In most real-world biological ecosystems where stimuli are often riddled with ambiguity, predictions are inherently probabilistic, and resources are time and computation constrained, organisms are forced to navigate the trade-offs between false-positives and false-negatives and the opportunity costs of (in)action. For example, if a feeding animal senses the presence of a possible predator, at the risk of losing its meal, it must decide whether the signal is real or a misdirection from a competitor, how imminent the threat is, and should it prepare to fight, flight, hide, or alternatively, can it afford to wait and invest additional resources to acquire more information. Organisms that can more optimally minimize the cost of false positives and protect against the downside risk of false negatives will be better adapted to the environment. In fact, organisms create niches (next essay) to minimize the uncertainty in the environment and optimize this trade-off.
A false positive (FP) occurs when a prediction is incorrectly classified as a positive and a false negative (FN) occurs when a prediction is incorrectly classified as a negative. For most real-world and informationally open predictions, it is not possible to eliminate both FPs and FNs. FP and FN overlap at the tails of their distributions and decreasing one increases the other. The optimal trade-off depends on the consequences of a prediction and the costs associated with the prediction. In most real world scenarios, the trade-off often skews towards the false positive. Returning to the example of a feeding animal, a FP results in a lost meal but a FN can lead to the loss of life. Similar to the biological world, our social institutions also tend to be false positive tolerant. In the tradition of law, a defendant is innocent unless proven guilty. The burden of proof – beyond reasonable doubt – resides with the prosecutor. It is preferable for someone guilty to be falsely categorized as free (false positive) than for a free person to be wrongly convicted (false negative).
Following the lead of law, medicine has the same FP tolerance and FN aversion. Every screening test, every biomarker, every radiological, and every clinical assessment has the underpinnings of the margin of safety – false positive tolerance. On the surface, this makes sense as the immediate downside risks of FNs (a missed diagnoses) outweigh the costs of FPs (more invasive testing or immediate treatment). However, within the context of the healthcare ecosystem that is strained under the pressures of population management, that lacks specific biomarkers (upcoming essay), and strongly incentivized to do, not miss, and treat, FPs cause insidious harm. In the Emergency Department, cursory triage screens lead to reflexive test ordering in under-discriminated and mis(categorized) patients of low specificity (high false positive) biomarkers such as lactate, d-dimer, and white blood cell count. Every lactate elevation is treated with broad spectrum intravenous antibiotics, d-dimer elevation in low-base rate patients are further evaluated by ionizing radiation and nephrotoxic effects of CT scans, and borderline or incidental findings on imaging studies are always tested and treated. In effect leading to needless hospitalizations, unnecessary treatments, and increasingly invasive testing. This is especially insidious in medically vulnerable populations such as the elderly, the immunosuppressed, the multimorbid, and the chronically ill. Although the breadth and depth of the effects are mostly unmeasured (upcoming essay), evidence does point towards false positives contributing to the epidemic of iatrogenesis and wasted resources.
In an inherently uncertain and dangerous world with resource constraints, all organisms risk drowning in false positives or perishing in false negatives. Although the biological world tends to be false positive tolerant, through the processes of learning, intelligence, and adaptability, fit organisms more optimally minimize uncertainty and optimize the trade-off between FPs and FNs. If learning is the process of making associations, then intelligence is making salient associations, and adaptability is having calibrated and ranged responses to variable and nuanced circumstances (upcoming essay). Learning, intelligence, and adaptability emerge out of the interaction between organisms and their ecological habitats. Many of these lessons of biology can broadly be applied to social institutions which represent a cumulative response – an adaptation (next essay) – to minimize uncertainty and identify risk. However, due to the lack of incentive schemes – selection pressures (upcoming essay) – healthcare mostly ignores and makes invisible the insidious effects of false positives.