Why are our Caesarean Sections High?
The number of caesarian sections (C-sections) in Kenya occasionally raises uproar on social media. Some people believe we have too many of them. In contrast, others believe most CSs are unnecessary and could have been avoided. According to KNBS data from 2022, Kenya had a total of 211,227 C-sections, 1,024,354 normal deliveries, 9,395 breeches, and 4,931 assisted vaginal deliveries. The total number of live births in that year was 1,249,907, which gives us a caesarian section rate of 16.9%. This figure replicates the 2022 figures from the Kenya Demographic and Health Survey, which found the national CS rate to be 16.5%. Interestingly, as the chart below shows, CS rates differ by county. Kirinyaga had a CS rate of 40%, Kiambu 33%, Tharaka Nithi 30%, Nairobi 28%, and Taita Taveta 27%. According to WHO, the CS rate in a population shouldn’t be more than 10%. So, who is responsible for these additional CSs?
High rates of CSs have been blamed on doctors and health facilities because they get to decide who goes to the theatre. Some doctors, however, contest suggesting that public health facilities only recommend CSs when absolutely necessary. Private hospitals, on the other hand, haven’t come out to defend themselves. The purpose of this post is to test both hypotheses. I will do that by running two models in a regression analysis. First, we measure the variance in CS rates in the country accounted for by the number of private facilities in each county. In the second, we check how much variance is explained by the number of public facilities in each county.
The table above is what you get when you run both models. The first model measures the effect of private facilities on the rate of CSs, while the second measures the impact of both public and private facilities. The R tells us the correlation, an essential component because we need to know the relationship between CS rate and the number of public and private facilities in a county. From the table, a higher number of private facilities in a county correlates with higher CS rates (0.477). As seen in model two, adding public facilities only slightly changes the effect. Adjusted R square tells us the variance, whereby the number of private facilities explains 21.1% of the variance in CS rates in the counties. Adding public facilities raises the variance to 22.3%. The R square change is the difference between the variances, which tells us that public facilities only explain 2.9% of the variance in CS rates.
Plotting these findings, you get the charts below. In the first chart, we notice that the relationship between the number of private health facilities and the rate of CSs is positive. In the second, the trendline is almost horizontal, suggesting no relationship between the number of public health facilities and CS rates (Having more public facilities in your county does not predict higher CS rates). This confirms the raw findings we saw above.
You also notice we have red dots and blue ones. The blue dots are the actual observed rates of CSs in the counties, while the red dots are the predicted rates. Using the model, for example, we can predict the CS rate of a county if it had 2,000 private health facilities by plugging in the coefficients. But that’s not as interesting as realizing that the actual CS rates in some counties are way beyond what is predicted by the model.
For instance, Kirinyaga County had a CS rate of 40%, Kiambu had 33.2%, Tharaka Nithi had 30.2%, and Nairobi had 28.1%. However, from the prediction model, the CS rate for Kirinyaga County is 17.5%, Kiambu 25.9%, Tharaka Nithi 13%, and Nairobi 35.4% (Ideally, these would have been the CS rates if the relationship between the number of private facilities and CS rate was perfect R=1). The difference between the actual rate of CSs and the predicted rates is interesting because it tells us that other factors could also determine the higher CS rates, giving rise to the disparity. These additional factors could also be driving higher CS rates in private facilities. Finding these factors means testing all variables we can think of. In this post, I’ll focus on the few where data exists.
We could assume, for example, that women who prefer CSs would rather go to private hospitals than face stun doctors in public facilities. There’s also the possibility that wealthier women opt for private facilities, where it’s much easier to request CSs and get them. I tested whether variables like poverty and HDI could predict higher CS rates, and the results were impressive. The table below is similar to the one above, which includes two additional models: the overall poverty rate and the Human Development Index.
HDI in the counties only explains 2.1% of the variance in CS rates and is insignificant, as shown in the last column. On the other hand, when you add the second model that includes private facilities in a county and its overall poverty rate, the correlation (R) with the CS rate jumps to 0.7, which, by all definitions, is strong. The correlation tables show the relationship between the overall poverty rate and the CS rate is significantly strong and negative (-0.69). This means that the poorer the counties, the lower the CS rates. Our first graph shows this, where counties like Wajir, Mandera, and Turkana have the lowest CS rates. Worth noting is that Kirinyaga County has the second lowest poverty in the country after Nairobi, with an overall poverty rate of 19.3%. From the table above, we also see that differences in poverty explain 27.5% of the variance in CS rates in the counties. Together with the number of private facilities, both variables explain 48% of the variance in CS rates.
From these, we now know that the higher the number of private facilities in a county, the higher the CS rates. We also know that higher poverty correlates negatively with CS rates. There are several implications for these results. First, one could argue that private health facilities do not open in poor counties. Second, wealthier individuals choose to give birth in private facilities, which might drive up their CS rates for either of these two reasons. The first is that more affluent women might choose CSs, and the second is that private facilities might recommend CS rates to wealthy women at higher rates to make more money. We cannot verify either of these hypotheses, but at least we know either could be true. One thing we can do, however, is to confirm whether wealthy counties have higher rates of C-sections. I used the fourth and highest wealth quintiles in the counties as measures of wealth and then ran a regression model.
The regression statistics show that wealth correlates strongly with CS deliveries at 0.6, and the relationship is significant. Wealth also explains 35% of the variance in CS rates in the counties, which is even higher than the 21.2% variance explained by the number of private health facilities in a county. The beta coefficients and the confidence intervals also look good. This finding is important because it confirms the negative correlation between poverty and CS rates. It also confirms our previous sentiments that wealth might drive CS rates up.
Interestingly, when you run both models (number of private facilities and wealth) together, the number of private facilities stops being a significant predictor of CS rates while wealth maintains a strong relationship. From the table below, the beta coefficient for the number of private facilities (0.044) is lower than that of wealth 0.57, suggesting that the latter has a more substantial effect. The confidence intervals and significance levels are also quite telling. Wealth has positive confidence intervals above 0, while the number of private health facilities does not. I believe this is because more wealth leads to more private facilities in a region and, subsequently, higher CS rates. Therefore, the number of private facilities stops being a strong predictor of CS rates because it’s just a proxy of wealth. This makes sense because Kirinyaga County has fewer private facilities than a county like Nyeri, yet it has higher CS rates. More CSs could be happening in the few private facilities in Kirinyaga, probably because the women can afford it.
Looking at global trends, we see that CS rates are rising. More developed countries are averaging CS rates of 30%. In comparison, the least developed countries have yet to bypass the recommended 10% average by WHO. These results further confirm our hypothesis that wealth could drive higher CS rates. Despite these findings, we cannot lose sight of the fact that CS rates above 10% are unnecessary and do not lead to fewer infant and maternal mortality rates.
To sum up, we have discovered that public hospitals might be telling the truth and do not conduct CSs unless necessary. Second, private hospitals are probably behind the high CS rates; however, they only explain a small portion of the variance and might be a strong proxy of affluence in a region, a stronger predictor of CS rates. Wealthier counties have more private health facilities and higher CS rates. This trend is global, whereby richer countries have been observed to have higher CS rates. However, this matter is not yet settled, and we need to look into more factors influencing CS rates.
