Researchers investigated population-level phenotypes occurring before and after diagnosis of schizophrenia using a large health insurance claims dataset.
“Mrs Blue” is a 70-year-old Caucasian female with a 45-year history of schizophrenia. She also has post-traumatic stress disorder (PTSD) stemming from a sexual assault in early adulthood and occasionally smokes marijuana. Her current psychotropic regimen includes aripiprazole, valproic acid, and citalopram, and she is overall stable with no psychiatric hospitalizations in the past 6 years. Prior to her schizophrenia diagnosis, she underwent a hysterectomy at age 20 for ovarian cancer. She has comorbid obesity, hypertension, hyperlipidemia, and osteoporosis. Three years ago, she was diagnosed with chronic renal failure. Her most recent BUN/Cr are 51/2.15, with a GFR of 26.
Schizophrenia affects more than 3 million individuals in the United States, and it is associated with substantial psychiatric, substance-related, and medical comorbidity.1 However, patterns of comorbidity in schizophrenia have not been systematically described using real-word data. In order to address this gap, Lu and colleagues leveraged a nationwide health insurance claims dataset from Aetna to assess population-level phenotypes associated with schizophrenia.2
The Current Study
Study authors conducted data-driven analyses on patient phenotypes before and after the index diagnosis of schizophrenia and compared data on phenotypes with matched controls without schizophrenia. The Aetna dataset contained 86 million participants from all 50 states, the District of Columbia, 5 populated territories, and Armed Forces Europe, with claims data from January 2008 to December 2019. ICD-9 and 10 codes form the basis of corresponding phenotype codes or phecodes. Schizophrenia was defined as at least 3 occurrences of phecodes 295 and 295.1. Inclusion criteria were documented birth year, biological sex, and zip code information. Participants also had to have been on the insurance plan for at least 12 months prior to their first schizophrenia diagnosis. Patients aged <15 years were excluded. Each patient in the schizophrenia cohort was matched to a non-schizophrenia control based on birth year, biological sex, and the first 3 digits of the zip code.
The authors analyzed 1890 different phenotypes based on phecodes (requiring at least 3 occurrences) before diagnosis for patients with a first documentation of schizophrenia between ages 15 and 29. Data for schizophrenia and controls were analyzed using Fisher’s exact test. Data on phenotypes occurring after the onset of schizophrenia were analyzed using Cox proportional hazard models. Patients were removed from these analyses if they had the phecode within 1 year of the first schizophrenia diagnosis. P-values were Bonferroni-corrected for multiple comparisons.
The authors identified 61,453 patients with schizophrenia (prevalence 0.7%), of whom 47% were male, and the average length of patient records was 6 years. They noted that 144 phecodes were significantly enriched in patients later diagnosed with schizophrenia. Some of the top phecodes were suicidal ideation (OR=51.2), bipolar (OR=36.1), antisocial/borderline personality (OR=18.4), substance addiction and disorders (OR=17.6), and PTSD (OR=16.2). Women who later developed schizophrenia were more likely to have type 2 diabetes, obstructive sleep apnea, and eating disorders; men were more likely to have acute renal failure, developmental delay, and rhabdomyolysis. The authors also noted that 402 phecodes were significantly enriched after the diagnosis of schizophrenia. Some of the top phecodes were coma/stupor/brain damage (HR=87.3), altered mental status (HR=86.2), alcoholism (HR=47.3), and septicemia (HR=30.4). Women were more likely to develop encephalopathy, epilepsy, and adverse drug events/allergies; men were more likely to develop impulse control disorder, esophageal bleeding, and acute osteomyelitis.
The authors concluded in these large-scale systematic analyses that there was a broad range of comorbidities enhanced in schizophrenia both before and after diagnosis, including PTSD, other anxiety disorders, and alcohol and other substance use disorders. There was a significant influence of sex on associated comorbidities. Study strengths include the large sample size and comprehensive data, consideration of comorbidities before and after the diagnosis of schizophrenia, and statistical correction for multiple comparisons. Study limitations include the potential lack of generalizability to other schizophrenia populations (as the present study was an employer-sponsored insurance claims dataset), the relatively short follow-up time, and the inherent nature of phecodes (which represent coarse categories of clinical phenotypes).
The Bottom Line
This study demonstrates the utility of large-scale, real-world data analyses to elucidate patterns of disease comorbidity, which can provide clinically useful information to treating clinicians.
Dr. Miller is a professor in the Department of Psychiatry and Health Behavior at Augusta University in Augusta, Georgia. He is on the Editorial Board and serves as the schizophrenia section chief for Psychiatric TimesTM. The author reports that he receives research support from Augusta University, the National Institute of Mental Health, and the Stanley Medical Research Institute.
1. Patel KR, Cherian J, Gohil K, Atkinson D. Schizophrenia: overview and treatment options. P T. 2014;39(9):638-645.
2. Lu C, Jin D, Palmer N, et al. Large-scale real-world data analysis identifies comorbidity patterns in schizophrenia. Transl Psychiatry. 2022;12(1):154.