How AI supports antimicrobial stewardship in primary care
Published on 2 September 2026
At National University Polyclinics (NUP), AI is helping clinicians identify why antibiotics are prescribed, giving antimicrobial stewardship teams clearer insights to guide intervention.
At a glance
- AI helps NUP to better understand the clinical indications behind antibiotic prescriptions.
- Clearer attribution allows stewardship teams to detect recurring inappropriate antibiotic prescriptions earlier.
- Embedded prescribing tools and real-time dashboards support more appropriate antibiotic use in primary care.
Antibiotics are routinely prescribed in primary care, but each prescribing decision carries implications not only for patient outcomes and potential side effects, but also for the growing global challenge of antimicrobial resistance.
At NUP, AI is being used to improve how prescribing decisions are interpreted, reviewed and acted on.
One longstanding challenge lies in determining the exact reason an antibiotic was prescribed, when a patient presents with multiple conditions during the same consultation. At the National University Health System (NUHS) Innovation Summit 2026, the NUP team shared how AI and data analytic tools can close that gap, giving stewardship teams a clearer picture of prescribing trends and where targeted interventions may be needed.
Better attribution makes prescribing data more meaningful
Effective antimicrobial stewardship starts with understanding not just how many antibiotics are being prescribed, but also why they are being prescribed.
With clearer attribution, stewardship teams can move beyond simply tracking antibiotic use to understanding what drives prescribing patterns. This includes pinpointing the conditions antibiotics are most frequently prescribed for, where diagnostic labelling and prescribing practices may not align with guidelines, and whether certain antibiotics are being used more frequently than expected.
Primary care records do not always hold definitive answers. A patient may present with respiratory symptoms, urinary complaints and skin issues in the same consultation. When multiple diagnoses are recorded during the same consultation, the reason for prescribing an antibiotic may not be clear.
When prescriptions can be linked more accurately to specific clinical conditions, the data becomes considerably more meaningful. Audits can be more targeted, prescribing variations can be assessed in context, and interventions can be directed where they are likely to have the greatest impact.
Dr Sky Koh, Family Physician, NUP and Adjunct Lecturer, NUS Yong Loo Lin School of Medicine (NUS Medicine).
AI converts clinical notes into actionable insights
Much of this information is captured in clinicians’ free-text notes rather than structured diagnosis fields, making it difficult to analyse at scale. AI bridges this gap by trawling these notes and identifying the most likely clinical indication associated with each antibiotic prescription.
Once converted into structured data, the information can be reviewed through dashboards, making it easier to interpret prescribing patterns and identify recurring gaps. Research co-led by Dr Koh has found that more than half of antibiotic prescriptions for uncomplicated urinary tract infections were inappropriate.
On the other hand, around one in three patients did not receive recommended treatment for acne care, as antibiotics were commonly prescribed alone without the recommended accompanying non-antibiotic creams.
In a recent study on antibiotic appropriateness for upper respiratory infections, it was found that patients with antibiotic allergies were also far more likely to be prescribed Watch antibiotics, which are broader-spectrum antibiotics with a higher potential to drive resistance.
Prescribing decisions are often shaped by an interplay of factors, including patients presenting with multiple symptoms during a single consultation, inconsistent documentation, antibiotic allergy labels and variations in clinical workflows.
Why indication mapping matters Patients may receive multiple diagnoses during a single consultation. Indication mapping helps determine which condition most likely prompted an antibiotic prescription.
|
Better insights support better prescribing
Clearer data is only part of the answer. Effective antimicrobial stewardship also entails supporting clinicians at the point of care.
At NUP, a piloted preference list customisation within the Epic electronic medical record system helps clinicians prescribe antibiotics more appropriately for urinary tract infections. The embedded tool provides guidance on the recommended antibiotic, dose, duration and frequency, in line with guidelines meted out by the Ministry of Health’s Agency for Care Effectiveness.
Results from the pilot demonstrated measurable improvement, achieving prescription accuracy of 86 per cent for urinary tract infections.
Embedding evidence-based guidance into clinical workflows makes recommended treatment options easier to access and encourages more consistent prescribing, particularly in busy primary care settings.
Looking ahead, Dr Koh will work with the NUP Clinical Informatics team to develop customised order panels in the Epic electronic medical record system for upper respiratory infections, helping clinicians make appropriate prescribing decisions more easily and consistently.
Continuous review strengthens stewardship over time
Stewardship efforts are most effective when prescribing patterns are reviewed on an ongoing basis. NUP’s AI-powered dashboards help track antimicrobial utilisation in real time and support reporting to regulatory authorities. The dashboards are actively monitored by the Antimicrobial Stewardship Programme team to identify prescribing gaps and opportunities for improvement.
Annual deep-dive audits have been conducted across various conditions, including urinary tract infections, acne and upper respiratory infections. Findings from those audits have identified prescribing gaps in primary care and informed subsequent improvement efforts.
Real-time review enables stewardship teams to act quickly, tailoring interventions to the gaps that need attention. Primary care remains at the forefront of antimicrobial stewardship, given the high volume of antibiotic prescriptions in these settings. Initiatives presented at the NUHS Innovation Summit 2026 show how AI can support antimicrobial stewardship in a practical way: by clarifying indications, revealing gaps, supporting better prescribing, and sustaining continuous review.
In consultation with Dr Sky Koh, Family Physician, NUP and Adjunct Lecturer, NUS Yong Loo Lin School of Medicine (NUS Medicine).
Related Articles
