Five-Decade Prevalence of Delirium in Pneumonia, Risk Factors, and Associated Mortality: A Systematic Review and Meta-Analysis

Abstract

Background: Delirium can occur in patients with pneumonia, but its prevalence is inconsistent across studies. Unreliable estimates and uncertainty regarding the significance of patient-specific vs. microbiological risk factors hinder delirium management and prognosis. Here, we provide robust estimates of delirium prevalence in patients with pneumonia, associated risk factors, and association with mortality.

Methods: We searched five databases (MEDLINE, Cochrane Library, Embase, PsycINFO, and Scopus), from inception to August 6, 2024. We included studies in adults hospitalized with pneumonia reporting delirium, encephalopathy, or altered mental status. Two investigators extracted data and assessed risk of bias. Summary rates were calculated using random-effects models. We performed prespecified analyses for diagnostic methods, microbiologic factors, clinical factors, and mortality, with sensitivity analysis among studies at low risk of bias. Registration: PROSPERO-CRD42023385571.

Results: Delirium prevalence across 126 studies was 22% (95% CI [18%-26%]), and higher in studies at low risk of bias (40% [24%-58%], n=11). Standardized assessments yielded higher rates than symptom- or ICD code-based assessments (p<0.05). Surprisingly, delirium rates did not differ by microbiological etiology (p=0.63), including COVID-19, nor by pneumonia origin (p=0.14). Predisposing factors included older age and neurologic and systemic comorbidities. Delirium was associated with increased mortality (OR 4.3 [3.24-5.76], p<0.001), without change over five decades (p = 0.32).

Interpretation: Delirium is highly prevalent and enduring in pneumonia. Our results emphasize patient- and care-related factors over microbiological causes, including COVID-19. Delirium's entrenched association with mortality, even considering covariates, reinforces the need to manage delirium as a convergent syndrome in pneumonia.

Keywords: delirium; encephalopathy; hospitalization; pneumonia; systematic review.

Figure 1.

Preferred Reporting in Systematic Review and Meta-Analysis (PRISMA) Flowchart

Figure 2.

Summary estimate for occurrence of delirium in patients hospitalized with pneumonia. Effect measures were calculated using random-effects models and expressed as proportions. 126 studies were included, and the overall summary rate was 0.22 (95% CI [0.18–0.26]).

Figure 3.

Delirium rates differ based on assessment methods and are higher in studies performed in intensive care units, but do not differ based on pneumonia origin or microbiological etiology. A. Delirium rates varied significantly according to the assessment method used, which explained 37% of variance in delirium rates across studies (R² = 36.8%, p<0 .0001). Given that the Diagnostic and Statistical Manual (DSM) assessment is considered the gold standard for delirium assessment, we compared delirium rates to those of the DSM. Delirium rates were similar to the DSM assessment for standardized assessments of mental status (AMS Scales, such as Glasgow Coma Scale or Richmond Agitation Sedation Scale, p=0.73) or validated Delirium scales (such as CAM, CAM-ICU, ICSDC, 4AT, see abbreviations in Supplementary eMethods, p=0.73). In contrast, rates of delirium were lower when assessed using nonstructured symptom assessments (i.e., symptom collection, p=0.03) or ICD codes (p<0.001). Symp. Collection = symptom collection: identification of relevant symptoms, e.g. chart report of altered mental status. Each bar represents a meta-analytic estimate of delirium rates, with the calculated 95% confidence interval. Post-hoc p-values in all panels are adjusted for multiple comparisons (Holm). Forest plots of all studies with n’s, meta-analytic proportion, CI, and heterogeneity measures for each subgroup analysis are provided in supplementary material (Supplementary Figure S4). B. Delirium rates varied significantly according to study setting, which explained 12% of variance in delirium rates across studies (R² 12.1%, p=0.002). Compared to the average of all studies, delirium rates were significantly higher for studies performed in the ICU (p=0.0002). (ICU = Intensive Care Unit, ED = Emergency Department). Conventions as in A, with forest plots in Supplementary Figure S5). C. Delirium rates did not vary significantly according to pneumonia origin, which explained only 2.4% of variance in delirium rates across studies (R² = 2.4%, p=0.15). (HCAP=Healthcare Acquired Pneumonia, CAP=Community Acquired Pneumonia). Conventions as in A, with forest plots in Supplementary Figure S6). D. Delirium rates did not vary significantly according to microbiological etiologies, which explained 0.0% of variance in delirium rates across studies (R² = 0.0%, p=0.63). Conventions as in A, with forest plots in Supplementary Figure S7).

Figure 4.

Delirium is associated with significantly increased mortality in patients with pneumonia and has not changed over time. A. Pooled effect of studies demonstrating the association between delirium and mortality using univariate analysis (n=41 studies, OR 4.3, 95% CI [3.24; 5.76], p<0.001). This association remained significant in studies applying multivariable models to quantify the relationship of delirium and mortality while controlling for clinical covariates (Table 1). B. Meta-regression analysis measuring the prevalence of delirium in patients hospitalized with pneumonia over time shows steady rates including during the COVID-19 era (R² = 0.08, F_1,124 = 1.45, p=0.23). C. Notably, the association of delirium and mortality over time remains unchanged (R² = 0.05, F_1,39 = 1.01, p=0.32). Each circle represents one study, and the size of the circles represents the study weight (studies with a higher precision, i.e., with a smaller standard error, receive a greater weight).