A comparison of Markov and generalized linear models of hospital mortality

This paper reports a comparison of three modeling strategies for the analysis of hospital mortality in a sample of general medicine inpatients in a Department of Veterans Affairs medical center. Logistic regression, a Markov chain model, and longitudinal logistic regression were evaluated on predictive performance as measured by the c-index and on accuracy of expected numbers of deaths compared to observed. The logistic regression used patient information collected at admission; the Markov model was comprised of two absorbing states for discharge and death and three transient states reflecting increasing severity of illness as measured by laboratory data collected during the hospital stay; longitudinal regression employed Generalized Estimating Equations (GEE) to model covariance structure for the repeated binary outcome. Results showed that the logistic regression predicted hospital mortality as well as the alternative methods but was limited in scope of application. The Markov chain provides insights into how day to day changes of illness severity lead to discharge or death. The longitudinal logistic regression showed that increasing illness trajectory is associated with hospital mortality. The conclusion is reached that for standard applications in modeling hospital mortality, logistic regression is adequate, but for new challenges facing health services research today, alternative methods are equally predictive, practical, and can provide new insights. ^

Cognitive Differences in Chart Reading: A Comparison of Nurses and Physicians

This pilot study compares the mental models of a patient constructed by nurses and physicians while reading an electronic medical record. Preliminary results suggest that the participants' summaries were both quantitatively and qualitatively different. The physician made more inferences and focused on deeper relationships in the record, whereas the nurse focused on the descriptive surface structure of the record.

The physical map of {\it Escherichia coli\/} K-12 strain MG1655: Construction, comparison with other strains and implications for mechanisms controlling oligonucleotide frequency

Complete NotI, SfiI, XbaI and BlnI cleavage maps of Escherichia coli K-12 strain MG1655 were constructed. Techniques used included: CHEF pulsed field gel electrophoresis; transposon mutagenesis; fragment hybridization to the ordered $\lambda$ library of Kohara et al.; fragment and cosmid hybridization to Southern blots; correlation of fragments and cleavage sites with EcoMap, a sequence-modified version of the genomic restriction map of Kohara et al.; and correlation of cleavage sites with DNA sequence databases. In all, 105 restriction sites were mapped and correlated with the EcoMap coordinate system.^ NotI, SfiI, XbaI and BlnI restriction patterns of five commonly used E. coli K-12 strains were compared to those of MG1655. The variability between strains, some of which are separated by numerous steps of mutagenic treatment, is readily detectable by pulsed-field gel electrophoresis. A model is presented to account for the difference between the strains on the basis of simple insertions, deletions, and in one case an inversion. Insertions and deletions ranged in size from 1 kb to 86 kb. Several of the larger features have previously been characterized and some of the smaller rearrangements can potentially account for previously reported genetic features of these strains.^ Some aspects of the frequency and distribution of NotI, SfiI, XbaI and BlnI cleavage sites were analyzed using a method based on Markov chain theory. Overlaps of Dam and Dcm methylase sites with XbaI and SfiI cleavage sites were examined. The one XbaI-Dam overlap in the database is in accord with the expected frequency of this overlap. The occurrence of certain types of SfiI-Dcm overlaps are overrepresented. Of the four subtypes of SfiI-Dcm overlap, only one has a partial inhibitory effect on the activity of SfiI. Recognition sites for all four enzymes are rarer than expected based on oligonucleotide frequency data, with this effect being much stronger for XbaI and BlnI than for NotI and SfiI. The latter two enzyme sites are rare mainly due to apparent negative selection against GGCC (both) and CGGCCG (NotI). The former two enzyme sites are rare mainly due to effects of the VSP repair system on certain di-tri- and tetranucleotides, most notably CTAG. Models are proposed to explain several of the anomalies of oligonucleotide distribution in E. coli, and the biological significance of the systems that produce these anomalies is discussed. ^

Comparison of a traditional and a multilevel Cox Proportional Hazards model

Hierarchically clustered populations are often encountered in public health research, but the traditional methods used in analyzing this type of data are not always adequate. In the case of survival time data, more appropriate methods have only begun to surface in the last couple of decades. Such methods include multilevel statistical techniques which, although more complicated to implement than traditional methods, are more appropriate. ^ One population that is known to exhibit a hierarchical structure is that of patients who utilize the health care system of the Department of Veterans Affairs where patients are grouped not only by hospital, but also by geographic network (VISN). This project analyzes survival time data sets housed at the Houston Veterans Affairs Medical Center Research Department using two different Cox Proportional Hazards regression models, a traditional model and a multilevel model. VISNs that exhibit significantly higher or lower survival rates than the rest are identified separately for each model. ^ In this particular case, although there are differences in the results of the two models, it is not enough to warrant using the more complex multilevel technique. This is shown by the small estimates of variance associated with levels two and three in the multilevel Cox analysis. Much of the differences that are exhibited in identification of VISNs with high or low survival rates is attributable to computer hardware difficulties rather than to any significant improvements in the model. ^

Bayesian generalized linear models for meta-analysis of diagnostic tests

With the recognition of the importance of evidence-based medicine, there is an emerging need for methods to systematically synthesize available data. Specifically, methods to provide accurate estimates of test characteristics for diagnostic tests are needed to help physicians make better clinical decisions. To provide more flexible approaches for meta-analysis of diagnostic tests, we developed three Bayesian generalized linear models. Two of these models, a bivariate normal and a binomial model, analyzed pairs of sensitivity and specificity values while incorporating the correlation between these two outcome variables. Noninformative independent uniform priors were used for the variance of sensitivity, specificity and correlation. We also applied an inverse Wishart prior to check the sensitivity of the results. The third model was a multinomial model where the test results were modeled as multinomial random variables. All three models can include specific imaging techniques as covariates in order to compare performance. Vague normal priors were assigned to the coefficients of the covariates. The computations were carried out using the 'Bayesian inference using Gibbs sampling' implementation of Markov chain Monte Carlo techniques. We investigated the properties of the three proposed models through extensive simulation studies. We also applied these models to a previously published meta-analysis dataset on cervical cancer as well as to an unpublished melanoma dataset. In general, our findings show that the point estimates of sensitivity and specificity were consistent among Bayesian and frequentist bivariate normal and binomial models. However, in the simulation studies, the estimates of the correlation coefficient from Bayesian bivariate models are not as good as those obtained from frequentist estimation regardless of which prior distribution was used for the covariance matrix. The Bayesian multinomial model consistently underestimated the sensitivity and specificity regardless of the sample size and correlation coefficient. In conclusion, the Bayesian bivariate binomial model provides the most flexible framework for future applications because of its following strengths: (1) it facilitates direct comparison between different tests; (2) it captures the variability in both sensitivity and specificity simultaneously as well as the intercorrelation between the two; and (3) it can be directly applied to sparse data without ad hoc correction. ^

Children with acute leukemia: A comparison of outcomes and cost-effectiveness from allogeneic blood stem cell and bone marrow transplantation

The relative merits of PBSCT versus BMT for children with standard and high risk hematologic malignancies remain unclear. In a retrospective single center study, we compared allogeneic peripheral blood stem cell transplantation (PBSCT) (n=30) with bone marrow transplantation (BMT) (n=110) in children with acute leukemia. We studied recipients of HLA matched sibling stem cells, and of stem cells from alternative donors (HLA mismatched and/or unrelated) and determined whether sourcing the stem cells from PB or marrow affected engraftment, incidence of acute and chronic GvHD, and disease-free survival at 1 year. Our results show a modest reduction in time to engraftment from PB stem cells and no greater risk of GvHD, but illustrate that the severity of the underlying disease is by far the greatest determinant of 1 year survival. Patients in the BMT group had a higher treatment success rate and lower costs than the recipients of the PBSCT within the standard but not the high risk disease group, where the treatment success rate and the cumulative costs were lower in the PBSCT group compared to the BMT group. Our current incremental cost-effectiveness ratio and analysis of uncertainty suggest that allogeneic transplantation of bone marrow grafts was a more cost-effective treatment option compared to peripheral blood stem cells in patients with standard risk childhood acute leukemia disease. For high risk disease our data are less prescriptive, since the differences were more limited and the range of costs much larger. Neither option demonstrated a clear advantage from a cost-effectiveness standpoint.^

Comparison of Bayesian and frequentist approaches in clinical outcomes between hospitals

Many public health agencies and researchers are interested in comparing hospital outcomes, for example, morbidity, mortality, and hospitalization across areas and hospitals. However, since there is variation of rates in clinical trials among hospitals because of several biases, we are interested in controlling for the bias and assessing real differences in clinical practices. In this study, we compared the variations between hospitals in rates of severe Intraventricular Haemorrhage (IVH) infant using Frequentist statistical approach vs. Bayesian hierarchical model through simulation study. The template data set for simulation study was included the number of severe IVH infants of 24 intensive care units in Australian and New Zealand Neonatal Network from 1995 to 1997 in severe IVH rate in preterm babies. We evaluated the rates of severe IVH for 24 hospitals with two hierarchical models in Bayesian approach comparing their performances with the shrunken rates in Frequentist method. Gamma-Poisson (BGP) and Beta-Binomial (BBB) were introduced into Bayesian model and the shrunken estimator of Gamma-Poisson (FGP) hierarchical model using maximum likelihood method were calculated as Frequentist approach. To simulate data, the total number of infants in each hospital was kept and we analyzed the simulated data for both Bayesian and Frequentist models with two true parameters for severe IVH rate. One was the observed rate and the other was the expected severe IVH rate by adjusting for five predictors variables for the template data. The bias in the rate of severe IVH infant estimated by both models showed that Bayesian models gave less variable estimates than Frequentist model. We also discussed and compared the results from three models to examine the variation in rate of severe IVH by 20th centile rates and avoidable number of severe IVH cases. ^

Quantitative comparison of late effects following photon versus proton external-beam radiation therapies: Toward an evidence-based approach to selecting a treatment modality

Radiation therapy has been used as an effective treatment for malignancies in pediatric patients. However, in many cases, the side effects of radiation diminish these patients’ quality of life. In order to develop strategies to minimize radiogenic complications, one must first quantitatively estimate pediatric patients’ relative risk for radiogenic late effects, which has not become feasible till recently because of the calculational complexity. The goals of this work were to calculate the dose delivered to tissues and organs in pediatric patients during contemporary photon and proton radiotherapies; to estimate the corresponding risk of radiogenic second cancer and cardiac toxicity based on the calculated doses and on dose-risk models from the literature; to test for the statistical significance of the difference between predicted risks after photon versus proton radiotherapies; and to provide a prototype of an evidence-based approach to selecting treatment modalities for pediatric patients, taking second cancer and cardiac toxicity into account. The results showed that proton therapy confers a lower predicted risk of radiogenic second cancer, and lower risks of radiogenic cardiac toxicities, compared to photon therapy. An uncertainty analysis revealed that the qualitative findings of this study are insensitive to changes in a wide variety of host and treatment related factors.

A comparison of techniques for forecasting utilization of Veterans Affairs hospitals

This study demonstrated that accurate, short-term forecasts of Veterans Affairs (VA) hospital utilization can be made using the Patient Treatment File (PTF), the inpatient discharge database of the VA. Accurate, short-term forecasts of two years or less can reduce required inventory levels, improve allocation of resources, and are essential for better financial management. These are all necessary achievements in an era of cost-containment.^ Six years of non-psychiatric discharge records were extracted from the PTF and used to calculate four indicators of VA hospital utilization: average length of stay, discharge rate, multi-stay rate (a measure of readmissions) and days of care provided. National and regional levels of these indicators were described and compared for fiscal year 1984 (FY84) to FY89 inclusive.^ Using the observed levels of utilization for the 48 months between FY84 and FY87, five techniques were used to forecast monthly levels of utilization for FY88 and FY89. Forecasts were compared to the observed levels of utilization for these years. Monthly forecasts were also produced for FY90 and FY91.^ Forecasts for days of care provided were not produced. Current inpatients with very long lengths of stay contribute a substantial amount of this indicator and it cannot be accurately calculated.^ During the six year period between FY84 and FY89, average length of stay declined substantially, nationally and regionally. The discharge rate was relatively stable, while the multi-stay rate increased slightly during this period. FY90 and FY91 forecasts show a continued decline in the average length of stay, while the discharge rate is forecast to decline slightly and the multi-stay rate is forecast to increase very slightly.^ Over a 24 month ahead period, all three indicators were forecast within a 10 percent average monthly error. The 12-month ahead forecast errors were slightly lower. Average length of stay was less easily forecast, while the multi-stay rate was the easiest indicator to forecast.^ No single technique performed significantly better as determined by the Mean Absolute Percent Error, a standard measure of error. However, Autoregressive Integrated Moving Average (ARIMA) models performed well overall and are recommended for short-term forecasting of VA hospital utilization. ^

A comparison of worker risk of injury while working on fully automated side loaders and other residential collection trucks at a large solid waste collection company

The study aim was to determine whether using automated side loader (ASL) trucks in higher proportions compared to other types of trucks for residential waste collection results in lower injury rates (from all causes). The primary hypothesis was that the risk of injury to workers was lower for those who work with ASL trucks than for workers who work with other types of trucks used in residential waste collection. To test this hypothesis, data were collected from one of the nation’s largest companies in the solid waste management industry. Different local operating units (i.e. facilities) in the company used different types of trucks to varying degrees, which created a special opportunity to examine refuse collection injuries and illnesses and the risk reduction potential of ASL trucks.^ The study design was ecological and analyzed end-of-year data provided by the company for calendar year 2007. During 2007, there were a total of 345 facilities which provided residential services. Each facility represented one observation.^ The dependent variable – injury and illness rate, was defined as a facility’s total case incidence rate (TCIR) recorded in accordance with federal OSHA requirements for the year 2007. The TCIR is the rate of total recordable injury and illness cases per 100 full-time workers. The independent variable, percent of ASL trucks, was calculated by dividing the number of ASL trucks by the total number of residential trucks at each facility.^ Multiple linear regression models were estimated for the impact of the percent of ASL trucks on TCIR per facility. Adjusted analyses included three covariates: median number of hours worked per week for residential workers; median number of months of work experience for residential workers; and median age of residential workers. All analyses were performed with the statistical software, Stata IC (version 11.0).^ The analyses included three approaches to classifying exposure, percent of ASL trucks. The first approach included two levels of exposure: (1) 0% and (2) >0 - <100%. The second approach included three levels of exposure: (1) 0%, (2) ≥ 1 - < 100%, and (3) 100%. The third approach included six levels of exposure to improve detection of a dose-response relationship: (1) 0%, (2) 1 to <25%, (3) 25 to <50%, (4) 50 to <75%, (5) 75 to <100%, and (6) 100%. None of the relationships between injury and illness rate and percent ASL trucks exposure levels was statistically significant (i.e., p<0.05), even after adjustment for all three covariates.^ In summary, the present study shows that there is some risk reduction impact of ASL trucks but not statistically significant. The covariates demonstrated a varied yet more modest impact on the injury and illness rate but again, none of the relationships between injury and illness rate and the covariates were statistically significant (i.e., p<0.05). However, as an ecological study, the present study also has the limitations inherent in such designs and warrants replication in an individual level cohort design. Any stronger conclusions are not suggested.^

Comparison of machine learning approaches and statistical methods for estimation of airborne chemical exposures

Accurate quantitative estimation of exposure using retrospective data has been one of the most challenging tasks in the exposure assessment field. To improve these estimates, some models have been developed using published exposure databases with their corresponding exposure determinants. These models are designed to be applied to reported exposure determinants obtained from study subjects or exposure levels assigned by an industrial hygienist, so quantitative exposure estimates can be obtained. ^ In an effort to improve the prediction accuracy and generalizability of these models, and taking into account that the limitations encountered in previous studies might be due to limitations in the applicability of traditional statistical methods and concepts, the use of computer science- derived data analysis methods, predominantly machine learning approaches, were proposed and explored in this study. ^ The goal of this study was to develop a set of models using decision trees/ensemble and neural networks methods to predict occupational outcomes based on literature-derived databases, and compare, using cross-validation and data splitting techniques, the resulting prediction capacity to that of traditional regression models. Two cases were addressed: the categorical case, where the exposure level was measured as an exposure rating following the American Industrial Hygiene Association guidelines and the continuous case, where the result of the exposure is expressed as a concentration value. Previously developed literature-based exposure databases for 1,1,1 trichloroethane, methylene dichloride and, trichloroethylene were used. ^ When compared to regression estimations, results showed better accuracy of decision trees/ensemble techniques for the categorical case while neural networks were better for estimation of continuous exposure values. Overrepresentation of classes and overfitting were the main causes for poor neural network performance and accuracy. Estimations based on literature-based databases using machine learning techniques might provide an advantage when they are applied to other methodologies that combine `expert inputs' with current exposure measurements, like the Bayesian Decision Analysis tool. The use of machine learning techniques to more accurately estimate exposures from literature-based exposure databases might represent the starting point for the independence from the expert judgment.^