Outcomes of subsyndromal delirium in ICU: a systematic review and meta-analysis

We conducted a systematic review and meta-analysis of prospective observational studies and clinical trials following the recommendations of the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group [11] and according to the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [12]. We searched the following data sources: Medline (1966 to 1 October 2016), Embase (1974 to 1 October 2016), CINAHL (1982 to 1 October 2016), the Cochrane Library (1 October 2016), and PsychINFO (1887 to 1 October 2016). The most recent search was performed on 1 October 2016. Reference lists of retrieved articles and of relevant review articles, as well as personal files, were hand searched. There was no language restriction. Search terms included: subsyndromal delirium, subclinical delirium, delirium, agitation which were cross-referenced with the terms intensive care, intensive care unit, ICU, critical care, critical illness, critically ill, sepsis, acute respiratory distress syndrome, multiple organ system failure, and mechanical ventilation. We considered the following criteria for study inclusion: 1) full-length reports published in peer-reviewed journals; 2) prospective observational cohorts or clinical trials of adult (>16 years old) patients admitted to the ICU; 3) use of validated screening or diagnostic instrument for delirium: CAM [13]; Confusion Assessment Method for the Intensive Care Unit (CAM-ICU) [3], ICDSC [2], DSM-IV TR [14], and DSM-V [5]; and 4) the relationship between SSD and at least one of the following outcomes reported: hospital and ICU LOS, MV duration, death in the ICU, conversion from subsyndromal delirium to delirium, or any post-hospital discharge outcomes. We excluded case studies or series and studies in which the majority of enrolled patients (or the largest subgroup) presented with the following conditions: primary central nervous system disorder (e.g., stroke, traumatic brain injury, central nervous system infections, brain tumors, recent intracranial surgery); underwent organ/tissue transplantation (patient subsets associated with pathophysiologically distinct forms of acute brain dysfunction); or experienced alcohol or substance withdrawal. Two investigators (JIFS and RBS) performed the study selection process including the initial search for the identification of references, the selection of potentially relevant titles for review of abstracts and, among these, those chosen for review of the full-length reports. All selections were decided by consensus.

Data extraction from the selected articles was independently performed by two authors (RBS and JIFS). The following data were recorded (when available): study characteristics (study location, period of enrollment, type of ICU, patient enrollment criteria, number of patients enrolled, methods used to identify delirium, duration of follow-up); patient characteristics (i.e., age, sex, premorbid cognitive and functional status, severity of illness scores, organ dysfunction scores and MV); and SSD prevalence and outcomes (i.e., conversion to delirium, death in the ICU and hospital, ICU and hospital LOS, and duration of MV).

To assess the methodological quality of the studies, we used the Newcastle–Ottawa Quality Assessment Scale (NOS) [15]. The scale evaluates three aspects of study methods: the selection of study groups (range 0–4), the comparability of groups (range 0–1), and the quality of outcome ascertainment (range 0–3). The total score ranges from 0 to 8, and an acceptable methodological design is reflected by a score >5.

We evaluated patient characteristics, prevalence of SSD and outcomes (conversion to delirium, mortality in the ICU and hospital, ICU and hospital LOS, and duration of MV) for patients with and without SSD. The main outcome of interest was mortality (ICU and hospital). The strength of the relationship between SSD and mortality was expressed as risk ratios (RRs) with 95% confidence intervals (CIs). We selected the risk ratio as a measure of effect for the binary outcome (death) since it is less prone to artificial inflation due to heterogeneity than risk difference. For continuous outcomes, we calculated the weighted standard mean difference (SMD) based on reported means or medians. We used the I ² test to describe the proportion of the total variation in the study estimates that is due to heterogeneity in the meta-analysis. We performed all analyses using Review Manager version 5.3 [16].

The initial search identified 6777 citations, and five studies were additionally identified as a result of reviewing the reference lists of others articles. Articles in duplicate were removed (n = 3878); most of duplicates occurred because PubMed citations were also included in the Embase library. After careful evaluation of the abstracts, 32 articles were retrieved and reviewed in detail. Disagreements (n = 2) between the two evaluators were solved by discussion and reaching a consensus. Finally, six studies met the inclusion criteria and were selected. Figure 1 depicts the flow diagram of the study search and selection process according to the PRISMA methodology.

Characteristics of the six studies selected are described in Table 1. Four studies [7, 8, 10, 17] exclusively evaluated critically ill surgical patients and the majority of patients (95%) had undergone elective surgery, one study evaluated a mixed population of critically ill patients [6], and one exclusively evaluated mechanically ventilated patients in the ICU [15]. The ICDSC was used for the diagnosis of delirium and SSD in four studies [6, 7, 9, 10]. The CAM was the instrument used in the two other studies [8, 17].

All studies were prospective, with two clinical trials [9, 10] and four observational studies [6‐8, 17]. Two studies were randomized clinical trials primarily designed to evaluate the use of antipsychotic drugs to prevent the conversion of SSD to delirium [9, 10]. The selected studies were well designed and the NOS quality assessment demonstrated a low bias risk in most of them (Table 1) [15]. The funnel plot of the included studies in the meta-analysis was performed and suggested a low publication bias (Additional file 1).

The main characteristics of those with no delirium, SSD, and delirium were well described in three observational studies [6‐8]. In general, they had similar baseline characteristics in each group with some exceptions. In the study of Breu et al. [7], which evaluated patients who underwent cardiac surgeries, those with no delirium were younger (65.9 ± 10.3 vs 67.5 ± 9.9 vs 69.7 ± 11.7 years, p < 0.01) and had less duration of extracorporeal circulation (91.4 ± 34.0 vs 109.6 ± 49.6 vs 113.2 ± 44.7 min, p < 0.01), when compared with SSD and delirium patients, respectively. In the study of Li et al. [8], patients with no delirium had received fewer units in blood transfusion (1.3 ± 2.9 vs 3.3 ± 5.4 vs 15.7 ± 13.6 units, p < 0.01) and presented intraoperative hypotension for a small period of time (13.3 ± 13.5 vs 24.6 ± 31.9 vs 81.4 ± 76.5 min, p = 0.01) when compared with SSD and delirium patients, respectively. In the study by Ouimet et al. [6], patients with no delirium were younger (age 60 ± 15 vs 65 ± 14 vs 64 ± 15 years, p < 0.01), had a higher proportion of surgical admission diagnoses (62.1% vs 47.5% vs 48.7%, p = 0.01), and had the lowest Acute Physiology and Chronic Health Evaluation (APACHE) II scores at admission (12.9 ± 6.9 vs 16.7 ± 7.8 vs 18.6 ± 8.0, p < 0.01) when compared with SSD and delirium patients, respectively.

SSD prevalence and hospital LOS were the most frequently reported outcomes (Table 1). Pooling all studies, a total of 2630 patients were evaluated; among them, SSD was identified in 950 (36%) patients.

Two studies evaluated the association of SSD with mortality [6, 7]. Ouimet et al. [6] reported an increased ICU mortality in the SSD group (10.6% vs 2.4%, p = 0.002) in comparison to patients with no delirium, but in a post-ICU follow-up and after adjusting for age, APACHE II score, and medication-induced coma, the mortality rate was similar in the SSD patients (hazard ratio 0.97 (0.61–1.55), p = 0.90) when compared to patients with no delirium. The study of Breu et al. [7] found that hospital mortality rates were comparable between SSD and patients with no delirium patients (4 (0.8-7.1) vs 5 (3.4-21.1), p = 0.05).

Hospital LOS was described and compared between SSD, delirium, and non-delirium patients in only three studies [6‐8]. Delirium was associated with longer hospital LOS when compared with non-delirium patients in all studies (Table 1). SSD was associated with longer hospital LOS when compared with non-delirium patients after meta-analysis performance (SMD 0.31 (95% CI 0.12–0.51), p = 0.002; I ² = 34%) (Fig. 2).

Only two studies described the ICU LOS [6, 7] and one of them described a gradient with increasing ICU LOS in the comparison of those with non-delirium, SSD, and clinical delirium patients (respectively 2.5 ± 2.1, 5.2 ± 4.9, and 10.8 ± 11.3 days, p < 0.01) [6]. One study evaluated the MV duration and described a non-clinically relevant increase in weaning time (10.0 ± 8.0 vs 11.0 ± 10.75 h, p < 0.01) in SSD patients when compared to those without delirium [7].

Two studies investigated the use of antipsychotic drugs to prevent the conversion from SSD to delirium [9, 10]. Al-Qadheeb et al. [9] described that the use of intravenous haloperidol 1 mg vs placebo every 6 h in SSD patients did not prevent conversion to delirium (12 (35.3%) vs 8 (23.5%), p = 0.29) or the time to first delirium occurrence (2 (2–3) vs 3 (2–4) days; p = 0.22), did not reduce delirium duration (2 (1–2) vs 3 (2–4) days, p = 0.261), ICU LOS (6.5 (4–8) vs 7 (4–9) days, p = 0.66), days on MV (4.5 (3–7) vs 5 (3–8), p = 0.79), or ICU mortality (9 (26.5%) vs 7 (20.6%), p = 0.55). In this study the sole observed difference was a reduced duration of agitation (0 (0–2) vs 2 (1–6) h, p = 0.008) in those receiving antipsychotics. Hakim et al. [10] described that the administration of risperidone (0.5 mg every 12 h) to elderly patients who experienced SSD after on-pump cardiac surgery was associated with a significantly lower prevalence of delirium by DSM criteria (7 (13.7%) vs 17 (34%), p = 0.031) and described fewer patients with ICDSC scores >3 (8 (15.7%) vs 19 (38%), p = 0.011) when compared with placebo. However, ICU and hospital LOS were comparable in both groups (2 (2–3) vs 3 (2–3) days, p = 0.517, and 6 (5–7) vs 6 (5–8) days, p = 0.056, respectively) as well as duration of clinical delirium (3 (2–4) vs 3 (3–4) days, p = 0.664).