Design: population-based registry cohort study with simulation against a random-mating scenario
Population: 4,244,585 Danish residents born 1969–2017; 549,541 parental couples and 822,209 offspring
Primary outcome: parental correlation for psychiatric and somatic disorders, and modelled effect on prevalence and heritability
Effect: overall partner correlation r = 0.28 (SE 0.003); schizophrenia r = 0.38, anorexia r = 0.11; prevalence inflated 1.8% and family-based heritability 12.5%

Danish national registry data covering 4,244,585 residents born between 1969 and 2017 identified 549,541 parental couples and 822,209 offspring, and measured how much partners resembled each other diagnostically. The overall correlation for psychiatric disorder between partners was 0.28. Within specific disorders it ranged from 0.38 for schizophrenia down to 0.11 for anorexia, and cross-disorder correlations ran from 0.01 to 0.26. Partners also resembled each other for somatic illness, but weakly (0.03 to 0.14), and that resemblance was almost entirely explained by their shared psychiatric comorbidity.

The analysis then compared observed prevalence and heritability against a simulated random-mating scenario. Non-random partnering inflated the prevalence of any psychiatric disorder by 1.8% and its family-based heritability by 12.5%.

Two things follow for clinic. First, when a patient with schizophrenia asks about their children, the familial risk figures in circulation are partly a product of this structure and not purely genetic transmission — the honest answer is that risk clusters in families for more than one reason. Second, when you take a family history, asking about one parent is not enough: the correlation of 0.28 means the other parent's history is not independent, and a service that assesses only the presenting parent will systematically underestimate what a child is growing up inside.

Take a family history from both sides, not the side that brought the patient in.
In perinatal and child services, ask about the partner's mental health as a matter of course.
Treat clustering in a family as a reason to offer the other parent an assessment, not only to counsel about genes.
When counselling about offspring risk, separate what is inherited from what is environmental — this study cannot tell them apart.

Why it matters

It changes what a familial risk figure means when you quote one to a patient asking about their children.

The statistics, in plain English

A correlation of 0.28 between partners is substantial for a trait nobody selects for consciously; 0.38 for schizophrenia is higher than most clinicians would guess. The standard errors are tiny because the sample is a whole country, so these estimates are precise — but precision is not the same as transportability, and Danish partnering patterns need not hold elsewhere. The 12.5% inflation of family-based heritability means a share of what twin and family studies attribute to genes is instead the structure of who has children with whom.