Results

NOTE: STILLBIRTH DATA IS SIMULATED

Overview

The following graphics and results show the different analytical methods we used to create a model that could analyze the relationship (or lack thereof) between rates of COVID-19 present in wastewater and rates of stillbirth at the municipal level in Alberta.

Correlation Plot

Figure 5. Correlation plot showing the relationship between simulated municipal-level stillbirth data and average viral copies per person of COVID-19. The x-axis (viral copies per person) is log-transformed.

Table 1. Excerpt of dataset used in the correlation plot showing the Location, n (number of paired ranks), r (correlation coefficient), and R2 (coefficient of determination).

The correlation plot (Figure 5) shows the correlation between our simulated stillbirth data and COVID-19 wastewater data, in units of viral copies per person. Here, the COVID-19 data has been log-transformed, because it was right-skewed. In Table 1, the statistic results from this plot are shown for each location. The number of datapoints to be compared is n, r is the correlation coefficient, and R2 is the coefficient of determination. All of our r and R2 values are quite close to zero. If our data were not simulated, we could say that the data does not adhere very well to a linear relationship (due to r being closer to 0 than -1 or +1), and the proportion of variance in the stillbirth data that is due to the predictor variable (COVID-19) is very low (R2 value approaching zero). This makes sense, as the variation in our dataset comes mostly from random number generation in R.

Linear Mixed Model

Figure 6. Linear Mixed Model Plot of simulated stillbirths and COVID-19 prevalence, by location. This shows the plot for Calgary individually on the right, and the other nine locations overlaid on the left. The x-axis (viral copies per person) is log-transformed.

The linear mixed model (LMM) plot (Figure 6), has location as a random effect and COVID-19 as a fixed effect on stillbirths, which is the response variable. None of these results are particularly significant or meaningful, but again, this model could be used to detect relationships between levels of COVID-19 in wastewater, and the number of stillbirths occurring in the province. As the COVID-19 data is right-skewed, it has been log-transformed.

Below, there are two different version of the LMM output in R: one with the COVID-19 data log-transformed (Figure 7), and one without the transformation (Figure 8). This shows the formulas for the analysis, as well as statistics for the random and fixed effects. Here, the fixed effects section indicates that the intercept is approximately 4, which means that with no fixed effects added, there are about 4 stillbirths per municipality per month. The effect of each additional unit of SARS-CoV-2 in the wastewater has an effect of -0.006347 (Figure 7) and -2.976e-16 (Figure 8). These are very small effects, and were this data not randomly simulated, we would very likely not have any kind of significant results. Furthermore, the residual variance in both figures far exceeds the variance due to location, showing that more of the variance in the sample is due to random chance than the location.

Figure 7. Linear Mixed Model output for COVID-19 and stillbirth analysis, with the COVID-19 data log-transformed.

Figure 8. Linear Mixed Model output for COVID-19 and stillbirth analysis.

Discussion

The analyses that we have shown above together provide a way to analyze the relationship between COVID-19 presence and rates of stillbirth. With the correlation plot, we can look, at the municipal level, for the degree of linearity between COVID-19 and stillbirth. The more linear the relationship, the stronger the effect of COVID-19 levels on rates of stillbirth. We could also explain this by saying that when the relationship is quite linear, more of the amount of variance in stillbirths can be explained by the effect of COVID-19.

The linear mixed model we have done has COVID-19 as a fixed effect on stillbirths, with location as a random effect. This allows us to account for the different levels of COVID-19 in the wastewater of different locations, as they can be different at the same point in time. It was also far easier than doing an individual analysis of each location, with better statistical power.

Overall, this series of analyses could, with proper data, investigate whether or not COVID-19 has an impact on stillbirth, and this could be done for Alberta, any other province, or even a country as a whole. Further developments of this model could include a time-lag analysis, which would examine whether the impacts of COVID-19 on pregnancy are immediate, or occur at a later point after infection.

Limitations

The limitations of this model are numerous. The first of which is our emmeans splicing of wastewater data. As can be seen in Figure 4, the trendline decreases dramatically from the earlier levels, and stays close to zero. This figure shows our best and most consistent data amongst all locations, and even then, reporting had decreased enough that the resulting model is not useful. Unfortunately, though this was our best attempt at connecting the two periods of data collection, it was not able to work with the provided data.

This model obviously cannot account for all of Alberta, as it only looks at ten municipalities in the province. These are definitely the largest municipalities, and account for a majority of the population, but we still miss out on many smaller municipalities, that also likely do not have wastewater monitoring operations.

Though the consistency of the Alberta wastewater data has improved substantially with the takeover by Alberta Health, allowing us to reasonable compare samples over space and time, it is not a perfect method. Wastewater data can be influenced by external factors, such as ratios of residential to industrial wastewater and extreme weather events (e.g. floods).

This model also unfortunately does not account for other adverse pregnancy outcomes, including miscarriages, low birth weight, early delivery, and maternal death. In the future, medical and public health research should further investigate the relationship that COVID-19 has to these outcomes, as the virus seems to be here to stay.

Page updated

Report abuse