In April 2018, I set out to perform what I thought would be a simple task: use previously published literature to explore the shape of the relationship between host biodiversity and disease risk. In doing so, I hoped to test three key hypothesized contingencies in the dilution effect of biodiversity – the frequently observed pattern of reduced disease risk in communities with increased species diversity. I believed the task would be simple because the data were already collected, several hypotheses were well established, and the statistical tests would not require complicated or unusual analytical approaches. I was wrong.
For me, this story starts in 2006, with a foundational paper in disease ecology that leveraged a simplified epidemiological model to generate clear, testable hypotheses amidst a disjointed and complicated field of research. In their groundbreaking 2006 paper, Effects of species diversity on disease risk, Felicia Keesing, Bob Holt, and Rick Ostfeld provided a synthesis aimed at addressing a key question in disease ecology: What is the mechanism by which biodiversity influences disease risk? With their simplified model, Keesing, Holt, and Ostfeld decomposed decades of research into 5 testable hypotheses, and in doing so, the authors took something almost immeasurably complicated (the ecology of plant and animal hosts, and the epidemiology of specialist, generalist, and vector-borne pathogens) and managed to identify specific mechanisms by which diversity can influence disease risk.
What resulted from this study was an explosion of new research, including empirical studies aimed at understanding mechanisms of dilution and a drawn-out debate regarding the generality of the relationship in nature. And through these debates and research, some critical gaps in our understanding of the biodiversity-disease relationship began to crystalize.
First, ecological theory and some empirical work started to suggest that the biodiversity-disease relationship might be scale-dependent, with dilution weakening (and possibly reversing) at large spatial scales. Second, some researchers suggested that, due to logistical constraints in study design, observational and experimental studies might omit important biodiversity levels, leading to an overestimation of dilution effects in nature. Finally, researchers suggested that biodiversity-disease relationships could be nonlinear, noting that nonlinearity in the biodiversity-disease relationship would indicate that where systems fall along biodiversity gradients could influence the nature of the relationship between biodiversity and disease risk. Importantly, these three contingencies could interact: whether the dilution effect was scale-dependent or sensitive to missing data might depend on the shape of the underlying biodiversity-disease relationship.
So, in 2018, we set out to test these contingencies in the relationship between biodiversity and disease using the published literature. First, we addressed whether the relationship between biodiversity and disease risk might be nonlinear. We searched the literature for every study of biodiversity and disease (prevalence or abundance) that included more than two unique measures of host diversity. Then, using a simple set of statistical tests, we assessed whether these biodiversity-disease relationships are nonlinear as predicted. They are. But this nonlinearity did not seem to alter the fact that, across most published studies, increasing biodiversity reduces disease. Next, we assessed whether nonlinear biodiversity-disease relationships are also scale-dependent. They are. But spatial scale was often related to study design, suggesting that to truly understand how spatial scale influences the dilution effect, we likely need more data. Last, we assessed whether missing data might bias observational studies to more commonly report dilution. It can. But accounting for missing data did not alter whether studies were likely to observe dilution, nor did accounting for missing data alter whether dilution was scale-dependent. Thus, missing data does not appear to be a factor influencing the frequency of dilution effects in the published literature.
What do these results mean? First, according to the published literature, they suggest that, despite key contingencies in the relationship between biodiversity and disease risk, the commonly observed “dilution effect” at small and intermediate spatial scales is real. As biodiversity is lost at small and intermediate spatial scales, disease risk most commonly increases. Second, these results, which leverage data from small-scale mesocosms all the way to global syntheses, suggest that spatial scale might moderate the relationship between biodiversity and disease risk: the shape of the biodiversity-disease relationship favors dilution, but this effect weakens as spatial scale increases. However, these results also highlight another important gap in our understanding of the biodiversity-disease relationship: Although some large-scale observational studies have been conducted, disease ecology still lacks global experiments, which could help explain the mechanisms of spatial moderation in the dilution effect. Together, these results suggest that biodiversity loss could exacerbate disease outbreaks at the scales in which humans are most likely to encounter disease, highlights important scales in which biodiversity conservation might be most useful for minimizing and mitigating these consequences, and identifies areas in the literature where we need more data.