Enhancing identification of nonaffective psychosis in register-based studies

Nationwide registers offer comprehensive, population-level data for monitoring and research, but diagnoses are made for clinical—not research—purposes, which introduces variability in diagnostic accuracy. The Finnish Quality of Psychosis Care Register (FQPCR) aims to evaluate quality of care and outcomes in people treated for schizophrenia and other primary psychotic disorders (nonaffective psychoses, NAP). For quality monitoring and research on early care, reliably identifying first-episode psychosis is essential. Common register-based practice uses a multi-year wash-out period to approximate first episodes, but optimal length and validity vary. Including diagnoses from both specialized psychiatric and primary care captures full care pathways but may reduce diagnostic reliability if first NAP diagnoses originate outside psychiatric services. This study asked: (1) In which situations might register-based NAP diagnoses be unreliable? (2) Can a 14-year wash-out reliably identify first episodes? (3) How do care settings and diagnosis confirmation at discharge relate to diagnostic stability and subsequent treatment? The purpose was to refine an algorithm for identifying NAP in health care registers and to assess whether the first register diagnosis approximates treatment onset.

Prior evidence indicates very early-onset NAP is rare and diagnostically challenging: nonaffective psychotic disorders under age 13 are uncommon and difficult to distinguish from other developmental or mood disorders. A Danish validation study reported low accuracy for very early-onset schizophrenia, especially in outpatient settings. Psychotic symptoms are common in dementia, typically arising from the dementia syndrome rather than new primary NAP, making differentiation important for care pathways. Emergency and early hospitalization diagnoses are known to change frequently by discharge. Historical validation studies of register-based psychosis diagnoses suggest few false positives but potentially lower sensitivity; however, these studies are older, focused on specialized services, and did not include primary care registers where diagnostic recording rates may be incomplete (about 60%). These gaps motivate updated validation and algorithmic refinements that incorporate primary care data and discharge-confirmed diagnoses.

Design and data sources: The study leveraged four linked Finnish nationwide registers using personal identification numbers: (1) Care Register for Health Care (CRHC) and (2) Register of Primary Health Care Visits (hospital care from 1996, specialized outpatient from 1998, primary care from 2011), (3) Register for Reimbursements for Prescription Medicines (all reimbursed purchases), and (4) Register for Special Reimbursement Right (higher reimbursement entitlement for severe mental disorders). Cohort definition: Individuals with NAP codes (ICD-10 F20–F29 or ICPC-2 P72 or P98) during 2010–2020 were identified. Those without a valid Finnish personal identity number were excluded. The analytic cohort focused on people with their first CRHC NAP diagnosis between 2010–2018 and no NAP diagnoses in 1996–2009 (14-year look-back, 12 years for specialized outpatient care), yielding n=49,165 prior to further exclusions derived from analyses. Hospital episodes and diagnosis confirmation: Inpatient entries were linked into continuous hospitalization periods when transfers occurred the same day; only hospitalizations lasting at least one night were counted. Psychiatric hospitalizations were identified via specialty codes (70, 70F, 70X, 70Z, 74, 75). Discharge diagnoses were defined as the last diagnosis of the psychiatric hospitalization. Preliminary diagnoses given at emergency departments or during hospitalization were compared against discharge diagnoses to assess stability. Outcomes and measures: Two-year follow-up (censoring by end of 2020) assessed: (a) receiving a new NAP diagnosis and (b) antipsychotic purchases (ATC N05A excluding lithium). Indicators of treatment prior to the first register NAP diagnosis included special reimbursement right and antipsychotic purchases at least one year before the first diagnosis. Statistical analysis: Logistic regression (R 4.1.3) predicted (1) new NAP diagnoses and (2) antipsychotic purchase within two years from sex, age, first diagnosis group (exclusive categories), treatment setting of the first diagnosis (e.g., psychiatric outpatient, primary care, nonpsychiatric hospitalization, inpatient preliminary vs discharge-confirmed), and hospital district (Helsinki vs other HUS vs other districts). Ethics and permissions: Data use was permitted by the Finnish Institute for Health and Welfare; separate ethical review or informed consent was not required for register-only research.

Cohort composition and first-diagnosis settings: Among people with first CRHC NAP diagnosis in 2010–2018 (n=49,165):

• First diagnosis settings: primary care 10,106 (21%); specialized psychiatric outpatient 10,973 (22%); other specialized health care 1,321 (3%); hospital discharge/during hospitalization/one day before 26,764 (54%). Age and prior diagnoses at first NAP diagnosis:
• Mean age 47 years; distribution right-skewed.
• 6% (2,757) before age 18; 0.8% (391) before age 13; 19 before age 7. Only a few under 7 had a psychiatric discharge diagnosis of NAP, and 5 (26%) had another NAP diagnosis within two years.
• 33% (16,222) received their first register NAP diagnosis at age ≥60.
• Prior to first NAP diagnosis, 64% (31,361) had any psychiatric diagnosis recorded. Common prior diagnoses: nonpsychotic depression 15,941 (32%); anxiety disorders 14,656 (30%); bipolar disorder 3,416 (7%); depressive disorder with psychotic symptoms 3,459 (7%); substance-induced psychosis 2,587 (5%). Dementia prior to first NAP diagnosis occurred in 2,278 (5%), rising to ~25% among those aged 80–90. Stability of preliminary inpatient diagnoses:
• First NAP diagnosis at emergency setting one day before/on psychiatric hospital admission (n=11,708): 61% had NAP as discharge diagnosis; if not NAP at discharge, common discharge diagnoses were substance-induced psychosis, other substance use, bipolar disorder, nonpsychotic depression, and anxiety disorders.
• First NAP diagnosis during hospitalization or one day before admission (excluding emergency; n=4,821): 63% had NAP as discharge diagnosis. Two-year outcomes by first-diagnosis setting:
• Probability of receiving another NAP diagnosis: discharge-confirmed NAP 74%; preliminary (not discharge-confirmed) 28%; nonpsychiatric hospitalization 32%; specialized psychiatric outpatient 79%; specialized nonpsychiatric outpatient 48%; primary care 56%.
• Antipsychotic purchases: discharge-confirmed NAP 88%; preliminary (not discharge-confirmed) 75%; nonpsychiatric hospitalization 72%; specialized psychiatric outpatient 79%; specialized nonpsychiatric outpatient 64%; primary care 73%. Regression results:
• Lower odds of a new NAP diagnosis for: preliminary inpatient (not discharge-confirmed) vs psychiatric outpatient (OR 0.11, 95% CI 0.10–0.11); first diagnosis from nonpsychiatric hospital care (OR 0.20, 95% CI 0.18–0.21); first diagnosis from primary care (OR 0.41, 95% CI 0.38–0.44). Older age also decreased probability, with high 2-year mortality in ages 80–90 (23%) and >90 (41%).
• Antipsychotic purchases: highest odds for discharge-confirmed inpatient group (OR 2.05, 95% CI 1.91–2.20). Lower odds for first diagnosis schizotypal disorder (OR 0.51, 95% CI 0.45–0.58) and ICPC-2-based diagnosis (OR 0.58, 95% CI 0.52–0.64). Exclusion criteria and effects:
• Exclusions determined: age <7 at first NAP diagnosis; prior dementia diagnosis; preliminary inpatient NAP diagnoses not confirmed by discharge diagnosis.
• Counts excluded: age <7 (n=19); prior dementia (n=2,278); preliminary inpatient not confirmed (n=5,836). Final sample: 41,032 (from 49,165). Mean age decreased from 47 to 44 years. Probability of a new NAP diagnosis within two years increased from 61% to 71%. Diagnostic distribution and sex proportions were largely unchanged. Indicators of treatment prior to first NAP diagnosis:
• 20% (9,810) had special reimbursement right before the first NAP diagnosis.
• 39% (15,838) had purchased antipsychotics at least one year before the first register diagnosis; half of these had a special reimbursement right, half did not, suggesting symptomatic treatment prior to formal diagnosis.
• Prior treatment markers were most common when the first diagnosis was schizophrenia or schizoaffective disorder and in ages 50–80, indicating that a 14-year wash-out may not always capture first-episode cases, particularly among older patients.

The study addressed the reliability of identifying NAP using health registers and the validity of approximating first episodes with a 14-year wash-out. Three situations produced unreliable NAP classification: diagnoses before age 7, NAP recorded after a prior dementia diagnosis, and preliminary inpatient NAP diagnoses not confirmed at discharge. Excluding these groups reduced the cohort size but improved the likelihood of subsequent NAP diagnoses and antipsychotic use, indicating greater diagnostic stability and cohort validity. Findings also showed that a substantial fraction had evidence of antipsychotic treatment or reimbursement before the first register NAP diagnosis, particularly among older adults and those first diagnosed with schizophrenia or schizoaffective disorder. Therefore, the first register diagnosis, even after a long wash-out, may not indicate the first treatment episode, especially at older ages. The results support prioritizing discharge-confirmed diagnoses for inpatient episodes and using all healthcare sectors (specialized and primary care) to capture comprehensive pathways, while recognizing potential lower diagnostic reliability in primary care. These refinements have direct implications for quality monitoring and epidemiologic research on psychosis.

Three exclusion criteria enhance the identification of nonaffective psychosis in register-based studies: (1) first diagnosis before age 7, (2) prior dementia diagnosis, and (3) preliminary inpatient NAP diagnoses not confirmed by discharge. Applying these criteria yields a younger cohort with higher diagnostic stability (greater probability of subsequent NAP diagnoses) without materially altering diagnostic distribution or sex balance. A 14-year wash-out does not always ensure capture of first-episode cases, particularly for older adults and those with first diagnoses of schizophrenia or schizoaffective disorder. For defining first-episode cohorts, limit to age ranges with complete prior diagnostic coverage and consider incorporating prior antipsychotic use and reimbursement rights to approximate treatment onset. Utilizing data from both specialized and primary care provides the most complete picture of NAP care. Future research should include updated validation studies of psychosis diagnoses in primary care registers and continued refinement of algorithms that integrate discharge-confirmed diagnoses and medication-based indicators of illness onset.

Only register variables were available; diagnostic recording in primary care general practitioner visits is incomplete (~60%). Data from private and occupational health care were unavailable (though private psychiatric hospital care is rare in Finland). Sensitivity of register-based NAP identification may be limited, and the accuracy of early childhood and emergency or preliminary inpatient diagnoses is variable. Mortality in older age groups affects follow-up ascertainment. Despite these constraints, national coverage and minimal attrition except emigration are strengths.