Where Primary Care Doctors Are Scarce, More Cancers Are Found Late

In the California and Texas counties with the fewest primary care physicians, 24.9% of adults diagnosed with one of the seven cancers the study tracked already had metastatic disease when it was found. In the counties with the most, the share was 19.7%.

Those figures come from a study published Oct. 19 in Cancer Medicine by researchers at UT Southwestern Medical Center. They pulled 708,451 adults diagnosed with breast, colorectal, liver, lung, ovarian, pancreatic or prostate cancer from 2015 to 2019 out of the two states’ cancer registries. Then they matched each patient to how many primary care doctors practiced in their county.

The link held after adjustment. It also held among the patients who arguably need a dense supply of doctors least: privately insured adults of working age.

How the study counted doctors

The exposure was county-level primary care physician density, measured per 3,500 residents. The authors picked that unit on purpose. Fewer than one primary care physician per 3,500 people is a criterion for a federal Health Professional Shortage Area designation, so the scale lines up with a threshold policymakers already use.

The physician counts come from HRSA’s Area Health Resources Files, sourced from the 2019 American Medical Association Masterfile. Only MDs and DOs practicing primary care were counted. Nurse practitioners and physician assistants were not.

After adjusting for age, sex, race and ethnicity, insurance, county poverty, rurality and state, each additional primary care physician per 3,500 residents went with 3% lower odds of metastatic cancer at diagnosis, an odds ratio of 0.97. Among privately insured patients aged 18 to 64, the drop was 6% per added physician. Among Medicare patients 65 and older, it was 2%.

Insurance status carried much larger odds ratios than density did. Compared with privately insured patients, uninsured patients had 2.81 times the odds of metastatic disease at diagnosis. Medicaid patients had 2.10 times the odds.

One doctor per 3,500 may not be enough

Sorted into bands, the counties didn’t fall on a straight line. Across all seven cancers combined, counties with 4 or more physicians per 3,500 people showed 13% lower odds than counties under the shortage line. The first step up was small, though. Counties with 1 to 1.9 physicians had an odds ratio of 0.96 against shortage counties for all cancers combined. Split by cancer type, the authors report no statistically significant difference at that level for six of the seven cancers.

They read that as a problem with the federal line. “This suggests that the HPSA threshold of 1 PCP may be inadequate,” they wrote, adding that a higher density “could possibly act as a mechanism to meaningfully lower metastatic cancer rates at diagnosis.”

Prostate cancer ran the other way. In several models, higher density went with higher odds of metastatic prostate cancer at diagnosis. The authors point to the 2012 change in U.S. Preventive Services Task Force screening guidance as one possible explanation, citing earlier work suggesting that areas dense with primary care followed it more closely and screened less.

Why would density matter for cancers that have no screening test? The authors cite an estimate that only 14% of U.S. cancers are diagnosed following a screening test. For the rest, someone has to hear the symptom, order the workup and make the referral. The paper suggests the association may reflect both pathways: more screening where it exists, and earlier recognition and referral of patients with symptoms.

What a county count can’t see

This is where a DPC reader should slow down. The study counts physicians as heads. A doctor with a small membership panel and a doctor with a full insurance-based panel each count as one.

The American Academy of Family Physicians puts the average DPC panel at about 413 patients. As simple arithmetic, a county sitting right at the shortage line, one physician per 3,500 residents, would need about eight and a half DPC physicians at that average to hold every resident on a panel. The paper doesn’t say which of its counted physicians ran membership practices.

DPC’s case for itself is access. AAFP reports that 99% of DPC practices offer same-day appointments, and timely evaluation of symptoms is one of the mechanisms the authors propose. You might reasonably expect that a patient who can see their doctor the same day gets a worrying symptom checked sooner. This study doesn’t test that, and it was never designed to.

The open question

The authors are blunt about their limits. The design is cross-sectional, so causality cannot be inferred, and they call the findings “hypothesis-generating.” Density was measured at the county level, and patients may get care outside their county. Advanced practice providers weren’t counted. The data cover two states, and the physician counts are from 2019.

For DPC, the gap that matters is narrower. When a physician leaves an insurance-based practice to open a membership practice in the same county, the study’s measure doesn’t move. The doctor still counts as one. What changes is how many residents that doctor can see quickly, and how many have to find someone else.

Whether that trade raises or lowers the odds of a late diagnosis for the county as a whole is a question this data can’t answer. Answering it would take registry cases linked to the practice model of each patient’s own primary care doctor, and this paper doesn’t have that link.