Development and validation of the Pediatric Stroke Quality of Life Measure

Development of the PSQLM

The PSQLM items were developed based on qualitative content analysis derived from individual interviews with the parents of 44 child stroke survivors, and six adolescent survivors who were at least 16 years old at the time of the interview. Children with both CSVT and AIS were included; they were at least 1 year after stroke and ranged in age from 2 to 18 years. Forty-five per cent (n=20) of parents interviewed had children aged between 2 years and 5 years; 36% (n=16) had children aged between 6 years and 10 years; and 18% (n=8) had children aged between 11 years and 18 years. Interviews explored parent and child QoL perspectives and addressed theoretically derived research interests based on the CHP conceptual model of QoL. Interviewees explored parents’ personal definitions of QoL, factors that enhanced and diminished QoL for their child, and the impact of the child's stroke on QoL. Interviews were audio-recorded and transcribed in a format compatible for qualitative analysis using QSR NVivo 8 data software (QSR International Pty Ltd, Doncaster, Vic., Australia). Based on the analysis, a series of charts was created compiling emergent themes and frequency, and items and domains of concern. Items identified added to or reduced one's QoL and crossed a wide range of domains. Many of the items reflected common themes not exclusive to this population, such as family support, good health, and accessibility within the community environment. However, there were concerns seemingly unique to this population that parents described as reducing QoL, including ongoing stress associated with the stroke event and recovery, fear of recurrent strokes, and changes in their child's development and capacity to learn, pay attention, make decisions, and not being bullied by peers. Enhancing QoL was the experience of the stroke, which brought families closer together, becoming more appreciative of what it means to be healthy, having financial support, and appropriate therapies and school programmes, and opportunities to participate in sports. Additional issues identified in the paediatric stroke literature and by the institutional children's stroke medical team were included in the final analyses. Items were assigned to domains of QoL encompassed within the CHP conceptual framework for QoL,28-30 namely (1) ‘being’ (who the child is perceived to be); (2) ‘belonging’ (the child's connection to people and places); and (3) ‘becoming’ (the child's nurtured growth and development). Together these three domains encompass the conceptual framework for QoL. This framework guided our initial development of interview questions, and subsequent categorization of items into one of the three QoL domains (see Fig. 1).

After organization of the QoL items within the novel questionnaire, we pilot tested the drafted instrument with 10 parents of children with AIS/CSVT attending our institutional children's stroke clinic. Parents provided feedback on content relevance, missing issues, and question wording. The comments were used to revise the novel questionnaire, which then underwent further pilot testing with three additional parents of children with AIS/CSTV.

The final instrument, named the PSQLM, consists of 75 items, each written in statement form (see Appendices S1 and S2 [online supporting information] for items and questionnaire format). To describe the QoL experienced by the child we consider the domains and subdomains along two dimensions: importance of the area of life as perceived by the child and enjoyment experienced. A basic QoL score is determined by the interaction between importance and enjoyment scores.28-30

The instructions on the PSQLM direct parents first to check off “Not applicable” if the item does not apply to their child. All items that are checked off as not applicable (not developmentally, culturally, or age-appropriate) are not included in scoring. This process eliminated the need to produce distinct questionnaires according to age range as seen in other generic assessments.

Parents were instructed to rate each applicable item on a 5-point Likert scale separately for importance (from a score of ‘5’ [extremely important] to a score of ‘1’ [not at all important]) and enjoyment/satisfaction (from a score of ‘5’ [extremely satisfied] to a score of ‘1’ [not at all satisfied]).

Items ranked as high importance and high satisfaction (each ranked at 5/5), yield an item score of 10. Items ranked as high importance (score of 5) but low enjoyment/satisfaction yield an item score of –10. Items ranked of moderate importance (score of 3) with moderate satisfaction (score of 3) yield a QoL score of 0. An item rated as high importance (5/5) with moderate satisfaction (score of 3/5) would yield a QoL score of 0.

Overall QoL mean scores are calculated as the average of the converted QoL item scores for the total number of items that were completed (i.e. were applicable) out of the 75-item questionnaire. Subdomain scores for being, belonging, and becoming are based on the average converted QoL item scores for those items that fall within each of the three subdomains of QoL-Being, QoL-Belonging, and QoL-Becoming.

Participants

Children with AIS/CSVT were identified in the Canadian Pediatric Ischemic Stroke Registry – Toronto site. All children up to 18 years of age with a diagnosis of AIS or CSVT who had either consented (>16y) or assented (>8y) with parent consent were eligible to participate in the study if they understood written English. Diagnostic criteria were (1) for AIS, acute onset of a clinical focal deficit (or seizures alone for infants) and computed tomography or magnetic resonance imaging showing infarction conforming to an established arterial territory; and (2) for CSVT, headache, diffuse, or focal neurological deficit or seizure, and magnetic resonance venography or computed tomography venography showing thrombus within venous channels. A parent from each household and children 10 years of age and older were invited to independently complete a parent proxy or self-report version of the questionnaire respectively. We did not invite younger children (aged 2–9y) to complete the questionnaires as we had not adapted the language and instructions by creating child versions more appropriate for younger participants.

Socio-economic status calculation methods

Socio-economic data were retrieved from 2006 Canadian census data corresponding to participant postal codes using Beyond 20/20 version 7.0 (Beyond 20/20 Inc., Ottawa, Canada). Unique census subdivision identifier codes were derived from each participant's postal code using the Canadian Census Analyser Postal Code Conversion Files.31 This is necessary as the Canadian Census does not use postal codes as a geographical unit, and this step is required to allow matching of postal codes to census data, for which socio-economic status data could then be retrieved. Based on this methodology, average and median family incomes were captured for each participant based on their census subdivision.

Validation methods

The instrument validation phase consisted of a mail out of the PSQLM and the Pediatric Quality of Life Inventory (PedsQL) questionnaire. The PedsQL is a standardized, generalized pediatric QoL questionnaire for children aged 2 to 18 years.32 The PedsQL measures QoL in terms of the impact of disease on an individual's physical, emotional, and social functioning. In contrast, the PSQLM assessment is not function based as there are no questions about how well or poorly a child performs in any of the domains. We hypothesized that high correlations between these two measures would not be expected given the differences in how the term ‘quality of life’ is operationalized as a construct between measures.

Self-report and parent versions of the PSQLM were mailed to all 153 children in the study population aged 10 to 18 years, for validation with parent proxy reports. Questionnaires were premarked with unique identifications associated with the participant's clinical and demographic information, to enable matching with the database of the Canadian Registry. Families were provided a postage-paid envelope to return the questionnaires. A sample of 72 participants (randomly selected from the original sample of 353) was asked to complete two PSQLM questionnaires 14 days apart to enable test–retest reliability analysis. For those participants, a second copy of the PSQLM was enclosed in a separate sealed envelope with instructions to complete 2 weeks after they completed and mailed the first copy. Only participants who returned both the initial and retest questionnaires were included in the data analysis for test–retest reliability.

In all children with AIS/CSVT the neurological deficit after stroke was assessed with serial examinations utilizing the Pediatric Stroke Outcome Measure (PSOM) in the children's stroke clinic from 1995. The PSOM is a standardized outcome measure used by paediatric neurologists to determine the neurological deficits (sensory motor, language, cognitive) of children after ischaemic stroke.33 The most recent PSOM summary score at the time of questionnaire completion was utilized.

Psychometric analyses

Data were analysed using PASW Statistics 18, Release Version 18.0.0 (formerly SPSS, Chicago, IL, USA). Parametric statistics were used to analyse those variables that were continuous. Descriptive statistics were calculated to define demographic and clinical data, including mean, SD, and range. Independent analyses of variance were performed to identify variance in PSQLM and PedsQL according to the degree of neurological deficit measured with the PSOM.

To establish construct validity, we tested domain-specific correlations using Spearman's rho correlation coefficients (r) comparing scores between subdomains of the PSQLM and the generic PedsQL. These were: PSQLM-Belonging with PedsQL-Social functioning; PSQLM-Becoming with PedsQL-School functioning and PSQLM-Being with both PedsQL-Physical Functioning; and PedsQL-Emotional Functioning. Finally, we also tested correlations between the total PSQLM with the total PedsQL score. We hypothesized that high correlations between these two measures would not be expected given the differences in the operationalization of the construct of quality of life in the two measures.

The reliability coefficient for test–retest reliability was analysed using the intraclass correlation coefficient (ICC), which is a reflection of both the correlation and agreement between the initial test and retest scores.34 ICC was also calculated to determine reliability in ratings between children and their parent proxy.

Respondents

Figure 2 shows the flow chart for the participants in the study. A total of 118 out of 353 (33%) packages were returned by mail. Of these, two were returned uncompleted with an attached note stating their child's ‘quality of life was fine’. Fifteen packages were returned unopened because of wrong addresses, leaving 101 of 353 (29%) valid returned packages.

One hundred parents and, independently, 44 children (aged 10–18y) returned completed proxy or self-report versions of the PSQLM. Ninety-seven per cent of respondents completed all items on the questionnaire. Clinical and demographic characteristics of the children of parent respondents are given in Table 1.

Most parent respondents were mothers (79.2%). Thirty-eight per cent (n=38/101) of the children were neonates at time of stroke. There were 69 (68%) children with arterial ischaemic-type stroke. There were 66 males (66%) and 35 (35%) females. Approximately half of the children had neurological deficits on the PSOM (52%), with moderate functional deficits in 27% and severe deficits in 25%. The mean age of the children of the parent proxy responders was 9 years and 9 months (SD 4.30). The average age of the children who completed PSQLM self-reports was 13 years and 6 months (SD 2.21). An average of 2 years 5 months had elapsed between the time the last PSOM was performed and the time of survey completion.

For test–retest reliability, 28 of 72 (39%) participants returned the initial PSQLM and 18 of 72 (25%) participants returned both the initial test and the retest. All questionnaires were completed by the same proxies initially and at retest, of which 15 of 18 (83%) were mothers. The mean age of children represented in the test–retest sample was 10 years and 9 months (SD 3.78).

PSQLM QoL scores

Descriptive statistics for the PSQLM overall QoL score and for the three subdomains of being, belonging, and becoming scores are given in Table 2. Using the scoring guidelines for the CHP's measure of QoL,28-30 scores that fell between 0.1 and 4.9 were coded as adequate QoL, greater than 5 as very good-to-excellent QoL, between 0 and –5 as poor QoL, and less than −5 as very poor QoL. Within the sample the ‘Overall PSQLM Parent Version QoL’ scores ranged from −6.69 to 9.78, and on the domains as follows: ‘being’ range −7.92 to 9.78; ‘belonging’ from −6.47 to 10.00; and ‘becoming’ from −5.91 to 10.00. These scores reflect that QoL was very good to excellent for some children, adequate for others, and poor or very poor for yet others. On the child version of the PSQLM the lower limit of the range of scores was not as severe as reported by parents. For the children overall PSQLM scores ranged between −1.91 to 9.45, with subdomains ranging from −2.39 to 10.00, suggesting that although QoL was perceived to be very good to excellent, some children did report a poor QoL. Scores on the PedsQL are interpreted on a scale of 0 to 100, with higher scores reflecting better QoL. Within our sample, PedsQL scores ranged from a low of 52.66 (SD 18.5), which reflects a moderately poor QoL, to a high of 87.17 (SD 11.18) reflecting a very good-to-high QoL.

Mean total score on the PSQLM (F_3,89=7.897, p<0.001) and PedsQL (F_3,74=19.600, p<0.001) varied with the degree of neurological deficit as measured on the PSOM (Table 3). There was a trend towards decreasing scores (lower QoL) on both the PedsQL and the PSQLM with increasing neurological deficit. Although mean PSQLM scores fell within the average range (0.71–4.65), more patients with severe neurological impairment were reported to have QoL within the poor category. Mean total score on the PSQLM and PedsQL was not significantly correlated with age at stroke (p=0.422 and p=0.423 respectively).

Reliability

As shown in Table 4, agreement between child-reported and parent-reported QoL scores based on 43 dyads (79.2% mothers) was moderate to excellent with ICCs ranging from 0.64 to 0.76. Test–retest reliability was excellent, with ICCs ranging from 0.91–0.94 (Table 5).

Internal consistency was excellent, with all ICCs exceeding 0.90 for the entire instrument (Cronbach's α=0.979, standardized α=0.986), and for each subdomain (being: α=0.954, standardized α=0.965; belonging: α=0.958, standardized α=0.969; becoming α=0.948, standardized α=0.964).

Validation

Construct validity was examined by Spearman's rho correlation coefficients (r). Results showed moderate correlations of PSQLM scores with domain-specific PedsQL scores, consistent with our a priori hypothesis that high correlations were not expected given the marked differences between measures in the operationalization of QoL. Specifically, correlations were, for belonging and PedsQL-Social functioning r=0.302 (p=0.003); for becoming and PedsQL-School functioning r=0.406 (p<0.001); for being and PedsQL-Physical Functioning r=0.375 (p<0.001); for being and PedsQL Emotional Functioning r=0.385 (p<0.001); and for overall QoL score and the PedsQL total score r=0.406 (p<0.001).

Content validation of questionnaire items was examined through consensus on validity of the selected items by parents of child stroke survivors and experts in the field. It was also assured by examination of the consistency of items within the three subdomains and overall PSQLM scores.