Eligible patients included those aged ≥ 18 years old who were admitted to a tertiary teaching hospital (1,200 beds) due to pneumonia between November 2014 and September 2015. CAP was defined as an acute community-acquired infection of the pulmonary parenchyma and was confirmed radiologically. The inclusion criteria comprised the presence of new chest radiographic infiltrates on admission, and at least one major or two minor criteria, as follows. The major inclusion criteria were cough, sputum, or fever (38°C), and the minor inclusion criteria were dyspnea, pleuritic chest pain, pulmonary consolidation, and leukocytosis > 12,000/mL. We excluded patients who had been admitted to the hospital within the previous 90 days, immunosuppressed patients (patients with acquired immunodeficiency syndrome, or those receiving chemotherapy and/or corticosteroids [equivalent prednisone dose: 20 mg/day]), and those who had required regular hemodialysis (> 2 sessions per week) within the previous 30 days. All patients who were transferred from a nursing home were excluded to rule out healthcare-associated pneumonia. Patients classified as “do not resuscitate” were excluded. The patients were divided into two groups: an elderly group (age ≥ 65 years) and a non-elderly group (age < 65 years). Malnutrition was assessed by clinical nutritionists within 48 hours of admission according to our hospital protocol, which was structured based on the Academy of Nutritional and Dietetics consensus criteria [13,14]. No single parameter is definitive for adult malnutrition; therefore, patients who were identified as having two or more of the following six characteristics were diagnosed with malnutrition: insufficient energy intake (duration = week, amount = % of target calorie), weight loss (duration = week, amount = % loss), loss of muscle mass (quadriceps and deltoids), loss of subcutaneous fat (triceps, chest), localized or generalized fluid accumulation (ankle, sacrum, and ascites), and diminished functional status measured using handgrip strength. Subsequently, malnutrition was classified into five criteria according to International Classification of Disease 9 clinical modification codes such as kwashiorkor, marasmus, protein calorie malnutrition, moderate malnutrition, or mild malnutrition. Chronic malnutrition was defined as a patient who had a reduced calorie intake < 75% for > 1 month and/or who lost more than 5% of their body weight in more than 1 month.

Data were collected and analyzed retrospectively. The following information was reviewed from patient medical records: demographic data, pneumonia severity index (PSI), risk factors, variables associated with nutritional status, the Charlson comorbidity index (CCI), the metabolic index, and clinical outcome. The survival of all patients was confirmed at each time point by confirming the loss of health insurance qualification.

Statistical analyses were conducted using R software version 3.0.1 (http://www.R-project.org). Descriptive statistics included frequency analyses (percentages) for categorical variables and means, and standard deviations for continuous variables. To test for differences in clinical and laboratory findings according to age group or nutrition group, the Wilcoxon rank-sum test was used for continuous variables, as appropriate. The chi-square test or Fisher’s exact test was used for discrete variables. Cox regression analyses were performed to identify predictors of 2-year mortality. After univariate analyses, significant and borderline (p ≤ 0.05) values were entered into stepwise backward multivariate Cox regression analyses. Two-way interaction terms were tested among the remaining significant variables. To approximate a random assignment in the elderly, the propensity score matching method was used to create sets of patient groups with and without malnutrition specifically, which were adjusted for their ages and rates of intensive care unit (ICU) admission [15]. We used a genetic matching algorithm that automates the process of searching for a good match result [16,17]. A genetic search algorithm can find a set of weights for each covariate such that the optimal balance can be achieved after matching. A genetic matching method was used to obtain the optimal ratio between the two groups without the ratio being limited to 1:1 or 1:2. The process was repeated until we obtained datasets with no statistically significant differences for age and rates of ICU admission. A two-sided p < 0.05 was considered statistically significant.

This study was approved by the Institutional Review Board of Pusan National University Yangsan Hospital (05-2017-180). The requirement of informed consent from the patients was waived due to the retrospective nature of this study.

The characteristics of the patients according to age group are shown in Table 1. The overall prevalence of malnutrition was 39.4%. The proportion of patients with malnutrition was significantly higher (53.4% vs. 11.9%, p < 0.001) and body mass index (BMI) was significantly lower (21.1± 3.5 kg/m² vs. 22.7± 4.0 kg/m², p = 0.004) in the elderly group than in the non-elderly group. In particular, the proportions of patients with kwashiorkor (9.2% vs. 0%, p = 0.011), protein calorie malnutrition (16.8% vs. 1.5%, p = 0.001), and malnutrition of a mild degree (22.1% vs. 10.4%, p = 0.044) were significantly higher in the elderly group than in the non-elderly group. The ratio of chronic malnutrition was also significantly higher (24.4% vs. 3.0%, p < 0.001) in the elderly group than in the non-elderly group. No significant difference in the metabolic index for nutrition was observed between the two groups. The CCI score was significantly higher (5.7 ± 2.1 vs. 2.5 ± 1.9, p < 0.001), and the PSI score was significantly lower (82.7 ± 33.2 vs. 90.3 ± 24.2, p < 0.001) in the elderly group than in the non-elderly group. The proportion of patients with a PSI classification above IV was lower in the elderly group than in the non-elderly group (PSI class IV and V: 39% vs. 44.8%, p = 0.039).

The discharge rate (81.7% vs. 97.0, p = 0.005), intermediate survival rate (1-year survival: 75.6% vs. 91.0%, p = 0.009), and long-term survival rate (2-year survival: 67.2% vs. 85.1%, p = 0.007) were significantly lower in the elderly group than in the non-elderly group (Table 2).

Age, nutritional status, PSI, and CCI were evaluated to identify independent risk factors for 2-year mortality. Univariate analyses showed that elderly (≥ 65 years) (odds ratio [OR], 2.70; 95% confidence interval [CI], 1.26 to 5.38; p = 0.005), malnutrition (OR, 3.82; 95% CI, 2.16 to 6.76; p < 0.001), CCI score (OR, 1.37; 95% CI, 1.24 to 1.52; p < 0.001), renal replacement therapy (OR, 3.87; 95% CI, 1.64 to 9.13; p = 0.002), and BMI (OR, 0.89; 95% CI, 0.82 to 0.98; p = 0.012) were associated with 2-year mortality. All of these variables were included in the Cox regression analysis model. Multivariate analyses showed that malnutrition (OR, 2.52; 95% CI, 1.39 to 4.60; p = 0.002) and CCI score (OR, 1.30; 95% CI, 1.17 to 1.45; p < 0.001) were associated with 2-year mortality (Table 3).

The characteristics of the elderly patients according to their nutritional status are shown in Table 4. Before propensity score matching, the malnourished group was significantly older (77.2 years vs. 73.5 years, p = 0.001). BMI and serum levels of total protein were lower in the malnourished group than in the well-nourished group (BMI: 19.8 ± 3.0 kg/m² vs. 22.5 ± 3.5 kg/m², p < 0.001; serum levels of total protein: 6.6 ± 0.8 g/dL vs. 6.8 ± 0.8 g/dL, p = 0.046). Use of a vasopressor and mechanical ventilation rates were significantly higher in the malnourished group at ICU admission (ICU admission: 14.3% vs. 1.6%, p = 0.009; use of vasopressor: 11.4% vs. 1.6%, p = 0.027; mechanical ventilation: 12.9% vs. 1.6%, p = 0.016). We performed propensity score matching analyses in the elderly group to reduce the effects of potential confounders, such as age and disease severity.

After propensity score matching, the final analyses included 87 patients, composed of a malnourished group (n = 37) and a well-nourished group (n = 50) (Fig. 1). The BMI (19.9 ± 3.2 kg/m² vs. 22.1 ± 2.8 kg/m², p = 0.002) was significantly lower in the malnourished group. No significant differences were observed regarding the PSI score, PSI class, CCI score, or the metabolic index. No significant differences were observed regarding the rates for ICU admission, use of a vasopressor, and mechanical ventilation (Table 4).

The clinical outcomes following propensity score matching are presented in Fig. 2. No significant difference in short-term outcomes was observed according to the presence of malnutrition. The hospital length of stay and the hospital mortality rates were not different between the malnourished and well-nourished groups (hospital length of stay: 15.2 ± 12.8 days vs. 13.5 ± 9.3 days, p = 0.474; hospital mortality: 5.4% vs. 2%, p = 0.389). However, hospital discharge status was significantly different between the groups. In the malnourished group, 70.3% of patients were discharged to home, which was lower than the rate in the well-nourished group (98%, p < 0.001), and 29.7% were transferred to long-term care facilities, which was higher than the rate in the well-nourished group (2%, p < 0.001). Moreover, long-term survival was also significantly lower in the malnourished group than in the well-nourished group (1-year survival: 67.6% vs. 90.0%, p = 0.009; 2-year survival: 59.5% vs. 86%, p = 0.005) (Fig. 2).

Age, nutritional status, PSI, and CCI were evaluated to identify independent risk factors for 2-year mortality in the elderly (Table 3). Univariate analyses showed that age (OR, 1.05; 95% CI, 1.00 to 1.10; p = 0.033), malnutrition (OR, 4.07; 95% CI, 1.94 to 8.51; p < 0.001), CCI score (OR, 1.33; 95% CI, 1.17 to 1.51; p < 0.001), renal replacement therapy (OR, 3.23; 95% CI, 1.26 to 8.30; p = 0.015), and BMI (OR, 0.91; 95% CI, 0.82 to 1.00; p = 0.049) were associated with 2-year mortality in the elderly. All of these variables were included in the Cox regression analysis model. Multivariate analyses showed that malnutrition (OR, 3.06; 95% CI, 1.44 to 6.50; p = 0.004), age (OR, 1.06; 95% CI, 1.01 to 1.12; p = 0.030), and CCI score (OR, 1.37; 95% CI, 1.18 to 1.58; p < 0.001) were associated with 2-year mortality in the elderly.