You are building a prediction model to develop a tool that can diagnose a particular disease so that individuals with the disease can receive treatment. The treatment is cheap and has no side effects. Patients with the disease who don't receive treatment have a high risk of mortality. It is of primary importance that your diagnostic tool has which of the following?

In this medical diagnosis scenario, the cost of a false negative (failing to diagnose a patient who has the disease) is extremely high (mortality). Conversely, the cost of a false positive (diagnosing a healthy patient) is low (cheap, no-side-effect treatment). Therefore, the primary goal is to minimize the number of false negatives. A low false negative rate ensures that the model identifies almost everyone with the disease, maximizing the number of patients who receive life-saving treatment. This is also known as maximizing recall or sensitivity.

A. High negative predictive value is desirable but secondary; directly minimizing the false negative rate is the primary goal given the high cost.
B. High positive predictive value aims to reduce false positives, which is not the priority here due to the low cost of treatment.
D. A low false positive rate is the opposite of the required strategy; we are willing to accept more false positives to avoid false negatives.

1. Ng, A. (n.d.). CS229 Machine Learning Course Notes: Evaluation Metrics. Stanford University. The notes on "Precision/Recall" discuss the trade-off, often using cancer detection as an example where high recall (low false negatives) is critical to avoid missing a diagnosis.

2. Powers, D. M. W. (2011). Evaluation: From Precision, Recall and F-Measure to ROC, Informedness, Markedness & Correlation. Journal of Machine Learning Technologies, 2(1), 37-63. Section 3.1 discusses how recall (sensitivity) measures the "proportion of actual positives which are correctly identified," which is precisely the goal of minimizing false negatives.