NR 503 Week 3 Bed Sharing SIDS Study Appraisal Example

Appraising the 2013 Carpenter Pooled Case-Control Study of Bed Sharing and SIDS, Its Claims Set Against What the Design Can Deliver, and the Alcohol Question Three Datasets Never Asked

Student Name

College of Nursing, Chamberlain University

NR 503: Population Health, Epidemiology and Statistical Principles

Instructor Name

Month Day, Year

Appraising the 2013 Carpenter Pooled Case-Control Study of Bed Sharing and SIDS, Its Claims Set Against What the Design Can Deliver, and the Alcohol Question Three Datasets Never Asked

Why This Study

Few questions in infant care produce as much disagreement as whether a baby should ever sleep in the parents' bed. The American Academy of Pediatrics recommends room sharing without bed sharing (Moon et al., 2022), while some breastfeeding advocates argue that bed sharing is low risk when no hazards are present. Carpenter et al. (2013) set out to settle the question for the lowest-risk families: babies who are breastfed and whose parents do not smoke. Their headline finding, that bed sharing raised the risk of sudden infant death syndrome (SIDS) about fivefold even in that group, is quoted widely in parent teaching. This appraisal asks what the study's design can and cannot support.

The Design as Performed

The authors describe the study as an individual-level analysis of five case-control datasets. That label is accurate, but what was done is worth stating fully. The authors combined data from five existing studies, themselves drawing on 19 studies across the United Kingdom, Europe and Australasia, into one dataset of 1,472 babies who died of SIDS and 4,679 living control babies of similar age, time and place. Cases were identified after death; controls were selected from the living population. Information on sleeping arrangements and other factors was collected by interviewing parents. Where information was missing, it was imputed, and the analysis used random effects logistic regression to adjust for confounding factors and allow for differences between the studies.

So the study is a comparison assembled after the outcome was already known. It cannot show that bed sharing came before a process leading to death in any way other than by asking parents what happened, and it can estimate odds ratios but not risks directly. The absolute risks in the abstract were estimated by combining the odds ratios with external population rates, which is a model, not a measurement.

The Claim Ledger

Each main claim is set beside what the design can deliver.

Table 1

Claims Set Against the Design

Note. Figures from Carpenter et al. (2013). CI = confidence interval.

When the Exposure Was Recorded

Exposure information came from parents' reports, collected after the event. For cases, bed sharing referred to the last sleep before death. For controls, the paper defines bed sharing as the baby being in bed with the parents when they woke on the day of the interview. The two reference periods are not the same kind of night. A case's last sleep was a single specific night, remembered after a tragedy; a control's reference sleep was the most recent night before an interview. Parents of a baby who died may remember the night in intense detail, and some may feel pressure to describe it in a particular way. Parents of a living baby are describing an ordinary morning.

That asymmetry is built into most SIDS case-control studies, and it cannot be fully removed. It can inflate or deflate the odds ratio depending on how each group reports, and the appraisal cannot say which way it pushes here. What it does show is that the exposure was measured differently in the two groups, which is the kind of error adjustment cannot fix.

Who Reached the Final Table

The most quoted estimate, the odds ratio of 5.1, applies to a narrow subgroup: babies under 3 months, breastfed, with non-smoking parents, a mother who took less than two units of alcohol in the prior 24 hours and used no drugs, and no other risk factors. The question is how many babies in that subgroup had those characteristics actually recorded.

The paper answers part of this itself. It states that questions on the mother's alcohol use in the last 24 hours and on illegal drug use were not asked in three of the five studies (Carpenter et al., 2013). For those studies, alcohol and drug use were imputed from other variables. So some babies counted in the no-alcohol, no-drugs subgroup were placed there by a statistical model rather than by anything a parent reported. The authors acknowledged the criticism and describe going back to original records for the key subgroup, but the records they found mostly covered alcohol use by others in the household rather than the mother. The key subgroup is therefore partly defined by imputation, and the odds ratio of 5.1 rests on it.

A Second Study, the Same Question

Blair et al. (2014) analyzed two UK case-control studies with 400 SIDS cases and 1,386 controls and reached a different answer for the low-risk group. Co-sleeping overall carried an odds ratio of 3.9, and the risk was very high on a sofa or next to a parent who had drunk more than two units of alcohol. But bed sharing in the absence of those hazards showed no significant risk overall (odds ratio 1.1, 95% CI 0.6 to 2.0), and for babies under 3 months the estimate was 1.6, with a confidence interval running from 0.96 to 2.7.

The two studies agree that hazards such as smoking, alcohol and sofas multiply the danger. They differ on the babies with none of those hazards, and one likely reason is exactly the measurement problem above: the UK studies asked about alcohol directly. The disagreement does not show that bed sharing is safe for young infants; Blair's estimate for babies under 3 months still leans toward risk. It shows that the size of the risk in the lowest-risk group is not settled.

One Thing Settled, One Not

Settled: bed sharing is associated with SIDS, and that association is much stronger when parents smoke, when the mother has been drinking or using drugs, or when the baby is young. Both studies support this, the finding is consistent with earlier work, and it is the basis of current guidance (Moon et al., 2022).

Not settled: the size of the risk, if any, for a breastfed baby under 3 months whose parents do not smoke, drink or use drugs, sleeping in an adult bed with no other hazards. Carpenter et al. (2013) estimate an odds ratio of 5.1, but it depends in part on imputed alcohol and drug data; Blair et al. (2014) found no significant increase. For parent teaching, that means the nurse can state clearly which conditions make bed sharing dangerous and can explain why the AAP advises against it for young infants, without claiming that one study has proved a fivefold risk for every family.

References

Blair, P. S., Sidebotham, P., Pease, A., & Fleming, P. J. (2014). Bed-sharing in the absence of hazardous circumstances: Is there a risk of sudden infant death syndrome? An analysis from two case-control studies conducted in the UK. PLoS ONE, 9(9), Article e107799. https://doi.org/10.1371/journal.pone.0107799

Carpenter, R., McGarvey, C., Mitchell, E. A., Tappin, D. M., Vennemann, M. M., Smuk, M., & Carpenter, J. R. (2013). Bed sharing when parents do not smoke: Is there a risk of SIDS? An individual level analysis of five major case-control studies. BMJ Open, 3(5), Article e002299. https://doi.org/10.1136/bmjopen-2012-002299

Moon, R. Y., Carlin, R. F., & Hand, I. (2022). Sleep-related infant deaths: Updated 2022 recommendations for reducing infant deaths in the sleep environment. Pediatrics, 150(1), Article e2022057990. https://doi.org/10.1542/peds.2022-057990