates to health care consumers
which two groups of people use consumer health resources most often?
chronically ill people and newly diagnosed people
what are barriers to consumer informatics? name 3
digital divide - poorer people don't have computer access
security issues - some information can't be put on the web
Impersonality of electronic interface - some health care needs to be more personal.
what is a disruptive technology?
one which allows people to do something by themselves that before, they could only do with the help of professionals.
4 types of patients
accepting - about 8% - let doctor decide everything
informed - 55% - doctor makes decision then they read about it
involved - 28% - shared decision-making
In control - 9% - want to be the one making the decision
explain Willingness to Pay
an approach to valuing human life based on the values implied by the choices people make every day to change their probabilities of living or dying. For example, a person's implicit valuation for life could be calculated by how much he'll pay for a car airbag that has a certain probability of reducing his chance of death.
a method where a computer program projects a two-dimensional image directly from a three-dimensional voxel array by casting rays from the eye of the observer through the volume array to the image plane.
what is the HON code?
a certification of web sites that is supposed to verify quality, but it's not very reliable.
who was Lee Lusted?
published some of the first articles in the 1950's about decision support
describe Leeds abdominal pain system
helped to make decision about whether an acute abdomen should go to surgery or not. listed possible diagnoses, assumed they were mutually exclusive and collectively exhaustive, and applied algorithms based on Bayes theorem.
explain forward and backward chaining
forward driven decision altorithms are data driven - if A and B and C, then likely result is D.
backward chaining starts with a hypothesis and then collects information to verify or reject it.
Mycin - was it forward or backward chaining?
backwards. it has a list of production rules and a rule engine that cycles through the rules and fires if criteria are met.
what is the chief problem with rule-based clinical decision support systems?
scale - there are so many rules that all interact, it gets slow and hard to maintain.
describe internist-1
a decision support system based on a "mind dump" of Jack Meyers, has a differential list whose probabilities vary as you put in new information. AKA QMR (quick medical record).
Describe the HELP system
designed at LDS hospital. was a CPR that included event-driven decision support. Was fore-runner of arden syntax (a shareable guideline representation formalism).
what is a clincial practice guideline?
a systematically developed statement that addresses a specific clinical scenario
what is a MLM
a medical logic module: used in authoring environments; if-then rules, chained or branched, flow-chart like algorithms. once integrated into runtime database, you get clinical decision support.
what is the EON project?
uses protege which is essentially an ontology editor to model knowledge. It's object-oriented. Has re-usable methods - propose and revise. example is athena project for hypertension advice from Stanford.
models for delivery of advice from a clinical decision support system
oracle - you go to it, submit your question and relevant data, it gives a plan.
attending system - critiques proposed plans rather than coming up with a plan.
what is a case manager?
coordinates multidisciplinary care in a health care system.
3 categories of information involved in planning a patient care system
1. patient specific data
2. organization-specific data
3. domain knowledge
WHat is TMIS?
technicon medical information system - one of the first out there,
what is a CHIN?
community health information network - an idea that failed because of too many competing political and financial interests
what is HEDIS
healthplan employer data and info set - developed by the NCOA, judges health plans by collecting data showing events between a presumed cause and outcome. NCQA, CDC, AHRQ, and HCFA involved. good tool for prevention monitoring.
what is a wildcard character?
in information retrieval systems, a method that allows a single or multiple-character expansion somewhere in the search string
von Neuman machine
a computer architecture that comprises a single processing unit, computer memory, and a memory bus.
What does ICPC stand for?
International classification of primary care
What is SNOMED?
Systematized nomenclature of Medicine - from American College of Pathology, importane because it's the largens, most comprehensive. Current version includes SNOMED CT (clinical terminology) and SNOMED RT (research terminology)
define representataion as it relates to standard vocabularies
describe something in as much detail as possible
how has managed care influenced CPR's
it leads to a push to decrease health care costs. CPR's reduce duplication of tests and efforts, provide decision support and more efficient use of available information. May simplify effects of capitation, DRG's, etc.
comment on difference between the internet, and internet technologies
you can use internet technologies such as a web browser without being on the internet. This may provice lots of built-in functionality, such as security
what is the Nyquist frequency?
frequency that data should be collected at, generally twice the expected frequency of change. For example, ekg's change at 150/second, so should sample at 300/second.
what's another term for logical-positivist?
what's another term for intuitionist-pluralist?
what is type I error?
a false positive result
what is type II error?
a false negative result
marginal cost-effectiveness ratio:
[C(b) - C(a)]/[O(b)-O[a)]
C= costs of two interventions
O = measured outcomes of interventions
are people's time included in direct or indirect costs?
direct. Indirect is things like time lost due to illness
what is a formative decision?
made as a result of studies undertaked while a resource is under development
what is a summative decision?
made after a resource has been installed.
what does IDN stand for?
integrated delivery network
What is CCC?
Center for Clinical computing - a CPR developed at Beth-Isreal hospital.
what is Diogene?
developed in geneva (?) a CPR that works via printed reports ordered by phone
describe a comparison-type study
an information resource is compared to something else under controlled conditions. Outcomes are measured. Statistical arguments are based on randomization, controls, etc. testing...testing...

Deck Info