A new study finds that adding social determinants of health to genetic models improves disease risk prediction across six common conditions. The findings draw on data from over 200,000 participants.
A genetic risk score alone does not tell the whole story. A study published in the American Journal of Human Genetics suggests that where and how a person lives — their income, education, housing, access to food — can sharpen disease risk predictions in ways that DNA alone cannot.
Biji et al. developed a method to quantify social determinants of health and fold them into standard genetic risk models. To do this, the researchers applied a statistical technique called multiple correspondence analysis to survey data from the All of Us Research Program, a national biobank maintained by the US National Institutes of Health.
What the study measured
The All of Us dataset captures responses from participants across the United States, including questions about neighbourhood conditions, employment, and healthcare access. From this, the researchers generated what they describe as "embeddings" — compressed numerical representations of each participant's social profile.
When added alongside polygenic risk scores, those social embeddings consistently improved risk prediction for six common complex diseases, according to the study. The researchers did not specify which six conditions in the abstract, but the diseases studied fall within the category of common, multi-factor illnesses — conditions shaped by both biology and environment.
Why this reaches beyond genetics
Polygenic risk scores have become a standard tool in precision medicine. They aggregate thousands of small genetic variants to estimate a person's inherited likelihood of developing a given disease. But the study found that genetic risk scores, used alone, leave meaningful predictive signal on the table.
Social determinants of health — long studied in epidemiology and public health — describe the conditions in which people are born, grow, work, and age. This study offers a scalable computational method to encode those conditions numerically, making them compatible with the quantitative frameworks used in genetic research, the authors reported.
The approach is described as scalable, meaning it does not require bespoke data collection for each new study. Researchers can apply the same method to any biobank that includes comparable survey instruments, the paper indicated.
For communities that face compounding disadvantages — including many people with albinism, who navigate higher healthcare costs, limited dermatology access, and barriers to specialist eye care — research that formally accounts for social context carries particular weight. Models that ignore lived conditions risk underestimating risk for those least served by existing health infrastructure.
The study does not address albinism directly. Its significance here lies in the methodological precedent: that social circumstance is a measurable, includable variable, not a background footnote.
Keywords
Core topics and entities mentioned in this summary.