Coding infections in primary careBMJ 2019; 367 doi: https://doi.org/10.1136/bmj.l6816 (Published 11 December 2019) Cite this as: BMJ 2019;367:l6816
- Alastair D Hay, professor of primary care
- Centre for Academic Primary Care, Bristol Medical School: Population Health Sciences, University of Bristol, Bristol BS8 2PS, UK
Antibiotics are one of the most commonly prescribed medicines in primary care worldwide, but as many as 23% in the United Kingdom1 and 25% in the United States2 are prescribed inappropriately. Governments concerned about the threat of antibiotic resistance have repeatedly called for improved stewardship to preserve antibiotic effectiveness for future generations.345
Reducing inappropriate prescribing requires behaviour change by clinicians. Providing peer-referenced prescribing data to individual clinicians is a proven method supporting behaviour change6 because it provides a strong counter to those who justify their prescribing by stating they deal with a different group of patients—for example, older patients with more long term chronic conditions, or more patients with acute problems. A peer-referenced individualised prescribing feedback system would work optimally if it met the following three criteria.
The first is that a diagnostic code is recorded every time an antibiotic is prescribed. Surprisingly, a substantial proportion of antibiotics prescribed in primary care are issued without a diagnosis code in the medical record. Strong evidence for this is provided in the linked paper by Ray and colleagues (doi:10.1136/bmj.l6461), which shows that 18% of the antibiotics prescribed to 130 million Americans in 2015 (collected from nearly one billion ambulatory care visits) had no coded indication.2 A recent UK study found an even higher percentage (36%) of antibiotic prescriptions without a coded diagnosis between 2013 and 2015.1
This could reflect diagnostic uncertainty. Most patients do not present with neatly differentiated symptoms that can be converted into a conclusive diagnosis of infection, so using a definitive diagnostic code would not reflect reality, even when an antibiotic is considered necessary. Worse, it could result in harm because subsequent consultations (especially with a different clinician) might put too much reliance on the original diagnosis and discourage reassessment. Therefore, improving the completeness of diagnostic coding could be achieved by increasing the use of “provisional” diagnostic codes (such as “suspected urinary tract infection”).
Ray and colleagues also show that longer consultations are more likely than shorter consultations to result in an antibiotic prescription without indication, perhaps reflecting more complex patients, “coding fatigue,” or insufficient consultation time.2
The second criterion is that all infections should be coded, not just those for which an antibiotic is being prescribed. The reason may not be immediately intuitive, but consider how clinicians might determine if their prescribing is appropriate using existing data. They could conduct an audit to compare prescribing against quality indicators, such as those published in 2011.7 Developed by a panel of European experts, these provide acceptable percentage ranges for prescribing, by infection. For example, they suggest no more than 30% of adults under 75 years with acute bronchitis should receive an oral antibiotic. Recommended percentages are also given for acute upper respiratory tract infection (≤20%), acute tonsillitis (≤20%), acute or chronic sinusitis (≤20%), acute otitis media (≤20%), acute urinary tract infection in women (≥80%), and pneumonia in adults younger than 65 (≥90%). Clearly, all infections must be coded to generate these percentages.
Having ensured that a diagnostic code is used every time an antibiotic is prescribed, and that all infections are coded, the third criterion would be to use a global measure of illness severity (such as mild, moderate, or severe) with each diagnostic code. Clinicians could then evaluate if their patients have more severe illness compared with other patients.
Using the above criteria, an individualised feedback system could provide clinicians with data on their use of diagnostic codes and antibiotics compared with peers. It would also allow clinicians to monitor their use of codes for more severe infections, such as tonsillitis and pneumonia, which might otherwise be used to justify prescribing decisions,8 as well as the proportion of patients with each condition for whom they prescribe. Those with responsibility for antimicrobial stewardship might wish to work with electronic health record providers and clinicians to encourage this more sophisticated diagnostic coding.
Of course individualised feedback as described is only one of a raft of antimicrobial stewardship strategies needed to improve prescribing, none of which will work in isolation. Others include better infection control, vaccination, and improved diagnostic precision, but the incentive for improving diagnostic coding is to provide information that can be used by clinicians to help them reflect and refine their prescribing behaviour.
I thank Matthew Thompson, Department of Family Medicine, University of Washington, for helpful contributions to earlier versions of this manuscript.
Competing interests: The BMJ has judged that there are no disqualifying financial ties to commercial companies.
The author declares the following other interests: I am a NIHR senior investigator for the National Institute for Health Research (NIHR). The views and opinions expressed are those of the author and not necessarily those of the National Health Service, the NIHR, or the Department of Health and Social Care.
The BMJ policy on financial interests is here: https://www.bmj.com/sites/default/files/attachments/resources/2016/03/16-current-bmj-education-coi-form.pdf.
Provenance and peer review: Commissioned; not peer reviewed.