Genetic Data From Medical Records May Predict Neuropsychiatric Conditions

Phenotyping patients using data from electronic medical records may result in viable classification of serious mental illnesses, according to study findings published in JAMA Psychiatry.

Data for this study were sourced from the Veterans Affairs (VA) hospitals. Genetic data included in electronic medical records were used to evaluate the polygenic risk scores (PRSs) for schizophrenia, bipolar disorder (BD), and depression. An embedded cohort (the Cooperative Studies Program [CSP] #572) of patients diagnosed with schizophrenia or BD I by clinicians was used to evaluate the accuracy of automated categorization after accounting for ancestry and patient characteristics.

The study population comprised 314,909 White individuals (mean age, 66.4 years) and 84,806 Black individuals (mean age, 58.0 years). Additionally, 697,749 individuals were enrolled in the Million Veterans Program (MVP; 9.0% women; median age, 61.0 years), and 3953 individuals with schizophrenia (7.3% women; median age, 56.0 years) and 5425 individuals with BD I (18.5% women; median age, 53.0 years) were enrolled in the CSP #572 cohort.

Among the CSP #572 cohort, 95.6% were correctly assigned to phecodes using PRS data. Misclassifications occurred among 2.8% of the schizophrenia cohort and 9.3% of the BD I cohort.

Misclassified BD I was observed more among Black individuals (odds ratio [OR], 2.24; 95% CI, 1.83-2.74; P <10^-14), men (OR, 1.60; 95% CI, 1.22-2.11; P <.01), and patients with lower University of California, San Diego, Performance-Based Skills Assessment scores (β, -0.09; 95% CI, -0.13 to -0.06; P <10^-6).

Among the general VA population, the rate of schizophrenia in the top decile of the disease PRS distribution was observed among 4% of White individuals (95% CI, 4.22-5.43; P <10^-131). In the top decile, individuals were 2.4-fold more likely to be diagnosed with schizophrenia, 2-fold more likely to be diagnosed with psychosis, 1.6-fold more likely to be diagnosed with BD, and 1.2-fold more likely to be diagnosed with major depression compared with those below the 90^th percentile, respectively. Similar trends were observed for BD and major depression.

Among Black individuals, risk for schizophrenia was 1.4-fold higher, risk for BD was 1.6-fold higher, and risk for major depression was 1.3-fold higher compared with the 90^th percentile of the schizophrenia PRS distribution, respectively.

Researchers observed a trend that individuals with more chronic illness had higher PRSs for neuropsychiatric conditions. Individuals who received inpatient treatment had higher PRSs than those who did not (OR per SD-unit increase, 1.25; 95% CI, 1.18-1.34; P <10^-11). Similar trends were observed for BP (OR per SD-unit increase, 1.14; 95% CI, 1.11-1.19; P <.10^-17) and major depression (OR per SD-unit increase, 1.04; 95% CI, 1.03-1.06; P <10^-8).

Evidence of pleiotropy was observed between schizophrenia and diabetes (P =7.07×10^-5), hearing loss (P =.000313), and osteoarthritis (P =.00875).

The study may have been limited by the exclusion of environmental predictors.

Study authors concluded, “Application of current neuropsychiatric PRSs to the MVP yielded results consistent with multiple, continuous liability distributions underlying schizophrenia, bipolar disorder, and major depression underscoring the advantages of multivariate and transdiagnostic approaches for studying these complex, heterogeneous clinical presentations.”

Disclosure: Multiple authors declared affiliations with industry. Please refer to the original article for a full list of disclosures.