The roles of deliberate practice and innate ability in developing expertise: evidence and implications
Abstract
Context
Medical education research focuses extensively on experience and deliberate practice (DP) as key factors in the development of expert performance. The research on DP minimises the role of individual ability in expert performance. This claim ignores a large body of research supporting the importance of innate individual cognitive differences. We review the relationship between DP and an innate individual ability, working memory (WM) capacity, to illustrate how both DP and individual ability predict expert performance.
Methods
This narrative review examines the relationship between DP and WM in accounting for expert performance. Studies examining DP, WM and individual differences were identified through a targeted search.
Results
Although all studies support extensive DP as a factor in explaining expertise, much research suggests individual cognitive differences, such as WM capacity, predict expert performance after controlling for DP. The extent to which this occurs may be influenced by the nature of the task under study and the cognitive processes used by experts. The importance of WM capacity is greater for tasks that are non-routine or functionally complex. Clinical reasoning displays evidence of this task-dependent importance of individual ability.
Conclusions
No single factor is both necessary and sufficient in explaining expertise, and individual abilities such as WM can be important. These individual abilities are likely to contribute to expert performance in clinical settings. Medical education research and practice should identify the individual differences in novices and experts that are important to clinical performance.
Introduction
The question of how people become experts has a long history. Traditional and popular views invoke innate differences in ‘talent and giftedness’ as likely explanations, but, as Ericsson et al.1 showed some time ago, the notion of innate talent as critical derives more from anecdotes about individuals, such as Tiger Woods, Bobby Fischer and Mozart, than it does from hard evidence. The main thesis of these studies is that the development of expertise requires extensive deliberate practice (DP).2-4 This alone is hardly surprising; what makes the observation more interesting is the claim that extensive DP is necessary and sufficient to the development of expertise.5, 6 According to Ericsson et al., innate individual ability plays little or no role in explaining expert performance.1, 2 Instead, accumulating 10 000 hours of DP eliminate individual differences in intelligence or personality – a claim that has been popularised by high-profile media articles and books.7, 8
Deliberate practice has represented a common theme in several recent articles in medical education journals, which have suggested that the implementation of training that fosters DP is sufficient to achieve high expert performance.9-13 In this view, DP is both the goal and the mechanism by which skills training interventions produce optimal learning outcomes. In concept learning14 and clinical skills,2 DP is invoked to account for the success of individual trainees, as well as to validate the efficacy of training processes. In general medicine too, the 10 000-hour rule has passed into mainstream discourse15 and is being used as a guiding principle13 in structuring training and education.
These applications of the DP principle by and large ignore the impact of individual differences. This is consistent with current medical education paradigms, which rarely address differences in individual ability, although, curiously, admissions and selection procedures are entirely concerned with measuring individual differences. Once a cohort of trainees has been selected, the goal is to train them to a minimum level of competence. Although common sense and empirical evidence16-18 suggest that significant individual differences remain among even expert doctors, efforts to distinguish the extents and sources of these differences are minimal. This position would be defensible if the only individual difference between experts that mattered was the extent of accumulated DP.
The notion that practice alone is sufficient for mastery (and indeed for elite performance) has popular appeal, but, like many such apparent universal truths, it may represent only a first approximation: innate individual differences may not be as irrelevant as has been implied. If there is evidence that individual differences do continue to exert an impact on expert performance even after extensive DP, then understanding how and why individual variation systematically influences expertise is crucial. Both education processes and research on the nature of learning will have to account for the impact of individual ability on learning and performance. Is there evidence to suggest that some innate individual differences predict expert performance even after controlling for DP?
This critical narrative review examines the relationships between expert performance, and DP and a critical innate cognitive ability: working memory (WM). We examine the literature that has explicitly tested how DP and WM account for the development of expertise and their relative contributions in explaining performance differences. We then contextualise the findings of the literature and suggest some implications.
Deliberate practice
Deliberate practice refers to strategic, focused, goal-oriented activities aimed at improving components of performance.1 It demands that the learner concentrates fully, analyses feedback, and executes repetitions with the aim of achieving a level of refine-ment just beyond his or her current level of performance.2, 5, 6 Deliberate practice is often executed in environments that allow for repetition and the provision of regular feedback.19, 20 Ericsson et al. first proposed the importance of accumulated DP in a structured study of expert musicians,1 among whom DP consisted of routine and dedicated practice on difficult tasks for 50–60 hours per week. They noted that the extensive accumulated DP was the distinguishing factor between expert and novice musicians.1 Typically using interviews, journaling and retrospective recall, other similar investigations of DP have been conducted in sports,21, 22 music,1, 23 aviation crisis management,24 chess25 and even insurance sales.26 In all of these domains, DP was shown to be significantly related to some aspect of expert performance and was often touted as the ultimate factor in differentiating true experts from non-experts.
Proponents of the DP model suggest that practice over a minimum of 10 years, or a cumulative 10 000 hours of DP, is the threshold for achieving expertise.1, 27 This commonly cited figure was first proposed by Chase and Simon28 in their seminal investigations of expertise in chess as the ‘10-year rule’. Although confirmatory studies suggest that this is a good approximation to the average required amount of DP, the evidence also points to wide variability around this mean.29 In Chase and Simon's original work,28 the estimated required time ranged from < 1000 to > 30 000 hours of practice. More recent research supports the claim for a wide range of accumulated practice. Gobet and Campitelli29 studied chess grand masters' patterns of practice, using Ericsson's methods and a large sample (n = 104) of chess players. The accumulated amount of DP among masters ranged from 3000 to 23 600 hours. Conversely, some non-masters had amassed > 20 000 hours of DP, but had not reached the same performance level. Gobet and Campitelli29 also found that masters did not differ from non-titled international players in the amount of DP they accumulated over the first 3 years, but already showed significant differences in their ratings. This large variation among experts leaves ample room for other factors, including innate individual cognitive differences.
However, the DP model proposed by Ericsson and Lehmann20 claims that innate intelligence, memory and other cognitive abilities have little impact on expert performance. Although the model acknowledges that differing levels of individual abilities may account for differences in performance among novices, the accumulation of DP and progression of expertise are assumed to reduce the effect of these factors. Differences in performance among experts are presumed to be attributable solely to differences in accumulated DP. This claim is not consistent with the extensive literature on individual cognitive differences, which shows that these factors are related to performance. Although many different cognitive factors vary among individuals, including general intelligence30 and sub-aptitudes such as visual–spatial ability, verbal reasoning, numeracy, auditory processing and abstract reasoning,31, 32 one of the most important is WM capacity,33 which, in turn, is related to measures of general intelligence.34-36
Working memory
Working memory refers to an individual's ability to store, access and manipulate recently acquired information for the purpose of task completion.37, 38 It refers to more than a short-term storage facility for recall. Rather, WM is composed of components that integrate sensory stimuli, coordinate executive control or attention, and store information temporarily.39 It is essentially the primary ‘arena of computation’ that allows for active cognitive processing of complex tasks.33
Importantly, WM is considered to be finite and limited,40 and shows considerable variation in capacity among individuals.41 These differences are typically measured according to success at digit span, complex span or operation span tasks, which often require the subject to recall numbers or letters while attempting to complete other tasks. For example, reading span tasks require individuals to read several sentences and to recall the last word of each sentence in the correct or reverse order.42 Despite their reliance on simple tasks, measures of WM are highly reliable and valid, with test–retest reliabilities of > 0.7 and high predictive validity against multiple outcomes (see Conway et al.42 for a review of psychometric qualities). Working memory capacity is highly predictive of performance on cognitive tasks such as reading comprehension,43 mathematics,44 multi-tasking45 and following directions during learning46 (see Unsworth and Engle for a review47). Moreover, although WM is tested as a separate factor in most intelligence tests, it has been shown to correlate independently and strongly with general intelligence48 and measures of creative problem solving (i.e. fluid intelligence).49
A number of studies point to the highly heritable and stable nature of WM capacity and suggest that it is largely genetically determined and thus that innate differences exist among individuals.50-52 Given the practical importance of WM in learning and performance in multiple areas, it is an obvious candidate for investigation in an exploration of individual differences in expert performance.
Methods
Studies containing the keywords ‘deliberate practice’, ‘working memory’, ‘individual cognitive differences’ and ‘expert performance’ were sought through PsycINFO, Google Scholar, and PubMed/MEDLINE. The abstracts of studies were reviewed by one author (KMK) to determine whether these studies addressed the question of DP versus individual differences in predicting performance using empirical methods. The database search was augmented by a review of the citations and bibliography of relevant papers to find other articles of interest. The results of the studies were summarised and discussed by two authors (GRN and KMK, or LEMG and KMK) to determine a final cohort of studies for inclusion. The criteria for selection required the study in question to explicitly test hypotheses related to the relative influence of WM or WM-related measures and accumulated DP in predicting performance.
The results of the articles retrieved are summarised according to the roles ascribed to DP, WM and the relative contributions of each in relation to expert performance. We also reconcile the results of the studies to explain differences in findings.
Results
Nine studies specifically tested the impact of DP or its surrogate measures and WM in predicting performance. These studies came primarily from the expertise and learning literature in psychology.
Deliberate practice and working memory
The two main models relating to DP and WM argue either that WM becomes less important as DP accumulates (the circumvention of limits model) or that WM remains a significant independent predictor of expertise (the independence model).
The circumvention of limits hypothesis argues that the effect of individual innate differences such as in WM are important predictors of performance at the novice stage,1, 2, 20, 53 but that these innate limits are overcome with the acquisition of extensive DP. The idea is that DP provides experts with cognitive strategies, such as shortcuts, heuristics and problem-solving schemas, that allow them to perform without excessively taxing limited WM capacity. Accordingly, the impact of innate individual differences in explaining differences among experts diminishes as hours of DP increase.54
A number of studies have tested the circumvention of limits hypothesis explicitly by comparing DP against individual differences and have shown some evidence for the sufficiency of DP to the exclusion or minimisation of cognitive ability or WM. Ruthsatz et al.55 noted a correlation between playing ability and general intelligence in novice band musicians, but not in more practised individuals. Similarly, Ericsson et al. found that differences in cognitive abilities did not relate to performance by expert violinists, although levels of DP did.1 Studies in chess showed that deliberate study by chess players was the critical factor for expertise development.25, 29 The ability to read ahead during copytyping by expert typists was related to levels of DP and not to surrogates of WM capacity.56 Specific evidence that DP is the most significant predictor of expert performance (and occasionally novice performance) to the exclusion of other innate abilities has also been shown in software programming57 and in awareness tasks in aviation.58 The mechanisms by which DP allows for the circumvention of WM capacity and other abilities are still debated. Ericsson and Lehmann20 suggest that bypassing WM limitations requires that experts reorganise their storage of concepts. One possible mechanism of reorganisation is the development of a ‘long-term’ WM,59 in which experts can create domain-specific memories that are retrieved as required and bypass WM. Whether this mechanism or some alternative organisation of memory is used is still under debate.60-62
By contrast, the independence model posits no interaction between WM and DP. In this view, the effects of WM and DP are additive in contributing to performance at both low and high levels of DP (i.e. the same for novices and experts). Whereas the circumvention of limits hypothesis proposes to remove individual differences altogether as a factor in explaining expert performance, the independence model proposes that WM capacity will remain a significant predictor of expert performance after DP has been accounted for.63
Independent effects of WM have been found in a number of areas of expertise, including sight reading by expert piano players,63 the ranking of hands by poker players,64 memory for positions and plays in baseball,65, 66 and second language acquisition.67 These studies typically controlled for self-reported surrogate measures for DP, such as domain knowledge, and tested the impact of WM capacity. For example, Meinz and Hambrick's study63 of expert piano players used procedures similar to those used in early DP studies in music to document the effects of accumulated DP as well as performance on four traditional measures of WM. They then examined players' performance in a sight-reading test (playing a piece of music without prior practice) as evaluated by two expert raters. Working memory capacity was significantly correlated with playing performance (r = 0.30) even after controlling for DP measures. Critically, DP and WM measures showed no correlations, which suggests they made independent contributions to predicting performance. Meinz et al.64 used a similar approach to study WM capacity in the context of poker performance. Participants with extensive experience of poker were evaluated for WM capacity, knowledge of poker rules, memory for game play, and the ability to evaluate card combinations to determine the probability of success. Working memory capacity was significantly associated with memory for game play (r = 0.30) and evaluation of cards (r = 0.26), but not with knowledge of poker rules or experience. The studies65, 66 in baseball knowledge measured various factors associated with knowledge and experience of baseball game play, as well as WM capacity, that were then used to predict participants' recall of play-by-play performance. Working memory was found to significantly predict recall of game play after controlling for knowledge and experience.
Returning to expertise in chess, although Grabner et al.68 did not explicitly test DP and WM capacity interactions in their study of chess expertise, the experts in their sample had an average of 10 000 hours of DP. Several measures of intelligence, including verbal and numerical ability, two measures that load heavily on WM capacity,49 correlated positively with players' international rankings, at 0.38 and 0.45, respectively, after controlling for DP.68
Reconciling individual differences and DP
Evaluating the relative merits of these models is difficult. Certainly, there is strong evidence for both the circumvention of limits and independence models. One factor that might reconcile the divergent results is consideration of the type of task used to assess expert performance. Firstly, it is not surprising that the acquisition of expertise in motor skills such as those required to play the violin may have more to do with practice than with innate cognitive skills. Still, playing chess does not involve motor skills. Another perspective is that some kinds of expertise do amount to the acquisition of routines and automaticity, which, again, is unlikely to relate to cognitive differences. Studies supporting the circumvention of limits hypothesis assessed expert performance on tasks that were explicitly routine for participants.1, 56, 57 Conversely, support for the independence model was found in studies that assessed performance on more complex, unfamiliar tasks.63-67
It is likely that the type of task under study will affect the presence or absence of an interaction between WM and DP. We propose that expert performance in routine, static tasks does not require extensive WM. Memory structures and the experience acquired from extensive DP are probably sufficient for most expert performance tasks, especially if these tasks are similar to tasks that have been previously performed. For example, study 2 in Ericsson et al.'s seminal paper1 was the first to examine prospectively the relationship between skilled performance and accumulated DP. The skilled performance task was described as:
‘…required subjects to give three successive performances of the Prelude No. 1 in C-major (“Wohltemperiertes Klavier”) by J S Bach. Subjects were given up to 15 min of practice to work out an interpretation and then attempted to replicate this interpretation three times while being as consistent as possible across performances. Musicians consider the piece to be technically very simple; at the same time it leaves room for musical interpretation.'1
The authors examined the correlation across performances in both amateurs and experts, finding that experts were more consistent.1 Arguably, replication and consistency are the hallmarks of routine expert performance in music. An expert musician need only recall the necessary patterns from long-term memory and apply them to resolve a routine task: repetitive performance of a simple piece. Investigations of DP in other domains similarly measured skill according to expert performance of routine tasks.55-58 The defining feature of expertise is that complex and challenging tasks for novices are made routine by extensive practice.
However, for more unfamiliar or more complex tasks that depart from routine problem solving, previous experience will fall short. Meinz and Hambrick's study63 of expert piano players used a sight-reading task that required musicians to play without rehearsal. Critically, the pieces were chosen for their range of difficulty and unfamiliarity to the participants.63 Although DP was a significant predictor of playing ability, WM capacity remained a consistently significant predictor for musicians with both low and high DP. Similarly, Hambrick and Engle65 found that although WM capacity was a significant predictor of baseball play recall for both participants with high and those with low levels of knowledge, the degree of prediction was greater for information not routinely used in game play. Thus, experts are likely to rely on all of their cognitive resources, including WM, in addition to strategies developed by DP to complete these unusual or non-routine tasks. Although DP will still be relevant across many task types, it is presumptive to conclude that the role of cognitive ability is similarly uniform across tasks, especially those as divergent as typing and chess, or simple repetition versus sight reading and the playing of previously unseen pieces of music. Table 1 presents summaries of the studies reviewed here, including information on the task type used to study the interaction between DP and WM.
| Study | Task | Findings on DP and individual differences | Model supported |
|---|---|---|---|
| Keith & Ericsson 56 | Typing speed, accuracy | DP predicted typing ability better than cognitive and motor skills measures | Circumvention of limits |
| Elliot Sim et al. 57 | Time and number of errors during project handover and software modification | Years of experience, but not general problem-solving skills, predicted performance | Circumvention of limits |
| Sohn & Doane 58 | Awareness and recall of instrument values in aviation simulation | WM capacity was a predictor in novices, but not in experts | Circumvention of limits |
| Ericsson et al. 1 | Repeated performance of pieces of music | DP, but not WM, was a significant predictor | Circumvention of limits |
| Meinz et al. 64 | Evaluation of winning-hand probability in poker | WM capacity was a significant predictor in both high- and low-domain knowledge players | Independence model |
| Meinz & Hambrick 63 | Sight reading of pieces of music | WM capacity was a significant predictor in both high- and low-DP individuals | Independence model |
| Payne et al. 67 | Proficiency in second-language comprehension | WM capacity was predictive of proficiency in both high- and low-experience speakers | Independence model |
| Hambrick & Oswald 66 | Transfer of baseball knowledge | WM capacity predicted performance on baseball play, information recall and transfer task; prediction was higher for the transfer task | Independence model; interaction with task type |
| Hambrick & Engle 65 | Recall of play-by-play information during baseball games | WM capacity was a significant predictor in both high- and low-knowledge participants. Knowledge less predictive of performance for non-routine information | Independence model; interaction with task type |
- DP, deliberate practice; WM, working memory.
One theoretical explanation for the task-dependent interaction between DP and WM is the dual-process theory of cognition, which claims that most cognitive processes (reasoning, decision making, perception, etc.) can occur through either of two complementary but qualitatively different modes of processing: System 1 and System 2.69, 70 The System 1 mode is rapid, non-analytic, unconscious, influenced by exemplars, and driven largely by pattern recognition. It is most apparent in rapid visual perception,71 but is also involved in more complex cognitive tasks such as medical diagnosis, when these tasks are relatively routine.72 The System 2 mode is slow, analytic, conscious, rule-based and characterised by the careful processing of information69 (see Evans for a review73). Although these systems are complementary, the application of both modes in particular tasks73 and throughout different levels of expertise is well established. As one becomes an expert, one relies more on System 1 and less on System 2 processing.74-76 Critically, the System 1 and 2 modes do not load on WM and other cognitive abilities equally.69, 70 System 1 processing is independent of WM capacity. By contrast, System 2 processing is related to WM capacity and correlated with measures of intelligence.77-79 Dual-processing models are increasingly supported by physiologic evidence that suggests System 2 processes are more energy-intensive80 and that they involve the activation of specific sections of the prefrontal cortex.81
The task-dependent usefulness of WM capacity for predicting performance is likely to be a function of the type of processing (System 1 versus System 2) used by experts to execute the task. For routine tasks, experts will use System 1 processing as they can rely on their extensive stores of experiential knowledge derived from practice. For unfamiliar and complex tasks, experts will shift to System 2 processing and draw on effortful WM-dependent processes.
There is already considerable evidence of this shift in expert clinical reasoning. Several studies have shown that medical experts rely primarily on System 1 processing during reasoning.72, 75, 82-85 However, when presented with ambiguous diagnostic problems84 and at critical moments during practice, clinicians switch to effortful reasoning with all the hallmarks of System 2 processing.86 Therefore, it is likely that WM may become a crucial factor in some parts of clinical practice. Although some tasks do become routine (i.e. physical examinations, management of common diseases), experts are also regularly required to perform clinical tasks that are functionally more difficult, as well as to resolve unusual or atypical patient presentations. For example, experienced surgeons report shifting from a speedy, automatic mode of operation to a slow and deliberately effortful mode in response to situational cues such as unexpected abnormalities.87 In these circumstances, surgeons display highly controlled and effortful execution of surgical skills that they otherwise describe as being fairly routine. Similar demands may be made in other specialties in which doctors may face difficult, unusual or novel problems. For these non-routine challenges, reliance on the hours of practice or experience accumulated may be insufficient to explain how clinicians resolve these problems. Awareness of WM capacity or other individual differences may be important in elucidating how experts function in these settings.
Discussion
Our review shows that expert performance in multiple domains may be affected by a number of factors, including a range of innate and environmental individual differences, as well as contextual factors. Reducing expert performance to only DP is a bold but unfounded proposition. Furthermore, although WM capacity is a significant and predictive cognitive limitation, we do not propose that WM is the only important individual difference or factor with which educators should concern themselves. Rather, our review of this literature suggests that some individual differences for particular tasks or contexts can play a role in either the acquisition of expertise or expert performance. Determining these individual differences, as well as the roles they play in novices and experts, will have implications for the structure of learning and practice.
Implications
The broadest implication of this review for medical education research and practice is that it calls for a reorientation to wider consideration of the individual differences that impact learning and performance. Individual differences are explicitly acknowledged as playing a significant role in the discourses of selection to medical training. Measures related to WM and intelligence,88 such as academic performance, aptitude testing and verbal reasoning, as well as other more general personal qualities, are assessed.89 Further assessments that measure individual differences continue to occur after selection and include licensing examinations of knowledge and skills, postgraduate selection into residency, the selection of chief residents or fellows, and staff selection. However, the formal importance attached to these measured individual differences diminishes as students progress into experts. Instead, individual differences in skill, ability and motivation have minimal impact in formal discussions around clinical expertise.
Nonetheless, there is sufficient evidence to suggest that overall individual differences in initial skill persist and remain meaningful after a long period of practice. A large study of hospital deliveries examined the complication rates associated with 1800 obstetricians over 18 years and over two million caesarean and vaginal deliveries.90 The authors found that an obstetrician's complication rate at graduation from residency was predictive of his or her complication rate after nearly two decades in practice after controlling for patient and environmental factors. Moreover, measures of initial skill explained more of the variation between doctors than did volume of practice, even in a subsample of doctors with over 15 years of experience each. This research does not specify the exact innate individual differences that affect performance. However, it does highlight individual variability that has meaningful impact on patient outcomes several years into practice and is not related to the amount of experience as measured by volume of practice or length of experience. Clearly, identifying the source of this individual ability – be it intelligence, WM, motivation or personality – is very important.
Some research has already begun to address medical education from the perspective of individual differences. For example, several studies have addressed the role of visual–spatial ability in surgical expertise91, 92 and the acquisition of surgical skills.93-95 These studies show that visual–spatial ability is a significant predictor of novice performance and rate of skill acquisition,96, 97 albeit that visual ability becomes less predictive94 for some tasks as expertise is acquired. This suggests that the baseline measurement of visual–spatial ability can enable education to tailor or focus training on areas in which students will have the greatest difficulty. This also requires some adjustment of timeline-related expectations as students with different abilities will take longer to achieve mastery.98
Furthermore, the task- or situation-dependent utility of individual abilities such as WM may be more frequent than imagined, given that medical practice can be highly varied and challenging. Given the evidence for processing shifts in clinical reasoning,86 future studies might integrate individual abilities such as WM capacity to clarify their various roles in predicting performance in specific problem-solving conditions.
This attention to individual differences in no way diminishes the role of DP. There are growing calls for the extensive incorporation of DP opportunities into medical curricula,99, 14 particularly in the areas of simulation-based education3 and improved feedback mechanisms. This emphasis is entirely warranted: DP is a crucial factor in developing expertise. However, it is unfeasible for any training programme to expose trainees to sufficient DP in all the possible presentations of a diagnosis or contexts in which a clinical skill must be performed. Furthermore, evidence that individual differences affect the rate at which students learn and how experts perform is compelling. Acknowledging this during training is especially important and lends support to the individualisation of training programmes. This shift in pedagogy will mean recognising that individuals will learn at different rates and that expectations on progress should be tied to realistic, grounded outcomes. The current movement in medical curricula to competency-based mastery models is thus encouraging as extensive DP and individual flexibility can be easily introduced into these programme models. Beyond education, the critical message is that we should not stop tracking individual variation upon certification or by some pre-set marker for sufficient practice. Experts will face unusual, atypical and non-routine challenges in the practice of medicine. How different individuals respond to these challenges and the factors that predict success or failure should be studied. Given that individual variability is real and has significant impact on patient-level performance, we should begin to pay attention to the factors that lead to meaningful differences.
Conclusions
Our review shows that emphasising DP to the exclusion of other factors in the study, development and education of experts is far too simplistic. There is sufficient evidence to suggest that individual differences, specifically in WM capacity, can play a role in the development of expert performance of non-routine or unfamiliar tasks. Although accumulated DP is predictive of performance, expertise is multifaceted and depends on a host of other factors, including individual ability. Medical education practice and research can benefit from uncovering individual differences that impact the acquisition of skill in novices and the development of expert skilled performance.
Contributors
all authors contributed to the conceptualising of this review. KMK conducted the review of the literature, wrote the first draft of this paper and edited subsequent drafts. All authors contributed to the critical revisions of this paper and approved the final manuscript for publication.
Acknowledgements
we would like to thank Dr Nicole Woods, The Wilson Centre, University of Toronto, Toronto, ON, Canada, for her comments and feedback on this paper.
Funding
none.
Conflicts of interest
none.
Ethical approval
not applicable.
