Physician judgement is a crucial adjunct to pneumonia severity scores in low-risk patients

RESULTS

The authors studied 1,472 patients with CAP over a 3-yr period. After exclusion of patients with CURB-65 scores 2–5, 565 patients were classified into CURB-65 score 0–1. 145 patients were discharged within 12 h of admission to hospital and 420 patients were admitted to hospital with mild CAP (CURB-65 score 0–1).

The demographic characteristics of the study groups are shown in table 1. There was a greater proportion of patients with chronic obstructive pulmonary disease (COPD) and chronic cardiac disease (COPD 21.2% versus 9.0%, p=0.01, and chronic cardiac disease 13.1% versus 4.8%, p=0.02) in those admitted to the hospital for >12 h.

Table 2 compares parameters of CURB-65 between the two groups. The only significant difference was that patients admitted to the hospital for >12 h had a respiratory rate of ≥30 breaths·min⁻¹. Small numbers of patients were discharged having had abnormal values for respiratory rate and systolic blood pressure on admission (table 2). In the majority of cases, these values improved with initial treatment (such as fluid management, nebuliser, analgesia and reassurance). Three out of the five patients with a respiratory rate of ≥30 breaths·min⁻¹ on admission had a diagnosis of COPD and received initial bronchodilator treatment. Of those patients with low systolic blood pressure <90 mmHg, two were young female patients with no other markers of systemic illness and one further patient's blood pressure improved over a few hours with initial treatment.

Despite similar CURB-65 scores, hospitalised patients often had multiple additional markers of severity. Table 3 compares other clinical and investigative parameters in both the groups. Patients admitted to the hospital for >12 h had an increased frequency of hypo- or hyperthermia, multilobar chest radiograph involvement, hyponatraemia, hypo- or hyperkalaemia, low albumin, acidosis and raised C-reactive protein levels, and were more hypoxaemic.

Reasons for hospitalisation of low-risk patients

A consensus among the reviewers was achieved in 100% of cases for assigning reasons for hospitalisation. The proportion of patients in each category is shown in figure 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1–

Reasons for hospital admission in patients with low CURB-65 (confusion, urea >7 mmol·L⁻¹, respiratory frequency ≥30 breaths·min⁻¹, systolic blood pressure <90 mmHg or diastolic blood pressure ≤60 mmHg, and age ≥65 yrs) scores.

Clinician concern

A significant proportion of hospitalised patients (39.3%) were classified as having high-risk clinical features requiring observation in hospital (clinician concern). Hypoxia requiring oxygen therapy was most common (31.4%) but other common reasons were unstable comorbidities (16.4%). These included ischaemic heart disease (acute coronary syndrome, arrhythmia and unstable angina), COPD, diabetic emergencies, severe anaemia, alcohol withdrawal and urinary retention.

5.2% had a secondary pneumonia complication such as parapneumonic effusion or lung abscess and 5.0% had metabolic abnormalities such as severe hyponatraemia (Na⁺ <130 mmol·L⁻¹), hypo- or hyperkalaemia (K⁺ <3.5 or >5 mmol·L⁻¹), or acute kidney injury requiring intravenous fluid therapy or in-patient hospital monitoring. Note that percentages are expressed as a proportion of the overall cohort of hospitalised patients (n=420). Percentages sum to >39.3% because some patients had more than one high-risk feature requiring hospitalisation.

Requiring additional investigations or treatment not related to severity of pneumonia

This group comprised 29.5% of the overall cohort and included patients that were hospitalised with none of the adverse features of severity. This group included patients requiring further investigations such as computed tomography scanning, ultrasound or bronchoscopy, where these investigations were not available rapidly as an outpatient service. 87.1% of these patients were discharged within 48 h of admission.

Unmet social needs

In 11.9% of cases, the reviewers could not identify any adverse clinical parameters to necessitate admission to hospital, but social circumstances were reported that would have made discharge from hospital inappropriate. These included elderly patients living alone without social support, patients whose home circumstances were unsafe or unsuitable and patients who were homeless.

No clear reason for hospitalisation

In 19.3% of cases, the reviewers could not identify a medical or social reason for the patient to be hospitalised.

CURB-65 was documented in 17.6% of patients admitted to hospital, compared with 26.9% of patients discharged from hospital (p=0.02). Of those where no clear reason could be established for hospitalisation, the CURB-65 score was documented in 17.2% of patients (p=0.1 when compared with patients discharged from hospital).

Secondary outcomes

Overall 30-day mortality was found to be 3.1% in those that were admitted (>12 h) compared with 0.7% in those that were discharged (≤12 h) (p=0.1). The one patient who died following discharge was re-admitted and died on day 24 from a cause unrelated to CAP.

In patients classified by the reviewers as “clinical concern”, mortality was 6.7%. This was significantly higher compared with 0% for patients awaiting investigations (p=0.009) and 0% for patients without a clear indication for hospitalisation (p=0.04). Mortality for patients with unmet social needs was 4.0% (two patients).

In a logistic regression analysis, parameters associated with 30-day mortality were chronic cardiac co-morbidity (adjusted odds ratio (aOR) 5.73, 95% CI 1.52–21.6; p=0.01); acidosis (pH<7.35) (aOR 5.14, 95% CI 1.44–18.3; p=0.01); hypoxia (aOR 9.86, 95% CI 2.39–40.7; p=0.002); and multilobar chest radiograph shadowing (aOR 4.54, 95% CI 1.21–17.1; p=0.03). The Hosner–Lemeshow goodness-of-fit test was p-value 0.8.

Mortality increased in low-risk patients according to the number of these additional adverse features (no additional adverse features 0.3% 30-day mortality (359 patients), one adverse feature (151 patients) 2.6%, two adverse features (45 patients) 11.1% and three adverse features (10 patients) 33.3%). No patient had all four additional adverse features. The mortality data are shown in figure 2.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2–

Additional severity markers and risk of mortality in patients with low CURB-65 (confusion, urea >7 mmol·L⁻¹, respiratory frequency ≥30 breaths·min⁻¹, systolic blood pressure <90 mmHg or diastolic blood pressure ≤60 mmHg, and age ≥65 yrs) scores with community-acquired pneumonia.

DISCUSSION

This was the first study to investigate reasons for hospitalisation among low-risk patients with CAP in an institution using the CURB-65 score. In this study, 74.3% of the patients reviewed who had low-risk pneumonia were admitted to the hospital. Most frequently, the reasons for hospitalisation were due to additional markers of severity, such as hypoxia or unstable comorbidities. Similar proportions were not severely ill but required additional investigations. It is possible that this group of patients could have been managed as outpatients if additional resources, such as rapid access to outpatient investigations, early clinic review or “hospital at home” style support were available. We did, however, identify 19.3% of patients who independent reviewers felt could have been safely managed as outpatients. Our results are similar to a study in the USA using the PSI, where 82% of low-risk CAP patients admitted had clinically justified reasons for hospital admission [23]. Potentially, therefore, by encouraging greater use of outpatient management and by providing additional resources for patients managed in the community, a significant number of hospital admissions for CAP could be avoided. Outpatient management of selected patients with CAP is safe, acceptable to patients and may be associated with significantly reduced hospital costs and complications [4].

In this study, where clinicians identified additional risk factors for mortality (such as hypoxia, acidosis, multilobar chest radiograph involvement or cardiac comorbidities), mortality was increased. The mortality rates for patients in whom clinicians identified additional risk factors (6.7%) were similar to those quoted for CURB-65 score 2 in some previous studies [12]. A small proportion of patients had more than one of these severity markers and had mortality rates equivalent to those of severe CAP.

CURB-65 score is one of the most widely used pneumonia prediction scoring systems in the world. It was validated primarily to predict 30-day mortality but has recently been recommended by national guidelines to help clinicians assess the need for hospital admission [11]. Although this study is the first to investigate reasons for hospitalisation using the CURB-65 score, others have studied reasons for hospitalisation of low-risk patients using the PSI. Aujesky et al. [16] studied 689 low-risk patients during a clinical trial that aimed to increase the proportion of patients treated in the community. In this study, the major reasons for hospitalisation were related to co-morbidities (71.5%) while patients with additional markers of severity (similar to the clinician concern category in this study) accounted for 29.3% of cases. Arnold et al. [23] showed that the majority of patients admitted with CAP with PSI I–II had extenuating clinical circumstances to justify their admission. Disease comorbidities (43%) and unmet social needs (18%) were the major categories of clinical factors justifying hospital admission for these patients. Used as the sole indicator for inappropriate hospitalisation, the PSI had a poor positive predictive value of only 16%. The corresponding value for the present study is very similar (19.3%). Hypoxia was not a major contributory factor in the study by Arnold et al. [23], perhaps because the PSI already incorporates oxygenation as a prognostically important factor to predict disease outcome. The study by Aujesky et al. [16] excluded patients with arterial hypoxaemia. In addition, because age >65 yrs contributes 1 point to the CURB-65 score, only one more abnormal variable is required to classify an elderly patient as requiring in-patient care. This may explain why our low-risk cohort was relatively young and why previous studies consistently showed that the PSI identifies a high proportion of patients as low risk [4].

Part of the objective of this study was to study the role of clinical judgement in applying the CURB-65 score in clinical practice. Our study suggests that the CURB-65 score may be under-utilised in clinical practice, as it was not documented in most of these low-risk patients. Out of the 19.3% patients hospitalised with no obvious justified reasons, only 17.2% had documented CURB-65 score in the notes. Had it been used, more in this group in conjunction with clinical judgement, we anticipate most of these patients could have been discharged.

However, in most cases, clinicians appeared to appropriately identify patients with additional risk factors not included in the CURB-65 score and these patients were justifiably admitted to hospital. Where clinicians identified a cause for concern, mortality was significantly increased above the level predicted by the CURB-65 score, to a level that requires hospital treatment based on current guidelines.

This study, however, has its limitations. Although patients were prospectively recruited, we determined reasons for hospitalisation retrospectively and this approach has inherent difficulties. We accounted for potential bias by using two blinded reviewers with a third independent reviewer where consensus was not reached. A similar study using a prospective design is desirable.