Attempts to use artificial intelligence (AI) in psychiatric disorders show moderate success, highlighting the potential of incorporating information from clinical assessments to improve the models. This study focuses on using large language models (LLMs) to detect suicide risk from medical text in psychiatric care.

To extract information about suicidality status from the admission notes in electronic health records (EHRs) using privacy-sensitive, locally hosted LLMs, specifically evaluating the efficacy of Llama-2 models.

We compared the performance of several variants of the open source LLM Llama-2 in extracting suicidality status from 100 psychiatric reports against a ground truth defined by human experts, assessing accuracy, sensitivity, specificity and F1 score across different prompting strategies.

A German fine-tuned Llama-2 model showed the highest accuracy (87.5%), sensitivity (83.0%) and specificity (91.8%) in identifying suicidality, with significant improvements in sensitivity and specificity across various prompt designs.

The study demonstrates the capability of LLMs, particularly Llama-2, in accurately extracting information on suicidality from psychiatric records while preserving data privacy. This suggests their application in surveillance systems for psychiatric emergencies and improving the clinical management of suicidality by improving systematic quality control and research.

Social determinants of health (SDoH), such as socioeconomics and neighborhoods, strongly influence health outcomes. However, the current state of standardized SDoH data in electronic health records (EHRs) is lacking, a significant barrier to research and care quality.

We conducted a PubMed search using “SDOH” and “EHR” Medical Subject Headings terms, analyzing included articles across five domains: 1) SDoH screening and assessment approaches, 2) SDoH data collection and documentation, 3) Use of natural language processing (NLP) for extracting SDoH, 4) SDoH data and health outcomes, and 5) SDoH-driven interventions.

Of 685 articles identified, 324 underwent full review. Key findings include implementation of tailored screening instruments, census and claims data linkage for contextual SDoH profiles, NLP systems extracting SDoH from notes, associations between SDoH and healthcare utilization and chronic disease control, and integrated care management programs. However, variability across data sources, tools, and outcomes underscores the need for standardization.

Despite progress in identifying patient social needs, further development of standards, predictive models, and coordinated interventions is critical for SDoH-EHR integration. Additional database searches could strengthen this scoping review. Ultimately, widespread capture, analysis, and translation of multidimensional SDoH data into clinical care is essential for promoting health equity.

Society of Thoracic Surgeons Congenital Heart Surgery Database is the largest congenital heart surgery database worldwide but does not provide information beyond primary episode of care. Linkage to hospital electronic health records would capture complications and comorbidities along with long-term outcomes for patients with CHD surgeries. The current study explores linkage success between Society of Thoracic Surgeons Congenital Heart Surgery Database and electronic health record data in North Carolina and Georgia.

The Society of Thoracic Surgeons Congenital Heart Surgery Database was linked to hospital electronic health records from four North Carolina congenital heart surgery using indirect identifiers like date of birth, sex, admission, and discharge dates, from 2008 to 2013. Indirect linkage was performed at the admissions level and compared to two other linkages using a “direct identifier,” medical record number: (1) linkage between Society of Thoracic Surgeons Congenital Heart Surgery Database and electronic health records from a subset of patients from one North Carolina institution and (2) linkage between Society of Thoracic Surgeons data from two Georgia facilities and Georgia’s CHD repository, which also uses direct identifiers for linkage.

Indirect identifiers successfully linked 79% (3692/4685) of Society of Thoracic Surgeons Congenital Heart Surgery Database admissions across four North Carolina hospitals. Direct linkage techniques successfully matched Society of Thoracic Surgeons Congenital Heart Surgery Database to 90.2% of electronic health records from the North Carolina subsample. Linkage between Society of Thoracic Surgeons and Georgia’s CHD repository was 99.5% (7,544/7,585).

Linkage methodology was successfully demonstrated between surgical data and hospital-based electronic health records in North Carolina and Georgia, uniting granular procedural details with clinical, developmental, and economic data. Indirect identifiers linked most patients, consistent with similar linkages in adult populations. Future directions include applying these linkage techniques with other data sources and exploring long-term outcomes in linked populations.

The expansion of electronic health record (EHR) data networks over the last two decades has significantly improved the accessibility and processes around data sharing. However, there lies a gap in meeting the needs of Clinical and Translational Science Award (CTSA) hubs, particularly related to real-world data (RWD) and real-world evidence (RWE).

We adopted a mixed-methods approach to construct a comprehensive needs assessment that included: (1) A Landscape Context analysis to understand the competitive environment; and (2) Customer Discovery to identify stakeholders and the value proposition related to EHR data networks. Methods included surveys, interviews, and a focus group.

Thirty-two CTSA institutions contributed data for analysis. Fifty-four interviews and one focus group were conducted. The synthesis of our findings pivots around five emergent themes: (1) CTSA segmentation needs vary according to resources; (2) Team science is key for success; (3) Quality of data generates trust in the network; (4) Capacity building is defined differently by researcher career stage and CTSA existing resources; and (5) Researchers’ unmet needs.

Based on the results, EHR data networks like ENACT that would like to meet the expectations of academic research centers within the CTSA consortium need to consider filling the gaps identified by our study: foster team science, improve workforce capacity, achieve data governance trust and efficiency of operation, and aid Learning Health Systems with validating, applying, and scaling the evidence to support quality improvement and high-value care. These findings align with the NIH NCATS Strategic Plan for Data Science.

The progression of long-term diabetes complications has led to a decreased quality of life. Our objective was to evaluate the adverse outcomes associated with diabetes based on a patient’s clinical profile by utilizing a multistate modeling approach.

This was a retrospective study of diabetes patients seen in primary care practices from 2013 to 2017. We implemented a five-state model to examine the progression of patients transitioning from one complication to having multiple complications. Our model incorporated high dimensional covariates from multisource data to investigate the possible effects of different types of factors that are associated with the progression of diabetes.

The cohort consisted of 10,596 patients diagnosed with diabetes and no previous complications associated with the disease. Most of the patients in our study were female, White, and had type 2 diabetes. During our study period, 5928 did not develop complications, 3323 developed microvascular complications, 1313 developed macrovascular complications, and 1129 developed both micro- and macrovascular complications. From our model, we determined that patients had a 0.1334 [0.1284, .1386] rate of developing a microvascular complication compared to 0.0508 [0.0479, .0540] rate of developing a macrovascular complication. The area deprivation index score we incorporated as a proxy for socioeconomic information indicated that patients who reside in more disadvantaged areas have a higher rate of developing a complication compared to those who reside in least disadvantaged areas.

Our work demonstrates how a multistate modeling framework is a comprehensive approach to analyzing the progression of long-term complications associated with diabetes.

Concern that self-harm and mental health conditions are increasing in university students may reflect widening access to higher education, existing population trends and/or stressors associated with this setting.

To compare population-level data on self-harm, neurodevelopmental and mental health conditions between university students and non-students with similar characteristics before and during enrolment.

This cohort study linked electronic records from the Higher Education Statistics Agency for 2012–2018 to primary and secondary healthcare records. Students were undergraduates aged 18 to 24 years at university entry. Non-students were pseudo-randomly selected based on an equivalent age distribution. Logistic regressions were used to calculate odds ratios. Poisson regressions were used to calculate incidence rate ratios (IRR).

The study included 96 760 students and 151 795 non-students. Being male, self-harm and mental health conditions recorded before university entry, and higher deprivation levels, resulted in lower odds of becoming a student and higher odds of drop-out from university. IRRs for self-harm, depression, anxiety, autism spectrum disorder (ASD), drug use and schizophrenia were lower for students. IRRs for self-harm, depression, attention-deficit hyperactivity disorder, ASD, alcohol use and schizophrenia increased more in students than in non-students over time. Older students experienced greater risk of self-harm and mental health conditions, whereas younger students were more at risk of alcohol use than non-student counterparts.

Mental health conditions in students are common and diverse. While at university, students require person-centred stepped care, integrated with local third-sector and healthcare services to address specific conditions.

The serotonin 4 receptor (5-HT4R) is a promising target for the treatment of depression. Highly selective 5-HT4R agonists, such as prucalopride, have antidepressant-like and procognitive effects in preclinical models, but their clinical effects are not yet established.

To determine whether prucalopride (a 5-HT4R agonist and licensed treatment for constipation) is associated with reduced incidence of depression in individuals with no past history of mental illness, compared with anti-constipation agents with no effect on the central nervous system.

Using anonymised routinely collected data from a large-scale USA electronic health records network, we conducted an emulated target trial comparing depression incidence over 1 year in individuals without prior diagnoses of major mental illness, who initiated treatment with prucalopride versus two alternative anti-constipation agents that act by different mechanisms (linaclotide and lubiprostone). Cohorts were matched for 121 covariates capturing sociodemographic factors, and historical and/or concurrent comorbidities and medications. The primary outcome was a first diagnosis of major depressive disorder (ICD-10 code F32) within 1 year of the index date. Robustness of the results to changes in model and population specification was tested. Secondary outcomes included a first diagnosis of six other neuropsychiatric disorders.

Treatment with prucalopride was associated with significantly lower incidence of depression in the following year compared with linaclotide (hazard ratio 0.87, 95% CI 0.76–0.99; P = 0.038; n = 8572 in each matched cohort) and lubiprostone (hazard ratio 0.79, 95% CI 0.69–0.91; P < 0.001; n = 8281). Significantly lower risks of all mood disorders and psychosis were also observed. Results were similar across robustness analyses.

These findings support preclinical data and suggest a role for 5-HT4R agonists as novel agents in the prevention of major depression. These findings should stimulate randomised controlled trials to confirm if these agents can serve as a novel class of antidepressant within a clinical setting.

This study serves as an exemplar to demonstrate the scalability of a research approach using survival analysis applied to general practice electronic health record data from multiple sites. Collection of these data, the subsequent analysis, and the preparation of practice-specific reports were performed using a bespoke distributed data collection and analysis software tool.

Statins are a very commonly prescribed medication, yet there is a paucity of evidence for their benefits in older patients. We examine the relationship between statin prescriptions for general practice patients over 75 and all-cause mortality.

We carried out a retrospective cohort study using survival analysis applied to data extracted from the electronic health records of five Australian general practices.

The data from 8025 patients were analysed. The median duration of follow-up was 6.48 years. Overall, 52 015 patient-years of data were examined, and the outcome of death from any cause was measured in 1657 patients (21%), with the remainder being censored. Adjusted all-cause mortality was similar for participants not prescribed statins versus those who were (HR 1.05, 95% CI 0.92–1.20, P = 0.46), except for patients with diabetes for whom all-cause mortality was increased (HR = 1.29, 95% CI: 1.00–1.68, P = 0.05). In contrast, adjusted all-cause mortality was significantly lower for patients deprescribed statins compared to those who were prescribed statins (HR 0.81, 95% CI 0.70–0.93, P < 0.001), including among females (HR = 0.75, 95% CI: 0.61–0.91, P < 0.001) and participants treated for secondary prevention (HR = 0.72, 95% CI: 0.60–0.86, P < 0.001). This study demonstrated the scalability of a research approach using survival analysis applied to general practice electronic health record data from multiple sites. We found no evidence of increased mortality due to statin-deprescribing decisions in primary care.

People under the care of mental health services are at increased risk of suicide. Existing studies are small in scale and lack comparisons.

To identify opportunities for suicide prevention and underpinning data enhancement in people with recent contact with mental health services.

This population-based study includes people who died by suicide in the year following a mental health services contact in Wales, 2001–2015 (cases), paired with similar patients who did not die by suicide (controls). We linked the National Confidential Inquiry into Suicide and Safety in Mental Health and the Suicide Information Database – Cymru with primary and secondary healthcare records. We present results of conditional logistic regression.

We matched 1031 cases with 5155 controls. In the year before their death, 98.3% of cases were in contact with healthcare services, and 28.5% presented with self-harm. Cases had more emergency department contacts (odds ratio 2.4, 95% CI 2.1–2.7) and emergency hospital admissions (odds ratio 1.5, 95% CI 1.4–1.7), but fewer primary care contacts (odds ratio 0.7, 95% CI 0.6–0.9) and out-patient appointments (odds ratio 0.2, 95% CI 0.2–0.3) than controls. Odds ratios were larger in females than males for injury and poisoning (odds ratio: 3.3 (95% CI 2.5–4.5) v. 2.6 (95% CI 2.1–3.1)).

We may be missing existing opportunities to intervene, particularly in emergency departments and hospital admissions with self-harm presentations and with unattributed self-harm, especially in females. Prevention efforts should focus on strengthening routine care contacts, responding to emergency contacts and better self-harm care. There are benefits to enhancing clinical audit systems with routinely collected data.

The focus on social determinants of health (SDOH) and their impact on health outcomes is evident in U.S. federal actions by Centers for Medicare & Medicaid Services and Office of National Coordinator for Health Information Technology. The disproportionate impact of COVID-19 on minorities and communities of color heightened awareness of health inequities and the need for more robust SDOH data collection. Four Clinical and Translational Science Award (CTSA) hubs comprising the Texas Regional CTSA Consortium (TRCC) undertook an inventory to understand what contextual-level SDOH datasets are offered centrally and which individual-level SDOH are collected in structured fields in each electronic health record (EHR) system potentially for all patients.

Hub teams identified American Community Survey (ACS) datasets available via their enterprise data warehouses for research. Each hub’s EHR analyst team identified structured fields available in their EHR for SDOH using a collection instrument based on a 2021 PCORnet survey and conducted an SDOH field completion rate analysis.

One hub offered ACS datasets centrally. All hubs collected eleven SDOH elements in structured EHR fields. Two collected Homeless and Veteran statuses. Completeness at four hubs was 80%–98%: Ethnicity, Race; < 10%: Education, Financial Strain, Food Insecurity, Housing Security/Stability, Interpersonal Violence, Social Isolation, Stress, Transportation.

Completeness levels for SDOH data in EHR at TRCC hubs varied and were low for most measures. Multiple system-level discussions may be necessary to increase standardized SDOH EHR-based data collection and harmonization to drive effective value-based care, health disparities research, translational interventions, and evidence-based policy.

There is a lack of standardised psychometric data in electronic health record (EHR)-based research. Proxy measures of symptom severity based on patients' clinical records may be useful surrogates in mental health EHR research.

This study aimed to validate proxy tools for the short versions of the Positive and Negative Syndrome Scale (PANSS-6), Young Mania Rating Scale (YMRS-6) and Montgomery–Åsberg Depression Rating Scale (MADRS-6).

A cross-sectional, multicentre study was conducted in a sample of 116 patients with first-episode psychosis from 12 public hospitals in Spain. Concordance between PANSS-6, YMRS-6 and MADRS-6 scores and their respective proxies was evaluated based on information from EHR clinical notes, using a variety of statistical procedures, including multivariate tests to adjust for potential confounders. Bootstrapping techniques were used for internal validation, and an independent cohort from the Treatment and Early Intervention in Psychosis Program (TIPP-Lausanne, Switzerland) for external validation.

The proxy versions correlated strongly with their respective standardised scales (partial correlations ranged from 0.75 to 0.84) and had good accuracy and discriminatory power in distinguishing between patients in and not in remission (percentage of patients correctly classified ranged from 83.9 to 91.4% and bootstrapped optimism-corrected area under the receiver operating characteristic curve ranged from 0.76 to 0.89), with high interrater reliability (intraclass correlation coefficient of 0.81). The findings remained robust in the external validation data-set.

The proxy instruments proposed for assessing psychotic and affective symptoms by reviewing EHR provide a feasible and reliable alternative to traditional structured psychometric procedures, and a promising methodology for real-world practice settings.

Natural language processing (NLP) methods hold promise for improving clinical prediction by utilising information otherwise hidden in the clinical notes of electronic health records. However, clinical practice – as well as the systems and databases in which clinical notes are recorded and stored – change over time. As a consequence, the content of clinical notes may also change over time, which could degrade the performance of prediction models. Despite its importance, the stability of clinical notes over time has rarely been tested.

The lexical stability of clinical notes from the Psychiatric Services of the Central Denmark Region in the period from January 1, 2011, to November 22, 2021 (a total of 14,811,551 clinical notes describing 129,570 patients) was assessed by quantifying sentence length, readability, syntactic complexity and clinical content. Changepoint detection models were used to estimate potential changes in these metrics.

We find lexical stability of the clinical notes over time, with minor deviations during the COVID-19 pandemic. Out of 2988 data points, 17 possible changepoints (corresponding to 0.6%) were detected. The majority of these were related to the discontinuation of a specific note type.

We find lexical and syntactic stability of clinical notes from psychiatric services over time, which bodes well for the use of NLP for predictive modelling in clinical psychiatry.

We tested the ability of our natural language processing (NLP) algorithm to identify delirium episodes in a large-scale study using real-world clinical notes.

We used the Rochester Epidemiology Project to identify persons ≥ 65 years who were hospitalized between 2011 and 2017. We identified all persons with an International Classification of Diseases code for delirium within ±14 days of a hospitalization. We independently applied our NLP algorithm to all clinical notes for this same population. We calculated rates using number of delirium episodes as the numerator and number of hospitalizations as the denominator. Rates were estimated overall, by demographic characteristics, and by year of episode, and differences were tested using Poisson regression.

In total, 14,255 persons had 37,554 hospitalizations between 2011 and 2017. The code-based delirium rate was 3.02 per 100 hospitalizations (95% CI: 2.85, 3.20). The NLP-based rate was 7.36 per 100 (95% CI: 7.09, 7.64). Rates increased with age (both p < 0.0001). Code-based rates were higher in men compared to women (p = 0.03), but NLP-based rates were similar by sex (p = 0.89). Code-based rates were similar by race and ethnicity, but NLP-based rates were higher in the White population compared to the Black and Asian populations (p = 0.001). Both types of rates increased significantly over time (both p values < 0.001).

The NLP algorithm identified more delirium episodes compared to the ICD code method. However, NLP may still underestimate delirium cases because of limitations in real-world clinical notes, including incomplete documentation, practice changes over time, and missing clinical notes in some time periods.

Anxiety and depression are frequently comorbid yet phenotypically distinct. This study identifies differences in the clinically observable phenome across a wide variety of physical and mental disorders comparing patients with diagnoses of depression without anxiety, anxiety without depression, or both depression and anxiety.

Using electronic health records for 14 994 participants with depression and/or anxiety in the Mayo Clinic Biobank, a phenotype-based phenome-wide association study (Phe2WAS) was performed to test for differences between these groups across a broad range of clinical diagnoses observed in the electronic health record. Additional analyses were performed to determine the temporal sequencing of diagnoses.

Compared to patients diagnosed only with anxiety, those diagnosed only with depression were more likely to have diagnoses of obesity (OR 1.75; p = 1 × 10−27), sleep apnea (OR 1.71; p = 1 × 10−22), and type II diabetes (OR 1.74; p = 9 × 10−18). Compared to those diagnosed only with depression, those diagnosed only with anxiety were more likely to have diagnoses of palpitations (OR 1.91; p = 2 × 10−25), benign skin neoplasms (OR 1.61; p = 2 × 10−17), and cardiac dysrhythmias (OR 1.45; p = 2 × 10−12). Patients with comorbid depression and anxiety were more likely to have diagnoses of other mental health disorders, substance use disorders, sleep problems, and gastroesophageal reflux relative to isolated depression.

While depression and anxiety are closely related, this study suggests that phenotypic distinctions exist between depression and anxiety. Improving phenotypic characterization within the broad categories of depression and anxiety could improve the clinical assessment of depression and anxiety.