Population Trends of New Prescriptions for Antihyperglycemics and Antihypertensives Between 2014 and 2022.

BACKGROUND: In the wake of pandemic-related health decline and health care disruptions, there are concerns that previous gains for cardiovascular risk factors may have stalled or reversed. Population-level excess burden of drug-treated diabetes and hypertension during the pandemic compared with baseline is not well characterized. We evaluated the change in incident prescription claims for antihyperglycemics and antihypertensives before versus during the pandemic. METHODS AND RESULTS: In this retrospective, serial, cross-sectional, population-based study, we used interrupted time series analyses to examine changes in the age- and sex-standardized monthly rate of incident prescriptions for antihyperglycemics and antihypertensives in patients aged ≥66 years in Ontario, Canada, before the pandemic (April 2014 to March 2020) compared with during the pandemic (July 2020 to November 2022). Incident claim was defined as the first prescription filled for any medication in these classes. The characteristics of patients with incident prescriptions of antihyperglycemics (n=151 888) or antihypertensives (n=368 123) before the pandemic were comparable with their pandemic counterparts (antihyperglycemics, n=97 015; antihypertensives, n=146 524). Before the pandemic, monthly rates of incident prescriptions were decreasing (-0.03 per 10 000 individuals [95% CI, -0.04 to -0.01] for antihyperglycemics; -0.14 [95% CI, -0.18 to -0.10] for antihypertensives). After July 2020, monthly rates increased (postinterruption trend 0.31 per 10 000 individuals [95% CI, 0.28-0.34] for antihyperglycemics; 0.19 [95% CI, 0.14-0.23] for antihypertensives). CONCLUSIONS: Population-level increases in new antihyperglycemic and antihypertensive prescriptions during the pandemic reversed prepandemic declines and were sustained for >2 years. Our findings are concerning for current and future cardiovascular health.

The Association Between Testosterone and Vascular Function in Reproductive-Aged Females With Chronic Kidney Disease.

Cardiovascular disease is the leading cause of death in women, and women with chronic kidney disease (CKD) experience especially elevated risk. This study examined the association between testosterone and vascular function in 61 reproductive-aged females with CKD. Testosterone levels and measures of vascular function were assessed, including pulse wave velocity, aortic augmentation, flow-mediated dilation (FMD), and velocity time integral. Multivariable linear regression analyses assessed the relationship between testosterone and each measure of vascular function. No associations were observed between testosterone and vascular function outcomes, although a significant positive association between testosterone-to-estradiol ratio and FMD was demonstrated. Although testosterone levels were not independently predictive of vascular function, the level of testosterone relative to estradiol was associated with FMD and may therefore influence endothelial function in the high-risk population of reproductive-aged female patients with CKD.

Alors que les maladies cardiovasculaires sont la cause principale de décès chez les femmes, les femmes atteintes d'une maladie rénale chronique (MRC) sont exposées à un risque particulièrement élevé. La présente étude vise à examiner l'association entre la testostérone et la fonction vasculaire de 61 femmes en âge de procréer atteintes d'une MCV. Nous avons évalué les concentrations de testostérone et les mesures de la fonction vasculaire, soit la vélocité de l'onde de pouls, l'augmentation de l'aorte, la dilatation médiée par le flux (DMF) et l'intégrale temps-vitesse. Les analyses multivariées de régression linéaire ont permis d'évaluer la relation entre la testostérone et chacune des mesures de la fonction vasculaire. Aucune association n'a été observée entre la testostérone et les résultats de la fonction vasculaire, bien qu'une association positive significative entre le ratio testostérone/œstradiol et la DMF ait été démontrée. Bien que les concentrations de testostérone n'étaient pas indépendamment prédictives de la fonction vasculaire, les concentrations de la testostérone relativement à l'œstradiol ont été associées à la DMF et peuvent par conséquent influencer la fonction endothéliale au sein de la population exposée à un risque élevé composée de patientes en âge de procréer atteintes d'une MRC.

Simulation of transportation of acute stroke patients in border regions.

Determining the optimal transportation for each stroke patient is critically important to achieve the best possible outcomes. In border regions the next comprehensive stroke center may be just across an international border, but bureaucratic and financial hurdles may prevent a simple transfer to the next stroke center. We hypothesized that in regions close to international borders, patients may benefit from an "open border, closed transfer scenario", meaning that patients in whom a large vessel occlusion (LVO) is detected in the primary stroke center will benefit from a transfer to the nearest stroke center offering endovascular thrombectomy-even if this may be across a national border. We used the Swiss-German-French trinational region as an example for a region with several international borders within close proximity to one another, and compared two feasible scenarios; (a) a "closed borders, open transfer" scenario, where the patient is transported to any center in the same country, (b) an "open border, closed transfer" scenario, where patients are always transported to the nearby primary stroke center first and then to the nearest comprehensive stroke center in either the same or a neighboring country and (c) and "open borders, open transfer" scenario. The outcome of interest was the predicted probability of acute ischemic stroke patients to achieve a good outcome using a conditional probability model which predicts the likelihood of excellent outcome (modified Rankin scale score of 0-1 at 90 days post-stroke) for patients with suspected LVO. Results were modeled in a virtual map from which the ideal transport concept emerged. For an exemplary LVO stroke patient in Germany, the probability of a good outcome was higher in an open border, closed transfer scenario than with closed borders, open transfer (33.1 vs. 30.1%). Moreover, time to EVT would decrease from 232 min in the first scenario to 169 min in an open border, closed transfer scenario. The catchment area of the University Hospital Basel was almost double the size in an open border, closed transfer scenario compared to closed borders (1674 km2 vs. 2897 km2) and would receive transfers from 3 primary stroke centers in other countries (2 in Germany and 1 in France). Stroke patients showed a higher likelihood of good outcome in the "open border" scenarios without transfer restrictions to a specific healthcare system. This probably has implications for stroke treatment in all border regions where EVT eligible stroke patients may benefit from transport to the closest EVT capable center whenever possible, regardless of whether this hospital is located in the same or a neighboring country/jurisdiction.

Risk of stroke within 3, 7, 14, 21 and 30 days after influenza vaccination in Alberta, Canada: A population-based study.

BACKGROUND AND PURPOSE: Influenza vaccination is associated with a longer-term protective effect against stroke; however, it has a short-term inflammatory response which may increase short-term risk of stroke. The aim was to investigate the association between influenza vaccination and short-term risk of stroke in adults. METHODS: Administrative data were obtained from the Alberta Health Care Insurance Plan for all adults in Alberta, Canada, from September 2009 to December 2018. The hazard of any stroke (acute ischaemic stroke, intracerebral haemorrhage, subarachnoid haemorrhage and transient ischaemic attack) within 3, 7, 14, 21 and 30 days of influenza vaccination compared to unexposed time was analysed using Andersen-Gill Cox models, with adjustment for age, sex, anticoagulant use, atrial fibrillation, chronic obstructive pulmonary disease, diabetes, hypertension, income quintile, and rural or urban home location. RESULTS: In the entire cohort consisting of 4,141,209 adults (29,687,899 person-years), 1,769,565 (42.7%) individuals received at least one vaccination. In total 38,126 stroke events were recorded with 1309 occurring within 30 days of a vaccination event. Influenza vaccination was associated with a significantly reduced hazard of stroke within 3 days (hazard ratio [HR] 0.83, 95% confidence interval [CI] 0.73-0.93), 7 days (HR 0.87, 95% CI 0.80-0.95), 14 days (HR 0.87, 95% CI 0.81-0.93), 21 days (HR 0.85, 95% CI 0.80-0.91) and 30 days (HR 0.66, 95% CI 0.65-0.68). CONCLUSIONS: An increased early risk associated with vaccination was not observed. The risk of stroke was reduced at all time points within 30 days after influenza vaccination.

Sex Differences of Hospital and Emergency Department Presentations of Stroke in Canada.

We provide an updated estimate of adult stroke event rates by age group, sex, and stroke type using Canadian administrative data. In the 2017-2018 fiscal year, there were an estimated 81,781 hospital or emergency department visits for stroke events in Canada, excluding Quebec. Our findings show that overall, the event rate of stroke is similar between women and men. There were slight differences in stroke event rate at various ages by sex and stroke type and emerging patterns warrant attention in future studies. Our findings emphasize the importance of continuous surveillance to monitor the epidemiology of stroke in Canada.

Exploring sex differences for acute ischemic stroke clinical, imaging and thrombus characteristics in the INTERRSeCT study.

Women, especially following menopause, are known to have worse outcomes following acute ischemic stroke. One primary postulated biological mechanism for worse outcomes in older women is a reduction in the vasculoprotective effects of estrogen. Using the INTERRseCT cohort, a multicentre international observational cohort studying recanalization in acute ischemic stroke, we explored the effects of sex, and modifying effects of age, on neuroradiological predictors of recanalization including robustness of leptomeningeal collaterals, thrombus burden and thrombus permeability. Ordinal regression analyses were used to examine the relationship between sex and each of the neuroradiological markers. Further, we explored both multiplicative and additive interactions between age and sex. All patients (n = 575) from INTERRseCT were included. Mean age was 70.2 years (SD: 13.1) and 48.5% were women. In the unadjusted model, female sex was associated with better collaterals (OR 1.37, 95% CIs: 1.01-1.85), however this relationship was not significant after adjusting for age and relevant comorbidities. There were no significant interactions between age and sex. In a large prospective international cohort, we found no association between sex and radiological predictors of recanalization including leptomeningeal collaterals, thrombus permeability and thrombus burden.

Adverse Events and Serological Responses After SARS-CoV-2 Vaccination in Individuals With Inflammatory Bowel Disease.

INTRODUCTION: We determined adverse events after 4 doses of severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) vaccine in those with inflammatory bowel disease (IBD), associations between antibodies and injection site reactions (ISR), and risk of IBD flare. METHODS: Individuals with IBD were interviewed for adverse events to SARS-CoV-2 vaccine. Multivariable linear regression assessed the association between antibody titers and ISR. RESULTS: Severe adverse events occurred in 0.03%. ISR were significantly associated with antibody levels after the fourth dose (geometric mean ratio = 2.56; 95% confidence interval 1.18-5.57). No cases of IBD flare occurred. DISCUSSION: SARS-CoV-2 vaccines are safe for those with IBD. ISR after the fourth dose may indicate increased antibodies.

Prediction of 90 day home time among patients with low baseline ASPECTS undergoing endovascular thrombectomy: results from Alberta's Provincial Stroke Registry (QuICR).

BACKGROUND: The benefit of endovascular thrombectomy (EVT) in stroke patients with a low baseline Alberta Stroke Program Early CT Score (ASPECTS, ≤5) is uncertain. We aim to use random forest regression modeling to predict 90 day home time in patients with low ASPECTS. METHODS: We used the Quality Improvement and Clinical Research (QuICR) provincial stroke registry and administrative data from southern Alberta to identify patients who underwent EVT in our center from July 2015 to November 2020. Baseline ASPECTS on non-contrast CT and CT angiography data were scored by a two physician consensus. The primary outcome was the predicted 90 day home time (the number of nights a patient is back at their premorbid living situation without an increase in level of care within 90 days of the stroke) using random forests regression. Estimates were generated using 200 bootstrapped datasets. Covariate contribution to home time was determined using partial dependence plots. RESULTS: Of 657 EVT patients, 85 (12.9%) had baseline ASPECTS ≤5 (mean age 70.9 years, 44.7% women, 93.9% good-moderate collaterals, 60% M1-middle cerebral artery occlusion). Using partial dependence estimates, mean predicted home times were similar in the low ASPECTS (44.3 days) versus higher ASPECTS (43.1) groups. Factors predicting lower 90 day home time in this population were diabetes mellitus (-8.8 days), hypertension (-5.7 days), and atrial fibrillation (-3.6 days). There was no meaningful difference in predicted 90 day home time by sex, baseline National Institutes of Health Stroke Severity Scale score, occlusion site, tandem lesion, collateral grade or thrombolysis. CONCLUSIONS: Patients with low ASPECTS who are selected for EVT using demographic and clinical profiles similar to higher ASPECTS patients achieved comparable outcomes.

Estimating the Number of Hospital or Emergency Department Presentations for Stroke in Canada.

BACKGROUND: Although age-standardized stroke occurrence has been decreasing, the absolute number of stroke events globally, and in Canada, is increasing. Stroke surveillance is necessary for health services planning, informing research design, and public health messaging. We used administrative data to estimate the number of stroke events resulting in hospital or emergency department presentation across Canada in the 2017-18 fiscal year. METHODS: Hospitalization data were obtained from the Canadian Institute for Health Information (CIHI) Discharge Abstract Database and the Ministry of Health and Social Services in Quebec. Emergency department data were obtained from the CIHI National Ambulatory Care Reporting System (Alberta and Ontario). Stroke events were identified using ICD-10 coding. Data were linked into episodes of care to account for readmissions and interfacility transfers. Projections for emergency department visits for provinces/territories outside of Alberta and Ontario were generated based upon age and sex-standardized estimates from Alberta and Ontario. RESULTS: In the 2017-18 fiscal year, there were 108,707 stroke events resulting in hospital or emergency department presentation across the country. This was made up of 54,357 events resulting in hospital admission and 54,350 events resulting in only emergency department presentation. The events resulting in only emergency department presentation consisted of 25,941 events observed in Alberta and Ontario and a projection of 28,409 events across the rest of the country. CONCLUSIONS: We estimate a stroke event resulting in hospital or emergency department presentation occurs every 5 minutes in Canada.

Modeling optimal patient transport in a stroke network capable of remote telerobotic endovascular therapy.

BACKGROUND: Telerobotic endovascular therapy (EVT) has the potential to decrease time to treatment and expand existing networks of care to more rural populations. It is currently unclear how its implementation would impact existing stroke networks. METHODS: Conditional probability models were generated to predict the probability of excellent outcome for patients with suspected large vessel occlusion (LVO). A baseline stroke network was created for California using existing intravenous thrombolysis (IVT) centers and comprehensive stroke centers (CSCs) capable of IVT and EVT. Optimal transport decisions and catchment areas were generated for the baseline model and three hypothetical scenarios through conversion of IVT centers at various distances from a CSC into centers capable of telerobotic EVT [i.e., hospitals ≥15 and <50 miles from a CSC were converted (Scenario 1), ≥50 and <100 miles (Scenario 2), and ≥100 miles (Scenario 3)]. Procedural times and success rates were varied systematically. RESULTS: Telerobotic EVT centers decreased median travel time for LVO patients in all three scenarios. The estimated number of robotically treated LVOs per year in Scenarios 1, 2, and 3 were 2,172, 740, and 212, respectively. Scenario 1 (15-50 miles) was the most sensitive to robotic time delay and success rate, but all three scenarios were more sensitive to decreases in procedural success rate compared to time delay. CONCLUSIONS: Telerobotic EVT has the potential to improve care for stroke patients outside of major urban centers. Compared to procedural time delays in robotic EVT, a decrease in procedural success rate would not be well tolerated.

Association between influenza vaccination and risk of stroke in Alberta, Canada: a population-based study.

BACKGROUND: Respiratory infection can be an immediate precursor to stroke and myocardial infarction. Influenza vaccination is associated with reduced risk of myocardial infarction and hospitalisation for cardiac disease, and influenza vaccination is strongly recommended for patients with heart disease. Evidence on whether the same protective association exists for stroke, and whether this potential effect is consistent across age and risk groups, is conflicting. We aimed to assess the risk of stroke after influenza vaccination in adults. METHODS: We obtained administrative data from the Alberta Health Care Insurance Plan (which covers all residents of Alberta, Canada) beginning on Sept 30, 2009, or May 15 of the year in which residents were recorded as being 18 years of age. Individuals were censored at the earliest of three events: death, recorded outmigration, or Dec 31, 2018. The outcome of interest was any stroke event, comprising acute ischaemic stroke, intracerebral haemorrhage, subarachnoid haemorrhage, and transient ischaemic attack. We used Andersen-Gill Cox models to analyse the hazard of any stroke event for individuals with recent (<182 days) influenza vaccination compared with those without recent influenza vaccination, with adjustment for age, sex, anticoagulant use, atrial fibrillation, chronic obstructive pulmonary disease, diabetes, hypertension, income quintile, and rural or urban home location. Two-way interaction terms between each individual covariate and vaccination status were used to assess for effect modification by risk factor. The association between vaccination and risk of each type of stroke was also modelled, adjusting for baseline covariates. FINDINGS: The study sample consisted of 4 141 209 adults (29 687 899 person-years of observation time) registered under the provincial health-care system between Sept 30, 2009, and Dec 31, 2018. 1 769 565 (42·73%) individuals received at least one vaccination during the study period, and 38 126 stroke events were recorded. Adjusted for demographics and comorbidities, recent influenza vaccination significantly reduced the hazard of stroke (hazard ratio 0·775 [95% CI 0·757-0·793]). This association persisted across all stroke types. We found effect modification by each covariate examined except for home location; however, vaccination was associated with a reduced risk of stroke overall across all ages and risk profiles with the exception of individuals without hypertension. INTERPRETATION: The risk of stroke is reduced among people who have recently been vaccinated against influenza compared with those who have not. This association extended to the entire adult population and was not limited to individuals with a baseline high risk of stroke. Further studies in a variety of settings are needed to evaluate whether influenza vaccination could be used as a public health strategy to prevent stroke. FUNDING: None.

Quantifying Improved Outcomes, Cost Savings, and Hospital Volume Changes From Optimized Emergency Stroke Transport.

BACKGROUND: A previously published conditional probability model optimizes prehospital emergency transport protocols for patients with suspected large-vessel occlusion by recommending the transport strategy, drip-and-ship or mothership, that results in a higher probability of an excellent outcome. In this study, we create generalized models to quantify the change in annual hospital patient volume, the expected annual increase in the number of patients with an excellent outcome, and the annual cost savings to a single-payer healthcare system resulting from these optimized transport protocols. METHODS: We calculated the expected number of patients with suspected large-vessel occlusion transported by ambulance over a 1-year period in a region of interest, using the annual stroke incidence rate and a large-vessel occlusion screening tool. Assuming transport to the closest hospital is the baseline transport policy across the region (drip-and-ship), we determined the change in annual hospital patient volume from implementing optimized transport protocols. We also calculated the resulting annual increase in the number of patients with an excellent outcome (modified Rankin Score of 0-1 at 90 days) and associated cost savings to a single-payer healthcare system. We then performed a case study applying these generalized models to the stroke system serving the Greater Vancouver and Fraser Valley Area, BC, Canada. RESULTS: In the Greater Vancouver and Fraser Valley Area, there was an annual increase of 36 patients with an excellent outcome, translating to an annual cost savings of CA$2 182 824 to the British Columbia healthcare system. We also studied how these results change depending on our assumptions of treatment times at the regional stroke centers. CONCLUSIONS: Our framework quantifies the impact of optimized emergency stroke transport protocols on hospital volume, outcomes, and cost savings to a single-payer healthcare system. When applied to a specific region of interest, these models can help inform health policies concerning emergency transport of patients with suspected large-vessel occlusion.

Changes in ischemic stroke presentations, management and outcomes during the first year of the COVID-19 pandemic in Alberta: a population study.

BACKGROUND: Pandemics may promote hospital avoidance, and added precautions may exacerbate treatment delays for medical emergencies such as stroke. We sought to evaluate ischemic stroke presentations, management and outcomes during the first year of the COVID-19 pandemic. METHODS: We conducted a population-based study, using linked administrative and stroke registry data from Alberta to identify all patients presenting with stroke before the pandemic (Jan. 1, 2016 to Feb. 27, 2020) and in 5 periods over the first pandemic year (Feb. 28, 2020 to Mar. 31, 2021), reflecting changes in case numbers and restrictions. We evaluated changes in hospital admissions, emergency department presentations, thrombolysis, endovascular therapy, workflow times and outcomes. RESULTS: The study included 19 531 patients in the prepandemic period and 4900 patients across the 5 pandemic periods. Presentations for ischemic stroke dropped in the first pandemic wave (weekly adjusted incidence rate ratio [IRR] 0.54, 95% confidence interval [CI] 0.50 to 0.59). Population-level incidence of thrombolysis (adjusted IRR 0.50, 95% CI 0.41 to 0.62) and endovascular therapy (adjusted IRR 0.63, 95% CI 0.47 to 0.84) also decreased during the first wave, but proportions of patients presenting with stroke who received acute therapies did not decline. Rates of patients presenting with stroke did not return to prepandemic levels, even during a lull in COVID-19 cases between the first 2 waves of the pandemic, and fell further in subsequent waves. In-hospital delays in thrombolysis or endovascular therapy occurred in several pandemic periods. The likelihood of in-hospital death increased in Wave 2 (adjusted odds ratio [OR] 1.48, 95% CI 1.25 to 1.74) and Wave 3 (adjusted OR 1.46, 95% CI 1.07 to 2.00). Out-of-hospital deaths, as a proportion of stroke-related deaths, rose during 4 of 5 pandemic periods. INTERPRETATION: The first year of the COVID-19 pandemic saw persistently reduced rates of patients presenting with ischemic stroke, recurrent treatment delays and higher risk of in-hospital death in later waves. These findings support public health messaging that encourages care-seeking for medical emergencies during pandemic periods, and stroke systems should re-evaluate protocols to mitigate inefficiencies.

Determinants of Leptomeningeal Collateral Status Variability in Ischemic Stroke Patients.

BACKGROUND: Collateral status is an indicator of a favorable outcome in stroke. Leptomeningeal collaterals provide alternative routes for brain perfusion following an arterial occlusion or flow-limiting stenosis. Using a large cohort of ischemic stroke patients, we examined the relative contribution of various demographic, laboratory, and clinical variables in explaining variability in collateral status. METHODS: Patients with acute ischemic stroke in the anterior circulation were enrolled in a multi-center hospital-based observational study. Intracranial occlusions and collateral status were identified and graded using multiphase computed tomography angiography. Based on the percentage of affected territory filled by collateral supply, collaterals were graded as either poor (0-49%), good (50-99%), or optimal (100%). Between-group differences in demographic, laboratory, and clinical factors were explored using ordinal regression models. Further, we explored the contribution of measured variables in explaining variance in collateral status. RESULTS: 386 patients with collateral status classified as poor (n = 64), good (n = 125), and optimal (n = 197) were included. Median time from symptom onset to CT was 120 (IQR: 78-246) minutes. In final multivariable model, male sex (OR 1.9, 95% CIs [1.2, 2.9], p = 0.005) and leukocytosis (OR 1.1, 95% CIs [1.1, 1.2], p = 0.001) were associated with poor collaterals. Measured variables only explained 44.8-53.0% of the observed between-patient variance in collaterals. CONCLUSION: Male sex and leukocytosis are associated with poorer collaterals. Nearly half of the variance in collateral flow remains unexplained and could be in part due to genetic differences.

Implementation of a provincial acute stroke pathway and its impact on access to advanced stroke care in Saskatchewan.

BACKGROUND: For ischaemic stroke, outcome severity is heavily time dependent. Systems of care need to be in place to ensure that patients with stroke are treated quickly and appropriately across entire health regions. Prior to this study, the province of Saskatchewan, Canada did not have a provincial stroke strategy in place. METHODS: A quality improvement project was undertaken to create and evaluate a provincial stroke strategy. The Saskatchewan Acute Stroke Pathway was created using a multidisciplinary team of experts, piloted at five stroke centres and then implemented provincially. The number of stroke alerts, door-to-imaging, door-to-needle, door-to-groin puncture times and treatment rates were collected at all centres. Improvements over time were analysed using run charts and individuals control charts. RESULTS: The number of stroke alerts province-wide trended upwards in the last 6 months of the study. There were no clear trends or shifts in the proportion of stroke alerts treated with alteplase or endovascular therapy. Across the province, the weighted mean door-to-imaging time decreased from 21 to 15 min, the weighted mean door-to-needle time decreased from 62 to 47 min and the mean door-to-groin puncture time decreased from 83 to 70 min. There was high variability in the degree of improvement from centre to centre. CONCLUSIONS: The implementation of a province wide acute stroke pathway has led to improvement in stroke care on a provincial basis. Further work addressing intercentre variability is ongoing.

Comparing regression modeling strategies for predicting hometime.

BACKGROUND: Hometime, the total number of days a person is living in the community (not in a healthcare institution) in a defined period of time after a hospitalization, is a patient-centred outcome metric increasingly used in healthcare research. Hometime exhibits several properties which make its statistical analysis difficult: it has a highly non-normal distribution, excess zeros, and is bounded by both a lower and upper limit. The optimal methodology for the analysis of hometime is currently unknown. METHODS: Using administrative data we identified adult patients diagnosed with stroke between April 1, 2010 and December 31, 2017 in Ontario, Canada. 90-day hometime and clinically relevant covariates were determined through administrative data linkage. Fifteen different statistical and machine learning models were fit to the data using a derivation sample. The models' predictive accuracy and bias were assessed using an independent validation sample. RESULTS: Seventy-five thousand four hundred seventy-five patients were identified (divided into a derivation set of 49,402 and a test set of 26,073). In general, the machine learning models had lower root mean square error and mean absolute error than the statistical models. However, some statistical models resulted in lower (or equal) bias than the machine learning models. Most of the machine learning models constrained predicted values between the minimum and maximum observable hometime values but this was not the case for the statistical models. The machine learning models also allowed for the display of complex non-linear interactions between covariates and hometime. No model captured the non-normal bucket shaped hometime distribution. CONCLUSIONS: Overall, no model clearly outperformed the others. However, it was evident that machine learning methods performed better than traditional statistical methods. Among the machine learning methods, generalized boosting machines using the Poisson distribution as well as random forests regression were the best performing. No model was able to capture the bucket shaped hometime distribution and future research on factors which are associated with extreme values of hometime that are not available in administrative data is warranted.