Factors affecting in-hospital mortality of non-tuberculous mycobacterial pulmonary disease

We examined data from the Diagnosis Procedure Combination (DPC) database in Japan from July 2010 to March 2014. The database includes administrative claims data and discharge abstract data from more than 1200 hospitals, which covered 50% of the total bed capacity of acute care hospitals during the survey period. Primary diagnoses and comorbidities are recorded using International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) codes, accompanied by text data in Japanese. The database also contains the following information: age, sex, body height and weight, smoking status, grade of activity of daily living expressed as Barthel index score on admission, discharge status, therapeutic procedures, and medication use during hospitalization. This study was approved by the Institutional Review Board of The University of Tokyo and was performed in accordance with the Declaration of Helsinki. The requirement for informed consent was waived because of the anonymous nature of the data.

We retrospectively extracted data for patients with ICD-10 code: A310 (pulmonary mycobacterial infection) and A319 (Mycobacterial infection, unspecified). We did not include patients who were diagnosed with NTM extra-pulmonary disease (ICD-10 code, A311 and A318) during the study period. We included only the last hospitalization of patients who required unscheduled hospitalization more than once during the study period. We excluded patients < 18 years of age and those with HIV infection (ICD-10 code, B20–B24). We identified comorbidities using ICD-10 codes, as shown in Table 1. Because it is difficult to distinguish between primary bronchiectasis and bronchiectasis secondary to NTM-PD [14], we did not include the comorbidity of bronchiectasis in our study.

To evaluate weight loss, we categorized BMI in accordance with the levels of severity of anorexia nervosa in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) [15].

We used either ICD-10 codes (R042, R048, and R049) or treatment with intravenous hemostatic agents (carbazochrome sodium sulfonate and/or tranexamic acid) as indicators of clinically significant hemoptysis or bloody sputum. We then excluded patients with diagnoses of hemorrhage from organs other than lungs (ICD-10 codes shown in Table 2) and those who had undergone endoscopic procedures for hemostasis of gastrointestinal tract bleeding. We considered that arterial embolization performed in patients with hemoptysis was bronchial artery embolization.

In addition, we extracted data concerning oral and intravenous treatment with various drugs, including corticosteroids and anti-mycobacterial agents. We evaluated prescription of the following antibiotics: rifamycins, including rifampicin and rifabutin; ethambutol; isoniazid; macrolides, including clarithromycin and azithromycin; aminoglycosides, including streptomycin, kanamycin, and amikacin; fluoroquinolones, including levofloxacin, moxifloxacin and sitafloxacin; and imipenem/cilastatin. We then assessed the combination of rifamycin, ethambutol, and clarithromycin, which is the recommended first line therapy for Mycobacterium avium complex pulmonary disease (MAC-PD) in Japan [16]. We also evaluated prescription of erythromycin, which has an anti-inflammatory effect in patients with chronic airway diseases.

We used χ² tests to compare baseline characteristics, treatments, and procedures during hospitalization between patients who died during hospitalization and those who survived. We performed multivariable logistic regression to identify risk factors for all-cause in-hospital mortality. We used multiple imputation by the chained equations technique to deal with missing data on BMI, smoking history, and Barthel index scores. We included all the variables analyzed in this study in the imputation model and created 20 imputed datasets. We fitted multivariable logistic regression analyses for in-hospital mortality with generalized estimating equations to account for within-hospital clustering [17]. We obtained one set of statistical results on each imputed dataset and integrated them using Rubin’s combination rules [18]. We set statistical significance at less than 0.05 for all analyses. We performed statistical analyses using Stata/MP version 14 (StataCorp, College Station, TX, USA).

We identified 16,280 patients with NTM-PD who required unscheduled hospitalization during the study period. We then excluded 24 patients aged < 18 years and 64 patients with HIV infection from the study, leaving 16,192 patients who were eligible for further analysis.

These patients characteristics are shown in Table 3. The median age was 78 years (interquartile range [IQR] 71–84) and 61.2% were women. The overall median BMI was 17.5 kg/m² (IQR 15.4–20.0), being 18.1 kg/m² (IQR 15.9–20.6) for men and 17.2 kg/m² (IQR 15.2–19.5) for women. The most common comorbidity was pulmonary infection other than NTM (50.8%). As for primary diagnosis during hospitalization, pulmonary diseases accounted for 68.3% (pulmonary infection 28.8%, NTM-PD 23.1%, and other pulmonary disease 16.4%). Malignant diseases and other non-pulmonary diseases accounted for 4.1 and 27.6%, respectively.

Treatments during hospitalization are shown in Table 4; 44.6% of the 16,192 patients had not received any antibiotics with antimicrobial activity against NTM during the hospitalization, 3.2% had received monotherapy with erythromycin, and 15.2% had been treated with combination therapy including rifamycin, ethambutol and clarithromycin. Corticosteroids were prescribed for 20.9% of all patients.

Overall, 3166 patients (19.6%) died during their unscheduled hospitalizations (Table 3). The median age of these patients was 80 years (IQR 74–85). The median BMI of all patients who died in hospital was 16.0 kg/m² (IQR 14.2–18.4), comprising 16.6 kg/m² (IQR 14.8–19.1) for men and 15.4 kg/m² (IQR 13.7–17.6) for women. The median length of hospital stay was 18 days (IQR 10–34) in all patients and 21 days (IQR 8–44) in patients who died during hospitalization. Two-thirds of the patients who required mechanical ventilation died during the hospitalization (Table 4).

Table 5 shows the results of the multivariable logistic regression analysis for all-cause in-hospital mortality. Higher in-hospital mortality was associated with male sex, lower BMI, lower Barthel index score, hemoptysis, and comorbidities, including pulmonary infection other than NTM, pulmonary aspergillosis, interstitial lung disease, pneumothorax, congestive heart failure, renal disease, and malignant disease.