Paper on Severity of Illness

This page was last updated on Saturday, May 26, 2018.  Check back for more updates throughout the semester. 

Purpose

The objective of this assignment is to show you how data from electronic health records can be analyzed using standard query language.  You are asked to analyze claims data and report the accuracy of a proposed severity index compared to established indices in the field.  By end of semester you are expected to submit the paper for publication.

The instructor will provide you with claims data from a national sample of hospitals.   You are asked to analyze these data and report the accuracy of several methods of measuring severity of illness.  You would need to use Standard Query Language to accomplish this task.  Before the analysis is done, you need to prepare a paper reporting your plans and methods.  Once the analysis is done you will write the results and discussion section of your paper.  You are asked to submit your paper for peer feedback by midterm and for publication by end of semester.

The idea of students contributing to the literature is nothing new.  For sometime, it has been clear that when students are asked to teach others they learn a great deal more than when they study the material for themselves.  This is why medical residents are asked to  "lean one, do one, and teach one."  Managers too can learn more from doing and teaching others.  The objective of asking you to write a paper is to help you learn in more detail the use of standard query language.

Steps in Completing the Paper

  1. Select a topic: All students are encouraged to write on the accuracy of co-morbidity indices but they can choose different diseases.  Some can focus on myocardial infarctions and cardiac events.  Others might want to focus on diabetes.  Still others may want to focus on different diseases.  Once you choose a disease you need to select from the data only cases in which the disease was present.  In claims data this is done through selecting cases in which particular diagnoses codes are present.  You need to search the web or PubMed to identify which codes you should select to obtain a comprehensive list of patients.  PubMed MI patients  Cancer patients
  2. Select the procedures you wish to test.  Use PubMed and Google Scholar to find relevant articles on measurement of co-morbidities, severity measures and prognostic indicators in the disease you are interested in.  Select at least 2 approaches that use claims data to predict prognosis of the patients.  One suggestion could be to go with Charleston Co-morbidity index and our proposed approach to measurement of severity.  
  3. Get the data from your instructor.  The data we use is available from Agency for Healthcare Quality and Research.  In order to make sure that you have access to the data, we have requested the data and it is available through your instructor. 
  4. Select a journal where you like to submit the paper:  Select the journal where the paper will be submitted and follow the format of the journal.  Some example journals are provided here: 
    • Health Services Research Journal  More►
    • Quality Management in Healthcare  More►
    • Health Care Financing review  More►
    • Journal of American Medical Informatics Association  More► 
  5. Check published literature:  Look at PubMed and also check Google Scholar to find out scientific progress in the field.  
  6. Do not make claims that are not supported by data.  Do not digress from your topic area.  Do not reference web sites.  The literature review should be comprehensive, i.e. refer to all relevant papers, but write less than 3-paragraphs long. 
  7. Draft paper: The draft of the paper must be finished by one week before the mid-term.   Write the introduction and section on methods and data sources ahead of completing the analysis of the data.  Analyze data available through the instructor.  Make sure that your method section refers to both how you selected patients with different diagnoses and how you measured their severity.  Write well.  Make sure your points are clear and are demonstrated with appropriate charts and figures.  Describe the source of the data and the results of your analysis.  Give a case example that supports your main finding.  Explain all terms when they are first mentioned and list them also at end of the paper.  Make sure that the draft paper has the following sections (See the journal you are submitting to for more detailed instructions):
    1. Title page:  Choose a title that appropriately describes the paper.  Include author's names but no contact information or other personal information. 
    2. Abstract:  The summary should be able to stand for itself and the reader should understand the scope of the paper from it.  It should refer to each of the major sections of the paper.  A well organized paper will allow you to take a sentence from each section in the paper to organize the summary.
    3. Introduction:  Discuss why this topic is important.  Provide a short review of the literature. 
    4. Methods:  Provide a step by step guide on how you analyzed the data, where the data comes from and other details of the study.  Describe what statistic you used to compare the accuracy of various indices. The test of validity should be conducted on cases not used in estimating parameters of the revised index. Different methods can be used in comparing two indices:
      • If predicting length of stay, then Analysis of variance can be used to measure the percent of length of stay explained by the different co-morbidity and severity indices.  For example, the following regression model will test if the Alemi’s approach to measurement of severity adds additional information beyond what is captured by Charlson’s approach.

        Length of stay= a (Charlson index) + b(Alemi index) + e

        In this equation, length of stay is the dependent variable.  The Charlson and Alemi indices are the independent variables and parameters a, and b are estimated from the data through regression analysis and e indicates a random error term.  If the parameter b is significantly different from zero then the hypothesis that the Alemi index does not add new information to Charlson’s prediction is rejected.   The percent of variance explained by each index is used to provide a benchmark of the accuracy of each index.
      • The accuracy of indices in predicting mortality rates can be compared using receiver operating curves.   PubMed►
    5. Results: Provide the findings in standalone tables and figures.  
    6. Discussion:  Discuss the limitation of the work and the importance of the findings.  Do not claim more than the data show.
    7. References and links:  Give references to literature cited following the format of the journal you are submitting to. 
    8. Reviewers comments:  Show peer review comments and extent of revisions done based on reviewer comments
  8. Get feedback:  The instructor will assign your work to two other students to review; likewise you need to review the work of two other students.  Use the following rubric to organize your response (make comments for each of the sections indicated.  Do not leave any of the sections blank) as well as expect your peer reviewer to follow the same rubric:  class="style4">
  9. Dates
    1.  Report of the paper was later than expected and whether your review was done in 24 hours.
  10. Begin with what worked well.  Point to specific sections of the paper and use adjectives liberally to praise the author.  This is the only place you are allowed to use adjectives, in all other sections avoid use of any adjectives.
  11. Presentation:
    1. Discuss whether each major point has been made with an appropriate visual aid. 
    2. Discuss the use of font size to mark paper sections and the hierarchy of ideas
    3. Discuss if color has been used appropriately to highlight significant points
    4. Discuss writing style and errors.  Discuss if the methods and sources of data were accurately and succinctly depicted.
    5. Discuss the organization of the paper.
    6. Discuss if references are linked to PubMed and other literature.
  12. Content:
    1. Make sure that relevant literature is reviewed.  Check that a search on PubMed does not identify articles that should have been reviewed but have not been reviewed.   Indicate what you searched for, list the top 10 relevant articles and discuss how many were included in the paper. 
    2. Discuss if the authors have followed the recommended outline and whether their departures from the outline makes sense.
    3. Check that the title is appropriate for the paper.  Make sure that the paper does not digress into unrelated materials.
    4. List what questions you had that were not answered by the paper.  Discuss if the paper's limitations were addressed
  13. What did you learn: 
    1. Discuss what you learned from the paper that you would try to do better in your own draft paper.  Note that your comments in this section should refer to how you would change your own paper.  The focus of improvement is on you and not the author of the paper you have reviewed.  Give a timetable by which you would complete the changes.
  14. Grade
    1. Rank order the two papers you have reviewed, indicating which in your opinion was best.
    2. Compare your work to the works you have reviewed and indicate which was best.
  • Revise the manuscript:  Using the feedback you have received and what you have learned from your review of work of others, revise the manuscript.  A significant revision must be done, even if you have no critical comments and even if the original draft was the best paper in the world.  Make sure that you include a section at the end of the paper referring to what was revised.
  • Present your work:  Present your findings to the entire class.
  • Post to the web or publish:  The instructor may ask you to combine your work with others in class and submit your work for publication. 
  • Timetable

    The first draft of the paper is due one week before the mid-term.  Your peer review of other student's work should be completed within 24 hours of receiving the paper.  Your final draft is due at midterm.  The final version of the paper is due after Spring break.  You must report where you have submitted the paper by end of semester. 

    Alternatives

    1. Students who wish to prepare a video instead of the paper can do so.  All videos should also be submitted to peer review before final submission to the instructor and posting to the web.
    2. If you have a different project that fits your career better, please let the instructor know.  The project must involve data from electronic health records and use of Standard Query Language to analyze the data.