Risk stratification for short‐term mortality at hospital admission for acute exacerbations of COPD

Abstract

Background and objective: Exacerbations of chronic obstructive pulmonary disease (ECOPD) are associated with increased in-hospital and short-term mortality. Developing an easy-to-use model to predict adverse outcomes will be useful in daily clinical practice and will facilitate management decisions. We aimed to assess mortality rates and potential predictors for short-term mortality after severe ECOPD. Classification and Regression Tree (CART) model was used to identify predictors of adverse outcome. Methods: A retrospective observational cohort study, including all patients admitted to Maastricht University Medical Center with ECOPD between June 2011 and December 2014 was performed. The last admission was taken into account, and its demographic, clinical and biochemical data were recorded. Results: A total of 364 hospitalized patients were enrolled. Mean (SD) age was 70.5 (10.2) years, 54.4% were male and mean FEV 1 45.2% (17.7) of predicted. The in-hospital and 90-day mortality were, respectively, 8.5 and 16.2%. Independent risk factors for 90-day mortality were: PaC0 2 (odds ratio (OR): 1.31; 95% confidence interval (CI): 1.00-0.35), age (OR: 1.09; CI: 0.06-0.11), body mass index (BMI) < 18.5 kg/m 2 (OR: 2.72; 95% CI: 0.53-1.47) and previous admission for ECOPD in last 2 years (OR: 1.29; 95% CI: −0.14, −0.65). The CART model selected PaCO 2 ≥ 9.1 kPa, age > 80 years, BMI < 18.5 kg/m 2 and previous admission for ECOPD as the most discriminatory factors. Conclusion: According CART analysis, high PaCO 2 and age, low BMI and previous admission for ECOPD in last 2 years were the strongest predictors of 90-day mortality in patients with severe ECOPD. In absence of any of these factors, no patients died, suggesting that this model indeed enables risk stratification.Data presented as n (%) or medium with standard deviation depicted after AE, unless otherwise stated. † (Pearson) chi-square. ABG, arterial blood gas; ALAT, alanineaminotranferase; ASAT, aminotransferase; BMI, body mass index; BP, blood pressure; CRP, C-reactive protein; CVP, central venous pressure; ECG, electrocardiogram; FiO 2 , fraction of inspired oxygen; LDH, lactic acid dehydrogenase; Na, sodium; NT proBNP, brain natriuremic peptide; SaO 2 , oxygen saturation measured in arterial blood gas; SpO 2 , oxygen saturation measured by pulse oximetry; WBC, white blood cell.Respirology (2019) 24, 765-776

Contemporary Concise Review 2019: Chronic obstructive pulmonary disease

“…For instance, Sprooten et al, in a contemporary cohort of patients hospitalized with COPD, showed that despite the progress in the management of COPD exacerbations over the past 30 years, severe COPD exacerbations remain highly lethal with 1 in 10 patients dying in hospital from complications related to their exacerbation (and 16% dying within 90 days of hospitalization) and 1 in 3 patients requiring mechanical ventilation for hypercapnic respiratory failure. 3 Importantly, the investigators found that these adverse events could be accurately predicted by measuring the patient's arterial carbon dioxide tension (PaCO 2 ) at the time of hospital admission. Mortality (from ventilatory failure) could be avoided (with modern treatment) when patients' PaCO 2 was less than 60 mm Hg, whereas mortality was extremely high when patients' PaCO 2 was 70 mm Hg or greater.…”

What's in the CART? A decision rule to predict short‐term mortality during COPD exacerbations

“…The severity of COPD exacerbations is largely determined by dyspnoea, and the most effective (life‐preserving) therapies are aimed at reducing the work of breathing and enhancing ventilation . Thus, Sprooten et al ’s finding that significant hypercapnia (at a threshold of 9.1 kPa or 69 mm Hg) was the leading predictor of mortality and that almost all survivors had PaCO 2 levels below 8.2 kPa (61 mm Hg) aligns with clinical experience . This is a relatively simple message but one that has important implications.…”

“…In a recent publication in Respirology , Sprooten et al address these concerns of existing decision rules by applying a method called Classification And Regression Tree (CART) to predict the risk of in‐hospital, as well as 90‐day, mortality in patients who were hospitalized for AECOPD . As noted previously, the current off‐the‐shelf decision rules are largely based on well‐established statistical principles such as multivariate regression analysis.…”

“…mortality) and chooses only those that have a high impact, discarding all the others. In Figure 2 of Sprooten et al ’s paper, the classification tree generated by the CART algorithm is shown for the exacerbation data set. The machine‐learning algorithm identified arterial carbon dioxide tension (PaCO 2 ) as the most important splitter variable, followed by age, body mass index (BMI) and a history of COPD hospitalization over the previous 2 years.…”

“…There is another important message in Sprooten et al ’s paper: severe COPD exacerbations remain largely undefeated. In Sprooten et al ’s contemporary cohort, nearly 1 in 10 patients died in hospital, and 16% died within 90 days of hospital admission.…”