Background: Predicting the outcomes of serious mental illnesses including bipolar disorder (BD) is clinically beneficial, yet difficult. Objectives: This study aimed to predict hospitalization and mortality for patients with incident BD using a deep neural network approach. Methods: We randomly sampled 20,000 US Veterans with BD. Data on patients’ prior hospitalizations, diagnoses, procedures, medications, note types, vital signs, lab results, and BD symptoms that occurred within 1 year before and at the onset of the incident BD were extracted as features. We then created novel temporal images of patient clinical features both during the prodromal period and at the time of the disease onset. Using each temporal image as a feature, we trained and tested deep neural network learning models to predict the 1-year combined outcome of hospitalization and mortality. Results: The models achieved accuracies of 0.766–0.949 and AUCs of 0.745–0.806 for the combined outcomes. The AUC for predicting mortality was 0.814, while its highest and lowest values for predicting different types of hospitalization were 90.4% and 70.1%, suggesting that some outcomes were more difficult to predict than others. Conclusion: Deep learning using temporal graphics of clinical history is a new and promising analytical approach for mental health outcome prediction.

Transformer is the latest deep neural network (DNN) architecture for sequence data learning, which has revolutionized the field of natural language processing. This success has motivated researchers to explore its application in the healthcare domain. Despite the similarities between longitudinal clinical data and natural language data, clinical data presents unique complexities that make adapting Transformer to this domain challenging. To address this issue, we have designed a new Transformer-based DNN architecture, referred to as Hybrid Value-Aware Transformer (HVAT), which can jointly learn from longitudinal and non-longitudinal clinical data. HVAT is unique in the ability to learn from the numerical values associated with clinical codes/concepts such as labs, and in the use of a flexible longitudinal data representation called clinical tokens. We have also trained a prototype HVAT model on a case-control dataset, achieving high performance in predicting Alzheimer’s disease and related dementias as the patient outcome. The results demonstrate the potential of HVAT for broader clinical data-learning tasks.

Previous studies suggested that human immunodeficiency virus (HIV) infected patients at risk of poor adherence were not distinguishable only based on the baseline characteristics. This study is to identify patient characteristics that would be consistently associated with poor adherence across regimens and to understand the associations between initial and long-term adherence. HIV treatment-naïve patients initiated on protease inhibitors, nonnucleoside reverse transcriptase inhibitors, or integrase strand transfer inhibitors were identified from the Veteran Health Administration system. Initial adherence measured as initial coverage ratio (ICR) and long-term adherence measured as thereafter 1-year proportion days covered (PDC) of base agent and complete regimen were estimated for each patient. The patients most likely to exhibit poor adherence were African-American, with lower socioeconomic status, and healthier. The initial coverage ratio of base agent and complete regimen were highly correlated, but the correlations between ICR and thereafter 1-year PDC were low. However, including initial adherence as a predictor in predictive model would substantially increase predictive accuracy of future adherence.

Many studies have estimated the association between the adherence to antiretroviral therapies and human immunodeficiency virus (HIV) patients’ virologic/immunologic outcomes. However, evidence is lacking on the causal effect of adherence on the outcomes. The goal of this study is to understand whether near perfect adherence is necessary to achieve optimal virologic outcome and also to investigate the effect of initial adherence to antiretroviral therapies on initial viral suppression by different regimens. A cohort study was conducted on HIV veterans initiating antiretroviral therapies in 1999 to 2015. The primary outcome was the first viral suppression occurred within 30 to 60 days since the index date. Multiple imputation was used to impute the missing value of virologic outcomes. The inverse probability of treatment weighting (IPTW) method was applied to estimate the viral suppression rate at each specific adherence category for each regimen category. Marginal structural models with IPTW were used to estimate the risk of viral suppression in lower-adherence categories in comparison to near-perfect adherence level ≥95%. Data showed that lower adherence caused lower viral suppression rate, with the association differentiated by the regimen. Patients on integrase strand transfer had the highest viral suppression rate, with patients on protease inhibitors having the lowest rate. Regardless of regimens, the viral suppression rate among patients at initial adherence of 75 to <95% was not statistically different from patients at adherence of ≥95%; however, the differences might be clinically significant.

Opioid use disorder is known to be under-coded as a diagnosis, yet problematic opioid use can be documented in clinical notes, which are included in electronic health records. We sought to identify problematic opioid use from a full range of clinical notes and compare the demographic and clinical characteristics of patients identified as having problematic opioid use exclusively in clinical notes to patients documented through ICD opioid use disorder diagnostic codes. We developed and applied a natural language processing (NLP) tool that combines rule-based pattern analysis and a trained support vector machine to the clinical notes of a patient cohort (n = 222,371) from two Veteran Affairs service regions to identify patients with problematic opioid use. We also used a set of ICD diagnostic codes to identify patients with opioid use disorder from the same cohort. The NLP tool achieved 96.6% specificity, 90.4% precision/PPV, 88.4% sensitivity/recall, and 94.4% accuracy on unseen test data. NLP exclusively identified 57,331 patients; 6997 patients had positive ICD code identifications. Patients exclusively identified through NLP were more likely to be women. Those identified through ICD codes were more likely to be male, younger, have concurrent benzodiazepine prescriptions, more comorbidities, and more care encounters, and were less likely to be married. Patients in both these groups had substantially elevated comorbidity levels compared with patients not documented through either method as experiencing problematic opioid use. Clinicians may be reluctant to code for opioid use disorder. It is therefore incumbent on the healthcare team to search for documentation of opioid concerns within clinical notes.

Cerebral microinfarcts are associated with cognitive impairment and dementia. Small vessel diseases such as cerebral arteriolosclerosis and cerebral amyloid angiography (CAA) have been found to be associated with microinfarcts. Less is known about the associations of these vasculopathies with the presence, numbers, and location of microinfarcts. These associations were examined in the clinical and autopsy data of 842 participants in the Adult Changes in Thought (ACT) study. Both vasculopathies were categorized by severity (none, mild, moderate, and severe) and region (cortical and subcortical). Odds ratios (OR) and 95% CIs for microinfarcts associated with arteriolosclerosis and CAA adjusted for possible modifying covariates such as age at death, sex, blood pressure, APOE genotype, Braak, and CERAD were estimated. 417 (49.5%) had microinfarcts (cortical, 301; subcortical, 249), 708 (84.1%) had cerebral arteriolosclerosis, 320 (38%) had CAA, and 284 (34%) had both. Ors (95% CI) for any microinfarct were 2.16 (1.46–3.18) and 4.63 (2.90–7.40) for those with moderate (n = 183) and severe (n = 124) arteriolosclerosis, respectively. Respective Ors (95% CI) for the number of microinfarcts were 2.25 (1.54–3.30) and 4.91 (3.18–7.60). Similar associations were observed for cortical and subcortical microinfarcts. Ors (95% Cis) for the number of microinfarcts associated with mild (n = 75), moderate (n = 73), and severe (n = 15) amyloid angiopathy were 0.95 (0.66–1.35), 1.04 (0.71–1.52), and 2.05 (0.94–4.45), respectively. Respective Ors (95% Cis) for cortical microinfarcts were 1.05 (0.71–1.56), 1.50 (0.99–2.27), and 1.69 (0.73–3.91). Respective Ors (95% Cis) for subcortical microinfarcts were 0.84 (0.55–1.28), 0.72 (0.46–1.14), and 0.92 (0.37–2.28). These findings suggest a significant association of cerebral arteriolosclerosis with the presence, number, and location (cortical and subcortical) of microinfarcts, and a weak and non-significant association of CAA with each microinfarct, highlighting the need for future research to better understand the role of small vessel diseases in the pathogenesis of cerebral microinfarcts.