Obstetrics is preventive medicineIn essence, obstetrics is a form of preventive medicine. Prenatal care and expert observation in labor are intended to identify and treat the women and babies who will have complications. That number is not a trivial number. The World Health Organization has arbitrarily declared that a C-section rate of 15% is needed to have low perinatal mortality and as we have seen from looking at international statistics, that number is more likely to be 20% or higher.
Most measurements and tests in obstetrics are screening tests. Screening tests identify those who are AT RISK for complications. Diagnostic tests actually make the diagnosis. We must decide where to make the cut off point for investigating the results of the screening test. So, for example, if we want to be know if a woman has gestational diabetes, we can give her glucola to drink and measure her blood sugar one hour later. However, it is up to us to decide what number is a blood sugar level that is too high and must be investigated further.
It is much easier to understand type I and type II error if you look at a graphic representation like the one below:
Let's assume that the curve on the left represents glucola test values for women who don't have gestational diabetes and the curve on the right represents glucola test values for women who do have gestational diabetes.
The first thing that you notice is that, in general, women who have gestational diabetes have high test results, and, in general, women who don't have gestational diabetes have low test results. The second thing you notice is that the curves overlap. There are women who don't have gestational diabetes who have high test results and there are women who have gestational diabetes who have low test results.
Now we have to decide where to set the cutoff point. Every woman who has test results over the cutoff point will have further testing for gestational diabetes and every women who has test results lower than the cutoff will be assumed to be normal.
Notice where the cutoff is drawn in the graphic, toward the end of the first curve. Notice what this means. There is a group of women who DO have gestational diabetes yet will not be identified; this is the type II error. Notice also that there is a group of women who DON'T have gestational diabetes who will be identified as at risk and will need further testing; this is the type I error.
Obviously, it is not acceptable for a gestational diabetes test to large numbers of women who actually have gestational diabetes. That means that the cutoff must be moved lower (pushed to the left). Look what happens then. We will reliably identify more women with gestational diabetes, which is desirable. However, we will also identify more women as having gestational diabetes when they don't actually have it.
Look very carefully. There is no place that we can put the cutoff that will have neither type I nor type II error. Decreasing the type I error (the number of women who actually have diabetes whom we fail to identify) REQUIRES that the type I error increase (the number of women who will be falsely identified as having gestational diabetes). There is no way to avoid it.
Any time you have to set a cutoff, and in the real world you always have to set a cutoff, you must compromise between type I and type II error. What is important to understand is that there will ALWAYS be type I and type II error. Therefore, the fact that there are unnecessary diagnostic tests and procedures is INEVITABLE, not an indication of poor medical practice. The only issue is where to draw the line (cutoff). So if you want to reduce the number of women who are erroneously identified as needing further diabetes testing, the trade off is that you will INEVITABLY fail to diagnose some women who do have gestational diabetes. Similarly, when you are looking at indications for C-section (estimated fetal size, decreased fetal heartrate, etc.) if you reduce that number of women who have unnecessary C-sections you will INEVITABLY increase the number of babies who die because they were not delivered by C-section.
That's why it is not particularly valuable to claim that there are too many C-sections, or too many women who have abnormal glucola results, but don't have gestational diabetes. If you want to reduce those numbers, you must specify where you think that the cutoff should be drawn AND you must explain why increasing the the number of women who should be treated but don't receive treatment (the type II error) is acceptable.