Screening can be defined as the identification of a disease in a patient before the onset of signs and symptoms. This is important because in many diseases, particularly chronic diseases, early detection coincides with early treatment, better outcomes, and lower costs to the healthcare provider.

Screening for some diseases has greater potential benefits than screening for others. In order for disease screening to be worthwhile, several conditions, as listed here, must be met (Martin et al., 2005):

• The outcome must be alterable at the time of identifying the disease
• The screening technique should be cost-effective
• The test should have high accuracy (see Chapter 3, Machine Learning Foundations for methods for measuring test accuracy in healthcare)
• The disease should carry a large burden on the population

An example of a popular screening problem and solution is using the Pap smear to screen for cervical cancer; women are recommended to undergo this cost-effective test every 1-3 years throughout most of their lives. Lung cancer screening is an example of a problem that has yet to find an ideal solution; while using x-rays to screen for lung cancer may be accurate and may lead to earlier detection in some cases, x-rays are costly and expose patients to radiation, and there is no strong evidence that early detection influences the prognosis or outcome (Martin et al., 2005). Increasingly, machine learning models are being developed in lieu of medical tests to screen for diseases including cancer, heart disease, and strokes (Esfandiari et al., 2014).