Where there are more dermatologists, melanoma is found earlier

Design

national ecological study linking State Cancer Profiles and Area Health Resource File data, multivariable analysis

Population

557 US National Cancer Institute health service areas, 2017-2021

Primary outcome

proportion of cutaneous melanoma diagnosed at regional or distant stage

Effect

versus no dermatologists, OR 0.89 for >0-1 per 100,000 (p=0.034), 0.89 for >1-2 (p=0.020), 0.84 for >2-4 (p<0.001), 0.80 for >4 (p<0.001)

Access to dermatology varies enormously, and stage at diagnosis is the single largest determinant of melanoma survival. This national ecological study linked State Cancer Profiles data with the Area Health Resource File across 557 US health service areas from 2017 to 2021, comparing the proportion of melanomas diagnosed at regional or distant stage against the number of dermatologists per 100,000 population.

The gradient was consistent. Against areas with no dermatologist, the odds of a melanoma being diagnosed late fell by 11% where there was up to one per 100,000 (OR 0.89), 11% for one to two, 16% for two to four (OR 0.84) and 20% above four per 100,000 (OR 0.80).

This is an ecological design and the authors say plainly that it cannot support causal inference. Areas with more dermatologists differ in income, insurance, education and much else, and areas with few cases were censored. It is also a US analysis, where dermatology access is mediated by insurance in ways that do not transfer. The value for a reader here is as an argument, not a finding: the case for teledermatology, for training primary care in dermoscopy, and for triage systems that get a pigmented lesion seen quickly rests on exactly this relationship, and in most of India the density figure is well below the lowest band studied.

• Read this as a service-level argument, not as evidence about any individual patient
• Prioritise fast-track pathways for pigmented lesions where specialist access is scarce
• Teledermatology triage is the practical lever where density cannot be changed quickly
• Train and equip primary care in dermoscopy rather than relying on referral alone
• Ecological associations of this kind routinely overstate what individual-level data would show

The statistics, in plain English

An odds ratio of 0.80 means one fifth lower odds of a late-stage diagnosis in the best-served areas compared with areas that have no dermatologist at all - a comparison between places, not between patients. This is the ecological fallacy in its classic form: a relationship that holds across regions need not hold for individuals within them, because the regions differ in everything else at the same time. Censoring areas with few cases removes exactly the sparsest, most underserved places, which will tend to flatten the gradient rather than exaggerate it.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.