BACKGROUND: The risk of all-cause mortality can inform decision-making for chronic disease prevention. We developed a predictive algorithm to estimate the 5-year risk of death among community-dwelling adults. METHODS: We derived and validated the Mortality Population Risk Tool (MPoRT) using data from population health surveys in Canada (the Canadian Community Health Survey) and the United States (the National Health Interview Survey), survey years 2001 to 2011, linked to vital statistics. The outcome was death within five years of the survey response. The algorithm was developed using data from Ontario respondents using a Cox proportional hazards model, then modified and re-estimated to allow cross-national assessment in Canada and the United States. Twenty-three prespecified predictors were assessed: seven sociodemographic, six behavioural, and ten general health and chronic disease. RESULTS: 527,369 respondents aged 20 to 105 years were included in the Canadian and United States development and validation cohorts, with 43,758 deaths during 3.68 million person-years follow-up. The final sex-specific MPoRT algorithms each contained 21 variables, showing strong discrimination (C-statistic: females 0.874 [0.871--0.877]; males 0.867 [0.865--0.871]) and good calibration overall and in 246 of 247 subgroups. Discrimination was modestly attenuated (0.01 decrease in C-statistic) in cross-national validation between Canada and the United States, with good calibration across all 71 subgroups. INTERPRETATION: MPoRT accurately discriminated all-cause mortality using only self-reported data, enabling broad application without clinical measures. While validation outside North America is needed to confirm broader applicability, MPoRT is designed for straightforward recalibration using routinely available national mortality data. This supports targeted chronic disease prevention strategies at both the population and individual levels, though the limitations inherent to self-reported predictors should be considered when interpreting predictions. ### Competing Interest Statement The authors have declared no competing interest. ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: Ottawa Health Science Network Research Ethics Board of the Ottawa Health Research Institute waived ethical approval for this work I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes Three sources of data were used for this study: Ontario CCHS (ICES Data): The dataset from this study is held securely in coded form at ICES. While data-sharing agreements prohibit ICES from making the dataset publicly available, access may be granted to those who meet pre-specified criteria for confidential access (available at www.ices.on.ca/DAS). The full dataset creation plan and underlying analytic code are available from the authors upon reasonable request. United States NHIS Data: The National Health Interview Survey (NHIS) public-use data files are freely available to researchers and the general public without special permissions or restricted access, and can be downloaded directly from the NCHS website (https://www.cdc.gov/nchs/nhis/data-questionnaires-documentation.htm). Restricted-use data files are available through the NCHS Research Data Center (RDC) subject to an application process. National Canadian CCHS Data: The Canadian version of the CCHS linked to mortality is available at Statistics Canada Regional Data Centres (RDCs). Access to these secure microdata files is restricted to affiliated researchers who apply and are approved through Statistics Canada (https://www.statcan.gc.ca/en/microdata/data-centres).

Among older adults hospitalized after a fall, receiving a DXA scan was associated with a decreased hazard of sustaining a hip fracture within 2 years. Older age, female sex, dementia, parkinsonism, mental health disorder and intermediate frailty were associated with a higher hazard of having a hip fracture post-discharge. Identify factors associated with increased hazard of having a hip fracture within 2 years of discharge following a fall-related hospitalization. We conducted a retrospective cohort study of all Ontario adults aged 65 + hospitalized after a fall between November 1, 2015 and October 31, 2020 using administrative health databases. We compared individuals who did or did not experience a hip fracture within 2 years of discharge based on socio-demographics, frailty, comorbidities, receipt of a dual-energy X-ray absorptiometry (DXA) scan, and family physician or geriatrician visits. We performed a Cox proportional hazard regression to identify factors associated with having a subsequent hip fracture, accounting for death as a competing risk. Among the 88,140 individuals who were discharged after a fall-related hospitalization, 4.6

Objective To examine the caregiving factors, sociodemographic characteristics, and self-reported health of caregivers who retired from the labor force to provide full-time, unpaid care to their care recipient. Design Matched case-control study. Setting and Participants Caregiver respondents from the Canadian Longitudinal Study on Aging baseline (2011-2015), follow-up 1 (2015-2018), and follow-up 2 (2018-2021) cycles. Caregiver respondents who indicated that they retired from the labor force to provide full-time care were matched 1:1 without replacement to caregiver respondents who did not retire to provide full-time care on birth year (±1 year), sex assigned at birth, and region of Canada. Methods Caregiving-related variables (eg, type of care and intensity, care recipient's sex, relationship to the caregiver, co-residence status), sociodemographic variables (eg, income, immigration, race, sexual orientation, marital status), and self-reported health were examined. Conditional logistic regression was used to model adjusted associations with being a caregiver who retired from the labor force to provide full-time care. Results There were 336 caregivers who retired to provide full-time care and were matched to 336 caregivers who did not retire to provide full-time care (mean age 70 years, 73% female; N = 672). Factors related to caregiving burden and type were strongly associated with retirement from the labor force to provide full-time care, whereas caregivers’ self-reported physical and mental health and household income were not associated with the decision to retire. Conclusions and Implications Caregivers' decisions to retire from the labor force were mostly driven by their care recipients’ care needs. These findings may suggest the need for flexible workplace policies that accommodate varied caregiving responsibilities to support caregivers. Increasing access to congregate residential and respite care may be another strategy that could reduce caregiver burden and retirement from the labor force.

Dementia is a major global health challenge and lifestyle modification is a key prevention strategy. Cardiovascular disease (CVD) is hypothesized to mediate lifestyle-dementia relationships, but empirical evidence is unclear. Mediation analysis offers insight into causal mechanisms beyond traditional associations. This scoping review synthesizes the limited available studies applying mediation analysis to examine whether CVD mediates associations between lifestyle factors (smoking, alcohol use, diet, physical activity) and cognitive outcomes in adults aged 45 and older. Of 1309 records screened, five studies met the inclusion criteria, reflecting a small, heterogeneous evidence base. Most examined physical activity (n = 4), with two reporting partial mediation by composite CVD risk scores. Evidence for diet (n = 2) and alcohol (n = 1) was inconclusive, and no studies assessed smoking. Overall, evidence for CVD as a mediator remains tentative, sparse, and inconsistent, highlighting major methodological gaps and an urgent need for robust studies to clarify whether cardiovascular health underpins lifestyle-related dementia risk. HIGHLIGHTS: Five studies were identified that used mediation analysis to explore the role of cardiovascular disease in the relationship between lifestyle risk factors and dementia. Cardiovascular disease may partially mediate the impact of physical activity on brain health. Diet and alcohol consumption showed no clear mediation effects by cardiovascular disease on cognition. Longitudinal, well-powered studies with robust mediation frameworks are urgently needed to evaluate vascular pathways and optimize dementia prevention strategies targeting modifiable lifestyle factors.

OBJECTIVE:To describe hospital-associated deconditioning experienced by long-term care (LTC) residents after hospitalization for a hip fracture in Ontario, Canada. DESIGN:Retrospective population-based cohort study using routinely collected data available through the Ontario Health Data Platform. SETTING AND PARTICIPANTS:LTC residents who were hospitalized for a hip fracture between June 1, 2018, and November 1, 2021. METHODS:Descriptive analyses were completed on resident age, sex, and comorbidities. We presented residents' length of hospital stay, admission to the intensive care unit, alternative level of care designation, and 30-day readmission following discharge. Deconditioning markers included both physical and psychological measures of resident functional status and cognition pre- and post-hip fracture hospitalization. RESULTS:We captured 4880 LTC residents who were hospitalized for a hip fracture between June 1, 2018, and November 1, 2021. Residents were on average 86.1 years old and predominately female (72%). Mean length of hospital stay was 6.97 days (SD 8.3), and 8% of residents died in hospital. Physical, cognitive, and psychological deconditioning were substantial postdischarge, especially in the ability to perform activities of daily living (10% dependent/totally dependent prehospitalization to 69% posthospitalization), balance while standing (20% severely impaired prehospitalization to 80% posthospitalization), health instability (6% with moderate to very high instability prehospitalization to 38% posthospitalization), and cognitive performance (1% with very severe impairment prehospitalization to 8% posthospitalization). CONCLUSIONS AND IMPLICATIONS:Markers of deconditioning demonstrate drastic declines in a variety of physical and psychological measures following hospitalization for a hip fracture among LTC residents.

INTRODUCTION:Concussions can have significant implications on the health and quality of life of older adults. As most concussion research previously focused on children, athletes and military populations, there is a need to better understand the concussion-specific treatments for adults aged 65 and older. The aim of our systematic review is to review the existing literature on the effectiveness of concussion treatments on outcomes in adults aged 65 and older. METHODS AND ANALYSIS:This systematic review will be conducted according to the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) guidelines and the Cochrane's Handbook for Systematic Reviews of Interventions. A comprehensive search of electronic databases (MEDLINE, Embase, CINAHL, AgeLine, APA PsycNet and Cochrane CENTRAL) will be performed and reference lists of included articles will be searched. We will conduct a two-step screening process and data extraction. The data analysis will integrate a narrative approach with vote-counting. The risk of bias in the included studies will be assessed, and the quality of evidence for each outcome will be evaluated using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach. ETHICS AND DISSEMINATION:The results of this systematic review will contribute to the current knowledge on concussion treatments and outcomes in older adults. This work is essential for identifying effective interventions and guiding future guidelines for this under-represented population. No ethical approval is needed for the review, and we plan to present the results at an international research conference and in a peer-reviewed journal. This protocol is registered in PROSPERO (CRD # pending).

Sleep plays an important role in cognitive function, such as emotional regulation, which is important for a higher quality of life. The study objective was to explore the association between baseline sleep quality and 6-year incidence of clinical depression. Using data from the Comprehensive cohort of the Canadian Longitudinal Study on Aging from baseline (2011-2015) to follow-up 2 (2018-2021), a total cohort of 13,997 middle aged and older Canadians free of self-reported diagnosis of clinical depression at baseline was obtained wherein the baseline mean age was 66 years (SD = 7.8), 50.3 % were female, 5.7 % were non-white, and 4.3 % had incident depression by follow-up 2. Sleep-related exposures included self-reported sleep quality (ranging from very dissatisfied to very satisfied), sleep duration, sleep frequencies (initiation difficulty, maintenance difficulty, daytime sleepiness), and symptoms of sleep disorders (sleep apnea, acting out dreams while asleep, restless leg syndrome). Multivariable regression models were used to assess the association between self-reported incident clinical depression and each sleep exposure, adjusting for several covariates (age, sex, sociodemographic, and health characteristics). Fully adjusted models demonstrated that individuals who were very dissatisfied with their sleep quality had 1.6 times the odds of reporting new depression while those with satisfied sleep had 1.2 times the odds of reporting new depression, compared to those who reported very satisfied sleep. Those with ≤4 h of sleep and those with 10+ hours of sleep had a higher predicted probability of having depression by 6-year follow-up, compared to those with 7 to 8 hours of sleep. This analysis suggests that poorer sleep quality may be associated with incident clinical depression, supporting the potential benefit of preventative measures and improvements to individuals' sleep in reducing the risk of depression.

OBJECTIVES:To examine transitions to a nursing home among residents of assisted living relative to community-dwelling home care recipients. DESIGN:Population-based retrospective cohort study emulating a target trial. SETTING AND PARTICIPANTS:Linked, individual-level health system data were obtained from older adults (aged ≥65 years) who made an incident application for a bed in a nursing home in Ontario, Canada, between April 1, 2014, and March 31, 2019, and were followed until December 31, 2019. METHODS:Residency in assisted living was compared with only community-dwelling home care. Any long-stay (≥90 days) and short-stay (<90 days) transitions to a nursing home were examined. Inverse probability weighted pooled logistic regression models were used to generate marginal cumulative incidence curves under each exposure status that were standardized by the covariates. RESULTS:This study included 10,012 residents of assisted living [mean (SD) aged 88.7 (6.26) years, 75% female] and 131,679 home care recipients [mean (SD) aged 84.8 (7.43) years, 63% female] who applied for a bed in a nursing home (N = 141,691; 95,744.6 person-years). There were 6049 transitions among applicants from assisted living and 85,190 transitions among applicants who were home care recipients to a nursing home. The 5-year absolute risk reduction was 110 transitions to a nursing home per 1000 older adult applicants if all applicants resided in assisted living (95% CI, 71-148). Residency in assisted living resulted in a 12.7% relative decrease in the 5-year risk of any transition to a nursing home had all applicants resided in assisted living (95% CI, 8.3%-17.1%). CONCLUSIONS AND IMPLICATIONS:Residents of assisted living were less likely to transition to a nursing home, despite equivalent clinical complexity and health care needs. The integration of assisted living into the continuum of care from the community to institutionalized nursing homes would better inform health system capacity and planning.

IntroductionTo develop and validate the Premature Mortality Population Risk Tool (PreMPoRT), a population-based risk algorithm that predicts the 5-year incidence of premature mortality among the Canadian adult population.MethodsRetrospective cohort analysis used six cycles of the Canadian Community Health Survey linked to the Canadian Vital Statistics Database (2000–2017). The cohort comprised 500 870 adults (18–74 years). Predictors included sociodemographic factors, self-perceived measures, health behaviours and chronic conditions. Three models (minimal, primary and full) were developed. PreMPoRT was internally validated using a split set approach and externally validated across three hold-out cycles. Performance was assessed based on predictive accuracy, discrimination and calibration.ResultsThe cohort included 267 460 females and 233 410 males. Premature deaths occurred in 1.40% of females and 2.05% of males. Primary models had 12 predictors (females) and 13 predictors (males). Shared predictors included age, income quintile, education, self-perceived health, smoking, emphysema/chronic obstructive pulmonary disease, heart disease, diabetes, cancer and stroke. Male-specific predictors were marital status, Alzheimer’s disease and arthritis while female-specific predictors were body mass index and physical activity. External validation cohort differed slightly in demographics. Female model performance: split set (c-statistic: 0.852), external (c-statistic: 0.856). Male model performance: split set and external (c-statistic: 0.846). Calibration showed slight overprediction for high-risk individuals and good calibration in key subgroups.ConclusionsPreMPoRT achieved the strongest discrimination and calibration among existing prediction models for premature mortality. The model produces reliable estimates of future incidence of premature mortality and may be used to identify subgroups who may benefit from public health interventions.

IntroductionA clinical prediction tool to estimate life expectancy in community-dwelling individuals living with dementia could inform healthcare decision-making and prompt future planning. An existing Ontario-based tool for community-dwelling elderly individuals does not perform well in people living with dementia specifically. This study seeks to develop and validate a clinical prediction tool to estimate survival in community-dwelling individuals living with dementia receiving home care in Ontario, Canada.Methods and analysisThis will be a population-level retrospective cohort study that will use data in linked healthcare administrative databases at ICES. Specifically, data that are routinely collected from regularly administered assessments for home care will be used. Community-dwelling individuals living with dementia receiving home care at any point between April 2010 and March 2020 will be included (N≈200 000). The model will be developed in the derivation cohort (N≈140 000), which includes individuals with a randomly selected home care assessment between 2010 and 2017. The outcome variable will be survival time from index assessment. The selection of predictor variables will be fully prespecified and literature/expert-informed. The model will be estimated using a Cox proportional hazards model. The model’s performance will be assessed in a temporally distinct validation cohort (N≈60 000), which includes individuals with an assessment between 2018 and 2020. Overall performance will be assessed using Nagelkerke’s R2, discrimination using the concordance statistic and calibration using the calibration curve. Overfitting will be assessed visually and statistically. Model performance will be assessed in the validation cohort and in prespecified subgroups.Ethics and disseminationThe study received research ethics board approval from the Sunnybrook Health Sciences Centre (SUN-6138). Abstracts of the project will be submitted to academic conferences, and a manuscript thereof will be submitted to a peer-reviewed journal for publication. The model will be disseminated on a publicly accessible website (www.projectbiglife.com).Trial registration numberNCT06266325(clinicaltrials.gov).

Introduction Avoidable hospitalizations are considered preventable given effective and timely primary care management and are an important indicator of health system performance. The ability to predict avoidable hospitalizations at the population level represents a significant advantage for health system decision-makers that could facilitate proactive intervention for ambulatory care-sensitive conditions (ACSCs). The aim of this study is to develop and validate the Avoidable Hospitalization Population Risk Tool (AvHPoRT) that will predict the 5-year risk of first avoidable hospitalization for seven ACSCs using self-reported, routinely collected population health survey data. Methods and analysis The derivation cohort will consist of respondents to the first 3 cycles (2000/01, 2003/04, 2005/06) of the Canadian Community Health Survey (CCHS) who are 18–74 years of age at survey administration and a hold-out data set will be used for external validation. Outcome information on avoidable hospitalizations for 5 years following the CCHS interview will be assessed through data linkage to the Discharge Abstract Database (1999/2000–2017/2018) for an estimated sample size of 394,600. Candidate predictor variables will include demographic characteristics, socioeconomic status, self-perceived health measures, health behaviors, chronic conditions, and area-based measures. Sex-specific algorithms will be developed using Weibull accelerated failure time survival models. The model will be validated both using split set cross-validation and external temporal validation split using cycles 2000–2006 compared to 2007–2012. We will assess measures of overall predictive performance (Nagelkerke R 2 ), calibration (calibration plots), and discrimination (Harrell’s concordance statistic). Development of the model will be informed by the Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD) statement. Ethics and dissemination This study was approved by the University of Toronto Research Ethics Board. The predictive algorithm and findings from this work will be disseminated at scientific meetings and in peer-reviewed publications.

BackgroundPancreatitis following endoscopic retrograde cholangiopancreatography (ERCP) can lead to significant morbidity and mortality. We aimed to develop an accurate post-ERCP pancreatitis risk prediction model using easily obtainable variables.MethodsUsing prospective multi-center ERCP data, we performed logistic regression using stepwise selection on several patient-, procedure-, and endoscopist-related factors that were determined a priori. The final model was based on a combination of the Bayesian information criterion and Akaike's information criterion performance, balancing the inclusion of clinically relevant variables and model parsimony. All available data were used for model development, with subsequent internal validation performed on bootstrapped data using 10-fold cross-validation.ResultsData from 3021 ERCPs were used to inform models. There were 151 cases of post-ERCP pancreatitis (5.0% incidence). Variables included in the final model included female sex, pancreatic duct cannulation, native papilla status, pre-cut sphincterotomy, increasing cannulation time, presence of biliary stricture, patient age, and placement of a pancreatic duct stent. The final model was discriminating, with a receiver operating characteristic curve statistic of 0.79, and well-calibrated, with a predicted risk-to-observed risk ratio of 1.003.ConclusionsWe successfully developed and internally validated a promising post-ERCP pancreatitis clinical prediction model using easily obtainable variables that are known at baseline or observed during the ERCP procedure. The model achieved an area under the curve of 0.79. External validation is planned as additional data becomes available.

Background Patients with dementia and their caregivers could benefit from advance care planning though may not be having these discussions in a timely manner or at all. A prognostic tool could serve as a prompt to healthcare providers to initiate advance care planning among patients and their caregivers, which could increase the receipt of care that is concordant with their goals. Existing prognostic tools have limitations. We seek to develop and validate a clinical prediction tool to estimate the risk of 1-year mortality among hospitalized patients with dementia. Methods The derivation cohort will include approximately 235,000 patients with dementia, who were admitted to hospital in Ontario from April 1st, 2009, to December 31st, 2017. Predictor variables will be fully prespecified based on a literature review of etiological studies and existing prognostic tools, and on subject-matter expertise; they will be categorized as follows: sociodemographic factors, comorbidities, previous interventions, functional status, nutritional status, admission information, previous health care utilization. Data-driven selection of predictors will be avoided. Continuous predictors will be modelled as restricted cubic splines. The outcome variable will be mortality within 1 year of admission, which will be modelled as a binary variable, such that a logistic regression model will be estimated. Predictor and outcome variables will be derived from linked population-level healthcare administrative databases. The validation cohort will comprise about 63,000 dementia patients, who were admitted to hospital in Ontario from January 1st, 2018, to March 31st, 2019. Model performance, measured by predictive accuracy, discrimination, and calibration, will be assessed using internal (temporal) validation. Calibration will be evaluated in the total validation cohort and in subgroups of importance to clinicians and policymakers. The final model will be based on the full cohort. Discussion We seek to develop and validate a clinical prediction tool to estimate the risk of 1-year mortality among hospitalized patients with dementia. The model would be integrated into the electronic medical records of hospitals to automatically output 1-year mortality risk upon hospitalization. The tool could serve as a trigger for advance care planning and inform access to specialist palliative care services with prognosis-based eligibility criteria. Before implementation, the tool will require external validation and study of its potential impact on clinical decision-making and patient outcomes. Trial registration NCT05371782.

OBJECTIVE:To examine transitions to an assisted living facility among community-dwelling older adults who received publicly funded home care services. DESIGN:Nested case-control study. SETTING AND PARTICIPANTS:Linked, population-level health system administrative data were obtained from adults aged 65 years and older who received home care services in Ontario, Canada, from April 1, 2018, to December 31, 2019. New residents of assisted living were matched on age, sex, and initiation date of home care (± 7 days) to community-dwelling home care recipients in a 1:4 ratio. METHODS:Clinical and functional status, health service use, sociodemographic variables, and community-level characteristics were examined; conditional logistic regression was used to model associations with a transition to an assisted living facility. RESULTS:There were 2427 new residents of assisted living who were matched to 9708 home care recipients [mean (SD) age 85.5 (6.02) years, 72% female]. Most of the new residents were concentrated in urban communities and communities with higher income quintiles. New residents had an increased rate of physician-diagnosed dementia [adjusted hazard ratio (aHR), 1.28; 95% CI, 1.14-1.43], mood disorders (aHR, 1.17; 95% CI, 1.05-1.29), and cardiac arrhythmias (aHR, 1.19; 95% CI, 1.07-1.32). They also had higher rates of mild cognitive impairment (aHR, 1.43; 95% CI, 1.24-1.66), 2 or more falls (aHR, 1.29; 95% CI, 1.11-1.51), participation in activities of long-standing interest in the past 7 days (aHR, 1.29; 95% CI, 1.11-1.50), and a lower rate of a spouse or partner unpaid caregiver vs a child (aHR, 0.66; 95% CI, 0.56-0.79). CONCLUSIONS AND IMPLICATIONS:New residents of assisted living were mostly women, were cognitively impaired, had clinical comorbidities that could increase their risk of injuries, and had caregivers who were their children. These findings stress the importance of upscaling memory and dementia care in assisted living to address the needs of this population.

Artificial intelligence (AI) has the potential to improve public health's ability to promote the health of all people in all communities. To successfully realize this potential and use AI for public health functions it is important for public health organizations to thoughtfully develop strategies for AI implementation. Six key priorities for successful use of AI technologies by public health organizations are discussed: 1) Contemporary data governance; 2) Investment in modernized data and analytic infrastructure and procedures; 3) Addressing the skills gap in the workforce; 4) Development of strategic collaborative partnerships; 5) Use of good AI practices for transparency and reproducibility, and; 6) Explicit consideration of equity and bias.

Background Modern health surveillance and planning requires an understanding of how preventable risk factors impact population health, and how these effects vary between populations. In this study, we compare how smoking, alcohol consumption, diet and physical activity are associated with all-cause mortality in Canada and the United States using comparable individual-level, linked population health survey data and identical model specifications. Methods The Canadian Community Health Survey (CCHS) (2003–2007) and the United States National Health Interview Survey (NHIS) (2000, 2005) linked to individual-level mortality outcomes with follow up to December 31, 2011 were used. Consistent variable definitions were used to estimate country-specific mortality hazard ratios with sex-specific Cox proportional hazard models, including smoking, alcohol, diet and physical activity, sociodemographic indicators and proximal factors including disease history. Results A total of 296,407 respondents and 1,813,884 million person-years of follow-up from the CCHS and 58,232 respondents and 497,909 person-years from the NHIS were included. Absolute mortality risk among those with a ‘healthy profile’ was higher in the United States compared to Canada, especially among women. Adjusted mortality hazard ratios associated with health behaviours were generally of similar magnitude and direction but often stronger in Canada. Conclusion Even when methodological and population differences are minimal, the association of health behaviours and mortality can vary across populations. It is therefore important to be cautious of between-study variation when aggregating relative effect estimates from differing populations, and when using external effect estimates for population health research and policy development.

Abstract Artificial intelligence (AI) has the potential to improve public health surveillance, health promotion, and population health management through improved targeting of interventions and policy to populations that are most in need, known as precision public health. To successfully realize this potential and use AI for public health functions it is important for public health organizations to thoughtfully develop strategies for AI implementation. Five key priorities for successful use of AI technologies by public health organizations are discussed in this commentary: 1) Contemporary data governance; 2) Investment in modernized data and analytic infrastructure and procedures; 3) Addressing the skills gap; 4) Development of strategic collaborative partnerships, and; 5) Use of AI best practices including explicit consideration of equity.

Background Most dementia algorithms are unsuitable for population-level assessment and planning as they are designed for use in the clinical setting. A predictive risk algorithm to estimate 5-year dementia risk in the community setting was developed. Methods The Dementia Population Risk Tool (DemPoRT) was derived using Ontario respondents to the Canadian Community Health Survey (survey years 2001 to 2012). Five-year incidence of physician-diagnosed dementia was ascertained by individual linkage to administrative healthcare databases and using a validated case ascertainment definition with follow-up to March 2017. Sex-specific proportional hazards regression models considering competing risk of death were developed using self-reported risk factors including information on socio-demographic characteristics, general and chronic health conditions, health behaviours and physical function. Results Among 75 460 respondents included in the combined derivation and validation cohorts, there were 8448 cases of incident dementia in 348 677 person-years of follow-up (5-year cumulative incidence, men: 0.044, 95% CI: 0.042 to 0.047; women: 0.057, 95% CI: 0.055 to 0.060). The final full models each include 90 df (65 main effects and 25 interactions) and 28 predictors (8 continuous). The DemPoRT algorithm is discriminating (C-statistic in validation data: men 0.83 (95% CI: 0.81 to 0.85); women 0.83 (95% CI: 0.81 to 0.85)) and well-calibrated in a wide range of subgroups including behavioural risk exposure categories, socio-demographic groups and by diabetes and hypertension status. Conclusions This algorithm will support the development and evaluation of population-level dementia prevention strategies, support decision-making for population health and can be used by individuals or their clinicians for individual risk assessment.

IntroductionData from population health surveys, administrative health records and environmental monitoring are increasingly being linked at the individual level. As these data become available to health researchers, there is an increasing need for methods which can make sense of large, noisy and heterogeneous data and can model complex relationships. Using these data, machine learning methods have the potential to produce population health risk algorithms with better performance than those developed with traditional statistical approaches. Objectives and ApproachThe objective of this work is to explore the use of machine learning methods for the development, validation and implementation of predictive risk algorithms designed specifically for population health planning purposes. Algorithms to predict risk of dementia and avoidable hospitalizations are in development using the Canadian Community Health Survey, geographic sociodemographic information, administrative health care utilization data and vital statistics. Methods being explored include naïve Bayes, gradient boosting, support vector machines and neural networks. ResultsRisk algorithms for population health should generally prioritize calibration over discrimination due to implications for resource allocation decisions. Approaches to minimize the risk of overfitting should be used and reweighting of unbalanced data avoided as it distorts the population-level nature of the data. It is important to be aware of propagating underlying bias in the data or exacerbating existing health inequities, which can be evaluated in part through assessment of calibration across relevant population subgroups. Approaches that consider multi-level data structures are needed to appropriately incorporate neighbourhood-level measures with individual-level information. To maximize population health impact and acceptability, model transparency and interpretability should be prioritized. ConclusionThere is tremendous potential for machine learning approaches to leverage large volumes of linked population data to produce predictive risk algorithms that will inform population health decision-making. Future work will explore use of complex environmental remote sensing and built environment data.