Single-dose Effectiveness of Mpox Vaccine in Quebec, Canada: Test-negative Design With and Without Adjustment for Self-reported Exposure Risk.

INTRODUCTION: During the 2022 mpox outbreak, the province of Quebec, Canada, prioritized first doses for pre-exposure vaccination of people at high mpox risk, delaying second doses due to limited supply. We estimated single-dose mpox vaccine effectiveness (VE) adjusting for virus exposure risk based only on surrogate indicators available within administrative databases (eg, clinical record of sexually transmitted infections) or supplemented by self-reported risk factor information (eg, sexual contacts). METHODS: We conducted a test-negative case-control study between 19 June and 24 September 2022. Information from administrative databases was supplemented by questionnaire collection of self-reported risk factors specific to the 3-week period before testing. Two study populations were assessed: all within the administrative databases (All-Admin) and the subset completing the questionnaire (Sub-Quest). Logistic regression models adjusted for age, calendar-time and exposure-risk, the latter based on administrative indicators only (All-Admin and Sub-Quest) or with questionnaire supplementation (Sub-Quest). RESULTS: There were 532 All-Admin participants, of which 199 (37%) belonged to Sub-Quest. With exposure-risk adjustment based only on administrative indicators, single-dose VE estimates were similar among All-Admin and Sub-Quest populations at 35% (95% confidence interval [CI]:-2 to 59) and 30% (95% CI:-38 to 64), respectively. With adjustment supplemented by questionnaire information, the Sub-Quest VE estimate increased to 65% (95% CI:1-87), with overlapping confidence intervals. CONCLUSIONS: Using only administrative data, we estimate one vaccine dose reduced the mpox risk by about one-third; whereas, additionally adjusting for self-reported risk factor information revealed greater vaccine benefit, with one dose instead estimated to reduce the mpox risk by about two-thirds. Inadequate exposure-risk adjustment may substantially under-estimate mpox VE.

Traditional methods hold their ground against machine learning in predicting potentially inappropriate medication use in older adults.

OBJECTIVES: Machine learning methods have gained much attention in health sciences for predicting various health outcomes but are scarcely used in pharmacoepidemiology. The ability to identify predictors of suboptimal medication use is essential for conducting interventions aimed at improving medication outcomes. It remains uncertain whether machine learning methods could enhance the identification of potentially inappropriate medication use among older adults compared to traditional methods. The aim of this study was 1) to compare the performances of machine learning models in predicting use of potentially inappropriate medications and 2) to quantify and compare the relative importance of predictors in a population of community dwelling older adults (>65 years) in the province of Quebec, Canada. METHODS: We used the Quebec Integrated Chronic Disease Surveillance System and selected a cohort of 1,105,295 older adults of whom 533,719 were potentially inappropriate medication users. Potentially inappropriate medications were defined according to the Beers list. We compared performances between five popular machine learning models (gradient boosting machines, logistic regression, naïve Bayes, neural networks, and random forests) based on ROC curves and other performance criteria, using a set of sociodemographic and medical predictors. RESULTS: No model clearly outperformed the others. All models except neural networks were in agreement regarding the top predictors (sex and anxiety-depressive disorders and schizophrenia) and the bottom predictors (rurality and social and material deprivation indices). CONCLUSIONS: Including other types of predictors (e.g., unstructured data) may be more useful for increasing performance in prediction of potentially inappropriate medication use.

Multimorbidity prevalence and health outcome prediction: assessing the impact of lookback periods, disease count, and definition criteria in health administrative data at the population-based level.

BACKGROUND: Health administrative databases play a crucial role in population-level multimorbidity surveillance. Determining the appropriate retrospective or lookback period (LP) for observing prevalent and newly diagnosed diseases in administrative data presents challenge in estimating multimorbidity prevalence and predicting health outcome. The aim of this population-based study was to assess the impact of LP on multimorbidity prevalence and health outcomes prediction across three multimorbidity definitions, three lists of diseases used for multimorbidity assessment, and six health outcomes. METHODS: We conducted a population-based study including all individuals ages > 65 years on April 1st, 2019, in Québec, Canada. We considered three lists of diseases labeled according to the number of chronic conditions it considered: (1) L60 included 60 chronic conditions from the International Classification of Diseases (ICD); (2) L20 included a core of 20 chronic conditions; and (3) L31 included 31 chronic conditions from the Charlson and Elixhauser indices. For each list, we: (1) measured multimorbidity prevalence for three multimorbidity definitions (at least two [MM2+], three [MM3+] or four (MM4+) chronic conditions); and (2) evaluated capacity (c-statistic) to predict 1-year outcomes (mortality, hospitalisation, polypharmacy, and general practitioner, specialist, or emergency department visits) using LPs ranging from 1 to 20 years. RESULTS: Increase in multimorbidity prevalence decelerated after 5-10 years (e.g., MM2+, L31: LP = 1y: 14%, LP = 10y: 58%, LP = 20y: 69%). Within the 5-10 years LP range, predictive performance was better for L20 than L60 (e.g., LP = 7y, mortality, MM3+: L20 [0.798;95%CI:0.797-0.800] vs. L60 [0.779; 95%CI:0.777-0.781]) and typically better for MM3 + and MM4 + definitions (e.g., LP = 7y, mortality, L60: MM4+ [0.788;95%CI:0.786-0.790] vs. MM2+ [0.768;95%CI:0.766-0.770]). CONCLUSIONS: In our databases, ten years of data was required for stable estimation of multimorbidity prevalence. Within that range, the L20 and multimorbidity definitions MM3 + or MM4 + reached maximal predictive performance.

Strategies aiming to improve statin therapy adherence in older adults: a systematic review.

BACKGROUND: Randomized clinical trials have shown that, under optimal conditions, statins reduce the risk of cardiovascular events in older adults. Given the prevalence and consequences of suboptimal adherence to statin among older adults, it is essential to document strategies designed to increase statin adherence in this population. The objective of this systematic review is to describe and summarize the effectiveness of interventions to improve statin adherence in older adults (≥ 65 years old). METHODS: This review followed PRISMA guidelines. Studies were identified from PubMed, PsycINFO, Embase, CINAHL and Web of Science. Study selection was conducted independently by four reviewers working in pairs. Included studies reported data on interventions designed to increase adherence to statin therapy in older adults and were original trials or observational studies. Interventions were pragmatically regrouped into 8 different categories going from patient to administrative level. Two reviewers extracted study data and assessed study quality independently. Given the heterogeneity between the included studies, a narrative critique and summary was conducted. RESULTS: Twelve out of the 2889 identified articles were included in the review. Our review showed that simplifying patients' drug regimen, administrative improvements and large-scale pharmacy-led automated telephone interventions show positive effects on patient adherence to statin therapy, with odds ratios between > 1.0 and 3.0, while education-based strategies and intensified patient care showed mixed results. CONCLUSIONS: Current evidence suggests that some interventions can increase statin adherence in older adults, which could help in the reduction of the risk of a cardiovascular event in this population.

Assessing the performance of group-based trajectory modeling method to discover different patterns of medication adherence.

It is well known that medication adherence is critical to patient outcomes and can decrease patient mortality. The Pharmacy Quality Alliance (PQA) has recognized and identified medication adherence as an important indicator of medication-use quality. Hence, there is a need to use the right methods to assess medication adherence. The PQA has endorsed the proportion of days covered (PDC) as the primary method of measuring adherence. Although easy to calculate, the PDC has however several drawbacks as a method of measuring adherence. PDC is a deterministic approach that cannot capture the complexity of a dynamic phenomenon. Group-based trajectory modeling (GBTM) is increasingly proposed as an alternative to capture heterogeneity in medication adherence. The main goal of this paper is to demonstrate, through a simulation study, the ability of GBTM to capture treatment adherence when compared to its deterministic PDC analogue and to the nonparametric longitudinal K-means. A time-varying treatment was generated as a quadratic function of time, baseline, and time-varying covariates. Three trajectory models are considered combining a cat's cradle effect, and a rainbow effect. The performance of GBTM was compared to the PDC and longitudinal K-means using the absolute bias, the variance, the c-statistics, the relative bias, and the relative variance. For all explored scenarios, we find that GBTM performed better in capturing different patterns of medication adherence with lower relative bias and variance even under model misspecification than PDC and longitudinal K-means.

A Bootstrap Approach for Evaluating Uncertainty in the Number of Groups Identified by Latent Class Growth Models.

The use of longitudinal finite mixture models such as group-based trajectory modeling has seen a sharp increase during the last few decades in the medical literature. However, these methods have been criticized, especially because of the data-driven modeling process, which involves statistical decision-making. In this paper, we propose an approach that uses the bootstrap to sample observations with replacement from the original data to validate the number of groups identified and to quantify the uncertainty in the number of groups. The method allows investigation of the statistical validity and uncertainty of the groups identified in the original data by checking to see whether the same solution is also found across the bootstrap samples. In a simulation study, we examined whether the bootstrap-estimated variability in the number of groups reflected the replicationwise variability. We evaluated the ability of 3 commonly used adequacy criteria (average posterior probability, odds of correct classification, and relative entropy) to identify uncertainty in the number of groups. Finally, we illustrate the proposed approach using data from the Quebec Integrated Chronic Disease Surveillance System to identify longitudinal medication patterns between 2015 and 2018 in older adults with diabetes.

An Alternative Perspective on the Robust Poisson Method for Estimating Risk or Prevalence Ratios.

The robust Poisson method is becoming increasingly popular when estimating the association of exposures with a binary outcome. Unlike the logistic regression model, the robust Poisson method yields results that can be interpreted as risk or prevalence ratios. In addition, it does not suffer from frequent nonconvergence problems such as the most common implementations of maximum likelihood estimators of the log-binomial model. However, using a Poisson distribution to model a binary outcome may seem counterintuitive. Methodologic papers have often presented this as a good approximation to the more natural binomial distribution. In this article, we provide an alternative perspective to the robust Poisson method based on the semiparametric theory. This perspective highlights that the robust Poisson method does not require assuming a Poisson distribution for the outcome. In fact, the method only assumes a log-linear relation between the risk or prevalence of the outcome and the explanatory variables. This assumption and the consequences of its violation are discussed. We also provide suggestions to reduce the risk of violating the modeling assumption. Additionally, we discuss and contrast the robust Poisson method with other approaches for estimating exposure risk or prevalence ratios. See video abstract at, http://links.lww.com/EDE/B987 .

Double robust estimation of optimal partially adaptive treatment strategies: An application to breast cancer treatment using hormonal therapy.

Precision medicine aims to tailor treatment decisions according to patients' characteristics. G-estimation and dynamic weighted ordinary least squares are double robust methods to identify optimal adaptive treatment strategies. It is underappreciated that they require modeling all existing treatment-confounder interactions to be consistent. Identifying optimal partially adaptive treatment strategies that tailor treatments according to only a few covariates, ignoring some interactions, may be preferable in practice. Building on G-estimation and dWOLS, we propose estimators of such partially adaptive strategies and demonstrate their double robustness. We investigate these estimators in a simulation study. Using data maintained by the Centre des Maladies du Sein, we estimate a partially adaptive treatment strategy for tailoring hormonal therapy use in breast cancer patients. R software implementing our estimators is provided.

Longitudinal plasmode algorithms to evaluate statistical methods in realistic scenarios: an illustration applied to occupational epidemiology.

INTRODUCTION: Plasmode simulations are a type of simulations that use real data to determine the synthetic data-generating equations. Such simulations thus allow evaluating statistical methods under realistic conditions. As far as we know, no plasmode algorithm has been proposed for simulating longitudinal data. In this paper, we propose a longitudinal plasmode framework to generate realistic data with both a time-varying exposure and time-varying covariates. This work was motivated by the objective of comparing different methods for estimating the causal effect of a cumulative exposure to psychosocial stressors at work over time. METHODS: We developed two longitudinal plasmode algorithms: a parametric and a nonparametric algorithms. Data from the PROspective Québec (PROQ) Study on Work and Health were used as an input to generate data with the proposed plasmode algorithms. We evaluated the performance of multiple estimators of the parameters of marginal structural models (MSMs): inverse probability of treatment weighting, g-computation and targeted maximum likelihood estimation. These estimators were also compared to standard regression approaches with either adjustment for baseline covariates only or with adjustment for both baseline and time-varying covariates. RESULTS: Standard regression methods were susceptible to yield biased estimates with confidence intervals having coverage probability lower than their nominal level. The bias was much lower and coverage of confidence intervals was much closer to the nominal level when considering MSMs. Among MSM estimators, g-computation overall produced the best results relative to bias, root mean squared error and coverage of confidence intervals. No method produced unbiased estimates with adequate coverage for all parameters in the more realistic nonparametric plasmode simulation. CONCLUSION: The proposed longitudinal plasmode algorithms can be important methodological tools for evaluating and comparing analytical methods in realistic simulation scenarios. To facilitate the use of these algorithms, we provide R functions on GitHub. We also recommend using MSMs when estimating the effect of cumulative exposure to psychosocial stressors at work.

On "Reflections on the concept of optimality of single decision point treatment regimes".

This is a discussion of "Reflections on the concept of optimality of single decision point treatment regimes" by Trung Dung Tran, Ariel Alonso Abad, Geert Verbeke, Geert Molenberghs, and Iven Van Mechelen. The authors propose a thoughtful consideration of optimization targets and the implications of such targets for the resulting optimal treatment rule. However, we contest the assertation that targets of optimization have been overlooked and suggest additional considerations that researchers must contemplate as part of a complete framework for learning about optimal treatment regimes.

Single-Dose Messenger RNA Vaccine Effectiveness Against Severe Acute Respiratory Syndrome Coronavirus 2 in Healthcare Workers Extending 16 Weeks Postvaccination: A Test-Negative Design From Québec, Canada.

BACKGROUND: In Canada, first and second doses of messenger RNA (mRNA) vaccines against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) were uniquely spaced 16 weeks apart. We estimated 1- and 2-dose mRNA vaccine effectiveness (VE) among healthcare workers (HCWs) in Québec, Canada, including protection against varying outcome severity, variants of concern (VOCs), and the stability of single-dose protection up to 16 weeks postvaccination. METHODS: A test-negative design compared vaccination among SARS-CoV-2 test-positive and weekly matched (10:1), randomly sampled, test-negative HCWs using linked surveillance and immunization databases. Vaccine status was defined by 1 dose ≥14 days or 2 doses ≥7 days before illness onset or specimen collection. Adjusted VE was estimated by conditional logistic regression. RESULTS: Primary analysis included 5316 cases and 53 160 controls. Single-dose VE was 70% (95% confidence interval [CI], 68%-73%) against SARS-CoV-2 infection; 73% (95% CI, 71%-75%) against illness; and 97% (95% CI, 92%-99%) against hospitalization. Two-dose VE was 86% (95% CI, 81%-90%) and 93% (95% CI, 89%-95%), respectively, with no hospitalizations. VE was higher for non-VOCs than VOCs (73% Alpha) among single-dose recipients but not 2-dose recipients. Across 16 weeks, no decline in single-dose VE was observed, with appropriate stratification based upon prioritized vaccination determined by higher vs lower likelihood of direct patient contact. CONCLUSIONS: One mRNA vaccine dose provided substantial and sustained protection to HCWs extending at least 4 months postvaccination. In circumstances of vaccine shortage, delaying the second dose may be a pertinent public health strategy.

Two-Dose Severe Acute Respiratory Syndrome Coronavirus 2 Vaccine Effectiveness With Mixed Schedules and Extended Dosing Intervals: Test-Negative Design Studies From British Columbia and Quebec, Canada.

BACKGROUND: The Canadian coronavirus disease 2019 (COVID-19) immunization strategy deferred second doses and allowed mixed schedules. We compared 2-dose vaccine effectiveness (VE) by vaccine type (mRNA and/or ChAdOx1), interval between doses, and time since second dose in 2 of Canada's larger provinces. METHODS: Two-dose VE against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection or hospitalization among adults ≥18 years, including due to Alpha, Gamma, and Delta variants of concern (VOCs), was assessed ≥14 days postvaccination by test-negative design studies separately conducted in British Columbia and Quebec, Canada, between 30 May and 27 November (epi-weeks 22-47) 2021. RESULTS: In both provinces, all homologous or heterologous mRNA and/or ChAdOx1 2-dose schedules were associated with ≥90% reduction in SARS-CoV-2 hospitalization risk for ≥7 months. With slight decline from a peak of >90%, VE against infection was ≥80% for ≥6 months following homologous mRNA vaccination, lower by ∼10% when both doses were ChAdOx1 but comparably high following heterologous ChAdOx1 + mRNA receipt. Findings were similar by age group, sex, and VOC. VE was significantly higher with longer 7-8-week versus manufacturer-specified 3-4-week intervals between mRNA doses. CONCLUSIONS: Two doses of any mRNA and/or ChAdOx1 combination gave substantial and sustained protection against SARS-CoV-2 hospitalization, spanning Delta-dominant circulation. ChAdOx1 VE against infection was improved by heterologous mRNA series completion. A 7-8-week interval between first and second doses improved mRNA VE and may be the optimal schedule outside periods of intense epidemic surge. Findings support interchangeability and extended intervals between SARS-CoV-2 vaccine doses, with potential global implications for low-coverage areas and, going forward, for children.

A comparison of confounder selection and adjustment methods for estimating causal effects using large healthcare databases.

PURPOSE: Confounding adjustment is required to estimate the effect of an exposure on an outcome in observational studies. However, variable selection and unmeasured confounding are particularly challenging when analyzing large healthcare data. Machine learning methods may help address these challenges. The objective was to evaluate the capacity of such methods to select confounders and reduce unmeasured confounding bias. METHODS: A simulation study with known true effects was conducted. Completely synthetic and partially synthetic data incorporating real large healthcare data were generated. We compared Bayesian adjustment for confounding (BAC), generalized Bayesian causal effect estimation (GBCEE), Group Lasso and Doubly robust estimation, high-dimensional propensity score (hdPS), and scalable collaborative targeted maximum likelihood algorithms. For the hdPS, two adjustment approaches targeting the effect in the whole population were considered: Full matching and inverse probability weighting. RESULTS: In scenarios without hidden confounders, most methods were essentially unbiased. The bias and variance of the hdPS varied considerably according to the number of variables selected by the algorithm. In scenarios with hidden confounders, substantial bias reduction was achieved by using machine-learning methods to identify proxies as compared to adjusting only by observed confounders. hdPS and Group Lasso performed poorly in the partially synthetic simulation. BAC, GBCEE, and scalable collaborative-targeted maximum likelihood algorithms performed particularly well. CONCLUSIONS: Machine learning can help to identify measured confounders in large healthcare databases. They can also capitalize on proxies of unmeasured confounders to substantially reduce residual confounding bias.

Evidence of the Different Associations of Prognostic Factors With Censoring Across Treatment Groups and Impact on Censoring Weight Model Specification: The Example of Anticoagulation in Atrial Fibrillation.

Inverse probability of censoring weights (IPCWs) may reduce selection bias due to informative censoring in longitudinal studies. However, in studies with an active comparator, the associations between predictors and censoring may differ across treatment groups. We used the clinical example of anticoagulation treatment with warfarin or a direct oral anticoagulant (DOAC) in atrial fibrillation to illustrate this. The cohort of individuals initiating an oral anticoagulant during 2010-2016 was identified from the Régie de l'assurance maladie du Québec (RAMQ) databases. The parameter of interest was the hazard ratio (HR) of the composite of stroke, major bleeding, myocardial infarction, or death associated with continuous use of warfarin versus DOACs. Two strategies for the specification of the model for estimation of censoring weights were explored: exposure-unstratified and exposure-stratified. The HR associated with continuous treatment with warfarin versus DOACs adjusted with exposure-stratified IPCWs was 1.26 (95% confidence interval: 1.20, 1.33). Using exposure-unstratified IPCWs, the HR differed by 15% in favor of DOACs (1.41, 95% confidence interval: 1.34, 1.48). Not accounting for the different associations between the predictors and informative censoring across exposure groups may lead to misspecification of censoring weights and biased estimate on comparative effectiveness and safety.

The impact of adjusting for pure predictors of exposure, mediator, and outcome on the variance of natural direct and indirect effect estimators.

It is now well established that adjusting for pure predictors of the outcome, in addition to confounders, allows unbiased estimation of the total exposure effect on an outcome with generally reduced standard errors (SEs). However, no analogous results have been derived for mediation analysis. Considering the simplest linear regression setting and the ordinary least square estimator, we obtained theoretical results showing that adjusting for pure predictors of the outcome, in addition to confounders, allows unbiased estimation of the natural indirect effect (NIE) and the natural direct effect (NDE) on the difference scale with reduced SEs. Adjusting for pure predictors of the mediator increases the SE of the NDE's estimator, but may increase or decrease the variance of the NIE's estimator. Adjusting for pure predictors of the exposure increases the variance of estimators of the NIE and NDE. Simulation studies were used to confirm and extend these results to the case where the mediator or the outcome is binary. Additional simulations were conducted to explore scenarios featuring an exposure-mediator interaction as well as the relative risk and odds ratio scales for the case of binary mediator and outcome. Both a regression approach and an inverse probability weighting approach were considered in the simulation study. A real-data illustration employing data from the Canadian Study of Health and Aging is provided. This analysis is concerned with the mediating effect of vitamin D in the effect of physical activity on dementia and its results are overall consistent with the theoretical and empirical findings.

Effectiveness of a workplace intervention reducing psychosocial stressors at work on blood pressure and hypertension.

OBJECTIVES: To assess the effectiveness of a workplace intervention reducing psychosocial stressors at work in lowering blood pressure and hypertension prevalence. METHODS: The study design was a quasi-experimental pre-post study with an intervention group and a control group. Post-intervention measurements were collected 6 and 36 months after the midpoint of the intervention. Participants were all white-collar workers employed in three public organisations. At baseline, the intervention and the control groups were composed of 1088 and 1068 workers, respectively. The intervention was designed to reduce psychosocial stressors at work by implementing organisational changes. Adjusted changes in ambulatory blood pressure and hypertension prevalence were examined. RESULTS: Blood pressure and hypertension significantly decreased in the intervention group while no change was observed in the control group. The differential decrease in systolic blood pressure between the intervention and the control group was 2.0 mm Hg (95% CI: -3.0 to -1.0). The prevalence of hypertension decreased in the intervention group, when compared with the control group (prevalence ratio: 0.85 (95% CI: 0.74 to 0.98)). CONCLUSIONS: Findings suggest that psychosocial stressors at work are relevant targets for the primary prevention of hypertension. At the population level, systolic blood pressure reductions such as those observed in the present study could prevent a significant number of premature deaths and disabling strokes.

Cumulative exposure to psychosocial stressors at work and global cognitive function: the PROspective Quebec Study on Work and Health.

OBJECTIVES: Psychosocial stressors at work have been proposed as modifiable risk factors for mild cognitive impairment (MCI). This study aimed to evaluate the effect of cumulative exposure to psychosocial stressors at work on cognitive function. METHODS: This study was conducted among 9188 white-collar workers recruited in 1991-1993 (T1), with follow-ups 8 (T2) and 24 years later (T3). After excluding death, losses to follow-up and retirees at T2, 5728 participants were included. Psychosocial stressors at work were measured according to the Karasek's questionnaire. Global cognitive function was measured with the Montreal Cognitive Assessment. Cumulative exposures to low psychological demand, low job control, passive job and high strain job were evaluated using marginal structural models including multiple imputation and inverse probability of censoring weighting. RESULTS: In men, cumulative exposures (T1 and T2) to low psychological demand, low job control or passive job were associated with higher prevalences of more severe presentation of MCI (MSMCI) at T3 (Prevalence ratios (PRs) and 95% CIs of 1.50 (1.16 to 1.94); 1.38 (1.07 to 1.79) and 1.55 (1.20 to 2.00), respectively), but not with milder presentation of MCI. In women, only exposure to low psychological demand or passive job at T2 was associated with higher prevalences of MSMCI at T3 (PRs and 95% CI of 1.39 (0.97 to 1.99) and 1.29 (0.94 to 1.76), respectively). CONCLUSIONS: These results support the deleterious effect of a low stimulating job on cognitive function and the cognitive reserve theory. Psychosocial stressors at work could be part of the effort for the primary prevention of cognitive decline.

Validation of case definitions of depression derived from administrative data against the CIDI-SF as reference standard: results from the PROspective Québec (PROQ) study.

BACKGROUND: Administrative data have several advantages over questionnaire and interview data to identify cases of depression: they are usually inexpensive, available for a long period of time and are less subject to recall bias and differential classification errors. However, the validity of administrative data in the correct identification of depression has not yet been studied in general populations. The present study aimed to 1) evaluate the sensitivity and specificity of administrative cases of depression using the validated Composite International Diagnostic Interview - Short Form (CIDI-SF) as reference standard and 2) compare the known-groups validity between administrative and CIDI-SF cases of depression. METHODS: The 5487 participants seen at the last wave (2015-2018) of the PROQ cohort had CIDI-SF questionnaire data linked to hospitalization and medical reimbursement data provided by the provincial universal healthcare provider and coded using the International Classification of Disease. We analyzed the sensitivity and specificity of several case definitions of depression from this administrative data. Their association with known predictors of depression was estimated using robust Poisson regression models. RESULTS: Administrative cases of depression showed high specificity (≥ 96%), low sensitivity (19-32%), and rather low agreement (Cohen's kappa of 0.21-0.25) compared with the CIDI-SF. These results were consistent over strata of sex, age and education level and with varying case definitions. In known-groups analysis, the administrative cases of depression were comparable to that of CIDI-SF cases (RR for sex: 1.80 vs 2.03 respectively, age: 1.53 vs 1.40, education: 1.52 vs 1.28, psychological distress: 2.21 vs 2.65). CONCLUSION: The results obtained in this large sample of a general population suggest that the dimensions of depression captured by administrative data and by the CIDI-SF are partially distinct. However, their known-groups validity in relation to risk factors for depression was similar to that of CIDI-SF cases. We suggest that neither of these data sources is superior to the other in the context of large epidemiological studies aiming to identify and quantify risk factors for depression.

Exploring polypharmacy with artificial intelligence: data analysis protocol.

BACKGROUND: Polypharmacy is common among older adults and it represents a public health concern, due to the negative health impacts potentially associated with the use of several medications. However, the large number of medication combinations and sequences of use makes it complicated for traditional statistical methods to predict which therapy is genuinely associated with health outcomes. The project aims to use artificial intelligence (AI) to determine the quality of polypharmacy among older adults with chronic diseases in the province of Québec, Canada. METHODS: We will use data from the Quebec Integrated Chronic Disease Surveillance System (QICDSS). QICDSS contains information about prescribed medications in older adults in Quebec collected over 20 years. It also includes diagnostic codes and procedures, and sociodemographic data linked through a unique identification number for each individual. Our research will be structured around three interconnected research axes: AI, Health, and Law&Ethics. The AI research axis will develop algorithms for finding frequent patterns of medication use that correlate with health events, considering data locality and temporality (explainable AI or XAI). The Health research axis will translate these patterns into polypharmacy indicators relevant to public health surveillance and clinicians. The Law&Ethics axis will assess the social acceptability of the algorithms developed using AI tools and the indicators developed by the Heath axis and will ensure that the developed indicators neither discriminate against any population group nor increase the disparities already present in the use of medications. DISCUSSION: The multi-disciplinary research team consists of specialists in AI, health data, statistics, pharmacy, public health, law, and ethics, which will allow investigation of polypharmacy from different points of view and will contribute to a deeper understanding of the clinical, social, and ethical issues surrounding polypharmacy and its surveillance, as well as the use of AI for health record data. The project results will be disseminated to the scientific community, healthcare professionals, and public health decision-makers in peer-reviewed publications, scientific meetings, and reports. The diffusion of the results will ensure the confidentiality of individual data.

Telephone versus web panel National Survey for monitoring adoption of preventive behaviors to climate change in populations: a case study of Lyme disease in Québec, Canada.

BACKGROUND: To monitor the adoption of climate change adaptive behaviors in the population, public health authorities have to conduct national surveys, which can help them target vulnerable subpopulations. To ensure reliable estimates of the adoption of these preventive behaviors, many data collection methods are offered by polling firms. The aim of this study was to compare a telephone survey with a web survey on Lyme disease with regard to their representativeness. METHODS: The data comes from a cross-sectional study conducted in the Province of Québec (Canada). In total, 1003 people completed the questionnaire by telephone and 956 filled in a web questionnaire. We compared the data obtained from both survey modes with the census data in regard to various demographic characteristics. We then compared the data from both samples in terms of self-reported Lyme disease preventive behaviors and other theoretically associated constructs. We also assessed the measurement invariance (equivalence) of the index of Lyme disease preventive behaviors across the telephone and web samples. RESULTS: Findings showed that neither the telephone nor the web panel modes of data collection can be considered more representative of the target population. The results showed that the proportion of item non-responses was significantly higher with the web questionnaire (5.6%) than with the telephone survey (1.3%), and that the magnitude of the differences between the two survey modes was nil for 19 out of the 30 items related to Lyme disease, and small for 11 of them. Results from invariance analyses confirmed the measurement invariance of an index of adaptation to Lyme disease, as well as the mean invariance across both samples. CONCLUSIONS: Our results suggested that both samples provided similar estimates of the level of adaptation to Lyme disease preventive behaviors. In sum, the results of our study showed that neither survey mode was superior to the other. Thus, in studies where adaptation to climate change is monitored over time, using a web survey instead of a telephone survey could be more cost-effective, and researchers should consider doing so in future surveys on adaptation to climate. However, we recommend conducting a pretest study before deciding whether to use both survey modes or only one of them.