Odds ratios IIBMJ 2010; 341 doi: https://doi.org/10.1136/bmj.c4971 (Published 15 September 2010) Cite this as: BMJ 2010;341:c4971
- Philip Sedgwick, senior lecturer in medical statistics
Researchers conducted a case-control study to investigate whether sleeping environment was associated with sudden infant death syndrome.1 The researchers aimed to include all cases of sudden unexpected death for infants aged between 7 and 364 days from five regions of England with a combined population of more than 17 million people. A total of 325 cases were identified through registers. For every case, four controls matched for age, locality, and time of sleep were included. Interviews were undertaken with the parents of all cases and controls to ascertain past sleeping environment.
Infant’s typical sleeping environment in relation to their parents was classified as usual room sleeper (shared room but not bed), usual solitary sleeper (slept in room separate from parents), or usual bed sharer (shared bed with parents for more than two nights per week). The results are presented in the table, which shows the odds ratios of sudden unexpected death for the categories of the infants’ usual sleeping environment.⇓
Which of the following statements, if any, are true?
a) The case-control study is population based
b) Usual room sharers are the reference category
c) The odds ratio of sudden infant death syndrome for usual room sharers was equal to 1.0
d) Usual bed sharers were 2.99 times as likely to experience sudden infant death syndrome than usual room sharers
Answers a, b, and d are all true, whereas c is false.
Case-control studies, as described in a previous statistical question, are observational and retrospective in design.2 3 The aim in this study was to identify potential risk factors for sudden infant death syndrome. Two groups of subjects were selected on the basis of outcome: one group of infants recorded as experiencing sudden infant death (cases) and one group of otherwise healthy controls. Past sleeping environment was compared between groups to ascertain whether it was associated with an increased risk or a reduced risk of sudden infant death syndrome.
The case-control study was population based (a is true). Population based case-control studies include all cases in the population during a given time period, or a sample of these cases. The authors of this study selected cases from five regions in England that had a combined population of more than 17 million people. Controls were also selected from these regions, chosen from registers of all subjects in the population. Case-control studies are typically hospital based—cases are identified from a hospital clinic for convenience and controls are frequently selected from another clinic at the same hospital. Controls in a population based case-control study will tend to be more representative of otherwise healthy individuals in the population than those in a hospital based study, which allows a more accurate comparison between cases and controls.
Odds ratios are the usual way of analysing case-control studies to ascertain whether exposure to a treatment or lifestyle choice is associated with a disease or outcome.4 When two possible exposures to the treatment or lifestyle choice exist, absent and present, an odds ratio comparing the odds of disease if treatment or lifestyle is present compared with the odds of disease if treatment or lifestyle is absent is typically presented.
The researchers investigated whether the infant’s sleeping environment was associated with sudden infant death syndrome. The proposed risk factor of sleeping environment had three categories—usual room sleeper, usual solitary sleeper, and usual bed sharer. One of these categories—usual room sleeper—was selected as the baseline or reference category against which the other two categories were compared (b is true). The 1.00 in the table is the convention used to represent the reference category—it does not represent an odds ratio (c is false). The categories of solitary sleeper and bed sharer were in turn each compared against the reference category and the separate odds ratios are presented in the table.
The odds ratio for sudden infant death syndrome for usual bed sharers compared with usual room sharers was 2.99. Therefore, the odds of sudden infant death syndrome for usual bed sharers were 2.99 times those of usual room sharers (d is true). Infants who were usual bed sharers were nearly three times as likely—that is, almost two times more likely—to experience sudden unexpected death compared with usual room sharers. The 95% confidence interval did not include unity; therefore, the odds of sudden unexpected death for usual bed sharers were statistically significantly different from those of usual room sharers at the 5% level of significance.5
The odds ratio for sudden unexpected death for infants who were usual solitary sleepers compared with those who were usual room sharers was 0.88. Although usual solitary sleepers were less likely to experience sudden infant death syndrome, the 95% confidence interval included unity so the odds for these infants were not statistically significantly different from those of usual room sharers at the 5% level of significance.
Cite this as: BMJ 2010;341:c4971
Competing interests: None declared.