Background Failure to follow-up abnormal test results can lead to patient harm.

Objectives We created and validated electronic trigger algorithms that analyzed electronic health record (EHR) data from a large Veterans Affairs (VA) network to identify patients with potential delays in diagnostic evaluation for multiple cancers.

Methods We developed five trigger algorithms to detect delays in diagnostic evaluation of possible bladder, breast, colorectal, hepatocellular, and lung cancer. Each used structured clinical data to identify patient records with red-flags (abnormal test results warranting further diagnostic evaluation). Red-flags included high-grade hematuria (>50 red blood cells/high powered field; bladder cancer trigger), abnormal mammograms (breast cancer trigger), iron deficiency anemia or positive fecal immunochemical tests (colorectal cancer trigger), elevated alpha-fetoprotein (hepatocellular trigger), or chest imaging flagged as suspicious for malignancy (lung cancer trigger). Algorithms excluded records where follow-up was unnecessary (e.g., hospice patient) and records where follow-up was documented within 30 (lung cancer trigger) or 60 days (all others). We validated triggers by applying them retrospectively to EHR data (see table 1 for timeframes and sample sizes).

Results The five triggers yielded PPVs ranging from 56.0–82.3%, NPVs ranging from 88.0–98.0%, sensitivity from 64.1–91.7%, and specificity from 81.1–96.5% (see table 1). We estimated that these triggers have the potential to identify 1192 diagnostic errors in the VA network studied per year.

Conclusions Our triggers have potential to identify large numbers of patients experiencing delays in diagnostic evaluation. Implementing prospective electronic trigger-based measurement systems using these algorithms could support health systems in reducing delays in delays in cancer diagnosis.