Diagnosing a disease is a step-by-step process of gathering clues, forming a list of possibilities, and narrowing that list until one explanation fits. It rarely happens in a single moment. Instead, it unfolds through conversation, physical examination, lab work, imaging, and sometimes tissue sampling, with each step refining what your doctor thinks is going on.
It Starts With Your Story
The most powerful diagnostic tool is the patient interview. Before any blood draw or scan, your doctor needs a detailed picture of what you’re experiencing, when it started, and how it’s changed. Clinicians are trained to walk through your symptoms systematically, covering the location of the problem, when and how it began, what it feels like, whether it spreads to other areas, how severe it is on a scale of 0 to 10, what makes it worse, and what makes it better.
Beyond the immediate complaint, your medical history fills in critical context. Past illnesses, surgeries, medications, supplements, allergies, family health patterns, and lifestyle factors like smoking, alcohol use, and exercise all shape which conditions are more or less likely. Two people with the same chest pain can point toward completely different diagnoses depending on whether one is a 28-year-old runner and the other is a 65-year-old with a family history of heart disease.
The Physical Examination
After the interview, your doctor examines your body using four core techniques: observation (looking for visible abnormalities like swelling, discoloration, or asymmetry), palpation (pressing on areas to check for tenderness, lumps, or organ enlargement), percussion (tapping on the chest or abdomen to listen for changes in sound that suggest fluid or solid masses), and auscultation (using a stethoscope to listen to heart sounds, lung sounds, and bowel activity).
A physical exam can confirm or rule out possibilities quickly. A doctor who hears a heart murmur, feels an enlarged liver, or notices a specific rash pattern already has strong directional evidence before ordering a single test.
Building a Differential Diagnosis
Doctors don’t typically zero in on one condition immediately. Instead, they generate a differential diagnosis: a ranked list of conditions that could explain your symptoms. This list gets built from the moment you describe your first complaint and is continuously updated as new information comes in.
The conditions considered early in this process matter enormously. Research on clinical reasoning shows that diagnoses not considered, or considered only late, are less likely to be detected even when supporting evidence eventually appears. This is because the working list of possibilities shapes which questions get asked, which tests get ordered, and how results are interpreted. Generating multiple possibilities early leads to more thorough questioning and more complete documentation.
As evidence accumulates, the list shrinks. When it narrows to one or two leading candidates, the process shifts from exploration to verification: checking whether the lead diagnosis coherently explains all the signs, symptoms, and risk factors present.
Laboratory Tests
Blood and urine tests are workhorses of diagnosis. A complete blood count measures your red blood cells, white blood cells, and platelets, helping identify infections, anemia, clotting disorders, and blood cancers. Metabolic panels check substances like electrolytes, blood sugar, and proteins to evaluate organ function, particularly in the kidneys and liver. Other tests measure hormones, antibodies, or markers of inflammation.
Some lab tests target very specific conditions. Hemoglobin A1c identifies Type 2 diabetes by reflecting your average blood sugar over the past two to three months. Glomerular filtration rate flags chronic kidney disease. Ejection fraction, measured through imaging, helps classify heart failure into subtypes that require different treatment approaches. Sweat chloride confirms cystic fibrosis. These measurable indicators, called biomarkers, let doctors move from suspicion to confirmation.
Certain tests require preparation. Cholesterol, triglyceride, and blood glucose tests commonly require fasting, meaning no food or drink other than water for several hours or overnight. A cortisol test may require you to rest beforehand and avoid eating or brushing your teeth. Some tests ask you to skip specific foods: a fecal occult blood test may require avoiding certain medications, while other specialized tests restrict foods like bananas, avocados, or walnuts. Always tell your provider about any medications or supplements you’re taking, as these can skew results.
Diagnostic Imaging
When doctors need to see inside your body, they choose from several imaging tools depending on what they’re looking for. X-rays are typically the first option. They’re fast, inexpensive, and excellent for spotting bone fractures, dislocations, joint problems, and misalignments. Even when a soft tissue injury is suspected, an X-ray may be ordered first to rule out a fracture.
CT scans combine X-ray technology with computer processing to create detailed cross-sectional images in about a minute. Their speed makes them ideal for emergencies. They’re commonly used after accidents or falls to detect blood clots, organ injuries, and subtle fractures that standard X-rays miss.
MRIs use magnetic fields instead of radiation and provide superior contrast for both bone and soft tissue. They’re especially useful for sports injuries and musculoskeletal problems: torn ligaments, cartilage loss, nerve compression, spinal injuries, and tendon ruptures. An MRI takes longer than a CT scan but reveals details that other imaging cannot.
Ultrasound uses sound waves to produce real-time images and is frequently used to examine soft tissues, blood flow, and organs like the heart, liver, and kidneys. It involves no radiation, making it the standard choice for monitoring pregnancy.
Biopsy and Tissue Analysis
Imaging can reveal a mass or irregular tissue, but it cannot determine whether cells are cancerous. For most cancers, the only way to make a definitive diagnosis is a biopsy, where a small piece of tissue is removed and examined under a microscope. The sample is sliced thin, placed on glass slides, and stained to make cellular structures visible.
A biopsy does more than confirm or rule out cancer. If cancerous cells are found, the analysis identifies where the cancer originated, what type it is, and how aggressive it is. Aggressiveness is expressed as a grade, typically on a scale of 1 to 4, based on how abnormal the cells look. This information directly shapes treatment decisions.
Biopsies are also used for non-cancerous conditions, including autoimmune diseases, chronic infections, and inflammatory conditions where tissue-level detail is needed to distinguish between similar-looking illnesses.
Where Diagnosis Can Go Wrong
Diagnosis is not infallible. A large study of over 2,400 hospitalized patients who either died or were transferred to intensive care found that 23 percent experienced a diagnostic error, and 17 percent of all patients in the study suffered harm as a result. Among patients who died, diagnostic error was a contributing factor in about 7 percent of cases. These numbers reflect a population that had already experienced bad outcomes, not the general hospital population, but they underscore how high the stakes can be when the process breaks down.
Using structured checklists is one approach that reduces errors. General checklists prompt doctors to pause and reflect on their reasoning, while symptom-specific checklists provide comprehensive lists of possible causes for a given complaint, acting as a reminder of frequently missed diagnoses. Both approaches encourage considering more possibilities earlier, which leads to more thorough investigation.
AI-Assisted Diagnosis
Artificial intelligence tools are entering clinical practice, particularly in fields that rely on image interpretation like radiology and dermatology. Results so far are mixed. Consumer-facing skin scanning apps have shown poor overall accuracy, averaging around 22 percent for correct diagnoses in one study using a diverse set of skin images. When simply distinguishing between cancerous and non-cancerous skin lesions, these apps detected cancer less than half the time on average.
Performance also varies by skin tone. AI algorithms trained primarily on lighter skin consistently perform worse on darker skin, a significant equity concern. In one comparison, accuracy dropped meaningfully when the same algorithms were tested on images of the darkest skin tones versus the lightest.
In more controlled clinical settings, AI performs better as a support tool than a standalone diagnostician. One study found that dermatologists using a neural network achieved 76.6 percent diagnostic accuracy compared to 62.8 percent without it, with the biggest improvements seen among less experienced doctors. The technology is most useful when it augments human judgment rather than replacing it.