Clinical Diagnosis and Differential Diagnosis

A differential diagnosis is a ranked list of possible diagnoses that could explain the patient's presenting complaint. Example: a 55-year-old man with chest pain and shortness of breath could have acute coronary syndrome (MI), pulmonary embolism, pneumonia, aortic dissection, or anxiety. Start with the worst-case (life-threatening) diagnoses: MI, PE, dissection. Then consider common diagnoses: pneumonia, anxiety, GERD. Use clinical features to narrow: Does he have chest pain radiating to the arm (MI), pleuritic pain (PE or pneumonia), or sharp pain between shoulder blades (dissection)? Is he hypoxic, hypotensive, or tachycardic? Does the EKG show ST elevation (MI) or normal (less likely acute). Test to rule out: EKG, troponin, chest X-ray, D-dimer (for PE). The goal is to narrow to the most likely diagnosis and rule out the dangerous ones. Narrow aggressively: if MI is ruled out by EKG and troponin, move to PE or pneumonia; order accordingly.

Bayesian reasoning uses conditional probability: the likelihood a patient has a disease given their features. Pre-test probability = how common is this disease in this population? Example: acute MI is common in a 60-year-old with chest pain and risk factors (diabetes, smoking), but rare in a healthy 25-year-old with chest pain. Likelihood ratio = how much does a test result change the probability? A positive troponin increases the likelihood of MI by 10–100x. A negative troponin decreases it by 10x. Post-test probability = likelihood after the test. If pre-test probability of MI is 20% in your patient, a positive troponin increases it to 70–90%, and a negative troponin drops it to 2–5%. Use this reasoning to avoid unnecessary testing (low pre-test probability + negative test = MI is ruled out; no need for more tests) and to focus on diagnosis (positive test + high post-test probability = treat for that diagnosis).

A good history includes: (1) Chief complaint—what brought the patient in today? (2) History of present illness (HPI)—onset (sudden vs. gradual), character (sharp, dull, throbbing), location (localized vs. diffuse), radiation (to arm, back, jaw), duration (minutes, hours, days), severity (1–10 scale), associated symptoms (nausea, vomit, fever, chills, sweats), triggers (food, exertion, positional, emotional), and what makes it better or worse. (3) Review of systems (ROS)—is there fever, chills, night sweats, weight loss, changes in appetite, vision, hearing, mood, sleep? (4) Past medical history (PMH)—previous diagnoses, surgeries, hospitalizations. (5) Medications—all drugs, doses, and adherence. (6) Allergies—drugs and reactions. (7) Family history—similar illness in relatives. (8) Social—smoking, alcohol, drug use, occupation, living situation, sexual history. Ask open-ended questions first ('Tell me about your pain') then close-ended to fill gaps ('Is it sharp or dull?'). Listen more than you talk; the patient will often reveal the diagnosis.

OPQRST stands for: O = Onset (sudden vs. gradual, when did it start?), P = Provocation/Palliating (what made it worse or better?), Q = Quality (sharp, dull, burning, throbbing?), R = Radiation (does it spread; to arm, back, jaw?), S = Severity (1–10 scale; how bad?), T = Timing (constant, intermittent, getting worse?). Example: MI pain is sudden onset, heavy/crushing quality, radiates to arm/jaw, 8–10 severity, unrelieved by antacids. GERD pain is gradual, burning, epigastric, relieved by antacids. Pulmonary embolism is sudden, pleuritic (worse with breath), shortness of breath is prominent. Gallstone is sudden, right upper quadrant, colicky (comes and goes), 1–2 hours duration. Use OPQRST to differentiate: the quality, radiation, and triggers often point to diagnosis.

Red flags vary by system: (Chest pain) Chest pain + diaphoresis, radiation to arm/jaw, hypotension = MI. Chest pain + dyspnea + hemoptysis + hypoxia = PE. Chest pain + severe back pain + hypotension = aortic dissection. (Abdominal pain) Abdominal pain + vomiting + inability to pass stool = bowel obstruction. Abdominal pain + severe, unrelenting, hypotension = ruptured AAA (abdominal aortic aneurysm). (Neurological) Sudden onset severe headache ('worst headache of life') = subarachnoid hemorrhage. Focal neurological deficit (weakness, speech loss, vision loss) = stroke. (General) Fever + stiff neck + rash = meningitis. Altered mental status in elderly = sepsis, stroke, or metabolic (hypoglycemia, hypoxia). Never dismiss red flag combinations; investigate immediately.

Chest X-ray (CXR) is read systematically: (1) ABCs: Airway (is the trachea centered?), Breathing (are lungs clear?), Circulation (is the cardiac silhouette normal?). (2) Look for infiltrates (white patches = pneumonia, pulmonary edema). (3) Look for effusion (fluid at lung bases, blunting of costophrenic angles = pneumonia with effusion, congestive heart failure). (4) Look for pneumothorax (black collapsed lung edge). (5) Look for cardiomegaly (enlarged heart suggests heart failure). (6) Look for specific patterns: 'butterfly' infiltrate (center of lung bases) = pulmonary edema (heart failure). Wedge-shaped infiltrate at periphery = PE with infarction. Normal CXR does not rule out pneumonia (10–20% of pneumonias have normal CXR early), PE (normal CXR in 25% of PE), or MI (CXR is usually normal in MI). Compare to prior CXR to assess for new findings.

Algorithms (CHEST PAIN protocol, SEPSIS bundle, STROKE alert) guide rapid diagnosis and treatment in common, time-sensitive conditions. CHEST PAIN: EKG within 10 minutes, troponin at 0 and 3 hours, serial vital signs. If troponin rises = MI, treat with dual antiplatelet therapy, heparin, and catheterization. SEPSIS: lactate, blood cultures, broad antibiotics within 1 hour, vasopressors if hypotensive. STROKE: CT head to rule out bleed, then tPA within 4.5 hours of onset or thrombectomy within 24 hours if eligible. Clinical decision support systems (EHR alerts) flag sepsis (elevated lactate + source), septic shock (hypotension requiring vasopressor), MI (troponin elevation + EKG changes), and guide next steps. These reduce missed diagnoses and speed treatment. However, over-reliance on algorithms can miss atypical presentations; always think critically and challenge the diagnosis if the patient does not fit.