How do primary care providers determine that a patient needs thyroid replacement therapy?

The first question has a testable answer. Diagnosis rests on biochemistry rather than on symptoms, because fatigue and weight gain are common and non-specific: an elevated thyroid-stimulating hormone with a low free thyroxine indicates overt primary hypothyroidism, while an elevated TSH with a normal free thyroxine is subclinical and its management depends on the degree of elevation, on symptoms, on antibody status and on context such as pregnancy. Say that the test should be repeated before treating, since a single abnormal result can reflect a transient illness or laboratory variation, and note that reference ranges shift with age and pregnancy. That gives the answer a decision rule rather than a description. Getting the biochemistry right first also matters for the second question, because the harms of unnecessary replacement are only assessable once you know what necessary looks like. Note as well that reference ranges shift with age and in pregnancy, which is a common source of over-diagnosis in older patients.

The middle question is where the discussion earns its place, because adding a small dose to help someone feel better sounds benign and is not. Levothyroxine suppresses endogenous production, so a trial that starts is difficult to interpret and often difficult to stop; excess exposure carries documented harms, including atrial fibrillation and reduced bone mineral density, which matter most in exactly the older population most likely to be offered a trial for tiredness; the true cause of the symptoms goes uninvestigated while the patient and clinician wait to see whether the thyroxine helps; and the placebo response in fatigue is substantial enough that the trial itself will not answer the question it was started to answer. Then the side-effect question follows naturally, since most adverse effects of levothyroxine are the effects of too much of it, and the practical answer is about titration against repeated testing, the interval before re-testing, and the interactions and timing that alter absorption.