Reaffirming a Positive Correlation Between Number of Vaccine Doses and Infant Mortality Rates: A Response to Critics

Abstract

Introduction In 2011, we published a study that found a counterintuitive, positive correlation, r = 0.70 (p < .0001), demonstrating that among the most highly developed nations (n = 30), those that require more vaccines for their infants tend to have higher infant mortality rates (IMRs). Critics of the paper recently claimed that this finding is due to "inappropriate data exclusion," i.e., the failure to analyze the "full dataset" of all 185 nations. Objective In the present study, we examine various claims postulated by these critics and the validity of their scientific methods, and we perform several investigations to assess the reliability of our original findings. Methods The critics select 185 nations and use linear regression to report a correlation between the number of vaccine doses and IMRs. They also perform multiple linear regression analyses of the Human Development Index (HDI) vs. IMR with additional predictors and investigate IMR vs. percentage vaccination rates for eight different vaccines. We perform odds ratio, sensitivity, and replication analyses. Results The critics' reanalysis combines 185 developed and Third World nations that have varying rates of vaccination and socioeconomic disparities. Despite the presence of inherent confounding variables, a small, statistically significant positive correlation of r = 0.16 (p < .03) is reported that corroborates the positive trend in our study. Multiple linear regression analyses report high correlations between IMR and HDI, but the number of vaccine doses as an additional predictor is not statistically significant. This finding is a likely consequence of known misclassification errors in HDI. Linear regression of IMR as a function of percentage vaccination rates reports statistically significant inverse correlations for 7 of 8 vaccines. However, several anomalies in the scatter plots of the data suggest that the chosen linear model is problematic. Our odds ratio analysis conducted on the original dataset controlled for several variables. None of these variables lowered the correlation below 0.62, thus robustly confirming our findings. Our sensitivity analysis reported statistically significant positive correlations between the number of vaccine doses and IMR when we expanded our original analysis from the top 30 to the 46 nations with the best IMRs. Additionally, a replication of our original study using updated 2019 data corroborated the trend we found in our first paper (r = 0.45, p = .002). Conclusions A positive correlation between the number of vaccine doses and IMRs is detectable in the most highly developed nations but attenuated in the background noise of nations with heterogeneous socioeconomic variables that contribute to high rates of infant mortality, such as malnutrition, poverty, and substandard health care.

Keywords: child immunization; coefficient of determination; confounders; hdi- human development index; hepatitis b vaccine; infant mortality rates; linear regression analysis; non-specific effects; vaccination rates; vaccines.

Figure 1. Scatter plot of infant mortality rates vs. number of vaccine doses with best-fit line and correlation coefficients

Reported by Miller-Goldman [1] and Nysetvold et al. [2].

Figure 2. Line of best fit for vaccination rates <60% (red) and for all vaccination rates (blue), with colored data points showing that high vaccination rates are neither necessary nor sufficient to cause low infant mortality rate

Figure 3. Correlation coefficients resulting from successive linear regression analyses of nations with incrementally worse infant mortality rates

Figure 4. Linear regression analysis of infant mortality rates and infant vaccine doses per nation reported by Goldman and Miller, 2019 data