I think this characterisation is incorrect. For example, the claim that smoking tobacco is an indicator for lung cancer, heart disease, etc, is imprecise but useful.
The actual actuarial and medical science claims are in the form of correlations with life expectancy and disease diagnosis rates in studies large enough to be overwhelmingly confident in the direction and magnitude of the correlations.
So "is an indicator" is a popular press bastardization of science, not the actual state of human knowledge. I would point you to reviews of these studies, but I don't have good medical journal search skills.
On the other hand, some activities which certainly seem to be science, are apparently useless, cosmogony, for example.
Progress in cosmogony seems to be in proposing ideas which can be shown not to work. Following up the dead ends in a maze is also science because until we show they are dead ends, we don't know that they are dead ends. There is no widely accepted cosmogony hypothesis, so I would argue that the field is in its infancy. When mature it should tell us something useful about the nature of fundamental physics.
If we assume metaphysical naturalism (though scientists needn't so assume), the descriptions of nature are observations and observation is independent of science.
Observations of nature are by means of apparatus both natural and man-made. Millikin is credited for measuring the fundamental charge of the electron, but that statement confounds the observation: "Millikan and Fletcher levitated a bunch of oil droplets of a certain size in an electrical apparatus and recorded meter readings" with its interpretation in light of Newtonian mechanics and Maxwellian electrodynamics and the work of others, including measuring the viscosity of air. Since the latter was poorly or incorrectly known, Millikin's calculations led to a slightly incorrect synopsis of his observations.
http://en.wikipedia.org/wiki/Oil-drop_experiment
In a similar way, our personal and scientific experience colors our communication of our observations. We fit observations into our heads and sometimes our eyes fail us (example: we are blinded by glare) and sometimes our heads fail us (example: a magician produces coins from containers we thought we knew to be empty). So observations, to be useful, need to be communicated. This can render them unreliable. So we reserve the term "fact" for observations we have confidence in and their interpretations in the best available scientific theory. It's a fact that DNA molecules encode heritable information in humans, even though you have never seen your parents DNA and likely have never seen a report comparing its sequence to your own.
It seems to me that science is better characterised as the construction of theoretical models which allow scientists to predict the probabilities of making certain specified observations, given certain other specified observations.
Those models, assuming they are reliable, are precise, useful and communicable descriptions of nature.
"Science has to be a precise, useful and communicable description of nature." was coined in reaction to posts that were vague "The universe is like a tomato with all the seeds as galaxies and the inside like the outside.", or useless "The universe conspires to make all electrons seem alike, but they have hopes and fears and ambitions, not that any measurement will show you." or gibberish "The donut turns rigidly in two ways at once and this allows it to triangulate on its center." Some aren't even descriptions of nature as they explicitly reject confrontation with observations and when such observations are raised as problematic, the poster simply denies the observations.
So, scientists aren't concerned with communicating descriptions of nature so much as with models, which entail no ontological commitments.
While crackpots are concerned only with styling themselves as knowledgable intellectuals and therefore have no need for a philosophy of what constitutes knowledge.
The nominalisation program of Field and Balaguer casts doubt on this assertion too.
Mark Balaguer has not argued against this point (that "the most precise and most useful descriptions are necessarily communicated in mathematics" ) because he has not defined precise in a way that does not involve mathematics and he has not defined another means to usefully summarize observations of nature. Mathematics is not just of field of abstract study, but is also a language.
Colyvan and Zalta's review of Balaguer's Platonism and anti-Platonism in Mathematics
Hartry H. Field's
Science Without Numbers: A Defence of Nominalism (1980) seems to be just that -- but the work is largely mathematical. That large parts of Universal Gravitation can be done without numbers is no surprise to any student of Newton. But both Newton's geometry and Field's relational language are mathematics.
Page one of review of Field's Science Without Numbers: A Defence of Nominalism
Field's Nominalism -- the belief that physics relationships are real, is anti-scientific, not anti-mathematical. For example, Field's primary example, Newtonian mechanics and gravitation, is wrong-headed since that is not the best summary of the behavior of the universe, and the current best understanding is quite different while still preserving Newtonian physics as workable approximation at human scales. Nominalizing incorrect summaries of nature beggars the notion of "real."