Course home





BHI Home Page

Transformational Technologies for
Biology, Medicine, and Health
Honors 222C & MEBI 498A, Spring, 2013

Project #1: Medical records & medical notes: Gibberish? Or a patient information source?

A team-based project (teams of ~2 students)

Due April 23, 1 pm.

There are two main questions for this project: (1) Can you understand a physician note from an EMR? (2) What happens if you try to use this information for patient record-keeping within a patient-based personal health record (a PHR)?

To investigate these two questions, you will (1) read through a set of medical records, including a small number of discharge summaries from the MIMIC-II database of ICU patients, and (2) work with a PHR system, namely, Microsoft's Health Vault.

For this project, our source of electronic medical record (EMR) data is The MIMIC-II database of ICU patients. This is a very large anonymized database of ~32,000 patients who visited the ICU at Beth Isreal Deaconess Medical Center in Boston over a 7 year period. (Beth Isreal is a primary teaching hospital for Harvard Medical School.) It includes high-resolution temporal data (waveforms), labs, orders, and physician and nursing notes for more than 40,000 ICU visits. I have obtained a copy of the "Clinical database -- demo version" for our studies. This demo version is a sampling (random, perhaps) of about 10% of the ICU visits, so it is more than 4000 patients / visits. It does not include any of the waveform data.

Notes about the Notes:

I will give each team of students (via email) a set of abour 130-150 patient visits from MIMIC-II. Although the data is technically provided as a PostGres relational databae, the files I will distribute are simply compressed comma-delimited text files, and therefore readable in any text viewer. (However, due to the length of text fields, do not try to import these files into Excell). Each visit is documented via a very large number of files, corresponding to labs, medications, events, procedures, demographics, ICD9 diagnoses, and textual clinical notes. Of these, it is only the Notes files that I will distribute (but you can ask me for others, if interested) via a zip file that includes ~100-140 files named "NOTEEVENTS-xxxxx.txt"

There are at least three sorts of Notes in each of these files:

(1) The discharge summary, which is a summary written by the physician (or possibly a team of physicians) at the time of discharge from the ICU. Note that in some cases, the patient may simply be discharged to the floor, or for observation. In the ICU, "discharged" doesn't necessarily mean "go home". The discharge summary is usually first, but not always (you may have to search for it.)

(2) Radiology reports. As we'll cover in module #2, these are reports from imaging procedures such as MRI, CT, and Ultrasounds. For project one, you will typically just skim these, although it might be interesting to note what the "reason for examination" or "underlying medical condition" are.

(3) Nursing progress notes. These are day-by-day (or more frequent) notes from the nursing staff. They should definitely be read when you are trying to understand all the details about the patient. These can have interesting insights that are quite different from the discharge summaries. They also contain day-to-day details that may NOT be interesting or relevant.

NOTE: Throughout these files you will see text inside of brackets, indicating de-identification. For example "On [**2866-7-2**], Mr. [**Known patient lastname 711**] passed away...." Here, the actual date and name of the patient have been removed and replaced by identifiers.

Details: Can you understand the notes?

From these notes files, each student in the team should randomly select 6 patients to read thru. Each of these patients must have a discharge summary, and all of these patient must survive through the end of the ICU stay. Because a fair number of these patients die in the ICU, or for whatever reason do not have a complete discharge summary, you may have to skim quite a few more than 12 to get this set. I'm not concerned about random selection, so you may pick-n-choose as strikes your fancy.

The first deliverable (see below) is a set of 6 one-paragraph summaries of what you can understand after reading your six patient visits. Don't spend too much time on these, and don't worry about understanding everything in this first pass (but do look up enough so that the paragraphs are coherent!)

Next, meet with you partner, and together, select two "interesting" patients from the set of 12 to study in greater detail. With these two patients, your aim is to understand as much as possible. Look up acronyms, use Wikipedia, and try to figure out as much as possible about the patient, their diagnoses, their medications, their treatment in the ICU, and their prognosis. At this stage, you should also skim any nursing notes and radiology reports that are included in the relevant NOTEEVENTS file. When you go to outside resources to help understand more, please make a note of what resources use used.

I'd strongly recommend doing this section collaboratively -- two heads are better than one, and much time can be saved by not looking up common acronyms twice, etc. As described in the deliverable section, you'll need to write up an essay describing what happened to these two patients in the ICU, and what their health status was (both before and after the ICU stay). However, before you do too much work writing up descriptions, please read through and think about all of the deliverables for this project (see below). As you delve into your two patients, you'll want to be taking notes to help you with those parts.

Details: Can you use the record's information in a PHR? :

In part one, the learning goal is simply to understand the information objectively, as well as to think about what makes this hard and complex for the average person to do. For part two of the assignment, the goal is to understand and imagine more of the subjective, personal impact this sort of information would have on the patient and the patient's caregiver(s).

Imagine that the patient data you looked is your own mother or father. Suppose that your are the primary "caregiver" for this patient -- this usually means you are the closest surviving healthy relative, and that you live in close physical proximity. (It also often means that you have full medical power-of attorney, so you can see all records, etc.) You've taken care of your mom/dad in the past, and it's expected that you'll do so again for this episode in the ICU. Thus, you'll be responsible for helping him/her with recovery, and more importantly, you have to help them and perhaps other close family members (e.g., your siblings) understand what has happened in the ICU, and what might happen in the future.

A tool that might help with these tasks is a personal health record system (a PHR), such as Microsoft Health Vault. For this project, a task is to create a pair of health vault accounts, and populate them with information from the two medical records you selected. This does not mean that you should attempt to copy all information from the discharge summaries into Health Vault (which wouldn't work well, anyways). Instead, the PHR should help you abstract out and focus on the main problems of your patient. Hopefully, you can use this tool to write down the "most important" information from the notes. You can then imagine making this portion of your Health Vault account accessible to your siblings or other family members, as a means of communication. Note that the discharge summary (and the nursing notes) often contain personal and family history information. These are important, and you must make your Health Vault information (and your essay) consistent with this information.

As described below, you will have to create a narrative that describes the patient status, and that uses portions of MS Health Vault to help with this task.

Deliverables: (all of these must be submitted to the Catalyst Drop Box for the class).

The first simple deliverable is due Wed, April 10th, 7pm:

For each of your 6 patients, write a single paragraph indicating important features of the patient and their ICU stay such as: primary diagnosis (or diagnoses), age & gender, status at time of discharge, medications, major procedures, etc. Do not worry if you can't capture all of these coherently, but try to get as much as seems reasonable. Unlike all other parts of the assignment, this one should be handed in individually. This portion of the assignment will count little toward your overall project grade, but is an important checkpoint to reach before carrying out the more time-consuming parts of the project.

Next, you must comprehensively understand the two patients, and enter information into MS HealthVault.

The remaining deliverables are both due on April 23rd, 1pm:

  • A set of screenshots, showing, for both patients, how you stored some of the relevant information into MS HealthVault. Please number these, as you may wish to refer to some in your essay (see below)
  • A 5-6 page essay, with four required sections, as described below. Single-spaced, 12 point font, 1 inch margins all around. Reference list does not count against the 6 page limit.

Required sections for essay:

(1) Executive summary, describing what you learned, and your big messages. You don't need to answer all of these in the executivie summary, but some example issues to address might be: Can you understand Physician notes? Would they be helpful for the patient and their families? Do PHRs work? Do you think patients should look at physician notes? (No more than 1/2 a page)

(2) Descriptions of each of the two selected patients and their ICU stay. Write these as if you were describing the situation to a close relative (i.e., a sibling). Spell out all acronyms in their first use, and include web pointer and citation to external resources, if these are useful. Where appropriate, refer to screenshots from MS HealthVault. (2-3 pages) Try to make these personable and story-like -- see grading rubric.

(3) An Informatics analysis: (2-3 pages)

  • What resources were important to carry out the task of understanding the note? How easy were these to use? Could "most people" use these?
  • What is the difference in langauge and emphasis between a physician note and a descriptive narrative, such as the one you wrote in section 2, for a family member?
  • What aspects of MS HealthVault helped or impeded your ability to record the most important aspects of this ICU stay? Were there any important aspects that could not easily be recorded in HealthVault? (and why couldn't they be recorded?)
  • Longitutinally (i.e., not for a single ICU visit) how useful do you think the notion of a PHR is? How useful might MS HealthVault, or perhaps some other PHR, be over the long run?

(4) Summary (less than a page): Reiterate your main message. If appropriate, include a self-reflective component -- what aspects of the learning in this project were the most surprizing to you? What aspects were the most difficult? (Since this is a team project, feel free to write separate paragraphs, possibly with names, if these self-reflective components were significantly different for the two members of the team.)

Optional sections & other ideas:

  • What might be the Impact on the patient of reading these notes? How might this affect their behavior?
  • Rather than the MS Health Vault model for PHRs, what would happen if you had Indivo-style technology? (See readings by Halamka and Mandl.) Would this change your main message and thinking about PHRs and physician notes?
  • What are the ethics of making the MIMIC-II database publically available? (Please note and read the data use agreement for accessing the full MIMIC-II database)

Grading rubric:

This is a team-based project, so there will only be one grade given per team.

The essay is a formal, academic work, and the writing must be your own. (However, this is a team project and a co-authored essay, so when I say "your own", I mean the 2nd person plural.) Please cite references (and web sites) as appropriate. Once again, be aware of student academic responsibility.

I have a pet peeve against the use of passive voice. Please write in active voice, using first person plural (as this is a team project) where appropriate. For example, the summary might have sentences that begin: "In summary, we felt that MS Health Vault performed....." Another pet peeve I have is to dislike sentences that are too long or convoluted. Especially when the ideas are complex, you should strive to keep your sentences simple.

2.5 -- 2.9
3.0 -- 3.3
3.4 - 3.7
3.8 -- 4.0
Patient description Most aspects of the patient diagnosis, prognosis, procedures, and medications are explained. Only minor clarity problems. Clearly written. Students explain most aspects of the patient visit, including Patient diagnosis, prognosis, procedures, and medications. Students explain all aspects of the patient health status . Students demonstrate that medications and diagnoses have been carefuly researched. Patient description seems realistic, personable. Patient description is clear and story-like. Students investigate aspects of the patient health status that are open to debate and less certain.
Health Vault information Students have used HV, but not many aspects of the patient state are documented. Students have used HV extensively to attempt to capture many aspects of the patient's health. Screen-shots are clear and fit the description text. In addition, students have explored how some of the ancillary apps might work with the patient & caregivers.  
Informatics analysis There are some clarity and coherence problems. Not all required components are well-discussed. Well-written. Student describes the revlevant information resources that were used for medical understanding. Students include an analysis of strengths and weaknesses of HV, as well as some broader thoughts about PHRs longitudinally. In addition, students include some ideas for improvements to HV, or describe why HV or PHRs broadly struggle to capture some sorts of knowledge. Analysis is provocative and imaginative. Students' analysis surprises me, and demonstrates insight into broader problems.
Quality of writing Poor structure; problems with argumentation and/or inappropriate paragraph organization. Frequent sentence-level problems. Straightforward construction of sentences and paragraphs. Clear argumentation. Some minor sentence-level problems. Very clear argumentation, with good intro and summary sections. Almost no sentence level problems.  


Last Updated:
April, '13

Contact the instructor at: