One of the biggest challenges in teasing out possible causation or directionality of an exposure and an observed phenomenon, it’s essential to consider confounding by indication. Although it’s described in the Medical Studies Core Topic Key Concepts page, it’s such an important consideration in both evaluating medical studies and in formulating questions for them that it deserves a special call-out — again and again and again.

So I’m writing three blog posts with mini case studies of confounding by indication because I REALLY want to drive home how important it is that reporters covering observational studies think hard about all the possible reasons a correlation might exist between an intervention or exposure and a subsequent intervention, medical condition or negative effect.

These are some of the most important questions to ask during an interview so that you can avoid misleading readers into thinking a spurious causal correlation exists when it doesn’t. (No, observational studies can almost never show causation, but they’re often reported — and read/interpreted — as such regardless.) Avoiding the trap of confounding by indication could prevent a great deal of harm and pain if researchers — and journalists — considered it more often.

In essence, confounding by indication means the most likely cause of a particular condition (often a chronic disease or an acute injury) or intervention is not related to the treatment or exposure a person had but to whatever CAUSED that exposure or treatment in the first place. But that explanation is abstract, so let’s get to the example.

A long-time debate in obstetrics is whether inducing labor increases the likelihood of cesarean deliveries, something the literature has appeared to suggest for decades. That higher rates of labor induction are correlated with higher rates of cesareans isn’t contested — plenty of studies show that relationship. But what if the reason a physician determines induction (indication) is necessary also is an existing risk factor for cesarean delivery? Researchers have tried to consider this as well in their studies, but only in the past few years have they designed studies adequately to really control for that possibility.

Gestational diabetes, for example, increases the risk of needing induction (though whether that’s necessary is also still debated) and increases the risk of a large-for-gestational age (LGA) babies — and LGA increases the risk of a cesarean delivery. So, in trying to determine whether induction increases cesarean risk, it’s necessary to consider why induction was indicated, whether that reason is also a risk factor for cesareans and then disentangle those interactions.

How that’s done depends on the medical question — watchful waiting is often used in OBGYN studies — but until it’s done, it’s not possible to definitively say whether it’s the induction or the reason(s) for induction that increase cesarean risk.

And that’s vitally important because assuming that induction alone can increase risk of cesarean influences whether a physician recommends an induction — and whether a woman accepts that recommendation. An induction and a trial of labor each have risks depending on a woman’s condition and gestational week, and cesarean deliveries carry their own risks.

It’s impossible to adequately weigh risks and benefits if you start with the assumption that one of your treatment options (induction) could increase other risks (cesarean delivery). Most early studies looking at induction and cesarean risk compared women who were induced to women who entered spontaneous labor and tried to control afterward for induction indications. Only more recently have studies taken a group of women who might all reasonably be induced for similar reasons and then assigned one group to induction and one to watchful waiting — thereby controlling for indication from the trial’s start.

Yet as more studies have done that, the evidence increasingly points to induction not increasing cesarean risk — and sometimes actually decreasing it. One of those studies produced such surprising results that became a central debate at the largest OBGYN annual conference last year, and it has the potential to dramatically change practice.

So, when reading an observational study, look for discussion of confounding by indication in their limitations section and look for possibilities they’ve missed — and ask them about it.

Tara Haelle (@TaraHaelle) is AHCJ's medical studies core topic leader, guiding journalists through the jargon-filled shorthand of science and research and enabling them to translate the evidence into accurate information.