Susceptibility of children born very preterm (VP; <32 weeks’ gestation) to a range of cognitive deficits (Anderson, Reference Anderson2014) and neuroanatomical alterations in frontal, hippocampal, and subcortical regions places them at risk for impaired memory functioning (Nam et al., Reference Nam, Castellanos, Simmons, Froudist-Walsh, Allin, Walshe, Murray, Evans, Muehlboeck and Nosarti2015; Omizzolo et al., Reference Omizzolo, Scratch, Stargatt, Kidokoro, Thompson, Lee, Cheong, Neil, Inder, Doyle and Anderson2014; Thompson et al., Reference Thompson, Adamson, Roberts, Faggian, Wood, Warfield, Doyle, Anderson, Egan and Inder2013; Vollmer et al., Reference Vollmer, Lundequist, Mårtensson, Nagy, Lagercrantz, Smedler and Forssberg2017), including in episodic memory (memory for past experiences; Tulving, Reference Tulving2002) and prospective memory (memory for future intended actions; Einstein & McDaniel, Reference Einstein and McDaniel1990). Despite their apparent propensity for vulnerability in these domains, there is lack of consensus regarding episodic memory in children born VP compared with their term-born peers (≥37 weeks’ gestation), and little research has focused on prospective memory. Further investigation of episodic and prospective memory in children born VP is warranted, especially given the functional importance of these skills and potential adverse consequences of deficits in these areas for daily life.
Episodic memory, defined as learning and recall of newly learned information (Marsh & Roediger, Reference Marsh, Roediger, Weiner, Healy and Proctor2012), is important for knowledge acquisition (Greenberg & Verfaellie, Reference Greenberg and Verfaellie2010), academic achievement (Hassevoort et al., Reference Hassevoort, Khan, Hillman, Kramer and Cohen2018) and identity development (Allebone et al., Reference Allebone, Rayner, Siveges and Wilson2015). Prospective memory is required to remember to perform future-intended actions, such as remembering to attend football practice at 4:00 pm (time-based), or to hand in an assignment when the school bell rings (event-based) (Einstein & McDaniel, Reference Einstein and McDaniel1990).
Episodic and prospective memory have protracted developmental trajectories, and deficits in these domains may become more apparent in later childhood when increasing demands are placed on these skills as children progress through school (Wehrle et al., Reference Wehrle, Kaufmann, Benz, Huber, O’Gorman, Latal and Hagmann2016). Episodic memory is underpinned by the medial temporal, prefrontal, and parietal lobes (Hebscher & Voss, Reference Hebscher and Voss2020; Preston & Wagner, Reference Preston, Wagner, Martinez and Kesner2007) and basal ganglia (Gershman & Uchida, Reference Gershman and Uchida2019) and improves throughout childhood (Gascoigne et al., Reference Gascoigne, Smith, Webster, Barton, Gill and Lah2013; Gott & Lah, Reference Gott and Lah2014), coinciding with maturation of these underlying frontal–hippocampal networks (Ghetti & Bunge, Reference Ghetti and Bunge2012; Lebel et al., Reference Lebel, Walker, Leemans, Phillips and Beaulieu2008). Effective prospective memory relies on adequate episodic memory functioning (Groot et al., Reference Groot, Wilson, Evans and Watson2002; Kliegel et al., Reference Kliegel, Martin, McDaniel and Einstein2002), and is governed by many of the same neural networks (Cona et al., Reference Cona, Scarpazza, Sartori, Moscovitch and Bisiacchi2015), but with greater recruitment of executive functions (Kliegel et al., Reference Kliegel, Martin, McDaniel and Einstein2002; Mahy et al., Reference Mahy, Moses and Kliegel2014).
Studies using direct assessment of episodic memory have shown higher rates of episodic memory impairments in children born VP compared with term-born children (Luu et al., Reference Luu, Ment, Allan, Schneider and Vohr2011; Omizzolo et al., Reference Omizzolo, Scratch, Stargatt, Kidokoro, Thompson, Lee, Cheong, Neil, Inder, Doyle and Anderson2014). Omizzolo et al. (Reference Omizzolo, Scratch, Stargatt, Kidokoro, Thompson, Lee, Cheong, Neil, Inder, Doyle and Anderson2014) found that 7-year-olds born VP had significantly poorer delayed recall, but not immediate recall, of verbal associative information, as well as impaired memory for spatial information across all stages of learning and memory (i.e., from encoding to retrieval). Luu et al. (Reference Luu, Ment, Allan, Schneider and Vohr2011) found that adolescents born VP performed significantly worse than term-born children at all stages of memory for both verbal associative learning and episodic memory for more complex geometric figures. Conversely, others failed to report associations between gestational age and episodic memory performance (e.g., de Amorim et al., Reference de Amorim, de Castro Magalhães, Malloy-Diniz and Campos2013; Nassar et al., Reference Nassar, Kaczkurkin, Xia, Sotiras, Pehlivanova, Moore, Garcia de la Garza, Roalf, Rosen, Lorch, Ruparel, Shinohara, Davatzikos, Gur, Gur and Satterthwaite2019; Rushe et al., Reference Rushe, Rifkin, Stewart, Townsend, Roth, Wyatt and Murray2001), owing to methodological variability, such as different assessment measures, inclusion of relatively high functioning target groups, or lack of subgroup analyses restricted to the most vulnerable children, born VP. More studies of episodic memory are needed in representative samples of children born VP, as this population is prone to altered development of brain structures considered to play a central role in episodic memory functioning, including the hippocampus (Thompson et al., Reference Thompson, Adamson, Roberts, Faggian, Wood, Warfield, Doyle, Anderson, Egan and Inder2013), basal ganglia, and thalamus (Omizzolo, et al., Reference Omizzolo, Scratch, Stargatt, Kidokoro, Thompson, Lee, Cheong, Neil, Inder, Doyle and Anderson2014).
There is good reason to believe that children born VP are also at risk of deficits in prospective memory. Prospective memory relies on episodic memory, which may be an area of vulnerability in this population, as discussed previously. Effective prospective memory also relies on executive functioning, in which children and adolescents born VP have high rates of impairment (Anderson, Reference Anderson2014), associated with white matter pathology and dysmaturation (Nam et al., Reference Nam, Castellanos, Simmons, Froudist-Walsh, Allin, Walshe, Murray, Evans, Muehlboeck and Nosarti2015; Vollmer et al., Reference Vollmer, Lundequist, Mårtensson, Nagy, Lagercrantz, Smedler and Forssberg2017). Yet, research on prospective memory in children born VP is scarce. Only two studies to date have examined prospective memory in this population, and both showed that children born VP had poorer prospective memory than their term-born peers at 7–9 years (Ford et al., Reference Ford, Griffiths, Neulinger, Andrews, Shum and Gray2016) and 13 years of age (Isaacs et al., Reference Isaacs, Vargha-Khadem, Watkins, Lucas, Mishkin and Gadian2003). However, both studies focused on short-term event-based tasks (response to an external cue within an assessment session). To date, time-based and long-term tasks (completed beyond an assessment session) have not been examined, despite these tasks being more sensitive to prospective memory deficits (Bedard et al., Reference Bedard, Taler and Steffener2018; de Mendonça et al., Reference de Mendonça, Felgueiras, Verdelho, Câmara, Grilo, Maroco, Pereira and Guerreiro2018). Comprehensive assessment of prospective memory is needed to clarify the nature of prospective memory functioning in children born VP and to inform targets for development of effective interventions.
This study aimed to examine functional memory in children born VP by examining episodic and prospective memory in VP and term-born groups of children at 13 years of age. Based on preliminary evidence to date, it was hypothesized that children born VP would demonstrate poorer episodic and prospective memory than term-born children.
Methods
Study participants and procedure
Participants were children born <30 weeks’ gestational age who enrolled in the VIBeS Plus study, a randomized controlled trial of a preventative care program designed to improve infant development, behavioral regulation, parent–child interactions and parent mental health (Spittle et al., Reference Spittle, Ferretti, Anderson, Orton, Eeles, Bates, Boyd, Inder and Doyle2009). A cohort of 120 eligible infants born VP were recruited at term-equivalent age from the Royal Women’s Hospital and Royal Children’s Hospital in Melbourne between January 2005 and January 2007. This included infants without congenital anomalies likely to affect neurodevelopment and for the purpose of the intervention, who were discharged from hospital by 4 weeks of age corrected for prematurity, lived within 100 km of the hospital and were born to English-speaking families. Participants were randomized to the intervention or standard follow-up conditions. No significant cognitive benefits of the intervention were observed at 2 (Spittle et al., Reference Spittle, Anderson, Lee, Ferretti, Eeles, Orton, Boyd, Inder and Doyle2010), 4 (Spencer-Smith et al., Reference Spencer-Smith, Spittle, Doyle, Lee, Lorefice, Suetin, Pascoe and Anderson2012), or 7–8-year follow-ups (Spittle et al., Reference Spittle, Barton, Treyvaud, Molloy, Doyle and Anderson2016). Therefore, the children born VP from the intervention and standard follow-up groups were combined in the current study.
A comparison group of children born at term (born ≥37 weeks’ gestation) was recruited from metropolitan Melbourne kindergartens at 4 years of age. Eligible participants were from English-speaking families and without history of significant congenital abnormalities or developmental difficulties.
The current study used data from the VIBeS Plus 13-year follow-up, which was approved by the Royal Children’s Hospital Human Research Ethics Committee and was completed in accordance with the Helsinki Declaration. Participants were assessed between September 2018 and March 2020. All children underwent a neuropsychological assessment at 13 years, corrected for prematurity to minimize bias in cognitive test scores in those born preterm (Wilson-Ching et al., Reference Wilson-Ching, Pascoe, Doyle and Anderson2014). Assessors were blind to perinatal history and previous assessment performances. At the 13-year follow-up, 81 of 118 (69%) surviving children born VP were assessed, while 26 of 41 (63%) children in the term group were assessed (see Figure 1). Data collection was terminated prematurely due to COVID-19 restrictions.
Figure 1. Flow diagram illustrating participant recruitment and allocation. Note. Number of participants lost to follow-up by reason for nonparticipation at each timepoint are cumulative from initial recruitment until 13 years (i.e., there were only two deaths overall).
Measures
Episodic memory
The Children’s Memory Scales (CMS; Cohen, Reference Cohen1997) Dot Locations and Word Pairs subtests, validated for use in children aged 5–16 years, were used to measure spatial and verbal episodic memory, respectively. Dot Locations involves learning the spatial location of an array of eight dots on a four-by-four grid across three learning trials with a 5-s exposure. A distractor trial is followed by a fourth recall trial and a delayed recall trial 20–30 min later. The Word Pairs subtest requires participants to immediately recall the second word from 14 abstract word pairs, heard over three learning trials, with feedback for errors. Delayed recall and recognition trials are administered 20–30 min later. Both subtests yield Learning, Immediate Recall and Delayed Recall scores, as well as a Recognition score for the Word Pairs subtest. Scores were converted to age-standardized scaled scores (M = 10, SD = 3).
Test-retest reliability coefficients range from .70 to .81 and .73 to .94 for Dot Locations and Word Pairs scores, respectively, in children aged 13–14 years (Cohen, Reference Cohen1997), indicating acceptable-to-excellent reliability (Groth-Marnat & Wright, Reference Groth-Marnat and Wright2016). Patterns of intercorrelations of CMS subtests provide evidence of adequate convergent and divergent validity and the CMS has demonstrated clinical sensitivity in detecting mild-to-moderate memory difficulties in children with neurodevelopmental disorders (Cohen, Reference Cohen1997).
Prospective memory
The PM Test was adapted for use with children from existing, behavioral prospective memory tests validated in adults (e.g., Royal Prince Alfred Prospective Memory Test [RPA-ProMem], Rivermead Behavioral Memory Test [RBMT] – 3rd edition, Memory for Intentions Screening Test [MIST]; Radford et al., Reference Radford, Lah, Say and Miller2011; Wilson et al., Reference Wilson, Greenfield, Clare, Baddeley, Cockburn, Watson, Tate, Sopena and Nannery2008; Woods et al., Reference Woods, Moran, Dawson, Catherine and Grant2008). The MIST has been previously adapted for use with children (Mills et al., Reference Mills, Garbarino and Raskin2021; Robey et al., Reference Robey, Buckingham-Howes, Salmeron, Black and Riggins2014). We adopted a similar approach, with a broader range of time delays across short-term (within assessment session) items, and inclusion of long-term (up to 1-week post-session) items, similar to the design of the RPA-ProMem.
The PM Test comprised eight items, with an equal weighting of time- and event-based items across six short-term items and two long-term items, respectively (see Supplementary Table 1 for full list of items). Thus, each item formed part of two scales (e.g., time-based and short-term). Items were designed to mimic everyday prospective memory tasks that may be encountered in the age range of participants in this study. For example, item 5 (Supplementary Table 1) requires the participant to remember to get a pen when they return to their desk from a standing activity. This item mimics real-world examples of prospective memory tasks, such as teachers asking students to prepare certain materials for an activity in a classroom environment, or instructions that parents might give their children when they transition between activities.
Short-term items were administered at set times throughout a 4-hr neuropsychological assessment and long-term items required participants to respond up to 1-week following the assessment. Event-based items required participants to respond to environmental cues (e.g., timer sound, when instructed to sit at the table), whilst time-based items required participants to monitor the time on their preferred device (personal watch, mobile phone, or surroundings) and respond at set times. Items were scored 0 (no response or incorrect response at incorrect time), 1 (partial correct response) or 2 (correct response at correct time). Scores were summed to yield a Total score (ranging from 0 to 16) as well as Time-based (0–8), Event-Based (0–8), Short-Term (0–12), and Long-Term (0–4) scores.
In cases in which a single time-based or event-based item was not administered due to factors unrelated to the child’s ability (e.g., administration error, written tasks unable to be completed due to neurosensory impairment, school/home assessments requiring protocol changes), a score was imputed using the median of the remaining scores of that scale (i.e., remaining time- or event-based scores). Subscale scores were recorded as missing if more than one item from the scale was not administered.
Parent-reported episodic and prospective memory
Parents rated their children’s functional memory on the Observer Memory Questionnaire – Parent Form (OMQ-PF; Gonzalez et al., Reference Gonzalez, Anderson, Wood, Mitchell, Heinrich and Harvey2008), validated for use with children aged 5–16 years. The OMQ-PF comprises 27 items of memory function in everyday scenarios that demand utilization of episodic memory (e.g., “does your child recall details of previous conversations?”) and prospective memory (e.g., “do you have to provide reminders for your child?”), as well as items of parents’ perceptions about their child’s memory function (e.g., “compared with other children of the same age, his/her memory ability is poor”). Items were scored on a 5-point Likert scale, with some reversed scored items, and summed to produce a total score ranging from 27 to 135, whereby higher scores reflect more optimal memory. The mean total score in a sample of 376 healthy Australian children who participated in a validation study was 107.26 (SD = 13.00) (Gonzalez et al., Reference Gonzalez, Anderson, Wood, Mitchell, Heinrich and Harvey2008). Total scores were considered invalid if two or more items of the questionnaire were incomplete.
The OMQ-PF has demonstrated sound internal consistency, with item-total coefficients ranging from .36 to .79 and Cronbach’s α of .92 (Gonzalez et al., Reference Gonzalez, Anderson, Wood, Mitchell, Heinrich and Harvey2008). In healthy children, the scale correlates primarily with measures of novel associative learning. The OMQ-PF is sensitive to memory impairment in children with temporal lobe epilepsy and has been shown to correlate more strongly with several memory measures in this population than in healthy controls.
Across all age-standardized measures (CMS subtests), scores greater than one standard deviation below the normative mean were classified as falling below age expectations; a cut-off that has been used to define impairment relative to typically developing children in this cohort previously (Spittle et al., Reference Spittle, Barton, Treyvaud, Molloy, Doyle and Anderson2016), other studies of children born VP (Pittet-Metrailler et al., Reference Pittet-Metrailler, Mürner-Lavanchy, Adams, Bickle-Graz, Pfister, Natalucci, Grunt and Borradori Tolsa2019), and other pediatric populations (Azevedo et al., Reference Azevedo, Kothur, Gupta, Webster, Dale, Wade, Gill and Lah2020; Perrin et al., Reference Perrin, Heller and Loe2019). A cut-off of one standard deviation acknowledges the potential functional implications of mild reductions from normative comparisons (Sohlberg & Mateer, Reference Sohlberg and Mateer2001). A cut-off of greater than one standard deviation below the term group mean was used when age-standardized normative data were unavailable (PM test and OMQ-PF).
General intellectual ability
Two Wechsler Abbreviated Scale of Intelligence – Second Edition (WASI-II; Wechsler & Zhou, Reference Wechsler and Zhou2011) subtests (Vocabulary and Matrix Reasoning) were used to obtain an estimate of full scale intellectual ability (FSIQ-2; M = 100, SD = 15). Participants with an estimated FSIQ-2 greater than two standard deviations below the test norm mean (i.e., <70) were categorized as having a general intellectual impairment.
Social risk
Social risk was assessed to describe the sociodemographic status of the sample using the Social Risk Index (SRI; Roberts et al., Reference Roberts, Howard, Spittle, Brown, Anderson and Doyle2008), reported by parents in a questionnaire at 13 years. The SRI is based on six factors, scored 0 (lower risk), 1 or 2 (higher risk), respectively: family structure (two caregivers, separated parents, single caregiver), highest level of primary caregiver education (tertiary, 11–12 years, <11 years), primary income earner employment status (full time, part time, unemployed) and occupation (skilled, semi-skilled, unskilled), language spoken at home (English, some English, no English), and maternal age at birth (>21 years, 18–21 years, <18 years). Scores from each factor were summed and used to categorize families into lower (0 to 1) or higher (≥2) social risk around the median, as has been done in this cohort previously (Roberts et al., Reference Roberts, Howard, Spittle, Brown, Anderson and Doyle2008; Spittle et al., Reference Spittle, Treyvaud, Lee, Anderson and Doyle2017).
Statistical analyses
Data were analyzed using Stata 16.0 (StataCorp, 2019). Univariable linear and logistic regressions were used for continuous and categorical variables, respectively, to compare VP and term groups on important demographic characteristics, as well as those who participated and those lost to current follow-up. In cases where the frequency was ≤5 for categorical participant characteristic variables, Fisher’s exact test was used.
Mean group differences in continuous memory outcome data were examined using linear regressions, adjusted for social risk due to observed between-group differences in social risk (see Results) and the known association between higher social risk and poorer long-term cognitive outcomes in children born VP (Doyle et al., Reference Doyle, Cheong, Burnett, Roberts, Lee and Anderson2015). Models adjusted for age at assessment for outcome measures that were not age-standardized (prospective memory and parent-reported memory measures). Group differences in rates of below age-expected performance (categorical memory data) were examined using logistic regressions, also adjusted for social risk for the same reason. All models were fitted with generalized estimation equations (GEEs) with an exchangeable correlation structure and robust standard errors to adjust for correlations between twins within the same family.
Sensitivity analyses excluded children with significant developmental delay (defined as FSIQ-2 of <70 and/or formal diagnosis of cerebral palsy) to examine the possible influence of a small proportion of impaired children on the results.
To account for multiple comparisons, all significance values were corrected using the Benjamini–Hochberg false discovery rate of .05 (Benjamini & Yekutieli, Reference Benjamini and Yekutieli2001). Interpretation of results considered the overall pattern and magnitude of between-group differences across outcome measures.
Results
Participant characteristics
Participant characteristics are summarized in Table 1. The VP and term groups were similar in terms of sex distribution, but the VP group had a lower mean IQ (mean difference = −10.1; 95% confidence intervals [CIs] −15.7 to −4.5). Children in the VP group were also more likely to be slightly older (0.2 years; −.01 to 0.4) and of higher family social risk (p = .05, Fisher’s exact test) (Table 1).
Table 1. Participant characteristics
Note. n = number; M = mean; SD = standard deviation; BPD = bronchopulmonary dysplasia; NEC = necrotizing enterocolitis; WMI = white matter injury; IVH = intraventricular hemorrhage; PVL = periventricular leukomalacia; CP = cerebral palsy; FSIQ-2 = estimate of intellectual ability based on two subtests.
a MRI data available for n = 72.
b Questionnaires returned for n = 79.
Higher social risk refers to the presence of multiple sociodemographic factors known to affect development based on family structure, primary caregiver education, primary income earner employment status and occupation, language spoken at home, and maternal age at birth.
Among term group participants, those assessed and those lost to follow-up did not differ on any important characteristics (Supplementary Table 2). Among the VP group, those assessed and those lost to follow-up were similar across most important characteristics, although VP non-participants were of higher neonatal medical risk, including lower mean birthweight (−140; −242 to −38) and higher rates of necrotizing enterocolitis (odds ratio [OR] = 2.9; CIs 1.0 to 8.4) (Supplementary Table 2).
Episodic memory (CMS word pairs and dot locations)
The VP group performed poorer than the term group on all measures of visuospatial and verbal episodic memory (see Table 2). While the rate of below age-expected performance was higher in the VP group than the term group on episodic measures (odds ratios [ORs] ranging from 2.2 to 7.2), these differences did not reach statistical significance.
Table 2. Memory outcomes and rates of below age-expected performance a in children born very preterm compared with term-born controls
Note. VP = very preterm; n = number; M = mean; SD = standard deviation; β = regression coefficient representing mean group differences; CI = confidence interval; OR = odds ratio; CMS = children’s memory scale; DL = dot locations subtest; WP = word pairs subtest; OMQ-PF = observer memory questionnaire-parent form; N/A = not applicable.
a Below age-expected performance defined as more than 1 SD below normative mean for age-standardized measures (CMS subtests), or more than 1 SD below the term group mean for all remaining measures (prospective memory test, OMQ-PF).
b n ranges from 79 to 81 due to missing outcome data.
c n ranges from 103 to 105 due to missing outcome and social risk data.
d Odds ratio unable to be computed due to cell size of zero in one group.
All models adjusted for social risk, while linear regression models comparing prospective and everyday memory also adjusted for corrected age at assessment.
Prospective memory
The VP and term groups performed similarly on measures of event-based and long-term prospective memory, but there was evidence that the VP group performed more poorly on measures of time-based and short-term prospective memory relative to the term group (Table 2). There was a higher proportion of children in the VP group performing more than 1 SD below the control group across time-based and short-term aspects of prospective memory (ORs of 2.5 and 3.0, respectively), but the evidence was weak.
Parent-reported everyday memory (OMQ-PF)
Parents of children born VP reported more everyday episodic and prospective memory difficulties in their children compared with parents of children born at term (Table 2). A significantly greater proportion of children born VP displayed elevated functional memory difficulties based on parent ratings, as the VP group had almost 12 times the odds to exhibit everyday difficulties than the term group.
All results remained the same after sensitivity analyses excluding children with significant developmental delay (n = 6) (Supplementary Table 3), and after adjusting for multiple comparisons.
Discussion
Our study provides strong evidence of memory difficulties in children born VP at 13 years relative to their term-born peers. We found episodic memory difficulties, encompassing learning, immediate recall, delayed recall and recognition, of both verbal and visual material in this population compared with term-born peers. We also documented selective prospective memory difficulties in VP 13-year-olds, involving time-based but sparing event-based prospective memory. Moreover, in our assessment, children born VP were more challenged by prospective memory tasks at short delays (within the assessment session) but not at long delays (up to a week post assessment). Between-group differences ranged from around half to three quarters of a standard deviation across episodic and prospective measures, respectively, indicating a moderate-to-strong effect. These findings were supported by parents of children born VP who reported greater challenges in their children’s ability to undertake everyday memory tasks compared with parents of term-controls, to the effect of just under one-and-a-half standard deviations.
Our study is novel given the comprehensive assessment of prospective memory, incorporating both event- and time-based measures, in children born VP. The study revealed poorer performance in time- but not event-based prospective memory in children born VP. This differential difficulty in prospective memory may be related to time-based tasks being cognitively more demanding than event-based tasks (Einstein et al., Reference Einstein, McDaniel, Richardson, Guynn and Cunfer1995; Mahy et al., Reference Mahy, Moses and Kliegel2014), and particularly sensitive to neurological impairments (Bedard et al., Reference Bedard, Taler and Steffener2018; de Mendonça et al., Reference de Mendonça, Felgueiras, Verdelho, Câmara, Grilo, Maroco, Pereira and Guerreiro2018). The difference in cognitive demand between time- and event-based tasks may be related to the presence or absence of external cues. Event-based tasks include external cues that likely trigger cued retrieval, while time-based tasks rely primarily on internal processes, executive functions and time-monitoring, and require self-initiated retrieval processes (Einstein et al., Reference Einstein, McDaniel, Richardson, Guynn and Cunfer1995; Mahy et al., Reference Mahy, Moses and Kliegel2014). In children born VP, time-based prospective memory difficulties might be related to poorer executive functioning (Anderson, Reference Anderson2014) associated with white matter pathology and dysmaturation (Nam et al., Reference Nam, Castellanos, Simmons, Froudist-Walsh, Allin, Walshe, Murray, Evans, Muehlboeck and Nosarti2015; Vollmer et al., Reference Vollmer, Lundequist, Mårtensson, Nagy, Lagercrantz, Smedler and Forssberg2017). Poorer episodic memory may also contribute to the observed prospective memory difficulties (Kinsella, Reference Kinsella2010; Kliegel et al., Reference Kliegel, Martin, McDaniel and Einstein2002; Mahy et al., Reference Mahy, Moses and Kliegel2014).
Unlike previous studies (Ford et al., Reference Ford, Griffiths, Neulinger, Andrews, Shum and Gray2016; Isaacs et al., Reference Isaacs, Vargha-Khadem, Watkins, Lucas, Mishkin and Gadian2003), we did not find evidence for event-based prospective memory difficulties in children born VP compared with their term-born peers. It is possible that the measure used in our study was not sufficiently challenging to detect subtle deficits in event-based prospective memory. For instance, half of the event-based items in our measure required children to respond to a timer sound (event-based cue), which might have been a too obvious trigger. In contrast, other studies have used more subtle cues, such as verbal cues in conversation (phrase: “Let’s try something different”; Ford et al., Reference Ford, Griffiths, Neulinger, Andrews, Shum and Gray2016) or even less obvious environmental cues (remembering to find an object whilst retracing a route; Isaacs et al., Reference Isaacs, Vargha-Khadem, Watkins, Lucas, Mishkin and Gadian2003).
Our finding that children born VP exhibit poorer performance in short-term (within assessment session) prospective memory is consistent with previous studies (Ford et al., Reference Ford, Griffiths, Neulinger, Andrews, Shum and Gray2016; Isaacs et al., Reference Isaacs, Vargha-Khadem, Watkins, Lucas, Mishkin and Gadian2003), but surprisingly there was no evidence of difficulties in long-term (up to 1-week post-session) prospective memory in this group. We expected to find deficits in long-term prospective memory, given the vulnerability in short-term prospective memory that has previously been reported and because long-term tasks are more sensitive in detecting prospective memory impairments than short-term tasks (Bedard et al., Reference Bedard, Taler and Steffener2018; de Mendonça et al., Reference de Mendonça, Felgueiras, Verdelho, Câmara, Grilo, Maroco, Pereira and Guerreiro2018).
A lack of between-group differences in long-term prospective memory in this study might again be due to the design of the measure. Short-term items were administered in a controlled assessment environment, compared with long-term items that provided participants with additional time for rehearsal, greater opportunity for exposure to external cues (e.g., twins inadvertently prompting one another or reminders from parents), and enabled them to employ additional strategies (e.g., to-do note or electronic reminder). In terms of scoring, there was a limited range of scores; with a maximum total score of four for long-term prospective memory, compared with 12 for short-term tasks. Leeway was also provided in scoring an item that required children to mail a questionnaire to account for postal delays. This made it challenging to differentiate between children who mailed their questionnaire on time or slightly late. Future research may consider the use of electronic media, although this could disadvantage children from lower socioeconomic backgrounds who might not have access to technological devices.
With regards to episodic memory, our study provides further insight into episodic memory difficulties in children born VP. Omizzolo et al. (Reference Omizzolo, Scratch, Stargatt, Kidokoro, Thompson, Lee, Cheong, Neil, Inder, Doyle and Anderson2014) found deficits in younger children born VP at all stages (i.e., from encoding to retrieval) of episodic memory for spatial information, but only in delayed recall for verbal information. Our study found below age-expected episodic memory performance at all stages of memory for both verbal and visual information; a similar pattern that has been observed in VP 16-year-olds (Luu et al., Reference Luu, Ment, Allan, Schneider and Vohr2011). Episodic memory difficulties in children born VP may be related to disrupted frontotemporal and cortical–subcortical networks, which are particularly critical for episodic memory (Hebscher & Voss, Reference Hebscher and Voss2020; Preston & Wagner, Reference Preston, Wagner, Martinez and Kesner2007). Cross-sectional studies in children born VP have reported a reduction in hippocampal volume compared with term controls that may increase with age: 3% reduction in hippocampal volume during infancy (Thompson et al., Reference Thompson, Omizzolo, Adamson, Lee, Stargatt, Egan, Doyle, Inder and Anderson2014), 6% reduction at 7 years (Omizzolo et al., Reference Omizzolo, Thompson, Scratch, Stargatt, Lee, Cheong, Roberts, Doyle and Anderson2013), and 12–15% at 15 years (Nosarti et al., Reference Nosarti, Al-Asady, Frangou, Stewart, Rifkin and Murray2002), with this magnitude of reduction persisting into adulthood (Aanes et al., Reference Aanes, Bjuland, Skranes and Løhaugen2015). Abnormalities in other brain regions relevant to episodic memory have also been identified in children born VP, including in the fornix (Kelly et al., Reference Kelly, Thompson, Cooper, Pham, Nguyen, Yang, Ball, Adamson, Murray, Chen, Inder, Cheong, Doyle and Anderson2020), basal ganglia and thalamus (Omizzolo et al., Reference Omizzolo, Scratch, Stargatt, Kidokoro, Thompson, Lee, Cheong, Neil, Inder, Doyle and Anderson2014). Based on these studies, we postulate that children born VP may grow into memory deficits, such that episodic memory difficulties do not resolve, but rather persist as the episodic memory system matures across childhood (Ghetti & Bunge, Reference Ghetti and Bunge2012). Longitudinal research is needed to elucidate whether children born VP grow into their memory deficits, in line with the altered maturational trajectory of the hippocampus.
Our study highlights that alongside episodic and prospective memory difficulties on testing, children born VP experience significant difficulties with memory in everyday life. The odds of below age-expected ability in everyday tasks that require episodic and prospective memory were twelve times higher in children born VP compared with term-born children, indicating that poor memory performance on direct assessment may be manifesting functionally in everyday situations. Of note, the greater magnitude of between-group differences based on parent-report compared with smaller group differences evident on direct memory assessment illustrates the potential benefit of a structured, one-to-one assessment setting with few distractions. The memory difficulties we identified in children born VP in an ideal assessment setting are likely to be more pronounced in everyday settings, such as classrooms. Consequences may be far-reaching, as episodic memory is important for learning new concepts (Greenberg & Verfaellie, Reference Greenberg and Verfaellie2010), academic progress (Hassevoort et al., Reference Hassevoort, Khan, Hillman, Kramer and Cohen2018), and development of self-identity (Allebone et al., Reference Allebone, Rayner, Siveges and Wilson2015). Prospective memory difficulties may have similar consequences, such as challenges with following through with classroom instructions and development of independence that is increasingly required as children progress through school (Henry et al., Reference Henry, Terrett, Altgassen, Raponi-Saunders, Ballhausen, Schnitzspahn and Rendell2014).
The wide-ranging memory difficulties we identified in children born VP highlight the need for intervention. Although few studies have evaluated memory interventions in children, some efficacy has been shown for compensatory strategies (e.g., electronic memory aids; Mahan et al., Reference Mahan, Rous and Adlam2017; Wilson et al., Reference Wilson, Emslie, Evans, Quirk, Watson and Fish2009), instructional learning strategies (e.g., errorless learning; Haslam et al., Reference Haslam, Hodder and Yates2011; Haslam et al., Reference Haslam, Wagner, Wegener and Malouf2017), and programmatic approaches that combine training of metacognition, attention, and strategy use (Catroppa et al., Reference Catroppa, Stone, Hearps, Soo, Anderson and Rosema2015). Broader memory rehabilitation literature suggests that interventions have most benefit when they combine approaches (e.g., internal and external strategies), target everyday memory complaints, and employ methods with clear functional relevance for everyday life (Parker et al., Reference Parker, Haslam, Fleming, Shum, Wilson, Winegardner, van Heugten and Ownsworth2017). For example, training children with everyday memory difficulties following acquired brain injury in both self-instruction and diary use has been found to significantly improve everyday functional outcomes, including frequency of diary use and ability to perform daily routines that require recall of information and events (Ho et al., Reference Ho, Epps, Parry, Poole and Lah2011). Our results suggest comprehensive interventions that foster memory skills and promote generalizability of these skills to everyday life are warranted for children born VP.
The use of a comprehensive measure of prospective memory in this study was novel and a major strength, enabling us to explore more functional aspects of memory as they may manifest for children born VP in everyday contexts. However, we acknowledge some methodological limitations. The lack of validated measures of prospective memory in children resulted in the use of an experimental measure, the sensitivity, and validity of which are yet to be established. The nature of the prospective memory measure also limited the degree of environmental control that could be used to reduce the influence of external factors on participant performance. For example, participants were free to employ additional strategies and may have been exposed to additional prompting beyond the assessment setting for long-term items. There was sample size imbalance across groups, with a smaller control group, which can affect statistical power and increase the Type II error rate. Unfortunately, recruitment ceased early in line with government regulations limiting face-to-face assessment during the COVID-19 pandemic.
It is also important to note two considerations in the interpretation of our data. Firstly, our findings are based on scores corrected for prematurity, given that use of chronological age underestimates true cognitive ability in children born preterm throughout childhood and adolescence, particularly at lower gestational ages (Wilson-Ching et al., Reference Wilson-Ching, Pascoe, Doyle and Anderson2014). Age correction is uncommon in clinical practice, and group differences would likely be even greater without this correction. Secondly, measures with US norms can result in inflated scores in Australian samples, as has been observed across various measures for example, the Bayley-III (Spencer-Smith et al., Reference Spencer-Smith, Spittle, Lee, Doyle and Anderson2015), original WASI (Beauchamp et al., Reference Beauchamp, Dooley and Anderson2013; Cheong et al., Reference Cheong, Anderson, Roberts, Burnett, Lee, Thompson, Molloy, Wilson-Ching, Connelly, Seal, Wood and Doyle2013), and WISC-IV (Hutchinson et al., Reference Hutchinson, De Luca, Doyle, Roberts and Anderson2013). This is the likely explanation for higher-than-expected IQ scores for the VP and control groups in our study.
We propose several directions for further research. Given development of executive functioning beyond 13 years (Anderson et al., Reference Anderson, Anderson, Northam, Jacobs and Catroppa2001), further research is needed to understand the developmental trajectories of functional memory systems that interact with executive functions. More broadly, our findings and emerging evidence of prospective memory difficulties in other pediatric populations (e.g., traumatic brain injury; Phillips et al., Reference Phillips, Shum, Mandalis, Parry, Benson, Morrow, Epps and Lah2018) highlight that research regarding functional memory in children is lacking and warrants further attention, which relies on further development and validation of pediatric prospective memory measures. Most importantly, intervention approaches are needed to support children with functional memory difficulties. These interventions may be informed by better understanding the mechanisms for poorer prospective memory in children born VP. To this end, examining the relationship between prospective memory functioning, episodic memory, and executive functions thought to underly effective prospective memory (i.e., working memory, planning, cognitive flexibility, inhibition, self-monitoring) may be an important starting point (Kliegel et al., Reference Kliegel, Martin, McDaniel and Einstein2002; Mahy et al., Reference Mahy, Moses and Kliegel2014). Finally, further research is required to understand whether the memory difficulties identified in children born VP are indicative of a specific vulnerability and primary memory deficit, or secondary to more generalized cognitive impairment.
This study emphasizes that the challenges associated with VP birth also often include functional memory difficulties, such as verbal and visuospatial episodic memory, time-based prospective memory, and everyday memory tasks. The continuation of functional memory difficulties beyond early childhood also indicates the importance for more long-term surveillance of cognitive and functional outcomes in children born VP, and underscores the importance for clinicians to consider prospective memory in pediatric neuropsychological assessment.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1355617722000170
Acknowledgements
We thank all families who participated in this study and all members of the VIBeS research team, particularly our research nurse, Merilyn Bear, as well as Isabelle Raiter, Sarit van Veen, Rachel Ellis, and Ngoc Nguyen for their contribution to data collection.
Funding statement
This study was funded by grants from the National Health and Medical Research Council (Project Grant ID 284512; Senior Research Fellowship ID 1081288 and Investigator Grant ID 1176077 [PJA]; Career Development Fellowship ID 1108714 [AJS]; Centre of Research Excellence Grants 1060733 and 1153176); The Cerebral Palsy Alliance Project Grant; Murdoch Children’s Research Institute, Myer Foundation, Allens Arthur Robinson Foundation, Thyne Reid Foundation, Victorian Government’s Operational Infrastructure Support Program, and Monash University Research Support Fund.
Conflicts of interest
None.
Open access
