Seasonality in six enterically transmitted diseases and ambient temperature.

We propose an analytical and conceptual framework for a systematic and comprehensive assessment of disease seasonality to detect changes and to quantify and compare temporal patterns. To demonstrate the proposed technique, we examined seasonal patterns of six enterically transmitted reportable diseases (EDs) in Massachusetts collected over a 10-year period (1992-2001). We quantified the timing and intensity of seasonal peaks of ED incidence and examined the synchronization in timing of these peaks with respect to ambient temperature. All EDs, except hepatitis A, exhibited well-defined seasonal patterns which clustered into two groups. The peak in daily incidence of Campylobacter and Salmonella closely followed the peak in ambient temperature with the lag of 2-14 days. Cryptosporidium, Shigella, and Giardia exhibited significant delays relative to the peak in temperature (approximately 40 days, P<0.02). The proposed approach provides a detailed quantification of seasonality that enabled us to detect significant differences in the seasonal peaks of enteric infections which would have been lost in an analysis using monthly or weekly cumulative information. This highly relevant to disease surveillance approach can be used to generate and test hypotheses related to disease seasonality and potential routes of transmission with respect to environmental factors.

INFERNO: a system for early outbreak detection and signature forecasting.

OBJECTIVE: Public health surveillance systems that monitor daily disease incidence provide valuable information about threats to public health and enable public health authorities to detect enteric outbreaks rapidly. This report describes the INtegrated Forecasts and EaRly eNteric Outbreak (INFERNO) detection system of algorithms for outbreak detection and forecasting. METHODS: INFERNO incorporates existing knowledge of infectious disease epidemiology into adaptive forecasts and uses the concept of an outbreak signature as a composite of disease epidemic curves. RESULTS: Four main components comprise the system: 1) training, 2) warning and flagging, 3) signature forecasting, and 4) evaluation. The unifying goal of the system is to gain insight into the nature of temporal variations in the incidence of infection. Daily collected records are smoothed initially by using a loess-type smoother. Upon receipt of new data, the smoothing is updated; estimates are made of the first two derivatives of the smoothed curve, which are used for near-term forecasting. Recent data and near-term forecasts are used to compute a five level, color-coded warning index to quantify the level of concern. Warning algorithms are designed to balance false detection of an epidemic (Type I errors) with failure to correctly detect an epidemic (Type II errors). If the warning index signals a sufficiently high probability of an epidemic, the fitting of a gamma-based signature curve to the actual data produces a forecast of the possible size of the outbreak. CONCLUSION: Although the system is under development, its potential has been demonstrated through successful use of emergency department records associated with a substantial waterborne outbreak of cryptosporidiosis that occurred in Milwaukee, Wisconsin, in 1993. Prospects for further development, including adjustment for seasonality and reporting delays, are also outlined.

Assessing the surveillance capability of Canada's national health surveys.

We assessed Canada's national health surveys as surveillance instruments, with emphasis on comparing the temporal structure of data sets with those generated by the US Behavioral Risk Factor Surveillance System (BRFSS). Only the Canadian Tobacco Use Monitoring Survey (CTUMS) has the BRFSS capability to generate continuous, uniform time series with monthly intervals. These time series can offer substantial extra value for retrospective analysis such as program evaluation in addition to surveillance. Expanding CTUMS is a simple option for providing an ongoing, uniform monthly survey instrument for non-tobacco variables. The Canadian Community Health Survey (CCHS) will generate monthly data, and could potentially generate useful continuous time series even though surveys at the health region and provincial levels will alternate annually. Reconfiguring the CCHS, or even implementing a provincial surveillance survey based on the BRFSS model are other viable options, but each option has associated tradeoffs or obstacles.

Effects of screening errors and differential mortality on the estimation of the incidence of dementia in the Canadian Study of Health and Aging.

The Canadian Study of Health and Aging produced an estimate of the incidence of dementia among elderly Canadians by following up, after 5 years, the undemented found in an initial prevalence survey. Initial and follow-up estimates could be biased by false-negative error in the screening tool used for subjects living in the community, and by erroneous classification of subjects who died in the interim. Here, we use a deterministic model to quantify those possible biases. We conclude that, using the estimates of the errors from control samples, the incidence among community subjects would be overestimated by 15%, and the incidence among the institutional subjects would be underestimated by 37%. The overall incidence would be underestimated by 14%. Most of the bias can be attributed to inaccuracies in the classification of deaths.

Likelihood and clinical trials.

The history of the application of statistical theory to the analysis of clinical trials is reviewed. The current orthodoxy is a somewhat illogical hybrid of the original theory of significance tests of Edgeworth, Karl Pearson, and Fisher, and the subsequent decision theory approach of Neyman, Egon Pearson, and Wald. This hegemony is under threat from Bayesian statisticians. A third approach is that of likelihood, stemming from the work of Fisher and Barnard. This approach is illustrated using hypothetical data from the Lancet articles by Bradford Hill, which introduced clinicians to statistical theory.

The impending Canadian prostate cancer epidemic.

PURPOSE: To model and forecast prostate cancer incidence and mortality in Canada to the year 2016. METHODS: Bivariate multiplicative models of prostate cancer incidence and mortality for Canadian men aged 45 years or older, linear in time and Weibull in age, were fitted using weighted non-linear regression. RESULTS: The number of incident cases of prostate cancer is forecast to increase from 11,355, observed in 1990, to 26,900 by the year 2010 and to 35,200 by the year 2016. The number of deaths are estimated to climb from 3,424, observed in 1991, to an estimated 6,300 by the year 2010, and to 7,800 by the year 2016. CONCLUSIONS: The dramatic increase in prostate cancer rates with age, coupled with the expected large increase in the elderly Canadian male population and steadily increasing prostate cancer incidence rates will produce very large increases in the number of men who will have prostate cancer over the next 20 years. This has important implications for health care delivery in the future.

Trends in mortality from melanoma in Canada and prediction of future rates.

A long term increase in incidence of and mortality due to malignant melanoma has been observed in all well documented white populations. The major identified cause of melanoma is sun exposure. One would expect predictions of future atmospheric ozone depletion to lead to an increase in ultraviolet radiation (UVR) and in the effects of sun exposure. We consider age-period data for Canadian malignant malanoma mortality. We fit a multiplicative exponential/logistic (MEL) model to the data and extrapolate to AD 2010 hence yielding point estimates of future rates. We obtain total mortality forecasts by multiplying rates by population estimates. We present standard errors for forecasts. We forecast that melanoma will be a much larger burden on the health care system in the early years of the next century than it is at present. We obtain an age-cohort model by a simple transformation of the age-period model. Also, we obtain unconditional probabilities of death due to melanoma both for age-period and age-cohort models. We discuss the assumptions underlying the MEL model that suggest possible relationships between UVR and melanoma.

PIP: The authors use a multiplicative exponential/logistic model and official data for 1951-1989 to extrapolate future trends in mortality from melanoma in Canada.

Monitoring statistics which have increased power over a reduced time range.

The problem of monitoring trends for changes at unknown times is considered. Statistics which permit one to focus high power on a segment of the monitored period are studied. Numerical procedures are developed to compute the null distribution of these statistics.

Time series valued experimental designs: A review.

A review is given of the literature on time-series valued experimental designs. Most of this literature is divided into two categories depending upon the factor status of the time variable. In one category, time is an experimental factor, and in the other it is a non-specific factor and enters the design in the context of replications. Analyses in both the time and frequency domain are reviewed. Signal detection models, Bayesian methods and optimal designs are surveyed. A discussion is also presented of application areas which include field trials and medical experiments. A main theme of the literature is that application of standard F-tests to highly correlated data can be misleading. A bibliography of relevant publications from 1949 onward is presented.

The effect of autocorrelated errors on change-detection statistics.

In this paper, regression models with error terms generated by lower order ARMA schemes are analyzed. Methods are discussed for estimating the parameters of the regression coefficients and the ARMA processes. The problem of detecting changes in the regression parameters is considered. A change-detection statistic proposed by MacNeill (1978) for regression problems is modified for application to ARMA processes. The effect of autocorrelated errors on this statistic is briefly discussed.