Neurocognitive functioning and health-related quality of life of children after pediatric intensive care admission: a systematic review

The systematic literature search was conducted on February 26, 2021 to identify relevant articles published 2015–2020 in the following electronic databases: Embase, Medline Ovid, Web of Science, Cochrane CENTRAL, and Google Scholar. Keywords used were pediatric intensive care, children, adolescents, neurocognitive functioning, and health-related quality of life (see Online Resource 1a for complete search strategy). The articles that resulted from the search were independently screened on title and abstract by two reviewers (JH + KD). Abstracts that did not overlap were discussed (JH + KD). Full-text screening was conducted by one reviewer (JH) and discussed with the other reviewer (KD). When full-text inclusion was debatable, another reviewer made the final decision about inclusion (KJ).

Studies were included when the following eligibility criteria were met: term born children to 18 years old who were admitted to a PICU; (neuro)cognitive functioning and/or HRQoL assessed with validated, age appropriate measurements with country-specific normative data or control groups in children who survived a PICU admission.

Exclusion criteria were the following: case reports and editorials; studies published in other languages than English; studies including neonates born preterm (gestational age < 36 weeks) admitted to the PICU or NICU when not reporting outcomes of term born children separately. When it was not clear whether a study included children born preterm (< 36 weeks), the authors were approached by e-mail to check. In case there was no response or when it remained unclear, the study was excluded. With regard to the neurocognitive outcome measurements, studies assessing neurocognitive functioning both by clinical and/or by proxy-reports of the parents or caregivers were included.

For all eligible studies, the following variables were extracted: population studied, age at PICU admission (or age at follow-up in case age at admission was not mentioned), sample size, study design, time between PICU admission and follow-up, measurements used (clinical tests, parent-reported or self-reported questionnaires), and main results. As to executing a meta-analysis of the results, this is discouraged based on the following conditions: both RCTs and observational studies are included in this systematic review, different approaches for analyses were used in studies involved, and different outcomes with different instruments were assessed [10, 11]. Under such conditions, a meta-analysis will not lead to a meaningful interpretation and was therefore not performed. A qualitative synthesis of the main results was done as follows: when the group mean of the patients was more than 1SD below normative data or mean score of healthy control children, scores were judged as impaired, when the group mean of patients was lower than normative data/healthy control data but within the normal range (within 1SD below the norm to average), they were judged as lower than average, and when scores were comparable or higher than the norm/healthy control data, they were judged as not impaired (see supplemental Table 1a and Online Resource 1b). To present the results in a structured way, studies in infants (group mean or median age less than 12 months) and children (group mean/median age 12 months or older) were separately reported as well as studies focusing on the short term (12 months or less) and studies focusing on the long term (more than 12 months) after PICU admission. The qualitative synthesis of HRQoL is divided into overall HRQoL, physical HRQoL, and psychosocial HRQoL, following the scale structure of most HRQoL instruments. Both domains emotional functioning and social functioning are represented in the psychosocial HRQoL scale. When emotional and/or social subscales were reported separately, we averaged the outcomes. Normative data or control group outcomes reported in the paper were used as references. When these were not mentioned, internationally accepted normative data were used as reference.

The methodological quality of the studies was assessed with appropriate scales [12]: the Cochrane Collaboration’s Tool (CCT) for RCTs, the Newcastle–Ottawa Scale (NOS) for cohort studies, and the Agency for Healthcare Research and Quality (AHRQ) methodology checklist for cross-sectional studies [12]. The Cochrane Collaboration’s tool consists of seven components that can be scored with ‘low risk,’ ‘high risk,’ or ‘unclear risk.’ For the current study, for each item judged as ‘high risk,’ one point was assigned, and for ‘low risk’ and ‘unclear risk’ zero points were assigned, which results in a maximum total score of seven points. The NOS consists of eight components and each component has a score of zero or one, with one component a score of zero, one, or two, and a maximum of nine points to achieve. The ARHQ counts 11 items scoring zero or one, which results in a maximum total score of 11 points.

A total of 3649 studies were identified. After reviewing title and abstract, 3350 studies were excluded. After full-text screening of the remaining 299 articles for eligibility, a total of 75 articles (38 on neurocognitive functioning, 33 on HRQoL, and 4 on both outcomes) met the inclusion criteria and were included in the qualitative data reviewing. Thirty-eight studies investigated neurocognitive functioning (Fig. 1), with sample sizes ranging between 17 and 786, and 32 with less than 100 study patients (Supplemental Table 1). Thirty-three studies examined HRQoL (Fig. 1), with sample sizes ranging between 6 and 1109, and 20 studies with less than 100 study patients (Supplemental Table 2). Four studies with sample sizes ranging between 25 and 43 investigated both neurocognitive functioning and HRQoL [13, 14]. Of the 75 articles included, 50 were prospective cohort studies, 6 retrospective cohort studies, 2 cross-sectional studies, and 17 RCTs (Supplemental Tables 1 and 2). Of these RCTs, ten studies reported effects of interventions during PICU admission and seven studies reported effects of interventions during follow-up. Two RCTs investigated psychosocial interventions, 15 RCTs examined the effect of disease management strategies.

The average risk of bias score of the 14 RCTs was 3.2 out of seven, with a range of zero to four. Most studies had a low risk of bias regarding the random sequence of the allocation to intervention groups. Almost all studies had problems with blinding participants to intervention groups. The average risk of bias score of the 50 cohort studies was 5.6 out of nine, with a range of three to eight. Selection bias was common due to inclusion criteria such as age restrictions. Furthermore, studies frequently had no adequate follow-up or had no description of the patients lost to follow-up. The risk of bias score of the two cross-sectional studies included in this systematic review were 4 and 6 out of 11 (Supplemental tables 1a, 1b, 2a, and 2b for individual scores of the studies).

A number of factors were reported to predict worse neurocognitive and HRQoL outcomes after PICU admission. These predictors include pre-existing factors (such as lower weight, lower SES), but also factors related to the management at the PICU (such as use of benzodiazepines and corticosteroids), of which the latter are modifiable. In infants at time of PICU admission, longer length of PICU stay was associated with worse general intelligence scores in the short term (≤ 12 months after PICU admission) [16, 17]. In the long term (> 12 months after PICU admission), worse scores in infants were predicted by lower weight [32], lower SES [32], and older age [40] predicted worse general intelligence scores. In children aged 12 months or older at the time of PICU admission factors that were associated with worse neurocognitive scores were higher baseline levels of inattention [25], higher serum levels of the biomarker neuron-specific enolase [25], longer length of PICU stay [30], and longer follow-up time [30] on the short term. In one study younger age at PICU admission was reported to predict worse scores for attention [25], but in another study higher age predicted worse scores for executive functioning [30]. This difference could be related to the difference in domains or to the difference in reason for admission, respectively, traumatic brain injury [25] and liver transplantation [30]. In the long term, longer length of PICU stay [51], more circulating phthalates [52] and use of benzodiazepines and corticosteroids during PICU admission [53], and higher premorbid developmental risk [51] were associated with worse neurocognitive scores. For HRQoL, factors associated with lower scores on the long term for infants were the need of mechanical ventilation [82] and neurological complications (intracranial ischemia or hemorrhage) [81]. For children aged 12 months or older, older age, longer length of stay and worse disease severity [60] were associated with lower HRQoL in the short term, and in the long term an elective PICU admission and neurological diagnosis [87]. In conclusion, these predictors are pre-existing factors or factors related to the management at the PICU, which may be modifiable.

Some studies have tried to influence these factors through psychosocial and disease management interventions, both during PICU admission and follow-up, to improve short-term and long-term outcomes related to neurocognitive functioning and HRQoL. Starting late with supplemental parenteral nutrition during the first week of critical illness compared to starting early had a positive effect on some neurocognitive domains [53, 54]. Hypothermia versus normothermia [75] and different sedation procedures did not reveal differences in HRQoL outcomes [56, 62, 63]. Two RCTs tried to improve HRQoL by rehabilitation [64, 66] and found that, after PICU discharge, an exercise program near home was slightly more effective than a program at the hospital [66] and that early protocolized rehabilitation (< 72 h of PICU admission) was not better than usual care [64]. A working memory training (Cogmed) improved working memory only immediately after the training, and the training had no effect on HRQoL outcomes [14]. Although some intervention studies tried to improve outcomes by affecting predictors, there is a lack of research on mitigating adverse PICS-p outcomes in critically ill children using RCTs. One of the under investigated fields is the importance of parents in the recovery of critically ill children after PICU discharge, investigating for example the effect of monitoring and empowering parents on the PICU through shared decision making and the influence on PICS-p outcomes after PICU discharge [90].

The survival rates at the PICU have increased in the last decades [91],which presents the challenge of limiting morbidity related to PICS-p. Some critically ill children obtained worse neurocognitive functioning and HRQoL outcomes compared with children from the general population. It is important to gain insight in this group of critically ill children and their parents to prevent cognitive problems or psychiatric problems such as anxiety, depression, or posttraumatic stress they may develop after PICU admission of the child [92]. Psychoeducation and parent support (teaching skills and emotional support) during PICU admission and directly after discharge are important [93]. During follow-up, all children and families should be monitored neurocognitive and psychosocially to find those at risk for impaired development and developing morbidities. Since some children and parents have delayed reactions with regard to psychosocial symptoms [94], this monitoring should be done both short term and long term [6]. Furthermore, psychosocial interventions should be offered to families at high risk for psychosocial problems. Although some studies reported predictors for parents at risk, for example parents with higher baseline stress levels, there is also a lack of multi-factorial predictive models for parental and children’s outcomes [93]. Such predictive models might guide (neuro)psychologists, involved in providing follow-up care, in selecting accurate neurocognitive tests, including assessment of attention, memory, and/or executive functioning, and validated HRQoL questionnaires and other psychosocial questionnaires to monitor psychiatric problems. Moreover, these predictive models might guide which psychosocial innervations might be deployed. However, there is also a lack of research into the effectiveness of psychosocial interventions critically ill children and their families with high risk for psychosocial problems.

With regard to the neurocognitive test battery used in the studies, only one research group assessed different domains of neurocognitive functioning (as part of a structured longitudinal clinical follow-up program) at different time points until the age of 18 years which was in a group of patients treated with ECMO [13, 14, 36, 38, 42, 43]. Overall, they found impairments in general intellectual functioning, attention, memory, and executive functioning. Interestingly, the other studies included in this review that investigated neurocognitive functioning assessed general intellectual functioning only, or exclusively investigated a certain domain of neurocognitive functioning (Table 1). General intellectual functioning is usually called intelligence and includes abilities (such as logical reasoning, problem-solving, learning, and verbal skills) that allows a person to understand the reality and to interact with it [95]. Although general intellectual functioning is predictive of school functioning, it is important to understand which, more specific, neurocognitive domain has been impaired. In neuropsychology, the model of hierarchy of neurocognitive domains has often been used to map developmental problems of the child. In this neurocognitive hierarchy, lower functions, such as sensation and attention, affect the outcomes of the higher neurocognitive domain memory. Impairments in memory functioning affects, in turn, the development of executive functioning [96]. When impairments exist in the lower domains of the hierarchy, patients will experience more consequences in daily life as it will affect the higher neurocognitive domains [96]. Therefore, it is important to assess all layers of the hierarchy to figure out where the problem exists.

HRQoL is usually assessed by the parent or caregiver, as young children are not able to report on their own HRQoL. Studies that also included children who were old enough to self-report, found that parents’ proxy-reports were worse than self-reports [78, 80]. Furthermore, the mental HRQoL of the parent itself was a predictor of HRQoL scores they reported for their child, which is a result of shared variance of the informant [60]. The value of proxy-reports is often debated. A systematic review examining the similarities and differences between reports of children and their parents concluded that it is not the question who of the informants is right, but that the discrepancies in outcomes between informants are rather a reflection of differences in their perspectives which gives valued information about the parent and the child [97]. Therefore, a multi-informant approach is recommended (both child and parent PROMs) in daily clinical practice.

A strength of the current study is that the literature search was conducted using rigorous methodology with the support of a librarian. Furthermore, it describes the legacy of a PICU admission on the crucial domains of daily life, namely cognitive, physical, emotional, and social health, described as PICS-p [98]. This systematic review investigated all these domains, including HRQoL which is the subjective evaluation of physical, emotional, and social health. Although one reviewer discussed all the full texts with another reviewer, a limitation is that not both reviewers read all the full texts. Furthermore, a second weakness is the large heterogeneity across studies with regard to study design, domains assessed, instruments used, analysis methods, and reporting results. Furthermore, most studies did not report standard scores, which makes it hard to compare results between studies. The majority of studies focused on specific diseases or diagnoses which makes it impossible to draw firm conclusions. Previous reviews excluded studies focusing on electively admitted children or on specific diagnosis [5, 89, 99], or excluded neonates admitted to the PICU [100]. Studies included in this systematic review are representative for the complete group of PICU survivors as no patient populations were excluded and most common reasons for a PICU admission were covered, particularly respiratory disease, cardiac disease, and neurologic disorders [101]. Lastly, the quality of studies varied with a number of studies having a high risk of bias, urging for high-quality future research.