Recognition of gasoline in fire debris using machine learning: Part II, application of a neural network.

The recognition of ignitable liquid (IL) residues in fire debris is a resource intensive but key part of an arson investigation. Due to the highly diverse and heavily loaded chemical matrix of fire debris samples, combined with the broad chemical composition of IL, the interpretation of the laboratory analysis results is a very challenging task for the forensic examiner. Fire debris samples are commonly analyzed using gas chromatography coupled to mass spectrometry (GC-MS). This method delivers both the total ion chromatogram (TIC) with the individually separated compounds and the underlying mass spectrum of each of the separated compounds. In this study, a completely new approach for the recognition of gasoline in fire debris samples is presented. First, the GC-MS data, including retention time, signal intensity, and mass spectrum is converted into a bitmap image. Five different data-to-image conversion approaches are tested, and their advantages and limitations are discussed. Subsequently, a convolutional neural network (CNN) is utilized to allocate the generated images to the classes "with gasoline" or "without gasoline". The applied approaches to generate a digital image and the pattern recognition of the CNN perform very well in the classification of unknown test samples. Depending on the data-to-image generation approach used, the rate of correct sample classification in the test dataset is between 95% and 98%. The machine learning approach in this study, as well as the complementary method presented in an accompanying article, are not only useful for the recognition of gasoline in fire debris but are equally applicable to any additional areas in which the interpretation of complex chromatographic and mass spectrometric is required.

Recognition of gasoline in fire debris using machine learning: Part I, application of random forest, gradient boosting, support vector machine, and naïve bayes.

The detection and identification of ignitable liquid (IL) residues in fire debris are two very challenging tasks in a fire investigation. To this day, the recognition of IL in fire debris includes the chemical analysis of the fire debris composition, followed by the examination and interpretation of the analysis result by a trained forensic examiner. Throughout the last decade, chemometrics and artificial intelligence have become increasingly important. In the present study, machine learning algorithms capable of recognizing gasoline residues in fire debris based on GC-MS data have been developed. Four methods, including random forest, gradient boosting, support vector machine, and naïve bayes are applied and used to classify fire debris samples into the two categories "with gasoline" or "without gasoline". A fifth method (logistic regression) did not converge due to well separated classes. A database comprising 360 measurements, including fire debris samples of real cases as well as fire debris samples spiked with known amounts of weathered gasoline (up to 99.6%), was available to train the machine learning algorithms (using 85% of the data) and to subsequently test the performance of the methods when classifying unknown samples (using 15% of the data). In general, the methods perform very well, as three of it succeeded to classify all test samples correctly without any false positive or false negative allocations. One (naïve bayes) was not trained enough to classify other (non-gasoline) IL correctly as "no gasoline". Furthermore, the random forest method reveals which chemical compounds are most relevant for the algorithm to classify the samples. In general, the presented approach is highly promising and could easily be extended or adapted to other types of IL. Similar to the neural network presented in the accompanying paper, such methods have the potential to serve as a fast screening technique for fire debris samples, thus supporting the forensic examiner by providing an additional independent opinion. Nonetheless, the definite identification of IL residues in fire debris always has to be accomplished by a forensic examiner.

Effectiveness of an HIV prevention program for secondary school students in Mongolia.

PURPOSE: To evaluate the effectiveness of a 3-year human immunodeficiency virus (HIV) prevention program for adolescents attending secondary school in Mongolia. METHODS: Comparisons of knowledge, attitudes, self-efficacy and safe sex practices of grade 10 students from schools with a peer education prevention program to grade 10 students from schools without the intervention. Peer education programs were launched in 2000 across Mongolia. In 2004, survey data was collected among 720 randomly selected students from eight schools with the peer education prevention program and compared with those of 647 students from eight schools without this intervention. Data was collected in Ulaanbaatar and three Mongolian provinces and analyzed using multilevel regression methods. RESULTS: Students of schools with the program were statistically significantly more knowledgeable, had less traditional attitudes, and had greater awareness of their self-efficacy in regards to HIV and sexual health. Students from schools with the peer education program were more likely to practice safe sex, though the difference was not statistically significant. However, safe sex practice was found to be statistically significantly safer in a subset of schools that had small teams of peer educators. CONCLUSION: Adolescents in Mongolia are sexually active and at risk for infection with HIV and other STIs. Peer education programs, particularly those that are managed by small teams, appear effective and should be implemented more broadly.

The DRG hospital payment system, surgical readmissions and cost containment.

We analyzed all adult surgical patients requiring readmission to the surgical service of an acute care academic hospital for a four-year period (1/1/85-12/31/88). We stratified surgical readmissions by the number of times the patient was readmitted to surgery (from one to five times). For surgical patients 41.1 per cent of the readmission population was readmitted more than once, only 4.4 per cent were readmitted five or more times. Patients requiring three or more admissions generally had the greatest hospital resource utilization, financial risk under DRG payment, and mortality, compared with other surgical readmissions. This analysis suggests that within the surgical readmission population resource parameters may differ by the number of readmissions per patient. Factors were identified which corresponded to a greater likelihood of surgical readmission, and possibly allow the focus of outpatient services which may reduce hospital inpatient costs in the future.

Hospital readmissions, otolaryngology, and the diagnosis related group hospital payment system.

An analysis of otolaryngologic patients requiring readmission was conducted at our institution during a 4-year period to determine the number of readmissions per patient and the time between discharge and hospital readmission. Readmitted otolaryngologic patients were found to have had greater hospital resource utilization, financial risk under diagnosis-related group payment, and mortality, compared with those patients not readmitted to our facility. For patients readmitted to otolaryngologic services (21.2% of total otolaryngologic patients), 20.4% of the readmissions occurred within 30 days of hospital discharge. Of these, 39.3% required one hospital readmission, 16.3% required two readmissions, and 46.4% of the patients called for three or more hospital admissions. Clinical factors were identified that resulted in a greater incidence of otolaryngologic readmission. Otolaryngologic patients readmitted to other clinical services were also studied. This analysis loads to the conclusion that inequities exist within the diagnosis related group hospital payment system vis-à-vis otolaryngologic readmissions. The results of these data also demonstrate leverage points in which we will be able to focus outpatient services for otolaryngologic patients requiring readmission and potentially decrease inpatient hospital expenditures in the days ahead.

Pediatric patients, DRG hospital payment, and comorbidities.

To determine whether the diagnosis related group (DRG) hospital payment system is causing financial pressure on pediatric hospitals similar to that shown in our previous work in other settings, we analyzed resource consumption for pediatric patients in any of the 251 DRGs not stratified by comorbidities. The new DRG prospective "all payor system" is in effect at our hospital. Analysis of 12,771 pediatric patients by payer (Medicaid and commercial insurance such as Blue Cross) in these DRGs for a 3-year period demonstrated that, as a group, pediatric patients with more comorbidities generated higher total hospital costs, a longer hospital length of stay, a greater percentage of procedures per patient, financial risk under DRG payment, more outliers (expensive patients), and a higher mortality rate than pediatric patients in the same DRGs with fewer comorbidities. This study confirms major inequities in DRG prospective hospital payment for many pediatric patients.

Limits of sensitivity of radiopertechnetate flow studies in the detection of asymmetrical renal perfusion.

A canine model has been devised using electromagnetic flowmeter probes to study the sensitivity of radiopertechnetate renal angiography. Highly significant correlations were obtained when the measured flow rates were compared with normalized values of the peak activity and uptake slopes for each kidney. Analysis of the data suggests the unilateral renal disease associated with less then 20% difference in flow between the kidneys may not be diagnosed reliably from the pertechnetate perfusion test.