Coronary calcification measures predict mortality in symptomatic women and men.

OBJECTIVE: To assess the prognostic value of absolute and sex-specific, age-specific and race/ethnicity-specific (Multi-Ethnic Study of Atherosclerosis, MESA) percentiles of coronary artery calcification in symptomatic women and men. METHODS: The study population consisted of 4985 symptomatic patients (2793 women, 56%) visiting a diagnostic outpatient cardiology clinic between 2009 and 2018 who were referred for cardiac CT to determine Coronary Artery Calcium Score (CACS). Regular care data were used and these data were linked to the databases of Statistics Netherlands for all-cause mortality data. Kaplan-Meier curves, multivariate Cox proportional hazards regression and concordance statistics were used to evaluate the prognostic value of CACS and MESA percentiles. Women were older compared with men (60 vs 59 years). RESULTS: Median CACS was 0 (IQR: 0-54) in women and 42 (IQR: 0-54) in men. After a median follow-up of 4.4 years (IQR: 3.1-6.3), 116 (2.3%; 53 women and 63 men) patients died. MESA percentiles did not perform better compared with absolute CACS (C-statistic 0.65, 95% CI 0.57 to 0.73, vs 0.66, 95% CI 0.58 to 0.74, in women and 0.59, 95% CI 0.51 to 0.67, vs 0.62, 95% CI 0.55 to 0.69, in men, for the percentiles and absolute CACS, respectively). CONCLUSIONS: In symptomatic individuals absolute CACS predicts mortality with a moderately good performance. MESA percentiles did not perform better compared with absolute CACS, thus there is no need to use them. Including degree of stenosis in the model might slightly improve mortality risk prediction in women, but not in men.

Development of a Pipeline for Adverse Drug Reaction Identification in Clinical Notes: Word Embedding Models and String Matching.

BACKGROUND: Knowledge about adverse drug reactions (ADRs) in the population is limited because of underreporting, which hampers surveillance and assessment of drug safety. Therefore, gathering accurate information that can be retrieved from clinical notes about the incidence of ADRs is of great relevance. However, manual labeling of these notes is time-consuming, and automatization can improve the use of free-text clinical notes for the identification of ADRs. Furthermore, tools for language processing in languages other than English are not widely available. OBJECTIVE: The aim of this study is to design and evaluate a method for automatic extraction of medication and Adverse Drug Reaction Identification in Clinical Notes (ADRIN). METHODS: Dutch free-text clinical notes (N=277,398) and medication registrations (N=499,435) from the Cardiology Centers of the Netherlands database were used. All clinical notes were used to develop word embedding models. Vector representations of word embedding models and string matching with a medical dictionary (Medical Dictionary for Regulatory Activities [MedDRA]) were used for identification of ADRs and medication in a test set of clinical notes that were manually labeled. Several settings, including search area and punctuation, could be adjusted in the prototype to evaluate the optimal version of the prototype. RESULTS: The ADRIN method was evaluated using a test set of 988 clinical notes written on the stop date of a drug. Multiple versions of the prototype were evaluated for a variety of tasks. Binary classification of ADR presence achieved the highest accuracy of 0.84. Reduced search area and inclusion of punctuation improved performance, whereas incorporation of the MedDRA did not improve the performance of the pipeline. CONCLUSIONS: The ADRIN method and prototype are effective in recognizing ADRs in Dutch clinical notes from cardiac diagnostic screening centers. Surprisingly, incorporation of the MedDRA did not result in improved identification on top of word embedding models. The implementation of the ADRIN tool may help increase the identification of ADRs, resulting in better care and saving substantial health care costs.

Outcomes in patients with a first episode of chest pain undergoing early coronary CT imaging.

OBJECTIVES: To investigate the impact of a CT-first strategy on all-cause and cardiovascular mortality in patients presenting with chest pain in outpatient cardiology clinics. METHODS: Patients with a first presentation of suspected angina pectoris were identified and their data linked to the registrations of Statistics Netherlands for information on mortality. The linked database consisted of 33 068 patients. CT-first patients were defined as patients with a CT calcium score and coronary CT angiography, within 6 weeks after their initial visit. Propensity score matching (1:5) was used to match patients with and without a CT-first strategy. After matching, 12 545 patients were included of which 2308 CT-first patients and 10 237 patients that underwent usual care. RESULTS: Mean age was 57 years, 56.3% were women and median follow-up was 4.9 years. All-cause mortality was significantly lower in CT-first patients (n=43, 1.9%) compared with patients without CT (n=363, 3.5%) (HR: 0.51, 95% CI 0.37 to 0.70). Furthermore, CT-first patients were more likely to receive cardiovascular preventative and antianginal medication (aspirin: 44.9% vs 27.1%, statins: 48.7% vs 30.3%, beta-blockers: 37.8% vs 25.5%, in CT-first and without CT-first patients, respectively) and to undergo downstream diagnostics and interventions (coronary interventions: 8.5% vs 5.7%, coronary angiography: 16.2% vs 10.6% in CT-first and without CT-first patients, respectively). CONCLUSIONS: In a real-world regular care database, a CT-first strategy in patients suspected of angina pectoris was associated with a lowering of all-cause mortality.

Deep neural networks reveal novel sex-specific electrocardiographic features relevant for mortality risk.

Aims: Incorporation of sex in study design can lead to discoveries in medical research. Deep neural networks (DNNs) accurately predict sex based on the electrocardiogram (ECG) and we hypothesized that misclassification of sex is an important predictor for mortality. Therefore, we first developed and validated a DNN that classified sex based on the ECG and investigated the outcome. Second, we studied ECG drivers of DNN-classified sex and mortality. Methods and results: A DNN was trained to classify sex based on 131 673 normal ECGs. The algorithm was validated on internal (68 500 ECGs) and external data sets (3303 and 4457 ECGs). The survival of sex (mis)classified groups was investigated using time-to-event analysis and sex-stratified mediation analysis of ECG features. The DNN successfully distinguished female from male ECGs {internal validation: area under the curve (AUC) 0.96 [95% confidence interval (CI): 0.96, 0.97]; external validations: AUC 0.89 (95% CI: 0.88, 0.90), 0.94 (95% CI: 0.93, 0.94)}. Sex-misclassified individuals (11%) had a 1.4 times higher mortality risk compared with correctly classified peers. The ventricular rate was the strongest mediating ECG variable (41%, 95% CI: 31%, 56%) in males, while the maximum amplitude of the ST segment was strongest in females (18%, 95% CI: 11%, 39%). Short QRS duration was associated with higher mortality risk. Conclusion: Deep neural networks accurately classify sex based on ECGs. While the proportion of ECG-based sex misclassifications is low, it is an interesting biomarker. Investigation of the causal pathway between misclassification and mortality uncovered new ECG features that might be associated with mortality. Increased emphasis on sex as a biological variable in artificial intelligence is warranted.

Routine clinical care data from thirteen cardiac outpatient clinics: design of the Cardiology Centers of the Netherlands (CCN) database.

BACKGROUND: Despite the increasing availability of clinical data due to the digitalisation of healthcare systems, data often remain inaccessible due to the diversity of data collection systems. In the Netherlands, Cardiology Centers of the Netherlands (CCN) introduced "one-stop shop" diagnostic clinics for patients suspected of cardiac disease by their general practitioner. All CCN clinics use the same data collection system and standardised protocol, creating a large regular care database. This database can be used to describe referral practices, evaluate risk factors for cardiovascular disease (CVD) in important patient subgroups, and develop prediction models for use in daily care. CONSTRUCTION AND CONTENT: The current database contains data on all patients who underwent a cardiac workup in one of the 13 CCN clinics between 2007 and February 2018 (n = 109,151, 51.9% women). Data were pseudonymised and contain information on anthropometrics, cardiac symptoms, risk factors, comorbidities, cardiovascular and family history, standard blood laboratory measurements, transthoracic echocardiography, electrocardiography in rest and during exercise, and medication use. Clinical follow-up is based on medical need and consisted of either a repeat visit at CCN (43.8%) or referral for an external procedure in a hospital (16.5%). Passive follow-up via linkage to national mortality registers is available for 95% of the database. UTILITY AND DISCUSSION: The CCN database provides a strong base for research into historically underrepresented patient groups due to the large number of patients and the lack of in- and exclusion criteria. It also enables the development of artificial intelligence-based decision support tools. Its contemporary nature allows for comparison of daily care with the current guidelines and protocols. Missing data is an inherent limitation, as the cardiologist could deviate from standardised protocols when clinically indicated. CONCLUSION: The CCN database offers the opportunity to conduct research in a unique population referred from the general practitioner to the cardiologist for diagnostic workup. This, in combination with its large size, the representation of historically underrepresented patient groups and contemporary nature makes it a valuable tool for expanding our knowledge of cardiovascular diseases. TRIAL REGISTRATION: Not applicable.