Introduction: Frailty is characterized by vulnerability to declining health and increased risk for adverse health outcomes. Measuring frailty would be beneficial for developing interventions and assessing healthcare resource needs. No standardized measurement tool for frailty has been established. The objective of this thesis was to evaluate the frailty of participants in the Canadian Longitudinal Study on Aging (CLSA). Methods: A Frailty Index (FI) was constructed for CLSA participants based on the cumulative deficit theory of frailty. Exploratory factor analysis was conducted to study the underlying constructs of frailty and identify key factors. A hypothesized measurement model for frailty was specified. The model was modified and tested using structural equation modelling (SEM) to improve goodness-of-fit. A new frailty measurement tool was created and the construct validity of the new tool and the Frailty Index were evaluated. Results: A FI was calculated for 20,874 CLSA participants (Mean 0.14 SD 0.07). The maximum FI value was 0.68. A model containing all hypothesized variables had good fit of the data, and all variables contributed significantly. A simplified model also showed good fit and included four domains: upper-body strength, lower-body strength, dexterity, and depressive symptoms. These results persisted in an independent dataset. A Simplified Frailty (SF) score was created based on this simplified model. The FI and SF scores showed significant agreement and associations with sociodemographic variables were as predicted. Conclusions: A FI was simple to construct in the CLSA, having good fit of the data and construct validity. These results are consistent with previous research on the cumulative deficit theory of frailty. A simplified frailty model revealed key domains of frailty and resulted in a potentially useful short screening tool. The FI is recommended as a valid and reproducible approach for measuring frailty in the CLSA and similar population datasets. / Thesis / Master of Science (MSc)

Objectives: To evaluate associations between public health measure adherence and changes in alcohol consumption and binge drinking during the COVID-19 pandemic and identify potential modifiers of these associations. Methods: A longitudinal study was conducted using the Canadian Longitudinal Study on Aging Baseline, Follow-Up 1, and COVID-19 Questionnaires (N=23,615). Multinomial logistic regression was used to examine associations between public health measure adherence (including self-quarantine, public gathering attendance, leaving home, mask-wearing, and handwashing) and 1) self-reported alcohol consumption changes during the first year of the pandemic 2) prospectively measured alcohol consumption changes from 2015-2018 to early pandemic (2020), and 3) prospectively measured changes in binge drinking events from 2015-2018 to early pandemic. Models were adjusted for sex, age, education, marital status, region, income, immigration status, and racial background. Results: During the first year of the pandemic, 13% (N=2,733) of participants self-reported increased alcohol use, while 13% (N=2,921) reported decreased use. Prospective measures suggested 19% (N=4,421) increased and 35% (N=7,971) decreased use, while 13% (N=1,427) increased and 18% (N=1,953) decreased binge drinking events. Compared to low public health measure adherence, high adherence was not associated with self-reported changes in alcohol use, with prospectively measured increased alcohol use, or with prospectively measured changes in binge drinking. However, high adherence was associated with higher odds of prospectively measured decreased alcohol use (aOR=1.17; 95% CI=1.05, 1.29). Conclusion: Higher public health measure adherence during the COVID-19 pandemic was not associated with increased alcohol use, suggesting that adherence did not negatively impact alcohol consumption. Findings may inform public health interventions throughout future pandemics. / Thesis / Master of Public Health (MPH)

Background: Obesity has been associated with increased health care use, but it is unclear whether this is consistent across all measures of obesity. The objectives of this thesis were to compare obesity defined by four anthropometric measures, body mass index (BMI), waist circumference (WC), waist-to-hip ratio (WHR), and percent body fat (%BF), and to estimate their associations with health care use among Canadian adults. Methods: A secondary data analysis was conducted from 30,097 individuals aged 45-85 years from the Canadian Longitudinal Study on Aging. Anthropometric measures were collected by trained research assistants and %BF, the reference standard, was measured using dual-energy x-ray absorptiometry. Obesity was defined as BMI≥30.0 kg/m2, WC≥88cm for females and ≥102cm for males, WHR≥0.85 for females and ≥0.90 for males, and %BF>35% for females and >25% for males. Approximately 18 months after baseline data collection, self-reported health care use in the past 12 months was collected, including any contact with a general practitioner, medical specialist, emergency department, and being a patient in a hospital overnight. Pearson correlation coefficients and sensitivity and specificity analyses were conducted to compare anthropometric measures to %BF. Relative risks and risk differences were calculated for measures of health care use, adjusted for sex, age, education, income, urban/rural, marital status, smoking status, and alcohol use. Secondary analyses were also stratified by sex and age. Results: The prevalence of obesity defined by BMI was 29%, by WC was 42%, by WHR was 62%, and by %BF was 73%. BMI and WC were highly correlated with %BF (r=0.75 and r=0.70, respectively) and WHR was weakly correlated with %BF (r=0.29). BMI and WC cut points demonstrated high specificity (>93%) and lower sensitivity (<58%) in predicting obesity defined by %BF. WHR cut points demonstrated high sensitivity (95%) and lower specificity (28%) in males, but lower sensitivity (44%) and high specificity (83%) in females in predicting %BF- defined obesity. There was an increased relative and absolute risk of health care use for all measures of obesity and all health care services. For example, WC-defined obesity was associated with increased relative risk (RR) of hospital overnight stay (RR: 1.40, 95% CI: 1.28- 1.54) and the risk difference (per 100) was 2.6 (95% CI:1.9-3.3). The risk of health care use was similar amongst females and males with obesity although relative risks and risk differences attenuated in the oldest adult group aged 75 and older compared to the youngest group aged 45- 54. Conclusion: The prevalence of obesity among Canadian adults varied substantially by anthropometric measure. BMI and WC have stronger correlations and concordance with %BF than does WHR, however all measures were positively associated with increased health care use. Further research should be conducted on obesity cut points to discern the best measure to predict health care use. / Thesis / Master of Public Health (MPH)

The business performance management (BPM) framework helps an organization continuously adjust and successfully execute its strategies. BPM helps increase flexibility by providing managers with an early alert about changes and, as a result, allows faster response to such changes. The Malcolm Baldrige National Quality Award (MBNQA) framework provides a basis for self-assessment and a systems perspective for managing an organization's key processes for achieving business results. The MBNQA framework is a more comprehensive framework and encapsulates the underlying constructs in the BPM framework. The objectives of this dissertation are fourfold: (1) to validate the underlying relationships presented in the 2008 MBNQA framework, (2) to explore the MBNQA framework at the dimension level, and develop and test constructs measured at that level in a causal model, (3) to validate and create a common general framework for the business performance model by integrating the practitioner literature with basic theory including existing MBNQA theory, and (4) to integrate the BPM framework and the MBNQA framework into a new framework (BPM-MBNQA framework) that can guide organizations in their journey toward achieving and sustaining competitive and strategic advantages. The purpose of this study is to achieve these objectives by means of a combination of methodologies including literature reviews, expert opinions, interviews, presentation feedbacks, content analysis, and latent semantic analysis. An initial BPM framework was developed based on the reviews of literature and expert opinions. There is a paucity of academic research on business performance management. Therefore, this study reviewed the practitioner literature on BPM and from the numerous organization-specific BPM models developed a generic, conceptual BPM framework. With the intent of obtaining valuable feedback, this initial BPM framework was presented to Baldrige Award recipients (BARs) and selected academicians from across the United States who participated in the Fall Summit 2007 held at Caterpillar Financial Headquarter in Nashville, TN on October 1 and 2, 2007. Incorporating the feedback from that group allowed refining and improving the proposed BPM framework. This study developed a variant of the traditional latent semantic analysis (LSA) called causal latent semantic analysis (cLSA) that enables us to test causal models using textual data. This method was used to validate the 2008 MBNQA framework based on article abstracts on the Baldrige Award and program published in both practitioner and academic journals from 1987 to 2009. The cLSA was also used to validate the BPM framework using the full body text data from all articles published in the practitioner journal entitled the Business Performance Management Magazine since its inception in 2003. The results provide the first cLSA study of these frameworks. This is also the first study to examine all the causal relationships within the MBNQA and BPM frameworks.

Introduction: L’inflammation chronique et l’hypercoagulabilité associées au cancer favorisent la survenue de thromboembolie. L’accident vasculaire cérébral (AVC) ischémique peut être le premier signe d’un cancer actif non diagnostiqué (ou occulte). La fréquence et les prédicteurs de cancer occulte en AVC ischémique demeurent cependant débattus. Nous avons d’abord effectué une revue systématique de la littérature afin de résumer les connaissances sur la fréquence et les prédicteurs de cancer en AVC ischémique. Nous avons ensuite effectué une étude de cohorte rétrospective appariée pour comparer le risque de cancer chez les individus ayant subi un AVC ischémique ou un accident ischémique transitoire (AIT) à celui d’individus sans AVC/AIT à travers des données de l’Étude longitudinale canadienne sur le vieillissement. Méthodes: Dans notre revue systématique, nous avons interrogé sept bases de données à la recherche d’articles publiés entre janvier 1980 et septembre 2019 rapportant des tumeurs malignes et des néoplasies myéloprolifératives diagnostiquées après un AVC ischémique (protocole PROSPERO: CRD42019132455). Dans notre cohorte appariée, nous avons utilisé les données de la cohorte globale (n=30 097) de l’Étude longitudinale canadienne sur le vieillissement, une grande cohorte populationnelle d’individus âgés de 45 à 85 ans au recrutement (2011 à 2015). Nous avons construit une cohorte rétrospective par appariement individuel exact sur l’âge (ratio 1:4) et avons utilisé des modèles à risques proportionnels de Cox pour estimer les rapports de risques instantanés de nouveau diagnostic de cancer avec et sans AVC/AIT préalable. Résultats: Pour notre revue systématique, nous avons dépisté 15 400 entrées et inclus 51 articles. L’incidence cumulée combinée de cancer dans la première année suivant un AVC ischémique était de 13,6 par millier (intervalle de confiance à 95% [IC 95%]: 5,6 à 24,8), plus élevée pour les études d’AVC cryptogénique (62,0 par millier; IC 95%: 13,6 à 139,3 vs 9,6 par millier; IC 95%: 4,0 à 17,3; p-value=0,02) et pour celles rapportant des tests de dépistage du cancer (39,2 par millier; IC 95%: 16,4 à 70,6 vs 7,2 par millier; IC 95%: 2,5 à 14,1; p-value=0,003). L’incidence de cancer après un AVC était généralement supérieure par rapport aux individus sans AVC et la plupart des cancers étaient diagnostiqués dans les premiers mois suivants l’AVC. Nous avons identifié plusieurs prédicteurs de cancer occulte, dont l’âge avancé, le tabagisme, l’infarctus de plusieurs territoires vasculaires cérébraux ainsi que l’élévation des d-dimères et de la protéine C-réactive. Pour notre étude de cohorte, nous avons respectivement inclus 920 et 3 680 individus dans les groupes avec et sans AVC/AIT. Nous avons observé une incidence supérieure de cancer dans la première année suivant l’AVC/AIT qui diminuait par la suite. Le risque instantané de nouveau diagnostic de cancer dans la première année suivant un AVC/AIT était significativement augmenté (rapport de risques instantanés=2,36; IC 95%: 1,21 à 4,61; p-value=0,012) par rapport aux individus appariés pour l’âge après ajustements. Les principaux types de cancer dans la première année étaient le cancer de la prostate (n=8, 57,1%) et le mélanome (n=2, 14,3%). Conclusion: Nous avons observé dans notre revue systématique et notre étude de cohorte une incidence de nouveau diagnostic de cancer suivant un AVC ischémique globalement faible, mais supérieure à celle d’individus sans AVC. La fréquence de nouveau diagnostic de cancer après un AVC était également supérieure en AVC cryptogénique et après un dépistage. Plusieurs prédicteurs peuvent être utilisés pour augmenter la probabilité prétest de cancer occulte en AVC ischémique. Toutefois, l’incidence de cancer post-AVC que nous rapportons est probablement sous-estimée en raison de limites méthodologiques des études méta-analysées. Des études prospectives de plus grande taille avec documentation systématique des diagnostics de cancer post-AVC sont nécessaires pour produire des estimations plus valides et précises qui pourront guider l’élaboration d’études randomisées et contrôlées de détection précoce de cancer en AVC ischémique. / Introduction: Cancer promotes thromboembolism through inflammation and hypercoagulability, and an ischemic stroke may be the first sign of an undiagnosed (occult) malignancy. The frequency and predictors of occult cancer in people with acute ischemic stroke, however, remains unclear. We first sought to summarize the existing published data regarding the frequency and predictors of cancer after an ischemic stroke in a systematic review. We also conducted a retrospective matched cohort study to compare the incidence of cancer in people who experienced an ischemic stroke or transient ischemic attack (TIA) to that of people without stroke, using data from the Canadian Longitudinal Study on Aging. Methods: For our systematic review, we searched seven databases from January 1980 to September 2019 for articles reporting malignant tumors and myeloproliferative neoplasms diagnosed after an ischemic stroke (PROSPERO protocol: CRD42019132455). For our matched cohort study, we used data from the comprehensive sub-group (n=30,097) of the Canadian Longitudinal Study on Aging, a large population-based cohort of individuals aged 45-85 years when recruited (2011-2015). We built a retrospective cohort with individual exact matching for age (1:4 ratio). We used Cox proportional hazards models to estimate hazard ratios of new cancer diagnosis with and without a prior stroke/TIA. Results: For our systematic review, we screened 15,400 records and included 51 articles. The pooled cumulative incidence of cancer within one year after an ischemic stroke was 13.6 per thousand (95% confidence interval [CI], 5.6 to 24.8), higher in studies focusing on cryptogenic stroke (62.0 per thousand; 95% CI, 13.6 to 139.3 vs 9.6 per thousand; 95% CI, 4.0 to 17.3; p- value=0.02) and those reporting cancer screening (39.2 per thousand; 95% CI, 16.4 to 70.6 vs 7.2 per thousand; 95% CI, 2.5 to 14.1; p-value=0.003). The incidence of cancer after stroke was higher overall compared to people without stroke. Most cases were diagnosed within the first few months after stroke. Several predictors of cancer were identified, namely older age, smoking, involvement of multiple vascular territories, as well as elevated C-reactive protein and d-dimers. For our cohort study, we respectively included 920 and 3,680 individuals in the stroke and non- stroke groups. We observed a higher incidence of cancer in the first year after stroke/TIA that declined afterwards. The hazard of new cancer diagnosis in the first year after stroke/TIA was significantly increased (hazard ratio=2.36; 95% CI, 1.21 to 4.61; p-value=0.012) as compared to age-matched non-stroke participants after adjustments. The most frequent primary cancers in the first year after stroke/TIA were prostate (n=8, 57.1%) and melanoma (n=2, 14.3%). Conclusion: We observed in both studies of our research project that the frequency of incident cancer after an ischemic stroke is low overall, but higher as compared to people without stroke. The frequency of new cancer diagnosis after stroke is also higher in cryptogenic stroke and after cancer screening. Several predictors may increase the yield of cancer screening after an ischemic stroke. The pooled incidence of post-stroke cancer is likely underestimated due to methodological issues in most studies of our review. Larger prospective studies with systematic ascertainment of cancer after stroke are needed to produce more valid and precise estimates of post-stroke cancer risk and guide randomized controlled studies of cancer screening in people with acute ischemic stroke.

No description available.