Contaminant or True Positive? How to Read a Blood Culture Result

First questionWho showed up?

Identity is the heaviest weight on the scale, because some organisms practically never lie and some lie almost every time.

There's a roster every microbiologist carries in their head. On one side, the organisms that mean business — recover them from blood and you believe them:

The left column means it. One bottle of S. aureus, of a streptococcus, of E. coli, of Candida — you report it, full identification and susceptibilities, and you ask for a repeat. No deliberation[1]. The right column is the cast of usual suspects: skin commensal flora and environmental drifters that hitch a ride during the draw[1][2].

But — and this is where the temptation is to sign out on autopilot — you never read the organism alone. You read it through the patient, because the lab requisition rarely tells you the whole story and you have to go looking for it. CoNS are the single most common contaminant we see and a genuine killer in exactly the patients who scare us: the ones with central lines, prosthetic valves, hardware, the neonates, the neutropenic[3][4][5]. Same bug. Opposite meaning. The patient decides which — which means before you call a CoNS "skin commensal," it's worth a glance at the ward it came from and a quick call to the unit if the ward is one of those.

Two names to tattoo on your memory:

• CoNS isn't always benign. Staphylococcus lugdunensis is a coagulase-negative staph by the bench definition, but it behaves like its vicious cousin S. aureus — it eats valves. Never wave it off as "just a skin commensal," and never let it get suppressed in a batch report as routine CoNS.
• Viridans strep keeps a double life. Often a contaminant; but in the right host — a febrile neutropenic with mucositis, or someone with a murmur and weeks of malaise — it's endocarditis announcing itself. Identity gives you the base rate. The patient tells you which way to bend it.

And resist the urge to over-read the species line. Pinning CoNS down to the exact species is satisfying on the bench, but on its own it hasn't reliably changed what clinicians do with the report[10]. It's a clue, not a confession.

Second questionHow many bottles, and did they agree?

This is the signal that decides more of our reports than any other — and the one most likely to be missing, because so many of our patients arrive with a single set drawn.

Picture the result two ways. The same organism blooming in both bottles of a set, or across two separate sets drawn from two separate sites — that's concordant growth, and it's the bug telling on itself, over and over. It pushes hard toward real infection[6][3]. Now picture the other version: one bottle positive, its partner stone silent. That's discordant growth, and for a skin commensal organism it is one of the most reassuring findings in all of microbiology.

Here's the number to keep in your back pocket. A lone bottle of CoNS in a two-bottle set — one up, one negative — carries roughly a 98% negative predictive value for true CoNS bacteraemia[6]. Read that again: in a patient with no hardware and no sepsis, that single bottle is contamination about nineteen times out of twenty. That one fact has spared more patients an unnecessary course of vancomycin than almost anything else in the stewardship toolkit — if the report makes the discordance clear instead of just stating "CoNS isolated."

Which leads to the rule we wish every ward internalised, and the one worth pushing in every CME and every collection-training session you run: draw enough sets, from separate sites. Two independent sets transform your ability to tell a true bacteraemia from a one-bottle skin commensal passenger[9]. One set isn't a blood culture. It's a coin flip in a lab coat — and when only one set arrives, that limitation belongs in your interpretive comment, because you are the only one who knows it's there.

Third questionHow fast did it grow?

The analyser hands you this one for free, and it's biology made visible: real bloodstream infection usually carries more organisms, so the bottle trips sooner.

The numbers are striking once you see them. Isolates that turn out to be real grow, on average, in around 18 hours. Contaminants dawdle in closer to 40[7]. The bench translation:

• Growth inside 24 hours leans pathogen. In one series, two-thirds of all true positives — and essentially none of the contaminants — had declared themselves by then[7].
• Growth after 48 hours, especially a textbook skin commensal organism, leans contaminant[7].

One caveat so the clock doesn't betray you: resistance slows the bug down. Methicillin-resistant staphylococci grow visibly later than their susceptible cousins, so a staph that's running a little behind schedule is not automatically off the hook[7]. Time-to-positivity is your best tiebreaker — not a verdict on its own.

Fourth questionWhose blood is this?

The exact same bottle means different things in different beds, because the patient sets the odds before the bug ever grows. And this is the information our requisitions are worst at giving us — so this is the question that most often means putting down the loop and chasing the ward, the file, or the treating unit for two minutes of history.

Lean toward "this is real," even for a low-virulence organism, when the patient is carrying a central line, a prosthetic valve, fresh surgery, an ICU or neonatal stay, a haematologic malignancy, or neutropenia[3][11][4]. Then layer the story on top: fever with a dropping blood pressure, a plausible source — a valve, a bone — and inflammatory markers (CRP, procalcitonin, a climbing white count) that all point the same way[11][12]. Each of those nudges a borderline CoNS from probably a skin commensal toward probably real.

And know your wards. Paediatric and oncology patients genuinely run a higher rate of true CoNS and other would-be contaminants, so an interpretive reflex calibrated on general adult medicine will quietly mislead you there[4][14]. The bench that reports for an oncology unit cannot use the same mental cut-off as the bench reporting for general medicine.

Now put them together

No single signal wins the argument. Their power is in agreement — which is exactly what the validated algorithms do: stack organism identity, time-to-positivity, the number of positives, and a quick clinical risk score into a single probability while susceptibilities are still cooking[13][12]. A classic CoNS model hit its sweet spot at about 62% sensitivity and 91% specificity only by weighing those features together, never one alone[12]. A paediatric ED model sorted children into clean tiers — 0%, 34%, 62% probability of true bacteraemia — with early growth (~19 hours) and an aggressive Gram morphology doing much of the heavy lifting[14].

So take the same four questions to two bottles sitting on your bench tonight.

Two patients, the same organism, opposite reports — and you can defend both. That's the whole craft.

One asymmetry to close on, because it overrides everything above: for a left-column organism — S. aureus, a streptococcus, an Enterobacterales, Candida — a single positive bottle is never dismissed. Ever. You report it and you ask for a repeat. The four-question framework is built for the grey zone. These organisms don't live there.

The decision: report, repeat, or suppress

Everything you just read exists to drive one of three moves — and all three are yours to make before the result leaves the lab.

Report it as significant when the story hangs together — a classic pathogen, concordant positives, fast growth, a fitting syndrome, or a line in the vein. Full identification, release susceptibilities, and where it matters, don't rely on the printed report alone — call the unit and ask for repeat cultures to prove it clears[13][3][11].

Sign out as equivocal and ask for a repeat when the signals quarrel — a borderline bug with one finger pointing each way. Report the isolate with an honest interpretive comment, hold the susceptibilities, and recommend a fresh set before anyone commits to broad-spectrum cover[9][8].

Suppress, with a documented reason, when everything converges on contamination: a single discordant bottle of a skin commensal organism, lazy growth, no sepsis, low-risk host. It's entirely reasonable to report "probable contaminant" and withhold the susceptibilities — but write down why, so the decision is yours on the record and not an unexplained silence the ward has to guess at[6][8][9].

None of this is academic hand-wringing. False positives drag whole chains behind them — needless antibiotics (vancomycin, again and again), longer stays, fatter bills — and labs that apply exactly these criteria have measurably cut both the misclassification and the reflex vancomycin[8][15][9]. And on our side of the bench there's one number that is purely ours to own: the long-standing target of keeping contamination at or below 3%[2]. Sit above it for long and you don't have an interpretation problem — you have a phlebotomy and collection-training problem, and fixing it is the one lever that lives entirely in the lab.

The checklist for 2 a.m.

When the bottle flags and the stain is drying, ask, in order:

1. Who showed up? Left-column bug → report and repeat. Right-column → keep reading.
2. How many bottles, and did they agree? Lone discordant skin commensal organism → ~98% likely contaminant — and say so in the comment.
3. How fast? Under 24 h leans pathogen; over 48 h leans contaminant — but remember the resistant slow-growers.
4. Whose blood is this? Lines, valves, neutropenia, a neonate, sepsis signs → trust the bug more, and chase the ward for the history the slip didn't give you.
5. So: report, repeat, or suppress — and when you genuinely can't tell, sign it out as equivocal, recommend a repeat set, and pick up the phone before the ward reaches for broad-spectrum cover on your word[13][2][8].

Read the bottle as a probability, not a verdict, sign the report you can defend at the morning round, and most of those calls will answer themselves long before the sun — or the susceptibilities — come up.

References

Bench to Bedside translates what the microbiology lab sees into what the ward should do. If a result has ever made you pause at the night bench — this is the place for you.