All topics › Evidence gaps controversies
When a headline says 'X doubles the risk of Y', how do you work out whether it actually matters?
When a headline says "X doubles the risk of Y" — folic acid "cuts risk by 70%", home birth "raises risk by 75%", omega-3 "halves early preterm birth" — how do you work out whether it actually matters? This topic teaches you the single most useful skill in this whole book: converting a scary or exciting headline into plain numbers.
A headline that says a risk "doubled" tells you nothing until you know what it doubled from. Doubling a risk of 5 in 1,000 gives 10 in 1,000 (5 extra cases); doubling a risk of 1 in 1,000,000 gives 2 in 1,000,000 (1 extra case). Always ask for both groups' absolute numbers — "how many in 1,000 with X, and how many in 1,000 without?" A Cochrane review of 35 studies found that relative-risk framings are perceived as larger and more persuasive than absolute ones, and can mislead [1]. Then check three more things: was the difference statistically significant (or could it be chance?), who was studied, and is this one study or a whole body of evidence?
"Risk reduced by 70%" sounds like 70 out of 100 people were helped. It usually means the risk went from some small number to a smaller number — and the headline doesn't tell you either one. A Cochrane systematic review of 35 studies on how health numbers are presented found that relative risk reduction, compared with absolute risk reduction or "number needed to treat", may be perceived to be larger and is more likely to be persuasive — but it is uncertain whether it helps people make decisions consistent with their own values, and it could lead to misinterpretation [1]. The effect held for both patients and health professionals [1]. Relative risk isn't a lie; it's an incomplete sentence.
When information is given as conditional probabilities or percentages, the base rate — how common the thing is in the first place — gets normalised away, and people (including doctors) systematically ignore it [2][3]. The classic demonstration: out of 1,000 women, 10 have breast cancer; 9 of those 10 get a positive mammogram; but 89 of the 990 women without cancer also get a positive result. So a positive mammogram means cancer in only 9 out of 98 women — about 1 in 11. When Gerd Gigerenzer put this to 160 gynaecologists, a majority (60%) answered 80–90% [2]. Stating the same facts as natural frequencies ("9 out of 98") instead of percentages makes the right answer dramatically easier to see [2][3]. This book follows that convention everywhere: X in 1,000 for both groups.
Tversky and Kahneman's landmark experiments showed that describing the same outcome as "200 of 600 people will be saved" versus "400 of 600 people will die" reverses most people's preferences, even though the numbers are identical [4]. Parenting headlines exploit the same machinery: "halves the risk" (gain frame) and "doubles the risk" (loss frame) push your emotions before your arithmetic has had a chance. Noticing the frame is half the defence [4].
The Birthplace in England study (64,538 low-risk women) found that for first-time mothers, the serious-adverse-event rate was 9.3 per 1,000 planned home births versus 5.3 per 1,000 planned hospital births — an adjusted odds ratio of 1.75 [6]. "75% higher risk" is arithmetically correct. In absolute terms it is 4 extra adverse events per 1,000 births. For second or later births there was no significant difference (2.3 vs 3.3 per 1,000) [6]. And US birth-certificate data showed neonatal mortality of 1.4 vs 0.3 per 1,000 (odds ratio 4.19) [6] — a headline could truthfully scream "home birth quadruples newborn deaths", and the absolute difference would be 1.1 extra deaths per 1,000. Whether 4 per 1,000 matters to you is a values question; the headline alone doesn't let you ask it. (Full analysis in this book's home-birth topic.)
In the MRC Vitamin Study, folic acid cut neural tube defects from 35 to 10 per 1,000 pregnancies in women with a previous affected pregnancy — a 71% relative reduction, or 25 fewer affected pregnancies per 1,000 [5]. This is the rare case where the dramatic relative figure is matched by a genuinely meaningful absolute one — because the base rate in that high-risk group was high. Note the catch: that 35-in-1,000 base rate was in women with a previous affected pregnancy. In the general population the base rate is far lower, so the absolute benefit per 1,000 women is far smaller. Same relative reduction, different base rate, different decision. (Full analysis in this book's prenatal-vitamins topic.)
A Cochrane review found omega-3 reduced early preterm birth (<34 weeks) from 46 to 27 per 1,000 — a relative risk of 0.58, i.e. "42% lower", which is 19 fewer per 1,000 [7]. For any preterm birth (<37 weeks): 134 to 119 per 1,000 — "11% lower", or 15 fewer per 1,000 [7]. Neither figure is unimpressive, but "halves the risk" and "19 fewer per 1,000" land very differently, and only the second lets you compare against the cost and hassle of the supplement. (Full analysis in this book's prenatal-vitamins topic.)
"Significant" in a study means "unlikely to be due to chance" — conventionally, a less than 5% probability the result is a fluke. It says nothing about size. A huge study can find a tiny, meaningless difference and call it significant; a small study can miss a real, large effect and report "no significant difference" — which means uncertain, not no effect. Example from this book: an earlier review found iron significantly reduced low birthweight (risk ratio 0.81); the updated review, with more trials, found risk ratio 0.84 (95% CI 0.69–1.03) — not statistically significant [8]. The estimate barely moved; what changed was the precision. "Significant" is about precision, not importance — always ask how big the absolute difference is, regardless of the p-value.
The folic acid trial that produced the famous 71% figure enrolled women with a previous neural tube defect pregnancy — a high-risk group [5]. The Birthplace home-birth findings differ between England, the Netherlands, and the US because the maternity systems differ [6]. Before applying any headline to yourself, ask: were these people like me — same country, same era, same risk profile? If not, the absolute numbers may not travel even if the relative ones do.
Single studies are routinely overturned by the next study. Bodies of evidence — systematic reviews that pool many studies — move more slowly and are more trustworthy, though they can still shift (the iron example above is a review shifting). The strongest position is a review of many RCTs; the weakest is a single observational study reported as a breakthrough. More on why studies contradict each other in the companion topic [why-parenting-studies-contradict-each-other].
This is the chapter's payoff: evidence literacy is a parental wellbeing tool, not just an intellectual exercise.
Fewer 3am panics. Most parenting headlines that spike your anxiety — "X doubles the risk of Y" — dissolve or shrink once you convert them to absolute numbers. Four extra events per 1,000 is a real finding worth knowing about; it is not the catastrophe the relative figure suggests. Being able to do that conversion yourself, in the moment, is the difference between informed concern and free-floating dread.
Better conversations with clinicians. When a midwife, GP, or health visitor quotes a risk, you now have one question that cuts through everything: "In absolute terms — how many in 1,000 with, and how many in 1,000 without?" Clinicians are used to this question, and a good one will welcome it. If they can't answer, that itself is information.
Spotting marketing dressed as science. Supplements, sleep gadgets, and parenting courses are routinely sold with relative-risk claims ("clinically proven to reduce X by 60%!") that omit base rates, sample sizes, and controls. The checklist below works on adverts too: no absolute numbers, no sale of your attention.
An honest gap: whether teaching these skills actually changes parents' decisions or reduces their anxiety has barely been studied — most risk-communication research was done in clinical and screening settings, not parenting ones [1]. This topic generalises from that literature, which is reasonable but untested. The claim here is modest: these are the same questions researchers themselves ask, so they're the right questions for you too.
(Adapted for this topic: headline phrasing vs absolute reality, from this book's own topics.)
| Headline-style claim | Absolute reality (per 1,000) | What the headline hides |
|---|---|---|
| Folic acid "cuts neural tube defects by 71%" | 35 → 10 per 1,000 (25 fewer) — in a high-risk group [5] | The base rate: in the general population it's far lower, so the absolute benefit is far smaller |
| Home birth "75% higher risk of serious events (first babies)" | 9.3 vs 5.3 per 1,000 (4 extra per 1,000) [6] | 4 per 1,000 is real but small; no significant difference for second+ births |
| US data: home birth "quadruples newborn deaths" | 1.4 vs 0.3 per 1,000 (1.1 extra per 1,000) [6] | A 4x relative risk on a rare outcome; US system differences limit generalisability |
| Omega-3 "cuts early preterm birth by 42%" | 46 → 27 per 1,000 (19 fewer per 1,000) [7] | For any preterm birth it's "11% lower": 134 → 119 per 1,000 (15 fewer) |
| "90% sensitive" mammogram, positive result | ~9 in 98 positives actually have cancer — about 1 in 11 [2] | Sensitivity is not the chance you have the disease; the base rate dominates |
The 3am checklist. When a headline grabs you, run through these before reacting: