Reading skills deficits in people with mental illness: A systematic review and meta-analysis

Abstract

Background

Good reading skills are important for appropriate functioning in everyday life, scholastic performance, and acquiring a higher socioeconomic status. We conducted the first systematic review and meta-analysis to quantify possible deficits in specific reading skills in people with a variety of mental illnesses, including personality disorders (PDs).

Methods

We performed a systematic search of multiple databases from inception until February 2020 and conducted random-effects meta-analyses.

Results

The search yielded 34 studies with standardized assessments of reading skills in people with one or more mental illnesses. Of these, 19 studies provided data for the meta-analysis. Most studies (k = 27; meta-analysis, k = 17) were in people with schizophrenia and revealed large deficits in phonological processing (Hedge’s g = −0.88, p < 0.00001), comprehension (Hedge’s g = −0.96, p < 0.00001) and reading rate (Hedge’s g = −1.22, p = 0.002), relative to healthy controls; the single-word reading was less affected (Hedge’s g = −0.70, p < 0.00001). A few studies in affective disorders and nonforensic PDs suggested weaker deficits (for all, Hedge’s g < −0.60). In forensic populations with PDs, there was evidence of marked phonological processing (Hedge’s g = −0.85, p < 0.0001) and comprehension deficits (Hedge’s g = −0.95, p = 0.0003).

Conclusions

People with schizophrenia, and possibly forensic PD populations, demonstrate a range of reading skills deficits. Future studies are needed to establish how these deficits directly compare to those seen in developmental or acquired dyslexia and to explore the potential of dyslexia interventions to improve reading skills in these populations.

Introduction

Reading is a complex process that requires the implementation of various skills simultaneously. To begin with, it requires recognition of the visual information necessary to extract the information from text [1]. The core reading skill is phonological processing, which involves recognition of the sound structure of the language, the decoding of written symbols into sounds (phonological awareness), and then their maintenance in working memory (phonological memory) [2]. Phonological processing facilitates the decoding of written information, which leads to word identification and subsequent extraction of meaning [3]. A failure to read each word correctly leads to problems with comprehension [4] as comprehension involves the processing of individual letters and words, and then putting them together to form meaning [5]. When one or more of these reading skills are impaired, and this impairment cannot be explained by general cognitive dysfunction or intelligence, this is referred to as dyslexia [6]. Overlaps between dyslexia and schizophrenia (SZ) have been suggested, based on previous findings of disruption in the processes that support skilled reading (e.g., deficits in language, auditory and visual perception, oculomotor control) in both disorders [7] but the nature and severity of reading skills deficits in SZ and other severe mental illnesses (MIs) remain unclear at present.

Reading skills are of enormous significance for a range of socioeconomic outcomes in modern societies, including academic performance, occupational achievement, and family and social relationships [8,9]. Furthermore, poor reading skills in children have been associated with increased antisocial behavior [10,11]. Likewise, in forensic populations, poor reading skills and dyslexia traits have been associated with increased anxiety and poor socialization, which, in turn, might explain their antisocial behavior [12,13]. In people with various MIs, undiagnosed reading problems, and dyslexia result in scholastic failure, in turn raising the risk for mood problems [14] and future criminal behavior [15]. Poor reading skills also pose a challenge for accessibility of mental health interventions [16] and predict poor psychosocial outcomes [17,18]. There is thus a need to consider reading deficits as a therapeutic target and address them, for example, with interventions used for dyslexia [7,19]. A thorough understanding of the pattern and magnitude of reading deficits in people with specific MIs is an important first step toward this goal.

The main aim of this systematic and meta-analytic review was to conduct a comprehensive analysis to delineate the nature and magnitude of reading impairments based on data from studies that employed standardized tools to assess reading skills in people with SZ, bipolar disorder, affective disorders (major depression, anxiety, mania), personality disorders (PDs; borderline personality disorder [BPD], antisocial personality disorder [ASPD], psychopathy), and general MIs (across diagnoses/not-specified). Our secondary aims were to examine whether (a) particular reading skill deficits were more strongly present when assessed with some tests compared to others, given that reading skills in different studies have been quantified using a variety of tests and batteries, and (b) groups with MIs and a forensic history show more pronounced deficits relative to those from nonforensic settings.

Methods

This systematic literature review and meta-analysis followed PRISMA guidelines [20]. Search terms and key articles were identified based on an exploratory search of databases and an internet search engine (Google Scholar). We then searched Academic Search Complete, CINAHL Plus, PsycINFO, PsycARTICLES, SocINDEX, MEDLINE via EBSCO Host and PubMed (up to Feb 2020) for all studies including reading assessment(s) in MIs (see Table 1 for the full search strategy and eligibility criteria). Manual searches were conducted using the relevant literature [7,17,21].

Table 1. Full search strategy per database and eligibility criteria.

Abbreviations: MI, Mental Illness; Intelligence Quotient, IQ

Two independent reviewers selected the studies (MV, BJ), and extracted and reviewed data for inconsistencies to reach a consensus (MV, LAW). Extracted data included tests and measures (Table 2), as well as participant characteristics, main findings, the language of assessment, and country (Table 3).

Table 2. Tests and measures used in the selected studies (k = 34) and diagnoses assessed. Studies involving forensic populations are in italics.

Abbreviations: BD, Bipolar Disorder; BDAE, Boston Diagnostic Aphasia Examination; CTOPP, Comprehensive Test of Phonological Processing (PA, Phonological Awareness; PM, Phonological Memory; RN, Rapid Naming; APA, Alternative Phonological Awareness; ARN, Alternative Rapid Naming); DD, Depressive Disorder; GORT, Gray Oral Reading Test; HC, Healthy Controls; ITBS, Iowa Test of Basic Skills; ITED, Iowa Test of Educational Development; JDT, Jacobson’s Decoding Test; LNNB, Luria-Nebraska Neuropsychological Battery; MI, Mental Illness; MST, Madison’s Spelling Test; MSVT, Madison’s Standardized Vocabulary Test; MWDT, Madison’s Word Decoding Test; NARA, Neale Analysis of Reading Ability; NDRT, Nelson–Denny Reading Test; PALPA, Psycholinguistic Assessments of Language Processing in Aphasia; PD, Personality Disorder; PIAT, Peabody Individual Achievement Test; RAN, Rapid Automatised Naming; RCBA, Reading Comprehension Battery for Aphasia; REALM, Rapid Estimate of Adult Literacy in Medicine; RNRT, Roentgen’s Nonwords Reading Test; RNST, Roeltgen’s Nonwords Spelling Test; SZ, Schizophrenia; SZAD, Schizoaffective Disorder; TOWRE, Test of Word Reading Efficiency; WJTA-III, Woodcock–Johnson III Tests of Achievement (BR, Broad Reading; BRS, Basic Reading Skills; RC, Reading Comprehension; PKG, Phoneme-Grapheme Knowledge); WRAT, Wide Range Achievement Test; WRMT-R, Woodcock Reading Mastery Test—Revised (BS, Basic Skills; PC, Passage Comprehension; PKG, Phoneme-Grapheme Knowledge).

Table 3 Summary of key data extracted from selected studies (k = 34).

Abbreviations: AdjD, Adjustment Disorder; AfD, Affective Disorder; Alc, Alcoholism; BD, Bipolar Disorder; BDAE, Boston Diagnostic Aphasia Examination; BPRS, Brief Psychiatric Rating Scale; CPZE, Chlorpromazine equivalents; CTOPP, Comprehensive Test of Phonological Processing (PA, Phonological Awareness, PM, Phonological Memory, RN, Rapid Naming, APA, Alternative Phonological Awareness, ARN, Alternative Rapid Naming); DD, Depressive Disorder; FC, Forensic Controls (history of violence without MI); GORT, Gray Oral Reading Test; HC, Healthy Controls; ITBS, Iowa Test of Basic Skills; ITED, Iowa Test of Educational Development; JDT, Jacobson’s Decoding Test; LNNB, Luria-Nebraska Neuropsychological Battery; Mac-CAT-CR, MacArthur Treatment Competence Assessment Tool for Clinical Research; MCCB, MATRICS Consensus Cognitive Battery; MI, Mental Illness; MST, Madison’s Spelling Test; MSVT, Madison’s Standardized Vocabulary Test; MWDT, Madison’s Word Decoding Test; NARA, Neale Analysis of Reading Ability; NDRT, Nelson–Denny Reading Test; NE, Not Examined; NR, Not Reported; PALPA, Psycholinguistic Assessments of Language Processing in Aphasia; PANNS, Positive and Negative Syndrome Scale; PD, Personality Disorder; PIAT, Peabody Individual Achievement Test; RAN, Rapid Automatised Naming; RCBA, Reading Comprehension Battery for Aphasia; REALM, Rapid Estimate of Adult Literacy in Medicine; RNRT, Roentgen’s Nonwords Reading Test; RNST, Roeltgen’s Nonwords Spelling Test; SZ, Schizophrenia; SZAD, Schizo-Affective Disorder; TOWRE, Test of Word Reading Efficiency; WJCog, Woodcock–Johnson Test of Cognitive Ability; WJTA-III, Woodcock–Johnson III Tests of Achievement (BR, Broad Reading, BRS, Basic Reading Skills, RC, Reading Comprehension, PKG, Phoneme-Grapheme Knowledge); WRAT, Wide Range Achievement Test; WRMT-R, Woodcock Reading Mastery Test—Revised (BS, Basic Skills, PC, Passage Comprehension, PKG, Phoneme-Grapheme Knowledge).

Studies including forensic populations are marked *Forensic*.

Studies included in the meta-analysis are marked “Meta-analysis”.

Bold entries indicates Visual aid to distinguish studies using a control group as a reference

Studies that reported means and standard deviations (s.d.) for patient and healthy control (HC) groups to permit the calculation of effect sizes were included in the meta-analysis (effect sizes also presented where only one study available). The remaining studies contributed only to the narrative synthesis (see Table 3 for details). Studies assessing individuals with conditions primarily classified as neurodevelopmental (ADHD, autism, learning difficulties, and intellectual disabilities) [82] were excluded.

Statistical analysis

The meta-analysis was conducted using Review Manager 5.3 Software—RevMan [83]. For eligible studies, effect sizes were calculated as Hedge’s g (standardized mean difference). A random-effects model was used as a more conservative approach. Heterogeneity was calculated as the I² measure of consistency for each meta-analytic calculation. Planned analyses included comparing each diagnosis (SZ, bipolar disorder, depression, anxiety, PDs, psychopathy), and unspecified general MI with healthy groups on specific reading skills (phonological processing and decoding; comprehension; single-word reading; rate, speed, accuracy, and fluency). For each reading skill, differences between tests to assess deficits in the patient group were calculated by investigating overlaps of confidence intervals of the summary effect sizes for each test. Risk of publication bias (none identified) was formally assessed via Egger’s and Begg’s tests and with funnel plots.

Results

Of 34 studies in total (Tables 2–3), 19 studies provided data for meta-analysis (Figure 1. PRISMA flowchart); five of these studies also presented composite scores (combining two or more measures) that are covered in the narrative synthesis. The remaining 15 studies contributed to the narrative synthesis only. The findings from the nonforensic and forensic samples are presented separately, followed by a direct comparison of forensic and nonforensic groups.

Figure 1. PRISMA flowchart.

Reading Skills in Nonforensic Populations

Schizophrenia

Phonological Processing and Decoding: Across seven studies (Figure 2(2.1)), SZ showed significantly poorer phonological processing compared to HC with a large effect size (Hedge’s g = −0.88, df = 24, p < 0.00001, CI = [−1.07, −0.70]). There was medium heterogeneity within the data (p = 0.001, I ² = 53%), with nonsignificant differences between the tests (p = 0.15, I ² = 32.3%).

2.1. Phonological processing and decoding

2.2. Comprehension

2.3 Single-word reading

2.4. Rate

2.5 Accuracy

2.6 Fluency

2.7 Speed

Figure 2. Reading deficits in schizophrenia (non-forensic population). Within each specific reading skill, the results are presented for each of the test(s)/measures used, followed by the analysis of differences between tests (last row). Negative values represent a poorer performance of people with schizophrenia in comparison to HC.

References: Arnott et al. [24]; Curzietti et al. [76]; Dondé et al. [18]; Gavilán and García-Albea [39]; Hayes and O’Grady [54]; Ho et al. [45]; Leonard et al. [36]; Light et al. [74]; Maj [62]; Martinez et al. [41]; Patrick et al. [59]; Potter and Nestor [73]; Revheim et al. [21]; Revheim et al. [17]; Walder et al. [22]; Whitford et al. [7]. Abbreviations: BDAE, Boston Diagnostic Aphasia Examination; CTOPP, Comprehensive Test of Phonological Processing (PA, Phonological Awareness, PM, Phonological Memory, RN, Rapid Naming, APA, Alternative Phonological Awareness, ARN, Alternative Rapid Naming); GORT, Gray Oral Reading Test; ITBS, Iowa Test of Basic Skills; ITED, Iowa Test of Educational Development; LNNB, Luria-Nebraska Neuropsychological Battery; NARA, Neale Analysis of Reading Ability; NDRT, Nelson–Denny Reading Test; PALPA, Psycholinguistic Assessments of Language Processing in Aphasia; RAN, Rapid Automatised Naming; RCBA, Reading Comprehension Battery for Aphasia; RNRT, Roentgen’s Nonwords Reading Test; RNST, Roeltgen’s Nonwords Spelling Test; WJTA-III, Woodcock-Johnson III Tests of Achievement Knowledge); WRAT, Wide Range Achievement Test; WRMT-R, Woodcock Reading Mastery Test-Revised. White circle —effect size for a particular study determining the difference between patients and controls. Black diamond —pooled effect size for particular test/subtest. Red diamond —overall effect size for diagnosis for a certain reading skill (e.g., comprehension) including all partial effect sizes.

Comprehension: Across 11 studies (Figure 2(2.2)), SZ showed poorer comprehension than HC with a large overall effect size (Hedge’s g = −0.96, df = 34, p < 0.00001, CI = [−1.15, −0.78]) and medium heterogeneity (p < 0.00001, I ² = 69%). The test differences were significant (p < 0.0001, I ² = 74.5%) with NDRT [49] and GORT-4 [40] showing the largest effect sizes for a comprehension deficit in SZ. In addition, three studies [17,21,41] reported lower Oral Reading Quotient from GORT-4 [40]. In other studies, retrospective assessment revealed that those with a current diagnosis were below the norm during 4th to 11th grade of school [44], with the most prominent deficit in the 11th grade, indicating a gradual decline [44,45]. A similar study on adolescents, who later developed psychosis, displayed a premorbid deficit in comprehension and sentence reading relative to HC [47].

Single-Word Reading: Across 10 studies [17,18,21,22,24, 36,59,62,73,74], there was a significant medium-size deficit (Figure 2(2.3)) in SZ relative to HC (Hedge’s g = −0.70, df = 9, p < 0.00001, CI = [−0.94, −0.46]). There was significant heterogeneity within the results (p = 0.01, I ² = 58%) but no test performed better than others (p = 0.20, I ² = 35.6%). Moreover, in two studies [62,63], both using LNNB—Reading subtest (see Table 2 for test descriptions) [61]—SZ showed a deficit compared to HC (data for meta-analysis not provided). In a third study [72], SZ scored markedly lower (M = 78.00, SD = 21.01) than the norm (M = 100) on WRAT-III [84].

Rate, Speed, Accuracy, and Fluency: Across five studies [17,21,24,76,85], there was a significant large effect of SZ diagnosis on reading rate (Hedge’s g = −1.22, df = 4, p = 0.002, CI = [−1.98, −0.46]) (Figure 2(2.4)). The effect of diagnosis [17,21,24,76] in accuracy failed to reach significance (Hedge’s g = −0.73, df = 3, p = 0.09, CI = [−1.56, 0.10]) (Figure 2(2.5)). There were, however, significant test differences for both rate (p = 0.04, I ² = 64.9%) and accuracy (p = 0.01, I ² = 77.1%), with the GORT-4 revealing large deficits [17,21], and the Alouette [75] showing no deficit [76] (Figures 2(2.4–2.5)). In fluency [17,18,21,41,77], there was a highly significant deficit in SZ (Hedge’s g = −2.03, df = 4, p < 0.00001, CI = [−2.82, −1.24]), but with large heterogeneity within results (84%) (Figure 2(2.6)). In reading speed (time taken to read certain content) [24,76], the effect of diagnosis was nonsignificant (Hedge’s g = −0.50, df = 1, p = 0.11, CI = [−1.11, −0.11]) (Figure 2(2.7)). In an additional study [77], 10–11% of SZ demonstrated nonfluencies (e.g., sound repetitions at beginning of word) in sentence and paragraph reading during the BDAE [38].

Composite Scores: Two studies [17,21] that examined Basic Reading Skills (phonological processing and single-word reading) and Phoneme-Grapheme Knowledge (phonological processing and orthography) composite scores from WJTA-III [35] showed different results, with only one of these showing a significant deficit in SZ [17]. Both studies [17,21] found significantly lower WJTA-III Broad Reading (phonological processing, comprehension, speed) scores in SZ, relative to HC. The study [22] that created a phonology composite score by combining the RNRT [33], RNST [33], WRAT-R [86], and the Controlled Oral Word Association Test (COWAT) [87] also reported a significant deficit in SZ relative to HC.

Reading-related skills

Vocabulary: Six studies [17,21,22,44,45,81] assessed reading-related skills in SZ. There was evidence of impaired vocabulary from an early age [44,45] and those with prodromal illness scored significantly below grade-norms when assessed by the ITBS [42] and ITED [43] as a part of their school performance. Vocabulary, assessed using the NDRT [49], was also impaired in two studies [17,21].

Spelling and Grammar: Spelling in RNST [33] was found to be adversely affected in male patients, while female patients scored similarly to HC [22]. Another study [44], which longitudinally assessed spelling together with grammar and other language-related skills by ITBS [42], found a significant decline in abilities at 11th grade in SZ. Similarly, SZ scored significantly lower in the WJTA-III [35] spelling subtest compared to HC [17]. Grammar was assessed exclusively in one study [22], using Caplan and Hildebrandt’s task [79], showing a stronger and significant deficit in male, relative to female, patients [22].

Orthography: Orthography processes are not reading abilities. However, in languages such as Chinese, orthography and semantics play an important role in reading, in contrast to alphabetical languages such as English where phonological processing plays a key role [81]. One study [81] that investigated orthography processes found significant deficits in orthography-phonology, but not in vocabulary when distinguishing real words from nonwords, in SZ compared to HC.

Affective disorders (depression, anxiety or mania)

Two studies [62,66] assessed single-word reading in depression, both using the REALM [64]. Of these, one study [62] showed a nonsignificant small deficit in people with depression (Hedge’s g = −0.30, df = 0, p = 0.37, CI = [−0.96, 0.36]) and, in the other study [66], all participants performed at 7–8th grade reading level.

Bipolar disorder

The earlier-mentioned study on adolescents [47] had also assessed comprehension premorbidly in a group who later developed nonpsychotic bipolar disorder and found them to have no deficit in comparison to HC.

Personality disorders/psychopathy

One study [30] assessed phonological processing and comprehension, using the Portuguese version of the PALPA [29], and showed medium-size deficits in both phonological processing (Hedge’s g = −0.55, df = 2, p = 0.004, CI = [−0.92, −0.18]) (Figure 3(3.1)) and comprehension (Hedge’s g = −0.47, df = 0, p = 0.05, CI = [−0.87, 0.39]) (Figure 3(3.2)) in people with diagnosed psychopathy (from community settings), compared with nonpsychopathic nonforensic controls.

3.1 Phonological processing and decoding.

3.2 Comprehension.

Figure 3. Reading deficits in community/nonforensic samples of people with psychopathy. Within each specific reading skill, the results are presented for each of the test(s)/measures used, followed by the analysis of differences between tests (last row). Negative values represent a poorer performance of people with personality disorder in comparison to healthy control.

Brites et al. [30]. Abbreviations: PALPA, Psycholinguistic Assessments of Language Processing in Aphasia. White circle —effect size for a particular study determining the difference between patients and controls. Red diamond —overall effect size for diagnosis for a certain reading skill (e.g., comprehension) including all partial effect sizes.

General mental illnesses (nonspecified/mixed)

Two studies [51,58] assessed comprehension and single-word reading while the third study [65] assessed single-word reading only. The first study [58] reported 9th-grade level comprehension as well as 9th-grade level single-word reading when assessed by WRAT-IV [88] in people with unspecific MIs. The second study [51], using the PIAT-comprehension subtest [50], reported 7th-grade comprehension, despite 9–10th grade for single-word reading, in psychiatric patients (majority with alcoholism or nonorganic psychoses). In the third study [65], 75% of the sample with MIs (mainly SZ and affective disorders) read below 7th grade when assessed by REALM [64].

Summary of Deficits in Nonforensic Populations

Overall, SZ was associated with pronounced deficits in phonological processing, comprehension, reading rate, and fluency (Figure 4), with deficits also present in reading-related skills. These deficits appear to be present often from an early age, with reading skills of SZ adults remaining below their achieved education levels. The single-word reading and speed were less impacted. There were few data in affective disorders, and only for single-word reading, showing a mild/nonsignificant deficit. Individuals with PDs/high psychopathy showed mild deficits in both phonological processing and comprehension (Figure 4). Comprehension and single-word reading skills of people with unspecified MIs from nonforensic settings were at secondary school levels, which, although below the norm, were better than those in SZ (Figure 4).

Figure 4. Interpretation of observed reading deficits in included diagnoses.

No deficit = nonsignificant differences between patients and healthy control (HC); Very mild deficit = Hedge’s g up to −0.30 and/or mixed results with the majority of samples scoring within the norm; Mild deficit = Hedge’s g up to −0.50 and/or reading skill at 9–10th-grade level; Moderate deficit = Hedge’s g up to −0.75 and/or reading skill at 7–8th grade level; Severe deficit = Hedge’s g over −0.75 and/or reading skill below 7th grade level. This interpretation considers whether the results were consistent or mixed. Empty circle  = Mixed evidence.

Reading Skills in Forensic Populations

Seven studies [15,26,27,30,68,70,71], all in PDs/psychopathy or general MIs, were found.

Personality disorders/psychopathy

Phonological Processing and Decoding: In the first study [30], the PALPA [29] phonological processing test showed a large deficit in the incarcerated group with diagnosed psychopathy relative to HC (Hedge’s g = −0.85, df = 2, p = 0.0001, CI = [−1.22, −0.47]) (Figure 5(5.1)). The second study [15], using the JDT [25] to examine decoding, showed marked impairment (Hedge’s g = −0.84, df = 0, p = 0.01, CI = [−1.51, −0.17]) in people with nonspecific PDs (and comorbid MIs), relative to HC.

5.1. Phonological processing and decoding

5.2. Comprehension. Psychopathy.

Figure 5. Reading deficits in forensic patients with psychopathy or personality disorders. Within each specific reading skill, the results are presented for each of the test(s)/measures used, followed by the analysis of differences between tests (last row). Negative values represent a poorer performance of people with psychopathy or personality disorder in comparison to healthy control.

Brites et al. [30]; Daderman et al. [15]. Abbreviations: JDT, Jacobson’s Decoding Test; PALPA, Psycholinguistic Assessments of Language Processing in Aphasia. White circle —effect size for a particular study determining the difference between patients and controls. Red diamond —overall effect size for diagnosis for a certain reading skill (e.g., comprehension) including all partial effect sizes.

Comprehension: One study [30] used the PALPA [29] and showed a large deficit in comprehension in incarcerated people with diagnosed psychopathy, compared to HC (Hedge’s g = −0.95, df = 0, p = 0.0003, CI = [−1.48, −0.43]) (Figure 5(5.2)). The other study [15] used a Swedish prose text [55] and found no deficit in PDs.

Single-word Reading: The first study [15] used a Swedish single-word reading test [28] and found significant impairment in PD inmates with comorbid MI and dyslexia, as well as in dyslexic inmates, in comparison to inmates without a PD diagnosis. In the second study [30], a diagnosis of psychopathy did not influence single-word reading as assessed by PALPA [29]. The third study [68] found literacy scores, as assessed by the TOWRE [67], to be below the norm in PD. None of these studies [15,30,68] provided data for effect size calculation.

Rate, Speed, Accuracy, and Fluency: Only one study [15] was found, showing that reading speed was negatively affected in 7 of 10 forensic PD participants, especially in those with comorbid dyslexia.

Reading-related skills

One study [15] showed that spelling was poorer in inmates with PD and dyslexia, as opposed to those with no comorbidities.

General mental illnesses (nonspecified/mixed)

Phonological Processing and Decoding: One study [27] used the JDT–Wordchains [25], the Word Attack test [89], and Phonological Choice [31], and revealed severely impaired phonological skills (below the 6th grade) in people with various MIs. The second study [26] examined correlations between psychopathic traits and phonological and decoding skills in forensic psychiatric patients, assessed with the “Pidgeon” test [34], the MWDT [28], and the JDT [25], and found positive correlations between the superficial item of the Psychopathy Checklist: Screening Version (PCL:SV) [90] and phonological processing and decoding of sentences (but not words). However, as the study did not include HCs or test normative scores, the findings are difficult to understand in terms of quantifying the deficit.

Comprehension: In one study [27] that used the Oral Close subtest of the WRMT-R [37], comprehension in inmates with MI was below 4th grade in 23% of Swedish native and in over 50% of non-native speakers. In another study [26] that used a silent paragraph reading test [56], no significant correlations between psychopathic traits and comprehension scores in people with nonspecified MIs were found.

Single-word Reading: There were two studies [70,71], both using the WRAT [69]. The first study [70] assessed people with various diagnoses (psychosis, mania, alcoholism, and ASPD). It found no significant differences between HC and psychosis (Hedge’s g = 1.42, df = 0, p = 0.68, CI = [−5.40, 8.24]), mania (Hedge’s g = 0.53, df = 0, p = 0.13, CI = [−0.15, 1.20]), or alcohol abuse (Hedge’s g = −0.49, df = 0, p = 0.10, CI = [−1.06, 0.09]) but single-word reading was significantly impaired in ASPD (Hedge’s g = −1.01, df = 0, p = 0.004, CI = [−1.69, −0.33]. The second study [71] found age-moderated differences in people with MIs and a history of violence, with people aged above 45 years scoring significantly better than those below 20 years.

Rate, Speed, Accuracy, and Fluency: One earlier-described study [26] found that, within those with MIs, reading speed [56] was positively correlated with affective and interpersonal traits (Factor 1, PCL:SV [90]).

Reading-related skills

In a study [26] involving Swedish inmates with MIs, neither spelling nor vocabulary scores significantly correlated with psychopathic traits.

Summary of Deficits in Forensic Populations

Overall, there was evidence of severe impairment in phonological processing and decoding in forensic populations with PDs/psychopathy (Figures 4 and 5), similar to that seen in SZ. There was also evidence of deficits in comprehension, single-word reading, and speed in this population (Figures 4 and 5). Studies on forensic patients with various MIs yielded mixed findings although one study [27] that examined inmates did show phonological processing and comprehension to be well below the norm.

Nonforensic versus Forensic Populations: Direct Comparison

Only one study [30] directly compared forensic and nonforensic groups. It used PALPA [29] and revealed a significant medium-size deficit in incarcerated individuals with psychopathy compared to nonincarcerated (community) sample with psychopathy in phonological processing and decoding (Hedge’s g = −0.49, df = 2, p = 0.03, CI = [−0.94, −0.04] (Figure 6(6.1)), and a large deficit in comprehension (Hedge’s g = −0.85, df = 1, p = 0.003, CI = [−1.43, −0.28]) (Figure 6(6.2)). These results support the findings from individual studies indicating severe reading deficits in incarcerated individuals with MI.

6.1 Phonological processing and decoding.

6.2 Comprehension.

Figure 6. Reading deficits in incarcerated vs community samples of people with a diagnosis of psychopathy. Within each specific reading skill, the results are presented for each of the test(s)/measures used, followed by the analysis of differences between tests (last row). Negative values represent a poorer performance of the forensic sample, compared to the nonforensic sample.

Brites et al. [30]. Abbreviations: PALPA, Psycholinguistic Assessments of Language Processing in Aphasia. White circle —effect size for a particular study determining the difference between patients and controls. Red diamond —overall effect size for diagnosis for a certain reading skill (e.g., comprehension) including all partial effect sizes.

Reading Skills Deficits in Mental Illness: Influencing Factors

Symptoms and medication

Of six studies in SZ [17,21,24,72,76,81] that examined the relationship between psychotic symptoms and reading skills, three [21,24,81] found a negative influence of positive and negative symptoms on phonological processing, comprehension, and orthography; and hallucinations negatively affected reading efficiency and speed in one study [76]. Five studies [17,22,41,63,85] examined the effect of antipsychotic dose as chlorpromazine equivalents; four [17,22,63,85] found no relationship with single-word reading, phonological processing, or comprehension, and one [41] found a negative influence of high dosage on fluency and comprehension. No significant association occurred between depressive symptoms and single-word reading [66].

Cognitive function

Six studies [17,21,41,44,54,70] examined the relationship between reading skills and general cognition in SZ. Verbal IQ significantly correlated with comprehension and vocabulary [44]. Lower premorbid IQ (single-word reading) predicted reading comprehension [17,54]. However, general IQ did not significantly predict any of the reading skills [41]. Similarly, working memory did not correlate with comprehension or reading rate in SZ and HC [21]. In forensic populations, full-scale IQ was significantly lower than single-word reading in individuals with SZ and bipolar disorder [70]. These results suggest that general verbal skills may influence comprehension but no marked impact of other cognitive abilities was found.

Education

In SZ, three studies [17,21,41] examined the influence of education and all found reading skills significantly below achieved academic levels. Six studies [18,24,39,62,76,81] matched their groups on education or entered it as a covariate [85], and all found significant impairments in various reading skills. Nonforensic populations with general MIs had single-word reading equivalent to their achieved education but their comprehension was lower [51]. Forensic PD also had comprehension below their education level [15].

Discussion

This systematic review and meta-analysis evaluated existing evidence to identify the type and degree of reading impairments in different MIs, the reading assessment tools that might most consistently detect them, and possible differences in the pattern of reading skills deficits in people with different MIs in forensic and nonforensic settings. Most of the reviewed studies (27/34) included people with SZ. There were seven studies of reading skills deficits in people with different MIs (PD or general MI) in forensic settings. Our findings are discussed below.

Effect of diagnosis in nonforensic samples

We observed significant deficits in multiple reading skills in SZ, resembling the pattern typically seen in dyslexia [6], and consistent with previous evidence for shared genetic and psychophysiological traits in SZ and dyslexia [7]. In our meta-analysis, both phonological processing and comprehension were greatly impaired. These impairments may be associated with ineffective use of contextual information [91] and contribute to poor speech in SZ, especially in close association with thought disorder [92]. Reading rate was low but the deficit in reading accuracy was lower. This indicates relatively preserved single-word reading skills, most likely because they are usually acquired before illness onset and remain intact [47]. In contrast, there was evidence for impairments in vocabulary and spelling, presumably as a result of disrupted scholastic experience. Disrupted scholastic experience during adolescence can affect complex skills such as comprehension [44,45,47], which could precipitate difficulties with processing complex written information in SZ. People with SZ showed reading skills well below their achieved education level (see Education). Reading skills deficits in SZ also do not seem to be explained by other aspects of cognition (see Cognitive Function) although more comprehensive investigations are needed to substantiate this. Our findings (Symptoms and Medication) further indicated that while symptoms and high antipsychotic doses may worsen reading skills, they do not fully explain the profile of reading skills deficits in SZ. Impairment in comprehension and vocabulary was present even before the onset of symptoms [44,45] together with deficient phonological processing, which has been related to disrupted visual processing in SZ since early age [21]. The symptoms can, however, aggravate deficits in reading skills, such as comprehension, which are acquired with experience, and also depend on the earlier acquired skills [93]. Recent data [94] suggest that some aspects of language production (e.g., slower articulation) that can affect reading skills assessments are particularly sensitive to dopamine-D2 receptor blocking antipsychotics. Furthermore, most studies in SZ included more men than women or men solely and also included people with schizoaffective disorder. Further studies need to comprehensively examine specific reading skills in both men and women with schizophrenia and schizoaffective disorder (separately) while taking medication, symptoms, cognition, education, and socioeconomic status into account.

Unlike in SZ and psychosis [51,58,65], nonpsychotic bipolar disorder, and affective disorders, seemed to have comprehension and single-word reading skills comparable to HC [30,47]. Although not all studies specified the type of PD, it seems that reading skill deficits may not be as prominent in nonforensic psychopathy as in SZ.

Effect of diagnosis in forensic samples

Our findings suggest only a weak or no deficit in nonforensic psychopathy but indicate a marked phonological processing and comprehension deficit in the incarcerated group. It is possible that PD/psychopathic individuals with good phonological processing and comprehension are more able to evade incarceration [30,95]. Nonetheless, marked reading deficits in the incarcerated group may have contributed to their poor adjustment within the community [27], which, in turn, increased the risk of incarceration. Men with MIs within forensic settings had significantly lower general reading abilities and spelling than women with MIs [27], consistent with the pattern seen in healthy samples [22].

Clinical implications

Comprehension has a significant influence on decision-making capacity in SZ [96], and this is likely to be true also for people with other MIs, especially within forensic populations. Dyslexia is often underdiagnosed in people with MIs, and this might explain their inability to complete higher education and obtain jobs [15], or the expression of socially unacceptable behaviors [27]. Furthermore, progression and engagement in therapeutic activities within mental health services often depend on good reading and language skills. This highlights a need to accurately identify reading deficits and develop specific programs to improve reading skills of people in psychiatric services. It may be possible to target reading deficits in SZ and other MIs by building on the less affected aspects, such as lexical knowledge (access to words) [97,98], and access to familiar information that can compensate for some of the reading deficits [99], while implementing interventions to ameliorate reading skills [100].

Effect of assessments

Significant between-test differences were found only in tests detecting deficits in comprehension, accuracy, and rate in SZ. In comprehension and rate, the NDRT and GORT-4, and in accuracy, the GORT solely, consistently detected large deficits while the Alouette (French) test detected no deficits (Figure 2). It is conceivable that certain deficits emerge more often/strongly in English compared to some other languages, as is the case in developmental dyslexia [101]. This possibility requires further study.

Conclusions

Our findings demonstrate pronounced deficits in phonological processing and comprehension in SZ and forensic PD/psychopathy. Reading skills in people with other MIs in nonforensic settings seem relatively unaffected. Among the tests, only the NDRT and GORT detected significantly stronger deficits in SZ than other measures. Considering the importance of good reading skills in everyday life, as well as for the clinical success of mental health services, there is a clear need to identify methods that can improve reading in SZ and forensic PD populations. These interventions could potentially build on relatively spared aspects of reading by implementing approaches already effective in dyslexia.