Life and Death and Measurement Error


Recently CBC Radio’s The Current broadcast a program on the debate over the treatment of premature babies. In it, the story of Dominica, who was born at 22 weeks and 6 days, is used to explore the wider considerations of how to treat babies born that early. According to the program, standard treatment in North America is to consider babies at 22 weeks “unviable” and thus perhaps not offered treatment or available.

However, a new study in the New England Journal of Medicine has found that a small margin of babies born at 22 weeks can survive if given medical treatment as Dominica was; that has many asking whether medical associations should change their advice.

Dominica is now 4 years old despite the odds and is apparently healthy. But the story illustrates, in addition to the significant moral, emotional, medical and public policy considerations, the implications of simple measurement, specifically, measuring the length of gestation of a baby. The Canadian Pediatric Society guidelines indicates that at 22 weeks “babies do not survive,” at 23 weeks survival is deemed possible although with a degree of pessimism, at 24 weeks the parents have “discretion” and can request intervention and at 25 weeks, the guidelines are to intervene.

Along with variations in each individual case and in the judgment of each medical professional, there is variation in measuring gestation length. The point of conception is itself often rounded up or down by several days or a couple of weeks. Given the importance of even a few days of gestation length in making decisions, the number of days of gestation are indexed up at midnight, which could shorten the actual gestation period by as much as a day (if, for example, conception was actually at 1:00 am).

Of course decision are not generally made on a hard and fast set of rules around gestation period, but given the spectrum of decisions of the 14 days from 22 weeks to 24 weeks, the imprecision of the estimation of gestation is arguably an important factor, if, for example, a case had an actual period of 23 weeks and 6 days when due to measurement error the case is considered as “22 weeks and 1 day.”

The program is found here: