It useful when comparing (and only when comparing) pregnant with non pregnant, especially with others criteria (typically, age).

The covariance between pregnancy and cause of death is what you want to seek, the only thing that would make that data unusable is a (very) common cause of death covariant with pregnancy, but unrelated to pregnancies. Like, for some reason, pregnant women would start reckless activities, but this behavior change would not be caused by pregnancy, either for physical or social reason.

It seems like that data would be very noisy data. Criteria for age would be one obvious thing, but it would also include social status, social support networks, hours on the road, frequency of visiting the doctor, distance to doctor and home, stress and access to leisure time, cultural factors, economical factors and many more.

The conclusions would also be conditions on such factors and how the interact with each other.