Patrick Brown

University of Toronto, Toronto, Ontario, Canada

Are you Patrick Brown?

Claim your profile

Publications (16)56.91 Total impact

  • [Show abstract] [Hide abstract]
    ABSTRACT: Our previous study found cancer detection rates were equivalent for direct radiography compared to screen-film mammography, while rates for computed radiography were significantly lower. This study compares prognostic features of invasive breast cancers by type of mammography. Approved by the University of Toronto Research Ethics Board, this study identified invasive breast cancers diagnosed among concurrent cohorts of women aged 50-74 screened by direct radiography, computed radiography, or screen-film mammography from January 1, 2008 to December 31, 2009. During the study period, 816,232 mammograms were performed on 668,418 women, and 3,323 invasive breast cancers were diagnosed. Of 2,642 eligible women contacted, 2,041 participated (77.3 %). The final sample size for analysis included 1,405 screen-detected and 418 interval cancers (diagnosed within 24 months of a negative screening mammogram). Polytomous logistic regression was performed to evaluate the association between tumour characteristics and type of mammography, and between tumour characteristics and detection method. Odds ratios (OR) and 95 % confidence intervals (CI) were recorded. Cancers detected by computed radiography compared to screen-film mammography were significantly more likely to be lymph node positive (OR 1.94, 95 %CI 1.01-3.73) and have higher stage (II:I, OR 2.14, 95 %CI 1.11-4.13 and III/IV:I, OR 2.97, 95 %CI 1.02-8.59). Compared to screen-film mammography, significantly more cancers detected by direct radiography (OR 1.64, 95 %CI 1.12-2.38) were lymph node positive. Interval cancers had worse prognostic features compared to screen-detected cancers, irrespective of mammography type. Screening with computed radiography may lead to the detection of cancers with a less favourable stage distribution compared to screen-film mammography that may reflect a delayed diagnosis. Screening programs should re-evaluate their use of computed radiography for breast screening.
    Breast Cancer Research and Treatment 08/2014; 147(2). DOI:10.1007/s10549-014-3088-2 · 4.20 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Purpose: To assess the associations of occupational heat and cold-related illnesses presenting in emergency departments in south western Ontario, Canada, with daily meteorological conditions using Bayesian inference. Methodology: Meteorological and air pollution data for the south western economic region of Ontario were gathered from Environment Canada and the Ministry of Environment. Daily heat and cold-related emergency department visits clinically attributed to work from 2004 to 2010 were tabulated. A novel application of Bayesian inference on a flexible Poisson time series model was undertaken to examine linear and non-linear associations between average, regional meteorological conditions and daily morbidity rates, to adjust for relevant confounders and temporal trends, and to consider potential interactions. Results: Bilinear associations were observed between regional temperatures and morbidities resulting from extreme temperature exposures. The median increase in the daily rate of emergency department visits for heat illness was 75% for each degree above 22 degrees C (posterior 95% credible interval (Cl) relative rate=1.56-1.99) in the daily maximum temperature. Below 0 degrees C, rates of occupational cold illness increased by a median of 15% for each degree decrease in the minimum temperature (posterior 95% CI 0.80-0.91); wind speed also had a significant effect. Conclusions: The observed associations can inform occupational surveillance and injury prevention programming, as well as public health efforts targeting vulnerable populations. Methodologically, the use of Bayesian inference in time series analyses of meteorological exposures is feasible and conducive to providing accurate advice for policy and practice. (C) 2014 The Authors. Published by Elsevier Inc.
    Environmental Research 05/2014; 132C:449-456. DOI:10.1016/j.envres.2014.04.022 · 3.95 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Purpose To assess the associations of occupational heat and cold-related illnesses presenting in emergency departments in south western Ontario, Canada, with daily meteorological conditions using Bayesian inference. Methodology Meteorological and air pollution data for the south western economic region of Ontario were gathered from Environment Canada and the Ministry of Environment. Daily heat and cold-related emergency department visits clinically attributed to work from 2004 to 2010 were tabulated. A novel application of Bayesian inference on a flexible Poisson time series model was undertaken to examine linear and non-linear associations between average, regional meteorological conditions and daily morbidity rates, to adjust for relevant confounders and temporal trends, and to consider potential interactions. Results Bilinear associations were observed between regional temperatures and morbidities resulting from extreme temperature exposures. The median increase in the daily rate of emergency department visits for heat illness was 75% for each degree above 22 °C (posterior 95% credible interval (CI) relative rate=1.56–1.99) in the daily maximum temperature. Below 0 °C, rates of occupational cold illness increased by a median of 15% for each degree decrease in the minimum temperature (posterior 95% CI 0.80–0.91); wind speed also had a significant effect. Conclusions The observed associations can inform occupational surveillance and injury prevention programming, as well as public health efforts targeting vulnerable populations. Methodologically, the use of Bayesian inference in time series analyses of meteorological exposures is feasible and conducive to providing accurate advice for policy and practice.
  • [Show abstract] [Hide abstract]
    ABSTRACT: Objective To describe the spatial distribution of incident cases of systemic lupus erythematosus (SLE) using geographic information systems (GIS). Methods Spatial analyses were carried out on 890 SLE patients and 541 psoriatic arthritis (PsA) patients (controls). Age- and sex-adjusted rates for SLE/PsA for each census tract were calculated using denominator population values from the Canadian census. Spatial variations in relative risk were estimated by modeling risk as the product of a time effect, an age effect, and a spatially autocorrelated risk surface to identify hot spots. Patients within the detected hot spot were compared to those outside the hot spot to identify explanatory factors. ResultsSLE patients were predominantly female (87.75%) and the incidence rate was highest among those 15–19 years of age (2.4 cases/100,000 person-years). In an SLE hot spot containing 59 patients, 100% of the patients were female and 49.1% (n = 29) were Caucasian, while outside of the hot spot, 86.9% (n = 722) of the patients were female and 68.4% (n = 568) were Caucasian. The proportion of cases of Chinese ethnicity was significantly greater within the hot spot. An interaction was found between Chinese ethnicity and residence within the hot spot, with the risk of SLE to the Chinese population found to be twice the risk to the non-Chinese population. ConclusionGIS was used to map SLE cases and a hot spot was identified after adjustment for age and sex. Ethnicity by itself did not confer an increased risk of SLE, but the interaction of ethnicity with location of residence significantly increased the risk of SLE.
    Arthritis & Rheumatology 06/2013; 65(6). DOI:10.1002/art.37897 · 7.87 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Purpose:To evaluate the performance of digital direct radiography (DR) and computed radiography (CR) compared with that of screen-film mammography (SFM) in large concurrent cohorts.Materials and Methods:This study was approved by the University of Toronto Research Ethics Board and did not require informed consent. Concurrent cohorts of women aged 50-74 years screened with DR (n = 220 520), CR (n = 64 210), or SFM (n = 403 688) between 2008 and 2009 were identified and followed for 12 months. Performance was compared between cohorts, with SFM as the referent cohort. Associations were examined by using mixed-effect logistic regression.Results:The cancer detection rate was similar for DR (4.9 per 1000; 95% confidence interval [CI]: 4.7, 5.2) and SFM (4.8 per 1000; 95% CI: 4.7, 5.0); however, the rate was significantly lower for CR (3.4 per 1000; 95% CI: 3.0, 3.9) (odds ratio, 0.79; 95% CI: 0.68, 0.93). Recall rates were higher for DR (7.7%; 95% CI: 7.6%, 7.8%) and lower for CR (6.6%; 95% CI: 6.5%, 6.7%) than for SFM (7.4%; 95% CI: 7.3%, 7.5%). Positive predictive value was lower for CR (5.2%; 95% CI: 4.7%, 5.8%) than for SFM (6.6%; 95% CI: 6.4%, 6.8%); however, the adjusted odds were not significant.Conclusion:Although DR is equivalent to SFM for breast screening among women aged 50-74 years, the cancer detection rate was lower for CR. Screening programs should monitor the performance of CR separately and may consider informing women of the potentially lower cancer detection rates.© RSNA, 2013.
    Radiology 05/2013; 268(3). DOI:10.1148/radiol.13122567 · 6.21 Impact Factor
  • Ye Li, Patrick Brown, Dionne C Gesink, Håvard Rue
    [Show abstract] [Hide abstract]
    ABSTRACT: This article presents a methodology for modeling aggregated disease incidence data with the spatially continuous log-Gaussian Cox process. Statistical models for spatially aggregated disease incidence data usually assign the same relative risk to all individuals in the same reporting region (census areas or postal regions). A further assumption that the relative risks in two regions are independent given their neighbor's risks (the Markov assumption) makes the commonly used Besag-York-Mollié model computationally simple. The continuous model proposed here uses a data augmentation step to sample from the posterior distribution of the exact locations at each step of an Markov chain Monte Carlo algorithm, and models the exact locations with an log-Gaussian Cox process. A simulation study shows the log-Gaussian Cox process model consistently outperforming the Besag-York-Mollié model. The method is illustrated by making inference on the spatial distribution of syphilis risk in North Carolina. The effect of several known social risk factors are estimated, and areas with risk well in excess of that expected given these risk factors are identified.
    Statistical Methods in Medical Research 04/2012; 21(5):479-507. DOI:10.1177/0962280212446326 · 2.96 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: To determine if the spatial pattern of gonorrhea observed for North Carolina was influenced by neighborhood-level sociocultural determinants of health, including race/ethnicity. A generalized linear mixed model with spatially correlated random effects was fit to measure the influence of socio-cultural factors on the spatial pattern of gonorrhea reported to the North Carolina State Health Department (January 1, 2005 to March 31, 2008). Neighborhood gonorrhea rates increased as the percent single mothers increased (25th to 75th neighborhood percentile Relative Rate 1.18, 95% CI 1.12, 1.25), and decreased as socioeconomic status increased (Relative Rate 0.89, 95% CI 0.84, 0.95). Increasing numbers of men in neighborhoods with more women than men did not change the gonorrhea rate, but was associated with decreased rates in neighborhoods with more men than women. Living in the mountains was protective for all race/ethnicities. Rurality was associated with decreased rates for Blacks and increased rates for Native Americans outside the mountains. Neighborhood-level sociocultural factors, primarily those indicative of neighborhood deprivation, explained a significant proportion of the spatial pattern of gonorrhea in both urban and rural communities. Race/ethnicity was an important proxy for social and cultural factors not captured by measures of socioeconomic status.
    Annals of epidemiology 04/2011; 21(4):245-52. DOI:10.1016/j.annepidem.2010.11.015 · 2.15 Impact Factor
  • Ye Li, Patrick Brown, Håvard Rue, Paul Fortin
    [Show abstract] [Hide abstract]
    ABSTRACT: Clinical data on the location of residence at the time of diagnosis of new lupus cases in Toronto, Canada, for the 40 years to 2007 are modelled with the aim of finding areas of abnormally high risk. Inference is complicated by numerous irregular changes in the census regions on which population is reported. A model is introduced consisting of a continuous random spatial surface and fixed effects for time and ages of individuals. The process is modelled on a fine grid and Bayesian inference performed by using integrated nested Laplace approximations. Predicted risk surfaces and posterior probabilities of exceedance are produced for lupus and, for comparison, psoriatic arthritis data from the same clinic. Simulations studies are also carried out to understand better the performance of the model proposed as well as to compare with existing methods.
    Applied Statistics 01/2011; 61(1). DOI:10.2307/41430951 · 1.42 Impact Factor
  • Source
    Patrick Brown, Hedy Jiang
    [Show abstract] [Hide abstract]
    ABSTRACT: A large number of factors can affect the statistical power and bias of analyses of data from large cohort studies, including misclassification, correlated data, follow-up time, prevalence of the risk factor of interest, and prevalence of the outcome. This paper presents a method for simulating cohorts where individual's risk is correlated within communities, recruitment is staggered over time, and outcomes are observed after different follow-up periods. Covariates and outcomes are misclassified, and Cox proportional hazards models are fit with a community-level frailty term. The effect on study power of varying effect sizes, prevalences, correlation, and misclassification are explored, as well as varying the proportion of controls in nested case-control studies.
    Biometrical Journal 10/2010; 52(5):604-15. DOI:10.1002/bimj.200900277 · 1.24 Impact Factor
  • 65th Annual Meeting of the Canadian-Rheumatology-Association; 06/2010
  • [Show abstract] [Hide abstract]
    ABSTRACT: Evidence from breast screening trials has shown that a significant reduction in breast cancer mortality from screening can be achieved by regular attendance. Few studies have evaluated the influence of nurses on compliance with breast screening recommendations. The cohort included 157,788 women ages 50 to 69 years who were screened at 1 of 9 regional cancer centers or 57 affiliated centers with nurses or 26 affiliated centers without nurses between January 1, 2002, and December 31, 2002, within the Ontario Breast Screening Program. These women were followed up prospectively for at least 30 months to compare compliance for annual and biennial screening recommendations among women who attended centers with and without nurses. The associations between type of screening center and the odds of compliance were modeled using mixed-effect logistic regression models. All P values are two-sided. Women attending a regional cancer center [odds ratios (OR), 1.96; 95% confidence interval (95% CI), 1.07-3.58] or affiliated center with nurses (OR, 1.75; 95% CI, 1.38-2.22) were significantly more likely to return within 18 months of their annual screening recommendation than women attending affiliated centers without nurses. In addition, women attending regional cancer centers (OR, 2.28; 95% CI, 1.34-3.89) or affiliated centers with nurses (OR, 2.30; 95% CI, 1.86-2.83) were significantly more likely to make a timely return within the recommended biennial screening interval of between 18 and 30 months. Breast screening programs should consider methods of integrating educational activities as provided by the nurses to improve compliance with screening.
    Cancer Epidemiology Biomarkers & Prevention 02/2010; 19(3):697-706. DOI:10.1158/1055-9965.EPI-09-0884 · 4.32 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The multidisciplinary cancer conference (MCC) provides an outlet for contributors in cancer care collectively to evaluate diagnosis and treatment options and to provide optimal patient care. The prevalence and perceived benefits of MCCs in Canada have not previously been described. Between February and March 2007, the Cancer Services Integration Survey, including four key statements concerning MCCs, was administered to cancer care providers and administrators in Ontario, Canada. A total of 1,769 responses were received with a response rate of 33%. Overall, 74% of respondents were aware of MCCs within their region, but only 58% were either regular MCC participants, or acknowledged participation of cancer providers in their institutions. Using multilevel modeling, physicians (OR 2.69, p-value < 0.01, 95% CI 1.62-4.57) and surgeons (OR 3.00, p-value < 0.01, 95% CI 1.52-6.20) both perceived greater benefit of MCCs for coordinating and improving patient plans than administrators. Although MCCs appear to positively influence patient care and interprofessional interactions, variability exists among cancer providers and administrators concerning their acceptance and perceived benefits. Further research should concentrate on further probing these trends, and exploring explanations and solutions for the inconsistent acceptance of MCCs into routine cancer care.
    Journal of Interprofessional Care 11/2009; 23(6):599-610. DOI:10.3109/13561820902921829 · 1.36 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: There is controversy about whether adding clinical breast examination (CBE) to mammography improves the accuracy of breast screening. We compared the accuracy of screening among centers that offered CBE in addition to mammography with that among centers that offered only mammography. The cohort included 290 230 women aged 50-69 years who were screened at regional cancer centers or affiliated centers within the Ontario Breast Screening Program between January 1, 2002, and December 31, 2003, and were followed up for 12 months. The regional cancer centers offer screening mammography and CBE performed by a nurse. All affiliated centers provide mammography but not all provide CBE. Performance measures for 232 515 women who were screened by mammography and CBE at the nine regional cancer centers or 59 affiliated centers that provided CBE were compared with those for 57 715 women who were screened by mammography alone at 34 affiliated centers that did not provide CBE. Sensitivity of referrals was higher for women who were screened at regional cancer centers or affiliated centers that offered CBE in addition to mammography than for women screened at affiliated centers that did not offer CBE (initial screen: 94.9% and 94.6%, respectively, vs 88.6%; subsequent screen: 94.9% and 91.7%, respectively, vs 85.3%). Mammography sensitivity was similar between centers that offered CBE and those that did not. However, women without cancer who were screened at regional cancer centers or affiliated centers that offered CBE had a higher false-positive rate than women screened at affiliated centers that offered only mammography (initial screen: 12.5% and 12.4%, respectively, vs 7.4%; subsequent screen: 6.3% and 8.3%, respectively, vs 5.4%). Women should be informed of the benefits and risks of having a CBE in addition to mammography for breast screening.
    CancerSpectrum Knowledge Environment 09/2009; 101(18):1236-43. DOI:10.1093/jnci/djp241 · 15.16 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The use of electronic health records (EHRs) to support the organization and delivery of healthcare is evolving rapidly. However, little is known regarding potential variation in access to EHRs by provider type or care setting. This paper reports on observed variation in the perceptions of access to EHRs by a wide range of cancer care providers covering diverse cancer care settings in Ontario, Canada. Perspectives were sought regarding EHR access and health record completeness for cancer patients as part of an internet survey of 5663 cancer care providers and administrators in Ontario. Data were analyzed using a multilevel logistic regression model. Provider type, location of work, and access to computer or internet were included as covariates in the model. A total of 1997 of 5663 (35%) valid responses were collected. Focusing on data from cancer care providers (N = 1247), significant variation in EHR access and health record completeness was observed between provider types, location of work, and level of computer access. Providers who worked in community hospitals were half as likely as those who worked in teaching hospitals to have access to their patients' EHRs (OR 0.45 95% CI: 0.24-0.85, p < 0.05) and were six times less likely to have access to other organizations' EHRs (OR 0.15 95% CI: 0.02-1.00, p < 0.05). Compared to surgeons, nurses (OR 3.47 95% CI: 1.80-6.68, p < 0.05), radiation therapists/physicists (OR 7.86 95% CI: 2.54-25.34, p < 0.05), and other clinicians (OR 4.92 95% CI: 2.15-11.27, p < 0.05) were more likely to report good access to their organization's EHRs. Variability in access across different provider groups, organization types, and geographic locations illustrates the fragmented nature of EHR adoption in the cancer system. Along with focusing on technological aspects of EHR adoption within organizations, it is essential that there is cross-organizational and cross-provider access to EHRs to ensure patient continuity of care, system efficiency, and high quality care.
    BMC Medical Informatics and Decision Making 08/2009; 9:38. DOI:10.1186/1472-6947-9-38 · 1.50 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: To determine factors that influence awareness of, and readiness to undergo, fecal occult blood testing (FOBT) for colorectal cancer (CRC) screening. Validated survey designed to ascertain respondents' stages of decision making regarding CRC screening using FOBT. Ontario. A total of 1013 people 50 years old and older drawn from all regions of the province using a random-digit dialing telephone protocol. Awareness of FOBT and readiness to undergo it for screening for CRC. Response rate was 69%. Results indicated that 54% of women and 45% of men had "heard of" FOBT, and 26% of women and 17% of men had heard of it but were still "not considering" FOBT screening. Only 17% of all respondents had "decided to have" FOBT screening. Demographic factors associated with having heard of FOBT were female sex, completion of college or higher education, and being married or living as married. Demographic factors associated with active consideration of FOBT among those who reported awareness of it were male sex and being married or living as married. Many people seemed uninformed about FOBT and not ready to undertake this type of screening. Results of this survey could help guide strategies and develop programs to make eligible people aware of CRC screening using FOBT and to motivate them to undergo testing.
    Canadian family physician Medecin de famille canadien 03/2009; 55(2):176-177.e4. · 1.40 Impact Factor
  • Source
    Patrick Brown, Lutong Zhou
    [Show abstract] [Hide abstract]
    ABSTRACT: The glmmBUGS package is a bridging tool between Generalized Linear Mixed Mod-els (GLMMs) in R and the BUGS language. It provides a simple way of performing Bayesian inference using Markov Chain Monte Carlo (MCMC) methods, taking a model formula and data frame in R and writing a BUGS model file, data file, and initial values files. Functions are provided to reformat and summarize the BUGS results. A key aim of the package is to provide files and objects that can be modified prior to calling BUGS, giving users a platform for cus-tomizing and extending the models to accommo-date a wide variety of analyses.