Codename “Dolphin” by fjzhangxiaoquan


									        User Needs Specification

       University of Maryland, CMSC434

Hyunyoung Song, John Brennan, Nima Negahban

            Sunday, May 29, 2011
1 Introduction
    In modern medical organizations, the average medical staff spends twenty percent of
their time monitoring high-dimensional, time-oriented data. This is partially due to the fact
that patient history is not kept electronically, as well as the insufficient support of temporal
dimension data storage and search system of the current system. Using a system that
already has extensive query mechanism for multiple patient history data, medical
researchers, doctors, and receptionists would get more insight into patient history patterns
using this new system.

    „Hippocrates Archive‟ will be medical history archive with temporal dimension. Not only
does it serve as database for patient history, but it provides temporal based query
mechanisms. The main focus of the system is in finding the „temporal pattern‟ among
patient history. A temporal pattern is a sequence of events and inter-event „TimeSpans‟.
The seven queries below, written in different BNF notations, find these different patterns.

E Pattern Queries
This is the base case that requires users to specify non temporal attributes of a single Event.

E-FT and E-VT Pattern Queries
E-FT consists of two Events related to one another by a fixed length of timespan (E-FT), or
of variable timespan(VT).

E*-VT Pattern Queries
Using this type of query, users can describe an event at four different levels of specificity.
(EventType, EventClass, EventName, Value range)
E*WC-VT Pattern Query
Using this query, we can query fixed non-temporal attributes and variable temporal
attribute query.
Ex. “Find patients whose Cholesterol was above 240 for a week”

(E*WC)F-VT Pattern Query
Impose Functional constraints among the event of the set.
Ex. “All cholesterol events within a week that fall between 240 and 260”

(E-VT-E)F Pattern Query
Non temporal attributes are assigned across the timespan, and functional relationships are
set between different sets.

2 User Scenarios
Aid new doctors with patient diagnosing
Dr. Kim Beanstock has been out of medical school for 3 years. She is working at a doctor‟s
office doing her residency, or PGY-3 (post graduate year #3). Dr. Beanstock walks into the
patient room, and sees the patient sitting down. She begins to ask the patient a series of
general questions to get an idea as to the conditions the patient is faced with. After some
time, Dr. Beakstock is still unsure as to what may have caused these symptoms. All other
doctors are busy, so she must find an answer to properly diagnose the patient. Dr.
Beanstock goes over to the computer in the room to boot up Hippocrates. She types in the
symptoms one by one—rash on left arm; ring shaped; mild itch. Dr. Beanstock clicks the
button to begin the process and the program begins. It cycles through all the patient data
on file trying to match up the patient‟s symptoms with a proper diagnosis. The program
spits back the patients that matched the query. It displays the most common diagnosis
with such symptoms which is Tinea corporis, also known as ringworm. The program also
displays a list of recommendations for treatment, this time it is an over-the-counter cream.

Help medical researchers find patterns in anonymous patient data
Chris Brown is a medical researcher with Johns Hopkins University. He has been studying
the possible causes of breast cancer for 6 years. There is so much data, it is hard to zone in
on the causes, other than hereditary. Chris hears about Hippocrates, and Johns Hopkins
provides him with a copy (all patient personal information is stripped out of the program).
Chris types “breast cancer” as his search criteria, and because he is studying this disease is
women ages 40-55, he refines his search to this age group. After generating the results,
there are well over 50,000 patients that match. The software highlights similar patterns
among the results returned. These patterns are listed in order by frequency. Chris notices
that the most common pattern is found in those that went on birth control between the ages
of 15 and 18. Chris is able to pinpoint a potential cause in less than 5 minutes. He is now
going to investigate this further.

Help doctors prescribe the right medicine
Doctor Covington has established his dermatology practice over the course of three decades.
However recently many of his patients are coming back to him with the same symptoms. It
seems that the medicine that he has been prescribing has not been effective for many of his
patients. The doctor loads up Hippocrates and lists all patients with the same diagnosis and
then sees what variations of prescriptions he has written in the past for previous patients
armed with the new data he is able to try new and proper prescriptions for his patients.

Help doctors conduct peer reviews
Doctor lewis-danfield is a pediatrician, his practice is extremely successful and maintaining
profitability mostly because he holds him to high ethical standards and makes time for
every patient and family. Due to his caring approach the good doctor has become
overwhelmed with patient appointments and fears that due to the high stress levels he
might be misdiagnosing or could have made better judgments. He asks his friend Doctor
Trevor Ramerson to use Hippocrates and based on a variety of sets of symptoms given,
compare what was the final diagnosis Ramerson gave his patients. The two discuss their
results on the phone and the good doctor sets his mind at ease.

Pattern Detection for Disease Outbreaks1
Dr. Song, Internist of Washington DC was granted a fund from the NIH (National Institute of
Health) to find out the outbreak of SARS2, seasonal flu, and to figure out how affective the
vaccination of respiratory disease was. If it hadn‟t been for Hippocrates Archive, this job is
worth a month of paper read through. Not that he have this temporal patient history archive
Dr. Song would just search how many cough syrup and liquid decongestant and
vaccinations he prescribed over the past year, past month, past week. From the information
that he retrieves, he would look for a cycle which would indicate outbreak of seasonal flu.
Then he would look for anomaly patterns. While looking at the regular pattern of patient‟s

visit, he realized in January 2004, abnormal amount of visit and medication had taken place.
However, this wasn‟t the only anomaly he was able to find. Before he would get back to NIH
sponsors about the result, he decided to search through the Hippocrates database with
different visual query. He decided to search for patient symptoms such as nausea and rash
during each of these anomaly periods. Dr. Song found 2 match in total. He now decides to
report about his search result that only took 2 hours to NIH.

Discover Family History and Social History
Jim who works at a health insurance company as a researcher has been assigned a task by
his supervisor to find out what could be genetic family disease by looking at a collection of
patient history „Hippocrates Archive‟. One of the diseases of interest is endometritis, he
queries the kind of prescription, blood test, lab test that family doctors and physicians
carried out. In case of endometritis, miscarriage or pelvic inflammatory disease, are the
cause of this disease. Using „Hippocrates Archive‟, Jim founds out that a serious of abortion
or miscarriage could be the cause of endometritis and vice versa. On the other hand, he
finds a cluster of women patients that were suffering from endometritis or taking likewise
medication and that they happened to be relatives.

3 Requirements
3.1    Data Requirements
The software uses information from two files—a people file (people.txt) and an events file
(myfile.txt). The 2 flat database files are tab delimited, and loaded when the program
begins (as opposed to a random access file or relational database that is accessed on-the-

People File Schema
Person id : int
First name : string
Last name : string
Age : int
Gender : enumeration : (Male/Female)

Events File Schema
Person id : int
DocID : int
Date : Date
EventType1 : string
EventType2 : string
EventType3 : string
Int: string

Further investigation into the experimental data will be required to design more concrete
user task and scenarios. Most of focus will be set to the date field. Currently, the date refers
to when event started. The origin of this date must be investigated further as well.

3.2    System Requirements
Must have:
Windows 2000 or greater
128 MB of RAM
10 MB of Hard Drive Space for Hippocrates Software
Pentium III 500MHZ or greater

Should have:
1 GB Storage space for the patient data
256 MB of RAM for optimal performance

Could have:
Internet connection to make research easier

4 References
[1] Journal of American Medical Association
[2] National Library of Medicine
[3] Bade, R., Schlechtweg, S., and Miksch, S. 2004. Connecting time-oriented data and
information to a coherent interactive visualization. In Proceedings of the SIGCHI Conference
on Human Factors in Computing Systems (Vienna, Austria, April 24 - 29, 2004). CHI '04.
ACM Press, New York, NY
[4] MomMD: Online magazine and association

To top