Mixture randomized item-response modeling: a smoking behavior validation study.

Misleading response behavior is expected in medical settings where incriminating behavior is negatively related to the recovery from a disease. In the present study, lung patients feel social and professional pressure concerning smoking and experience questions about smoking behavior as sensitive and tend to conceal embarrassing or threatening information. The randomized item-response survey method is expected to improve the accuracy of self-reports as individual item responses are masked and only randomized item responses are observed. We explored the validation of the randomized item-response technique in a unique experimental study. Therefore, we administered a new multi-item measure assessing smoking behavior by using a treatment-control design (randomized response (RR) or direct questioning). After the questionnaire, we administered a breath test by using a carbon monoxide (CO) monitor to determine the smoking status of the patient. We used the response data to measure the individual smoking behavior by using a mixture item-response model. It is shown that the detected smokers scored significantly higher in the RR condition compared with the directly questioned condition. We proposed a Bayesian latent variable framework to evaluate the diagnostic test accuracy of the questionnaire using the randomized-response technique, which is based on the posterior densities of the subject's smoking behavior scores together with the breath test measurements. For different diagnostic test thresholds, we obtained moderate posterior mean estimates of sensitivity and specificity by observing a limited number of discrete randomized item responses.

Latent growth modeling of IRT versus CTT measured longitudinal latent variables.

Latent growth models are often used to measure individual trajectories representing change over time. The characteristics of the individual trajectories depend on the variability in the longitudinal outcomes. In many medical and epidemiological studies, the individual health outcomes cannot be observed directly and are indirectly observed through indicators (i.e. items of a questionnaire). An item response theory or a classical test theory measurement model is required, but the choice can influence the latent growth estimates. In this study, under various conditions, this influence is directly assessed by estimating latent growth parameters on a common scale for item response theory and classical test theory using a novel plausible value method in combination with Markov chain Monte Carlo. The latent outcomes are considered missing data and plausible values are generated from the corresponding posterior distribution, separately for item response theory and classical test theory. These plausible values are linearly transformed to a common scale. A Markov chain Monte Carlo method was developed to simultaneously estimate the latent growth and measurement model parameters using this plausible value technique. It is shown that estimated individual trajectories using item response theory, compared to classical test theory to measure outcomes, provide a more detailed description of individual change over time, since item response patterns (item response theory) are more informative about the health measurements than sum scores (classical test theory).

A Box-Cox normal model for response times.

The log-transform has been a convenient choice in response time modelling on test items. However, motivated by a dataset of the Medical College Admission Test where the lognormal model violated the normality assumption, the possibilities of the broader class of Box-Cox transformations for response time modelling are investigated. After an introduction and an outline of a broader framework for analysing responses and response times simultaneously, the performance of a Box-Cox normal model for describing response times is investigated using simulation studies and a real data example. A transformation-invariant implementation of the deviance information criterium (DIC) is developed that allows for comparing model fit between models with different transformation parameters. Showing an enhanced description of the shape of the response time distributions, its application in an educational measurement context is discussed at length.

Multilevel IRT using dichotomous and polytomous response data.

A structural multilevel model is presented where some of the variables cannot be observed directly but are measured using tests or questionnaires. Observed dichotomous or ordinal polytomous response data serve to measure the latent variables using an item response theory model. The latent variables can be defined at any level of the multilevel model. A Bayesian procedure Markov chain Monte Carlo (MCMC), to estimate all parameters simultaneously is presented. It is shown that certain model checks and model comparisons can be done using the MCMC output. The techniques are illustrated using a simulation study and an application involving students' achievements on a mathematics test and test results regarding management characteristics of teachers and principles.

Influenza surveillance in the Pacific Northwest 1976-1980.

Between June 1976 and June 1980 active year-round surveillance for influenza was carried out in Seattle in order to establish an early warning system. This report compares yield by different community groups and age. Waves of influenza virus infection appearing in three successive springs were followed in each instance by epidemics with the same subtype virus(es) in the following winter. These included two co-circulating A/H3N2 variants (A/Victoria/75 and A/Texas/77) in spring 1977 and winter 1977-1978, A/H1N1 (A/USSR) in spring 1978 and H1N1 (A/Brazil) winter 1978-1979, and type B influenza in spring 1979 and winter 1979-1980. Despite intensive surveillance through the summer and fall, the first isolate was not obtained until early December each year. Young adults (18-30) were as good sources for influenza viruses as children (less than 18).

Serum cortisol levels predict infarct size and patient mortality.

We have investigated prospectively the serum cortisol response to acute myocardial infarction in 70 consecutive patients admitted to a coronary care unit and we have shown that the levels are significantly raised early in the course of the illness and prior to elevation of the cardiac specific enzyme fraction, creatine kinase MB. The magnitude of the cortisol response is related to the size of the ensuing infarction (rs = 0.54) as calculated from the total creatine kinase MB release (P < 0.001) and very high levels (> 2000 mumol/l) are predictive of mortality (P < 0.05). Serum cortisol levels may have a role in the early identification of myocardial infarction and in predicting those patients with a poor prognosis.

Stochastic EM for estimating the parameters of a multilevel IRT model.

An item response theory (IRT) model is used as a measurement error model for the dependent variable of a multilevel model. The dependent variable is latent but can be measured indirectly by using tests or questionnaires. The advantage of using latent scores as dependent variables of a multilevel model is that it offers the possibility of modelling response variation and measurement error and separating the influence of item difficulty and ability level. The two-parameter normal ogive model is used for the IRT model. It is shown that the stochastic EM algorithm can be used to estimate the parameters which are close to the maximum likelihood estimates. This algorithm is easily implemented. The estimation procedure will be compared to an implementation of the Gibbs sampler in a Bayesian framework. Examples using real data are given.

Bayesian tests of measurement invariance.

Random item effects models provide a natural framework for the exploration of violations of measurement invariance without the need for anchor items. Within the random item effects modelling framework, Bayesian tests (Bayes factor, deviance information criterion) are proposed which enable multiple marginal invariance hypotheses to be tested simultaneously. The performance of the tests is evaluated with a simulation study which shows that the tests have high power and low Type I error rate. Data from the European Social Survey are used to test for measurement invariance of attitude towards immigrant items and to show that background information can be used to explain cross-national variation in item functioning.

A Multivariate Multilevel Approach to the Modeling of Accuracy and Speed of Test Takers.

Response times on test items are easily collected in modern computerized testing. When collecting both (binary) responses and (continuous) response times on test items, it is possible to measure the accuracy and speed of test takers. To study the relationships between these two constructs, the model is extended with a multivariate multilevel regression structure which allows the incorporation of covariates to explain the variance in speed and accuracy between individuals and groups of test takers. A Bayesian approach with Markov chain Monte Carlo (MCMC) computation enables straightforward estimation of all model parameters. Model-specific implementations of a Bayes factor (BF) and deviance information criterium (DIC) for model selection are proposed which are easily calculated as byproducts of the MCMC computation. Both results from simulation studies and real-data examples are given to illustrate several novel analyses possible with this modeling framework.

Modes of action of poliovirus vaccines and relation to resulting immunity.

Parenteral inactivated poliovirus vaccine (IPV) chiefly induces formation of serum antibody. Infection, with oral poliovirus vaccine ( OPV ) or wild poliovirus, also induces development of secretory IgA antibody. In addition, infection results in virus shedding and, hence, in possible spread of virus to contacts (more likely with wild than with vaccine virus). IPV induces a high level (90%-95%) of protection against disease, which presumably is mediated by serum antibody that prevents CNS invasion resulting from viremia. IPV induces little protection against infection but does modify the related virus shedding--chiefly that from the oropharynx. Infection, whether by OPV or wild virus, induces a high level of protection against disease. It also results in appreciable prevention or modification of infection because of the development of secretory IgA antibody. This effect is directly protective for the vaccinee and also benefits the community, since the exposed vaccinee can play little or no part in the spread of wild virus. One major question remaining concerns the maintenance of immunity. Lifelong immunologic memory assures an enhanced serum antibody response to any infection that occurs, but will it be sufficiently rapid that when preinfection antibody cannot be detected newly formed antibody will block blood-borne viral invasion of the CNS? If not, booster doses of vaccine are indicated. OPV boosters might be indicated in any case to reinforce protection against infection and so maintain herd immunity.

Herd immunity and measles.

The basic concept of herd immunity is directly applicable only under very special conditions. The agents of disease must be restricted to a single host species within which transmission occurs by relatively direct contact, and infection induces solid immunity. Also outbreaks must occur only in randomly mixing populations. In free-living populations, susceptibles are not distributed homogeneously but tend to cluster in subgroups defined by age and by such factors as ethnicity and socioeconomic status. The requisite for occurrence of epidemics, namely a large enough number of susceptibles in frequent contact with each other, exists in virtually all large populations, regardless of the total proportion of the population that is immune. Experience with measles illustrates these conditions. Total prevalence of immunity of greater than or equal to 90% in developing countries does not prevent annual epidemics among the susceptibles, most of whom are children younger than three years of age. Where vaccination is widely practiced, as in the United States since 1962, measles has continued to occur in poorly immunized subgroups that are characterized by low educational level and economic status, very young age, or religious beliefs forbidding acceptance of vaccine. Ultimate success of a systematic immunization program requires knowledge of distribution of susceptibles by age and subgroup and maximal effort to reduce the concentration of susceptibles throughout the community rather than aiming to reach any specific proportion of the overall population.

Is a rhinovirus vaccine possible?

Renewal of support for efforts to develop a rhinovirus vaccine seems justified in the light of newer epidemiologic and immunologic studies. The major contribution of RV to acute upper respiratory disease in all age groups but especially in young children emphasizes the public health importance of an effective vaccine. Epidemiologic surveillance of RV infections in widely separated areas has identified two relevant phenomena. First, in each area, certain serotypes were more frequently encountered and tended to persist. Such "common" serotypes accounted for a disproportionate share of the infections recognized and, hence, constitute special targets for immunization. Second, a clear increase over time in the proportion of RV isolates representing higher numbered (types 56-89) serotypes or untypable strains (potentially new serotypes) suggests that new serotypes continue to emerge as the result of progressive antigenic shift. The common origin of the multitudinous RV serotypes so suggested is consistent with the extensive antigenic cross-relations which are becoming evident. Systematic cross-testing with monospecific antisera, especially when high titer sera are employed, has revealed an appreciable number of one-way and reciprocal relations. Largely fortuitous observations of naturally or experimentally infected humans have revealed many additional cross-relations manifested by concurrent response to heterologous RV and presumably attributable to sensitizations resulting from prior RV infections. A model for this has been provided by rabbits immunized sequentially with different potent RV immunogens. Available information as displayed in figure 2 indicates that extensive cross-relations do exist and that sizeable groups of closely related serotypes may be identified. More intensive search for heterotypic response to infection of man coupled with selective use of the rabbit model should define the full extent and strength of cross-relations and identify completely the more closely related groups of serotypes as the basis for formulation of a broadly effective RV vaccine containing a limited number of serotypes. The cross-relations described are based entirely on development of serum neutralizing antibody, the presence of which in man is clearly correlated with relative protection against infection and disease. While this protection may well prove to be mediated largely by concomitant nasal secretory antibody, it is not unreasonable to expect that the antigenic cross-relations also would be manifested in secretory antibody response. These important questions concerning secretory antibody can be best explored in a limited series of volunteer trials with selected cross-related RV serotypes in which homotypic and heterotypic protection could be correlated with serum and nasal secretory antibody. The state of current knowledge, as I view it, is sufficient to justify initiation of such trials at any time.