Health Informatics Assignment Week 19 to 21

Week 19 Questions

1. With the advent of full-text searching, should the National Library of Medicine abandon human indexing of citations in MEDLINE? Why or why not?
2. Explain why you think PMC is or is not a good idea.
3. How would you aggregate the clinical evidence-based resources described in the chap-ter into the best digital library for clinicians?
   

   Answer:

   Virtual libraries provide a number of the identical services, however their focus tends to be at the digital components of content material. The developing amount of scientific records available in IR systems and digital libraries calls for new strategies to choose that that is exceptional to apply for medical choices. The philosophy guiding this technique is Evidence-Based Medicine (EBM), which can be considered a set of equipment to tell medical choice making. It lets in scientific revel in to be incorporated with exceptional medical technological know-how. An opportunity attitude, specifically when groups and/or proprietary collections are concerned, is the digital library. Digital libraries may present many characteristics with brick and mortar libraries, but additionally tackle a few additional demanding situations. The digital libraries have the whole capacity of, noting that the underlying technologies were evolved with federal leadership, that files face both technical and operational demanding situations. Digital libraries and commercial publishing ventures want mechanisms to ensure that documents have persistent identifiers. The unique structure for the web predicted through the net engineering challenge force was to have every uniform resource locator (URL), the deal with entered into an internet browser or used in a web link, connected to a uniform resource name (URN) that could be continual. The combination of URN and a URL, a uniform resource identifier (URI), would offer a persistent right of entry to digital objects. The resource for resolving URNs and URIs never implemented on a large scale. One method that has all started to see massive adoption through publishers, specifically scientific journal publishers, is the digital object identifier. The DOI has these days been given the status of a popular by way of the NISO with the designation Z39.84. The DOI itself is notably simple, which includes a prefix that is assigned by the IDF to the publishing entity and a suffix that is assigned and maintained with the aid of the entity. For example, the DOI for articles from the journal of the American Medical Informatics Association have the prefix 10.1197 and the suffix jamia.M####, where #### is a variety of assigned by the journal editors. Likewise, all publications within the digital library of the association for Computing machinery have the prefix 10.1145 and a unique identifier for the suffix (e.g., 345508.345539 for the paper Hersh et al., 2000). Publishers are recommended to facilitate decision by way of encoding the DOI into their URLs in a well-known manner, e.g., http://doi.Acm.Org/10.1145/345508.345539.

4. Devise a curriculum for teaching clinicians and patients the most important points about searching for healthrelated information.
5. Find a consumer-oriented Web page and determine the quality of the information on it.
6. What are the limitations of recall and precision as evaluation measures and what alternatives would improve upon them?
   

   Answer:

   Measures based on significance contain their drawbacks. Though no one disputes that users expect applications to recover valuable content, this is not obvious that the sum of related retrieved images seems to be the true expression of how well a software works. This has been recognized whether medical users are reluctant to be worried about certain steps whenever they are actually finding another reference to a specific query and are willing to provide it irrespective about how many certain related information we lose or even how many non-relevant documentations they obtain (decrease precision). This acknowledges that if interventions which use a more situational view of relevance could not be established to test user engagement, so the only substitutes can be recall and precision. Throughout the years, a variety of user-oriented tests were performed aiming at clinical knowledge consumers. Acknowledging the weaknesses of memory and accuracy in testing medical users of IR systems, the reason for such experiments is that the main purpose to use an IR system is to seek a response to a question. Although the user definitely needs to locate appropriate documentation to examine the problem, the quantities of these documentation becomes less essential than whether query is addressed effectively. Assessment focuses on the analysis of tests is extremely popular mostly as framework for developing methodologies for recall. Research teams could use comparisons to strong future in a systematic format with the same experimental set-up, consequently allowing for comparing the results. Furthermore, user-oriented assessment is complex and expensive, even though it is extremely effective, and therefore challenging to duplicate. That consistency and optimization is what allows the set of tests so enticing. Nonetheless, attributable to its isolation from fact, there seem to be a number of disadvantages to evaluating catalogue-based evaluation. Test databases tests allow a variety of hypotheses: that report significance is separate of one another; that certain records are extremely applicable. Alternate measures like positive predictive value and related graphs of Precision / Recall have been used less commonly. Most bioinformatics experiments create and test classification algorithms to be implemented to extremely imbalanced data, wherein the number of negatives greatly surpasses the number of positives. ROC is a common and powerful tool for measuring binary classification efficiency. CROC, CC, and PRC were proposed as replacements to ROC or are used less often. In the detailed review, designers demonstrate from many viewpoints, the gaps between the different steps. Just the PRC adjusts with the positive-negative ratio. As PRC plots express the sensitivity of classification models to datasets through simple visual indications and allow a precise and intuitive understanding of the functional output of classifiers. The findings of the separate study strongly suggest use of PRC charts as being the most descriptive method for image examination.

7. Select a concept that appears in two or more clinical terminologies and demonstrate how it would be combined into a record in the UMLS Metathesaurus.
8. Describe how you might devise a system that achieves a happy medium between of intellectual property and barrier-free access to the archive of science.

Week 20 Questions

Questions

1. Researchers in medical AI have argued that there is a need for more expert knowledgein medical decision-support systems, but developers of Bayesian systems have argued that expert estimates of likelihoods are inherently flawed and that advice programs must be based on solid data. How do you account for the apparent difference between these views? Which view is valid? Explain your answer.
2. Explain the meaning of Internist-1/QMR’s frequency weights and evoking strengths.What does it mean for a finding to have a frequency weight of 4 and an evoking strength of 2? How do these parameters relate to the concepts of sensitivity, specificity, and predictive value that were introduced in Chapters 2 and 3?
3. Let us consider how deDombal and other developers of Bayesian systems have usedpatient-care experience to guide the collection of statistics that they need. For example, consider the database in the following table, which shows the relationship between two findings (f₁ and f₂) and a disease (D) for 10 patients.

In the table, ,D signifies the absence of disease D. A 0 indicates the absence of a finding or disease, and a 1 indicates the presence of a finding or disease. For example, based on the above database, the probability of finding f1 in this population is 7/10570 percent.

Refer back to Chapters 2 and 3 as necessary in answering the following questions:

1. What are the sensitivity and specificity of each of f1 and f2 for the disease D? What is the prevalence of D in this 10-person population?
2. Use the database to calculate the following probabilities:

   p[f₁uD]

   p[f₁u,D]

   p[f₂uD]

   p[f₂u,D]

   p[D] p[,D]

3. Use the database to calculate p[Duf₁ and f₂].
4. Use the probabilities determined in b to calculate p[Duf1 and ,f2] using a heuristicmethod that assumes that findings f1 and f2 are conditionally independent given a disease and the absence of a disease. Why is this result different from the one in c? Why has it generally been necessary to make this heuristic approximation in Bayesian programs?

4. In an evaluation study, the decision-support system ONCOCIN provided adviceconcerning cancer therapy that was approved by experts in only 79 percent of cases (Hickam et al., 1985b). Do you believe that this performance is adequate for a computational tool that is designed to help physicians to make decisions regarding patient care? What safeguards, if any, would you suggest to ensure the proper use of such a system? Would you be willing to visit a particular physician if you knew in advance that she made decisions regarding treatment that were approved by expert colleagues less than 80 percent of the time? If you would not, what level of performance would you consider adequate? Justify your answers.
   

   Answer:

   In clinical decision making, it shall be kept in mind that various decision-making tools shall be employed so that proper decisions by the clinicians could be made. The performance needs to be enhanced in the case of ONCOCIN. Various measures shall be introduced so that the better decision making could be practiced. Among the active research obstacles is a need to improve the analytical methods used among medical professionals to encode the vast array of information used during critical thinking. Albeit well-established strategies including the use of frameworks or guidelines operate to store empirical or interpretive information, there are still many difficult problems. Likewise, humans have an impressive ability to comprehend information changes over time, to evaluate seasonal variations and also to propose recommendations of progression of the disease or illness reaction to previous therapeutics. Researchers seek to progress computer-based modelling methods for these tasks. It is equally obvious that merely giving computers loads of factual information does not enable them qualified in a sector unless they're already professional in implementing the information properly. Specifically, in this field, enhanced knowledge of human problem-solving behavior is enabling workers to explore decision-supporting tools that further precisely simulate the mechanism through which professional practitioners shift from observations through diagnoses or plan development. MDT consultations have been used to customize professional decision making on treatments available, however these knowledge and experience could not be transferred digitized amongst centers. Developers established a computer vision solution to better optimize medical decision-making, engineered to anticipate MDT choices regarding cancer treatments. The difference among MDT- and guideline-based judgments about adjuvant chemotherapy suggests that several non-clinicopathological factors, such as patient discretion and accessibility of resources, are factored through local experts therapeutic decisions making. The EON program is a realistic representation from one of the latest decision-support programs based on guidelines that offer clinical guidance for care in compliance to predetermined guidelines. The structures mentioned here exemplify wildly divergent configurations in certain detail. QMR is mainly used for a stand-alone system, DXplain is a self-contained system, but usually often accessible via the World Wide Web, and EON provides a series of software modules designed to be incorporated into broader clinical decision support systems.

5. A large international organization once proposed to establish an independent laboratory—much like Underwriters Laboratory in the United States—that would test medical decision-support systems from all vendors and research laboratories, certifying the effectiveness and accuracy of those systems before they might be put into clinical use. What are the possible dimensions along which such a laboratory might evaluate decision-support systems? What kinds of problems might such a laboratory encounter in attempting to institute such a certification process? In the absence of such a credentialling system for decision-support systems, how can health-care workers feel confident in using a clinical decision aid?

Week 21 Questions

1. What are two advantages and two limitations of including visual material in the following teaching programs:
   1. A simulated case of a patient who is admitted to the emergency unit with a gun-shot wound
   2. A lecture-style program on the anatomy of the pelvis
   3. A reference resource on bacteria and fungi
2. You have decided to write a computer-based simulation to teach students about the management of chest pain.
   1. Discuss the relative advantages and disadvantages of the following styles of pres-entation: (1) a sequence of multiple-choice questions, (2) a simulation in which the patient’s condition changes over time and in response to therapy, and (3) a program that allows the student to enter free-text requests for information and that provides responses.
   2. Discuss at least four problems that you would expect to arise during the process ofdeveloping and testing the program.
   3. For each approach, discuss how you might develop a model that you could use toevaluate the student’s performance in clinical problem solving.
3. Examine two clinical simulation programs. How do they differ in their presentationof history taking or physical examination of the patient?
4. Select a topic in physiology with which you are familiar, such as arterial blood–gasexchange or filtration in the kidney, and construct a representation of the domain in terms of the concepts and subconcepts that should be taught for that topic. Using this representation, design a teaching program using one of the following methods: (1) a didactic approach, (2) a simulation approach, or (3) an exploration approach.
5. Describe at least three challenges you can foresee in dissemination of computer-basedmedical education programs from one institution to another.
6. Discuss the relative merits and problems of placing the computer in control of the teaching environment, with the student essentially responding to computer inquiries, versus having the student in control, with a much larger range of alternative courses of action.