Tools to Simplify Follow-up of Patient Outcomes - SIIM News Fall 2008

Tarik K. Alkasab, MD, PhD

As I took a shift as the radiology resident in the emergency department one evening, a mysterious case popped up on the PACS queue. An abdominal CT scan of an essentially asymptomatic 35-year-old man showed a single hypodense lesion either in or directly adjacent to his iliopsoas muscle. I chatted about the case with some colleagues, and we basically agreed on the differential diagnosis of fluid collection versus cancer and debated whether its apparent relationship to the muscle body made one or the other diagnosis more likely. I dictated the case, giving a differential diagnosis slightly favoring an abscess over a neoplasm, and signed off the report. Though my job was finished for the case, I was not: I wanted to know how the biopsy would come out.

Most radiologists frequently seek information on the subsequent clinical course of patients whose studies they have interpreted, both out of intellectual curiosity and in an effort to improve future diagnoses. Sometimes, radiologists ask referring physicians about the outcome of further workup when they see them later. Other times, for a particularly interesting case, one might make a note of the patient’s medical record number (MRN) for a later search through the electronic medical record (EMR) for subsequent operative notes, pathology reports, or imaging results. One senior radiologist in our department famously keeps a calendar in his breast pocket full of notes on patients he is following. This process of following up on patient outcomes is a critical part of self-evaluation and improvement for radiologists, especially those in training. However, as workloads increase, time for noting cases and then searching for new information has become more and more limited. Our project is based on the idea that radiology information systems should make this process as efficient as possible.

Though the information systems used at our institution, the Massachusetts General Hospital, have some facilities for noting cases for follow-up, most members of the radiology department have found them to be unsatisfactory, and they are used only sporadically. Frequently, the issue is that the cumbersome user interface for capturing cases interrupts a busy, intense workflow. Therefore, as a prototype, we sought to create a simple system that would allow radiologists to quickly record basic information about cases from within the PACS and then permit easy tracking of these cases via links to the EMR system from any computer. We wished to accomplish this with minimal infrastructure using a scripting tool to create small programs layered unobtrusively on top of the existing information systems without requiring any modification or re-configuration.

The challenge was to create a simple database of cases stored in a secure, protected area from within a PACS workstation. We created a three component system – we call it “RaceTrack” (for “radiology case tracking”) – to accomplish this, consisting of a case-capture tool to extract case information and simple user input from the PACS window to a text file; a case-loading tool for creating database entries based on the text files; and a look-up tool to automatically find medical record information from within the database.

We created these tools using AutoHotkey (www.autohotkey.com), an open-source Windows scripting environment that can produce small programs that respond to keyboard events. We chose a Microsoft Excel file stored on a secure, shared storage area to serve as the database.

Figure 1: The case capture tool quickly captures information about a case a radiologist wishes to track. Note that the accession number and medical record number are automatically extracted from the PACS environment.

Case capture was the first obstacle. We were able to create a script such that when the radiologist presses a key combination in the PACS environment, the RaceTrack case capture tool extracts the MRN and accession number for the patient being viewed and displays them in a small dialog box (Figure 1). The user then enters a brief description of the case (usually five words or so), and dismisses the dialog box. The information is stored in a text file on a temporary storage area.

We next created the case-loading tool, whose function is to read the data from the text files and automatically create a new row for each case to be tracked in the radiologist’s personal Microsoft Excel file (Figure 2). This can be run from any computer in the hospital network. Finally, from within the Excel file, a simple script allows the radiologist to select any of the stored cases and use a single keystroke to bring up relevant records in the web-based EMR system or images in the separate web-based PACS. The Excel file also includes columns for noting information on outcomes for further tracking or research.

Figure 2: Each radiologist has a personal Excel file containing the list of followed cases. Note that the accession number is a link to the examination in the web-based PACS, and a radiologist can go directly to the patient’s EMR entry from the patient’s medical record number.

So when I realized that I wanted to know the outcome of my case, I pressed the case capture key combination, and a dialog box popped up in my PACS session with the accession number of the study and the medical record number filled in. I entered a brief description of the case: “Weird iliopsoas thing, abscess vs. met,” pressed “return” to write the information to a text file, and went on with my work. The next day, I ran the case-loading tool, which moved all of my cases of interest from the previous night into my (now lengthy) Excel spreadsheet. Every week or so, I pull up the Excel spreadsheet via virtual private network from home and use the scripts to link directly to the EMR for the patient, and see if a diagnosis has been found yet.

Therefore, our goal of facilitating patient outcome tracking has been achieved. However, there remain many possible improvements to such a system. The two-step process for creating a record in the database should be reduced to a single step. Also, the spreadsheet-as-database could be replaced with a true relational database system, which would allow more sophisticated tracking and searching of the list of tracked cases. We plan to implement these and other improvements to the system.

And the mysterious iliopsoas lesion? My review of the records a couple of months later found that it was a metastasis from the patient’s previously undiagnosed testicular cancer.

Dr. Alkasab is a resident in the Department of Radiology, Massachusetts General Hospital and Harvard Medical School, Boston, MA.

Editor’s note: This article is based on a scientific poster presentation at the SIIM 2008 Annual Meeting in Seattle.