Book

# Applied Logistic Regressio

Authors:

## Abstract

From the reviews of the First Edition."An interesting, useful, and well-written book on logistic regression models . . . Hosmer and Lemeshow have used very little mathematics, have presented difficult concepts heuristically and through illustrative examples, and have included references."—Choice"Well written, clearly organized, and comprehensive . . . the authors carefully walk the reader through the estimation of interpretation of coefficients from a wide variety of logistic regression models . . . their careful explication of the quantitative re-expression of coefficients from these various models is excellent."—Contemporary Sociology"An extremely well-written book that will certainly prove an invaluable acquisition to the practicing statistician who finds other literature on analysis of discrete data hard to follow or heavily theoretical."—The StatisticianIn this revised and updated edition of their popular book, David Hosmer and Stanley Lemeshow continue to provide an amazingly accessible introduction to the logistic regression model while incorporating advances of the last decade, including a variety of software packages for the analysis of data sets. Hosmer and Lemeshow extend the discussion from biostatistics and epidemiology to cutting-edge applications in data mining and machine learning, guiding readers step-by-step through the use of modeling techniques for dichotomous data in diverse fields. Ample new topics and expanded discussions of existing material are accompanied by a wealth of real-world examples-with extensive data sets available over the Internet.
... A model with AUC of 0.5 is no better than a coin toss; AUCs in the range 0.7-0.8 are regarded as reflecting a prediction model with acceptable discrimination, 0.8-0.9 as excellent discrimination and above 0.9 as outstanding discrimination. 16 We used likelihood ratio tests to compare models and the final model was chosen as the model with the minimum Akaike Information Criterion (AIC) 17 among competing models. ...
... For the final model, we assessed the goodness of fit of the silicosis risk predictions applied to the Modelling group graphically and quantitatively by the Hosmer-Lemeshow and Stukel goodness-of-fit tests. 16,18,19 We estimated the accuracy characteristics of the final model (sensitivity, specificity, PPV, NPV and per cent correctly classified) for silicosis classifications, according to different thresholds of predicted risk, focusing on the subgroup of deceased claimants in the Modelling group, which was considered to be most comparable to the deceased claimants in the Prediction group, including with regard to likely prevalence of silicosis. ...
Article
The Qhubeka Trust was established in 2016 in a legal settlement on behalf of former gold miners seeking compensation for silicosis contracted on the South African mines. Settlements resulting from lawsuits on behalf of gold miners aim to provide fair compensation. However, occupational exposure and medical records kept by South African mining companies for their employees have been very limited. Some claimants to the Qhubeka Trust died before medical evaluation was possible, thus potentially disadvantaging their dependants from receiving any compensation. With medical evaluation no longer possible, a statistical approach to this problem was developed. The records for claimants with medical evaluation were used to develop a logistic regression prediction model for the likelihood of silicosis, based on the potential predictors: cumulative exposure to respirable dust, age, years since first exposure, years of life lost prematurely, vital status at 31 December 2019, and a history of tuberculosis diagnosis. The prediction model allowed estimation of the likelihood of silicosis for each miner who had died without medical evaluation and is a novel approach in this setting. In addition, we were able to quantitatively evaluate the trade-offs of different silicosis risk classification thresholds in terms of true and false positives and negatives. Significance: • A statistical approach can be used for risk estimation in settings where the outcome of interest is unknown for some members of a class. • The likelihood of silicosis in deceased miners without medical evaluation in the Qhubeka Trust can be accurately estimated, using information from finalised claims. • Strategies for classifying the silicosis status of deceased miners without medical evaluation in the Qhubeka Trust can be assessed in a rigorous, quantitative framework.
... X i is a vector of control variables, including socio-demographic and socio-economic variables. Even though some previous studies suggest to include all clinically and intuitively relevant variables in the model regardless of their statistical significance, such a procedure can introduce numerically unstable estimates in the model and difficulties for other researchers to reproduce the results (Hosmer et al., 2013). Due to these shortcomings, we follow seven steps to purposefully select the model (cf. ...
... Due to these shortcomings, we follow seven steps to purposefully select the model (cf. Hosmer et al., 2013, for an in depths description). ...
Article
Full-text available
Formal digital credit raises hopes to decrease the gender gap in financial inclusion. However, up until now, it remains unknown whether these hopes are justified. Using nationally representative household surveys from Kenya, the present study aims to fill this gap. We find strong indication that formal digital credit, contrasting to expectations, has led to an increase in the gender gap in financial inclusion. We further find indication that the pervasive gender gap in the formal digital credit market is largely attributable to gender differences in socio‐economic variables in combination with a lack of contract term heterogeneity in that market. The paper suggests that policies to strengthen women's position in society and/or to encourage contract term heterogeneity in the formal digital credit market could decrease the gender gap in financial inclusion.
... Overall statistical significance tests for models were carried out to evaluate the quality of the logistic regression results (i.e., model fit). The Omnibus Test of Model Coefficients (32) was used, and the Hosmer and Lemeshow goodness of fit test (33). Moreover, Bootstrapping (34) was used to confirm statistically significant results for each predictor coefficient (B). ...
... In both time intervals receiving a DBI > 1 was associated with a statistically significant higher OR for readmission when compared with non-exposed patients. Logistic stepwise regression was carried out to take account of possible confounders, e.g., CCI (33,53), together with bootstrapping to ensure that the prediction model was robust. Model discrimination testing was also performed. ...
... The area under the receiver operating characteristic curve (AUC) values were calculated to represent the proprioceptive acuity scores. A score of 1 indicated 100% accuracy, and 0.5 indicated that the accuracy was attributed to chance (28,29). Significant differences between body sides and study groups were analyzed by one-way ANOVA. ...
Article
Full-text available
Background: Bilateral proprioception deficits were reported in stroke survivors. However, whether bilateral proprioception deficits exist in the ankle joint after stroke was unclear. Ankle proprioception is a significant predictor of balance dysfunction after stroke, and previous studies to date are lacking appropriate evaluation methods. Objectives: We want to determine whether the active movement extent discrimination apparatus (AMEDA) is a reliable tool for assessing ankle proprioceptive acuity in stroke survivors and the presence of deficits in ankle proprioception on the affected and unaffected sides in patients after stroke. Methods: Bilateral ankle proprioception was assessed in 20 stroke patients and 20 age-matched healthy controls using AMEDA. Test-retest reliability was assessed using the intraclass correlation coefficient (ICC). Results: The ICC in the affected and unaffected sides was 0.713 and 0.74, respectively. Analysis of variance revealed significant deficits in ankle proprioception in subacute stroke survivors vs. healthy controls (F = 2.719, p = 0.045). However, there were no significant differences in proprioception acuity scores between the affected and unaffected sides in patients after stroke (F = 1.14, p = 0.331). Conclusions: Stroke survivors had bilateral deficits in ankle proprioceptive acuity during active movements compared with age-matched healthy controls, underscoring the need to evaluate these deficits on both sides of the body and develop effective sensorimotor rehabilitation methods for this patient population. The AMEDA can reliably determine bilateral ankle proprioceptive acuity in stroke survivors.
... Log ratios and p-values were computed for all binary associations using a Wald chi-squared test, with a 95% confidence interval (40). Following the strategy described in Hosmer et al., variables with p < 0.25 in bivariate analyses were then included in a multivariable model (41). The multivariable model was run using a generalized linear model (R version 3.2.4, ...
Article
Full-text available
Background Injection drug use (IDU) is the leading risk factor for hepatitis C virus (HCV) transmission in the U.S. While the general risk factors for HCV transmission are known, there is limited work on how these factors interact and impact young people who inject drugs (YPWID). Methods Project data were drawn from a study of 539 New York City (NYC) residents ages 18-29 who were recruited via Respondent-Driven Sampling and, reported past-month non-medical use of prescription opioids and/or heroin. Analyses are based on a subsample of 337 (62%) who reported injecting any drug in the past 12 months. All variables were assessed via self-report, except HCV status, which was established via rapid antibody testing. Integrating the observed statistical associations with extant literature on HCV risk, we also developed a qualitative system dynamics (SD) model to use as a supplemental data visualization tool to explore plausible pathways and interactions among key risk and protective factors for HCV. Results Results showed a 31% HCV antibody prevalence with an overall incidence of 10 per 100 person-years. HCV status was independently correlated with having shared cookers with two or more people (AOR = 2.17); injected drugs 4–6 years (AOR = 2.49) and 7 or more years (AOR = 4.95); lifetime homelessness (AOR = 2.52); and having been incarcerated two or more times (AOR = 1.99). These outcomes along with the extant literature on HCV risk were used to develop the qualitative SD model, which describes a causal hypothesis around non-linearities and feedback loop structures underlying the spread of HCV among YPWID. Conclusions Despite ongoing harm reduction efforts, close to a third of YPWID in the community sample have been exposed to HCV, have risks for injection drug use, and face challenges with structural factors that may be preventing adequate intervention. The qualitative SD model explores these issues and contributes to a better understanding of how these various risk factors interact and what policies could potentially be effective in reducing HCV infections.
... We described categorical variables using frequencies, and compared socioeconomic, work-related, home-related, and COVID-19-related characteristics by loneliness status using chi-square tests and Fisher's exact tests (for questions with low cell sizes). We used a purposeful selection approach to build a multivariable logistic regression model to explain increased feelings of loneliness (Hosmer et al., 2013). Since this approach retains important confounders in addition to significant covariates, the use of a purposeful selection approach is advantageous compared to other selection techniques when building an explanatory model (Bursac et al., 2008). ...
Article
The Coronavirus Disease 2019 (COVID-19) pandemic had many negative consequences, one of which was the increase of loneliness. We aimed to explore associations between sociodemographic, work-related, home-related, and COVID-19-related characteristics and increased feelings of loneliness among adults living in the United States (US). We analyzed cross-sectional baseline data from The Quickly Understanding Impacts of COVID-19 Study (The QUICk Study) collected from May to October 2020 using online surveys completed by a sample of adults living in the US. We used chi-square tests, Fisher exact tests, and logistic regression to identify characteristics associated with increased loneliness. The study sample included 577 adults living in the US. Approximately 37% of the sample reported feeling lonelier than usual over the past month. Younger age, sexual minority status, lower education level, depression, living alone, part-time employment status, and student employment status were significantly associated with increased feelings of loneliness. Depression, younger age, and living alone remained significantly associated with increased feelings of loneliness in the multivariable logistic regression analysis. In the US, young adults, adults with depression, and adults who live alone may have been more likely to experience increased feelings of loneliness during the early COVID-19 pandemic.
... This reason could also potentially explain the better performances for T3 as this item directly related to students' spatial behaviours within one of the primary task spaces (the small red circle around the phone which is used to make the MET calls, see Figure 1). In contrast, the models outperformed the baseline model regarding each collaboration assessment item (T4-T6) with an acceptable level of differentiation (Hosmer Jr et al., 2013) (AUC of around 0.74). This finding was expected as students' spatial behaviours has been frequently found to contain meaningful insights about their interaction and collaboration in physical learning spaces (Riquelme et al., 2020;Saquib et al., 2018;Yan et al., 2021). ...
Article
Full-text available
Article
Full-text available
Background In professional sports, injuries resulting in loss of playing time have serious implications for both the athlete and the organization. Efforts to quantify injury probability utilizing machine learning have been met with renewed interest, and the development of effective models has the potential to supplement the decision-making process of team physicians. Purpose/Hypothesis The purpose of this study was to (1) characterize the epidemiology of time-loss lower extremity muscle strains (LEMSs) in the National Basketball Association (NBA) from 1999 to 2019 and (2) determine the validity of a machine-learning model in predicting injury risk. It was hypothesized that time-loss LEMSs would be infrequent in this cohort and that a machine-learning model would outperform conventional methods in the prediction of injury risk. Study Design Case-control study; Level of evidence, 3. Methods Performance data and rates of the 4 major muscle strain injury types (hamstring, quadriceps, calf, and groin) were compiled from the 1999 to 2019 NBA seasons. Injuries included all publicly reported injuries that resulted in lost playing time. Models to predict the occurrence of a LEMS were generated using random forest, extreme gradient boosting (XGBoost), neural network, support vector machines, elastic net penalized logistic regression, and generalized logistic regression. Performance was compared utilizing discrimination, calibration, decision curve analysis, and the Brier score. Results A total of 736 LEMSs resulting in lost playing time occurred among 2103 athletes. Important variables for predicting LEMS included previous number of lower extremity injuries; age; recent history of injuries to the ankle, hamstring, or groin; and recent history of concussion as well as 3-point attempt rate and free throw attempt rate. The XGBoost machine achieved the best performance based on discrimination assessed via internal validation (area under the receiver operating characteristic curve, 0.840), calibration, and decision curve analysis. Conclusion Machine learning algorithms such as XGBoost outperformed logistic regression in the prediction of a LEMS that will result in lost time. Several variables increased the risk of LEMS, including a history of various lower extremity injuries, recent concussion, and total number of previous injuries.
Article
Full-text available
Background: Public engagement in the research of environmental epidemiological problems is becoming an important measure to empower citizens to identify the local environmental and health problems and to explain different environmental exposures affect estimates for males and females. This HORIZON2020 CitieS-Health Kaunas Pilot study examines the relationship between urban built and social environment, health behaviors, and health in men and women. Methods: This cross-sectional study included 1086 18-74-year-old participants residing in 11 districts of Kaunas city, Lithuania. Using GIS, we measured traffic flow, noise, NO2, PM2.5, PM10, and greenness NDVI for the participants' home addresses, determined participants' perceptions of environmental quality, linked this information with personal sociodemographic data, and used multivariate logistic regression to assess the associations with health issues (physician-diagnosed chronic disease and self-rated general health) in men and women. Results: Men and women similar rated the quality of the neighborhood environment, except for air pollution and satisfaction with the public transport in the district. The traffic-related health associations were stronger for women than for men. The prevalence of poor health increased with the increasing age of men and women, yet no significant differences between gender health risks were found in the total sample. Perceived air pollution, irregular visits to green space, and chronic diseases were consistently associated with poor health risks in men and women, yet part-time jobs and low income had a higher impact on women's poor health. Conclusions: Quality of the built neighborhood, air pollution, irregular visits to the green space, and chronic disease had a joint effect on the magnitude of the prevalence of poor health in men and women. Our results suggest that decreasing air pollution and improving the urban built neighborhood supporting citizens' physical activity in green spaces, might reduce health risks for all.
Article
Full-text available
The Screening Scale of Pedophilic Crime Scene Behavior (SSPC) is a seven-item structured rating scale assessing pedophilic sexual arousal. In the current study, we cross-validated the scale’s convergent validity using multiple measures of sexual interest in children (clinical diagnosis of pedophilia, the high fixation/low social competence type of the MTC:CM4, and phallometric assessment of sexual interests toward children) in two independent samples (USA and Canada). In both samples and in relation to all three criteria, the SSPC showed acceptable (phallometry) to excellent (clinical assessment) diagnostic accuracy. Furthermore, the SSPC showed incremental validity in relation to the Screening Scale for Pedophilic Interest and at times outperformed it in convergent validity analyses. The current study also provides psychometric information that can help users choose an appropriate SSPC cutoff score.
Article
Ovarian cancer (OC) is the eighth most common cancer worldwide and is usually diagnosed in advanced stages. The relationship between treatment in high-volume hospitals (HVHs) and survival in OC has been documented by multiple studies, which showed that superior treatment and survival outcomes are associated with surgical expertise and multidisciplinary resources. To our study, 135 first-time patients treated in the years 2019–2020 in the Department of Oncology of Poznań University of Medical Sciences were enrolled. Th analysis showed a significant dependency between being treated in a HVH from the beginning of one’s diagnosis and the scope of the first intervention. Additionally, among patients treated in our centre, a significant portion of patients underwent laparoscopy, and from one year to another the number of laparoscopies performed increased. This may indicate that more patients began to qualify for neoadjuvant treatment. Patients benefit the most from surgery in a centre with more experience in treating ovarian cancer. In the future, we will be able to expand this study by using data from patients treated before 2019 and analysing larger cohorts of patients. This might enable us to update the rates of overall survival (OS), objective response rate (ORR) and progression-free survival (PFS).
Article
Although relatively rare, sexual homicide has stimulated research efforts to better understand and intervene in the prevention of such crime. However, specific high-risk victims such as sex-trade workers have been understudied. To contribute to the limited scientific literature on this topic, the current study examines the characteristics of the sexual victimization of sex-trade workers using a sample of 402 lethal and non-lethal cases. Bivariate and multivariate analyses indicated that the lethal outcome associated with the sexual victimization of sex-trade workers may not be associated with sadism and paraphilic behaviors, but rather by excessive violence and victim's vulnerabilities. Conjunctive analysis further indicated that the combination of excessive violence, victim's intoxication, situational factors, and crime characteristics was associated with the lethal sexual victimization of sex-trade workers. Findings suggest strategies for the development of situational prevention and intervention strategies.
Article
Full-text available
Background As the incidence of gastric cancer (GC) increases sharply in adults aged over 40 years, screening of this high-risk population is important. This study aimed to explore knowledge level of GC related risk factors and symptoms, and to identify influencing factors associated with intention toward GC screening among people aged 40 years old and above in China. Methods A cross-sectional, web-based survey was conducted among people aged 40 years old and above between October 2021 and March 2022 in Southeastern China. The participants' knowledge was assessed by a series of questions about risk factors (24-item scale) and warning symptoms (14-item scale). Results A total of 2547 complete responses were received. The mean age was 47.72 (±7.20) years and near 60% were male. Respondents had a moderate level of knowledge about risk factors and warning symptoms of GC. The total mean knowledge score was 23.9 (±9.8) out of a possible score of 38. Majority (80%) of respondents reported intention to be screened for GC in the next 5 years. The most influential predictors of screening intention were income level (OR = 2.13, 95% CI: 1.36–3.32), perceived benefits (OR = 1.99, 95% CI: 1.33–2.73), perceived severity (OR = 1.68, 95% CI: 1.20–2.34), ever took GC screening (OR = 1.63, 95% CI: 1.28–2.08), perceived poor overall health (OR = 1.59, 95% CI: 1.19–2.11), and perceived barriers (OR = 1.56, 95% CI: 1.17–2.09). Other significant factors were ever diagnosed with chronic gastric diseases, total knowledge score, and cues-to-action. The major reasons for not willing to take screening were “endoscopy is uncomfortable” (29.6%), “worry about screening results” (23.6%), and “have no symptoms” (21.3%). Conclusion High-risk population aged 40 years and above expressed high intention to receive GC screening. Intervention to improve health promotion and reduce the barriers to uptake of GC screening among high-risk populations in China is warranted.
Article
Article
Full-text available
Background Patients with pituitary lesions experience decrements in quality of life (QoL) and treatment aims to arrest or improve QoL decline. Objective To detect associations with QoL in trans-nasal endoscopic skull base surgery patients and train supervised learning classifiers to predict QoL improvement at 12 months. Methods A supervised learning analysis of a prospective multi-institutional dataset (451 patients) was conducted. QoL was measured using the anterior skull base surgery questionnaire (ASBS). Factors associated with QoL at baseline and at 12-month follow-up were identified using multivariate logistic regression. Multiple supervised learning models were trained to predict postoperative QoL improvement with five-fold cross-validation. Results ASBS at 12-month follow-up was significantly higher (132.19,SD = 24.87) than preoperative ASBS (121.87,SD = 25.72,p<0.05). High preoperative scores were significantly associated with institution, diabetes and lesions at the planum sphenoidale / tuberculum sella site. Patients with diabetes were five times less likely to report high preoperative QoL. Low preoperative QoL was significantly associated with female gender, a vision-related presentation, diabetes, secreting adenoma and the cavernous sinus site. Top quartile change in postoperative QoL at 12-month follow-up was negatively associated with baseline hypercholesterolemia, acromegaly and intraoperative CSF leak. Positive associations were detected for lesions at the sphenoid sinus site and deficient preoperative endocrine function. AdaBoost, logistic regression and neural network classifiers yielded the strongest predictive performance. Conclusion It was possible to predict postoperative positive change in QoL at 12-month follow-up using perioperative data. Further development and implementation of these models may facilitate improvements in informed consent, treatment decision-making and patient QoL.
Article
Full-text available
Background: The current study aimed to validate the utility of previously established validity indicators derived from MOXO-d-CPT's continuous performance test. Method: Healthy simulators feigned impairment after searching online for relevant information, an ecologically valid coaching condition (n = 39). They were compared to ADHD patients (n = 36) and healthy controls (n = 38). Results: Simulators performed significantly worse than ADHD patients in all MOXO-d-CPT indices, as well as a scale that integrates their contributions (feigned ADHD scale). Three indices (attention, hyperactivity, and impulsivity) and the latter scale exhibited adequate discriminative capacity. Higher education was associated with an exaggerated impairment among simulators, easing their detection. Conclusion: Similarity between the current study and a previous study which examned the utlity of the MOXO-d-CPT validity indicators, increases our confidence in the efficacy of the latters embedded validity indicators. Though the findings provide initial validation of these validity indicators, generalizing beyond highly functioning participants necessitates further research.
Article
Full-text available
The selection of influential predictor factors with maximum model accuracy is the main goal of the regression domain. The present study is conducted to integrate an innovative method, that is, “a hybrid of relaxed lasso and ridge regression,” with a logistic regression model in the context of dichotomous factors. The efficacy of the proposed approach is illustrated using both simulated and real-life data. The results suggested that HRLR-logistic selected the best subset compared to standard logistic, Lasso, and Ridge regression. Based on the Akaike information criterion (3065.85) and the Bayesian information criterion (3151.46), the proposed approach is proved to have the highest efficiency for cesarean section data. In addition, the study identified the elements that contribute to the cesarean section in Pakistan. It is evidenced that woman’s literacy level (β = 0.5828), place of delivery (β = 0.8990), availability of nurse as an assistant (β = 0.7370), and care during the first two days of delivery (β = 0.7837) are remarkable factors associated with cesarean section.
Article
During the COVID-19 pandemic possible substance use disorders (SUD) were exacerbated from increased stress and isolation. Experiences of symptomology differ widely by occupations. Objectives: The objectives were to determine if there is a temporal relationship between COVID-19 vulnerability and possible SUDs among first responders, and to examine the association with neighborhood vulnerability. We conducted an analysis with two distinct cohorts dependent on time of entry: 1) First responders that began counseling prior to COVID-19 and 2) First responders that began counseling after the start of COVID-19. Data were collected at intake from first responders seeking mental health services between 2017 and 2021 at an organization in Dallas/Fort Worth, Texas. The study sample included 195 mostly male (75%) first responders (51% law enforcement officers; 49% emergency medical technicians/firefighters). Bivariate models tested unadjusted relationships between covariates and possible SUD. Adjusted models consisted of a two-level multivariable logistic regression models. Nearly 40% (n = 77) screened positive for a possible SUD. Those beginning counseling after COVID-19 did not have higher odds of SUDs. For every unit increase in neighborhood Severe COVID-19 Health Risk Index at a first responder’s residential location there was an increase in the odds of a possible SUD (AOR = 3.14, 95% CI: 1.47, 6.75). Our study highlights the degree to which personal and residential vulnerability to COVID-19 impacted first responders. The increased occupational stress of this population, and an established pattern of maladaptive coping, elucidates the need for preventative and clinical approaches to strengthen the resilience of this population.
Article
Full-text available
The aims of this study were to identify factors that a) predict whether people experience housing related discharge delay (HRDD) from a mental health inpatient unit; and b) predict the length of HRDD for people affected. By identifying the groups most affected by HRDD, clinicians and policy makers can prioritise and address barriers to timely discharge at both an individual and systemic level. A case control study using a detailed medical record review was conducted in one Australian mental health service. Demographic, clinical, contextual and systemic variables were collected for patients with HRDD in one calendar year ( n = 55) and a random comparison sample ( n = 55). Logistical and multiple regression analyses were conducted to identify variables that predict HRDD and length of HRDD. A model that correctly predicted 92% of HRDD and 78% of non-HRDD cases using five variables was developed. These variables were: diagnosis of schizophrenia or other psychotic disorder, physical comorbidity, having a history of violence or aggressive behaviour, being employed and being involved as a defendant in the justice system. The first three variables increased the likelihood of HRDD, while the second two reduced the likelihood of HRDD. For people who experienced HRDD, the only variable that predicted length of delay was staff reported difficulty finding appropriate support services. This model can be used to rapidly identify patients who might be at risk of HRDD and commence coordinated actions to secure appropriate housing and supports to facilitate timely discharge, thereby addressing a current practice gap. These findings highlight the intersection between health, housing and disability services in the lives of people with serious mental illness, and the need for a whole of government approach to investment and integration to address systemic barriers to suitable housing and supports.
Article
Formularies are intended to simplify clinical decision‐making by collecting evidence‐based information on drugs and their dosages. This study assessed the characteristics of sources used to support drug dosages and reference intervals for mammals in a specific exotic animal formulary, and how the sources had changed over five editions. Each reference supporting drug dosages and reference intervals in the sections for ferrets, rabbits, rodents, hedgehogs and miniature pigs in all five editions of the formulary was evaluated and classified by two independent investigators in terms of the type of source cited. Univariable and multi‐variable logistic regression models were built to evaluate changes between editions and sections. In total, 1338 references supporting drug dosages and 180 references supporting reference intervals were included from all editions of the formulary. Primary sources were cited by 525 (39.2%) and 39 (21.7%) of the drug and reference interval references, respectively. For drug dosages, the current edition of the formulary (2018) cited a higher proportion of primary rather than secondary sources compared with the first edition (odds ratios 3.4, 95% confidence interval 2.1 to 5.6), while for reference intervals there were no significant changes between editions. In the current edition of the formulary, the 168 secondary sources cited for drug dosages included 78 (46.4%) textbooks, 63 (37.5%) reviews, 14 (8.3%) personal communications and 7 (4.2%) other formularies. A large proportion of references supporting drug dosages and reference intervals in the evaluated sections cited secondary sources. Although modest improvements have been observed over time, practitioners should be aware that the evidence supporting several drugs and dosages was limited, and assess the information within the formulary critically.
Chapter
This paper presents the results of a comparative study of machine learning techniques when predicting deep vein thrombosis. We used the Ri-Schedule dataset with Electronic Health Records of suspected thrombotic patients for training and validation. A total of 1653 samples and 59 predictors were included in this study. We have compared 20 standard machine learning algorithms and identified the best-performing ones: Random Forest, XGBoost, GradientBoosting and HistGradientBoosting classifiers. After hyper-parameter optimization, the best overall accuracy of 0.91 was shown by GradientBoosting classifier using only 15 of the original variables. We have also tuned the algorithms for maximum sensitivity. The best specificity was offered by Random Forests. At maximum sensitivity of 1.0 and specificity of 0.41, the Random Forest model was able to identify 23% additional negative cases over the screening practice in use today. These results suggest that machine learning could offer practical value in real-life implementations if combined with traditional methods for ruling out deep vein thrombosis.
Article
Full-text available
Background: Syria’s civil conflict, which began in 2011, led millions of Syrians to migrate to countries all over the world, including Turkey. Considering the fact that war-caused migrations may affect the citizens of the host countries and immigrants from diverse perspectives, It is important to make scientific research on the outcomes of migration after the Syrian civil war. Objective: This paper investigates the relationship between chronic diseases, such as, cardiac disease, diabetes, and hypertension, and covariates, including socioeconomic status, war-related conditions, risky health behaviours, health services use, and health literacy, using survey data on 7 202 Syrian refugees from 4 068 households living out of camp settlements in Turkey. Methods: Logistic regressions were employed to examine the relationship between the chronic diseases and covariates, which include socioeconomic status, war-related conditions, risky health behaviours, health services use, and health literacy. Findings: The results reveal that pre-migration income, health behaviours, such as, tobacco consumption, body mass index, daily activity, health-care use, and health literacy are the most important factors for one or more chronic diseases. Conclusion: Considering the effects of risk factors on chronic diseases among Syrian refugees, it is critical to take preventive steps for negative outcomes.
Article
The COVID-19 pandemic spread all over the world, starting in China in late 2019, and significantly affected life in all aspects. As seen in SARS, MERS, COVID-19 outbreaks, coronaviruses pose a great threat to world health. The COVID-19 epidemic, which caused pandemics all over the world, continues to seriously threaten people's lives. Due to the rapid spread of COVID-19, many countries' healthcare sectors were caught off guard. This situation put a burden on doctors and healthcare professionals that they could not handle. All of the studies on COVID-19 in the literature have been done to help experts to recognize COVID-19 more accurately, to use more accurate diagnosis and appropriate treatment methods. The alleviation of this workload will be possible by developing computer aided early and accurate diagnosis systems with machine learning. Diagnosis and evaluation of pneumonia on computed tomography images provide significant benefits in investigating possible complications and in case follow-up. Pneumonia and lesions occurring in the lungs should be carefully examined as it helps in the diagnostic process during the pandemic period. For this reason, the first diagnosis and medications are very important to prevent the disease from progressing. In this study, a dataset consisting of Pneumonia and Normal images was used by proposing a new image preprocessing process. These preprocessed images were reduced to 15x15 unit size and their features were extracted according to their RGB values. Experimental studies were carried out by performing both normal values and feature reduction among these features. RGB values of the images were used in train and test processes for MLAs. In experimental studies, 5 different Machine Learning Algorithms (MLAs) (Multi Class Support Vector Machine (MC-SVM), k Nearest Neighbor (k-NN), Decision Tree (DT), Multinominal Logistic Regression (MLR), Naive Bayes (NB)). The following accuracy rates were obtained in train operations for MLAs, respectively; 1, 1, 1, 0.746377, 0.963768. Accuracy results in test operations were obtained as follows; 0.87755, 0.857143, 0.857143, 0.877551, 0.938776.
Article
Full-text available
Article
This study examines how institutional varieties at the subnational (state) level influence firm‐level innovation in an emerging economy—India. Knowledge of how institutional varieties influence firm‐level innovation is derived principally from country‐level studies involving multiple developed countries. Research on emerging economies is sparse and tends to follow country‐level approaches involving multiple countries. Research involving a single emerging economy where there are substantial institutional varieties between regions is thin. The institutional varieties of some emerging countries are so striking that they can be viewed as several countries within a country, for example, India, China. This study contributes to the innovation literature on the role of institutional varieties on firm‐level innovation by focusing on a different level of analysis—a single, emerging economy with substantial institutional varieties across the different states of India. Innovation in emerging economies is a topic of increasing academic interest. A multilevel study involving regional‐ and firm‐level factors is employed. Firm‐level data are from the World Bank Enterprise Survey and regional‐level data are from statistical agencies in India. The results confirm that institutional varieties have major impacts on firm‐level innovation. The research, policy, and managerial implications are discussed.
Article
Full-text available
Background and Aims Little is known about individual differences in Hallucinogen Persisting Perceptual Disorder (HPPD). This study investigated visual processing style and personality across two HPPD types (HPPD I and HPPD II) and a Non-HPPD group. Methods An online survey was delivered to participants sourced from online HPPD and psychedelic user groups and forums ( N = 117). Using one-way ANOVA, respondents were compared across four measures of individual difference. Using logistic regression, a range of visual symptoms and experiences were investigated as potential predictors of group categorisation. Results The HPPD I group had higher absorption and visual apophenia scores than the other groups and was predicted by higher drug use. The HPPD II group showed significantly higher trait anxiety than both other groups. Across the HPPD groups, HPPD II categorisation was also predicted by increased negative precipitating experiences, lack of prior knowledge and pre-existing anxiety diagnoses. Conclusions Anxiety, negative precipitating experiences and lack of prior knowledge are associated with negative experiences of persistent visual symptoms following hallucinogen use, whilst higher absorption and visual apophenia are associated with positive or neutral experiences. Together these findings indicate that differences in personality may play a role in determining an individual's experience of HPPD, highlighting the role of individual difference research in expanding knowledge around HPPD.
Article
Full-text available
Objectives: Research spanning the fields of clinical, social and health psychology suggests that death anxiety is an important construct. However, no comprehensive, psychometrically adequate measure of the construct exists. The current studies outline the development of a new measure of death anxiety, the Death Anxiety Beliefs and Behaviours Scale (DABBS), which is the first measure to specifically assess unhelpful beliefs and behaviours that may underlie fears of death. Methods: In Study 1, items were piloted in a large community sample (N = 505). In Studies 2A and 2B, exploratory and confirmatory factor analyses were performed using a treatment-seeking (N = 200) and non-treatment-seeking sample (N = 200). These analyses resulted in the final 18-item scale. Results: The DABBS demonstrated good construct validity, criterion validity, internal consistency and test-retest reliability. In Study 3, the DABBS effectively distinguished participants with clinically significant death anxiety and distress from those without, demonstrating excellent discriminant validity. Conclusions: The present data indicate that the DABBS is a valid and reliable measure of affect, beliefs and behaviours relating to death anxiety, in a community sample of adults and among those seeking mental health treatment. Given the increasing recognition of the importance of death anxiety, the DABBS offers a useful research and clinical tool.
Article
Full-text available
Background Major depressive disorder (MDD) is a common mental illness, characterized by persistent depression, sadness, despair, etc., troubling people’s daily life and work seriously. Methods In this work, we present a novel automatic MDD detection framework based on EEG signals. First of all, we derive highly MDD-correlated features, calculating the ratio of extracted features from EEG signals at frequency bands between $$\beta$$ β and $$\alpha$$ α . Then, a two-stage feature selection method named PAR is presented with the sequential combination of Pearson correlation coefficient (PCC) and recursive feature elimination (RFE), where the advantages lie in minimizing the feature searching space. Finally, we employ widely used machine learning methods of support vector machine (SVM), logistic regression (LR), and linear regression (LNR) for MDD detection with the merit of feature interpretability. Results Experiment results show that our proposed MDD detection framework achieves competitive results. The accuracy and $$F_{1}$$ F 1 score are up to 0.9895 and 0.9846, respectively. Meanwhile, the regression determination coefficient $$R^2$$ R 2 for MDD severity assessment is up to 0.9479. Compared with existing MDD detection methods with the best accuracy of 0.9840 and $$F_1$$ F 1 score of 0.97, our proposed framework achieves the state-of-the-art MDD detection performance. Conclusions Development of this MDD detection framework can be potentially deployed into a medical system to aid physicians to screen out MDD patients.
Article
Full-text available
Determining the presence of tumor in biopsies and the decision‐making during resections is often dependent on intraoperative rapid frozen‐section histopathology. Recently, stimulated Raman scattering microscopy has been intro‐ duced to rapidly generate digital hematoxylin‐and‐eosin‐stained‐like images (stimulated Raman histology) for intra‐ operative analysis. To enable intraoperative prediction of tumor presence, we aimed to develop a new deep residual convolutional neural network in an automated pipeline and tested its validity. In a monocentric prospective clinical study with 94 patients undergoing biopsy, brain or spinal tumor resection, Stimulated Raman histology images of intraoperative tissue samples were obtained using a fiber‐laser‐based stimulated Raman scattering microscope. A residual network was established and trained in ResNetV50 to predict three classes for each image: (1) tumor, (2) non‐ tumor, and (3) low‐quality. The residual network was validated on images obtained in three small random areas within the tissue samples and were blindly independently reviewed by a neuropathologist as ground truth. 402 images derived from 132 tissue samples were analyzed representing the entire spectrum of neurooncological surgery. The automated workflow took in a mean of 240 s per case, and the residual network correctly classified tumor (305/326), non‐tumorous tissue (49/67), and low‐quality (6/9) images with an inter‐rater agreement of 89.6% (κ = 0.671). An excellent internal consistency was found among the random areas with 90.2% (Cα = 0.942) accuracy. In conclusion, the novel stimulated Raman histology‐based residual network can reliably detect the microscopic presence of tumor and differentiate from non‐tumorous brain tissue in resection and biopsy samples within 4 min and may pave a prom‐ ising way for an alternative rapid intraoperative histopathological decision‐making tool.
Article
Full-text available
Introduction Chlorhexidine cord care is an effective intervention to reduce neonatal infection and death in resource constrained settings. The Federal Ministry of Health of Ethiopia adopted chlorhexidine cord care in 2015, with national scale-up in 2017. However, there is lack of evidence on the provision of this important intervention in Ethiopia. In this paper, we report on the coverage and determinants of chlorhexidine cord care for newborns in Ethiopia. Methods A standardized Nutrition International Monitoring System (NIMS) survey was conducted from January 01 to Feb 13, 2020 in four regions of Ethiopia (Tigray, Amhara, Oromia, and Southern Nations, Nationalities and Peoples Region [SNNPR]) on sample of 1020 women 0–11 months postpartum selected through a multistage cluster sampling approach. Data were collected using interviewer-administered questionnaires in the local languages through home-to-home visit. Accounting for the sampling design of the study, we analyzed the data using complex data analysis approach. Complex sample multivariable logistic regression was used to identify the determinants of chlorhexidine cord care practice. Results Overall, chlorhexidine was reportedly applied to the umbilical cord at some point postpartum among 46.1% (95% confidence interval [CI]: 41.1%– 51.2%) of all newborns. Chlorhexidine cord care started within 24 hours after birth for 34.4% (95% CI: 29.5%– 39.6%) of newborns, though this varied widely across regions: from Oromia (24.4%) to Tigray (60.0%). Among the newborns who received chlorhexidine cord care, 48.3% received it for the recommended seven days or more. Further, neonates whose birth was assisted by skilled birth attendants had more than ten times higher odds of receiving chlorhexidine cord care, relative to those born without a skilled attendant (adjusted odds ratio [AOR]: 10.36, 95% CI: 3.73–28.75). Besides, neonates born to mothers with knowledge of the benefit of chlorhexidine cord care had significantly higher odds of receiving chlorhexidine cord care relative to newborns born to mothers who did not have knowledge of the benefit of chlorhexidine cord care (AOR: 39.03, 95% CI: 21.45–71.04). Conclusion A low proportion of newborns receive chlorhexidine cord care in Ethiopia. The practice of chlorhexidine cord care varies widely across regions and is limited mostly to births attended by skilled birth attendants. Efforts must continue to ensure women can reach skilled care at delivery, and to ensure adequate care for newborns who do not yet access skilled delivery.
Article
Full-text available
This study analyzed spectral variations of the particulate matter (PM hereafter)-exposed pine trees using a spectrometer and a hyperspectral imager to derive the most effective spectral indices to detect the pine needle exposure to PM emission. We found that the spectral variation in the near-infrared (NIR hereafter) bands systemically coincided with the variations in PM concentration, showing larger variations for the diesel group whereas larger dust particles showed spectral variations in both visible and NIR bands. It is because the PM adsorption on needles is the main source of NIR band variation, and the combination of visible and NIR spectra can detect PM absorption. Fourteen bands were selected to classify PM-exposed pine trees with an accuracy of 82% and a kappa coefficient of 0.61. Given that this index employed both visible and NIR bands, it would be able to detect PM adsorption. The findings can be transferred to real-world applications for monitoring air pollution in an urban area.
Article
Background Questionnaires have been used in the past 2 decades to predict the diagnosis of vertigo and assist clinical decision-making. A questionnaire-based machine learning model is expected to improve the efficiency of diagnosis of vestibular disorders. Objective This study aims to develop and validate a questionnaire-based machine learning model that predicts the diagnosis of vertigo. Methods In this multicenter prospective study, patients presenting with vertigo entered a consecutive cohort at their first visit to the ENT and vertigo clinics of 7 tertiary referral centers from August 2019 to March 2021, with a follow-up period of 2 months. All participants completed a diagnostic questionnaire after eligibility screening. Patients who received only 1 final diagnosis by their treating specialists for their primary complaint were included in model development and validation. The data of patients enrolled before February 1, 2021 were used for modeling and cross-validation, while patients enrolled afterward entered external validation. Results A total of 1693 patients were enrolled, with a response rate of 96.2% (1693/1760). The median age was 51 (IQR 38-61) years, with 991 (58.5%) females; 1041 (61.5%) patients received the final diagnosis during the study period. Among them, 928 (54.8%) patients were included in model development and validation, and 113 (6.7%) patients who enrolled later were used as a test set for external validation. They were classified into 5 diagnostic categories. We compared 9 candidate machine learning methods, and the recalibrated model of light gradient boosting machine achieved the best performance, with an area under the curve of 0.937 (95% CI 0.917-0.962) in cross-validation and 0.954 (95% CI 0.944-0.967) in external validation. Conclusions The questionnaire-based light gradient boosting machine was able to predict common vestibular disorders and assist decision-making in ENT and vertigo clinics. Further studies with a larger sample size and the participation of neurologists will help assess the generalization and robustness of this machine learning method.
Article
Full-text available
Objective: This study examined whether body shape and composition obtained by three-dimensional optical (3DO) scanning improved the prediction of metabolic syndrome (MetS) prevalence compared with BMI and demographics. Methods: A diverse ambulatory adult population underwent whole-body 3DO scanning, blood tests, manual anthropometrics, and blood pressure assessment in the Shape Up! Adults study. MetS prevalence was evaluated based on 2005 National Cholesterol Education Program criteria, and prediction of MetS involved logistic regression to assess (1) BMI, (2) demographics-adjusted BMI, (3) 85 3DO anthropometry and body composition measures, and (4) BMI + 3DO + demographics models. Receiver operating characteristic area under the curve (AUC) values were generated for each predictive model. Results: A total of 501 participants (280 female) were recruited, with 87 meeting the criteria for MetS. Compared with the BMI model (AUC = 0.819), inclusion of age, sex, and race increased the AUC to 0.861, and inclusion of 3DO measures further increased the AUC to 0.917. The overall integrated discrimination improvement between the 3DO + demographics and the BMI model was 0.290 (p < 0.0001) with a net reclassification improvement of 0.214 (p < 0.0001). Conclusions: Body shape measures from an accessible 3DO scan, adjusted for demographics, predicted MetS better than demographics and/or BMI alone. Risk classification in this population increased by 29% when using 3DO scanning.
Article
Full-text available
Suicide is a critical public health problem. Over the past decade, suicide rates have increased among Black and Latinx adults in the U.S. Though depression is the most prevalent psychiatric contributor to suicide risk, Black and Latinx Americans uniquely experience distress and stress (e.g., structural adversity) that can independently operate to worsen suicide risk. This makes it important to investigate non-clinical, subjective assessment of mental health as a predictor of suicide ideation. We also investigate whether social support can buffer the deleterious impact of poor mental health on suicide ideation. We analyzed data from 1,503 Black and Latinx participants of the Washington Heights Community Survey, a 2015 survey of residents of a NYC neighborhood. Multivariable logistic regression was conducted to examine the effect of subjectively experienced problems with anxiety and depression on suicide ideation independent of depression diagnosis, and the role of social support as a moderator. Estimated prevalence of past two-week suicide ideation was 5.8%. Regression estimates showed significantly increased odds of suicide ideation among participants reporting moderate (OR = 8.54,95% CI = 2.44–29.93) and severe (OR = 16.84,95% CI = 2.88–98.46) versus no problems with anxiety and depression, after adjustment for depression diagnosis. Informational support, i.e., having someone to provide good advice in a crisis, reduced the negative impact of moderate levels of anxiety and depression problems on suicide ideation. Findings suggest that among Black and Latinx Americans, subjective feelings of anxiety and depression account for a significant portion of the suicide ideation risk related to poor mental health. Further, social support, particularly informational support, may provide protection against suicide ideation.
Chapter
Machine learning has an application in healthcare, one such application is disease prediction. Because of industrialization and construction, the pollution in the cities has been increased, which gave birth to the critical diseases. We could easily identify that the living habits of a human being is changed because of working condition and work pressure. Identification of disease at an earlier stage is a very crucial job. The early prediction of a disease by a doctor is a challenging task because it requires a huge amount of data to make accurate prediction and to take preventive action. The data are collected over time in a hospital utilized to predict disease, and further, we can do mortality analysis to utilize hospital resources. If we could predict a pandemic, like COVID‐19 and number of patients who are going to be affected accurately, then accordingly, we can plan resources and suggest precautions to prevent it from spreading. Prediction models are based on analyzing data using k‐means clustering, K‐Nearest Neighbor (KNN), and Convolution Neural Network (CNN). Disease prediction consists data of living style and routine check‐up information of a person. The CNN prediction is recorded to be near about 84.5%. The prediction algorithm, which utilizes the benefits of CNN and big data analytics, reported a 94.8% accuracy.
Article
The practice of excess charge where healthcare providers bill Medicare more than the allowed limit, is pervasive in the United States. Previous research has argued that it is possibly used to set private insurance prices, at times to account for inadequate Medicare reimbursements, ultimately leading to high price variations and causing inequities in healthcare service delivery. The objective of this study is to predict a provider’s excess charge quartile and identify the features predictive of their membership in that billing group. We employ distinct multi-class prediction models for three common medical procedures with the highest degree of price variation. The models incorporate four different dimensions of healthcare service- healthcare provider, medical procedure, medical practice, and socioeconomic and demographics of a provider’s patient base. Our analysis of the top predictive features demonstrates that billing patterns are characteristic of a provider’s overall practice rather than specific medical procedures. Also, our results bear important public health implications in the US as our study reveals socioeconomic and demographic factors to be significant predictors of a provider’s billing patterns. The features used in our prediction model can be adapted to other country-specific settings as well. Our work thus provides policymakers with a data-driven foundation for using a holistic framework that has not yet been utilized in healthcare pricing decisions.
Article
Full-text available