Predicting Readmissions for Asthma and COPD: How AI Can Help

By Campion Quinn, MD

Artificial intelligence is taking healthcare by storm—and it’s not just hype. A new study by Lopez et al. [1]shows how AI can predict hospital readmissions for asthma and chronic obstructive pulmonary disease (COPD), which account for a significant share of healthcare costs and suffering. This exciting research demonstrates how AI can sift through mountains of patient data to help clinicians identify which patients might need extra care after leaving the hospital.

A New Way to Predict Readmissions

Asthma and COPD are tricky to manage. Both diseases can cause flare-ups that land patients in the hospital. Some patients, unfortunately, return for additional care within weeks. Predicting who’s likely to come back has been challenging—until now.

Using electronic health records (EHRs), the researchers applied several AI models, including one called a multilayer perceptron (MLP). This deep learning model outperformed four machine learning techniques in identifying high-risk patients. With a sensitivity and specificity of 79%, the MLP model balanced catching actual risks and minimizing false alarms.

What the Data Tells Us

The study uncovered fascinating patterns. Patients with high white blood cell counts and certain chronic conditions, like heart failure, were more likely to be readmitted. Inhaled medications administered during the initial hospitalization—such as corticosteroids and beta-agonists—also affected the predictions.

These insights suggest that AI can go beyond surface-level symptoms to uncover underlying risk factors. That’s the magic of deep learning—it identifies patterns that might not be obvious, even to experienced clinicians.

Addressing Healthcare Disparities

The study also highlighted some concerning disparities. Black and Hispanic patients were more likely to be readmitted, reflecting broader inequities in healthcare. While predictive algorithms like this can help address these gaps, they need careful monitoring to avoid reinforcing existing biases.

Bringing AI into Clinical Practice

So, what does this mean for practicing physicians? Predictive models like the MLP can support clinical decision-making by flagging patients who may need closer follow-ups or additional care. But for AI to truly make a difference, it must be integrated thoughtfully into clinical workflows.

Physicians don’t need to become data scientists, but collaboration with experts in machine learning can ensure these tools are effective, transparent, and bias-free.

Conclusion

The future of healthcare lies in tools like AI that complement—not replace—clinical judgment. This study shows how predictive models can help physicians proactively manage chronic conditions like asthma and COPD. With AI by your side, you’ll have better insights to guide patient care—and ultimately, better outcomes for those who need it most.

 [1] Lopez, K., Li, H., Lipkin-Moore, Z. et al. Deep learning prediction of hospital readmissions for asthma and COPD. Respir Res 24, 311 (2023). https://doi.org/10.1186/s12931-023-02628-7