Electronic trigger-tools utilize electronically stored health data to generate clinical alerts that can improve the quality and efficiency of healthcare.
A recent report demonstrates that this can be successfully done. An automated trigger-tool scanned electronic health records to identify patients at risk of delayed cancer diagnosis. Clinicians were alerted to patients with a high-likelihood of cancer diagnosis who had not been appropriately followed-up. The tool was studied on approximately 300 000 patients over a 1-year period with abnormal test results that were suspicious for prostate and colo-rectal cancer.
To ensure that the electronic trigger-tool did not overwhelm healthcare providers with inappropriate alerts, the specificity of the trigger-tool was improved by including a broad set of information in the automated algorithm, modeled on the decision making process of experienced clinicians. In the study, this information included:
- Abnormal test results.
- Patient demographics (e.g. age).
- “Red flag” criteria (i.e. additional diagnostic information related to the diagnosis that either raised or decreased the probability of cancer).
- Clinical exclusion criteria (e.g. terminal illness, known cancer).
- Expected follow-up and timing.
The positive predictive power of the trigger for correctly identifying missed follow-up was between 60-70% for the different tests studied, indicating that the specificity criteria worked well. The tool was also clinically relevant. For patients with missed follow-up for prostate, 11% were found to have prostate cancer. For patients with missed follow-up for colo-rectal cancer 3.6% were found to have cancer and additional 8.9% had at least one neoplastic polyp.
We are slowly reaching a milestone in healthcare where large volumes of clinical data are accessible in databases and other electronic systems. We must ensure that this data is used to improve the quality and efficiency of patient care, as well as to deliver a return on investment for the healthcare providers that generate and record this data.
E-trigger tools can enable this, and can potentially be applied for a wide range of diagnoses, in many clinical settings, and incorporating a wider range of clinical data than is practical for clinicians to evaluate. E-trigger tools will be particularly useful for low frequency events that occur for large numbers of patients, where manual chart reviews would be impractical.
It is important that e-trigger tools do not overwhelm clinicians with inappropriate alerts. At the hospital where I work, I regularly receive automated notifications of potential drug interactions. Unfortunately, the majority of these alerts are clinically irrelevant because our pharmacy e-trigger tool cannot access the additional clinical information that I utilize while making treatment decisions. This lack of specificity means that the alerts are often distracting rather than helpful, potentially crossing the line from improving quality of care to compromising it.
Ultimately, e-trigger tools should be able to access most of the information that clinicians use to make decisions, as well as additional information that is useful but difficult or impractical for clinicians to quickly evaluate. Information that is stored in free-text fields, even within electronic health records, is currently inaccessible to most e-trigger tools. It can be expected that the ability of electronic systems to search free-text fields will improve in the future, allowing the “narrative” form of record keeping that most clinicians are familiar with to be accessible to electronic tools.
The ability to start utilizing large clinical data sets to off-load repetitive work from clinicians and to improve the quality of care has been demonstrated. It’s time to put these tools into use for the good of our patients.