Introduction
The American Psychiatric Association's
Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV), characterizes
delirium, also known as
acute brain dysfunction, as including four key features [
1]. The definition includes (1) acute onset (usually hours to days) with a fluctuating course during the day, (2) a disturbance of consciousness, (3) change in cognition and (4) evidence of a causative medical condition, substance intoxication or medication side effect determined from the history, physical examination or laboratory tests. Delirium is common in adult intensive care, and there are validated tools for measurement as well as known risk factors and associated adverse neurocognitive outcomes [
2],[
3]. A validated diagnostic method for delirium in the pediatric population has yet to become part of daily clinical practice. This is problematic, as the correct diagnosis of delirium in children enables early intervention to avoid the adverse effects of undiagnosed and untreated delirium. Moreover, the prevalence, risk factors, treatments and outcomes of pediatric delirium have not been determined. To study them, accurate, practicable screening tests are required.
To determine what is known about the diagnosis of delirium in the pediatric intensive care unit (PICU), we conducted a systematic review of the literature on the accuracy of diagnostic tests for delirium in children in the PICU. A secondary objective was to determine the risk factors for, outcomes of, and treatments for delirium in the PICU in the studies we identified for inclusion. We hypothesized that we would find validated and accurate diagnostic tests for delirium in the PICU. We identified only five accuracy studies, each of which had a high risk of bias and examined in isolation a different screening test for delirium. We make suggestions for future research needed in this field.
Discussion
Delirium in adult intensive care is common (incidence estimates range from 45% to 87%), and there are well-validated screening tools for diagnosis. It has known risk factors and is associated with increased intensive care and hospital lengths of stay, increased mortality, and long-term cognitive impairment [
2],[
3]. In our systematic review, we found only five studies of the diagnostic accuracy of tests for delirium in the PICU [
9]-[
12],[
17]. In these five studies, delirium was common, with a prevalence of 13% to 28% (excluding the study of "clinical suspicion," in which partial and differential verification bias limited sensitivity for delirium), including hyperactive, hypoactive and mixed subtypes. In general, the studies were small and single-centered and had methodological weaknesses and high risk of bias. In each study, researchers examined a different index test, and, other than the "clinical suspicion" index test study, the researchers did not examine risk factors, treatments or outcomes of delirium. The sensitivity and specificity of the PAED (retrospectively in patients ≥1 year of age), the CAP-D (prospectively in all age groups) and the p-CAM-ICU (prospectively in patients developmentally ages ≥5 years) were high; however, these data should be interpreted cautiously, given the study limitations, including high risk of bias and small numbers, and, in addition, one group could not confirm high sensitivity for the PAED [
11]. There were no direct comparisons of the index tests, except in one study in which both the CAP-D and PAED were compared. In that study, the CAP-D had much better accuracy. Interrater reliability was examined only for the p-CAM-ICU and CAP-D(R), and the reliability was excellent.
We are aware of two other systematic reviews of delirium in children, neither of which critically examined studies of the diagnostic accuracy of tests for delirium in the PICU. Neither of these two reviews referenced any studies that we did not include [
18],[
19]. In one, the authors searched for publications about pediatric delirium (not limited to the PICU) published between 1980 and 2009 and found only small case series and case reports. They concluded, "Delirium is an important but neglected disorder of childhood associated with significant morbidity and high mortality. Current clinical practice for management is based on slim empirical evidence" ([
18], p 337). In the other systematic review, the authors focused on delirium in the PICU and searched the literature published between March 2009 and March 2011 (that is, following the first systematic review dates). In that review, the authors found only two observational studies, one of which examined p-CAM-ICU and was included in our present review [
10] and one of which was the follow-up study to the "clinical suspicion" study included in our present review [
14]. The authors of the latter review concluded that "there are still important, yet unresolved issues, regarding pathophysiology and biomarkers, risk factors, early detection, and appropriate treatment" ([
19], p 1106). We are also aware of two narrative reviews of delirium in the PICU [
20],[
21]. Neither of those reviews references any study not included in our present review, nor is either of them focused on the diagnostic accuracy of delirium screening tools in the PICU.
The importance of delirium in adult intensive care is clear. Delirium is both common and commonly overlooked, and it has adverse consequences, including prolonged ventilation, increased hospital and ICU lengths of stay, increased health care costs, long-term cognitive impairment and physical disability. The duration of delirium is independently associated with mortality [
2],[
3],[
22]. The main screening tools for adults include the CAM-ICU (pooled sensitivity 76% to 80%, pooled specificity 96%) and the Intensive Care Delirium Screening Checklist (ICDSC; pooled sensitivity 74% to 80%, pooled specificity 75% to 82%), which are used to assess delirium over the course of 1 minute (CAM-ICU) and 8 to 24 hours (ICDSC), respectively [
3]. The CAM-ICU requires interaction with the patient, but the ICDSC does not (observation only) [
3].
Similarly, in our present review, we found that delirium in the PICU is common and that it likely has adverse consequences, including increased hospital and PICU lengths of stay and possibly increased mortality. The main screening tools either require (p-CAM-ICU) or do not require (PAED, CAP-D, CAP-D(R)) patient interaction. Both the p-CAM-ICU and the CAP-D(R) are quick screening tools (taking less than 2 minutes to complete) [
17],[
21], have high sensitivity (78% and 94%, respectively) and high specificity (91% and 79%, respectively) and have excellent interrater reliability. However, methodological risk of bias and the limited number of single-center studies preclude making conclusions about their true performance characteristics. The weaknesses of the PAED include its low sensitivity for hypoactive delirium, variable sensitivity in the two studies (91% and 50%) [
9],[
11] and the major methodological biases of the favorable study (retrospective methodology and imputed data for >16% of patients).
There are limitations to this systematic review. The quality of information in systematic reviews is only as good as the included studies. The quality of our five included studies was only modest, with small patient numbers and high risk of bias. The studies are the first in the field and likely reflect the difficulty of performing costly, time-consuming clinical research in critical care. Nevertheless, the results of our systematic review suggest that investment in high-quality comparative and confirmatory studies are needed. The small studies we identified did not allow for examination of accuracy in different subgroups of patients based on age, diagnosis and so forth. The literature search methods for diagnostic accuracy studies are not well worked out. We followed recommendations in the literature, but it is possible that we missed important studies. We did not screen for descriptive studies of delirium in the PICU not focused on accuracy. Thus, It is possible that there are data about prevalence, risk factors, treatment and outcomes that we did not find in our literature search. The strengths of our study include its systematic search strategy, the use of published recommendations for assessing accuracy studies and their quality in systematic reviews, and the use of prespecified study screening and data collection tools employed independently by three and two authors, respectively (to minimize the risk of subjective bias in study selection and quality assessment). Our conclusions are therefore based on a transparent methodology using evidence-based guidelines for systematic assessment of the available literature.
Key messages
Delirium is common in pediatric intensive care, with a prevalence of 13% to 28%.
High-quality research to determine the accuracy of delirium screening tools in the PICU are required before prevalence, risk factors, treatments and outcomes can be determined.
There are two promising screening tools for pediatric delirium in intensive care that warrant further study: p-CAM-ICU and CAP-D(R).
Competing interests
The authors declare that they have no competing interests.