Projects, designs, and writings on health IT

2014-08-10

Scheduler: A housestaff compliance tool

6:33 PM Posted by David, MD , No comments

I built this with Eugene Gitelman in 2014 at the Hospital of the University of Pennsylvania.

The Problem

Scheduling hundreds of residents in two-week rotations can be difficult. There are duty-hour regulations (80-hour work week, 16-hour maximum shift for interns, 28-hour maximum for residents), and there are practical problems, like missing a day when transitioning from night shift to day shift. Formerly, 4 internal medicine chiefs spent 4 full-time weeks building and checking the schedule.

The Solution

Scheduler grabs the schedules from amion and systemically checks for violations to the multitude of rules. This was implemented in July 2014. Now, the chiefs have real-time duty-hour violations checking and they are approaching zero scheduling errors.

Technologies

Ruby on Rails

2014-04-27

What an EMR Built on Twitter Would Look Like

5:54 PM Posted by David, MD , , 1 comment

This post originally appeared in The Health Care Blog 2014-04-27



What an EMR Built on Twitter Would Look Like


By DAVID DO, MD

WILD PREDICTION: It won’t be long before every patient has a Twitter feed, and doctors subscribe to them for real-time updates.

This is a time when the demands of being a physician are changing, and we need to leverage technology to maintain awareness of a huge number of patients. There is also increasing need for handoffs and communication between providers.

Here’s the bottom line: how can we improve technology when doctors seem so resistant? They are not happy with their EMRs, and rightly so, because they were built to do too much for too many.

Current system is inefficient

The EMR has become essential for documentation, billing, medical reasoning, and communication, among other things. Currently, documentation is built on a system of daily progress notes. If I consult a cardiologist about a case, he needs to go through each note, containing narratives, laboratory values, vital signs, and physical exams.

A patient with a seven-day hospital stay may have twenty notes that need synthesis to put together the story–this can take hours per patient!

In an age where more providers are involved in a patient’s care (whether due to duty hour restrictions, or the increasing presence of specialists for every problem), this inefficiency is not acceptable.

Providers hate redundant documentation

Not only is this system of documentation difficult for doctors to interpret, it is tedious to produce. Documentation has become a chore, and I’ll explain why. For patients in the hospital, documentation often exists in two separate streams, the daily progress notes (located in the official chart or the official EMR) and the summary/signout document (often located in a Word document on a hard drive in the hospital).

Billing and legal is based off the progress notes. The summary/signout document is often unofficial, developed from a need for a more up-to-date and comprehensive summary of what is happening with a patient. Think of it as a “final report”, synthesizing all the data, organizing ideas and describing the plan.



This also leads to poor communication

Because the workflow is built around the summary and signout document, daily progress notes have become ceremonial and rote. It therefore becomes unreliable—can I trust that the provider meant to check this box as normal?

Further, communication between providers suffers because the accessible document is often the progress note.

The solution comes from social media

That’s where the innovation of social media comes in. I see the major innovation of social media as of organizing the immense amounts of information. In the 90′s you could join chat rooms about a particular interest or hobby.

On Twitter, you are participating in an infinite number of chatrooms at once, and you can be part of a conversation on #SXSW14 or @THCBstaff just by searching.

Here’s how a Twitter-based EMR would work:

All clinical data, events, notes, and annotations are included in a single live feed. This stream will serve as the centerpiece of all functions, with a few carefully designed filters.



With this conceptual framework, there are several natural improvements on the existing EMR.

  • The Clinical Awareness view: By watching the raw live feed, you are immediately aware when your patient’s CT scan is complete or laboratory values are available.
  • The Patient Timeline view: By looking at a patient’s feed, you can learn about the entire course of illness. It’s much faster that flipping through pages of notes that may have been copied forward.
  • Progress Note Generation: Doctors and nurses need only focus on documenting in the Twitter feed. Progress notes are automatically produced from all the interval data.
  • Conversation view: different specialists can have a discourse about a difficult case, and it’s built into the EMR. This is the meaning of patient-centered care.
In medicine we spend a tremendous amount of time and money on antiquated software that acts like a digital version of a paper chart. We can learn a few lessons from the innovators in social media about how to manage large amounts of data.

David Do, MD is a resident physician at the Hospital of the University of Pennsylvania. He is an agile software developer and Chief Technology Officer at Symcat.com.

2014-04-22

How the practice of prepopulating progress notes died

2:51 AM Posted by David, MD , No comments

Nostalgia

I spoke to a senior resident who reminded me that overnight interns used to print and populate progress notes for the daytime team before their arrival. I got the sense he was implying "they don't make them [interns] like they used to". Curious about how this practice could have disappeared entirely, I did some investigation. As it turns out, interns used to handwrite vital signs (and their 24-hour ranges) on every single progress note.


Temp: (99.3 - 101.1) HR: 67-82O2: 91-92BP: 90-125/ 50-60RR: 16-20

Even with a computerized EMR, and digitized vital signs, it can be incredibly time consuming to look at graphs for 5 vital signs x 8 patients and translate it to the above format.

How this practice died

Once we pushed the IT department to automatically populate vital signs on the progress notes, we experienced tremendous time-savings. Printing your own progress notes in the morning became trivial, and the practice of pre-populating them became of marginal benefit.

Interns jobs are still very much automatable

Interns, come in earlier than the rest of the medical team to collect data. I estimate 75% of my time is dedicated towards data collection; that is, looking at paper and digital sources of data and scribbling it on the progress note. (When we shift two electronic provider documentation, I will spend my time flipping between screens and using copy-paste.)

Perhaps 10% of my time is dedicated to synthesizing data. What might be causing this sodium value to be so high? How might a high sodium value be causing this patient's confusion? How can I adjust the insulin regimen to keep my patient's glucose better controlled? What should I do with the results of this chest x-ray?

Finally, the last 15% percent of my time will be spent at the bedside, asking a few questions and updating patients about their results.

75% Data collection
10% Data synthesis
15% Patient interaction

Why this matters

Data collection still takes the majority of my time every morning, but luckily, it's the most automatable step. In fact, I think its entirely automatable. IT departments need to start prioritizing workflow and time-saved. It's my hope that through this design blog, we can shift some of that data collection time towards patient interaction.


2014-01-16

Rethinking the signout

5:33 PM Posted by David, MD , No comments

What is Signout?

The Signout has a standard universal structure but is implemented in many forms, from dedicated apps to excel spreadsheets. The goals are always the same--to provide a comprehensive but quick view of a patient and the up-to-date status of the hospitalization, including clinical status, relevant test results, and pending test results.
A good Signout document reflects the thought process behind a patient's care. It is a living document, and is therefore always relevant, as opposed to the static H&P and daily progress notes.
Here I outline how the Signout could be the central component of the EMR, with various views of the same fields that are tailored to specific tasks.













2014-01-14

Idea 1: Admit patients twice as fast

10:44 PM Posted by David, MD , 10 comments
In any existing EMR, users find themselves doing the digital equivalent of gathering data from disparate sources to evaluate a single patient. Here, I present a basic user-centered design that could drastically improve workflow and ensure that no information is missed.

2013-08-11

What the Recent Data Breach Says About the State of Health IT

5:48 PM Posted by David, MD , No comments
This post originally appeared the The Health Care Blog 2013-08-11

What the Recent Data Breach Says About the State of Health IT

By David Do, MD

Recently officials at Oregon Health Sciences University discovered that residents in several departments were storing patient information on Google Drive, and had been doing so for the past two years. They treated this discovery as a breach of privacy and notified 3000 patients about the incident.

While I don’t condone the storage of patient information on unapproved services like Gmail or Google Drive, this incident pretty much highlights the sorry state of information systems within the hospital and the unfulfilled need by physicians for tools that facilitate workflow and patient care.

It says something that the Oregon residents felt compelled to take such a drastic action. I don’t know what punishment – if any – those responsible were given by administrators for their “crimes.” I’ll leave it to readers to make up their own minds about the wisdom of the unauthorized workaround and the appropriateness of any punishment. But I do know that the message the incident sends is a very clear one.

We’re screwing this up. There is really no earthly reason why it should be any more difficult to share a patient record than it is to share a Word doc, a Powerpoint or yes, even a cloud-based Google Drive spreadsheet.

Why the Breach Happened

What’s going on here? Let’s say I admit a patient to the hospital.  Our friend was hospitalized here just last month, and like many patients, he has dementia or is poorly educated, and does not know the names of the medications he takes. Unfortunately, I don’t have the ability to see what he takes or how he was treated during the prior admission because the records in the computer are there for documentation’s sake and don’t contain any meaningful information. This is clearly a problem for me.

Therefore I will spend time calling outside facilities to gather information and repeat several tests and imaging procedures.

Medical care has become a team sport, and residents have developed systems for keeping track of their patients and communicating to other physicians. It takes some time to think about and process each patient that comes in, to consolidate all the information. Ultimately, I need to boil that information down to a five-minute description on the patient, their problems, the status of their current admission, and what needs to happen before they go home.  We do this in the form of a signout document.




Figure: The signout document has four to five columns and includes the To Do list for each patient.

The EMR does not have a good way to store information in this format, and  additionally I have no way of editing this in real-time to communicate with my
coworkers what still needs to be done. That’s why residents were storing their  signouts in Google Drive.

What providers need here is simple data management. We need to store and access this list from different computers. We need the ability to enter a subset of those data  using a custom form, and the ability to print subsets of those data to create a To Do lists, rounding sheets, or progress notes.

What we can learn from this breach

In the end, there was no actual breach of data. Don’t blame the residents for what  happened in Oregon. They have, better than anyone, demonstrated an unfulfilled
need for good medical software that is user-tested.  Products that develop through hospital IT departments are slow, non-intuitive, and buggy. Hospitals need to invest
in health IT, above and beyond that need to meet the requirements of Health IT  legislation. We need designers, software product mangers, and user testers from
Google.

David Do, MD is a graduate of The Johns Hopkins University School of Medicine and a resident physician at the Hospital of the University of Pennsylvania. He is an agile software developer and Chief Technology Officer at Symcat.com.