A simple medical code is reshaping our fight against superbugs.
When the world first grappled with COVID-19, healthcare systems faced an unprecedented challenge. In urgent care clinics, where patients often arrive with coughs, fevers, and breathing difficulties, doctors confronted a critical question: how to treat these mysterious symptoms while avoiding the overprescription that fuels antibiotic-resistant bacteria. The answer, it turns out, may lie in an unexpected place—the international classification system used to code every patient visit.
The COVID-19 pandemic created a paradox in antibiotic prescribing. Initially, public health measures like masking and social distancing caused a dramatic drop in common respiratory infections, leading to fewer antibiotic prescriptions in urgent care settings4 . One study at two academic urgent care clinics showed antibiotic prescribing dropped from 17% of visits before COVID-19 to just 11% during the early pandemic months4 .
of urgent care visits
of urgent care visits
This positive trend, however, proved temporary. As pandemic restrictions eased, many countries saw a troubling rebound in antibiotic use. Research from Singapore showed antibiotic prescription rates climbed from 3.5% in 2022 to 4.0% in 2023—a 9.5% relative increase—driven largely by respiratory conditions2 . Similarly, European data revealed substantial jumps in antibiotic prescriptions across six countries in 2022, with increases ranging from 12.0% in France to 39.3% in Belgium3 .
This rollercoaster of prescribing patterns revealed a critical gap in our understanding: how do we accurately identify which patients truly have COVID-19 versus other infections in urgent care settings? The answer has profound implications for antibiotic stewardship—the effort to ensure antibiotics are used appropriately to combat the growing threat of antimicrobial resistance.
In healthcare, every diagnosis is assigned a code from the International Classification of Diseases (ICD) system. These codes help track health trends, determine payments, and—crucially for researchers—identify which conditions are being treated. When COVID-19 emerged, new ICD-10 codes were created to identify these cases. But how these codes are applied in practice varies significantly, creating challenges for understanding true prescribing patterns1 6 .
The central problem is this: different healthcare providers may use different COVID-19 codes for similar patient presentations. Some might use specific codes for confirmed COVID-19, while others might use broader codes for "COVID-19-like illness" or related symptoms. This variability makes it difficult to accurately track how often antibiotics are prescribed for COVID-19 patients—a crucial metric since antibiotics are ineffective against viruses like SARS-CoV-2.
| Code Type | Description | Sensitivity | Specificity | Antibiotic Prescribing Rate |
|---|---|---|---|---|
| COVID-19 specific codes | Codes for confirmed COVID-19 | Lower | Higher | Very low (<1%)4 |
| COVID-19-like illness (CLI) codes | Codes for suspected COVID-19 or related symptoms | Higher | Lower | Moderate6 |
| Combined definitions | Including both COVID-19 and CLI codes | Highest | Lowest | Varies6 |
To understand how COVID-19 code definitions impact antibiotic prescribing rates, researchers at Stanford University conducted a detailed analysis across two academic urgent care clinics1 . Their investigation provides fascinating insights into this complex issue.
The research team employed a retrospective approach, examining all telemedicine and office visits at two urgent care clinics from January 2020 through September 2021. They extracted ICD-10 encounter codes alongside SARS-CoV-2 testing data from electronic medical records1 .
The key innovation of their methodology was comparing two different approaches to identifying COVID-19 cases:
The researchers then calculated the sensitivity of COVID-19 ICD-10 codes against a positive NAAT test and determined antibiotic prescribing rates for each group.
The study uncovered crucial limitations in relying solely on billing codes to identify COVID-19 encounters and assess antibiotic prescribing1 . The sensitivity of COVID-19 ICD-10 codes—their ability to correctly identify true COVID-19 cases—varied significantly throughout the pandemic6 .
Most strikingly, the research revealed that broader code definitions (including both COVID-19 and COVID-19-like illness codes) provided the highest sensitivity but the lowest specificity6 . This tension between accurately capturing all COVID-19 cases versus correctly excluding non-COVID cases directly impacted calculated antibiotic prescribing rates.
| Encounter Type | Pre-COVID-19 Antibiotic Prescribing Rate | COVID-19 Period Antibiotic Prescribing Rate | Change |
|---|---|---|---|
| All respiratory encounters | 17% | 11% | -35%4 |
| Tier 3 respiratory encounters (antibiotics almost never indicated) | 16% | 5% | -69%4 |
| COVID-19-coded encounters | N/A | <1% | N/A4 |
| Non-COVID-19 respiratory encounters | 17% | 14% | -18%4 |
Perhaps the most encouraging finding was that COVID-19-coded encounters were rarely associated with antibiotic prescriptions—less than 1% in the Stanford study4 . This suggests that when clinicians identified patients as having COVID-19, they generally avoided inappropriate antibiotic use.
Understanding how studies like the Stanford analysis are conducted reveals the complexity of antibiotic stewardship research. Here are the key tools and methods researchers use:
Standardized classification systems that categorize diseases and health problems, essential for identifying specific patient populations4
A classification system categorizing encounters based on whether antibiotics are almost always (Tier 1), sometimes (Tier 2), or almost never (Tier 3) indicated4
| Metric | Pre-Implementation | Post-Implementation | Change |
|---|---|---|---|
| Day of therapy/outpatient-day | 0.85 ± 0.41 | 0.72 ± 0.39 | -15%5 |
| Favorable clinical outcomes | 42.7% | 62.1% | +20%5 |
| Target antibiotic consumption | Higher | Lower | Significant reduction5 |
Post-Prescription Authorization (PPA) is a stewardship strategy where certain antibiotics can be freely prescribed initially but require specific approval for continued use beyond a set time frame5 .
The COVID-19 pandemic served as a natural experiment for antibiotic stewardship, revealing both challenges and opportunities4 . The variability in how COVID-19 encounters are defined and coded highlights the need for standardized approaches in both clinical practice and research.
Moving forward, antimicrobial stewardship programs must account for the unique impact of the pandemic and carefully consider how they stratify respiratory visits to avoid erroneous trends when comparing to pre-COVID-19 data4 . The low antibiotic prescribing rates for clearly identified COVID-19 cases suggest that improved diagnostic accuracy can reduce inappropriate antibiotic use.
As one study noted, "The big question remains as to whether the lower antibiotic prescribing rates will be sustained beyond the COVID-19 pandemic"4 . The answer may depend on building upon lessons learned during the pandemic—including how we define, track, and respond to respiratory illnesses in urgent care settings.
The precision of a medical code might seem like an administrative technicality, but as the COVID-19 pandemic revealed, these details profoundly influence clinical decisions that affect us all—determining when antibiotics are prescribed, and when they're appropriately withheld in the ongoing battle against antimicrobial resistance.