Dynamic ElecTronic hEalth reCord deTection (DETECT) of individuals at risk of a first episode of psychosis: a case-control development and validation study

Abstract

Background Many individuals who will experience a first episode of psychosis (FEP) are not detected before occurrence, limiting the effect of preventive interventions. The combination of machine-learning methods and electronic health records (EHRs) could help address this gap.Methods This case-control development and validation study is based on EHR data from IBM Explorys. The IBM Explorys Platform holds standardised, longitudinal, de-identified, patient-level EHR data pooled from different health-care systems with distinct EHRs. The present EHR-based studies were retrospective, matched (1:1), casecontrol studies compliant with RECORD, STROBE, and TRIPOD statements. The study included individuals in the IBM Explorys database who at some point between 1990 and 2018 had a diagnosis of FEP followed by schizophrenia, and psychosis-free matched control individuals from a random subsample of the full cohort. For every individual in the FEP cohort, the individual in the control cohort was matched to have a similar date for inclusion in the database and a similar total observation time. Individuals in the FEP cohort had their index date defined as the first diagnosis of psychosis or the first prescription of antipsychotic medication. Individuals in the control cohort had their index date defined to occur the same number of days after inclusion in the database as their matching FEP individual. The FEP and control cohorts were both randomly split into development and validation datasets in a ratio of 7:3. The subset of individuals in the validation dataset who had all their health-care encounters at providers that were not seen in the development dataset made up the external validation subset. A novel recurrent neural network model was developed to predict the risk of FEP 1 year before the index date by employing demographics and medical events (in the categories diagnoses, prescriptions, procedures, encounters and admissions, observations, and laboratory test results) dynamically collected in the EHR as part of clinical routine. We named the recurrent neural network Dynamic ElecTronic hEalth reCord deTection (DETECT). The main outcomes were accuracy and area under receiver operating characteristic curve (AUROC). Decision-curve analyses and dynamic patient journey plots were used to evaluate clinical usefulness. FindingsThe FEP and control cohorts each comprised 72 860 individuals. 102 030 individuals (51 015 matching pairs) were randomly allocated to the development dataset and the remaining 43 690 to the validation dataset. In the validation dataset, 4770 individuals had all their encounters outside of the 118 790 health-care providers that were encountered in the development dataset. The data from these individuals made up the external validation subset. The median follow-up (observation time before index date) was 6•0 years (IQR 3•0-10•4). In the development dataset, DETECT's prognostic accuracy was 0•787 and AUROC was 0•868. In the validation dataset, DETECT's prognostic accuracy was 0•774 and AUROC was 0•856. In the ...

Real-world long-term outcomes in individuals at clinical risk for psychosis: The case for extending duration of care

“…This study further advances knowledge by demonstrating that a possible way of overcoming these challenges may be to leverage EHRs, which are increasingly adopted across several primary and secondary health care systems. EHR not only represent real-word clinical information but can incorporate automatic detection or prognostic algorithms [63] , [64] , [65] , [66] , translating stratified and precision medicine approaches in this field. Overall, these results will inform future studies such as the proposed 26-site ProNET cohort study and global health policies relating to CHR-P service development as well as worldwide clinical guidelines.…”

Early Intervention Services for First Episode of Psychosis in South London and the Maudsley (SLaM): 20 Years of Care and Research for Young People

“…A complementary approach may be to embed precision psychiatry in these refined healthcare systems (31), leveraging risk calculators that automatically screen EHR to detect individual risk of developing FEP (32,33). These algorithms can also dynamically incorporate clinical information as it emerges (34,35), for example access to A&E over follow-up (on average 3.01 over a 6 year period in EI patients, compared to an annualized median of 0.24 in the general population (36)). SLaM EI has recently piloted and implemented these tools in clinical routine practice (37).…”

“…Incidence of Psychosis Within SLaMThe general population aged[16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35] years in Lambeth was 115,751 in 2007 and 128,751 in 2017; in Southwark 105,232 in 2007 and 117,309 in 2017; in Lewisham 88,953 in 2007 and 96,378 in 2017; in Croydon 97,787 in 2007, and 100,612 in 2017 (21), total population of 407,722 in 2007 and 443,050 in 2017. Incidence of psychosis (in 2017) was: Lambeth 71.9 cases per 100,000 personyears, Southwark 69.6 cases per 100,000 person-years, Lewisham 71.3 cases per 100,000 person-years and Croydon 58.3 cases per 100,000 person-years, all substantially greater than the national average of 34.9 cases per 100,000 person-years.…”

Is large-scale population screening coming to psychiatry?

“…Using a 1:1 case-control design on one of the largest real-world databases of electronic health records (EHRs), covering approximately 15% of the US population in both primary and secondary care, Lars Lau Raket and colleagues 4 developed a risk prediction model to detect individuals at risk for developing a first episode of psychosis. The Dynamic ElecTronic hEalth reCord deTection (DETECT) model was able to identify individuals with first episode of psychosis who later developed schizophrenia with a prognostic accuracy of 0•774 and an area under the receiver operating characteristics (AUROC) of 0•856 in the validation dataset, and with a prognostic accuracy of 0•724 and AUROC of 0•799 in the external validation subset.…”