Evaluating a novel approach to stimulate open science collaborations: a case series of "study-a-thon" events within the OHDSI and European IMI communities.

Objective: We introduce and review the concept of a study-a-thon as a catalyst for open science in medicine, utilizing harmonized real world, observation health data, tools, skills, and methods to conduct network studies, generating insights for those wishing to use study-a-thons for future research. Materials and Methods: A series of historical study-a-thons since 2017 to present were reviewed for thematic insights as to the opportunity to accelerate the research method to conduct studies across therapeutic areas. Review of publications and experience of the authors generated insights to illustrate the conduct of study-a-thons, key learning, and direction for those wishing to conduct future such study-a-thons. Results: A review of six study-a-thons have provided insights into their scientific impact, and 13 areas of insights for those wishing to conduct future study-a-thons. Defining aspects of the study-a-thon method for rapid, collaborative research through network studies reinforce the need to clear scientific rationale, tools, skills, and methods being collaboratively to conduct a focused study. Well-characterized preparatory, execution and postevent phases, coalescing skills, experience, data, clinical input (ensuring representative clinical context to the research query), and well-defined, logical steps in conducting research via the study-a-thon method are critical. Conclusions: A study-a-thon is a focused multiday research event generating reliable evidence on a specific medical topic across different countries and health systems. In a study-a-thon, a multidisciplinary team collaborate to create an accelerated contribution to scientific evidence and clinical practice. It critically accelerates the research process, without inhibiting the quality of the research output and evidence generation, through a reproducible process.

Investigating the impact of development and internal validation design when training prognostic models using a retrospective cohort in big US observational healthcare data.

OBJECTIVE: The internal validation of prediction models aims to quantify the generalisability of a model. We aim to determine the impact, if any, that the choice of development and internal validation design has on the internal performance bias and model generalisability in big data (n~500 000). DESIGN: Retrospective cohort. SETTING: Primary and secondary care; three US claims databases. PARTICIPANTS: 1 200 769 patients pharmaceutically treated for their first occurrence of depression. METHODS: We investigated the impact of the development/validation design across 21 real-world prediction questions. Model discrimination and calibration were assessed. We trained LASSO logistic regression models using US claims data and internally validated the models using eight different designs: 'no test/validation set', 'test/validation set' and cross validation with 3-fold, 5-fold or 10-fold with and without a test set. We then externally validated each model in two new US claims databases. We estimated the internal validation bias per design by empirically comparing the differences between the estimated internal performance and external performance. RESULTS: The differences between the models' internal estimated performances and external performances were largest for the 'no test/validation set' design. This indicates even with large data the 'no test/validation set' design causes models to overfit. The seven alternative designs included some validation process to select the hyperparameters and a fair testing process to estimate internal performance. These designs had similar internal performance estimates and performed similarly when externally validated in the two external databases. CONCLUSIONS: Even with big data, it is important to use some validation process to select the optimal hyperparameters and fairly assess internal validation using a test set or cross-validation.

Prediction of 90-day mortality after surgery for colorectal cancer using standardized nationwide quality-assurance data.

BACKGROUND: Personalized risk assessment provides opportunities for tailoring treatment, optimizing healthcare resources and improving outcome. The aim of this study was to develop a 90-day mortality-risk prediction model for identification of high- and low-risk patients undergoing surgery for colorectal cancer. METHODS: This was a nationwide cohort study using records from the Danish Colorectal Cancer Group database that included all patients undergoing surgery for colorectal cancer between 1 January 2004 and 31 December 2015. A least absolute shrinkage and selection operator logistic regression prediction model was developed using 121 pre- and intraoperative variables and internally validated in a hold-out test data set. The accuracy of the model was assessed in terms of discrimination and calibration. RESULTS: In total, 49 607 patients were registered in the database. After exclusion of 16 680 individuals, 32 927 patients were included in the analysis. Overall, 1754 (5.3 per cent) deaths were recorded. Targeting high-risk individuals, the model identified 5.5 per cent of all patients facing a risk of 90-day mortality exceeding 35 per cent, corresponding to a 6.7 times greater risk than the average population. Targeting low-risk individuals, the model identified 20.9 per cent of patients facing a risk less than 0.3 per cent, corresponding to a 17.7 times lower risk compared with the average population. The model exhibited discriminatory power with an area under the receiver operating characteristics curve of 85.3 per cent (95 per cent c.i. 83.6 to 87.0) and excellent calibration with a Brier score of 0.04 and 32 per cent average precision. CONCLUSION: Pre- and intraoperative data, as captured in national health registries, can be used to predict 90-day mortality accurately after colorectal cancer surgery.

Vascular and metabolic risk factor differences prior to dementia diagnosis: a multidatabase case-control study using European electronic health records.

OBJECTIVE: The objective of the study is to compare body mass index (BMI), systolic/diastolic blood pressure (SBP/DBP) and serum total cholesterol levels between dementia cases and controls at multiple time intervals prior to dementia onset, and to test time interval as a modifying factor for these associations. DESIGN: Case-control study. SETTING: Six European electronic health records databases. PARTICIPANTS: 291 780 cases at the date of first-recorded dementia diagnosis, compared with 29 170 549 controls randomly selected from the same databases, age matched and sex matched at this index date. EXPOSURE: The following measures were extracted whenever recorded within each dataset: BMI (kg/m2), SBP and DBP (mm Hg) and serum total cholesterol (mmol/L). Levels for each of these variables were defined within six 2-year time intervals over the 12 years prior to the index date. MAIN OUTCOMES: Case-control differences in exposures of interest were modelled for each time period and adjusted for demographic and clinical factors (ischaemic/unspecified stroke, type 2 diabetes mellitus, acute myocardial infarction, hypertension diagnosis, antihypertensive medication, cholesterol-lowering medication). Coefficients and interactions with time period were meta-analysed across the six databases. RESULTS: Mean BMI (coefficient -1.16 kg/m2; 95% CI -1.38 to 0.93) and SBP (-2.83 mm Hg; 95% CI -4.49 to -1.16) were lower in cases at diagnosis, and case-control differences were greater in more recent time periods, as indicated by significant case-x-time interaction and case-x-time-squared interaction terms. Time variations in coefficients for cholesterol levels were less consistent between databases and those for DBP were largely not significant. CONCLUSION: Routine clinical data show emerging divergence in levels of BMI and SBP prior to the diagnosis of dementia but less evidence for DBP or total cholesterol levels. These divergences should receive at least some consideration in routine dementia risk screening, although underlying mechanisms still require further investigation.

Accuracy of an automated knowledge base for identifying drug adverse reactions.

INTRODUCTION: Drug safety researchers seek to know the degree of certainty with which a particular drug is associated with an adverse drug reaction. There are different sources of information used in pharmacovigilance to identify, evaluate, and disseminate medical product safety evidence including spontaneous reports, published peer-reviewed literature, and product labels. Automated data processing and classification using these evidence sources can greatly reduce the manual curation currently required to develop reference sets of positive and negative controls (i.e. drugs that cause adverse drug events and those that do not) to be used in drug safety research. METHODS: In this paper we explore a method for automatically aggregating disparate sources of information together into a single repository, developing a predictive model to classify drug-adverse event relationships, and applying those predictions to a real world problem of identifying negative controls for statistical method calibration. RESULTS: Our results showed high predictive accuracy for the models combining all available evidence, with an area under the receiver-operator curve of ⩾0.92 when tested on three manually generated lists of drugs and conditions that are known to either have or not have an association with an adverse drug event. CONCLUSIONS: Results from a pilot implementation of the method suggests that it is feasible to develop a scalable alternative to the time-and-resource-intensive, manual curation exercise previously applied to develop reference sets of positive and negative controls to be used in drug safety research.

Antidepressants and heart-rate variability in older adults: a population-based study.

BACKGROUND: Tricyclic antidepressants (TCAs) and selective serotonin reuptake inhibitors (SSRIs) may be associated with lower heart rate variability (HRV), a condition associated with increased mortality risk. We aimed to investigate the association between TCAs, SSRIs and HRV in a population-based study. METHOD: In the prospective Rotterdam Study cohort, up to five electrocardiograms (ECGs) per participant were recorded (1991-2012). Two HRV variables were studied based on 10-s ECG recordings: standard deviation of normal-to-normal RR intervals (SDNN) and root mean square of successive RR interval differences (RMSSD). We compared the HRV on ECGs recorded during use of antidepressants with the HRV on ECGs recorded during non-use of any antidepressant. Additionally, we analysed the change in HRV on consecutive ECGs. Those who started or stopped using antidepressants before the second ECG were compared with non-users on two ECGs. RESULTS: We included 23 647 ECGs from 11 729 participants (59% women, mean age 64.6 years at baseline). Compared to ECGs recorded during non-use of antidepressants (n = 22 971), SDNN and RMSSD were lower in ECGs recorded during use of TCAs (n = 296) and SSRIs (n = 380). Participants who started using TCAs before the second ECG had a decrease in HRV and those who stopped had an increase in HRV compared to consistent non-users (p < 0.001). Starting or stopping SSRIs was not associated with HRV changes. CONCLUSION: TCAs were associated with a lower HRV in all analyses, indicating a real drug effect. For SSRIs the results are mixed, indicating a weaker association, possibly due to other factors.

Combining multiple healthcare databases for postmarketing drug and vaccine safety surveillance: why and how?

A growing number of international initiatives (e.g. EU-ADR, Sentinel, OMOP, PROTECT and VAESCO) are based on the combined use of multiple healthcare databases for the conduct of active surveillance studies in the area of drug and vaccine safety. The motivation behind combining multiple healthcare databases is the earlier detection and validation, and hence earlier management, of potential safety issues. Overall, the combination of multiple healthcare databases increases statistical sample size and heterogeneity of exposure for postmarketing drug and vaccine safety surveillance, despite posing several technical challenges. Healthcare databases generally differ by underlying healthcare systems, type of information collected, drug/vaccine and medical event coding systems and language. Therefore, harmonization of medical data extraction through homogeneous coding algorithms across highly different databases is necessary. Although no standard procedure is currently available to achieve this, several approaches have been developed in recent projects. Another main challenge involves choosing the work models for data management and analyses whilst respecting country-specific regulations in terms of data privacy and anonymization. Dedicated software (e.g. Jerboa) has been produced to deal with privacy issues by sharing only anonymized and aggregated data using a common data model. Finally, storage and safe access to the data from different databases requires the development of a proper remote research environment. The aim of this review is to provide a summary of the potential, disadvantages, methodological issues and possible solutions concerning the conduct of postmarketing multidatabase drug and vaccine safety studies, as demonstrated by several international initiatives.

Minimum bandwidth requirements for recording of pediatric electrocardiograms.

BACKGROUND: Previous studies that determined the frequency content of the pediatric ECG had their limitations: the study population was small or the sampling frequency used by the recording system was low. Therefore, current bandwidth recommendations for recording pediatric ECGs are not well founded. We wanted to establish minimum bandwidth requirements using a large set of pediatric ECGs recorded at a high sampling rate. METHODS AND RESULTS: For 2169 children aged 1 day to 16 years, a 12-lead ECG was recorded at a sampling rate of 1200 Hz. The averaged beats of each ECG were passed through digital filters with different cut off points (50 to 300 Hz in 25-Hz steps). We measured the absolute errors in maximum QRS amplitude for each simulated bandwidth and determined the percentage of records with an error >25 microV. We found that in any lead, a bandwidth of 250 Hz yields amplitude errors <25 microV in >95% of the children <1 year. For older children, a gradual decrease in ECG frequency content was demonstrated. CONCLUSIONS: We recommend a minimum bandwidth of 250 Hz to record pediatric ECGs. This bandwidth is considerably higher than the previous recommendation of 150 Hz from the American Heart Association.

New normal limits for the paediatric electrocardiogram.

AIMS: Previous studies that determined the normal limits for the paediatric ECG had their imperfections: ECGs were recorded at a relatively low sampling rate, ECG measurements were conducted manually, or normal limits were presented for only a limited set of parameters. The aim of this study was to establish an up-to-date and complete set of clinically relevant normal limits for the paediatric ECG. METHODS AND RESULTS: ECGs from 1912 healthy Dutch children (age 11 days to 16 years) were recorded at a sampling rate of 1200 Hz. The digitally stored ECGs were analysed using a well-validated ECG computer program. The normal limits of all clinically relevant ECG measurements were determined for nine age groups. Clinically significant differences were shown to exist, compared with previously established normal limits. Sex differences could be demonstrated for QRS duration and several amplitude measurements. CONCLUSIONS: These new normal limits differ substantially from those commonly used and suggest that diagnostic criteria for the paediatric ECG should be adjusted.

A new approach to mechanical simulation of lung behaviour: pressure-controlled and time-related piston movement.

A mechanical lung simulator is described (an extension of a previous mechanical simulator) which simulates normal breathing and artificial ventilation in patients. The extended integration of hardware and software offers many new possibilities and advantages over the former simulator. The properties of components which simulate elastance and airway resistance of the lung are defined in software rather than by the mechanical properties of the components alone. Therefore, a more flexible simulation of non-linear behaviour and the cross-over effects of lung properties is obtained. Furthermore, the range of lung compliance is extended to simulate patients with emphysema. The dependency of airway resistance on lung recoil pressure and transmural pressure of the airways can also be simulated. The new approach enables one to incorporate time-related mechanics such as the influence of lung viscosity or cardiac oscillation. The different relations defined in the software can be changed from breath to breath. Three simulations are presented: (1) computer-controlled expiration in the artificially ventilated lung; (2) simulation of normal breathing; and (3) simulation of viscoelastance and cardiac influences during artificial ventilation. The mechanical simulator provides a reproducible and flexible environment for testing new software and equipment in the lung function laboratory and in intensive care, and can be used for instruction and training.

PEDMEANS: a computer program for the interpretation of pediatric electrocardiograms.

The interpretation pediatric electrocardiograms (ECGs) is complicated because of the strong age-dependency of the diagnostic criteria. We wanted to develop and evaluate a computer program for the interpretation of pediatric 12-lead ECGs. Continuous age-dependent normal limits were established based on ECGs from 1,912 healthy Dutch children. Additionally, a reference interpretation was obtained for 1,718 ECGs recorded at the Sophia Children's Hospital. The total set of ECGs was divided in a training set of 1076 ECGs and a test set of 642 ECGs. All ECGs were recorded at a sampling rate of 1,200 Hz. Based on the normal limits and the training set, diagnostic rules were formalized in an iterative process by using expert interviews and automatic rule induction. The resultant rules were evaluated on the test set. The performance of the program, on our study population, appears to justify its use in a clinical setting. Preferably, the program should also be evaluated in other clinical centers.

Effects of fluticasone propionate on methacholine dose-response curves in nonsmoking atopic asthmatics.

Methacholine is frequently used to determine bronchial hyperresponsiveness (BHR) and to generate dose-response curves. These curves are characterized by a threshold (provocative concentration of methacholine producing a 20% fall in forced expiratory volume in one second (PC20) = sensitivity), slope (reactivity) and maximal response (plateau). We investigated the efficacy of 12 weeks of treatment with 1,000 microg fluticasone propionate in a double-blind, placebo-controlled study in 33 atopic asthmatics. The outcome measures used were the influence on BHR and the different indices of the methacholine dose-response (MDR) curve. After 2 weeks run-in, baseline lung function data were obtained and a MDR curve was measured with doubling concentrations of the methacholine from 0.03 to 256 mg x mL(-1). MDR curves were repeated after 6 and 12 weeks. A recently developed, sigmoid cumulative Gaussian distribution function was fitted to the data. Although sensitivity was obtained by linear interpolation of two successive log2 concentrations, reactivity, plateau and the effective concentration at 50% of the plateau value (EC50) were obtained as best fit parameters. In the fluticasone group, significant changes occurred after 6 weeks with respect to means of PC20 (an increase of 3.4 doubling doses), plateau value fall in forced expiratory volume in one second (FEV1) (from 58% at randomization to 41% at 6 weeks) and baseline FEV1 (from 3.46 to 3.75 L) in contrast to the placebo group. Stabilization occurred after 12 weeks. Changes for reactivity were less marked, whereas changes in log, EC50 were not significantly different between the groups. We conclude that fluticasone is very effective in decreasing the maximal airway narrowing response and in increasing PC20. However, it is likely that part of this increase is related to the decrease of the plateau of maximal response.