Background: Many missed and delayed cancer diagnoses result from breakdowns in communication and coordination of abnormal findings suspicious for cancer, which often first emerge in the primary care setting. Our previous work in the VA has shown that delays in the follow-up of abnormal test results persist despite reliable delivery of test results through the electronic health record. Methods to detect these delays and identify "high risk" patients are underdeveloped and need to be optimized for use within Patient Aligned Care Teams (PACTs). We conducted pilot work to determine whether the use of electronic queries, or "triggers," can proactively identify patients at risk of delayed cancer diagnosis. Triggers prompted review of selected medical records with evidence of possible care delays (e.g., a chart with no documented follow-up of an abnormal chest X-ray after 30 days). More than half the charts identified by the triggers were confirmed on chart review to have missed follow-up (positive predictive values [PPVs] >50%). However, the processes by which our team confirmed these delays and communicated them to providers were inefficient and resource intensive. Objectives: Building on our pilot work, we propose to develop and test an innovative automated surveillance intervention to improve timely diagnosis and follow-up of five common cancers in primary care practice (colorectal, prostate, lung, hepatocellular, and breast). Our methodology will use the VA Informatics and Computing Infrastructure (VINCI) to trigger medical records with evidence of potential delays in follow-up of abnormal test results. To guide our work, we will use an 8-dimension, socio-technical model built on principles from clinical informatics and human factors.
Our specific aims are to: 1) Evaluate the accuracy of a VINCI- based "real-time" automated surveillance system to identify patients at risk of missed or delayed diagnosis of 5 common cancers. 2) Establish how to integrate "real-time" surveillance and communication of information about at-risk patients into the point of PACT care through adoption of informatics and human factors engineering principles. 3) Evaluate effects of the automated surveillance intervention on timeliness of the diagnostic process and cost-effectiveness as compared with usual care. Methods: Study sites include facilities in VISN 12.
In Aim 1, we will use an iterative approach to develop and test algorithms to "trigger" records lacking documented follow-up action after pre-defined diagnostic clues for cancer. Data elements needed to operationalize our triggers already exist as part of the Corporate Data Warehouse. We will apply trigger algorithms to test cohorts, compare their output against manual chart reviews to confirm delays and use these data to modify the algorithms to improve trigger PPVs. The finalized triggers will be applied to validation cohorts to determine the final PPVs through the same methods.
In Aim 2 we will use interviews, task analysis, participatory design techniques, and usability testing to ensure that the automated intervention will fit within the workflow of real-world clinical practice. We will determine the technical requirements to transmit data to the PACTs, explore the best ways of communicating the information to the PACT team, and conduct usability testing to evaluate notification designs.
In Aim 3 we will conduct a cluster randomized controlled trial with VISN 12 PACTs randomly assigned to intervention or usual care. Intervention will consist of: 1) daily data extraction withn the VINCI platform to identify patients at risk of diagnostic delays;and 2) automated communication to PACT teams in VISN 12 about which of their patients are experiencing potential delays. Our outcomes are the median time in days from diagnostic clue to follow-up action (e.g., time to colonoscopy after a positive hemoccult) and the proportion of patients receiving appropriate and timely follow-up care. To determine cost effectiveness of the intervention, we will use a measure of incremental cost per additional delayed cancer diagnosis case averted. Our findings will provide important information on the effectiveness and value of automated interventions to identify and reduce cancer-related diagnostic delays.
Our study establishes automated methods for VA primary care settings to identify and potentially reduce diagnostic delays, a significant patient safety issue that has been understudied in comparison to other patient safety topics. The proposed intervention can serve as a model to mitigate communication and coordination breakdowns in the diagnosis of cancer and other diseases that are often first detected in primary care.