The patient is a 30-year-old male who seems healthy. The patient has to do a blood test for some administrative forms (namely, proof of immunity to rubella and varicella), and wonder what else could be worthwhile to test.

What blood tests are worth doing for a healthy 30-year old male?

A comment deleted by a moderator suggested ​"Glucose, HbA1c, HDL and LDL cholesterol, TSH, T4, creatinine, urea, sodium, potassium".

The patient has no concern in particular, and cost is not an issue.

  • My thinking here was to consider common diseases that you can have that can go unnoticed. E.g. it is known that many people have undiagnosed diabetes, many people have undiagnosed hypothyroidism. Also kidney function can be impaired quite a bit (e.g. due to diabetes) without that leading to symptoms. To make the question better defined, one can ask how to choose some given number tests (say 10), of blood tests such that the survival probability after, say, ten years is optimized. This can be calculated from the known statistics. Mar 10, 2016 at 4:56
  • @CountIblis And like many comment answers it could have been made into a good answer. Why not just do so?
    – Carey Gregory
    Mar 10, 2016 at 5:09
  • @CareyGregory Ok, I'll do that. Mar 10, 2016 at 5:11
  • Is cost meant to be considered in this? It looks as though you are going to have you antibodies titered against a common ID panel, most of which also include other pathogens as a matter of course. Do you have any concerns in particular? In general I always try to stay away from such tests as hypochondria-like behavior can easily be induced in such a setting, and normally doesn't improve outcome.
    – Atl LED
    Mar 10, 2016 at 19:55
  • @AtlLED No concern in particular, cost is not an issue. Mar 10, 2016 at 21:10

2 Answers 2


The best place to find this answer is the US Preventive Services Task Force. They are an "independent, volunteer panel of national experts in prevention and evidence-based medicine" who make recommendations for what kinds of routine care are worth doing (in other words, do more good than harm) for what sorts of people (e.g. which age, sex, etc.). They give their recommendations grades based on how strong the evidence is and how substantial the benefit is.

They created an app to filter the recommendations based on someone's age, sex, pregnancy status, tobacco use, and if they are sexually active. For a 30 year old male, here are the blood tests that are recommended (and the answer to your question).

Grade A (recommended with substantial benefit/certainty):

  • HIV screening
  • Syphilis screening if at increased risk

Grade B (recommended with moderate benefit/certainty):

  • Hepatitis B screening if at high risk
  • Hepatitis C screening
  • Tuberculosis screening if at increased risk

Let's consider common diseases that a healthy 30 year old can have without noticing any symptoms. E.g. it is known that many people have undiagnosed diabetes, many people have undiagnosed hypothyroidism. Also kidney function can be impaired quite a bit (e.g. due to diabetes) without that leading to symptoms. If we focus on these issues then you could choose to the following test: Glucose and HbA1c to see if the person has diabetes, TSH and T4 to detect hypothyroidism, and creatinine, urea, sodium, potassium to detect problems with the kidneys. Also, measuring HDL and LDL cholesterol can be useful as quite a few young people have too high cholesterol levels.

Now, to make the question better defined, one can ask how to choose some given number of blood tests such that some chosen health criterion, say, the survival probability after ten years is optimized. This can in principle be calculated from the known statistics. To see how to set up this calculation, consider doing just one blood test for disease X.

The patient is in this case selected from a pool of people who do not have any significant symptoms of disease X. So, if X represents diabetes, the patient is currently not complaining about excessive thirsts, feeling tired etc. If X represents kidney disease then the patient is not at the stage where the kidney function is so low that it causes symptoms. This means that the probability that the patient will be found to be suffering from X should be derived from the appropriate conditional probability that conditions on the patient not having any significant symptoms (the symptoms are mild enough for it to be compatible to having no complaints).

For any chosen X you can then calculate the health criterion (e.g. survival after ten years) in the event of a positive test compared to not doing the test. So, this depends on the known effects of early treatment, the probability for detecting X will then yield the expected improvement for this health outcome.

  • 8
    IMO, this line of reasoning just doesn't work. First, your chosen conditions are whatever you thought of off the top of your head. More importantly, the principles of population screening are much more complicated and take into account whether identification of an abnormality is likely to improve outcomes, a question that requires careful study design and data analysis to account for lead-time bias, etc. The principles outlined in the landmark paper by Wilson and Jungner (1968) are further nuanced in newer studies, e.g. Harris, 2011.
    – Susan
    Mar 10, 2016 at 8:54
  • @Susan I've rewritten the answer. One has to condition on the available information. What matters is that if this is repeated many times, the patients do better as judged by the chosen health criterion compared to any alternative choice for X. Omitting information used to condition the probabilities on (e.g. observations that the patient looks a bit overweight) can influence the health gains of the tests, but you are still going to find the optimal strategy given that you decide not to consider the weight of the patient. Mar 10, 2016 at 17:55
  • 3
    Anyway, different issue : as I read the question, it's a practical one -- what has been shown to be helpful? This answer is about a theoretical model of how to determine this (though I would maintain that this is better done by way of empirical data..... the requisite "known statistics" in your model aren't known at all, and the risks of screening are yet another issue to be considered), but I don't think actually answers the OPs questions except by a couple guesses in the first paragraph.
    – Susan
    Mar 10, 2016 at 23:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.