Date of Award

January 2022

Document Type

Open Access Thesis

Degree Name

Medical Doctor (MD)



First Advisor

Toral Surti


The cognitive impairments of schizophrenia drive the functional disability of the illness but are difficult to treat. One barrier to effective cognitive interventions may be medical illnesses that compromise cognition and are over-represented in people with schizophrenia. Obstructive sleep apnea (OSA) is treatable, causes reversible impairments in many cognitive domains also affected by schizophrenia, and is likely under-diagnosed in people with schizophrenia. We have estimated the prevalence of OSA in schizophrenia, both by self-report and with a predictive model, and characterized the associations between OSA and cognition and functional capacity in schizophrenia, using a large dataset of 3942 patients with schizophrenia collected by the Veterans Administration Cooperative Studies Program (CSP) #572 “Genetics of Functional Disability in Schizophrenia and Bipolar Illness”. Neuropsychological tests included TMT-A, BACS Symbol Coding, Category Fluency, verbal learning, working memory and NAB Mazes. Functional capacity measures were the UCSD Performance Skills Assessment Battery (UPSA-B) and the Everyday Functioning Battery- Advanced Finances (EFB-AF). Phi correlations were used to assess associations of self-reported OSA (R-OSA) with demographic and clinical factors. Self-reported diagnosis may underestimate prevalence of OSA in this sample, so a clinical prediction model was also used to calculate predicted prevalence of OSA (P-OSA). Each participant’s composite cognitive score (CCS) was calculated by averaging their age- and gender-corrected T-scores for each cognitive test, with higher scores indicative of better performance. T-tests compared assessments between reported and non-reported OSA (R-OSA v. nR-OSA) and predicted and non-predicted OSA (P-OSA v. nP-OSA). ANOVAs were used to examine differences in CCS, UPSA-B, and EFB-AF among R-OSA, predicted-and-not-reported OSA (PnR-OSA), and No-OSA. Binary logistic regression models of PnR-OSA with sociodemographic and clinical variables were used to characterize this vulnerable subgroup. The reported prevalence of OSA was 14.4% (n=566). R-OSA patients were more likely to have a college education, be married, and be functionally independent. The predicted prevalence of OSA was 71.9% (n=2834). R-OSA patients had higher CCS than nR-OSA, whereas P-OSA patients had lower CCS than nP-OSA (p’s<0.0002). R-OSA patients performed better than nR-OSA in speed of processing assessments, whereas P-OSA individuals performed worse than nP-OSA in speed of processing, verbal learning, and working memory (p’s<0.0005). R-OSA had higher UPSA-B and EFB-AF than nR-OSA (p’s<0.0001). P-OSA patients had a lower EFB-AF than those with nP-OSA (p=0. 003). PnR-OSA patients performed worse than both R-OSA and No-OSA on CCS and EFB-AF, and worse than R-OSA on UPSA-B (p’s <0.05). Veterans with PnR-OSA tended to be older, male, smokers, unmarried, and have higher BMI and less education that the rest of the sample. Our analyses suggest only 20% of OSA in schizophrenia is diagnosed. Self-reported OSA was associated with better performance on cognitive and functional measures, whereas predicted OSA was associated with worse performance on these measures. People with higher cognitive capacity may be more likely to seek medical care, while those with less cognitive capacity are at greater risk for having co-occurring medical conditions that further compromise cognition. Patients vulnerable to under-diagnosis likely have the most to gain from treatment of their OSA.


This is an Open Access Thesis.

Open Access

This Article is Open Access