When Dr. Jasmine Teng reviewed the medical charts of cancer patients at Peter Mac in Melbourne, she saw a pattern buried in plain sight—dozens of cases of immune-related colitis, a serious bowel inflammation triggered by life-saving immunotherapies, were being identified slowly, manually, and often too late. Now, her team has flipped the script with Australia’s first digital tool capable of rapidly flagging these cases using existing hospital data. Immune checkpoint inhibitors have revolutionized cancer care, offering hope to patients with melanoma, lung cancer, and more—but up to half of those treated face immune-related colitis, a side effect that can derail therapy or, in severe cases, become life-threatening. Until now, identifying affected patients meant sifting through reams of clinical notes, a process too slow for timely intervention or large-scale research.
That’s changing. Teng and her colleagues at the National Centre for Infections in Cancer (NCIC) and the Centre for Health Services Research in Cancer (CHSRC) have built a 'digital phenotype'—a computer algorithm trained to scan Electronic Medical Records and pinpoint patients with immune-related colitis with high accuracy. The tool, validated for the first time in Australian cancer patients and detailed in JCO Clinical Cancer Informatics, doesn’t just speed up diagnosis; it unlocks a new frontier in precision oncology. By automating case detection, it allows clinicians to study the true incidence of colitis, map complex treatment pathways, and, crucially, begin searching for biological markers that predict who is most at risk.
The implications ripple far beyond administrative efficiency. With this tool, researchers can now assemble large, well-defined patient cohorts in weeks, not years—accelerating studies into why some immune systems turn on the gut and others don’t. This could lead to preemptive strategies: adjusting immunotherapy doses, introducing protective agents, or monitoring high-risk patients more closely from day one. As Teng puts it, the digital phenotype doesn’t just react to illness—it helps anticipate it. And because it runs on data already collected during routine care, it can be scaled across hospitals without burdening staff. For a system under pressure, that’s transformative.
The tool’s success at Peter Mac sets a benchmark for digital health innovation in oncology. It’s a quiet revolution, powered not by flashy new machines but by smarter use of what’s already there. As immunotherapy use grows, so will the need to manage its risks—intelligently, equitably, and in real time. This algorithm is more than a detector; it’s a bridge to safer, more personalized cancer care. And it’s already showing the way.