Study design and population

This retrospective observational study included all adult patients hospitalized with CAP between January 1, 2018 and December 31, 2023 at Flinders Medical Centre and Royal Adelaide Hospital in the Adelaide metropolitan area, South Australia.

Ethics

Ethical approval was obtained from the Southern Adelaide Human Research Ethics Committee and the Central Adelaide Human Research Ethics Committee (18887).

Data collection

Data for this study were extracted from the electronic medical records (EMRs). CAP was identified by the International Classification of Diseases and Related Health Problems, Tenth Revision, Australian modification (ICD‑10 AM) codes (J12 to J18.9).¹⁰ We specifically excluded patients who were identified as COVID‑19–positive on ICD‑10‑AM coding (U07.1 or J12.82). Hospital‑acquired pneumonia cases, defined as symptoms emerging 48 hours or longer after hospital admission or within 7 days postdischarge, and patients who tested positive for SARS‑CoV‑2 infection (ascertained through positive polymerase chain reaction tests) were also excluded.

We collected demographic and clinical information, including comorbidities, Charlson comorbidity index (CCI), frailty status assessed by the Hospital Frailty Risk Score (HFRS),¹¹ and socioeconomic status (SES) evaluated using the index of relative socioeconomic disadvantage (IRSD), with higher values indicative of better SES. The nutritional status was assessed using the Malnutrition Universal Screening Tool (MUST).¹² Patients with an HFRS of 5 or higher were classified as frail while those with a MUST score of 1 or higher were classified as malnourished, as previously reported.¹¹ Pneumonia severity was assessed using the CURB‑65 (confusion, uremia, respiratory rate, BP, age ≥65 years) score on admission,¹³ which includes the following variables: confusion, urea levels higher than 7 mmol/l, respiratory rate of 30/min or more, systolic blood pressure below 90 mm Hg or diastolic blood pressure of 60 mm Hg or lower, and age of 65 years or more. Severe respiratory failure was defined as the requirement for advanced respiratory support, including high‑flow oxygen therapy (defined as the need for 100% humidified oxygen delivered at flow rates up to 60/l/min), noninvasive ventilation, or invasive mechanical ventilation.

Vitamin C levels measured during hospitalization were extracted from EMRs. Measurements were conducted in a central laboratory using samples protected from light by an aluminum foil and transported on ice. Vitamin C levels were determined using high‑performance liquid chromatography with a detection limit of 3 μmol/l, and a reference range (RR) of 28–114 μmol/l. Levels below 11 μmol/l indicated vitamin C deficiency. Concentrations below this threshold indicate severe vitamin C deficiency and are more likely to be associated with scurvy.¹⁴ Additional extracted parameters at the time of admission included: hemoglobin (in g/l; RR, 135–175 g/l for men and 115–155 g/l for women), C‑reactive protein (CRP; in mg/l; RR, 0–8 mg/l), white blood cell count (measured as the number of cells × 10⁹/l; RR, 4–11 × 10⁹/l), platelet count (measured as the number of cells × 10⁹/l; RR, 150–450 × 10⁹/l), creatinine (in μmol/l; RR, 45–90 μmol/l), and albumin (in g/l; RR, 34–48 g/l). The neutrophil‑to‑lymphocyte ratio, which has been associated with poor clinical outcomes in CAP, was calculated by dividing the neutrophil count by the lymphocyte count measured on admission. The primary aim was to determine the prevalence of vitamin C deficiency among patients hospitalized with CAP. Secondary aims were to evaluate whether vitamin C status was associated with differences in length of hospital stay (LOS), risk of ICU admission, duration of ICU stay, development of severe respiratory failure, need for vasopressor support, in‑hospital and 30‑day mortality, and 30‑day readmission rates according to vitamin C status.

Statistical analysis

The normality of data was determined using the Shapiro–Wilk test. Descriptive statistics were used to summarize the data. Continuous variables were presented as mean (SD) or median (interquartile range [IQR]) depending on distribution, and compared using the t tests or the Wilcoxon rank‑sum tests, as appropriate. Categorical variables were summarized as frequencies and percentages, and compared using the χ² tests. Correlation between continuous variables (eg, plasma vitamin C and CRP concentrations) was assessed using the Spearman rank correlation coefficient, given the non‑normal distribution of the data. A multilevel Poisson regression model with a random intercept for hospital was used to assess hospital and ICU LOS according to vitamin C status, accounting for hospital level variance. The incidence rate ratio (IRR) quantified the relative change in LOS associated with vitamin C deficiency. For count outcomes, such as LOS (measured in days), the IRR reflected the multiplicative change in the expected number of hospital days. An IRR greater than 1 indicated a longer expected LOS, whereas an IRR below 1 indicated a shorter expected LOS. No specific threshold was used to define a “long” stay; instead, the differences in LOS were interpreted on a continuous scale.

Categorical outcomes, including ICU admission, mortality, and readmissions were analyzed using multilevel logistic regression models with random intercepts for hospital to account for clustering within hospitals. The models were adjusted for age, sex, CCI, CURB‑65 score, HFRS, IRSD, CRP and albumin levels, and time‑to‑sampling of blood for the vitamin C test. In addition, sensitivity analysis was performed by restricting analyses to patients who had moderate‑to‑severe pneumonia (CURB‑65 score ≥2) and using vitamin C as a continuous variable. Statistical significance was set at P below 0.05, and analyses were performed using STATA software version 18.0 (Stata Corp., College Station, Texas, United States) and Python version 3.12.4 (Python Software Foundation, Fredericksburg, Virginia, Unites States).

Outcomes

Among the patients hospitalized with CAP, those with vitamin C deficiency demonstrated differences in several clinical outcomes, as compared with those without deficiency.

Median (IQR) LOS was longer in the vitamin C–deficient patients than the nondeficient individuals (10.4 [5.1–19.5] vs (8.7 [4.6–13.9] d), though this difference was insignificant in the unadjusted analysis. However, after adjustment for age, sex, comorbidities, frailty, pneumonia severity, and biochemical parameters, vitamin C deficiency was independently associated with a 23% longer hospital LOS (IRR, 1.23; 95% CI, 1.06–1.42; P = 0.005; Table 1).

Admission to the ICU occurred in 12.2% of the vitamin C–deficient patients, though this value was similar as the one recorded for the nondeficient group (P = 0.41). In adjusted regression, there was an insignificant trend toward increased odds of ICU admission among the vitamin C–deficient patients (odds ratio [OR], 18.09; 95% CI, 0.87–372.89; P = 0.06; Table 1).

ICU LOS was longer in the vitamin C–deficient individuals. Median (IQR) ICU stay was 102 (92–395) hours for the deficient patients, as compared with 41.5 (30–481.5) hours for the nondeficient individuals. This difference was significant in both unadjusted (IRR, 1.74; 95% CI, 1.61–1.89; P <⁠0.001) and adjusted analyses (IRR, 3.07; 95% CI, 2.18–4.3; P <⁠0.001; Table 1).

Multivariable models did not identify significant associations between vitamin C deficiency and mortality or readmissions, though point estimates suggested higher odds of mortality and lower odds of readmissions among the deficient patients (Table 1). Similarly, there were no differences in the risk of severe respiratory failure or need for vasopressor support between the 2 groups (Table 1).

Sensitivity analysis

Sensitivity analysis confirmed that total hospital LOS was considerably prolonged among the patients with moderate‑to–severe pneumonia (CURB‑65 score ≥2; OR, 1.39; 95% CI, 1.05–1.85; P = 0.02). In addition, when analyzed as a continuous variable, higher plasma vitamin C levels were associated with shorter hospital LOS, with each 1-µmol/l increase in vitamin C corresponding to a 0.3% reduction in LOS (IRR, 0.997; 95% CI, 0.995–0.998; P <⁠0.001).

Limitations

This study has several limitations. First, vitamin C testing was performed in only a small subset of the CAP patients, and selection bias may have influenced the results, as those tested may have differed systematically from those not tested. Second, vitamin C levels were measured at a median of 7 days after admission, a time when they may have already declined due to hospitalization and acute illness. While we adjusted for time‑to‑testing in multivariable models, this timing may have still limited the accuracy of deficiency classification. Third, we lacked information on vitamin C supplementation before or during hospitalization, which may have influenced plasma levels. Additionally, although we adjusted for several important confounders, residual confounding cannot be excluded due to the retrospective study design and small sample size. Importantly, for binary outcomes with a low number of events, multivariable logistic regression models yielded wide confidence intervals and imprecise effect estimates, likely due to a low events‑per‑variable ratio. These findings should therefore be interpreted with caution and considered exploratory. Further studies with larger cohorts are warranted to validate these associations and confirm the role of vitamin C in this clinical context.

Conclusions

This study suggests that nearly half of hospitalized CAP patients present with vitamin C deficiency, a condition associated with poor clinical outcomes. Further research is needed to confirm these findings and to assess whether vitamin C supplementation could improve outcomes in CAP patients.