Predicting treatment resistance from first-episode psychosis using routinely collected clinical information

Leveraging deep neural network and language models for predicting long-term hospitalization risk in schizophrenia

Early warning of long-term hospitalization in schizophrenia (SCZ) patients at the time of admission is crucial for effective resource allocation and individual treatment planning. In this study, we developed a deep learning model that integrates demographic, behavioral, and blood test data from admission to forecast extended hospital stays using a retrospective cohort. By utilizing language models, our developed algorithm efficiently extracts 95% of the unstructured electronic health records data needed for this work, while ensuring data privacy and low error rate. This paradigm has also been demonstrated to have significant advantages in reducing potential discrimination and erroneous dependencies. By utilizing multimodal features, our deep learning model achieved a classification accuracy of 0.81 and an AUC of 0.9. Key risk factors identified included advanced age, longer disease duration, and blood markers such as elevated neutrophil-to-lymphocyte ratio, lower lymphocyte percentage, and reduced albumin levels, validated through comprehensive interpretability analyses and ablation studies. The inclusion of multimodal data significantly improved prediction performance, with demographic variables alone achieving an accuracy of 0.73, which increased to 0.81 with the addition of behavioral and blood test data. Our approach outperformed traditional machine learning methods, which were less effective in predicting long-term stays. This study demonstrates the potential of integrating diverse data types for enhanced predictive accuracy in mental health care, providing a robust framework for early intervention and personalized treatment in SCZ management.

Multimodal prediction of the need of clozapine in treatment resistant schizophrenia; a pilot study in first-episode psychosis

As many as one third of the patients diagnosed with schizophrenia do not respond to first-line antipsychotic medication. This group may benefit from the atypical antipsychotic medication clozapine, but initiation of treatment is often delayed, which may worsen prognosis. Predicting which patients do not respond to traditional antipsychotic medication at the onset of symptoms would provide fast-tracked treatment for this group of patients. We collected data from patient records of 38 first-episode psychosis patients, of whom seven did not respond to traditional antipsychotic medications. We used clinical data including medical records, voxel-based morphometry MRI data and inter-subject correlation fMRI data, obtained during movie viewing, to predict future treatment resistance. Using a neural network model, we correctly predicted future treatment resistance in six of the seven treatment resistance patients and 25 of 31 patients who did not require clozapine treatment. Prediction improved significantly when using imaging data in tandem with clinical data. The accuracy of the neural network model was significantly higher than the accuracy of a support vector machine algorithm. These results support the notion that treatment resistant schizophrenia could represent a separate entity of psychotic disorders, characterized by morphological and functional changes in the brain which could represent biomarkers detectable at early onset of symptoms.

Predicting treatment resistance in positive and negative symptom domains from first episode psychosis: Development of a clinical prediction model

BACKGROUND

Treatment resistance (TR) in schizophrenia may be defined by the persistence of positive and/or negative symptoms despite adequate treatment. Whilst previous investigations have focused on positive symptoms, negative symptoms are highly prevalent, impactful, and difficult to treat. In the current study we aimed to develop easily employable prediction models to predict TR in positive and negative symptom domains from first episode psychosis (FEP).

METHODS

Longitudinal cohort data from 1027 individuals with FEP was utilised. Using a robust definition of TR, n = 51 (4.97 %) participants were treatment resistant in the positive domain and n = 56 (5.46 %) treatment resistant in the negative domain 12 months after first presentation. 20 predictor variables, selected by existing evidence and availability in clinical practice, were entered into two LASSO regression models. We estimated the models using repeated nested cross-validation (NCV) and assessed performance using discrimination and calibration measures.

RESULTS

The prediction model for TR in the positive domain showed good discrimination (AUC = 0.72). Twelve predictor variables (male gender, cannabis use, age, positive symptom severity, depression and academic and social functioning) were retained by each outer fold of the NCV procedure, indicating importance in prediction of the outcome. However, our negative domain model failed to discriminate those with and without TR, with results only just over chance (AUC = 0.56).

CONCLUSIONS

Treatment resistance of positive symptoms can be accurately predicted from FEP using routinely collected baseline data, however prediction of negative domain-TR remains a challenge. Detailed negative symptom domains, clinical data, and biomarkers should be considered in future longitudinal studies.

White blood cell counts, ratios, and C-reactive protein among individuals with schizophrenia spectrum disorder and associations with long-term outcomes: a population-based study

BACKGROUND

Immune mechanisms are associated with adverse outcomes in schizophrenia; however, the predictive value of various peripheral immune biomarkers has not been collectively investigated in a large cohort before.

OBJECTIVE

To investigate how white blood cell (WBC) counts, ratios, and C-Reactive Protein (CRP) levels influence the long-term outcomes of individuals with schizophrenia spectrum disorder (SSD).

METHODS

We identified all adults in the Central Denmark Region during 1994-2013 with a measurement of WBC counts and/or CRP at first diagnosis of SSD. WBC ratios were calculated, and both WBC counts and ratios were quartile-categorized (Q4 upper quartile). We followed these individuals from first diagnosis until outcome of interest (death, treatment resistance and psychiatric readmissions), emigration or December 31, 2016, using Cox regression analysis to estimate adjusted hazard ratios (aHRs).

RESULTS

Among 6,845 participants, 375 (5.5 %) died, 477 (6.9 %) exhibited treatment resistance, and 1470 (21.5 %) were readmitted during follow-up. Elevated baseline levels of leukocytes, neutrophils, monocytes, LLR, NLR, MLR, and CRP increased the risk of death, whereas higher levels of lymphocytes, platelets, and PLR were associated with lower risk. ROC analysis identified CRP as the strongest predictor for mortality (AUC=0.84). Moreover, elevated levels of leukocytes, neutrophils, monocytes, LLR, NLR and MLR were associated with treatment resistance. Lastly, higher platelet counts decreased the risk of psychiatric readmissions, while elevated LLR increased this risk.

CONCLUSIONS

Elevated levels of WBC counts, ratios, and CRP at the initial diagnosis of SSD are associated with mortality, with CRP demonstrating the highest predictive value. Additionally, certain WBC counts and ratios are associated with treatment resistance and psychiatric readmissions.

Development and initial evaluation of a clinical prediction model for risk of treatment resistance in first-episode psychosis: Schizophrenia Prediction of Resistance to Treatment (SPIRIT)

BACKGROUND

A clinical tool to estimate the risk of treatment-resistant schizophrenia (TRS) in people with first-episode psychosis (FEP) would inform early detection of TRS and overcome the delay of up to 5 years in starting TRS medication.

AIMS

To develop and evaluate a model that could predict the risk of TRS in routine clinical practice.

METHOD

We used data from two UK-based FEP cohorts (GAP and AESOP-10) to develop and internally validate a prognostic model that supports identification of patients at high-risk of TRS soon after FEP diagnosis. Using sociodemographic and clinical predictors, a model for predicting risk of TRS was developed based on penalised logistic regression, with missing data handled using multiple imputation. Internal validation was undertaken via bootstrapping, obtaining optimism-adjusted estimates of the model's performance. Interviews and focus groups with clinicians were conducted to establish clinically relevant risk thresholds and understand the acceptability and perceived utility of the model.

RESULTS

We included seven factors in the prediction model that are predominantly assessed in clinical practice in patients with FEP. The model predicted treatment resistance among the 1081 patients with reasonable accuracy; the model's C-statistic was 0.727 (95% CI 0.723-0.732) prior to shrinkage and 0.687 after adjustment for optimism. Calibration was good (expected/observed ratio: 0.999; calibration-in-the-large: 0.000584) after adjustment for optimism.

CONCLUSIONS

We developed and internally validated a prediction model with reasonably good predictive metrics. Clinicians, patients and carers were involved in the development process. External validation of the tool is needed followed by co-design methodology to support implementation in early intervention services.

Association of symptom severity and cerebrospinal fluid alterations in recent onset psychosis in schizophrenia-spectrum disorders - An individual patient data meta-analysis

Neuroinflammation and blood-cerebrospinal fluid barrier (BCB) disruption could be key elements in schizophrenia-spectrum disorderś(SSDs) etiology and symptom modulation. We present the largest two-stage individual patient data (IPD) meta-analysis, investigating the association of BCB disruption and cerebrospinal fluid (CSF) alterations with symptom severity in first-episode psychosis (FEP) and recent onset psychotic disorder (ROP) individuals, with a focus on sex-related differences. Data was collected from PubMed and EMBASE databases. FEP, ROP and high-risk syndromes for psychosis IPD were included if routine basic CSF-diagnostics were reported. Risk of bias of the included studies was evaluated. Random-effects meta-analyses and mixed-effects linear regression models were employed to assess the impact of BCB alterations on symptom severity. Published (6 studies) and unpublished IPD from n = 531 individuals was included in the analyses. CSF was altered in 38.8 % of individuals. No significant differences in symptom severity were found between individuals with and without CSF alterations (SMD = -0.17, 95 %CI -0.55-0.22, p = 0.341). However, males with elevated CSF/serum albumin ratios or any CSF alteration had significantly higher positive symptom scores than those without alterations (SMD = 0.34, 95 %CI 0.05-0.64, p = 0.037 and SMD = 0.29, 95 %CI 0.17-0.41p = 0.005, respectively). Mixed-effects and simple regression models showed no association (p > 0.1) between CSF parameters and symptomatic outcomes. No interaction between sex and CSF parameters was found (p > 0.1). BCB disruption appears highly prevalent in early psychosis and could be involved in positive symptomś severity in males, indicating potential difficult-to-treat states. This work highlights the need for considering BCB breakdownand sex-related differences in SSDs clinical trials and treatment strategies.

Clozapine use in the first two years after first-episode psychosis in a real-world clinical sample

BACKGROUND

Approximately 20-30% of patients with schizophrenia fail to respond to antipsychotic treatment and are considered treatment resistant (TR). Although clozapine is the treatment of choice in these patients, in real-world clinical settings, clinicians often delay clozapine initiation, especially in first-episode psychosis (FEP).

AIM

The main aim of this study was to describe prescription patterns for clozapine in a sample of patients diagnosed with FEP and receiving specialized treatment at a university hospital. More specifically, we aimed to determine the following: (1) the proportion of patients who received clozapine within two years of disease onset, (2) baseline predictors of clozapine use, (3) time from starting the first antipsychotic to clozapine initiation, (4) concomitant medications, and (5) clozapine-related adverse effects.

METHODS

All patients admitted to a specialized FEP treatment unit at our hospital between April 2013 and July 2020 were included and followed for two years. The following variables were assessed: baseline sociodemographic characteristics; medications prescribed during follow-up; clozapine-related adverse effects; and baseline predictors of clozapine use. We classified the sample into three groups: clozapine users, clozapine-eligible, and non-treatment resistant (TR).

RESULTS

A total of 255 patients were consecutively included. Of these, 20 (7.8%) received clozapine, 57 (22.4%) were clozapine-eligible, and 178 (69.8%) were non-TR. The only significant variable associated with clozapine use at baseline was the Global Assessment of Functioning (GAF) score (R2=0.09, B=-0.07; OR=0.94; 95% CI: 0.88-0.99; p=0.019). The median time to clozapine initiation was 55.0 (93.3) days. The most common side effect was sedation.

CONCLUSIONS

A significant proportion (30.2%) of patients in this cohort were treatment resistant and eligible for clozapine. However, only 7.8% of the sample received clozapine, indicating that this medication was underprescribed. A lower baseline GAF score was associated with clozapine use within two years, suggesting that it could be used to facilitate the early identification of patients who will need treatment with clozapine, which could in turn improve treatment outcomes.

[Management of treatment resistance-Treatment-resistant schizophrenia]

Despite a very high prevalence and substantial impairments among affected individuals, treatment-resistant schizophrenia (TRS) has not been sufficiently researched in clinical research in the field of psychiatric disorders and the pathophysiology is still poorly understood. A better clinical and pathophysiological understanding of this heterogeneous and severely affected population of people with persistent symptoms in different domains is necessary in order not only to be able to intervene early but also to develop novel therapeutic strategies or individualized treatment approaches. This review article presents the state of the art criteria of the pharmacological TRS, neurobiological disease models and predictive factors for TRS as well as the phenomenon of pseudo-treatment resistance and the clinical management of TRS. In the future, not only the use of operationalized criteria and definitions of TRS in longitudinal studies and randomized-controlled trials (RCTs) are paramount, but also the observation of trajectories with the integration of multimodal longitudinal phenotyping and the longitudinal collection of clinical routine data in academic research, which will be possible in the newly created German Center for Mental Health (DZPG).

Inflammatory blood cells and ratios at remission for psychosis relapse prediction: A three-year follow-up of a cohort of first episodes of schizophrenia

BACKGROUND

The clinical course following a first episode of schizophrenia (FES) is often characterized by recurrent relapses, resulting in unfavorable clinical and functional outcomes. Inflammatory dysregulation has been implicated in relapse risk; however, the predictive value of inflammatory blood cells in clinically remitted patients after a FES has not been previously explored.

METHODS

In this study, we closely monitored 111 patients in remission after a FES until relapse or a three-year follow-up endpoint. The participants were recruited from the multicenter 2EPS Project. Data on inflammatory blood cells and ratios were collected at baseline and at the time of relapse or after three years of follow-up.

RESULTS

Monocyte counts (OR = 1.91; 95 % CI = 1.07-3.18; p = 0.009) and basophil counts (OR = 1.09; 95 % CI = 1.01-1.12; p = 0.005) at baseline were associated with an increased risk of relapse, while the platelet-lymphocyte ratio (OR = 0.98; 95 % CI = 0.97-0.99; p = 0.019) was identified as a protective factor. However, after adjusting for cannabis and tobacco use during the follow-up, only monocyte counts (OR = 1.73; 95 % CI = 1.03-2.29; p = 0.027) and basophil counts (OR = 1.08; 95 % CI = 1.01-1.14; p = 0.008) remained statistically significant. ROC curve analysis indicated that the optimal cut-off values for discriminating relapsers were 0.52 × 10^9/L (AUC: 0.66) for monocytes and 0.025 × 10^9/L (AUC: 0.75) for basophils. When considering baseline inflammatory levels, no significant differences were observed in the inflammatory biomarkers at the endpoint between relapsers and non-relapsers.

CONCLUSION

This study provides evidence that higher monocyte and basophil counts measured at remission after a FES are associated with an increased risk of relapse during a three-year follow-up period.

Development of an individualized risk calculator of treatment resistance in patients with first-episode psychosis (TRipCal) using automated machine learning: a 12-year follow-up study with clozapine prescription as a proxy indicator

About 15-40% of patients with schizophrenia are treatment resistance (TR) and require clozapine. Identifying individuals who have higher risk of development of TR early in the course of illness is important to provide personalized intervention. A total of 1400 patients with FEP enrolled in the early intervention for psychosis service or receiving the standard psychiatric service between July 1, 1998, and June 30, 2003, for the first time were included. Clozapine prescriptions until June 2015, as a proxy of TR, were obtained. Premorbid information, baseline characteristics, and monthly clinical information were retrieved systematically from the electronic clinical management system (CMS). Training and testing samples were established with random subsampling. An automated machine learning (autoML) approach was used to optimize the ML algorithm and hyperparameters selection to establish four probabilistic classification models (baseline, 12-month, 24-month, and 36-month information) of TR development. This study found 191 FEP patients (13.7%) who had ever been prescribed clozapine over the follow-up periods. The ML pipelines identified with autoML had an area under the receiver operating characteristic curve ranging from 0.676 (baseline information) to 0.774 (36-month information) in predicting future TR. Features of baseline information, including schizophrenia diagnosis and age of onset, and longitudinal clinical information including symptoms variability, relapse, and use of antipsychotics and anticholinergic medications were important predictors and were included in the risk calculator. The risk calculator for future TR development in FEP patients (TRipCal) developed in this study could support the continuous development of data-driven clinical tools to assist personalized interventions to prevent or postpone TR development in the early course of illness and reduce delay in clozapine initiation.

The association between clinical, sociodemographic, familial, and environmental factors and treatment resistance in schizophrenia: A machine-learning-based approach

BACKGROUND

Prediction of treatment resistance in schizophrenia (TRS) would be helpful to reduce the duration of ineffective treatment and avoid delays in clozapine initiation. We applied machine learning to identify clinical, sociodemographic, familial, and environmental variables that are associated with TRS and could potentially predict TRS in the future.

STUDY DESIGN

Baseline and follow-up data on trait(-like) variables from the Genetic Risk and Outcome of Psychosis (GROUP) study were used. For the main analysis, we selected patients with non-affective psychotic disorders who met TRS (n = 200) or antipsychotic-responsive criteria (n = 423) throughout the study. For a sensitivity analysis, we only selected patients who met TRS (n = 76) or antipsychotic-responsive criteria (n = 123) at follow-up but not at baseline. Random forest models were trained to predict TRS in both datasets. SHapley Additive exPlanation values were used to examine the variables' contributions to the prediction.

STUDY RESULTS

Premorbid functioning, age at onset, and educational degree were most consistently associated with TRS across both analyses. Marital status, current household, intelligence quotient, number of moves, and family loading score for substance abuse also consistently contributed to the prediction of TRS in the main or sensitivity analysis. The diagnostic performance of our models was modest (area under the curve: 0.66-0.69).

CONCLUSIONS

We demonstrate that various clinical, sociodemographic, familial, and environmental variables are associated with TRS. Our models only showed modest performance in predicting TRS. Prospective large multi-centre studies are needed to validate our findings and investigate whether the model's performance can be improved by adding data from different modalities.