CURB-65 is a five-point score for grading community-acquired pneumonia, a lung infection caught outside hospital, in adults. It gives one point each for confusion, a blood urea above 7 mmol/L, a breathing rate of 30 or more a minute, low blood pressure and age 65 or over, and comes from a 2003 paper by Lim and colleagues in Thorax. We checked the death rates that travel with it. A per-score set repeated in online calculators and references, running from 0.6% at a score of 0 to 27.8% at scores of 4 and 5, is not in that paper. The paper's own table counts 1 death among 137 patients at a score of 0 and 26 among 65 at a score of 4. Only 7 of the 718 patients in its main cohort scored 5.
The paper counts adults in UK, New Zealand and Dutch hospitals in the 1990s and 2000 who died within 30 days. Like the percentages beside the AF stroke score, its figures describe a cohort and forecast nothing for one person. Nothing here is medical advice.
Death rates in the 2003 abstract contradict the paper's own table
Lim's paper built the score on a derivation cohort of 718 patients, 69 of whom died, and tested it on a separate validation cohort of 214, with 20 deaths. Its Figure 2 defines confusion as "a Mental Test Score of 8 or less, or new disorientation in person, place or time". Low blood pressure means systolic (the higher number) under 90 mmHg or diastolic 60 or below.
Table 4 reports every score separately; the abstract puts score 4 at 41.5% and score 5 at 57%, where the table records 26 of 65 (40%) and 1 of 7 (14%). Its 3% at a score of 2, below its 3.2% at 1, looks like a lost digit. The paper never explains why its abstract and table differ. With 7 and 3 patients scoring 5, one death moves the percentage by 14 or 33 points.
Mortality figures repeated online match no cell in the paper
Lim's Table 4 counts sit beside the repeated set below; the combined column is our arithmetic.
| Score | Derivation deaths | Validation deaths | Both cohorts (ours) | Repeated set |
|---|---|---|---|---|
| 0 | 1 of 137 | 0 of 36 | 0.6% | 0.6% |
| 1 | 4 of 187 | 0 of 54 | 1.7% | 2.7% |
| 2 | 17 of 184 | 5 of 60 | 9.0% | 6.8% |
| 3 | 20 of 138 | 9 of 42 | 16.1% | 14.0% |
| 4 | 26 of 65 | 5 of 19 | 36.9% | 27.8% |
| 5 | 1 of 7 | 1 of 3 | 20.0% | 27.8% |
Only the score-0 figure coincides, with a total the paper never prints. The rest cannot be mistyped cells: 6.8% of 184 patients would be 12.5 deaths. We found the set in neither the paper nor the validation studies we opened, and could not trace its origin. Calculover.com, on a page updated in July 2026, already warns that six-row tables "circulate online but are not reliably traceable to the primary paper".
Until 18 October 2026 our CURB-65 pneumonia severity tool credited the set to Lim 2003 with scores 4 and 5 "pooled", gave each score a care setting, counted an empty urea box as no point and labelled it "BUN > 19 mg/dL", which would score a decimal 19.1 to 19.6 that Lim's rule does not. The revision that day shows Table 4's counts, switches to CRB-65 when urea is blank and takes three urea units. With age and breathing rate ticked and a BUN of 19.4 mg/dL, it gives a CURB-65 of 2 and 17 deaths among 184; a BUN of 20 gives 3, and 20 among 138.
A urea at the threshold works out at 19.6 mg/dL of urea nitrogen
UK and Singapore labs report urea in millimoles per litre (mmol/L), a count of molecules. US labs usually report blood urea nitrogen (BUN) in milligrams per decilitre (mg/dL), which weighs only its nitrogen. Each urea molecule, CO(NH₂)₂, carries two nitrogen atoms of 14.0067 g/mol, so 1 mmol/L of urea is 2 × 14.0067 ÷ 10 = 2.80134 mg/dL of BUN, and Lim's 7 mmol/L is 19.61 mg/dL. The medical site iatrox already prints 19.6.
Whole-number BUN readings trip the criterion within rounding of the millimole cut
On a whole-number BUN report, "above 19" and "20 or more" are the same cut, and both sit where 7 mmol/L falls. Rounding blurs this edge. A reported BUN of 20 mg/dL, for instance, can represent a true value anywhere from 19.5 to 20.5. On our check of 101 decimal BUN values from 15.0 to 25.0 mg/dL, "above 19" scored six that Lim's rule would not, from 19.1 to 19.6.
US wordings belong to a different rule. The 2019 ATS/IDSA guideline lists "blood urea nitrogen level ≥ 20 mg/dl" as a minor criterion for severe pneumonia, and its 2007 predecessor wrote ">20 mg/dL". The bigger trap is urea weighed whole, at 60.055 g/mol, where 7 mmol/L is about 42 mg/dL. Read against "BUN > 19", every whole value from 20 to 42, 23 in all, would score a point Lim's rule withholds.
Urea results from three Singapore public-hospital labs come in millimoles per litre
The test catalogues of SGH, NUH and SKH all give serum urea in mmol/L. NUH lists "Blood Urea Nitrogen (BUN), Serum" as an alias and codes the test LOINC 14937-7, urea nitrogen in moles per volume, so a Singapore "BUN" can be a mmol/L number. NICE in England words its criterion the same way: "raised blood urea nitrogen (over 7 mmol/litre)".
A reference interval describes healthy people; Lim's cut came from a study. NUH's interval for men aged 50 and over runs to 9.2 mmol/L, and SKH's adult interval, updated 17 June 2025, to 7.8. A result of 8.0 sits inside both and above Lim's cut. Private labs and polyclinics were not checked.
Bedside findings alone make up the shorter CRB score
Leave out urea and the score becomes CRB-65, from 0 to 4, needing no blood test; Lim's paper already reports it. Among its derivation cohort, 2 of 167 patients with a CRB-65 of 0 died, 37 of 455 with 1 or 2, and 30 of 96 with 3 or 4. A German study of 1,967 patients by Bauer and colleagues in 2006 found CURB, CRB and CRB-65 gave "comparable predictions". CRB-65 could be worked out for 90% of patients and CURB for 65%, because urea results were missing.
Risk bands kept from 2014 sit beside NICE's new place-of-care wording
NICE published guideline NG250 on 2 September 2025. It "updates and replaces" CG191 of December 2014, NG138 and NG139. For England, it uses CRB65 in primary care and CURB65 in hospital. The recommendations keep risk boxes tagged [2014]: a CURB65 of 3 to 5 is "high risk (more than 15% mortality risk)". The place-of-care wording is tagged [2025] and now names virtual wards and same-day emergency care. In the US, the 2019 ATS/IDSA guideline prefers the Pneumonia Severity Index.
NICE asks for clinical judgement together with the CURB65 score, "bearing in mind this can be affected by other factors, for example, comorbidities or pregnancy".
View author profile → · Editorial policy
