The added value of temporal data and the best way to handle it: A use-case for atrial fibrillation using general practitioner data.

INTRODUCTION: Temporal data has numerous challenges for deep learning such as irregularity of sampling. New algorithms are being developed that can handle these temporal challenges better. However, it is unclear how the performance ranges from classical non-temporal models to newly developed algorithms. Therefore, this study compares different non-temporal and temporal algorithms for a relevant use case, the prediction of atrial fibrillation (AF) using general practitioner (GP) data. METHODS: Three datasets with a 365-day observation window and prediction windows of 14, 180 and 360 days were used. Data consisted of medication, lab, symptom, and chronic diseases codings registered by the GP. The benchmark discarded temporality and used logistic regression, XGBoost models and neural networks on the presence of codings over the whole year. Pattern data extracted common patterns of GP codings and tested using the same algorithms. LSTM and CKConv models were trained as models incorporating temporality. RESULTS: Algorithms which incorporated temporality (LSTM and CKConv, (max AUC 0.734 at 360 days prediction window) outperformed both benchmark and pattern algorithms (max AUC 0.723, with a significant improvement using the 360 days prediction window (p = 0.04). The difference between the benchmark and the LSTM or CKConv algorithm decreased with smaller prediction windows, indicating temporal importance for longer prediction windows. The CKConv and LSTM algorithm performed similarly, possibly due to limited sequence length. CONCLUSION: Temporal models outperformed non-temporal models for the prediction of AF. For temporal models, CKConv is a promising algorithm to handle temporal data using GP data as it can handle irregular data.

Routine healthcare disruptions: a longitudinal study on changes in self-management behavior during the COVID-19 pandemic.

BACKGROUND: The outbreak of COVID-19 had a significant impact on routines and continuity of professional care. As frequent users of this professional care, especially for people with chronic diseases this had consequences. Due to barriers in access to healthcare, an even greater appeal was made on the self-management behaviors of this group. In the present study, we aim to investigate the extent to which self-management changed during the recent pandemic, and which factors contributed to these changes. METHODS: The Dutch 'National Panel of people with Chronic Illness or Disability' was used to collect self-reported data of people with at least one chronic disease. Self-management was assessed with the Partners in Health questionnaire at two time points: before the crisis in 2018 and during the second wave of crisis in Autumn 2020. Paired t-tests were used to analyze changes in self-management. Potential associating factors on three levels - patient, organization and environment - were assessed in 2020 and their impact on self-management changes was tested with multinomial logistic regression. RESULTS: Data from 345 panel members was available at two time points. In the majority of people, self-management behaviors were stable (70.7%). About one in seven experienced improved self-management (15.1%), and a similar proportion experienced deteriorated self-management (14.2%). Sex, physical disability, mental health and daily stressors due to COVID-19 (patient level), changes in healthcare access (organization level), and social support (environment level) were significantly associated with experienced changes in self-management. CONCLUSIONS: People with chronic diseases experienced different trajectories of self-management changes during COVID-19. We need to be aware of people who seem to be more vulnerable to a healthcare crisis and report less stable self-management, such as those who experience mental health problems or daily stressors. Continuity of care and social support can buffer the impact of a healthcare crisis on self-management routines of people with chronic diseases.

Prediction of heart failure 1 year before diagnosis in general practitioner patients using machine learning algorithms: a retrospective case-control study.

OBJECTIVES: Heart failure (HF) is a commonly occurring health problem with high mortality and morbidity. If potential cases could be detected earlier, it may be possible to intervene earlier, which may slow progression in some patients. Preferably, it is desired to reuse already measured data for screening of all persons in an age group, such as general practitioner (GP) data. Furthermore, it is essential to evaluate the number of people needed to screen to find one patient using true incidence rates, as this indicates the generalisability in the true population. Therefore, we aim to create a machine learning model for the prediction of HF using GP data and evaluate the number needed to screen with true incidence rates. DESIGN, SETTINGS AND PARTICIPANTS: GP data from 8543 patients (-2 to -1 year before diagnosis) and controls aged 70+ years were obtained retrospectively from 01 January 2012 to 31 December 2019 from the Nivel Primary Care Database. Codes about chronic illness, complaints, diagnostics and medication were obtained. Data were split in a train/test set. Datasets describing demographics, the presence of codes (non-sequential) and upon each other following codes (sequential) were created. Logistic regression, random forest and XGBoost models were trained. Predicted outcome was the presence of HF after 1 year. The ratio case:control in the test set matched true incidence rates (1:45). RESULTS: Sole demographics performed average (area under the curve (AUC) 0.692, CI 0.677 to 0.706). Adding non-sequential information combined with a logistic regression model performed best and significantly improved performance (AUC 0.772, CI 0.759 to 0.785, p<0.001). Further adding sequential information did not alter performance significantly (AUC 0.767, CI 0.754 to 0.780, p=0.07). The number needed to screen dropped from 14.11 to 5.99 false positives per true positive. CONCLUSION: This study created a model able to identify patients with pending HF a year before diagnosis.

CHARGE-AF in a national routine primary care electronic health records database in the Netherlands: validation for 5-year risk of atrial fibrillation and implications for patient selection in atrial fibrillation screening.

AIMS: To validate a multivariable risk prediction model (Cohorts for Heart and Aging Research in Genomic Epidemiology model for atrial fibrillation (CHARGE-AF)) for 5-year risk of atrial fibrillation (AF) in routinely collected primary care data and to assess CHARGE-AF's potential for automated, low-cost selection of patients at high risk for AF based on routine primary care data. METHODS: We included patients aged ≥40 years, free of AF and with complete CHARGE-AF variables at baseline, 1 January 2014, in a representative, nationwide routine primary care database in the Netherlands (Nivel-PCD). We validated CHARGE-AF for 5-year observed AF incidence using the C-statistic for discrimination, and calibration plot and stratified Kaplan-Meier plot for calibration. We compared CHARGE-AF with other predictors and assessed implications of using different CHARGE-AF cut-offs to select high-risk patients. RESULTS: Among 111 475 patients free of AF and with complete CHARGE-AF variables at baseline (17.2% of all patients aged ≥40 years and free of AF), mean age was 65.5 years, and 53% were female. Complete CHARGE-AF cases were older and had higher AF incidence and cardiovascular comorbidity rate than incomplete cases. There were 5264 (4.7%) new AF cases during 5-year follow-up among complete cases. CHARGE-AF's C-statistic for new AF was 0.74 (95% CI 0.73 to 0.74). The calibration plot showed slight risk underestimation in low-risk deciles and overestimation of absolute AF risk in those with highest predicted risk. The Kaplan-Meier plot with categories <2.5%, 2.5%-5% and >5% predicted 5-year risk was highly accurate. CHARGE-AF outperformed CHA2DS2-VASc (Cardiac failure or dysfunction, Hypertension, Age >=75 [Doubled], Diabetes, Stroke [Doubled]-Vascular disease, Age 65-74, and Sex category [Female]) and age alone as predictors for AF. Dichotomisation at cut-offs of 2.5%, 5% and 10% baseline CHARGE-AF risk all showed merits for patient selection in AF screening efforts. CONCLUSION: In patients with complete baseline CHARGE-AF data through routine Dutch primary care, CHARGE-AF accurately assessed AF risk among older primary care patients, outperformed both CHA2DS2-VASc and age alone as predictors for AF and showed potential for automated, low-cost patient selection in AF screening.

Exploring the Impact of Child Maltreatment and Interparental Violence on Violent Delinquency in an International Sample.

Young people are exposed to violence regularly in their homes, schools, and communities. Such exposure can cause them significant physical, mental, and emotional harm, with long-term effects lasting well into adulthood. Of particular concern is violence within the family, where children are victimized by their parents. Research shows that direct and indirect childhood exposure to violence and maltreatment within the family increases the risk of subsequent violent delinquent behavior. Social learning theory and attachment theory place parenting at the center of the "cycle of violence," and "intergenerational transmission of violence" claims that experiencing violence in childhood will lead to the perpetration of violence in adolescence. Although much research has been done, these assertions have never been tested on a large international sample of young people. The current article fills this void by analyzing surveys completed by 57,892 students who were 12 to 16 years old from 25 countries as part of the International Self-Report Delinquency Study (ISRD3). Structural equation modeling (SEM) is used to test the direct and indirect effects of child maltreatment and interparental violence on self-reported violent delinquency. Mediating effects are proposed for attachment to parents, parental social control (measured by parental knowledge, parental monitoring, and child disclosure), and parental moral authority. Analysis suggests direct effects of child maltreatment and interparental violence, as well as mediating effects of parental monitoring, parental knowledge, and parental moral authority. Child disclosure and attachment to parents do not affect violent juvenile offending. Being a victim of both child maltreatment and interparental violence is found to exacerbate the effect on violent offending. The results support the cross-national generalizability of the "cycle of violence" argument that children tend to reproduce the behavior of their parents.

Applying machine learning on health record data from general practitioners to predict suicidality.

BACKGROUND: Suicidal behaviour is difficult to detect in the general practice. Machine learning (ML) algorithms using routinely collected data might support General Practitioners (GPs) in the detection of suicidal behaviour. In this paper, we applied machine learning techniques to support GPs recognizing suicidal behaviour in primary care patients using routinely collected general practice data. METHODS: This case-control study used data from a national representative primary care database including over 1.5 million patients (Nivel Primary Care Database). Patients with a suicide (attempt) in 2017 were selected as cases (N = 574) and an at risk control group (N = 207,308) was selected from patients with psychological vulnerability but without a suicide attempt in 2017. RandomForest was trained on a small subsample of the data (training set), and evaluated on unseen data (test set). RESULTS: Almost two-third (65%) of the cases visited their GP within the last 30 days before the suicide (attempt). RandomForest showed a positive predictive value (PPV) of 0.05 (0.04-0.06), with a sensitivity of 0.39 (0.32-0.47) and area under the curve (AUC) of 0.85 (0.81-0.88). Almost all controls were accurately labeled as controls (specificity = 0.98 (0.97-0.98)). Among a sample of 650 at-risk primary care patients, the algorithm would label 20 patients as high-risk. Of those, one would be an actual case and additionally, one case would be missed. CONCLUSION: In this study, we applied machine learning to predict suicidal behaviour using general practice data. Our results showed that these techniques can be used as a complementary step in the identification and stratification of patients at risk of suicidal behaviour. The results are encouraging and provide a first step to use automated screening directly in clinical practice. Additional data from different social domains, such as employment and education, might improve accuracy.

Transforming prevention systems in the United States and the Netherlands using Communities That Care Promising prevention in the eyes of Josine Junger-Tas.

Josine Junger-Tas introduced the Communities That Care (CTC) prevention system to the Netherlands as a promising approach to address the growing youth violence and delinquency. Using data from a randomized trial of CTC in the United States and a quasi-experimental study of CTC in the Netherlands, this article describes the results of a comparison of the implementation of CTC in 12 U.S. communities and 5 Dutch neighborhoods. CTC communities in both countries achieved higher stages of a science-based approach to prevention than control communities, but full implementation of CTC in the Netherlands was hampered by the very small list of prevention programs tested and found effective in the Dutch context.