[T]he risk of an autism diagnosis in a second-born child rose more than three-fold when the child was conceived within 12 months of the birth of the first baby . . . second-borns conceived between 12 and 23 months after a first child was born had twice the risk of being diagnosed with autism compared to babies conceived a full three years after an older sibling was born. . . . For the new study, researchers scrutinized birth records from 662,730 second-born children from California, all of whom were born between 1992 and 2002 and none of whom had an older sibling with autism. By age 6, 3,137 of the second-borns had received a diagnosis of autism, according to data provided by California’s Department of Developmental Services. Of those, 2,747 occurred in children born less than 36 months after their siblings[.]
Similar risk factors have been identified for schizophrenia, which like autism, is also associated independently with advanced paternal age, suggesting again that both conditions may have a similar disease model for an epidemiology perspective. Given these findings, I'd be willing to guess that bipolar disorder, which also seems to have a similar disease model, also has an elevated "Irish twin" risk.
Given the fact that it isn't uncommon for autistic children to have siblings with subclinical autistic traits, a factor that tends to understate the degree to which autism is hereditary, the other possibility that comes to mind is a genes x environment interaction. It may be that many of the second children diagnosed with autism actually did share genes with an older system that pre-disposed each to autism, but that in utero stresses associated with a swift second pregnancy caused the autism-like symptoms to cross the line from sub-clinical to clinically diagnosable as a result, while autism traits expressed only at a subclinical level in the first child of the pair.