Standardized criteria and automated alerts are effective tools in facilitating early detection and diagnosis of HF and CKD. EMR prompts can be implemented at various stages of patient care. For example, alerts may notify providers when a patient becomes eligible for a specific GDMT intervention or when new laboratory results indicate declining renal function, prompting timely treatment adjustments. By providing real-time information to all members of the care team, such alerts improve communication and coordination among clinicians, ensuring that everyone is aware of the latest patient updates and can work together to provide optimal care. The PROMPT-HF trial evaluated whether an EMR-based clinical decision support tool could enhance the initiation or optimization of GDMT in 1310 patients with HFrEF. Clinicians who received the automated alerts in the EMR system were more likely to prescribe additional classes of GDMT (adjusted RR 1.41 (1.03–1.93); p = 0.03) and to increase the dose of classes the patient was already receiving compared with those who did not receive alerts.[23] The trial showed that 80% of clinicians found the alerts to be “very helpful” and that a low-cost, scalable intervention can improve the quality of care for HF patients by prompting clinicians to follow the guidelines. This may lead to better long-term health outcomes and reduced hospitalizations.[23]

Similarly, the PROMPT-AHF trial examined whether EMR could increase GDMT prescriptions for patients hospitalized with acute HF.[24] The primary outcome of increased GDMT prescriptions at discharge occurred in 34% of patients in both the alert and no-alert groups, showing no significant difference (adjusted RR: 0.95 (0.81, 1.12), P = 0.99). However, patients in the alert group were more likely to have an increase in MRA prescriptions (adjusted RR: 1.54 (1.10, 2.16), P = 0.01).[24]

The BETTER CARE-HF trial found that an automated EMR alert doubled MRA prescribing in eligible HF patients with reduced ejection fraction (HFrEF) compared to usual care (29.6% vs. 11.7%; relative risk: 2.53; P < 0.0001), achieving an absolute improvement of 17%.[25] The alert also outperformed an automated message intervention (relative risk: 1.67; P = 0.002), underscoring the importance of timely delivery at the point of care. Overall, the findings show that well-designed EMR alerts can significantly increase prescriptions of life-saving therapies when implementation science is effectively applied.[25]

The REVEAL-HF study evaluated whether providing clinicians with 1-year mortality risk estimates for hospitalized heart failure patients would improve clinical decision-making and patient outcomes. The results showed no significant difference in the primary outcome, composite of 30-day all-cause readmission and 1-year mortality, between the alert and the usual care groups (38.9% vs 39.3%; P = 0.89), respectively.[26]

While much of the current evidence focuses on leveraging EMR-based alerts to optimize GDMT in HF, similar strategies are increasingly being explored for the management of CKD. For instance, Peralta et al. evaluated an electronic clinical decision support system (eCDSS) embedded in the EMR for CKD management across primary care practices. They found that providers who received automated eCDSS prompts were significantly more likely to recognize CKD (adjusted OR: 2.4; 95% CI: 1.7–3.3) and order appropriate laboratory monitoring (adjusted OR: 1.6; 95% CI: 1.2–2.1) compared with those in usual care.[27] Likewise, Sonoda et al. reported that EMR-based alerts guiding drug prescribing for patients with impaired kidney function reduced the rate of inappropriate dosing of renally cleared or nephrotoxic medications from 12.5% to 4.4% (relative risk reduction ≈65%), and improved adherence to safe prescribing recommendations by nearly 30%.[28]

Recently, there has been a significant push towards utilizing machine learning (ML) techniques to predict different stages of CKD. Based on data collected from medical records and registries for multiple variables including but not limited to (age, sex, hemoglobin, albumin, blood glucose, and serum creatinine), different ML algorithms, such as logistic regression, random forest, decision trees, and XGBoost, demonstrated high accuracy ranging from 85 to 97%, with sensitivity and specificity ranging from 56% to 80% and 95% to 99%, respectively, according to the CKD stage.[29, 30]

However, challenges remain in generalizing and validating ML models across different clinical settings. A clinical research or registry generally collects a wide range of patient information and biomarkers that may not be routinely collected in real-world clinical settings. Integrating complex prediction models into clinical practice also requires addressing interpretability and potential issues with generalizability to earn healthcare professionals’ trust.[31]

Additionally, EMR alerts can notify healthcare providers in real time about important events related to a patient’s care, such as when a patient becomes eligible for a specific intervention or treatment, or when new lab results or test results for a patient become available. By providing this real-time information to all members of the patient’s care team, EMR alerts can improve communication and coordination among the healthcare providers involved in the patient’s care. This ensures that everyone is aware of the latest updates and can work together cohesively to provide optimal care for the patient.[32]

As the use of EMR systems becomes more widespread, they may generate alerts or reminders that are not clinically relevant or may provide outdated information on medication dosages or interactions, encountering a challenge known as “alert fatigue.” Alert fatigue occurs when an excessive volume of prompts from EMRs and other systems, such as patient monitors, overwhelms caregivers. As a result, providers may become desensitized to these alerts, increasing the risk that important warnings are overlooked.[33]

Implementing an alert system differentiates alerts by severity, ensuring only high-severity alerts are interruptive, and applying human factors engineering principles can improve the design and usability of alerts. Hard-stop alerts in EMR systems can be effective tools for improving healthcare processes and outcomes, with studies showing an 88% improvement in process measures and a 79% improvement in health outcomes.[34] Hard-stop alerts should be reserved for critical safety issues, include third-party override options, and be designed with substantial user input.

Regularly auditing and optimizing alerts, involving clinicians in the design process, and allowing customization options can enhance the system’s effectiveness and help manage alert fatigue.[35, 36]