Sensitivity, specificity and fetal monitoringI've spent the better part of the last few hours doing calculations and drawing tables, trying to figure out the best way to explain some very difficult statistical concepts. I apologize in advance for how confusing this is. I welcome any suggestions on how to better illustrate the concepts, and, of course, any corrections of math mistakes.
For every screening test, there is a sensitivity rate and a specificity rate and these numbers dramatically affect the way that the test can or cannot be used. In the context of a screening test like fetal monitoring (which is screening for fetal distress), specificity is the proportion of times that result a result indicating no fetal distress actually reflects the fact that there is no fetal distress. In contrast, a sensitivity is the proportion of times that an abnormal fetal heart rate actual indicates real fetal distress. The primary problem with electronic fetal monitoring is that it has a relatively low specificity rate. In contrast, intermittent auscultation has a relatively low sensitivity rate.
In studies, continuous fetal monitoring appears to provide no advantage over strict intermittent auscultation, but that does NOT mean that electronic fetal monitoring is useless, and it does NOT mean that intermittent auscultation is just as good as electronic monitoring. To understand why this is the case, you need to know something about the way that sensitivity and specificity affect outcomes. In the following examples, I am going to assume that electronic fetal monitoring has a sensitivity rate of 99.9% and a specificity rate of 88.8%. I will also assume that intermittent auscultation has a sensitivity rate of 88.8% and a specificity rate of 99.9%. These are not actual values, but they serve as representations of the trade offs made between EFM and intermittent auscultation. We are also going to assume that every woman who shows signs of fetal distress, on EFM or intermittent auscultation, will have a C-section.
Let's take a look at a sample population of 1 million low risk women in labor, and let's assume (for ease of calculation) that this population will have a 1% rate of true fetal distress. What would the results be if all the woman in this group had EFM during labor?
The table shows that for our hypothetical low risk population of 1 million women monitored with EFM:
9,990 babies saved, 116,880 unnecessary C-sections, and 10 neonatal deaths
Compare that to 1 million women monitored using intermittent auscultation, which has a lower sensitivity level, but a higher specificity level.
Look at the dramatic difference:
8,880 babies saved, 99 unnecessary C-sections, and 110 neonatal deaths
Using intermittent auscultation resulted in more than 100,000 fewer C-sections, but an additional 100 babies died. You can make an argument (and many people do) that the life of 1 baby is not worth 1000 unnecessary C-sections, and hence, intermittent auscultation should be substituted for EFM. Of course, that means acknowledging that 100 babies would die who might otherwise be saved.
But that's not the end of the discussion. All screening tests perform better in at risk populations as opposed to the general population.
Let's create a hypothetical population of high risk women and assume that this group will have a 10% rate of true fetal distress. It should work out the same way, right? No, not even close. The performance of screening tests improves as the risk increases within the population, even though the sensitivity and specificity remain the same.
If our hypothetical group of 1 million high risk women were to be monitored with EFM, we might see something like the table below.
While the number of unnecessary C-sections is the almost the same, the number of lives saved has increased dramatically:
99,900 babies saved, 100,800 unnecessary C-sections, 100 neonatal deaths.
Once again, we can compare that to 1 million women monitored using intermittent auscultation, which has a lower sensitivity level, but a higher specificity level.
These change due to the increased risk level is dramatic:
88,800 babies saved, 900 unnecessary C-sections, 11,200 neonatal deaths
This result is clearly unacceptable.
We can summarize the data in a final table, showing the results of switching from EFM to intermittent monitoring.
In a low risk group, it may be feasible to replace EFM with intermittent auscultation. There will be a dramatic reduction in unnecessary C-sections, but there will be a small rise in preventable neonatal deaths. In contrast, in a high risk group, there is a comparable reduction in unnecessary C-sections, but there is a massive increase in neonatal deaths.
Looking at these numbers (even though they are all hypothetical), it is possible to make several claims. First, in order to determine how intermittent auscultation compares with EFM, it must be tested in hundreds of thousands of women. No studies like that have been done. Second, in small studies (like the ones that have been done to date) we would expect to see only small differences or even no difference in mortality rates. That does not mean that the two techniques are comparable. Third, the risk level of the population has a dramatic impact on the relative merits of the two methods of monitoring. There is no role for routine intermittent monitoring in high risk patients.