A systematic review on improving cognition in schizophrenia: which is the more commonly used type of training, practice or strategy learning?

Background The purpose of this article was to conduct a review of the types of training offered to people with schizophrenia in order to help them develop strategies to cope with or compensate for neurocognitive or sociocognitive deficits. Methods We conducted a search of the literature using keywords such as “schizophrenia”, “training”, and “cognition” with the most popular databases of peer-reviewed journals. Results We reviewed 99 controlled studies in total (though nine did not have a control condition). We found that drill and practice training is used more often to retrain neurocognitive deficits while drill and strategy training is used more frequently in the context of sociocognitive remediation. Conclusions Hypotheses are suggested to better understand those results and future research is recommended to compare drill and strategy with drill and practice training for both social and neurocognitive deficits in schizophrenia.


Background
About 80% of individuals with a diagnosis of schizophrenia struggle with a variety of neurocognitive and sociocognitive deficits [1,2]. The neurocognitive domains typically affected include speed of processing, attention/vigilance, working memory, verbal learning, reasoning and problem solving [3,4], whereas social cue perception, affect recognition, attribution, and theory of mind are the sociocognitive domains most affected [5,6]. Cognitive dysfunctions are considered to be core features of schizophrenia, since they are strongly correlated with poor functional outcome [7][8][9] as well as being better predictors of general outcome and rehabilitation than positive symptoms [10,11]. Although pharmacological and psychological treatments can effectively reduce [12] positive symptoms of schizophrenia, they do little to improve cognition [7]. Thus, using cognitive retraining or remediation to create significant improvements has received more attention in recent years [7,13]. According to T Wykes, V Huddy, C Cellard, SR McGurk and P Czobor [14], there are two types of training: 1) "drill and practice," where there is no explicit component, meaning that learning is based on repeating a task that becomes gradually more difficult and where participants implicitly learn the strategy by trial and error, and 2) "drill and strategy," where the focus is to teach the explicit use of a determined strategy (see also [12]). While explicit learning impairments have been consistently reported in schizophrenia literature [15,16], there is still a debate over impairments to implicit learning. For example, some studies report that implicit learning is intact for tasks such as probabilistic classification learning (e.g., [17]), weather prediction (e.g., [18]), and artificial grammar learning (e.g., [19]), while others report an impairment in colour pattern learning but not in letter string learning [20]. Adding to this conundrum are a variety of different training procedures currently being tested, both for drill and strategy (includes explicit and implicit learning) and for drill and practice (implicit learning only). These training procedures focus on a variety of different targets therefore, in this review, we will focus on neurocognitive and sociocognitive domains. For this reason we will not include studies aiming solely to reduce positive or negative symptoms or to improve upon social skills. Contrary to the recently published meta-analyses focusing on efficacy of cognitive training [14,21], this review will analyze and describe which training paradigms were most used to improve neurocognitive and sociocognitive deficits, whether they be drill and practice or drill and strategy methods.

Review protocol
Inclusion criteria: 1) outcome: either neurocognition or sociocognition, 2) date and journal: peer-reviewed journals from 1995 up to 2013, 3) language: English or French, 4) diagnosis: majority (≥70%) of participants with a schizophrenia diagnosis (others include schizoaffective disorders and first-episode psychosis). We excluded all training types that aimed solely to reduce positive or negative symptoms, improve social skills, increase metacognition, etc. Nevertheless, studies that targeted sociocognition or neurocognition while also aiming to reduce symptoms or improve social skills as secondary objective, were included. Finally, we removed studies that used the training or remediation for evaluation rather than for treatment (i.e., studies assessing the deficits at baseline with no intention of remediation or intervention) as well as meta-analyses and reviews. Our goal was to review studies that had a therapeutic outcome. Since the main objective of our article is to provide a descriptive listing of the training offered and not to conduct an efficacy analysis, we included studies that did not have control conditions. Given the large number of articles included (n = 99), and the fact that our definitions of the types of training were inclusive, the first three authors read, classified, and compared their ratings for each article to ensure reliability of the results. Articles were classified in two categories, according to the targeted deficits: i) Sociocognitive, which included topics such as emotional recognition, Theory of Mind, attributional style, and social cue recognition; ii) Neurocognitive, which included areas such as executive functioning, memory and attention. Importantly, social functioning was excluded from the dichotomy of classification as most, if not all studies, ultimately aim to improve upon work and functional outcomes of individuals. Furthermore, we compared the results of our literature search with articles listed in the meta-analyses of T Wykes, V Huddy, C Cellard, SR McGurk and P Czobor [14], O Grynszpan, S Perbal, A Pelissolo, P Fossati, R Jouvent, S Dubal and F Perez-Diaz [22] and A Medalia and AM Saperstein [23] to ensure that we did not miss any relevant articles.

Article retrieval
We conducted a literature review using the following databases: PsychINFO (1995 to May 2013), MEDLINE (R) (1995 to May 2013) and MEDLINE Daily Update (R). Using the title keywords "schizophrenia and (training or remediation or intervention or practice) and (soci* a or neuro* or cogniti* or metacogniti* or problem-solving or visual or memory)" , we obtained 465 results from all databases. To ensure further precision we added the following filters: a) "limit to English and French language" (to ensure understanding of the content) which yielded 172 results, b) "limit to peer-reviewed journals" resulting in 164 results. The final manipulation was to remove all duplicates, which left us with a total of 121 articles to investigate. Upon final removal of all articles that did not meet our criteria, we reviewed 99 articles. The last date of search for articles was January 2014.

Results
Results are presented in Tables 1, 2 and 3, divided according to the aim of the studies: improving neurocognitive deficits, sociocognitive deficits or both. These were further subdivided by either drill and practice or drill and strategy training methods. First, we will describe the studies that focus on a single area of cognition (i.e., Table 1 for neurocognition and Table 2 for sociocognition) as treatment targets and that used a single training type (drill and practice or drill and strategy). Then, we will describe the results of studies with multiple aims in terms of neurocognitive and sociocognitive deficits (Table 3). There is an important distinction to be made between the targeted deficitswhich is how we classified the studies between neurocognition, sociocognition, or bothand the measured variables. Indeed, it is often the case that a variable is measured to assess the impact of the training without having been specifically targeted by the training, which, therefore, gives a sense of the generalization of the results. As seen more explicitly in Table 2, many of the studies aiming to improve sociocognition also measure the impact of the training on more neurocognitive variables.

Neurocognitive deficits
We identified a total of 62 studies pertaining to neurocognitive training. Of these, 58 included randomized controlled trials or placebo conditions, while four had no control. At first glance (see Table 1), it appears that for people with schizophrenia drill and practice training is used more frequently to train neurocognitive deficits (i.e., drill and practice = 35 studies, 33 with controls and two without; drill and strategy = 27 studies, 25 with controls and two without).
Examining the drill and strategy studies, a pattern rapidly emerges when the methods of training are considered.        Twelve of 27 studies used group therapy in their training rather than individual computerized training with therapist assistance. However, there does not seem to be a link between the method of training (individual or group) and the outcome measures. Though it is not the goal of our review, it is important to note that all articles with drill and strategy approaches to training reported between-group improvements of the targeted deficits. Furthermore, eight of the 17 studies with follow up measures at either three, four or six months also reported sustained gains in cognition [25,32,37,41,45,47,48,50,64]. Drill and practice studies most commonly used computerized tasks, done individually. However, there was more variety in the methods of training, for example, at least five studies used pencil-and-paper procedures [60,67,69,73,75]; though Lopez-Luengo utilized both pen-and-paper and audio] while five others used a combination of audio and visual tasks [62,63,77,78,83] to reduce the deficits. Furthermore, most studies using drill and practice methodologies (all except [61,69]) reported between-group improvements in cognition between the experimental and control groups, at least for some measures.
The studies we analyzed targeted a variety of neurocognitive deficits -memory, attention/vigilance, reasoning, verbal learning -yet overall, across studies, no single deficit stood out as being resistant to implicit training. Therefore, it would seem that most domains of neurocognition respond well to drill and practice training, even though only seven studies had follow ups at six months, six [52,55,60,63,64,82] confirming that the gains were maintained and one [65] showing that only the affect recognition benefits were not maintained at the 1-year follow up.

Sociocognitive deficits
In contrast to studies focusing on neurocognition, those aiming to improve sociocognitive deficits used mostly drill and strategy approaches (i.e., drill and practice = two studies with control groups; drill and strategy = 21 studies, 18 with controls and three without). Importantly, all studies included a variety of visual aids such as vignettes, Powerpoint presentations or videos of social situations. Furthermore, visual presentations and explanations by the therapist about the goal of the training were often done in group settings. This method allows modelling by the therapist but also incorporates group exercises and practice as well as role-plays.
Interestingly, for sociocognition, whether the training paradigm was drill and strategy (e.g. [97]) or drill and 210 practice (e.g. [107]), there was a general concern to assess whether remediation of a specific type of deficit would generate generalizable results, not only to functional outcomes but also to broader domains of social cognition such as Theory of Mind.

Studies that aimed to improve both neuro and sociocognition
It is more difficult to find a pattern in the types of training when the target deficits are broader and span across both neurocognitive (such as memory and attention) and sociocognitive domains (such as social perception and emotion recognition). However, most use drill and strategy paradigms that generally combine computerassisted programs for neurocognition, and guided practice, modeling and role-play for sociocognition. There is also a mix of individualized and group approaches that seem, again, to follow the trend that neurocognition is trained individually while sociocognition is trained in groups, and   this is true for both drill and practice as well as drill and strategy.

Discussion
The purpose of this article was to review the type of trainingwhether drill and practice or drill and strategymost often offered in clinical studies to people with schizophrenia to help overcome neurocognitive or sociocognitive deficits. We included articles with varying scientific value for both neurocognitive and sociocognitive training; nine of the 99 articles we reviewed had no control condition. However, since we are not presenting a thorough analysis of the efficacy or effectiveness of these training methods (see 14 for details, [22,23]), we opted to include them for descriptive purposes. Although we found a variety of training modalities offered, some more behavioral, some using computer training, real-life situations, indirect training, etc., we were able to determine if a training paradigm was drill and practice or drill and strategy in nature, and which of these methods was used more frequently to improve neurocognitive or sociocognitive deficits. We also planned to describe the patterns and modalities used to train the targeted deficits (i.e., neuroor sociocognitive). In our literature search, we found that drill and practice training programs were used more frequently for improving neurocognitive deficits. Of the 62 studies we reviewed, 35 used procedures that mostly involved errorless learning, a type of training where the degree of difficulty of the task increases with the performance of the participant and where no conscious effort is necessary to improve. Studies using drill and strategy (n = 27) seemed particularly interested in the impact of the training on other variables outside of neurocognition, such as symptoms and quality of life. This was not the case for the drill and practice approaches. Another difference was that studies using drill and strategy training almost always measured executive functioning (n = 15), whereas studies using drill and practice training did not. However, we could not determine whether one specific domain of neurocognition was more easily retrained than another with drill and practice vs. drill and strategy procedures. Furthermore, most studies were of short duration and only a few had follow up measures (e.g., drill and strategy n = 8 [25,32,37,41,45,[47][48][49]; drill and practice n = 7 [52,55,60,[63][64][65]82]. This could be improved upon in future studies, since it is difficult under these circumstances to decide whether the observed effects are maintained over time or not. When attempting to put the findings on neurocognitive deficits into context, we wondered why drill and practice training would be used more often to retrain neurocognitive deficits. The answer may lie in the way these functions interact in our cognitive processes. Some domains, like attention and speed of information processing, seem more implicit by naturethe bottom-up approach. We could posit that these functions are not used consciously and a person would not need to inherently know "how" to use the functions; instead they would simply perform the task repetitively and unconsciously. However, this might imply that drill and practice procedures would only improve neurocognitive deficits, which might not be the case, as judged by the results reported in recent meta-analyses [14,22]. Furthermore, since implicit learning has been reported as being generally intact in schizophrenia [84], some, like Fisher and colleagues [62], suggest that high levels of repetition (e.g., more than 1,000 rehearsals) and a high percentage of reward schedule (e.g.: 85%), will allow for neurological improvements. Yet, studies using drill and strategy procedures in their training methods also seem to generate consistent positive outcomesthe top-down approach. Of note, Wykes and colleagues [14] suggested that drill and strategy training include elements that are explicitly learned (through modeling, explanation or role-playthe "strategy") but also elements inevitably linked with repetition (the "drill") and considered implicit learning, which might explain why they are effective.
Tentatively, we suggest that since drill and strategy learning is thought to allow better integration of the rules and, thus, greater association between the various training elements [122], changes in cognition tend to occur over time. Blairy and colleagues [25], who also reported long-lasting improvements on memory and executive functions after explicit training, hypothesized that participants learned to bind different aspects of the Note. CRT = cognitive remediation training, CBT = cognitive behavioral therapy, TAU = treatment-as-usual, MATRICS = Measurement and Treatment Research to Improve Cognition in Schizophrenia. experiment together and that it allowed for better consolidation in memory. Thus, at this time, we cannot draw a conclusion about whether certain domains of neurocognition respond better to one type of training over another. Further studies must be conducted, preferably comparing different forms of training with each other and adding follow up measures to assess whether the benefits of training remain stable through time.
Social cognition is considered by many researchers to have a strong relationship with positive functional outcomes [123,124]. Concurrently, the meta-analysis by McGurk [12] reported that programs using strategy coaching (drill and strategy training) for sociocognitive deficits had strong effects on functional outcomes as well as on the targeted social cognition skills. Consistent with this, we found that drill and strategy training was more frequently used for sociocognitive retraining. It seems intuitive that learning and integrating a social skill requires that it be practiced in a social setting, which was consistent with our findings when analyzing the studies. Most used group settings, where participants received their training then performed and practiced the learned techniques with a therapist to correct the behavior and give feedback. Moreover, it was also reported that integrating rehearsals into the training yields greater functional outcome improvements [23]. Indeed, sociocognitive studies tend to measure social functioning or social adjustment following training more often than studies aiming to improve upon neurocognitive deficits. Yet, a growing field around implicit learning in social cognitive psychology [125] suggests that drill and practice or other forms of more implicit training might be useful for sociocognition as well.
The collection of studies of Bell and colleagues on work and social outcomes using drill and practice [53,55] hint at the importance of generalizing the benefits of training to real-life situations, such as the ability to find and maintain work or to increase work productivity in the form of hours and money earned. However, both of these studies integrated the drill and strategy approach with a program of supported employment, creating a hybrid retraining program which has been efficient in the past [14]. Indeed, while improving cognitive deficits is commendable, functional outcomes are issues that should not be dismissed when considering the difficulties faced by individuals suffering from schizophrenia when trying to reintegrate the work force or create a social network.
We have also discovered that training programs usually target cognitive improvements "at large", rather than specifically focusing on the individual deficits highlighted by the person's profile, most likely to allow more people to receive the training without the need for specific neuropsychological or sociocognitive evaluations. We suggest choosing one type of training over another depending on the overall goal one is trying to achieve: drill and practice for precise deficits and drill and strategy to obtain general gains. More studies are needed to determine if drill and practice could be useful for sociocognition as well.
Furthermore, specific training methodologies seem to benefit specific domains of social cognition. For example, though it appears that Social Cognition and Interaction Training (even when including the family in the training sessions) improves Theory of Mind (ToM), group practices and Powerpoint presentations detailing the concepts of ToM did not improve ToM but did improve emotion recognition. We suggest that ToM is a more complex construct of sociocognition and requires more precise and detailed training than emotion recognition. Horan and colleagues [93] suggest that even defining the different concepts contained within ToM, such as appreciation of humour, is difficult and the training for it is more challenging. Furthermore, a recent meta-analysis of social cognition training in schizophrenia [21] also reported inconsistent effect sizes when ToM is targeted, suggesting that the key elements needed in the training for ToM must be better identified.
When the objectives of the training are broader, meaning that they aim to improve both neurocognitive and sociocognitive deficits through drill and strategy, the variables measured are also more varied and often include certain measures of functional or occupational outcome. Furthermore, these studies often tend to combine training with other types of intervention such as cognitive-behavior therapy, supportive therapy or occupational therapy.
Overall, our review summarizes the current state of research into cognitive training in schizophrenia. In neurocognition, drill and practice training is used more frequently and with a variety of different procedures such as auditory training [62] or target discrimination [74]. Tailoring the training to specifically address precise deficits might be one of the key benefits of drill and practice training. However, from the studies we evaluated, drill and strategy training was more easily generalized to all neurocognitive deficits. Indeed, a recent meta-analysis on the benefits of cognitive remediation in schizophrenia noted that this modality of training produces stable benefits on global cognition [14]. We suggest choosing one type of training over another depending on the overall goal one is trying to achieve: drill and practice for precise deficits and drill and strategy to obtain general gains in neurocognition.

Limitations
There are a few limitations to our review. First, to reflect current trends, we included only studies published between 1995 and 2013, although interest in cognition remediation started as early as the end of the 1970's [126].
Second, the fact that drill and practice or drill and strategy training can involve multiple strategies and training techniques (e.g., times eye tracking, computer programs, paper-pencil tasks, errorless learning, group learning, and various modalities of feedback) prevented us from describing them in detail and some of these specific strategies might explain differences in outcomes. Our goal was to describe what was being offered, not to promote one approach in particular. We also did not include studies described as "metacognitive", a term that involves cognitive biases, at times social and/or neurocognitive, that are linked to the symptoms of psychosis [127] for example, focusing on the cognitive bias of jumping to conclusions as linked to delusions. It is important to note that these types of training are not the only modalities offered to help overcome neurocognitive or sociocognitive deficits. Occupational therapy [128], social skills training [129], as well as certain forms of metacognitive psychotherapies [130] have also been documented.

Conclusion
Future research is warranted to compare both drill and strategy and drill and practice programs with one another under control and experimental conditions, as well as to highlight the benefits and limitations of each. This would help to identify which type of deficit would benefit more from which training or to isolate particular participant profiles that respond best to a specific training strategy. Moreover, we suggest that more focus be brought to targeting participants' specific deficits to tailor the training to those needs. This would increase the potential impact and generalization to "real-life" situations, both in the context of neuro and sociocognitive retraining. Finally, we propose investigating the benefits of both neurocognitive and sociocognitive training in the context of comorbidity. It is well know that schizophrenia is often comorbid with social anxiety (in 30% of cases; [131]) and substance abuse (in 50% of cases; [132]), to name a few. It is conceivable that the interplay of those disorders could be a substantial challenge for training. Nevertheless, very few studies have examined the impact of these presentations and doing so would be of paramount importance as it could increase the ecological validity and generalizability of the results.
Endnote a *stands for truncation.