To the Editor: We note with interest the recent commentary by Cretikos et al on the predictive value of a high respiratory rate for adverse outcomes.1 We wish to provide empirical evidence from Australian patients with pneumonia in support of their view that simple clinical parameters are good predictors of adverse outcomes.

We examined data from a prospective cohort of consecutive patients presenting to the Royal Melbourne Hospital Emergency Department with radiologically confirmed, community-acquired pneumonia between 2003 and 2006.2 In an earlier study of a subset of these patients,3 we found that hypotension and tachypnoea were strongly associated with death and/or the need for respiratory/inotropic support (odds ratios, 8.0 and 3.5, respectively).

In the full cohort (n = 740), we examined factors associated with either admission to the intensive care unit (ICU) or mortality (106 patients were in one of these two categories). Respiratory rate was documented in 712 patients (96%). A combination of tachypnoea (≥ 24 breaths/min) and/or hypotension (systolic blood pressure ≤ 90 mmHg) had similar predictive value for the risk of ICU admission and/or death to the recommended system of risk stratification, the Pneumonia Severity Index (PSI)4 (Box). The combination of respiratory rate and systolic blood pressure performed better than either sign alone in ruling out at-risk patients if both were normal (ie, a high negative predictive value), although almost a quarter of patients with either clinical sign had adverse outcomes.

The PSI is based on 20 individual clinical and laboratory parameters, and evidence suggests that it is poorly documented in patients’ records.5 Our data relate to patients with community-acquired pneumonia from a single centre and thus have limited statistical power for making comparisons. However, they suggest that these two routinely measured clinical parameters can be used to stratify patients at risk of adverse outcomes at the time of presentation. We support efforts to incorporate simple clinical indicators into systems that can identify seriously unwell patients early in the course of illness.

To the Editor: Cretikos et al make a strong case for routinely recording patients’ respiratory rate (RR) in acute wards.1

In a prospective study designed to evaluate the prognostic value of RR in acutely hospitalised patients aged over 75 years, we analysed data from all admissions to a single 14-bed acute-care geriatric unit between 15 May and 15 November 2007. Clinical data were recorded on admission and blood tests were performed the next morning. The Charlson score was used to assess comorbidity.2

Features of the 195 admissions during the study period (13 patients were admitted twice) are reported in the Box. The main reasons for admission were falls (15%), left ventricular failure (11%), pneumonia (11%), cancer complications (9%), pyelonephritis (7%) and stroke (5%). Twenty-nine patients died in hospital, including six from cancer complications, five from pneumonia, four from pyelonephritis and four from left ventricular failure.

Based on univariate logistic regression analysis at the 0.05 significance level, the following variables were predictive of death during hospitalisation: being male (odds ratio [OR], 2.42, Wald test P = 0.03); increased Charlson score (OR, 1.50 for each additional point between 3 and 13; P < 0.001); decreased systolic blood pressure (OR, 6.71 for systolic blood pressure < 100 mmHg; P =0.004); abnormal heart rate (< 60 beats/min or > 100 beats/min) (OR, 3.63; P = 0.003); increased RR (OR, 1.81 for each additional 5 breaths/min between 14 and 44; P < 0.001); abnormal blood sodium level (< 137 mmol/L or > 143 mmol/L) (OR, 2.82; P = 0.01) and raised C-reactive protein level (OR, 2.67 for C-reactive protein level > 45 mg/L; P = 0.02).

After multivariate logistic regression analysis with stepwise backward elimination, the only remaining factors that were significant predictors of death during hospitalisation were Charlson score (OR, 1.53 for each additional point between 3 and 13; P < 0.001) and RR (OR, 1.83 for each additional 5 breaths/min between 14 and 44 breaths/min; P < 0.001).

RR on admission was associated with an area under the ROC (receiver operating characteristic) curve of 0.73 (95% CI, 0.64–0.82) for the prediction of death during hospitalisation. An RR of ≥ 20 breaths/min had a sensitivity of 97% (95% CI, 80%–100%), a specificity of 28% (95% CI, 21%–35%) and a negative likelihood ratio of 0.12 (95% CI, 0.02–0.82) for prediction of death during hospitalisation. Only one patient (admitted for stroke) with an RR of < 20 breaths/min at admission died in hospital. Higher RR cut-off values increased specificity but reduced sensitivity. For instance, an RR of ≥ 30 breaths/min had a specificity of 90% (95% CI, 84%–94%), but a sensitivity of 38% (95% CI, 21%–58%), leading to a positive likelihood ratio of 3.68 (95% CI, 1.93–7.04) for prediction of death during hospitalisation.

In conclusion, besides comorbidity (Charlson score), RR was the most useful predictor of death in acutely hospitalised patients aged over 75 years. Our results extend the evidence base promoting regular documentation of RR in acute-care departments.1

To the Editor: The concept that respiratory rate (RR) is a key vital sign is hardly new,1 but it is being re-emphasised.2,3 The recent article by Cretikos et al3 highlights the diagnostic relevance of a raised RR for serious adverse events. The authors make sensible recommendations regarding the need to educate hospital staff about the importance of measuring patients’ RR.

However, they do not specify how this rate should be measured. Indeed, most textbooks of general medicine, and even respiratory medicine, fail to provide guidance on this or to define an abnormal rate. A widely used book on clinical examination4 suggests measuring RR while feeling the pulse, and quotes a normal (adult) resting range of 16–25 breaths/min, but no source for this information is provided.

There are few reports of true normal resting RR measurements obtained by covert observation. Respiratory physiologists have long known that RR commonly increases and becomes more regular as soon as a subject becomes aware of the measurement. This is especially so if a mouthpiece is in place. Rates as low as 8 breaths/min may be seen at rest, and the normal adult range quoted by physiologists is 11–14 breaths/min.5 Bradypnoea is usually defined as a rate less than 8 breaths/min and tachypnoea as a rate greater than 18–20 breaths/min.

There is no gold standard method for accurate measurement of RR in clinical practice. Of course, when patients are being monitored, particularly with a nasal cannula, it should be easy to obtain the rate. Oximetry is not a surrogate measure of RR, although it is often easier to record. Inductance bands around the chest provide a simple non-invasive way to measure RR. In addition to the absolute rate, an irregular and erratic rate is of concern.

In settings in which formal monitoring is not being conducted, RR is the one “vital” sign that must be assessed when the patient is resting quietly, unaware of its measurement, and not conversing with staff. Duplicate measurements should be made over an interval of at least 1 minute. An RR of over 20 breaths/min, particularly if irregular, is noteworthy. Tachypnoea is intimately linked with the sensation of breathlessness, and hence the patient’s respiratory sensations should be assessed. In some patients with a normal RR at rest, marked tachypnoea may be precipitated by mild exertion, such as walking a few paces.