COMMENT|ONLINE FIRST

In-hospital mortality of patients with severe bloodstream infection: only the tip of the iceberg

Mathias W Pletz, Stefan Hagel, Sebastian Weis

Lancet Infect Dis Published:January 19, 2022

DOI:https://doi.org/10.1016/S1473-3099(21)00619-8

Many studies on severe infections use in-hospital mortality, 14-day mortality, or 30-day mortality, as classic endpoints. However, in the era of high-performance intensive care, these endpoints no longer seem appropriate. Modern intensive-care medicine has massively reduced early mortality in patients with sepsis.¹ Although this reduction is gratifying, it should not obscure the fact that these patients are still far from being healed.A reduction in early mortality might even mask differences in antibiotic efficacy. In the MERINO trial,² which revealed that mortality in patients with extended-spectrum β-lactamase bloodstream infections was nearly three times higher in people treated with piperacillin–tazobactam than in those treated with meropenem, most patients did not die during antibiotic therapy, and instead died after the (assumed to be successful) completion of such therapy. Also, the few studies that had long-term follow-up of patients after severe infection (ie, pneumonia³ and sepsis⁴) showed substantially higher all-cause mortality in the months after infection compared with during treatment, correlated with the initial severity of the infection. The reasons for this increased long-term mortality are still poorly understood. Current concepts argue that the cause is relapsing infections and superinfections caused by immunoparalysis, or an altered microflora, or cardiovascular events caused by inflammation, or a combination of these. However, patients with bloodstream infections often have severe chronic comorbidities that are generally associated with a poor overall prognosis. For example, in a US multicentre study in patients with sepsis, the most common underlying causes of death were solid cancer (21·0%), chronic heart disease (15·3%), haematological cancer (10·3%), dementia (9·7%), and chronic lung disease (9·0%). Conditions that qualify for hospice care (most commonly end-stage cancer) were present on hospice admission in 121 of 300 (40·3%; 95% CI 34·7–46·1) patients whose death was associated with sepsis, which contributed to the high long-term mortality rates in sepsis.⁵The BLOOMY scores⁶ are therefore an important step into the right direction, as they capture not only 14-day mortality risk, but also 6-month mortality risk in patients hospitalised with bloodstream infections. By contrast with the widely used Pitt Bacteraemia Score,⁷the BLOOMY scores draw on a large cohort of 2334 patients, use the strengths of modern data-analysis methods such as machine learning, and were validated in a subsequent cohort of 1023 patients. In the study,⁶ the BLOOMY 14-day score was not significantly different to the Pitt Bacteraemia Score (AUROC 0·83 BLOOMY 14-day vs 0·78 Pitt Bacteraemia Score; p=0·1170), whereas there was a significant difference with the BLOOMY 6-month score (0·73 BLOOMY 6-month vs 0·61 Pitt Bacteraemia Score; p=0·0012). This result is noteworthy, because for the past 3 decades, the Pitt Bacteraemia Score has been considered the gold standard by which to measure acute severity of illness and predict mortality in patients with bloodstream infections.⁸By contrast with the established scoring systems for sepsis (ie, Sequential Organ Failure Assessment [SOFA], quick SOFA, APACHE II) and pneumonia (ie, Pneumonia Severity Index, CURB, CURB-65), which focus exclusively on host factors, the new BLOOMY scores showed that pathogen species and resistance also have an effect on mortality. This addition is another important advantage of the BLOOMY scores, which thus bridge the gap between a rather host-centred view of intensivists and a more pathogen-focused view of microbiologists; an excellent example of the value of an interdisciplinary infectious diseases approach for these patients.The BLOOMY scores have disregarded only one easily measured risk factor: the patient respiratory rate. Increased respiratory frequency is very easy to detect clinically, without a laboratory, and has been proven to be an important predictive parameter, not only in all scoring systems for pneumonia,⁶ but also in scoring systems for cross-system sepsis.⁹ It is highly likely that including this parameter would have further increased the performance of the BLOOMY scores.Nevertheless, the study authors have provided important data that might help to stimulate further research in bloodstream infections, not only to assess mortality in the short term, but also mortality in the long term. Further studies in infectious diseases need to systematically capture long-term mortality risk and causes of death after hospital discharge. Only then can the scientific infectious diseases community identify the underlying pathomechanisms and use them to develop new diagnostic and therapeutic strategies that will cure, in the long term, patients with bloodstream infections. The development of the BLOOMY scores is an important step on this path.