Microbiology and prognostic prediction model of bloodstream infection in patients with hematological malignancies.

BACKGROUND

In recent years, with the continuous development of treatments for hematological malignancies (HMs), the remission and survival rates of patients with HMs have been significantly improved. However, because of severe immunosuppression and long-term recurrent neutropenia during treatment, the incidence and mortality of bloodstream infection (BSI) were all high in patients with HMs. Therefore, we analyzed pathogens' distribution and drug-resistance patterns and developed a nomogram for predicting 30-day mortality in patients with BSIs among HMs.

METHODS

In this retrospective study, 362 patients with positive blood cultures in HMs were included from June 2015 to June 2020 at West China Hospital of Sichuan University. They were randomly divided into the training cohort (n = 253) and the validation cohort (n = 109) by 7:3. A nomogram for predicting 30-day mortality after BSIs in patients with HMs was established based on the results of univariate and multivariate logistic regression. C-index, calibration plots, and decision curve analysis were used to evaluate the nomogram.

RESULTS

Among 362 patients with BSIs in HMs, the most common HM was acute myeloid leukemia (48.1%), and the most common pathogen of BSI was gram-negative bacteria (70.4%). The final nomogram included the septic shock, relapsed/refractory HM, albumin <30g/l, platelets <30×109/l before BSI, and inappropriate empiric antibiotic treatment. In the training and validation cohorts, the C-indexes (0.870 and 0.825) and the calibration plots indicated that the nomogram had a good performance. The decision curves in both cohorts showed that the nomogram model for predicting 30-day mortality after BSI was more beneficial than all patients with BSIs or none with BSIs.

CONCLUSION

In our study, gram-negative bacterial BSIs were predominant in patients with HMs. We developed and validated a nomogram with good predictive ability to help clinicians evaluate the prognosis of patients.