Explore the vast range of titles available.

Innovative models of collaborative palliative care are urgently needed to meet gaps in end-of-life care among people with heart failure. We sought to determine whether regionally organized, collaborative, home-based palliative care that involves cardiologists, primary care providers and palliative care specialists, and that uses shared decision-making to promote goal- and need-concordant care for patients with heart failure, was associated with a greater likelihood of patients dying at home than in hospital. We conducted a population-based matched cohort study of adults who died with chronic heart failure across 2 large health regions in Ontario, Canada, between 2013 and 2019. The primary outcome was location of death. Secondary outcomes included rates of health care use, including unplanned visits to the emergency department, hospital admissions, hospital lengths of stay, admissions to the intensive care unit, number of visits with primary care physicians or cardiologists, number of home visits by palliative care physicians or nurse practitioners, and number of days spent at home. Patients who received regionally organized, collaborative, home-based palliative care (n = 245) had a 48% lower associated risk of dying in hospital (relative risk 52%, 95% confidence interval 44%–66%) compared with the matched cohort (n = 1172) who received usual care, with 101 (41.2%) and 917 (78.2%) patients, respectively, dying in hospital (number needed to treat = 3). Additional associated benefits of the collaborative approach included higher rates of clinician home visits, longer time to first hospital admission, shorter hospital stays and more days spent at home. Adoption of a model of regionally organized, collaborative, home-based palliative care that uses shared decision-making may improve end-of-life outcomes for people with chronic heart failure.

Background Supervised machine learning is increasingly being used to estimate clinical predictive models. Several supervised machine learning models involve hyper-parameters, whose values must be judiciously specified to ensure adequate predictive performance. Objective To compare several (nine) hyper-parameter optimization (HPO) methods, for tuning the hyper-parameters of an extreme gradient boosting model, with application to predicting high-need high-cost health care users. Methods Extreme gradient boosting models were estimated using a randomly sampled training dataset. Models were separately trained using nine different HPO methods: 1) random sampling, 2) simulated annealing, 3) quasi-Monte Carlo sampling, 4-5) two variations of Bayesian hyper-parameter optimization via tree-Parzen estimation, 6-7) two implementations of Bayesian hyper-parameter optimization via Gaussian processes, 8) Bayesian hyper-parameter optimization via random forests, and 9) the covariance matrix adaptation evolutionary strategy. For each HPO method, we estimated 100 extreme gradient boosting models at different hyper-parameter configurations; and evaluated model performance using an AUC metric on a randomly sampled validation dataset. Using the best model identified by each HPO method, we evaluated generalization performance in terms of discrimination and calibration metrics on a randomly sampled held-out test dataset (internal validation) and a temporally independent dataset (external validation). Results The extreme gradient boosting model estimated using default hyper-parameter settings had reasonable discrimination (AUC=0.82) but was not well calibrated. Hyper-parameter tuning using any HPO algorithm/sampler improved model discrimination (AUC=0.84), resulted in models with near perfect calibration, and consistently identified features predictive of high-need high-cost health care users. Conclusions In our study, all HPO algorithms resulted in similar gains in model performance relative to baseline models. This finding likely relates to our study dataset having a large sample size, a relatively small number of features, and a strong signal to noise ratio; and would likely apply to other datasets with similar characteristics.

Canada's approach to refugee resettlement includes government sponsorship, a pioneering private sponsorship model and a third blended approach. Refugees are selected and supported differently in each approach including healthcare navigation. Little is known about how well private sponsors facilitate primary care navigation and whether this changed during the large-scale 2015 Syrian resettlement initiative characterized by civic and healthcare systems engagement. Population-based cohort study of resettled refugees arriving in Ontario between April 1, 2008 and March 31, 2017, with one-year follow-up, using linked health and demographic administrative databases. We evaluated associations of resettlement model (GARs, Privately Sponsored Refugees [PSRs], and Blended-Visa Office Referred [BVORs]) by era of arrival (pre-Syrian and Syrian era) and by country cohort, on measures of primary care (PC) navigation using adjusted Cox proportional hazards and logistic regression. There were 34,591 (pre-Syrian) and 24,757 (Syrian era) resettled refugees, approximately half of whom were GARs. Compared with the reference group pre-Syrian era PSRs, Syrian PSRs had slightly earlier PC visits (mean = 116 days [SD = 90]) (adjusted hazard ratios [aHR] = 1.19, 95% CI 1.14-1.23). Syrian GARs (mean = 72 days [SD = 65]) and BVORs (mean = 73 days [SD = 76]) had their first PC visit sooner than pre-Syrian era PSRs (mean = 149 days [SD = 86]), with respective aHRs 2.27, 95% CI 2.19-2.35 and 1.89, 95% CI 1.79-1.99. Compared to pre-Syrian PSRs, Syrian GARs and BVORs had much greater odds of a CHC visit (adjusted odds ratios 14.69, 95% CI 12.98-16.63 and 14.08, 95% 12.05-16.44 respectively) and Syrian PSRs had twice the odds of a CHC visit. Less timely primary care and lower odds of a CHC visit among PSRs in the first year may be attributed to selection factors and gaps in sponsors' knowledge of healthcare navigation. Improved primary care navigation outcomes in the Syrian era suggests successful health systems engagement.

Whether virtual health care has changed access to services for patients living far from a specialist physician is unknown. We aimed to determine whether driving distances between patients and their specialists had changed following increased availability of virtual care in Ontario, such that specialists saw patients from farther away. We performed a population-based cohort study using linked health and administrative databases. We included all specialist physicians working in Ontario from Jan. 1, 2019, to Nov. 30, 2019 (pre–virtual care period) and from Jan. 1, 2022, to Nov. 30, 2022 (virtual care period), and their patients. Outcomes were measures of proximity between specialists and their patients including differences in 90th-percentile driving distance, mean driving time, and the proportion of patients with driving times longer than 60 minutes between time periods. We used multivariable linear regression models to compare outcomes across physician specialties, adjusting for physician age, sex, practice size, and location. We included 11 096 specialists (4232 surgical and 6864 medical; 0.8% rural). After adjustment, we found no meaningful changes in the 90th-percentile driving distance between time periods for surgical (difference 6.7 km, 95% confidence interval [CI] −4.1 km to 17.5 km) or medical specialties (difference 1.3 km, 95% CI −6.6 km to 9.2 km). For surgical specialists, the proximity measures of mean driving time increased by 5 minutes (95% CI 1 min to 10 min) and the proportion of patients living more than 60 minutes away increased by 2.1% (95% CI 0.7% to 3.9%), but we saw no significant change for medical specialists. After expansion of virtual care, the distance between specialists and patients did not meaningfully change. To make virtual care more accessible, especially for those living in rural areas, attention should be paid to other factors such as referral patterns and the role of patients in determining the type of visit they prefer.

ABSTRACTBackgroundThe implementation and clinical impact of machine learning–based early warning systems for patient deterioration in hospitals have not been well described. We sought to describe the implementation and evaluation of a multifaceted, real-time, machine learning–based early warning system for patient deterioration used in the general internal medicine (GIM) unit of an academic medical centre. MethodsIn this nonrandomized, controlled study, we evaluated the association between the implementation of a machine learning–based early warning system and clinical outcomes. We used propensity score–based overlap weighting to compare patients in the GIM unit during the intervention period (Nov. 1, 2020, to June 1, 2022) to those admitted during the pre-intervention period (Nov. 1, 2016, to June 1, 2020). In a difference-indifferences analysis, we compared patients in the GIM unit with those in the cardiology, respirology, and nephrology units who did not receive the intervention. We retrospectively calculated system predictions for each patient in the control cohorts, although alerts were sent to clinicians only during the intervention period for patients in GIM. The primary outcome was non-palliative in-hospital death. ResultsThe study included 13 649 patient admissions in GIM and 8470 patient admissions in subspecialty units. Non-palliative deaths were significantly lower in the intervention period than the pre-intervention period among patients in GIM (1.6% v. 2.1%; adjusted relative risk [RR] 0.74, 95% confidence interval [CI] 0.55–1.00) but not in the subspecialty cohorts (1.9% v. 2.1%; adjusted RR 0.89, 95% CI 0.63–1.28). Among high-risk patients in GIM for whom the system triggered at least 1 alert, the proportion of non-palliative deaths was 7.1% in the intervention period, compared with 10.3% in the pre-intervention period (adjusted RR 0.69, 95% CI 0.46–1.02), with no meaningful difference in subspecialty cohorts (10.4% v. 10.6%; adjusted RR 0.98, 95% CI 0.60–1.59). In the difference-indifferences analysis, the adjusted relative risk reduction for non-palliative death in GIM was 0.79 (95% CI 0.50–1.24). InterpretationImplementing a machine learning–based early warning system in the GIM unit was associated with lower risk of non-palliative death than in the pre-intervention period. Machine learning–based early warning systems are promising technologies for improving clinical outcomes.

Youth have reported worsening mental health during the COVID-19 pandemic. We sought to evaluate rates of pediatric acute care visits for self-harm during the pandemic according to age, sex and mental health service use. We conducted a population-based, repeated cross-sectional study using linked health administrative data sets to measure monthly rates of emergency department visits and hospital admissions for self-harm among youth aged 10–17 years between Jan. 1, 2017, and June 30, 2022, in Ontario, Canada. We modelled expected rates of acute care visits for self-harm after the pandemic onset based on prepandemic rates. We reported relative differences between observed and expected monthly rates overall and by age group (10–13 yr and 14–17 yr), sex and mental health service use (new and continuing). In this population of about 1.3 million children and adolescents, rates of acute care visits for self-harm during the pandemic were higher than expected for emergency department visits (0.27/1000 population v. 0.21/1000 population; adjusted rate ratio [RR] 1.29, 95% confidence interval [CI] 1.19–1.39) and hospital admissions (0.74/10 000 population v. 0.43/10 000 population, adjusted RR 1.72, 95% CI 1.46–2.03). This increase was primarily observed among females. Rates of emergency department visits and hospital admissions for self-harm were higher than expected for both those aged 10–13 years and those aged 14–17 years, as well as for both those new to the mental health system and those already engaged in care. Rates of acute care visits for self-harm among children and adolescents were higher than expected during the first 2 and a half years of the COVID-19 pandemic, particularly among females. These findings support the need for accessible and intensive prevention efforts and mental health supports in this population.

To assess how changes in outpatient services during the first year of the COVID-19 pandemic were related to acute healthcare use (emergency department or hospitalizations) for individuals with asthma or chronic obstructive pulmonary disease (COPD). We conducted an observational study using health administrative data in Ontario (Canada) from January 2016 to March 2021 on all adults with diagnosed asthma or COPD. We used monthly time series auto-regressive integrated moving-average (ARIMA) and pre-pandemic monthly rates (January 2016 to February 2020) to calculate projected rates (i.e., a pandemic had not occurred) during the pandemic (March 2020 to March 2021), and Quasi-Poisson models with two-way interaction to estimate crude and adjusted rate ratios. In the first pandemic year, in individuals with asthma or COPD, outpatient visit rates started lower than projected (Mar-May 2020), returned to projected in the middle of the year (Jun-Aug 2020) and then rose to higher than projected between Sep 2020 and Mar 2021: observed rates of 80,293 per 100,000 persons vs. projected 74,192 (95% CI: 68,926-79,868) in individuals with asthma, and 92,651 vs. projected 85,871 (95% CI: 79,975-92,207) in individuals with COPD. Acute care rates remained below projected during the first pandemic year. While pulmonary function test (PFT) rates remained below projected during the first pandemic year, in both populations, a decrease in acute care visits during the pandemic, compared to pre-pandemic, was noted during months with the highest PFT rates (interaction p-values < 0.0001). Despite asthma and COPD being ambulatory-care sensitive conditions, lower rates of outpatient visits during the beginning of the pandemic were not associated with increased rates of acute care use. Lower PFT rates were associated with higher acute care visit rates, suggesting that access to PFT during pandemic is likely important for individuals with asthma or COPD.

Background Topic models are a class of unsupervised machine learning models, which facilitate summarization, browsing and retrieval from large unstructured document collections. This study reviews several methods for assessing the quality of unsupervised topic models estimated using non-negative matrix factorization. Techniques for topic model validation have been developed across disparate fields. We synthesize this literature, discuss the advantages and disadvantages of different techniques for topic model validation, and illustrate their usefulness for guiding model selection on a large clinical text corpus. Design, setting and data Using a retrospective cohort design, we curated a text corpus containing 382,666 clinical notes collected between 01/01/2017 through 12/31/2020 from primary care electronic medical records in Toronto Canada. Methods Several topic model quality metrics have been proposed to assess different aspects of model fit. We explored the following metrics: reconstruction error, topic coherence, rank biased overlap, Kendall’s weighted tau, partition coefficient, partition entropy and the Xie-Beni statistic. Depending on context, cross-validation and/or bootstrap stability analysis were used to estimate these metrics on our corpus. Results Cross-validated reconstruction error favored large topic models (K ≥ 100 topics) on our corpus. Stability analysis using topic coherence and the Xie-Beni statistic also favored large models (K = 100 topics). Rank biased overlap and Kendall’s weighted tau favored small models (K = 5 topics). Few model evaluation metrics suggested mid-sized topic models (25 ≤ K ≤ 75) as being optimal. However, human judgement suggested that mid-sized topic models produced expressive low-dimensional summarizations of the corpus. Conclusions Topic model quality indices are transparent quantitative tools for guiding model selection and evaluation. Our empirical illustration demonstrated that different topic model quality indices favor models of different complexity; and may not select models aligning with human judgment. This suggests that different metrics capture different aspects of model goodness of fit. A combination of topic model quality indices, coupled with human validation, may be useful in appraising unsupervised topic models.

Increased rates of pediatric eating disorders have been observed during the COVID-19 pandemic, but little is known about trends among adults. We aimed to evaluate rates of emergency department visits and hospital admissions for eating disorders among adolescents and adults during the pandemic. We conducted a population-based, repeated cross-sectional study using linked health administrative data for Ontario residents aged 10–105 years during the prepandemic (Jan. 1, 2017, to Feb. 29, 2020) and pandemic (Mar. 1, 2020, to Aug. 31, 2022) periods. We evaluated monthly rates of emergency department visits and hospital admissions for eating disorders, stratified by age. Compared with expected rates derived from the prepandemic period, emergency department visits for eating disorders increased during the pandemic among adolescents aged 10–17 years (7.38 v. 3.33 per 100 000; incidence rate ratio [IRR] 2.21, 95% confidence interval [CI] 2.17–2.26), young adults aged 18–26 years (2.79 v. 2.46 per 100 000; IRR 1.13, 95% CI 1.10–1.16) and older adults aged 41–105 years (0.14 v. 0.11 per 100 000; IRR 1.15, 95% CI 1.07–1.24). Hospital admissions for eating disorders increased during the pandemic for adolescents (8.82 v. 5.74 per 100 000; IRR 1.54, 95% CI 1.54–1.54) but decreased for all adult age groups, especially older adults aged 41–105 years (0.21 v. 0.30 per 100 000; IRR 0.72, 95% CI 0.64–0.80). Emergency department visits for eating disorders increased among adolescents, young adults and older adults during the pandemic, but hospital admissions increased only for adolescents and decreased for all adult groups. Differential rates of acute care use for eating disorders by age have important implications for allocation of inpatient mental health resources.

Research has demonstrated that there is lower operative mortality at hospitals with higher surgical volume. Using administrative data from Medicare, this study found lower mortality associated with each of eight procedures when performed by surgeons who undertook the operation more frequently. Lower mortality with surgeons who operate frequently. For many surgical procedures, patients at hospitals where a high number of such procedures are performed (high-volume hospitals) have lower mortality rates than those at hospitals that are less experienced with the procedures. 1 – 4 In one recent study of the national population of Medicare recipients, we found strong relations between hospital volume and operative mortality associated with 14 high-risk cancer resections and cardiovascular procedures. 5 Despite the considerable body of research in this area, little is known about the mechanisms underlying the observed associations between volume and outcome. Because they tend to be much larger facilities, high-volume hospitals have a broader . . .