Desktop Medicine

Concepts of disease are essential for defining medicine. By the 20th century, the dominant concept was pathology in an individual, the foundation for the bedside model of medicine. Bedside medicine organizes the patient-physician relationship around the chief concern, which guides the focus of the history taking and physical examination; medical training that emphasizes laboratory-based sciences and a physical diagnosis; and a bedside presentation.

Today, however, a new model has emerged: desktop medicine. This term describes how a desk with a networked computer is transforming medical science and, in turn, medical practice. The desktop is the space in which researchers discover risk factor–based diseases and where physicians and patients go to gain information to diagnose and treat diseases. In developed nations, desktop diseases such as dyslipemia occupy a substantial portion of a physician's practice, are leading causes of morbidity and mortality, and have attracted the attention of policy makers. Medicare will soon require an annual personalized health risk assessment.¹

Desktop diseases are discovered when studies show a factor (eg, blood pressure) is associated with a negative health outcome (eg, stroke), and then a clinical trial shows that an intervention affecting that risk factor reduces the risk of that outcome event.² Key technologies are networked computers that perform rapid multivariate analyses of large data sets. These sciences and technologies enable researchers to discover the characteristics of persons at risk and to create prediction models that assess whether a patient is at sufficient risk that a physician ought to intervene. For example, the National Cholesterol Education Program's risk assessment tool integrates 7 factors to determine a person's 10-year risk for myocardial infarction.³

In desktop medicine, the clinician gathers risk factors by taking the patient's history and physical examination and by reviewing published clinical studies. The clinician then uses these risk factors to determine whether the patient is at sufficient risk to recommend treatment. This exercise of gathering risk factors and then assessing how well they predict health outcomes and the benefits of reducing those risk factors is a clinical-actuarial correlation. The Fracture Risk Assessment criteria⁴ for the diagnosis of osteoporosis illustrate this. A physician gathers a patient's 12 clinical risk factors, enters data about those risk factors into an online model, and receives the patient's 10-year probability of a fracture, and then determines whether to recommend treatment.⁴

Desktop medicine has begun to transform how physicians diagnose bedside diseases. Risk measurements compete with signs and symptoms and encompass progressively milder stages of disease. For example, Alzheimer disease is transforming from a diagnosis based on disabling cognitive declines to a biomarker of neurodegeneration. Concepts of treatment as risk management are also transforming the care of bedside diseases. Patients who recover from a bedside disease often enter into years of monitoring for other diseases (eg, colitis that requires screening for cancer).⁵

The salience of risk in desktop disease discovery, diagnosis, and treatment suggests that the Medical College Admission Test should measure skills in probabilistic reasoning and decision making, thereby encouraging students to major in desktop sciences such as statistics and psychology. The core medical curriculum needs revision as well. The US Medical Licensing Examination should test knowledge of epidemiology, decision sciences, and biomarker-focused laboratory sciences, and how well students apply probability to clinical practice and managing information.

The desktop and bedside medicine models differ in the role of the patient's chief concern to organize the clinical encounter (Table). The desktop encounter begins with an approach called running the numbers first.⁶ This involves performing a risk assessment before soliciting the patient's chief concern.

Advocates of this approach contend that when physicians begin with the chief concern they can neglect the care of desktop diseases and thus inadequately treat these diseases, such as failing to intensify treatment in patients with uncontrolled hypertension. Critics argue that this approach is at odds with the principles of primary care.⁷ However, physicians need skills in how to incorporate desktop and bedside models into the office visit and how to shape patients' expectations for a visit, especially for patients with both bedside and desktop diseases.

Bedside diseases are categorical. Disease is either present or it is not. In contrast, desktop diseases are dimensional because risk is a continuum. The argument follows that when risk data are available, physicians should discuss disease not as a category but as a probability.⁸ Rather than a disease label compelling treatment (eg, I have cancer, remove it), a risk estimate allows patients and physicians to practice clinical-actuarial correlation (eg, my chance of cancer death is too low to justify surgery).

This approach presents challenges. Because patients have more access to their own risk data via electronic resources and self-measurement of biomarkers, physicians lose exclusive control over organizing the medical encounter. In addition, both patients and physicians have cognitive biases in how they reason through risk information. Each may transform calculated risks into markedly different values. This personalized representation can affect decision making in a manner contrary to the goals of risk reduction.⁹

To address these challenges, medical training needs to include how to help patients to appreciate their relevant risks and effectively manage these risks. Just as bedside medicine developed methods to help physicians and patients understand and appreciate symptoms (eg, how many flights of stairs can you climb before you get short of breath?), desktop medicine needs to develop techniques to help patients think about and act on their risks. This desktop approach will include skills that cultivate the expectation of the opposite of risk, which is the probability of a future good outcome.

Clinical-actuarial correlation and running the numbers first identify patients who need interventions to reduce risks, but patients often fail to adopt or adhere to those interventions. Instead, patients may maintain behaviors that achieve short-term goals but result in long-term harms. Essential to desktop treatment is physicians improving their skills in how to change these patients' behaviors. Approaches such as payments for medication adherence will require physicians to learn how to talk with patients about using monetary incentives to treat disease.¹⁰

Desktop medicine does not so much change medicine as explain the way it is. Educating and training physicians to practice desktop medicine is especially important for the care of elderly patients who have competing risks.