How should we screen for gestational diabetes?

In our learning health system, we discovered that a new screening approach had increased diagnoses without improving overall outcomes. So remarkably, we switched back.

By Sascha Dublin, MD, PhD, an associate investigator at Kaiser Permanente Washington Health Research Institute (KPWHRI) and an internal medicine physician at Washington Permanente Medical Group, and Gaia Pocobelli, PhD, a research associate at KPWHRI

When pregnant women have gestational diabetes, it means they have too much sugar in their blood, which can cause problems. Babies may grow too large, causing injuries during birth to the mother or baby and raising the risk of cesarean delivery. After birth, babies may experience hypoglycemia (low blood sugar). And uncontrolled diabetes can increase the risk of stillbirth — which is rare, but obviously of great concern.

Being overweight increases the risk of diabetes, and with the obesity epidemic, gestational diabetes is becoming more common. Health care providers want to do the best possible job of screening pregnant women for this problem, so that we can help women control their blood sugar and have healthier babies. But how do we do that?

Right now, medical experts aren’t sure. The traditional approach, which is currently followed by most U.S. health care providers, has been a two-step process, where all women receive a basic screening test and only some women (those with an abnormal result) go on to a longer and more intensive test. But in 2010, a major international organization — the International Association of the Diabetes in Pregnancy Study Groups (IADPSG) — proposed a dramatic change: that all pregnant women undergo the longer, more intensive test. This has been controversial and most U.S. providers haven’t switched to this approach.

What do we do when all the evidence isn’t in yet?

So what should health care providers recommend when the evidence does not point to a single best path forward? We build the road as we travel, taking every opportunity to discover what works — and what doesn’t. And at Kaiser Permanente Washington, where we strive to be a “learning health system,” we’re using research methods to learn from our own data and then applying what we learn to improve care.

Gestational diabetes screening is a prime example. In 2010, Kaiser Permanente Washington Women’s Health leaders decided to change the organization’s guidelines to follow the new IADPSG recommendations mentioned above; we began screening all pregnant women with the single, intensive test. At the same time, our new guideline also lowered blood sugar target levels for recommending medication — and emphasized insulin as a first-line treatment.

But because there weren’t yet clear answers about whether outcomes were better with the one- or two-step approach, our organization wanted to study what was happening within our system. So with help from our colleagues Drs. Paula Lozano, Susan Warwick, Jane Dimer, and David McCulloch, we applied for and were awarded a grant from Group Health Foundation’s Momentum Fund (previously called the Partnership for Innovation).

Our study was large and rigorous

Many aspects of our study give us confidence in the validity of our results. First, it was a large study and took into account many characteristics of the women and their pregnancies. Most prior studies compared health outcomes in women and babies before and after a switch from a two-step to a one-step process. This can be confusing or inaccurate if there were other changes in care around the same time.

In our research, we were able to study a side-by-side comparison group where no switch in screening process had occurred — that is, pregnant women insured by Kaiser Permanente Washington who were cared for by outside, contracted providers. We found that nearly all the contracted physicians kept using two-step screening, whereas providers in our own group practice switched nearly completely to one-step screening. Thanks to the external control group, we could take into account background trends in health outcomes, making it more likely that what we found reflects true effects of the switch in screening process.

Here’s what we discovered: Following the guideline change in our group practice, diagnoses of gestational diabetes increased by 41 percent. Patients had a dramatic increase in the use of insulin, as well as increases in outpatient fetal monitoring, induced labor, and low blood sugar in newborns. Disappointingly, there was no benefit in terms of the outcomes we had most hoped to impact. We saw no reduction in cesarean deliveries, very high birthweight babies (more than 8 pounds, 13 ounces), or admissions to the newborn intensive care units.

In other words, we found that the screening change had made pregnancy and birth more “medicalized” — without improving important outcomes for babies or mothers.

Stepping back to move toward better outcomes

After hearing our results, our system’s Women’s Health leaders revised our guidelines, returning to two-step screening. This is a remarkable step because one of the hardest parts of American medicine is “de-implementation,” — that is, to stop doing a medical test or treatment even when it has been proven ineffective.

To help everyone benefit from what we’ve learned, we presented our results at the Society for Maternal-Fetal Medicine’s Annual Pregnancy meeting in January. And this month our findings were published online by the journal Obstetrics & Gynecology as “One-Step Approach to Identifying Gestational Diabetes Mellitus: Association with Perinatal Outcomes.” Our coauthors are Onchee Yu, Sharon Fuller, James Fraser, Paige Wartko, Lu Chen, PhD, and Katherine Newton, PhD, of KPWHRI; David McCulloch, MD, and Susan Warwick, MD, of the Kaiser Permanente Washington medical group; and Jane Dimer, MD, of Swedish Medical Center, all in Seattle.

To update Kaiser Permanente Washington’s guidelines, we also worked with Kate Zeichner, Emily Ghilarducci, Karen Madwed, Andrew Warner, Nadia Salama, Jessica Gorman, Avra Cohen, Kim Riddell, Avantika Waring, Angie Sparks, Maria Ciani, Evangeline Erskine, Annelise Gaaserud, James Greene, Neeta Makhija, Robyn Mayfield, and Ann Stedronsky.

It was great to work closely with our clinical leaders in designing and carrying out this project, and it’s been inspiring to see how much our findings mean to the clinical teams, who were eager to find out if the new approach was helping women and babies. They were incredibly receptive to our findings, and willing to take on the challenge of rethinking how to deliver care based on what we had learned. This truly was an example of the “learning health system” — how we can adapt and improve care in response to new knowledge made possible by the combination of rigorous research methods, great clinical data, and committed leadership.