Homework: Coding Medical Data
                              
            (Introduction to Medical Informatics)
         (http://www.cpmc.columbia.edu/edu/textbook)

The purpose of this homework is to make you more aware of
the challenges involved in coding medical data. The goal is
to convert a narrative report, which is readable only to
human beings, into a form that is usable by automated
systems (e.g., for clinical research databases and automated
decision-support systems).

Take the admission note on page 41 of Medical Informatics:
Computer Applications in Health Care by Shortliffe and
coeditors. Code it as a series of facts organized as
follows:
     
     (patient, parameter, value, time, modifier1, modifier2, ...)
     (patient, parameter, value, time, modifier1, modifier2,...)
     (patient, parameter, value, time, modifier1, modifier2, ...)
     ...

"Patient" is a unique identifier (medical record number) for
the patient, and may assumed to be "1" for this execise.
"Parameter" is what the observer was looking at, and "value"
is what he or she saw. "Time" is when it was seen, or when
it was true. "Modifier" is any additional information that
helps describe the fact (e.g., certainty of the observation,
side it occurred on, degree). There may be zero, one, or
more modifiers per fact.

There will be many ways to code the note. You will have to
decide such things as (1) what to code as a separate
parameter and what to stick into the modifier fields; (2)
what to store as an aggregate value, and what to break into
separate facts; (3) how to represent a negative finding.
Here is an example of how you might code "the patient
reported having had his cholesterol checked last year, and
he thinks the result was 240."
     
     (1, cholesterol, 240, 1992-01-01 - 1992-12-31, certainty=possible)

The followig abbreviations are used in the text:

ID = identification
     y/o = year old
     CC = chief complaint
     HPI = history of present illness
     pt = patient
     MD = physician
     
     ER = emergency room
     TB = tuberculosis
     PMH = past medical history
     hx = history
     NCD = normal childhood disease
     meds = medications
     NVD = normal vaginal delivery
     ROS = review of systems
     Gen'l = general
     PTA = prior to admission
     x (with a bar over it) = except


Issues for class discussion on Homework: Coding Medical Data

- vocabulary
     myalgia
     symptoms (value = myalgia)
     HPI: myalgia

- duplicate headache as CC and HPI?

- many uses of value - where to encode the relation
     measurement_of (age is 42)
     type_of (non-productive cough)
     treatment_of (belladonna for migraine)
     allergic_to vs treated_with predisone

- data that varies with time
     age = 42
     "in the past 10 days"

- onset_of
     do symptoms stop that day?
     if not, then how long do they go?

- how to represent relationships among findings
     belladonna treats migraine (diagnosis)
     belladonna failed to change symptoms

- parsing "non-productive cough but is a non-smoker"
     two separate findings

- how to represent changes
     symptoms worsened

- using multiple rows vs list

- represent time as absolute vs relative

- modifiers cannot contain independent observations

Additional assignment:

Given only the following allowed queries, write the
pseudocode necessary to retrieve the data from your
assignment.
     
     get_record given pt and parameter
     get_record given pt and parameter and value
     get_record given parameter
     get_record given parameter and value

1. did the patient feel nauseous on 5/26/89?
     yes

2. was the patient getting worse on 6/1/89?
     yes

3. how much does the patient smoke (packs per day)?
     0 PPD

4. why was belladonna prescribed to the patient?
     to treat migraine

5. how bad was the patient's headache?
     severe

6. what patients have had frontal headaches?
     (unclear)