How not to screw up your baby

Writing this from paternity leave as a first time dad, as any self-respecting prospective parent, I did a bunch of anxiety-ridden last minute research into early childcare.

TL;DR It’s pretty hard to screw up a baby. The few strong findings - like sleeping on the stomach being clearly dangerous - will make it to your paediatrician’s advice fairly fast, so you’ll hear about them. With other things, the effects are more minor and heterogeneity among children means you’ll likely to do what works out for you, anyway. If you enjoy optimising as yours truly, read on, however.

The basics

There are roughly three grades of strength of evidence in paediatrics, based on how well a study design can separate causation from correlation:

• A - Randomised control trials. Paediatrics is unusually hard to study because you cannot ethically randomise most parenting behaviours, but some studies do exist. For example for breastfeeding, instead of enforcing babies to a dietary group, mothers in the treatment group are nudged towards the treatment. Especially good are meta-analyses of RCTs, pooling their effect sizes.
• B - Observational studies. These vary in quality and design, but the most important subgroups are:
  • Sibling studies - Even though the treatment allocation is not random, we can assume many of the confounding factors such as the mother’s background to be the same.
  • Cohort/cross-sectional studies - The quality of these depends on how convincingly they include control factors (income, race etc).
  • Case control studies - Used for when the outcome is rare, like SIDS. For a set of cases such as SIDS deaths, a control group is selected (possibly matched on other factors), and differences between the groups are estimated. For example, if the incidence of parents smoking in the SIDS group is larger than control, this suggests a causal link is possible.
  • Case studies - Simple reports that something has occurred, without a strong suggestion of a causal link.
• C - Low quality evidence. Animal and cell culture studies, survey data, anecdotes, folk wisdom, anthropological observations, psychological theory, philosophy, appealing to nature etc. I am clumping all of these together, while a separate hierarchy could of course be built here.

The majority of baby books are vibes based (category C), often marketed by someone with a PhD espousing their philosophy, with a dearth of peer-reviewed evidence. I found only two high quality audiobooks [1][2] that summarise solid science, while being reader-friendly.

Parents often care about three things: survival, feeding and sleeping. These also tend to be the bitterest points of emotionally loaded debate online.

1. Survival. Roughly 0.1 to 1 children per 1000 die from SIDS before one year of age. Before safe sleep campaigns, this was more than double, so following those recommendations is a good first step to follow. Some surprising preventions appear to be effective: for example, offering a pacifier at night time can more than halve the odds of SIDS. However, once obvious prevention methods have been applied, it’s not fruitful to further worry about SIDS - the residual risk starts approximating everyday risks such as driving with an infant.
2. Feeding. In developed countries, formula is a solid alternative to breastmilk. The stigma attached to formula feeding derives from historical observational studies (of evidence category B) noting that breastfed babies have significantly better life outcomes. However, there’s a significant confounding factor here - the mother’s motivation and socioeconomic status. Once controlling for these, either in RCTs (A) [10] or sibling studies (best of B) [6][7][8][9] , the difference in long term outcomes that people mostly stress about (IQ, obesity) either vanishes or is small. That said, breastmilk does provide several short term benefits, so should be the default choice. If it does not work out, however, no reason to fret.
3. Sleeping. The most emotionally divisive subject seems to be sleep training - whether to separate the baby to own room and let them “cry it out” or other variations if they are fussy. Based on all kinds of quality studies [4] (including category A), this is fine, and improves both the child’s and their parent’s sleep, with no evidence of long term harm. However, since it contradicts sharply with philosophies such as attachment parenting (based on psychological theory, rather than paediatric studies), some parents find it horrific. Sleeping habits are also largely culturally determined, so sleep training is more popular in individualistic Western countries.

Institutional lag

Advice given in maternity wards often reflects long-standing practice and cultural norms as much as current evidence. Pacifier use is a good example: despite strong evidence of a protective association with SIDS and RCTs usually showing no causal link with early weaning [5], many parents are still strongly advised against it. This appears to be a case where institutional practice lags behind the evidence.

Throughout all of parenting community forum events (perekool) - lectures presented by midwives in our hospital during the pregnancy - the use of pacifiers was demonised as unnecessary, harmful for the latch efficiency and causing early weaning.

Overall, I’ve found midwives to be unreasonably opinionated about outdated practices such as this. We joke with my wife that during our hospital stay, for every 8 hour shift we received a new midwife, and a new set of instructions how to take care of the baby. And this is in Estonia, with one of the lowest child mortality rates in the world and a well functioning paediatrics system, so I assume it’s worse elsewhere. My suggestion - if you are already reading metaanalyses, you can probably ignore much of the midwives’ advice.

Just take it easy

Given the studies, there are at least three reasons to not stress too much about them:

• Studies are notably biased towards baby outcomes rather than the mother’s. For things like breastfeeding, sleep training or pacifier use, there’s an obvious trade-off between parental and filial comfort, especially if there are any deviations from the normal, such as breastfeeding difficulties. However, if parental comfort is pushed too far in the form of sleep deprivation or depression, it seems commonsense it will also eventually hurt the baby. Therefore, “try for the best, but then just do what works” seems like solid advice.
• Studies identify average effects, but children are markedly different. For example, with sleep training, some babies adapt to it quickly, others cry uncontrollably for hours if left unattended. Again, “just do what works for your child” seems like the way to go.
• Especially when listening to Cribsheet [2], it was notable how many tested effects were inconclusive or very small in size. This means that whatever you decide to do, it will probably work out fine. The goal is not to follow every recommendation perfectly, but to understand which ones actually matter — particularly when your sleep-deprived judgment is being bombarded with confident care instructions from people who have known you for five minutes.

Sources

[1] The Science of Mom, Alice Callahan
[2] Cribsheet, Emily Oster
[3] Do pacifiers reduce the risk of sudden infant death syndrome? A meta-analysis
[4] Effectiveness of behavioral sleep interventions on children’s and mothers’ sleep quality and maternal depression: a systematic review and meta-analysis
[5] Risks and Benefits of Pacifiers
[6] Improved Estimates of the Benefits of Breastfeeding Using Sibling Comparisons to Reduce Selection Bias
[7] Is Breast Truly Best? Estimating the Effect of Breastfeeding on Long-term Child Wellbeing in the United States Using Sibling Comparisons
[8] Breastfeeding and risk of overweight in childhood and beyond: a systematic review with emphasis on sibling-pair and intervention studies
[9] Effect of breast feeding on intelligence in children: prospective study, sibling pairs analysis, and meta-analysis
[10] Breastfeeding during infancy and neurocognitive function in adolescence: 16-year follow-up of the PROBIT cluster-randomized trial