Taking Radiologist Recommendations to the Next Level

2019 Imaging Innovation Award Winner: Process Improvement for Follow-up Radiology Report Recommendations of Lung Nodules

By Radiology Group of Abington

Communication of imaging results and follow-up recommendations to patients and primary care providers (PCPs) is a challenge for healthcare systems. In addition, tracking whether a patient’s follow-up has been completed is another significant gap in care coordination. Patients are often unaware of or cannot even understand the significance of radiology findings or follow-up recommendations reported after imaging procedures. In addition, patients may not have a primary physician listed at time of imaging if the first encounter is in the emergency room or if their primary care physician or specialist works in a different EHR platform. Communication of imaging results to different healthcare providers is challenging with the myriad of existing EHR systems, which often lack interoperability with other clinical entities. Description of lung nodules in radiology reports can vary widely if a standardized lexicon is not used. Moreover, follow-up recommendations by radiologists can be varied for certain size lung nodules because an individual’s risk factors to develop lung cancer may not be known at the time of dictation.

Aims and objectives. The goal of this project was to develop a better automated tracking method and communication tool to reduce the likelihood that needed follow-up studies are missed by patients and clinicians. Secondary outcomes we hoped to achieve were to track the patient condition at follow-up imaging. Also, we tried to determine which communication pathway was better at increasing the likelihood of follow-up being completed: a letter to the PCP or a letter to the patient.

Leadership and project management. Monthly meetings with a larger multidisciplinary team included two physician patient safety officers, a radiology administrative director, a physician who served as chief medical information officer, a champion radiologist, a surgical resident and a senior hospital administrator. This group designed the process of letter notification to the PCP and patient, identifying which clinical radiologic follow-up was to be queried for proof of concept. We chose lung nodule(s) due for follow-up as the primary focus since this scenario has the most widely accepted, evidence-based recommendations. A smaller working group consisted of the champion radiologist and an analyst with IT and nursing experience (the latter employed by the billing company) who met weekly to review identified cases. The radiology champion met with the primary care physicians and their office managers at monthly meetings to educate them about these follow-up letters that they and their patients would be receiving in the mail. The chief medical officer strongly emphasized that the PCP would be responsible for determining if follow-up imaging would be needed after review of the patient’s risk factors and clinical history, even if the PCP was not the one ordering the original imaging study. The PCP was the most central care coordinator best equipped at managing patient problem lists and orchestrating needed follow-up since the PCP had the most complete clinical history and relationships with referring specialists.

Key steps. Initially using a natural language processing (NLP) commercial software, we could identify the radiology reports that were overdue for follow-up based on data from 2016. However, we also realized the laborious work needed to track the patients, opting instead to use an export into Excel spreadsheets from the NLP which contained the full report and other key elements needed for tracking. The export only contained NLP identified reports with follow-ups detected as overdue. As we began taking exports from older studies, we realized that the Excel spreadsheets became quite voluminous with multiple patients and that access to prior reports and their time stamps was needed. We also needed to reference the report data exported from the NLP with data from the billing system to get addresses for the patients and the PCPs. This was done manually in spreadsheets with lookup functions. At this point, we hired a computer programmer.

The first iteration of the custom designed analytics system was used to review radiology reports and significantly decreased the time requirement of reviewing and assigning intervals for follow-up. Once brought into the new system, the profile logic in the customized analytics system was applied to data-mine the overdue reports for the lung profiles. We implemented automation of Lung-RADS report follow-up due dates to enhance system intelligence, using Lung-RADS categories 1 through 4 to calculate the due date and to automate closing when the patient returned for follow-up. We further enhanced the software by tracking which lung nodules had resolved, improved or worsened at the time of follow-up. With our tracking of clinical conditions such as worsening of lung finding at follow-up imaging, we could identify patients who returned for follow-up with new lung nodules so that they could be placed in a separate category within the analytics system as patients who are at higher risk.

Positive outcomes. We saw steady improvement in rates of completed imaging follow-ups over the course of the project. For example, the percentage of completed follow-ups more than doubled, rising from 26.5% in 2015 (baseline) to 41.3% in 2016 and, finally, to 59.7% in 2017. This coincided with our increase in targeted communications: Within 2017 alone, for example, we sent 43 letters in July and went on to average more than 250 per month from August through October. Later we refined our follow-up profiles to include searching for follow-up terminology within a certain word count. At this time, the system output averaged more than 200 letters sent per month. Further, while close to 70% of patients were stable at follow-up and 5% had improved, 11% were found to be worsening, while 6% had a new lung nodule and 4.2% had a new abnormality. We also measured the follow-up completed numbers by using the PCP letter send date or the patient letter send date to find which method of notification was more effective. We found the average return of patients improved by 10% when letters were sent to the PCP (45%) as compared to being only sent to the patient (35%).

Innovative elements. When we first started the lung nodule tracking program, we be-lieved that a commercially available NLP would be the sole answer to this issue. We quickly discovered that the complexity of tracking interval follow-ups in order to com-municate the information to patients and their PCPs efficiently rendered the NLP soft-ware quite limited in its ability to achieve our objectives. We realized that, since all radi-ology reports, patient demographics, and primary physician contact information was centrally located, a customized, internally developed software solution was the only path for us to utilize our resources wisely. We also gained knowledge on critical trends by utilizing the filters on the user interface to perform analysis and graph the data. We could thereby quickly make changes within a few weeks after identifying opportunities with each software iteration. Acceptance by primary care physicians of the notification letter was a challenge but more easily accomplished after meeting the office managers and PCP administrative meetings. Many of these physicians never knew that their patients had radiology findings that needed follow-up, and some were resistant to take on this responsibility of addressing the situation since they did not order the initial study. Enlisting the aid of senior leadership to convince the primary care physician to take on this role was essential in order to close existing communication gaps at the root of the problem.

.

Submitted by Philip Lim, MD, chair of radiology at Abington Hospital/Jefferson Health in Pennsylvania.


2019 Imaging Innovation Awards Winners

Winners Turn Good Ideas Into Outstanding Advancements
By Radiology Business Staff | Read Introduction

Alerting Busy Providers Without Needlessly Interrupting Their EHR Workflows
By UT Southwestern Medical Center at Dallas and Parkland Health & Hospital System | Read Case Study

Harnessing AI to ‘Make it Easier for Radiologists to Practice Better’
By Radiology Partners | Read Case Study

Taking Charge of Change to Standardize Care Across Many Sites
By Integrated Imaging Consultants | Read Case Study

Precisely Targeting Gadolinium for MS Patients Who Truly Need It
By the Department of Radiology at the Hospital of the University of Pennsylvania | Read Case Study