We read with interest the article by Barlow et al.1 The CURB-65 criteria currently recommended by the British Thoracic Society (BTS) based on the study by Lim et al2 are useful and more pragmatic than other criteria, as shown in their study. However, CURB has a lower specificity in older patients and addition of the age 65 criterion to CURB adds nothing to the sensitivity and little to the specificity in hospitalised patients.3 Increasing the urea cut off point by 2 units produces better specificity but at the expense of reduced sensitivity.4 Recently proposed SOAR criteria (systolic BP, oxygenation, age and respiratory rate) are at least as useful as CURB-65 in older patients3 but require additional information on arterial and inspired gas oxygen tensions. Better rules are therefore required for populations that include elderly patients. We hypothesised that using (1) age 85 as another cut off level and (2) two levels of urea cut off points at 7 mmol/l and 11 mmol/l in the scoring system would improve the assessment of severity in community acquired pneumonia (CAP).

We therefore modified CURB-65 and formulated a new rule (CURB-age) where:

• the presence of new confusion scores 1;

• urea >7 mmol/l but ⩽11 mmol/l scores 1;

• urea >11 mmol/l scores 2;

• respiratory rate ⩾30/min scores 1;

• either diastolic blood pressure ⩽60 mm Hg or systolic blood pressure <90 mm Hg scores 1;

• age ⩾65 and <85 scores 1;

• age ⩾85 scores 2.

Since the maximum possible score becomes 7, we defined severe pneumonia as a score ⩾4 for the CURB-age criteria compared with ⩾3 for CURB-65.

The subjects were 189 patients (median age 75 years, range 17–96, 56.1% men) who were included in two prospective observational studies of CAP.3 Detailed methodology has been reported previously.3 5 Using CURB-65 there were 109 non-severe cases (57.7%) and 80 severe cases (42.3%) and by CURB-age criteria there were 125 non-severe cases (66.1%) and 64 severe cases (33.9%). There were 5 deaths in each of the non-severe groups and 22 deaths in each of the severe groups. We examined the sensitivity, specificity, positive predictive values (PPV) and negative predictive values (NPV) of 6 week mortality and their corresponding 95% Pearson-Clopper exact confidence intervals for both CURB-65 and CURB-age criteria (table 1). The CURB-age criteria showed a significantly higher specificity (p = 0.0001, McNemar test).

A simple modification improves the specificity and PPV without losing the sensitivity of CURB-65 criteria and without requiring any additional information. It is as simple as CURB-65 and provides higher accuracy in identifying those who died over SOAR and CURB-65 criteria with significantly higher specificity. We combined the data from two cohorts of patients with CAP from two time periods, with the second cohort being elderly patients only (⩾65 years). It is reassuring that the CURB-age criteria better identified severe pneumonia in this older cohort. In the study in which the CURB-65 criteria were developed and validated, the median age of patients was 64 years.2

Our findings have important clinical implications. The current BTS guidelines recommend that severe CAP should be treated with intravenous antibiotics. These are more likely to produce untoward side effects such as antibiotic-associated diarrhoea than oral antibiotics, especially in older adults, and their use should be limited to truly severe CAP in older patients. Although the number of patients in our study is comparable to the original validation cohort reported by Lim et al2 (189 vs 214), larger studies are needed to test the validity of these modified criteria.