Design
A mixed-methods prospective cohort design will be used to examine children’s psychological and behavioral responses over a 3-year period post-PICU hospitalization (study group n = 110), and post-ENT day surgery (comparison group n = 110). A subgroup of parents of children in the study group (n = 20) and the comparison group (n = 20) will also be interviewed at 1 and 3 years post-discharge using an interpretive descriptive approach [44, 45]. Parents will be invited to include their child for a portion of each interview to allow a fuller understanding of the impact of hospitalization. The complementarity of the quantitative and qualitative data will provide a comprehensive understanding of children’s psychological and behavioral responses to PICU hospitalization [46]. We will use standardized questionnaires with all participants (n = 220). With open-ended interviews we will explore a subgroup of participants’ (n = 40) perceptions of the impact of PICU and day-surgery hospitalization on the child at 1 year (close to the study mid-point), and 3 years (the final data collection point) post-discharge. This will allow us to probe participant responses to questionnaire items [45, 47, 48]. In addition, interviewing parents in both groups will allow us to capture a broad scope of hospital experiences. Multiple types of data will enhance the potential for developing new insights in the field and exploring alternative interpretations of the data, ultimately enhancing data validation [47, 49]. Data will be collected throughout the follow-up period (Figure 2). To ensure sufficient time for enrolment and follow-up, this study will take place over a 5-year period across 3 Canadian pediatric hospitals.
An ENT day-surgery comparison group was chosen for several reasons. They are an otherwise healthy group of children who have nominal contact with the medical system. They will undergo a hospitalization experience resulting in a minimal length of stay, low invasive procedure scores, and low severity of illness scores—variables to which our PICU group will have higher exposure. In addition, they will act as controls for expected maturational changes in children over the course of the 3-year follow-up.
Setting
The study will take place in 3 university-affiliated pediatric hospitals located across 3 Canadian provinces: The Montreal Children’s Hospital (MCH) in Montreal, Quebec, The Hospital for Sick Children (HSC) in Toronto, Ontario, and the IWK Health Centre (IWK) in Halifax, Nova Scotia. All hospitals have PICUs that admit children from infancy to 17 years with a wide range of medical and surgical diagnoses. Comparison groups will be drawn from hospital ENT day surgery units at the MCH and the IWK.
Study recruitment will take place in the PICU (study group). The first data collection point post-PICU will be on the hospital ward prior to discharge. Children in the comparison group will be recruited through the hospital day surgery unit. The first data collection point will be at the hospital following surgery. Data will subsequently be collected using standardized measures that parents complete at home. Parents asked to participate in follow-up interviews will be purposively selected, as we wish to recruit a heterogeneous sample that reflects the general PICU (and ENT) population. Interviews will take place following questionnaire completion at years 1 and 3 in the families’ homes, or at the study site from which they were recruited, based on parent preference.
Population
The target population is children aged 3 to 12 years who have suffered a critical illness requiring PICU admission for a minimum of 24 hours, and are expected to survive. Children must speak English or French to participate, one parent must speak, read and write English or French, and the child must be identified as ready for discharge. While the PICU population is heterogeneous, medical diagnoses have not been related to psychological sequelae. Rather, illness severity and exposure to invasive procedures have been identified as significant predictors of post-traumatic stress symptoms 6 months post-discharge [11]. To capture the full range of admissions, we will only exclude children who have had a previous PICU admission, and those who suffer severe brain injury and neurological impairment (because they cannot be evaluated using standardized measures). On the day we begin recruiting at each site (i.e., at study inception only), we will exclude children who have been in the PICU longer than 60 days (1.1% of MCH and 0.96% of HSC admissions in 2011), as this will preclude our ability to gather baseline BASC-2 scores. Afterwards, all children, regardless of admitting diagnosis and length of stay, and their parents (or primary caretaker) will be potential candidates.
The comparison group will include children aged 3 to 12 years who have undergone ENT day surgery. Children must speak English or French, and one parent must speak, read and write English or French to participate. Any imbalance that may occur between groups will be adjusted for in the statistical analyses. Children who have experienced a previous PICU admission, or who are neurologically impaired will be excluded from the comparison group (as per the study group). We will not exclude children in the comparison group who require PICU hospitalization during the 3-year follow-up period, because this would result in an overly healthy group at baseline. Rather, we will censor these children at the time of PICU admission in order to retain information about them for as long as possible in the study [50].
Procedures
All PICU admissions and those booked for ENT day surgery will be screened for eligibility. A Research Assistant (RA) will approach parents of eligible children to explain the study, answer questions, and obtain parent consent and child assent from those who agree to participate. A subgroup of 20 PICU and 20 ENT parents will be invited to participate in face-to-face interviews at year 1 and 3 follow-up points. All parents will be asked at enrolment if they would agree to be contacted regarding participation in these interviews, and 40 parents will be subsequently be selected (see Sample size estimation). Following enrolment, demographic and hospitalization data will be collected from the child’s medical chart and verified with parents. Hospitalization information will include medical diagnosis, number of previous hospitalizations, length of stay, pediatric risk of mortality (PRISM-3 [51]) score as an indicator of illness severity, and the child’s invasive procedure score (IPS [11]). Parents will complete the State-Trait Anxiety Inventory (STAI [52, 53]) as an indicator of parent anxiety, and the BASC-2 [54] as an indicator of their child’s usual (i.e., pre-admission) emotional and behavioral characteristics. PICU admissions are frequently unplanned, making it impossible to gather data prior to admission. The BASC-2 items ask about child characteristics over the previous 6 months; therefore, parents will complete it immediately following their child’s PICU or day surgery stay, and will be asked to think about their child’s usual behavior during the 6-month period prior to hospitalization. For children who remain in the PICU 60 days post-admission and who are not ready for discharge, parents will complete the BASC-2 at that time to minimize recall bias. In this way, we expect all baseline scores to reflect the child’s usual pre-hospital characteristics.
The second period of data collection will take place 6 weeks post-PICU or day surgery hospitalization (Figure 2). Parents will complete two questionnaires requiring a total of 20 to 30 minutes: the STAI, and the Post-Hospital Behavior Questionnaire (PHBQ [55, 56]), which evaluates changes in the child’s behavior since hospitalization. Children aged 6 years and older will complete the Children’s Critical Illness Impact Scale, written (CCIIS [24, 57]) or pictorial (Y-CCIIS [39]) version depending on reading ability, which takes 10 to 20 minutes to complete. Parents will receive follow-up telephone calls from the site-based RA to ensure questionnaires were received, answer any questions, and encourage parents to return them by mail. Those who do not return questionnaires within 3 weeks will be offered the possibility of a home courier pick-up.
Data collection will continue at 6 months, 1, 2, and 3 years post-discharge (Figure 2). Reminder telephone calls will ensure complete follow-up, and parents will be asked whether their child has required further hospital admissions. The IPS (based on chart review) will be calculated for each readmission, and a cumulative (follow-up) IPS score will be generated. Parents will be asked to spend approximately 45 minutes at each follow-up point completing 3 questionnaires: the BASC-2; the Parenting Stress Index (PSI [58]) to assess stress in the parent–child system as well as life stresses beyond parental control; and the SDQ [59, 60], which screens child behavior and its impact on the child and others. Children will complete the CCIIS again at the 6-month visit to measure distress post-discharge. Children aged 8 years and older will complete the BASC-2 Self-Report Profile (SRP), an indicator of their usual emotional and behavioral characteristics.
Questionnaires will be mailed back to participants’ respective recruitment sites, allowing site-based RAs to keep track of mailings and follow-up requirements. The BASC-2 (primary outcome variable) has clinical cut-off scores, and while scores cannot be interpreted in isolation (the child would require further assessment by a clinical psychologist or psychiatrist), parents will be informed if their child’s behavioral symptoms index composite score falls within a clinically significant range; specifically, if the child receives a t-score >70 on either the Parent Report Scale or the Child Self Report Scale (i.e. scores ≥95% percentile on the overall behavioral symptoms index of the parent version, and the emotional symptoms index of the child version). Parents will receive a telephone follow-up call, a letter explaining their child’s score, and information regarding community resources they can contact for support. We will track care received through professional mental health services, and explore potential effects in our analyses should changes in children’s patterns of recovery be observed.
Face-to-face, 1-hour interviews will take place at the end of years 1 and 3. At the initial interview, the RA will ask open-ended, semi-structured questions designed to elicit a narrative from parents about their experience of events and their perceptions of their child’s return to daily life. Interviewers will also probe parents’ responses to questionnaire items. At the second interview, semi-structured questions will be tailored to participants’ previous interview and questionnaire responses. At each interview, parents will be encouraged to invite their child to be present for the first 20 to 30 minutes to comment on their experiences of returning to daily life after hospitalization. The purpose of including the child is to obtain a fuller understanding of the impact of hospitalization. The interview focus will then shift to include the parents only so that they have an opportunity to share their own narratives without the child present. Parents’ wishes to include or exclude their child will be respected.
Measures
All measures have been tested for their psychometric properties and translated from English into French. The primary outcome measure of children’s psychological and behavioral responses following hospitalization is the BASC-2. The BASC-2 assesses positive (adaptive) and negative (clinical) dimensions of emotional (e.g., anxiety, depression, somatisation), behavioral (e.g., hyperactivity, withdrawal, aggression), and adaptive functioning (e.g., social skills, leadership, adaptability) in children aged 2 ½ to 18 years [54]. It is a multidimensional measure with excellent psychometric properties, and is well-suited to the heterogeneous population of children admitted to the PICU.
Secondary outcome measures include the SDQ, which characterizes children’s behavioral responses and contains an impact supplement that provides complementary information [59, 60]. Parents are asked whether they feel their child has an emotional, concentration or behavioral problem and, if so, to describe the chronicity, distress, social impairment and burden to others of that problem. The Harter Scales of Perceived Competence (referred to as Harter, hereafter) provide self-report information on perceived competence from the child’s perspective, reflecting an essential component of psychological well-being [61, 62].
Predictors of children’s psychological and behavioral responses following hospitalization include child, parent and hospitalization characteristics. We will measure child characteristics at baseline including age, number of previous hospitalizations, and previous emotional and behavioral characteristics (using the BASC-2 as an indicator of pre-admission behavior), as well as distress 6 weeks post-discharge (CCIIS [24, 57] and PHBQ [55, 56]). Parent anxiety will be measured in hospital and at 6 weeks post-discharge (STAI [52, 53]). Data regarding illness severity (PRISM-3 [51]), number of invasive procedures the child is exposed to (IPS [11]), and length of stay will be gathered during hospitalization. These variables may be used to identify high risk children.
Potential mediators of children’s long-term responses include parenting stress and other life stresses (PSI [58]), number of hospital readmissions, cumulative number of invasive procedures (IPS [11]) related to hospital readmissions, and number of professional counselling sessions during follow-up.
Data analysis
In this mixed-methods study, quantitative data will be analyzed using SAS version 9.3 (SAS Institute, Cary, NC). To describe the two groups of participants, demographic and clinical characteristics at baseline will be summarized using means or medians for continuous data and proportions for categorical data. Psychological and behavioral response scores for PICU and ENT groups will be calculated and compared at each follow-up time point, both statistically and graphically.
To address our primary objective, we will compare group response scores on the BASC-2 at 3 years post-hospitalization, as well as at 6 months, 1 year and 2 years post-hospitalization using independent two-sample T-tests. We will then compare psychological and behavioral responses within patients and between groups at 6 months, 1, 2, and 3 years post-discharge using a mixed effect linear regression model to take the random effect of time into account and to adjust for baseline characteristics [63].
To address our secondary objectives, we will: (a) compare children’s problem behaviors on the SDQ within patients and between groups at 6 months, 1, 2 and 3 years post-hospitalization using a mixed effect linear regression model to take the random effect of time into account and to adjust for child characteristics at baseline; (b) compare children’s perceived competence on the Harter within patients and between groups at 6 months, 1, 2 and 3 years post-hospitalization using a mixed effect linear regression model to take the random effect of time into account and to adjust for child characteristics at baseline; (c) explore baseline child and PICU-based factors and potential mediators predictive of poor psychological and behavioral response scores 6 months, 1, 2 and 3 years post-PICU hospitalization using correlation analysis and multivariate linear regression. Statistical analyses will be conducted following the verification of assumptions underlying those analyses. In our previous work, PICU children were categorized as high or low risk in a secondary data analysis based on a positive skew in the distribution of invasive procedure score data [12]. We do not wish to assume a similar distribution based on changes in the PICU population over the past decade, and study limitations identified in our previous secondary data analysis; therefore, we will examine the data and determine whether non-linear terms or transformations are needed in the regression models.
To address our final secondary objective, qualitative (interview) data will be analyzed concurrently with quantitative and qualitative data collection. This will foster data complementarity by allowing us to probe questionnaire responses [46, 64], generate data that cannot be captured by questionnaire measures alone [46], and ultimately produce deeper insights into children’s psychological and behavioral responses to PICU hospitalization [49]. Audiotapes of interviews will be transcribed verbatim, read in their entirety, and coded line-by-line. Open-coding will be used with the initial semi-structured interviews to identity themes in participants’ narratives [65]. Prominent themes will be flagged for incorporation into the second semi-structured interview, along with items of concern identified on the completed questionnaires. Semi-structured scripts for the second interview will be tailored to themes identified in the first interview and the questionnaire items we wish to probe further. We will use open and axial coding with second interviews to connect the emergent categories [65]. For all interviews, data will be analyzed using the constant comparison method [66]. Comments by children present during the interviews will be highlighted. NVivo software will be used to support data management and sorting [67].
Sample size estimation
The sample size estimate is based on psychosocial and behavioral response scores on the BASC-2 at 3 years post-PICU discharge. A sample size of 126 participants (63/group) will achieve 80% power at a 0.05 significance level, allowing us to detect a medium effect size of 0.5 between groups using an independent two-sample T-test. Baseline child characteristics may differ in PICU and ENT groups; therefore, we will adjust for potential confounding using multivariate linear regression and, in anticipation of this, have added an additional 20% to our sample size. Finally, we estimated a 30% attrition rate at 3 years (attrition in our 2002 cohort study was 13% at 6 months [11]). Therefore, we aim to enroll no fewer than 110 children per group (n = 220).
We will conduct interviews with a subset of 40 parents (20/group) of children enrolled in the larger cohort. Purposive sampling will be used to select parents whose children reflect the heterogeneous group normally admitted to the PICU, as well as children admitted for typical ENT day surgeries. We anticipate this sample size will allow for a fuller understanding of children’s experiences [46, 49], and allow us to reach data saturation [47].
Ethical considerations
The study and all consent and assent forms were reviewed by the Research Ethics Boards of participating institutions. Participants are asked to provide written consent (parent), and verbal or written assent (child) according to provincial law. They can withdraw from the study at any time during the 3-year follow-up period without affecting patient care. Parents of children whose overall score on the BASC-2 falls within a clinically significant range (≥95% percentile on the overall behavioral or emotional symptoms index) at any of the follow-up time points will be informed and directed to local community health resources.
Knowledge dissemination
A report of the study findings will be shared with all participants, and they will be invited to a presentation of the findings. We will submit a copy of the report and invitations to the presentation of study findings to the parent/child councils and family support groups in participating hospitals and community partners, including rehabilitation and transition facilities. To promote broader community engagement, findings will be shared with clinicians and researchers at all sites, and telehealth presentations will facilitate dissemination to a larger professional audience. Study findings will be published in peer-reviewed journals and disseminated via academic meetings.