Publications by Year: 2019

We present DR-Train, the first long-term open-access dataset recording dynamic responses from in-service light rail vehicles. Specifically, the dataset contains measurements from multiple sensor channels mounted on two in-service light rail vehicles that run on a 42.2-km light rail network in the city of Pittsburgh, Pennsylvania. This dataset provides dynamic responses of in-service trains via vibration data collected by accelerometers, which enables a low-cost way of monitoring rail tracks more frequently. Such an approach will result in more reliable and economical ways to monitor rail infrastructure. The dataset also includes corresponding GPS positions of the trains, environmental conditions (including temperature, wind, weather, and precipitation), and track maintenance logs. The data, which is stored in a MAT-file format, can be conveniently loaded for various potential uses, such as validating anomaly detection and data fusion as well as investigating environmental influences on train responses.

Shared decision making (SDM)-when clinicians and patients make medical decisions together-is moving swiftly from an ethical ideal toward widespread clinical implementation affecting millions of patients through recent policy initiatives. We argue that policy initiatives to promote SDM implementation in clinical practice carry the risk of several unintended negative consequences if limitations in defining and measuring SDM are not addressed. We urge policy makers to include prespecified definitions of desired outcomes, offer guidance on the tools used to measure SDM in the multitude of contexts in which it occurs, evaluate the impact of SDM policy initiatives over time, review that impact at regular intervals, and revise SDM measurement tools as needed.

This study characterizes the prevalence and content of US state statutes governing treatment decisions for decisionally incapacitated pregnant women.

BACKGROUND: Recipients of implantable cardioverter defibrillator (ICD) generator replacement with multiple medical comorbidities may be at higher risk of adverse outcomes that attenuate the benefit of ICD replacement. The aim of this investigation was to study the association between the Charlson comorbidity index (CCI) and outcomes after ICD generator replacement.

METHODS: All patients undergoing first ICD generator replacement at Mayo Clinic, Rochester and Beth Israel Deaconess Medical Center, Boston between 2001 and 2011 were identified. Outcomes included: (a) all-cause mortality, (b) appropriate ICD therapy, and (c) death prior to appropriate therapy. Multivariable Cox regression analysis was performed to assess association between CCI and outcomes.

RESULTS: We identified 1421 patients with mean age of 69.6 ± 12.1 years, 81% male and median (range) CCI of 3 (0-18). During a mean follow-up of 3.9 ± 3 years, 52% of patients died, 30.6% experienced an appropriate therapy, and 23.6% died without experiencing an appropriate therapy. In multivariable analysis, higher CCI score was associated with increased all-cause mortality (Hazard ratio, HR 1.10 [1.06-1.13] per 1 point increase in CCI, P < .001), death without prior appropriate therapy (HR 1.11 [1.07-1.15], P < .0001), but not associated with appropriate therapy (HR 1.01 [0.97-1.05], P = .53). Patients with CCI ≥5 had an annual risk of death of 12.2% compared to 8.7% annual rate of appropriate therapy.

CONCLUSIONS: CCI is predictive of mortality following ICD generator replacement. The benefit of ICD replacement in patients with CCI score ≥5 should be investigated in prospective studies.

OBJECTIVES: To more clearly define the landscape of digital medical devices subject to US Food and Drug Administration (FDA) oversight, this analysis leverages publicly available regulatory documents to characterise the prevalence and trends of software and cybersecurity features in regulated medical devices.

DESIGN: We analysed data from publicly available FDA product summaries to understand the frequency and recent time trends of inclusion of software and cybersecurity content in publicly available product information.

SETTING: The full set of regulated medical devices, approved over the years 2002-2016 included in the FDA's 510(k) and premarket approval databases.

PRIMARY AND SECONDARY OUTCOME MEASURES: The primary outcome was the share of devices containing software that included cybersecurity content in their product summaries. Secondary outcomes were differences in these shares (a) over time and (b) across regulatory areas.

RESULTS: Among regulated devices, 13.79% were identified as including software. Among these products, only 2.13% had product summaries that included cybersecurity content over the period studied. The overall share of devices including cybersecurity content was higher in recent years, growing from an average of 1.4% in the first decade of our sample to 5.5% in 2015 and 2016, the most recent years included. The share of devices including cybersecurity content also varied across regulatory areas from a low of 0% to a high of 22.2%.

CONCLUSIONS: To ensure the safest possible healthcare delivery environment for patients and hospitals, regulators and manufacturers should work together to make the software and cybersecurity content of new medical devices more easily accessible.

PURPOSE: Physical activity improves outcomes across a broad spectrum of cardiovascular disease. The safety and effectiveness of exercise-based interventions in patients with implantable cardioverter-defibrillators (ICDs) including cardiac resynchronization therapy defibrillators (CRT-Ds) remain poorly understood.

METHODS: We identified clinical studies using the following search terms: "implantable cardioverter-defibrillators"; "ICD"; "cardiac resynchronization therapy"; "CRT"; and any one of the following: "activity"; "exercise"; "training"; or "rehabilitation"; from January 1, 2000 to October 1, 2015. Eligible studies were evaluated for design and clinical endpoints.

RESULTS: A total of 16 studies were included: 8 randomized controlled trials, 5 single-arm trials, 2 observational cohort trials, and 1 randomized crossover trial. A total of 2547 patients were included (intervention groups = 1215 patients, control groups = 1332 patients). Exercise interventions varied widely in character, duration (median 84 d, range: 23-168 d), and follow-up time (median 109 d, range: 23 d to 48 mo). Exercise performance measures were the most common primary endpoints (87.5%), with most studies (81%) demonstrating significant improvement. Implantable cardioverter-defibrillator shocks were uncommon during active exercise intervention, with 6 shocks in 635 patients (0.9%). Implantable cardioverter-defibrillator shocks in follow-up were less common in patients receiving any exercise intervention (15.6% vs 23%, OR = 0.68; 95% CI, 0.48-0.80, P < .001). (Equation is included in full-text article.)O2 peak improved significantly in patients receiving exercise intervention (1.98 vs 0.36 mL/kg/min, P < .001).

CONCLUSION: In conclusion, exercise interventions in patients with ICDs and CRT-Ds appear safe and effective. Lack of consensus on design and endpoints remains a barrier to broader application to this important patient population.

RATIONALE: Multiple clinical trials support the effectiveness of cardiac resynchronization therapy (CRT); however, optimal patient selection remains challenging due to substantial treatment heterogeneity among patients who meet the clinical practice guidelines.

OBJECTIVE: To apply machine learning to create an algorithm that predicts CRT outcome using electronic health record (EHR) data avaible before the procedure.

METHODS AND RESULTS: We applied machine learning and natural language processing to the EHR of 990 patients who received CRT at two academic hospitals between 2004-2015. The primary outcome was reduced CRT benefit, defined as <0% improvement in left ventricular ejection fraction (LVEF) 6-18 months post-procedure or death by 18 months. Data regarding demographics, laboratory values, medications, clinical characteristics, and past health services utilization were extracted from the EHR available before the CRT procedure. Bigrams (i.e., two-word sequences) were also extracted from the clinical notes using natural language processing. Patients accrued on average 75 clinical notes (SD, 29) before the procedure including data not captured anywhere else in the EHR. A machine learning model was built using 80% of the patient sample (training and validation dataset), and tested on a held-out 20% patient sample (test dataset). Among 990 patients receiving CRT the mean age was 71.6 (SD, 11.8), 78.1% were male, 87.2% non-Hispanic white, and the mean baseline LVEF was 24.8% (SD, 7.69). Out of 990 patients, 403 (40.7%) were identified as having a reduced benefit from the CRT device (<0% LVEF improvement in 25.2%, death by 18 months in 15.6%). The final model identified 26% of these patients at a positive predictive value of 79% (model performance: Fβ (β = 0.1): 77%; recall 0.26; precision 0.79; accuracy 0.65).

CONCLUSIONS: A machine learning model that leveraged readily available EHR data and clinical notes identified a subset of CRT patients who may not benefit from CRT before the procedure.