The scientific method was pioneered by Galileo Galilei who did the first recorded gedankenexperiment (thought experiment) as well as the first experimental test of a physical law. He kept detailed records of his thoughts, experiments, and observations, including the observation while in church that a hanging lamp with a longer chain had a longer period and swung more slowly than a similar lamp with a shorter chain. Through a series of experiments he also determined that the period of the pendulum was independent of the mass of the lamp, that the period varied directly as the square root of the length of the chain, and that the period was independent of the length of the arc. (This latter statement only holds true for arc angle lengths for which the sine of the angle is approximately equal to the angle, expressed in radians, i.e. up to about 10 degrees, which is why grandfather pendulum clocks are tall and thin.)
Now let us address the science behind medical diagnosis and treatment. Almost all correct medical treatments and predictions were arrived at by observation and induction,and almost all the incorrect treatments arrived at by deduction or using false analogies. Induction (pace David Hume) is the reasoning that says that since the sun has risen in the east for over 10,000 years, it will rise in the east tomorrow. Deduction is the reasoning that says that the proper way to prevent calcium kidney stones is to decrease the amount of calcium in the diet (a statement which has been shown to be false: clinical studies have shown that you should INcrease the amount of calcium in the diet). We have to rely on induction because the "laws" of biology and life cannot be expressed by mathematical equations, and there are few if any axioms from which medical "laws" can be deduced. We also sometimes have difficulty defining what we mean by "normal", and how "abnormal" you have to be before a doctor declares that you have a "disease".
And let us not forget the male analog of a tree falling in a forest and making no sound because there is no one there to hear it: Many men feel that you are only sick if a doctor tells you that you are sick, so if you never see a doctor.............then you are never sick. This is one reason why men get so depressed when they have a heart attack---their denial has crashed into the wall of reality. They then also often become extreme believers in the virtues of a healthy diet, weight loss, and daily exercise.
We also have a macro/micro problem: If I want to improve the overall health of a nation of 10 million people, and lengthen the average life expectancy of its citizens, I would recommend certain tests, but for a given individual, since I have no advance way of knowing the ultimate cause of his/her death, I cannot recommend those same tests with the same degree of certainty. I have a number of elderly patients who have refused annual mammograms, and a number of patients (curiously enough not the same people) who have refused an annual stool for occult blood and periodic colonoscopies, and so far none of the first group has developed breast cancer, and none of the second (with one exception) has developed colon cancer. (The exception was a female who had had lymphoma, and did agree to annual pap smears.) In addition, to the best of my knowledge, none of my patients who refused the flu vaccine have died from influenza.
The basis of medical reasoning by induction is that if a certain treatment for a certain disease diagnosed by a certain method "cured" more patients than did non-treatment, then the same treatment for the same disease (if indeed you do have the same disease as the treated group) will cure you. We are here generalizing that we are all human beings, and deliberately neglecting all differences in sex, age, weight, height, hair color, skin color, ethnic background, geography, living conditions, time of year, religious belief, overall health, and prior medical history. Whether or not any of these conditions makes a difference for the particular treatment for a particular disease has not, in general, ever been looked at, and we are also assuming that if a drug is given it will have the same chemical constituents and potency as the medicine used to treat the cured group. This is a very large assumption. We also conveniently ignore the fact that the test group was not on any other medicines, since the test group is almost always pharmacologically null, and most patients over the age of 50 are taking at least three prescription or over-the-counter drugs. To show you how disease outcome predictions have changed, I just read a medical summary of the third patient to survive an infection of the brain with the rabies virus, which we were taught as recently as 10 years ago is 100% fatal.
I am forever reminding my medical residents that common diseases occur commonly, but can occur with uncommon symptoms. For instance, an acute attack of (ocular) glaucoma can present as intense abdominal pain. If a brunette says she is coughing up red hair, then you should check her mediastinum for a dermoid cyst. If a patient complains of red urine or blood-like red strands in the stool, inquire about recent ingestion of beets. If the computer says that a female patient has severe diabetes because she is spilling sugar and protein in her urine, has an elevated blood sugar, and has not had her period for two months, you must first rule out pregnancy, because in most computer disease diagnostic trees, pregnancy is not listed as a disease, and therefore the computer cannot recognize the constellation of symptoms for what it really signifies.
The field of medicine is an art as well as a science. A well-trained physician can look at a patient, recall the last seen visage of the patient and decide that "this patient looks sick", using some internal heuristic and gestalt.
Unfortunately that is not an acceptable computer diagnosis, so we are forced to invent a disease that fits.
I tell my residents that if the admitted patient is alive when I come in to do my morning rounds, then they have succeeded in the first part of their job, which is to ask for each and every symptom and abnormal test "what disease or condition can the patient have that can kill him/her if I don't make the diagnosis, and how can I diagnose and treat it?". This requires applied induction, because the resident must ask him/herself what condition that he/she has either seen or read about that can evince these symptoms. I might add that the real challenge is to know which abnormal tests should be ignored, because they are not clinically significant to the presenting case.
As a final note, a physician's ego should never interfere with the treatment of the patient. A patient's disagreeing with either the diagnosis or proposed treatment is never a reason for the physician so show anger or disbelief. I have many, many male diabetic patients who should be taking insulin but they refuse: if they take insulin they really have diabetes, but if they only take pills then they just have "a sugar problem".