Looking for something? Start here.
Custom Search


Want The Unnecesarean in your inbox? Enter your email address:



« "More Business of Being Born" Explores Cesareans and VBAC | New Yorkers, Did You Get Your Maternity Information Booklet While Pregnant? »

Practice Variation: Induction Rates at University vs. Community Hospitals

Bookmark and Share


Guest post by Jessica Turon


The Basics

Many studies on the rising cesarean and induction rates focus on maternal risk factors, but it’s also clear that rates vary a lot from hospital to hospital. Does hospital variation reflect differing patient populations? Or is the variation due to different hospital characteristics, patients aside? Jill asked me to review this article, which looks at whether hospital type (university versus community) is associated with the rates of term labor inductions and cesareans if certain maternal characteristics are held constant.


The Boring Part

The paper is a retrospective cohort study based on Ohio birth certificate data in 2006-2007. (Did you know birth certificates had all that data on them? The one in your files at home sure doesn’t!) The investigators included only births from 20 weeks through 42 weeks, in known hospitals, and excluded fetal deaths and births with major congenital anomalies, for n=283,370.* The analyses of induction and cesarean rates further limited the study population to births from 37-42 weeks, for n=244,464.


The Results

As expected, they found that university hospitals had a somewhat higher-risk population in terms of maternal morbidities, and adjusted for these in the analyses. The comparison of labor induction rates at university versus community hospitals was notable:

  • at 37 weeks, the odds were 1.7:1, or about a 50% higher risk at community hospitals**
  • at 38 weeks the odds were 1.8:1, or about a 54% higher risk
  • and at 39 to ≤42 weeks, the odds were 2.0:1, or about a 60% higher risk

Unlike induction, cesarean rates showed no significant differences between community and university hospitals.


What’s it Mean?

Overall, this study is of good enough quality to show that in Ohio in 2006-7, community hospitals were much likelier to induce women at term than university hospitals.

More broadly, this study is a good piece of evidence about how hospital characteristics – not just maternal characteristics – can influence hospital-level rates of various obstetric practices. This is a topic that definitely needs more study!

Does this study provide information on what kind of hospital to deliver at, if you’re planning a hospital birth? The answer is “not really” – these data are more useful for policy-makers, hospital administrators, and so forth. The association found here was significant, but not huge, and the study looked at only one hospital characteristic.*** It’s probably more helpful to ask questions about the individual hospitals you’re considering, like whether they have midwives on staff, or their specific induction and cesarean rates.

You might be wondering why cesarean rates didn’t go up along with inductions, given what’s known about the association between them. The authors wondered too, and gave some background on how that association has been studied. This study can’t comment on the relationship, though, because the relevant data here is cross-sectional. In other words, we don’t know if the women having the inductions were the ones having the cesareans, so we can’t see the relationship between the two in the individual woman. Although we’d expect higher cesarean rates at the community hospitals to accompany their higher induction rates, this paper can’t explain why we don’t get that result.


Data Wonk Notes

*Oddly, the study looked at differences in maternal morbidity by hospital type among women giving birth between 21 and 42 weeks (n=283,370) and then used that data in the logistic regression to adjust for the smaller, presumably healthier population giving birth only at term (n=244,464). If this were corrected, it would likely have the effect of making the hospital types look more similar, which in turn would mean less adjustment, meaning the odds ratio would be lower.

**The paper makes a common mistake in interpreting the odds ratio that I have attempted to correct. They write “During gestational week 37, women who delivered in community hospitals were 70% more likely to undergo induction than those in university hospitals (27% vs 19%; aOR, 1.7; 95% CI, 1.5–1.8)” (page 346.e3). Unfortunately, this is the incorrect interpretation of an odds ratio (versus a risk ratio); more information here. I used this tool to correct it for the “% more likely” information I gave above. [They follow this up with “80% of the women were more likely to undergo induction in week 38 of gestation  (31% vs 21%; aOR, 1.8; 95% CI, 1.7–2.0)” which I am assuming is an editing error as it makes no sense otherwise.]

***They also looked at teaching versus non-teaching hospitals as an exposure and found similar but weaker results. There wasn’t a chart for this data or any explanation of confounders or adjustment.


And finally

If “Epidemiology Trail” were a game, I would probably be dead of dysentery. Did I get something wrong? Please tell me in the comments!


Snyder CC, Wolfe KB, Loftin RW, et al. The influence of hospital type on induction of labor and mode of delivery. Am J Obstet Gynecol 2011;205:346.e1-4.


PrintView Printer Friendly Version

EmailEmail Article to Friend

Reader Comments (3)

Thank you Jessica for taking the time to review and explain this article. Unfortunately, this article has as one of its conclusions that “Recently, the accepted doctrine that an induction of labor confers an increased risk of cesarean delivery has been challenged in the literature.” And their summary version of the article states that “These findings support recent studies that suggest an induction of labor is not so significant a risk factor for cesarean delivery as previously thought.”

There are several concerns with the conclusions that the authors made in this study. Jessica correctly points out that we don’t know the cesarean delivery rate for the women who were induced. More importantly, if we look at the demographics displayed in Table 1 of the article we find the same flaw present in many of these studies. The demographics are represented by mean values and not by the inherent risk for cesarean delivery. For a better understanding of why inherent risk is more important than mean values watch the video at http://www.youtube.com/watch?v=tjklQZRQKss (38 minutes) but skip the first 24 minutes and 40 seconds and watch for four minutes. I will bet that if we knew the inherent risk of the women who labored and had more information about the women who didn’t, we would find that inducing labor increased the risk of cesarean delivery in both the university and community groups.

Until the concept of inherent risk is better understood by all obstetrical care providers, obstetricians will continue to follow mistaken evidence right into the operating room. Get informed!

October 13, 2011 | Unregistered CommenterGustavo San Roman, M.D.

I'd love to see more such studies-- thank you for sharing.

It's so tough to extract solid indirect conclusions from studies like this, and I really wish they wouldn't try. There are so many reasons why the C/S rate may not have been higher in the community hospitals, despite the higher induction rates, that could have nothing or little to do with the validity (or invalidity) of a correlative or causal relationship between inductions and C/S.

Just for an example quickly fabricated off the top of my head (worth about what you paid for it)...

I find it likely that university hospitals have more (quantity) and more experienced surgeons than community hospitals do. So if we assume for the sake of argument that inductions are truly significantly correlated to C/S, I think it's quite possible that the link is not exactly causal, but it instead sort of pseudo-causal among those with the ability and comfort level to perform C/S frequently, quickly and "well." That is, perhaps it is the case that induction can be causal when it comes to C/S, but that more often, induction (more frequently than uninduced labor) leads to a "gray area," where labor may stall, etc. or lead to distress that is not sustained or dire, but perhaps open to a judgment call. In such cases, if there are more dedicated surgeons more closely monitoring these mothers, I can imagine that a C/S would be more likely to be called. If, however, there are fewer surgical resources readily at hand, it might be more likely that these "gray areas" would be waited out, whether intentionally or just by necessity (if the only available surgeon(s) is/are in surgery). And then, of course, more of those "gray areas" might lead to vaginal births.

But this is just one hypothesis. It might also be possible (not having read the study) that community hospitals might be more likely to use methods and/or "degrees" of induction that are milder, and thus less likely to lead to or be correlated with increased C/S. For example, if they have fewer resources, these hospitals may be more likely to try an induction, and if it doesn't seem to "take," stop it and send the woman home. Again, I haven't read the study, but perhaps it's possible that, given the induction disparity goes up after 39 weeks, that those women, being closer to term, might have a better chance of being "ready" and ending up with a vaginal birth. This might only account for part of the lack of difference in C/S rates, but it could be something.

Another, related idea-- I wouldn't be surprised if the women at the community hospitals received fewer U/S during pregnancy than the university-based women, even if the greater number of university-based higher-risk women were accounted for. Again-- resources. Since late "dating" U/S are notoriously inaccurate and seem more likely to move due dates up than back, I wouldn't be surprised if women at university hospitals got more C/S without TOL (for "big baby"-- at least as a factor) and/or if their inductions were more likely to lead to C/S because their babies more likely to be iatrogenically premature ("misdated" by U/S). Thus, though the induction rate itself might be lower in university hospitals, if they tend to induce women whose babies are less likely to respond "positively," they'd have more C/S than expected, even if induction does tend to lead to C/S. So the C/S rate at the community hospitals could be "too high" in part as a result of a "too high" rate of inductions, but they could "look normal" compared to university hospitals if university hospitals have a "too high" percentage of inductions leading to C/S. You know what I mean. Just some examples, some of which might be disputable.

The point being that there isn't nearly enough information to conclude that induction isn't "really" correlated or causal when it comes to C/S. My wild hypotheses are in addition to the main problem you already noted-- are women getting C/S the same ones who were induced? What's the connection there? If you don't even know that, then how can you draw any sort of conclusion about correlation between the two procedures? It would be like looking at two different populations and noting that though one had much higher rates of smoking, they had the same number of deaths from lung cancer, and then concluding that perhaps smoking is not so correlated to lung cancer as previously thought. But if you don't know who of the cancer deaths in each population were smokers, how can you even begin to say that? And further-- if you don't even have a good "control" group yet for what is a "normal"/"acceptable" rate of lung cancer deaths (as we don't yet, really, with C/S and inductions), then what are you even doing concluding at all?

It's certainly possible the low-smoking group lives in an area with one major industry, where most folks work in the same air-polluting factory, and the high-smoking group lives in pristine high desert. If you don't even have a great idea of what lung cancer rate is "normal," the high-smoking group could have (in retrospect) a very high rate of lung cancer deaths compared to what "should" be, but if the low-smoking group has the same rate, though it might be quite high as well, you could conclude that there is no correlation with smoking (or no correlation with polluting factories, for that matter) instead of that there are strong correlations with both.

October 16, 2011 | Unregistered CommenterDreamy

Hi Dreamy,

Thanks for the comment. In the article, the authors speculate more on this induction-cs 'mismatch' than they do on the actual question of why the community hospitals have such higher induction rates. It has piqued my curiosity on the body of research about the induction-cs correlation, though, as they gave a few pointers to some criticisms of the studies that established that belief. Not a body of lit I'm familiar with.

I wonder if you'd care to speculate in the same vein as you did above about the actual findings of the study regarding the higher odds of induction at community hospitals. How do the kind of resource-based arguments you made for the induction-cs connection (or lack thereof) reflect on the discrepancy in induction rates?

Thanks again for commenting!


October 18, 2011 | Unregistered CommenterJMT
Comments for this entry have been disabled. Additional comments may not be added to this entry at this time.