Strong Predictive Model Research Articles

Forecasting Emergency Paediatric Asthma Hospital Admissions in Trinidad and Tobago: Development of a Local Model Incorporating the Interactions of Airborne Dust and Pollen Concentrations with Meteorological Parameters and a Time-Lag Factor

Respiratory diseases such as asthma and rhinitis are multifaceted disorders which are exacerbated by various factors including: gender, age, diet, genetic background, biological materials, allergens (pollen and spores), pollutants, meteorological conditions and dust particles. It is hypothesized that, the number of valid physician diagnosed cases of paediatric asthma, which has resulted in emergency room visits in Trinidad can be expressed as a function of the magnitude of pollen counts, particulate matter (PM10), and selected meteorological parameters. These parameters were used to develop a 7-day predictive model for paediatric asthma admittance. The data showed no obvious, strong correlations between paediatric asthma admissions and dust concentrations, and paediatric asthma admissions and pollen concentrations, when considered in isolation or in a linear fashion. However, using polynomial regression analysis, which looked at combinations of interactions, a strong 7-day predictive model for paediatric asthma admissions, was developed. The model was tested against actual data collated during the study period and showed a strong correlation (R2 = 0.85) between the regression model and the actual admissions data.

MEDIATING IMPACT OF MANUFACTURING TECHNOLOGY, LEAN AND STRATEGIC FLEXIBILITY ON MANUFACTURING PERFORMANCE

In respect of Malaysian lean manufacturer; there has been insufficient research on what are the winning formulae to overcome the mounting challenges from the business environment. In this study, three identified strategies i.e lean, manufacturing technology and strategic flexibility, that could positively improve manufacturing performances, are investigated in depth. The research approach used was quantitative with data analyzed using SmartPLS. Results from this study also emphasize several interesting quasi - paradoxical relationships which implied that Malaysian lean manufacturers prefer a cautious approach towards manufacturing technology implementation, preferring to incorporate technology in a more stable business environment and view strategic flexibility as unfavorable toward performance improvement. In conclusion, this study manages to produce a strong predictive strategy-performance model, which explains the mediation impact of manufacturing strategies on external environmental factors and manufacturing performance.

Merchandise sales rank in professional sport

Purpose – The purchase drivers of merchandise sales rank in professional sport are examined at both a conceptual and an empirical level. The paper aims to discuss these issues. Design/methodology/approach – A database was constructed for all 30 clubs in the National Hockey League based on a conceptual model of relevant variables. Both public and private data sources were accessed, covering an extensive 12-season period (1999-2011), including the 2004-2005 lockout when the season was cancelled. Principal-components analysis was used to reduce the number of variables for regression analysis to distinguish relatedness and to gauge the influence of those variables on merchandise sales rank. Findings – The results reveal that six club-based factors impact merchandise sales rank: Overall Fan Satisfaction, Media Exposure, On-Field Performance, Strength of a Club’s Brand, Local Market Dynamics, and Fan Capacity to Pay. These six categories of purchase drivers form a strong predictive model of merchandise sales rank in the National Hockey League. Research limitations/implications – The resulting model could be extended in future research by adding extra categories to the conceptual framework and by developing alternative or better measures of the variables the authors use. It could also be tested with other sales data as the dependent variable since the study was limited to ranking data on merchandise data for the National Hockey League clubs for each year. Future research could use the actual merchandise dollars for each club in a league to test the model. A further extension would be to model subparts of merchandise (such as jersey sales, impacts of brand changes, etc.). A similar area of future research would be to look at the role of individual athletes as opposed to clubs in driving merchandise sales rank or volume. Practical implications – For managers in professional sport, the results suggest that there are steps that can be taken to improve merchandise sales rank (and, by extension, merchandise sales volume). Practitioners can develop and follow strategies in this regard. Results also suggest that practitioners should put the achievement of high Regional Television Ratings – the strongest influencing variable on merchandise sales rank – as a priority. Social implications – The results confirm that brand is important when attempting to increase merchandise sales rank. Club managers need to be cognizant of their brand and its impact on merchandising in all decisions. Originality/value – The business of professional sport is evolving globally with new sources of revenue, including merchandising increasing in prominence. This research explores the drivers of merchandise sales rank in professional sport and provide direction on key antecedents. The study proposes and tests a conceptual model.

Clinical factors, including All Patient Refined Diagnosis Related Group severity, as predictors of early rehospitalization after COPD exacerbation.

Background:Patients hospitalized for chronic obstructive pulmonary disease (COPD) exacerbations carry a high risk for early rehospitalization. We wished to identify the basic clinical factors associated with a high risk of rehospitalization, and to see how well the standardized All Patient Refined Diagnosis Related Group (APR-DRG) severity of illness (SOI) subclassification predicted rehospitalization if combined with other simple clinical measures.Methods:We identified adult patients aged ≥40 years discharged from a major hospital in the Southwestern USA with a COPD discharge diagnosis during the study index period (1 October 2009 to 30 September 2010). Patients readmitted within 30 days (“early rehospitalization”) and 90 days (“any rehospitalization”) were each compared with those not rehospitalized. Clinical parameters (including demographics, comorbidities) and recent healthcare utilization were examined for their association with rehospitalization. Factors independently associated with rehospitalization were then combined with the index admission APR-DRG SOI assessment using conditional linear regression to find the best models in terms of the highest C-statistic.Results:Among 306 patients hospitalized for COPD, 62 (20.3%) had a rehospitalization within 90 days and 28 (9.2%), an early readmission. An APR-DRG SOI subclassification ≥3 was a modest independent predictor of early or any readmission, with adjusted odds ratios ranging from 2.09 to 3.33. Models that combined the APR-DRG SOI subclassification with clinical factors present before the index hospitalization had strong C-statistics of ≥0.80. Good models without the APR-DRG SOI subclassification but including a history of recent hospitalizations before the index hospitalization were also identified.Conclusions:An APR-DRG SOI subclassification of ≥3 for the index COPD admission is associated with an increased risk of early rehospitalizations, and can be combined with a few historical clinical factors to create strong predictive models for rehospitalization. This study demonstrates that hospitals can use commonly collected clinical information to help identify COPD patients at a high risk of failure after discharge.

An Ensemble Learning for Predicting Breakdown Field Strength of Polyimide Nanocomposite Films

Using the method of Stochastic Gradient Boosting, ten SMO‐SVR are constructed into a strong prediction model (SGBS model) that is efficient in predicting the breakdown field strength. Adopting the method of in situ polymerization, thirty‐two samples of nanocomposite films with different percentage compositions, components, and thicknesses are prepared. Then, the breakdown field strength is tested by using voltage test equipment. From the test results, the correlation coefficient (CC), the mean absolute error (MAE), the root mean squared error (RMSE), the relative absolute error (RAE), and the root relative squared error (RRSE) are 0.9664, 14.2598, 19.684, 22.26%, and 25.01% with SGBS model. The result indicates that the predicted values fit well with the measured ones. Comparisons between models such as linear regression, BP, GRNN, SVR, and SMO‐SVR have also been made under the same conditions. They show that CC of the SGBS model is higher than those of other models. Nevertheless, the MAE, RMSE, RAE, and RRSE of the SGBS model are lower than those of other models. This demonstrates that the SGBS model is better than other models in predicting the breakdown field strength of polyimide nanocomposite films.

Using Baseball Data as a Gentle Introduction to Teaching Linear Regression

This effort describes a successful classroom exercise to introduce simple and multiple linear regression to working professional MBA students. The exercise starts by exploring the relationship between a baseball team’s payroll with its winning percentage. The exercise then continues with the introduction of additional predictor variables so that the students are able to build a strong predictive model for winning percentage. Student feedback consistently praises the exercise as an effective way to learn about linear regression.

Modeling and forecasting the distribution of Vibrio vulnificus in Chesapeake Bay.

To construct statistical models to predict the presence, abundance and potential virulence of Vibrio vulnificus in surface waters of Chesapeake Bay for implementation in ecological forecasting systems. We evaluated and applied previously published qPCR assays to water samples (n = 1636) collected from Chesapeake Bay from 2007-2010 in conjunction with State water quality monitoring programmes. A variety of statistical techniques were used in concert to identify water quality parameters associated with V. vulnificus presence, abundance and virulence markers in the interest of developing strong predictive models for use in regional oceanographic modeling systems. A suite of models are provided to represent the best model fit and alternatives using environmental variables that allow them to be put to immediate use in current ecological forecasting efforts. Environmental parameters such as temperature, salinity and turbidity are capable of accurately predicting abundance and distribution of V. vulnificus in Chesapeake Bay. Forcing these empirical models with output from ocean modeling systems allows for spatially explicit forecasts for up to 48 h in the future. This study uses one of the largest data sets compiled to model Vibrio in an estuary, enhances our understanding of environmental correlates with abundance, distribution and presence of potentially virulent strains and offers a method to forecast these pathogens that may be replicated in other regions.

Boosted regression trees

Purpose– The purpose of this paper is to apply boosted regression trees (BRT) to a heterogeneous data set of residential property drawn from a jurisdiction in Malaysia, with the objective to evaluate its application within the mass appraisal environment in Malaysia. Machine learning (ML) techniques have been applied to real estate mass appraisal with varying degrees of success.Design/methodology/approach– To evaluate the performance of the BRT model two multiple regression analysis (MRA) models have been specified (linear and non-linear). One of the weaknesses of traditional regression is the need to a priori specify the functional form of the model and to ensure that all non-linearities have been accounted for. For a BRT model the algorithm does not require any predetermined model or variable transformations, making the process much simpler.Findings– The results show that the BRT model outperformed the MRA-specified models in terms of the coefficient of dispersion and mean absolute percentage error. While the results are encouraging, BRT models still lack transparency and suffer from the inability to translate variable importance into quantifiable variable effects.Practical implications– This paper presents a useful alternative modelling technique, BRT, for use within the mass appraisal environment in Malaysia. Its advantages include less intensive data cleansing, no requirement to specify the predictive underlying model, ability to utilise categorical variables without the need to transform them and not as data hungry, as for example, MRA.Originality/value– This paper adds to the knowledge in this area by applying a relatively new ML model, BRT to residential property data from a jurisdiction in Malaysia. BRT has shown promise as a strong predictive model when applied in other disciplines; therefore this research empirically tests this finding within real estate valuation.

Evaluating sample allocation and effort in detecting population differentiation for discrete and continuously distributed individuals

One of the most pressing issues in spatial genetics concerns sampling. Traditionally, substructure and gene flow are estimated for individuals sampled within discrete populations. Because many species may be con- tinuously distributed across a landscape without discrete boundaries, understanding sampling issues becomes para- mount. Given large-scale, geographically broad conserva- tion efforts, researchers are looking for guidance as to the trade-offs between sampling more individuals within a population versus few individuals scattered across more populations. Here, we conducted simulations that address these issues. We first established two archetypical patterns of dispersion: (1) individuals within discrete populations, and (2) continuously distributed individuals with limited dispersal. We used genotypes generated from a spatially- explicit, individual-based program and simulated genetic structure in individuals from nine different population sizes across a landscape that either had barriers to movement (defining discrete populations) or isolation-by-distance patterns (defining continuously distributed individuals). Then, given each pattern of dispersion, we allocated sam- ples across four different sampling strategies for each of the nine population sizes in various configurations for sampling more individuals within a population versus fewer individuals scattered across more populations. We assessed the population genetic substructure with both the population-based metric, FST, and an individual-based metric, DPS regardless of the true pattern of dispersion to allow us to better understand the effect of incorrectly matching the metric and the distribution (e.g., FST with continuously distributed individuals, and vice versa). We show that sampling many subpopulations (or sampling areas), thus sampling fewer individuals per subpopulation, overestimates measures of population subdivision with the population-based metric for both patterns of dispersion. In contrast, using the individual-based metric gives the opposite results: sampling too few subpopulations, and many individuals per subpopulation, produces an under- estimate of the strength of isolation-by-distance. By com- paring all results, we were able to suggest a strong predictive model of a chosen genetic structure metric for elucidating the sampling design trade-offs given each pat- tern of dispersion and configuration on the landscape.

Predicting mild traumatic brain injury patients at risk of persistent symptoms in the Emergency Department

Objective: To identify factors that can predict which emergency department (ED) patients with mTBI are likely to develop persistent post-concussion symptoms (PPCS).Design: A matched case-control study was conducted at a Level 1 trauma centre between June 2006 and July 2009. Patients diagnosed with mTBI in the ED and diagnosed at a concussion management programme with at least one PPCS (85 cases) were compared to patients diagnosed with mTBI in the ED (340 controls) to determine if factors assessed at the time of ED presentation could predict patients likely to develop persistent symptoms.Results: Multivariable hierarchical logistic regression with variables indicating increased risk for PPCS (prior mTBI, history of depression, history of anxiety, multiple injury, forgetfulness/poor memory, noise sensitivity, or light sensitivity) resulted in a final predictive model including prior mTBI, history of anxiety, forgetfulness/poor memory and light sensitivity. The final model had a specificity of 87.9% and a sensitivity of 69.9%.Conclusions: A strong prediction model to identify those ED patients with mTBI at risk for PPCS was developed and could be easily implemented in the ED; therefore, helping to target those patients who would potentially benefit from close follow-up.

Development and Validation of a Necrotizing Soft-Tissue Infection Mortality Risk Calculator Using NSQIP

Necrotizing soft-tissue infections (NSTI) are a group of uncommon, rapidly progressive infections requiring prompt surgical debridement and systemic support. A previous attempt to define risk factors for mortality from NSTI had multiple limitations. The objective of this study was to develop and validate a 30-day postoperative mortality risk calculator for patients with NSTI using NSQIP. The NSQIP Participant Use Files (2005-2010) were used as the primary data source. Patients diagnosed with NSTI were identified by ICD-9 codes. Multiple logistic regression analysis identified key preoperative variables predicting mortality. Bootstrap analysis was used to validate the model. In 1,392 identified NSTI cases, demographics were as follows: 42% were female, median age was 55 years (interquartile range 46 to 63 years), and median body mass index was 32 kg/m(2) (interquartile range 26 to 40 kg/m(2)). Thirty-day mortality was 13%. Seven independent variables were identified that correlated with mortality: age older than 60 years (odds ratio [OR]= 2.5; 95% CI 1.7-3.6), functional status (partially dependent: OR= 1.6; 95% CI 1.0-2.7; totally dependent: OR= 2.3; 95% CI 1.4-3.8), requiring dialysis (OR= 1.9; 95% CI 1.2-3.1), American Society of Anesthesiologists class 4 or higher (OR= 3.6; 95% CI 2.3-5.6), emergent surgery (OR= 1.6; 95% CI 1.0-2.3), septic shock (OR= 2.4; 95% CI 1.6-3.6), and low platelet count (<50K/μL: OR= 3.5; 95% CI 1.6-7.4; <150K/μL but >50K/μL: OR= 1.9; 95% CI 1.2-2.9). The receiver operating characteristic area was 0.85 (95% CI 0.82-0.87), which indicated a strong predictive model. Using bootstrap validation, the optimism-corrected receiver operating characteristic area was 0.83 (95% CI 0.81-0.86), which represents the model performance in future patients. The model was used to develop an interactive risk calculator. This risk calculator has excellent predictive ability for mortality in patients with NSTI. This simple interactive tool can aid physicians and patients in the decision-making process.

Automated flare prediction using the AdaBoost algorithm

We propose a flare prediction method based on the AdaBoost algorithm, which constructs a strong prediction model from a combination of several basic models. Three predictors, extracted from the photospheric magnetograms, are applied as features to predict the occurrence of flares with a certain level over 24 hours following the time when the magnetogram is recorded. To demonstrate the effectiveness of the proposed method, comparisons of experimental results with respect to some existing methods are given. The results show that an improvement is achieved in predicting the occurrences of large flares.

Predictive Power of Standard Variables on Postsecondary Achievement in Arkansas

Today, as many as 25 to 40 percent of students who attend college qualify for some form of remedial education program provided by postsecondary institutions (Kaye, Lord, Bottoms, Presson, &amp; Cornet, 2006). Many colleges and universities view the inclusion of remediation as an integral part of their educational mission. However, the costs of such remedial programs are high. While estimates of actual costs vary greatly, most assume a national remediation expense of over one billion dollars annually. In Arkansas, the cost of remediation is over $53 million, up over 65% from costs about a decade ago (ADHE, 2006). The remediation policy in Arkansas relies on the premise that there is a single measure, namely ACT, which can predict the level of college readiness and, subsequently, the need for remediation. This study examined the predictive power of the ACT to determine college success, the relationship between the ACT and other measures of student achievement, and explored combinations of variables to determine if a stronger predictive model could be identified.

Earnings-Dividend Relationship in Corporate Nigeria; A Test of Predictive Efficacy

Prompted by the need for an empirical evaluation of the relationship between corporate net earnings and dividend payouts of quoted Nigerian firms as well as the search for a strong predictive model for this relationship, the study evaluates the predictive efficacies of current- and one- year- lagged earnings regression models among the Nigerian quoted firms. Applying the ordinary least square regression analysis on one hundred and four (104) firms selected as the study sample, the results indicate that dividend payouts are relatively more sensitive to current earnings per share compared to past earnings per share. Further, the percentage change in dividend payouts attributable to changes in current earnings per share is found to be relatively higher than that attributable to changes in past earnings per share, thus providing evidence that current earnings model is relatively more effective in predicting the dividend payouts of Nigerian quoted firms. The study recommends strong information dissemination to all stakeholders in the Nigerian capital market in order to improve market efficiency and potential benefits derivable from the market by all participants.

A time-series modeling method based on the boosting gradient-descent theory

The forecasting of time-series data plays an important role in various domains. It is of significance in theory and application to improve prediction accuracy of the time-series data. With the progress in the study of time-series, time-series forecasting model becomes more complicated, and consequently great concern has been drawn to the techniques in designing the forecasting model. A modeling method which is easy to use by engineers and may generate good results is in urgent need. In this paper, a gradient-boost AR ensemble learning algorithm (AREL) is put forward. The effectiveness of AREL is assessed by theoretical analyses, and it is demonstrated that this method can build a strong predictive model by assembling a set of AR models. In order to avoid fitting exactly any single training example, an insensitive loss function is introduced in the AREL algorithm, and accordingly the influence of random noise is reduced. To further enhance the capability of AREL algorithm for non-stationary time-series, improve the robustness of algorithm, discourage overfitting, and reduce sensitivity of algorithm to parameter settings, a weighted kNN prediction method based on AREL algorithm is presented. The results of numerical testing on real data demonstrate that the proposed modeling method and prediction method are effective.

Predicting Response of Vitreous Hemorrhage after Intravitreous Injection of Highly Purified Ovine Hyaluronidase (Vitrase) in Patients with Diabetes

To develop a predictive model for patients with diabetes who are most likely to have vitreous hemorrhage clearing by 3 months after a single, intravitreous injection of highly purified, preservative-free, ovine hyaluronidase (Vitrase; ISTA Pharmaceuticals, Inc., Irvine, CA). Post hoc data analysis was performed on two randomized, double-masked, placebo-controlled, phase 3 clinical trials of a single intravitreous injection of Vitrase for severe vitreous hemorrhage. Vitreous hemorrhage density was scored using a 0 to 4 vitreous hemorrhage grading scale in 12 radial segments of the fundus ("clock hours"). Reduction in total hemorrhage point score (DeltaTHPS) between baseline and 1 month after injection was analyzed as a predictor of vitreous hemorrhage outcome at 3 months. A strong predictive model was demonstrated by receiver operating characteristic (ROC) curve analysis; area under the curve (AUC) = 0.845 (P < 0.0001). The DeltaTHPS was higher in hyaluronidase-treated subjects than in saline-treated control subjects. Median DeltaTHPS was 8.0 and 6.0 in subjects treated with 55 IU (68 USP) and 75 IU (93 USP) of hyaluronidase respectively, versus 2.0 in saline control subjects (P < 0.0001). discussion. The DeltaTHPS at 1 month provides quantitative guidance for predicting the outcome of a single intravitreous ovine hyaluronidase injection in patients with diabetes and severe vitreous hemorrhage (ClinicalTrials.gov numbers, NCT00198510 and NCT00198497).

Delayed Graft Function: Utility of Predictive Models

To the Editor: We read with interest the article by Grossberg et al. (1). They have questioned the utility of a weighted nomogram designed to predict the risk of delayed graft function (DGF) published by Irish et al. (2) in 2003. Irish et al. designed a scoring system based on a large patient cohort (13 846), which was validated in 20 684 transplantations in the Scientific Renal Transplant Registry. Grossberg et al. applied this nomogram to a cohort of 169 patients and found that the ability to predict DGF was poor. Although Irish et al. recommended using their nomogram in a single case scenario; the model lacks high sensitivity and specificity at any given threshold. For instance, at 0.2 threshold the sensitivity is 79.75 and the specificity is 53.12; and at 0.4 threshold the sensitivity is 28.97 and the specificity is 89.24. Grossberg et al. did not quantify the diagnostic potential of the tested nomogram when applied to their patient cohort. On calculation of sensitivity and specificity of the nomogram when applied to Grossberg study population, we found that these indices vary from the original values (Figure 1). These sensitivity and specificity values at any given threshold are not high enough to suggest a strong predictive model. However this may be a reflection of the suboptimal diagnostic potential of the original nomogram. The predictive ability of a model should be reproducible when applied to a population of similar characteristics. The 95% confidence intervals of the predictive indices when applied to two patient populations do not overlap (Figure 1). This suggests that the two groups may not be similar in their characteristics. The tested nomogram provides percent risk of developing DGF. For majority of the cases the percent risk derived from the nomogram will not be high or low. Thus, the model may be accurate in predicting the percentage risk for a patient, but it is not very useful if the vast majority of the percentage risks are ‘near average’. The 95% confidence intervals of the predictive indices when applied to two patient populations do not overlap. The utility of predictive model lies in its applicability to a generalized population, so that interventions could be introduced for patients at risk. In our opinion, a patient population that is large enough to represent any random patient population may be required to design a robust predictive model. Postimplantation factors such as ischemia/reperfusion injury and immune activation may have an influence on DGF, which in turn may have an impact on the diagnostic potential of pretransplantation predictive models. In summary, a unique and reliable model is not yet available to predict DGF, and stronger models can be designed through additional studies. In addition to prediction the problem of DGF may be minimized by exploring the mechanistic pathways and by conducting clinical trials, which use strategies to correct early defects at cellular and molecular level.

Land-use history, forest conversion, and soil organic carbon in pine plantations and native forests of south eastern Australia

Land-use history – the number, type, and duration of previous land uses – is relevant to many questions regarding land-use effects on soil carbon, but is infrequently reported. We examine the importance of land-use history variables, as well as topographic and edaphic variables, on soil C in a range of forest types – native forest, pine plantations, secondary forest and rehabilitated forest – at three contrasting locations in south eastern Australia. Our comparisons include a novel forest conversion of exotic pine plantations to native, broadleaf forest. Using nested ANOVAs, we detected few differences in soil C concentration indices (total C, microbial biomass C, K 2SO 4–C) and C content among eucalypt-dominated vegetation and pine plantations within each location (0–10 cm depth). However, planned contrasts indicated a 30% decrease in soil C content with conversion of native forest to pine plantation of age 37 years. The reverse land-use change – pine plantation to native, broadleaf forest – was associated with a decrease in soil C concentration and content at one location (40%; age 12–13 years) and no detectable changes at another (to age 7 years). Variable effect between locations of this novel land-use change on soil C could be due to differences in potential productivity, conifer species, and plantation age. We used correlation coefficients and general linear models to identify widely applicable variables for predicting soil C concentration and content at local scales (≤ 20 km 2). Within-location relationships with topographic variables were weak and infrequent relative to those with edaphic and land-use history variables. Soil texture was strongly correlated with soil C at each location, although the relative significance of different particle size fractions differed among locations. Electrical conductivity appeared more widely applicable since it was included in C models at two locations. Combining land-use history and edaphic variables produced strong predictive models for soil C concentrations and content at two locations (total r 2 0.83 to 0.95). Positive relationships were indicated between soil C and ‘age of current vegetation’ at one location, and negative relationships were indicated with ‘number of land uses’ at another. These data highlight a potential predictive role for land-use history variables in local-scale assessments of soil C in forested landscapes.

Predicting Toxicity Equivalents in Fish Tissue

Toxicity equivalents is a measure of “dioxin-like” toxicity contributed by polychlorinated biphenyls (PCBs), polychlorinated dibenzo-p-dioxins (dioxins), and polychlorinated dibenzofurans (furans). Calculation of toxicity equivalents require the use of analytical procedures that are expensive and slow, making them impractical for routine analysis. U.S. Environmental Protection Agency National Fish Tissue Study (2003) data were used to determine the most significant predictors of toxicity equivalents with multiple regression analysis. The strongest predictive model (P < 0.0001, R2 = 0.97) included five compounds (PCB-118; PCB-126; 2,3,7,8-TCDD; 1,2,3,6,7,8-HxCDD; 2,3,4,7,8-PeCDF). However, the required lower limit of detection for an analytical method measuring these congeners is 0.1 ppt and would not provide much benefit over the current analytical method. An alternative model (P < 0.0001, R2 = 0.68) that included three PCBs (PCB-138, PCB-153, PCB-118) would require a limit of detection of 1,000 ppt and be more practical. This research demonstrates that the measurement of selected compounds can be used to estimate toxicity equivalents and consequently serve as the impetus for the development of lower cost, rapid analytical methods for analysis of fish.

Predictors of atrial fibrillation after off-pump coronary artery bypass graft surgery

Objective: Atrial fibrillation is the most common complication after coronary artery bypass graft surgery. This arrhythmia may lead to hemodynamic compromise, prolonged hospitalization, and increased risk for cerebral thromboembolism. Older age is the only variable consistently associated with the development of postoperative atrial fibrillation; however, no strong predictive model exists. The purpose of this study was to identify perioperative characteristics associated with new-onset atrial fibrillation in patients undergoing off-pump coronary artery bypass grafting. Design: Prospective, observational. Setting: University tertiary care hospital. Participants: One hundred sixty consecutive patients undergoing off-pump coronary artery bypass grafting. Interventions: None. Measurements and Main Results: Incidence of postoperative atrial fibrillation was the major outcome. Atrial fibrillation occurred in 33 patients (20.6%). Multivariate analysis identified reintervention (odds ratio 26.8), revascularization of the ramus medianus (odds ratio 3.9), and age (odds ratio 1.069 per year) as the only independent predictors of postoperative atrial fibrillation. All patients were in sinus rhythm at hospital discharge. One hospital death was noted. Conclusions: Despite the less invasive approach, the incidence of postoperative atrial fibrillation is high after off-pump coronary artery bypass grafting. Older age, grafting of the ramus medianus, and a redo operation were predictors of new-onset postoperative atrial fibrillation. It is possible that left atrial stretching with heart dislocation during revascularization of the lateral wall could lead to postoperative atrial fibrillation.