Association Between the Duration of Untreated Psychosis and Selective Cognitive Performance in Community-Dwelling Individuals With Chronic Untreated Schizophrenia in Rural China

Key Points

Question

Is chronic untreated psychosis associated with decreases in cognitive performance?

Findings

In this case-control study of 197 individuals with chronic untreated schizophrenia and 220 individuals without mental illness in rural China, the median duration of untreated schizophrenia was 23 years; a longer duration of untreated psychosis was associated with significantly lower cognitive performance on 3 tests assessing different aspects of executive functioning.

Meaning

This study’s findings suggest that reduced cognitive performance can emerge and continue to progress long after the development of psychosis and may reflect earlier neurodevelopmental and later neurodegenerative etiological factors.

Abstract

Importance

Cognitive deficits constitute core features of schizophrenia, but the trajectories of cognitive difficulties in chronic untreated schizophrenia remain unclear.

Objective

To assess the association of neuropsychological deficits with duration of untreated psychosis in individuals with chronic untreated schizophrenia.

Design, Setting, and Participants

Community-dwelling individuals with chronic untreated schizophrenia (untreated patient group) and individuals without mental illness (control group) were recruited from predominantly rural communities in Ningxia, China between June 20, 2016, and August 6, 2019, and administered the Structured Clinical Interview for DSM-IV, the Mini-Mental State Examination, an 8-test version of the MATRICS Consensus Cognition Battery adapted for use in individuals with low levels of education, and a measure of social cognition.

Main Outcomes and Measures

Comparison of cognitive test scores between the two groups and association of cognitive test scores with duration of untreated schizophrenia.

Results

The patient group included 197 individuals with chronic untreated schizophrenia (101 men [51.3%]; mean [SD] age, 52.1 [11.8] years; median [interquartile range] years of schooling, 3 [0-6] years; median [interquartile range] years of untreated psychosis, 22.9 [14.9-32.8] years). The control group included 220 individuals (118 men [53.6%]; mean [SD] age, 52.1 [11.2] years; median [interquartile range] years of schooling, 4 [0-6] years). The untreated patient group performed significantly worse than the control group on all cognitive measures (adjusted partial Spearman correlation coefficient [Spearman ρ] ranged from −0.35 for the revised Chinese version of the Reading the Mind in the Eyes Test to −0.60 for the Brief Visuospatial Memory Test-Revised; P < .001 for all comparisons). Longer durations of untreated psychosis were associated with lower performance in 3 MATRICS Consensus Cognition Battery measures assessing different aspects of executive functioning (Brief Visuospatial Memory Test–Revised [ρ = −0.20; P = .04]; Brief Assessment of Cognition in Schizophrenia, Symbol Coding subtest [ρ = −0.35; P < .001]; and Neuropsychological Assessment Battery, Mazes subtest [ρ = −0.24; P = .01]). The median duration of untreated psychosis (22.9 years) was associated with estimated score reductions in the 3 measures of 34% (95% CI, 10%-52%), 43% (95% CI, 28%-55%), and 57% (95% CI, 31%-73%), respectively.

Conclusions and Relevance

The findings of this study suggest that long-term untreated schizophrenia was associated with decreases in selective cognitive abilities; both neurodegenerative pathology and neurodevelopmental dysfunction may be factors in cognition in persistent psychosis. Expanding research to include cohorts of patients from underserved rural communities in low- and middle-income countries may provide new insights about the etiological factors, disease course, and management of schizophrenia.

This cross-sectional study uses neuropsychological assessments to examine the association between the duration of untreated psychosis and cognitive performance among community-dwelling individuals with chronic untreated schizophrenia in rural China.

Introduction

Cognitive deficits constitute core features of schizophrenia,^1,2,3,4 encompassing general intellectual deficits^5,6,7 and more specific deficits, such as processing speed and working memory.⁵ These deficits occur in verbal and nonverbal cognitive domains^1,2,8; in premorbid,^5,9,10,11 prodromal,^{12,13,14,15,16,17} first-episode,^8,18 chronic,^2,19,20 and remitted stages of illness^21,22; and in familial and clinical high-risk cohorts.^12,23 They are independently associated with functional outcomes, which supports the view that schizophrenia is a neurodevelopmental cognitive disorder.^{2,24,25,26,27,28,29}

Less is known about the trajectory of cognitive deficits after the development of psychosis. Cognitive differences that occur between the prodrome stage and the conversion to schizophrenia are usually small and/or restricted to a few cognitive domains,^29,30 which is consistent with the view that most cognitive deficits in patients with schizophrenia occur before the onset of psychosis, stabilize after the onset,^{20,31,32,33,34} and remain relatively unchanged thereafter.^32,35 However, some studies report long-term improvements in cognitive functioning after the first episode of psychosis,³⁶ and others report differential rates of decrease over time in different cognitive domains.^25,37,38,39

It is important to clarify the trajectory of cognitive functioning in individuals with chronic untreated psychosis. First, most research involves patients who received treatment with antipsychotic medications, so the cognitive deficits reported do not reflect the unmodified features of psychosis. Although antipsychotic medications may provide modest benefits for cognition,⁴⁰ it is unclear whether they forestall further cognitive decrease. Second, the cognitive worsening observed in patients with chronic untreated schizophrenia suggests that neurodegenerative models may explain some aspects of schizophrenia better than traditional neurodevelopmental models,⁴¹ implicating different pathophysiological mechanisms and treatment targets. Third, higher mortality and morbidity and reduced longevity in patients with chronic schizophrenia⁴² raise unanswered questions about pathological aging; the identification of continuing cognitive decreases would help to clarify this important issue.^35,43,44

Untreated psychosis has long been associated with worse clinical outcomes in patients with schizophrenia.⁴⁵ Studies, primarily performed in high-income countries in which the duration of untreated psychosis is usually less than 5 years, have reported that a longer duration of untreated psychosis is associated with more severe positive and negative symptoms and lower rates of remission,^{46,47,48,49,50} reductions in global and social functioning,^48,49 and increases in social anxiety,⁵¹ hippocampal atrophy, and other structural changes in the brain.^52,53 However, meta-analyses have reported limited or no association between the duration of untreated psychosis and neuropsychological performance or structural brain measures.^54,55,56,57 The trajectory of cognitive functioning in individuals with a long duration of untreated psychosis (ie, 5-50 years) remains unclear.

Chronic untreated psychosis is relatively common among individuals in the rural communities of low- and middle-income countries with limited health care resources,⁵⁰ but there are few reports about these patients. One 14-year follow-up study of individuals with schizophrenia in rural China found that a longer duration of untreated psychosis was associated with more severe psychiatric symptoms and social dysfunction.⁵⁸ A national public health initiative in China^59,60 has facilitated the identification and treatment of community-dwelling individuals with serious mental illness. We used this program to identify a distinct sample of 197 community-dwelling individuals with untreated schizophrenia. This sample provided an opportunity to assess whether cognitive performance decreases as the duration of untreated psychosis extends beyond the 2- to 5-year duration of untreated psychosis that is typically reported in studies from high-income countries.

Methods

This article reports interim findings from an ongoing study assessing cognitive, clinical, and functional domains of untreated psychosis among individuals in China. The study began in June 2015 and is projected to end in May 2021. This study was approved by the institutional review boards of the Shanghai Mental Health Center, New York University, and Columbia University. Written informed consent was obtained from all participants. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for case-control studies.

Participants

Untreated individuals with schizophrenia were identified through the registry of persons with serious mental illness in the Ningxia Hui Autonomous Region in northwest China. This provincial-level registry is part of the National Continuing Management and Intervention Program for Psychoses,⁶¹ which aims to register and treat all individuals with serious mental illness nationwide. Individuals in the Ningxia registry who had never received antipsychotic medication or were virtually drug naive (defined as no receipt of antipsychotic medications in the last 5 years and lifetime receipt of antipsychotic medications of <16 days) were identified via phone contact with the patients’ guardians. If potentially eligible, a home or local health clinic visit was arranged; after receiving informed consent from the patient and their guardian or adult family members, the adapted Chinese version of the Structured Clinical Interview for DSM-IV (SCID-IV)⁶² was administered by a trained psychiatrist. Drug-naive or virtually drug-naive individuals with a diagnosis of schizophrenia without intellectual disability or organic mental disorder were eligible. A detailed discussion of medical history was conducted with both the patient and guardian(s) to identify the year and month of the first onset of hallucinations, delusions, or disorganized thinking. The duration of untreated psychosis was defined as the interval between symptom onset and study enrollment. Common reasons for not seeking treatment were the cost of hospitalization, the lack of recognition of psychotic symptoms as indications of mental illness, the patients’ denial of the presence or severity of illness, and the belief that treatment could be harmful.

Individuals without mental illness recruited for the control group were identified from local health clinic registries (which included almost all permanent residents) in communities around Ningxia. Individuals who were similar to the patients in age, sex, years of education, location of residence (rural or urban), and ethnicity (self-identified as Han or Hui) and who provided written informed consent were asked to complete the SCID-IV to exclude those with current or previous psychotic or organic mental disorders or with family histories of intellectual disability or organic mental disorders.

Cognitive Assessment

Individuals in the untreated patient group and the control group were administered the Mini-Mental State Examination (MMSE; score range, 0-39, with lower scores indicating greater cognitive impairment) and the Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS) Consensus Cognition Battery (MCCB) by trained psychologists or public health fellows.

The Chinese version of the MCCB includes 9 tests⁶³ (described in the order in which they were administered in the battery): (1) the Trail Making Test, Part A (TMT-A) is a timed paper-and-pencil test in which the respondent draws a line to connect 25 consecutively numbered circles placed irregularly on a sheet of paper (scored based on amount of time [seconds] to complete test; score range, 0-300 seconds, with higher scores indicating greater cognitive deficit); (2) the Brief Assessment of Cognition in Schizophrenia, Symbol Coding subtest (BACS Symbol Coding) is a timed paper-and-pencil test in which the respondent uses a key to write digits that correspond with symbols (score range, 0-110 items correct, with higher scores indicating better cognitive performance); (3) the Hopkins Verbal Learning Test–Revised (HVLT-R) is an orally administered test in which a list of 12 words from 3 taxonomic categories is presented, and the respondent is asked to recall as many words as possible after each of 3 learning trials (score range, 0-36, with higher scores indicating better verbal recall); (4) the Wechsler Memory Scale, third edition, Spatial Span subtest (WMS-III Spatial Span) uses a board on which 10 cubes are irregularly spaced, with the respondent tapping cubes in the same (or reverse) sequence as the test administrator (score range, 0-18, with higher scores indicating better performance); (5) the Neuropsychological Assessment Battery, Mazes subtest (NAB Mazes) comprises 7 timed paper-and-pencil mazes of increasing difficulty that measure foresight and planning (index score range, 0-26, with higher scores indicating better performance); (6) the Brief Visuospatial Memory Test–Revised (BVMT-R) is a visual memory test with 6 geometric designs in 6 locations on a card that respondents learn over 3 trials, with correct designs and locations summed for a total recall score (score range, 0-36, with higher scores indicating better total recall); (7) the Category Fluency, Animal Naming subtest (CF Animal Naming) is an oral test in which respondents name as many animals as possible within 60 seconds (score based on total number of animals named, with higher numbers indicating better performance); (8) the Continuous Performance Test–Identical Pairs version (CPT-IP) is a computer-administered measure of sustained attention, in which the respondent presses a response button for consecutive matching numbers (score range, 0-5 for each of 3 trials, with higher scores indicating better sustained attention); and (9) the Mayer-Salovey-Caruso Emotional Intelligence Test, Managing Emotions subtest (MSCEIT Managing Emotions) is a paper-and-pencil multiple-choice test that assesses the respondent’s management of emotions (score range, 50-150, with higher scores indicating better management of emotions). The Letter-Number Span test, which is included in the English-language version of the MCCB, was removed from the Chinese version because it uses the English alphabet.

A pilot study found that most situations presented in the MSCEIT Managing Emotions⁶⁴ were unfamiliar to rural respondents, so we substituted an alternative social cognition test. This test combined a translated version of the Reading the Mind in the Eyes Test (RMET), which includes 36 photographs of the eyes of persons with white ancestry,⁶⁵ with a separate Chinese version, which includes 34 photographs of the eyes of persons with Asian ancestry,⁶⁶ to create a revised 70-item Chinese version (RMET-CV-R). In the RMET-CV-R, respondents view the black-and-white photographs of eyes and select which of 4 terms best describes the mental state of that person (score range, 0-70, with higher scores indicating better emotional discrimination). The 8 MCCB tests plus the RMET-CV-R assessed 7 distinct cognitive domains: processing speed, verbal learning, working memory, reasoning, visual learning, social cognition, and attention and vigilance.

Additional changes in the administration of the MCCB tests were needed to make them suitable for this study’s sample. The Chinese conorming study of the MCCB⁶³ was limited to predominantly urban-dwelling individuals aged 20 to 59 years with more than 5 years of education. However, most untreated patients in the present study were older individuals with less than 5 years of education (more than 30% of patients had never attended school) who lived in rural counties. Few patients had ever participated in psychological testing, less than 20% of patients had ever used a computer (which was required for administration of the CPT-IP), and a single reading of the MCCB instructions was often insufficient for them to understand the test expectations. If we had used the age and educational criteria of the Chinese conorming study, we would have needed to exclude 65% of identified untreated individuals with schizophrenia, producing an unrepresentative sample and perpetuating the theoretical and ethical problem of excluding the most vulnerable individuals with disabilities from research and models examining the etiological factors and disease management of individuals with serious mental illness.

We chose to administer this cognitive battery to all consenting individuals with schizophrenia (regardless of age or years of education) but revised the administration procedures to improve the validity of test scores. The formal scored portion of each test remained unchanged, but test administrators repeated or rephrased instructions for each test up to 6 times until respondents understood the test expectations. If respondents were uncertain, the administrators asked them to describe the test expectations before beginning the formal test. We also added a training maze before administering the NAB Mazes, and we trained respondents to use a computer mouse before starting the CPT-IP.

Despite these adaptations, several patients (and some individuals in the control group) had difficulty understanding the test requirements or were otherwise unable to complete some of the tests. Standard practice when scoring the MCCB is to use the worst score (ie, 300 seconds for the TMT-A and 0 points for all other tests); however, given our sample’s demographic differences from participants in almost all previous studies that used the MCCB, it was important to determine whether respondents understood the assessment tasks. Therefore, we developed test-specific rules to distinguish test results with confirmed validity (ie, respondents understood the test expectations) from test results with indeterminate validity (ie, respondents may not have understood the test expectations) based on interviewers’ assessments of the respondents’ comprehension of the required task and their ability to complete the task. For tests that required multiple trials (ie, the HVLT-R, WMS-III Spatial Span, BVMT-R, and CPT-IP), each trial had to have test results with confirmed validity for the overall test results to be considered to have confirmed validity. When assessing cognitive domains, if any of the 3 MCCB tests associated with the processing speed domain (ie, the TMT-A, BACS Symbol Coding, and CF Animal Naming) had test results with confirmed validity, the overall results for the processing speed domain were considered to have confirmed validity. Details of the changes made in our adapted version of the MCCB will be presented in a subsequent article.

Statistical Analysis

We used χ² and Wilcoxon rank sum tests to compare demographic characteristics between groups. A statistical test based on the Spearman partial correlation coefficient⁶⁷ was used to detect differences in cognitive test results between groups (individuals in the untreated patient group were assigned a value of 1 and individuals in the control group were assigned a value of 0), controlling for sex, age, ethnicity (Han vs Hui), years of education, and location of residence (urban vs rural).

The association of the duration of untreated psychosis with cognitive performance among untreated patients was assessed using Spearman partial correlation tests, which were controlled for sex, years of education, and age at psychosis onset. We adjusted for sex and years of education because these variables were significantly associated with cognitive performance in our sample. Chronological age is the sum of the duration of untreated psychosis and the age at onset, so controlling for age would have masked the consequences of the duration of untreated psychosis. In contrast, the age at onset is associated with cognitive variables but does not overlap with the duration of untreated psychosis; controlling for age at onset made it feasible to formulate valid interpretations about the cognitive differences between patients with different durations of untreated psychosis. When assessing the association of the duration of untreated psychosis with the results of the 8 MCCB tests and the RMET-CV-R, we adjusted for multiple testing using the Benjamini and Hochberg method, with statistical significance set at a false discovery rate (FDR) of less than 0.05.⁶⁸

The effect size of the duration of untreated psychosis was assessed using generalized linear models, which accommodated various distributions of the outcome variables and estimated the consequences of the factors of interest, adjusted for covariates, including age at onset, sex, and 4 educational levels based on total years of education received (0 years, 1-3 years, 4-6 years, and ≥7 years). Based on the distributions of cognitive measures associated with the duration of untreated psychosis, we used models for outcomes with variance proportional to the mean score of the cognitive measure or with a constant coefficient of variation. Using a logarithmic function to link mean cognitive measures to the duration of untreated psychosis and covariates, we converted the estimated coefficient of the duration of untreated psychosis to the mean ratio of the cognitive outcome (adjusted for covariates) for each 5-year increase in the duration of untreated psychosis.

To compare findings when using all test results with those obtained when limiting the analyses to test results with confirmed validity (ie, tests with indeterminate validity were excluded), 2 separate sets of analyses were conducted; for analyses using all test results, test results with indeterminate validity were assigned the worst possible scores. Data used in this interim analysis were collected from June 20, 2016, to August 6, 2019. Analyses were performed using SAS version 9.4 (SAS Institute Inc). All statistical tests were 2-tailed with statistical significance set at P<.05.

Results

The untreated patient group comprised 197 individuals with chronic untreated schizophrenia; of those, 101 individuals (51.3%) were male, with a mean (SD) age of 52.1 (11.8) years, a median age at psychosis onset of 27.3 years (IQR, 22.0-33.8 years), a median duration of untreated psychosis of 22.9 years (IQR, 14.9-32.8 years), and a maximum duration of untreated psychosis of 58.9 years (Table 1). The median education level was 3.0 years (IQR, 0-6.0 years; 59 individuals [29.9%] had never attended school). Twelve patients (6.1%) were virtually drug naive, and 185 patients (93.9%) had never received antipsychotic medication. The control group comprised 220 individuals without mental illness; of those, 118 individuals (53.6%) were male, with a mean (SD) age of 52.1 (11.2) years and a median educational level of 4.0 years (IQR, 0-6.0 years; 69 individuals [31.4%] had never attended school).

Table 1. Demographic Characteristics of Participants.

Characteristic	No. (%)		P valuea
	Untreated patient group (n = 197)	Control group (n = 220)
Sex
Male	101 (51.3)	118 (53.6)	.63
Female	96 (48.7)	102 (46.4)
Age, y
Range	19-81	26-78	.94
Mean (SD)	52.1 (11.8)	52.1 (11.2)
Median (IQR)	52.0 (44.0-61.0)	52.0 (45.0-58.5)
Ethnicity
Han	116 (58.9)	128 (58.2)	.89
Hui	81 (41.1)	92 (41.8)
Location of residence
Urban	15 (7.6)	13 (5.9)	.49
Rural	182 (92.4)	207 (94.1)
Education, y
Quartile
0	59 (29.9)	69 (31.4)	.09
1-3	49 (24.9)	36 (16.4)
4-6	42 (21.3)	65 (29.5)
≥7	47 (23.9)	50 (22.7)
Range	0-12	0-14	.96
Mean (SD)	3.7 (3.4)	3.7 (3.4)
Median (IQR)	3.0 (0-6.0)	4.0 (0-6.0)
Marital status
Never married	46 (23.4)	1 (0.5)	<.001
Currently married	89 (45.2)	208 (94.5)
Previously married	62 (31.4)	11 (5.0)
Current living situation
Lives alone	54 (27.4)	3 (1.4)	<.001
Lives with family members	143 (72.5)	217 (98.6)
Self-reported family financial condition in previous month
Excellent	0	2 (0.9)	<.001
Good	8 (4.1)	53 (24.1)
Average	40 (20.3)	108 (49.1)
Poor	106 (53.8)	52 (23.6)
Very poor	43 (21.8)	5 (2.3)
Self-reported level of physical activity in previous 3 mo (0-6)b
Range	0-6	3-6	<.001
Mean (SD)	3.2 (2.5)	6.0 (0.2)
Median (IQR)	3.0 (0-6.0)	6.0 (6.0-6.0)
Age at psychosis onset, y
Range	10.7-56.3	NA	NA
Mean (SD)	28.6 (9.0)	NA
Median (IQR)	27.3 (22.0-33.8)	NA
Duration of untreated psychosis, y
Range	0.2-58.9	NA	NA
Mean (SD)	23.5 (12.3)	NA
Median (IQR)	22.9 (14.9-32.8)	NA

Abbreviations: IQR, interquartile range; NA, not applicable.

^aP values were derived from χ² tests for group differences in categorical variables and Wilcoxon rank sum tests for group differences in continuous variables.

^bSelf-reports of the level of physical activity in each of the last 3 months were rated as 0 (significantly reduced), 1 (somewhat reduced), or 2 (not reduced), resulting in a total score of 0 to 6 with higher scores representing higher levels of physical activity.

Individuals in both the untreated patient group and the control group were predominantly residents of rural counties (more than 328 individuals [>80%] were farmers). The groups did not substantially differ by sex, age, years of education, location of residence, or ethnicity; however, individuals in the untreated patient group were less likely to be currently married or living with family members and more likely to have lower income and lower levels of physical activity compared with individuals in the control group.

In the control group, 1925 of 1980 test results (97.2%; range, 94.1%-100%) for the 8 MCCB tests and the RMET-CV-R had confirmed validity. In the untreated patient group, 1270 of 1773 test results (71.6%; range, 48.7%-81.2%) had confirmed validity (Table 2). After controlling for covariates, the analysis limited to test results with confirmed validity (in the untreated patient group, samples sizes ranged from 96 for the RMET-CV-R to 189 for the MMSE; in the control group, sample sizes ranged from 207 for the CPT-IP to 220 for the MMSE) found that individuals in the untreated patient group had worse performance than those in the control group on all cognitive measures (the adjusted Spearman partial correlation coefficient [Spearman ρ] ranged from −0.35 for the RMET-CV-R to −0.60 for the NAB Mazes; FDR<0.001; P < .001 for all comparisons). When test results with indeterminate validity were included (ie, 197 test results in the untreated patient group and 220 test results in the control group), these results were more significant (Spearman ρ ranged from −0.55 for the CPT-IP to −0.65 for the BVMT-R; FDR<0.001; P < .001 for all comparisons) (eTable in the Supplement). These results indicate the discriminant validity of the adapted version of the MCCB used in this study.

Table 2. Cognitive Measures in Untreated Patient Group vs Control Group^a.

Test (theoretical score range)	Untreated patient group				Control group				Spearman ρb
	No. of tests	Score, mean (SD)	Score, median (IQR)	Score range	No. of tests	Score, mean (SD)	Score, median (IQR)	Score range
MMSE (0-39), total	189	17.3 (10.1)	17 (9-25)	0-38	220	26.3 (7.6)	27 (21-32)	4-38	−0.56
Overall MCCB
Test results with confirmed validity (0-9), No.	197	6.5 (3.0)	8 (5-9)	0-9	220	8.8 (0.7)	9 (9-9)	4-9	−0.57
Cognitive domains with confirmed validity (0-7), No.	197	5.1 (2.2)	6 (4-7)	0-7	220	6.9 (0.4)	7 (7-7)	4-7	−0.58
Specific MCCB testsc
TMT-A (0-300), s	143	167 (97)	140 (76-300)	33-300	210	95 (81)	66 (46-96)	16-300	0.57
BACS Symbol Coding (0-110), No.	136	15.2 (12.2)	15 (3-23)	0-51	205	25.1 (15.8)	25 (13.0-35.0)	0-69	−0.54
HVLT-R (0-36), total	157	9.2 (6.8)	9 (3-14)	0-27	220	15.5 (6.5)	15 (11-20)	0-33	−0.51
WMS-III Spatial Span (0-18), total	156	7.4 (4.7)	7 (4-11)	0-18	220	12.4 (3.9)	13 (10-15)	3-22	−0.58
NAB Mazes (0-26), subtest	160	2.9 (4.0)	1 (0-4)	0-19	214	7.7 (5.6)	7 (3-11)	0-24	−0.59
BVMT-R (0-36), total	158	5.3 (5.2)	4 (1-8)	0-22	219	13.1 (8.3)	12 (6-19)	0-32	−0.60
CF Animal Naming, No.	152	8.7 (5.4)	8 (4-13)	0-21	220	13.6 (4.3)	14 (10.5-16.0)	0-26	−0.48
CPT-IP (0-5), meand	112	0.60 (0.67)	0.33 (0.06-1.05)	0-2.96	207	1.05 (0.81)	0.92 (0.34-1.61)	0-3.29	−0.43
RMET-CV-R (0-70), No.c	96	25.4 (9.8)	23 (18-32)	11-55	210	30.3 (11.9)	29 (21-41)	7-57	−0.35

Abbreviations: BACS Symbol Coding, Brief Assessment of Cognition in Schizophrenia, Symbol Coding subtest; BVMT-R, Brief Visuospatial Memory Test–Revised; CF Animal Naming, Category Fluency, Animal Naming test; CPT-IP, Continuous Performance Test–Identical Pairs version; HVLT-R, Hopkin’s Verbal Learning Test–Revised; MCCB, Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS) Consensus Cognition Battery; MMSE, Mini-Mental Status Examination; NA, not applicable; NAB Mazes, Neuropsychological Assessment Battery, Mazes subtest; RMET-CV-R, Reading the Mind in the Eyes Test, revised Chinese version; TMT-A, Trail Making Test, Part A; WMS-III Spatial Span, Wechsler Memory Scale, third edition, Spatial Span subtest.

^aIncludes test results with confirmed validity on the MMSE and 9 cognitive tests (8 tests from the MCCB plus the RMET-CV-R).

^bSpearman partial correlation coefficient for the association between cognitive outcome and group (the untreated patient group was assigned a value of 1, and the control group was assigned a value of 0), controlled for sex, age, years of education, location of residence (rural vs urban), and ethnicity (Han vs Hui). Based on the statistical test for Spearman partial correlation coefficients,⁶⁷ all P values were less than .001.

^cFalse discovery rates <.001.

^dCalculated using d′ reaction time.

Analyses limited to test results with confirmed validity found that scores on the BACS Symbol Coding (n = 136; Spearman ρ = −0.35; P < .001), NAB Mazes (n = 160; Spearman ρ = −0.24; P = .01), and BVMT-R (n = 158; Spearman ρ = −0.20; P = .04) were significantly lower in patients with a longer duration of untreated psychosis (Table 3). Analyses including all 197 test results identified 4 additional tests in which lower cognitive performance was significantly associated with a longer duration of untreated psychosis: the TMT-A (Spearman ρ = 0.14; P = .03), HVLT-R (Spearman ρ = −0.16; P = .03), CPT-IP (Spearman ρ = −0.17; P = .03), and RMET-CV-R (Spearman ρ = −0.16; P = .03). The 2 overall measures of respondents’ difficulty with understanding tasks (ie, the number of MCCB test results with indeterminate validity and the number of MCCB domains with test results of indeterminate validity) increased significantly with a longer duration of untreated psychosis (for the number of MCCB test results, Spearman ρ = 0.18; P = .01; for the number of MCCB domains with indeterminate validity, Spearman ρ = 0.19; P = .009).

Table 3. Spearman Partial Correlation Coefficients for Association Between Duration of Untreated Psychosis and Cognitive Outcomes^a .

Test	Untreated patients with test results with confirmed validity			All untreated patients (N = 197)b
	No. of tests	Spearman ρ	P value	Spearman ρ	P value
MMSE, total	189	−0.14	.05	−0.10	.16
Overall MCCB
Test results with indeterminate validity, No.	NA	NA	NA	0.18	.01
Cognitive domains with indeterminate validity, No.	NA	NA	NA	0.19	.009
Specific MCCB testsc
TMT-A, s	143	0.10	.43	0.17	.03
BACS Symbol Coding, No.	136	−0.35	<.001	−0.32	<.001
HVLT-R, total	157	−0.08	.44	−0.16	.03
WMS-III Spatial Span, total	156	−0.07	.47	−0.10	.15
NAB Mazes, index	160	−0.24	.01	−0.24	.003
BVMT-R, total	158	−0.20	.04	−0.24	.003
CF Animal Naming, No.	152	−0.09	.43	−0.12	.11
CPT-IP, meand	112	−0.19	.10	−0.17	.03
RMET-CV-R, No.c	96	−0.004	.97	−0.16	.03

Abbreviations: BACS Symbol Coding, Brief Assessment of Cognition in Schizophrenia, Symbol Coding subtest; BVMT-R, Brief Visuospatial Memory Test–Revised; CF Animal Naming, Category Fluency, Animal Naming test; CPT-IP, Continuous Performance Test–Identical Pairs version; HVLT-R, Hopkin’s Verbal Learning Test–Revised; MCCB, Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS) Consensus Cognition Battery; MMSE, Mini-Mental Status Examination; NA, not applicable; NAB Mazes, Neuropsychological Assessment Battery, Mazes subtest; RMET-CV-R, Reading the Mind in the Eyes Test, revised Chinese version; TMT-A, Trail Making Test, Part A; WMS-III Spatial Span, Wechsler Memory Scale, third edition, Spatial Span subtest.

^aIncludes test results with confirmed validity and all test results of untreated patients on the MMSE and 9 cognitive tests (8 tests from the MCCB plus the RMET-CV-R).

^bThe worst possible scores were assigned to tests with indeterminate validity. For the TMT-A, the worst possible score was 300 seconds. For all specific MCCB tests and the RMET-CV-R, the worst possible score was 0.

^cP values were adjusted for false discovery rates.

^dCalculated using d′ reaction time.

After adjusting for covariates, for each 5-year increase in the duration of untreated psychosis, the mean number of MCCB test results and the mean number of MCCB domains with test results of indeterminate validity increased by 8.7% (95% CI, 1.7%-16.2%) and 8.9% (95% CI, 1.7%-16.6%), respectively, which suggested greater difficulty in completing the tests (Table 4). Among the 3 MCCB tests that were significantly associated with the duration of untreated psychosis in analyses limited to test results with confirmed validity, for each 5-year increase in the duration of untreated psychosis, the mean score reductions on the BVMT-R, BACS Symbol Coding, and NAB Mazes were 8.8% (95% CI, 2.3%-14.8%), 11.6% (95% CI, 6.9%-16.1%), and 16.7% (95% CI, 7.8%-24.7%), respectively.

Table 4. Covariate-Adjusted Mean Ratios^a.

Test	Untreated patients with test results with confirmed validity		All untreated patients (N = 197)b
	No. of tests	Mean ratio (95% CI)	Mean ratio (95% CI)
Overall MCCB
Test results with indeterminate validity, No.	NA	NA	1.09 (1.02-1.16)
Cognitive domains with indeterminate validity, No.	NA	NA	1.09 (1.02-1.17)
Specific MCCB tests
TMT-A, s	NA	NA	1.07 (1.01-1.14)
BACS Symbol Coding, No.	136	0.88 (0.84-0.93)	0.87 (0.82-0.92)
HVLT-R, total	NA	NA	0.93 (0.88-0.99)
NAB Mazes, index	160	0.83 (0.75-0.92)	0.81 (0.73-0.91)
BVMT-R, total	158	0.91 (0.85-0.98)	0.89 (0.83-0.96)
CPT-IP, meanc	NA	NA	0.86 (0.78-0.95)
RMET-CV-R, No.	NA	NA	0.92 (0.86-0.99)

Abbreviations:: BACS Symbol Coding, Brief Assessment of Cognition in Schizophrenia, Symbol Coding subtest; BVMT-R, Brief Visuospatial Memory Test–Revised; CPT-IP, Continuous Performance Test–Identical Pairs version; HVLT-R, Hopkin’s Verbal Learning Test–Revised; MCCB, Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS) Consensus Cognition Battery; NA, not applicable; NAB Mazes, Neuropsychological Assessment Battery, Mazes subtest; RMET-CV-R, Reading the Mind in the Eyes Test, revised Chinese version; TMT-A, Trail Making Test, Part A.

^aMean ratios for each 5-year increase in the duration of untreated psychosis for 7 cognitive tests with statistically significant associations with the duration of untreated psychosis in untreated patients, adjusted for age at psychosis onset, sex, and years of education. The Mini-Mental Status Examination, the Category Fluency, Animal Naming Test, and the Wechsler Memory Scale, third edition, Spatial Span subtest were not included in this analysis because these tests were not significantly associated with the duration of untreated psychosis.

^bThe worst possible scores were assigned to tests with indeterminate validity. For the TMT-A, the worst possible score was 300 seconds. For all specific MCCB tests and the RMET-CV-R, the worst possible score was 0.

^cCalculated using d′ reaction time.

Among 7 tests significantly associated with the duration of untreated psychosis in analyses that included all test scores, for each 5-year increase in the duration of untreated psychosis, the mean TMT-A score increased 7.2% (95% CI, 0.8%-14.1%), indicating greater dysfunction, and the mean score reductions were 13.2% (95% CI, 8.0%-18.2%) on the BACS Symbol Coding, 6.7% (95% CI, 0.7%-12.3%) on the HVLT-R, 18.7% (95% CI, 8.7%-27.5%) on the NAB Mazes, 10.8% (95% CI, 3.9%-17.1%) on the BVMT-R, 13.7% (95% CI, 4.6%-21.8%) on the CPT-IP, and 8.1% (95% CI, 1.2%-14.4%) on the RMET-CV-R.

Based on these estimates, the median duration of untreated psychosis (22.9 years) was associated with performance reductions in the 3 MCCB measures with significant results based on the analysis of test results with confirmed validity of 34% (95% CI, 10%-52%) for the BVMT-R, 43% (95% CI, 28%-55%) for the BACS Symbol Coding, and 57% (95% CI, 31%-73%) for the NAB Mazes. Corresponding performance reductions in the 7 measures with significant results based on the analysis of all test results ranged from 27% (95% CI, 3%-45%) for the HVLT-R to 61% (95% CI, 34%-77%) for the NAB Mazes.

Discussion

To our knowledge, this study is the first to assess cognition in a relatively large sample of individuals who had active psychosis that remained untreated over a long period, with a median duration of untreated psychosis of 22.9 years and a maximum duration of untreated psychosis of 58.9 years. The study provides a new understanding of the trajectory of chronic psychotic illness unaltered by medications or hospitalization. Our results suggest that untreated psychosis is associated with progressively lower performance on a battery of cognitive tests. This finding differs from shorter-term studies of the duration of untreated psychosis, which generally report unchanging impairment of cognitive function (or short-term cognitive decrease^19,25) and no association between the duration of untreated psychosis and cognition.⁵⁶ It is important to note, however, that this study raises psychometric and interpretive questions about the adaptation of cognitive tests that were originally normed for a different population and about the assessment of individuals with schizophrenia in low- and middle-income countries, which could not be addressed fully in this article.

With this caveat in mind, these results support the hypothesis that a longer duration of untreated psychosis is associated with lower cognitive performance, and the findings are consistent with studies reporting that decreases in cognitive performance among patients with chronic psychosis are likely to be selective, unlike the broader cognitive decreases reported in the premorbid, prodromal, and first-episode stages of the illness.^19,25 The comparison of individuals in the untreated patient group with those in the control group indicated significantly lower performance in the untreated patient group for all cognitive measures assessed, which is consistent with the broad cognitive and performance deficits typically seen in patients with schizophrenia. However, among the untreated patients, when using our more conservative analyses, which were limited to test scores with confirmed validity, the association of lower cognitive functioning with a longer duration of untreated psychosis was only significant in 3 of 9 cognitive tests. The conceptual coherence of these 3 selective deficits adds to their potential importance; the deficits included visual learning (BVMT-R), problem solving (NAB Mazes), and processing speed (BACS Symbol Coding), which are all components of executive functioning.

Progressive cognitive performance decreases in patients with chronic untreated psychosis can be substantial, but the mechanisms underlying these selective decreases remain unclear. The decreases may reflect the trajectories of selective premorbid cognitive vulnerabilities that extend beyond the development of psychosis,^11,69 consistent with a neurodevelopmental etiology. Executive dysfunctions associated with a longer duration of untreated psychosis may, alternatively, reflect a progressive process emerging after psychosis onset instead of (or in addition to) a longstanding neurodevelopmental problem,⁴¹ possibly involving neurodegenerative pathophysiology or accelerated aging.

Owing to the lower level of education and relatively older age of the predominantly rural sample in the present study, use of the standard inclusion criteria for administering the MCCB would have excluded 65% of identified community-dwelling individuals with untreated schizophrenia. To make it possible to administer the MCCB to all participants, regardless of age or years of education, we revised the test administration procedures and developed test-specific criteria to assess whether respondents understood the required tasks, then classified test results as having confirmed validity or indeterminate validity. This step is important in studies of individuals in China and other low- and middle-income countries. The inclusion criteria and validation studies for use of the MCCB in these countries are typically based on urban populations with higher educational levels who are receiving psychiatric treatment^70,71; these cohorts are not demographically representative of the residents of rural communities that account for approximately one-half of the populations of low- and middle-income countries, which are locations in which underserved patients with the most disabilities reside. Notably, such validation procedures may facilitate future research in which tests and conventions developed in western countries need to be adapted for use in populations that are geographically isolated, economically disadvantaged, and have low educational levels.

More research is needed to confirm the validity of this method of expanding the target group that can receive the MCCB, but the significantly lower scores of individuals in the untreated patient group compared with those in the control group (which persisted after adjustment for potential confounding variables) for all 8 MCCB tests and the RMET-CV-R (P < .001 for all comparisons) support the discriminant validity of our adapted version of the MCCB. These differences were present when limiting analyses to test results with confirmed validity and when using all test results by assigning the worst possible score to test results with indeterminate validity. Employing test-specific validity measures to limit the test results analyzed made our findings more conservative. Whether the findings of less restrictive analyses that include all test results (eg, the 4 additional MCCB tests that were significantly associated with the duration of untreated psychosis when analyzing all test results) should be accepted or considered false-positive findings will need to be explored in future studies.

Limitations

This study has several limitations. First, despite the discriminant validity of our adapted version of the MCCB, further research will be needed to assess its validity in a population for which the original MCCB was not normed (ie, in individuals with lower educational levels or older age). In particular, additional investigation will be required to explain the cognitive scores of individuals in both the untreated patient and control groups, which were substantially lower than those reported in previous studies.

Second, our cross-sectional correlation of the duration of untreated psychosis with cognitive measures did not assess longitudinal changes in individual patients. Confirming the association between the duration of untreated psychosis and decreasing cognitive performance requires comparing longitudinal data from individuals with schizophrenia who remain untreated with demographically matched individuals without mental illness.

Third, the current analysis did not include a control group of treated patients matched with untreated patients by age at onset and duration of illness, making it difficult to directly control for age in our estimates of the effect sizes for the duration of untreated psychosis. Fourth, untreated patients with chronic psychosis are not representative of all patients with chronic psychosis, so the trajectory of their cognitive performance may be different from the trajectory observed if all chronic patients remained untreated. Fifth, deficits in declarative memory, which may emerge during aging (especially in patients with neurodegenerative conditions), were not comprehensively assessed by the 8 MCCB tests or the RMET-CV-R.

Conclusions

The results of this study indicate that selective reductions in cognitive performance in a distinct sample of individuals with chronic untreated schizophrenia are associated with prolonged duration of untreated psychosis. These findings, if confirmed, could potentially alter the view that cognition in psychosis is mainly characterized by stable deficits that are explained by neurodevelopmental etiological factors.

Supplement.

eTable. Comparison of the MMSE, Each of the 8 Cognitive Tests in the Adapted MCCB, and the REMT-CV-R Between All Untreated Patients and Individuals in the Control Group