If you read that paper, they posit that a likely "cause" is that unmarried people tend to have cognitive impairments diagnosed later relative to when someone (i.e. a spouse) is living with them.