Machine Learning in Epidemiology and Health Outcomes Research.

Machine learning approaches to modeling of epidemiologic data are becoming increasingly more prevalent in the literature. These methods have the potential to improve our understanding of health and opportunities for intervention, far beyond our past capabilities. This article provides a walkthrough for creating supervised machine learning models with current examples from the literature. From identifying an appropriate sample and selecting features through training, testing, and assessing performance, the end-to-end approach to machine learning can be a daunting task. We take the reader through each step in the process and discuss novel concepts in the area of machine learning, including identifying treatment effects and explaining the output from machine learning models.

The Screen for Child Anxiety-Related Emotional Disorders Is Sensitive but Not Specific in Identifying Anxiety in Children with High-Functioning Autism Spectrum Disorder: A Pilot Comparison to the Achenbach System of Empirically Based Assessment Scales.

Validated brief screening instruments are needed to improve the detection of anxiety disorders in autism spectrum disorder (ASD). The Screen for Child Anxiety-Related Emotional Disorders (SCARED), a 41-item parent- and self-reported scale measuring anxiety, was compared to the Achenbach System of Empirically Based Assessment (ASEBA) scales. One hundred participants with a clinical diagnosis of high-functioning ASD, aged 8-18 years, and their parents completed the above scales. We hypothesized that the SCARED would be useful in screening for anxiety and its results for total scores of anxiety would converge with ASEBA syndrome scales for anxiety and internalizing disorders. Significant correlations were shown between the SCARED and the Child Behavior Checklist (CBCL) and Youth Self-Report (YSR) across a broad spectrum of scales. The CBCL syndrome scale for anxious/depressed showed the highest correlation and predicted anxiety scores on the SCARED. While many of the YSR scales significantly correlated with child ratings of anxiety, none of the scales predicted the SCARED child scores. Differences in self and parent reports suggest that parents interpret externalizing behaviors as signs of anxiety in ASD, whereas youth may describe internalized symptoms as anxiety. Females were more likely to self-report anxiety than males. Results support the use of the SCARED as a screening tool for anxiety in high-functioning ASD, but it should be supplemented with other tools to increase the specificity of its results.

Using cluster analysis of cytokines to identify patterns of inflammation in hospitalized patients with community-acquired pneumonia: a pilot study.

INTRODUCTION: Patients with severe community-acquired pneumonia (CAP) are believed to have an exaggerated inflammatory response to bacterial infection. Therapies aiming to modulate the inflammatory response have been largely unsuccessful, perhaps reflecting that CAP is a heterogeneous disorder that cannot be modulated by a single anti-inflammatory approach. We hypothesize that the host inflammatory response to pneumonia may be characterized by distinct cytokine patterns, which can be harnessed for personalized therapies. METHODS: Here, we use hierarchical cluster analysis of cytokines to examine if patterns of inflammatory response in 13 hospitalized patients with CAP can be defined. This was a secondary data analysis of the Community-Acquired Pneumonia Inflammatory Study Group (CAPISG) database. The following cytokines were measured in plasma and sputum on the day of admission: interleukin (IL)-1ß, IL-1 receptor antagonist (IL-1ra), IL-6, CXCL8 (IL-8), IL-10, IL-12p40, IL-17, interferon (IFN)γ, tumor necrosis factor (TNF)α, and CXCL10 (IP-10). Hierarchical agglomerative clustering algorithms were used to evaluate clusters of patients within plasma and sputum cytokine determinations. RESULTS: A total of thirteen patients were included in this pilot study. Cluster analysis identified distinct inflammatory response patterns of cytokines in the plasma, sputum, and the ratio of plasma to sputum. CONCLUSIONS: Inflammatory response patterns in plasma and sputum can be identified in hospitalized patients with CAP. Characterization of the local and systemic inflammatory response may help to better discriminate patients for enrollment into clinical trials of immunomodulatory therapies.

Process control charts in infection prevention: Make it simple to make it happen.

BACKGROUND: Quality improvement is central to Infection Prevention and Control (IPC) programs. Challenges may occur when applying quality improvement methodologies like process control charts, often due to the limited exposure of typical IPs. Because of this, our team created an open-source database with a process control chart generator for IPC programs. The objectives of this report are to outline the development of the application and demonstrate application using simulated data. METHODS: We used Research Electronic Data Capture (REDCap Consortium, Vanderbilt University, Nashville, TN), R (R Foundation for Statistical Computing, Vienna, Austria), and R Studio Shiny (R Foundation for Statistical Computing) to create an open source data collection system with automated process control chart generation. We used simulated data to test and visualize both in-control and out-of-control processes for commonly used metrics in IPC programs. RESULTS: The R code for implementing the control charts and Shiny application can be found on our Web site (https://github.com/ul-research-support/spcapp). Screen captures of the workflow and simulated data indicating both common cause and special cause variation are provided. CONCLUSIONS: Process control charts can be easily developed based on individual facility needs using freely available software. Through providing our work free to all interested parties, we hope that others will be able to harness the power and ease of use of the application for improving the quality of care and patient safety in their facilities.

Disinfectant sprays versus wipes: Applications in behavioral health.

We evaluated the utility and usability of a spray and a wipe-based cleaner/disinfectant in a behavioral health setting where health care providers institute cleaning and disinfection procedures. The spray-based version resulted in significantly lower adenosine triphosphate readings after use compared with the wipe version. Staff surveys indicated that users preferred the spray version.

Opioid overdose leading to intensive care unit admission: Epidemiology and outcomes.

PURPOSE: There is a scarcity of studies assessing the patient population admitted to the intensive care unit (ICU) with opioid overdose. We sought to characterize the epidemiologic features and outcomes of this patient population. MATERIALS AND METHODS: This is a retrospective cohort study of adult patients admitted to the ICU at University of Louisville Hospital for opioid overdose. We reviewed each patient's hospital record for demographic data, comorbidities, opioid used, coingestions, and outcomes. RESULTS: We included 178 adult patients, of which 107 (60%) were females. The median age was 41 years (interquartile range [IQR], 23). Oxycodone and hydrocodone were the 2 most commonly abused opioids. Benzodiazepines were the most common drug coingested, followed by amphetamines. Tobacco smoking, chronic pain, and alcoholism were the most frequent comorbidities identified. Mental disorders were also common. Most patients required invasive mechanical ventilation (84.8%). Median ICU length of stay was 3 days. Eighteen patients (10.1%) died in the hospital, whereas 6 patients (3.4%) were discharged to a nursing home. Patients who had any coingestion were significantly more likely to undergo invasive mechanical ventilation (91% vs 77%; P=.014) and had longer ICU length of stay (3 [IQR, 2] vs 2 [IQR, 1.8] days; P=.024). CONCLUSION: Opioid overdose is a common cause of ICU admission and affects a relatively young population. Most have respiratory failure requiring mechanical ventilation. It is associated with a relatively high inhospital mortality. Coingestions appear to have an impact on outcomes.

Understanding why low-risk patients accept vaccines: a socio-behavioral approach.

BACKGROUND: Vaccines are one of the most important public health interventions. Understanding factors associated with vaccine acceptance is critical. The objectives of this study were to evaluate the impact of the three constructs of the Theory of Planned Behavior (TPB) on the intention to be vaccinated among healthy individuals being seen for pre-travel care, and to evaluate if behavioral intention was associated with vaccine acceptance. METHODS: We surveyed individuals seeking vaccination at the University of Louisville Vaccine and International Health and Travel Clinic. Linear and two stage least squares regression models were used to define the associations between constructs of the TPB and the intention to be vaccinated, as well as the association between the intention to be vaccinated and vaccine acceptance. RESULTS: A total of 183 individuals were included in the analysis. None of the constructs of the TPB were associated with intention to be vaccinated. Behavioral intention was not associated with vaccination acceptance. CONCLUSIONS: This study suggests that the TPB does not predict the intention to get vaccinated among individuals attending our Vaccine and International Health and Travel Clinic. It will be critical to define better predictors of vaccine uptake in healthy, low-risk individuals to increase vaccine acceptance.

Efficacy of a novel skin antiseptic against carbapenem-resistant Enterobacteriaceae.

Infections caused by carbapenem-resistant Enterobacteriaceae (CRE) are increasing on a global scale. Because of the need for CRE transmission prevention and control, we sought to evaluate the efficacy of a silver-based skin antiseptic against these organisms. Using a human skin analog, a third party laboratory conducted efficacy testing. The results suggest that this product provides antimicrobial activity against CRE on human skin. Because of the unique properties, this antiseptic may be useful for daily bathing of hospitalized patients to assist in the control of CRE.

Visual grids for managing data completeness in clinical research datasets.

Missing data arise in clinical research datasets for reasons ranging from incomplete electronic health records to incorrect trial data collection. This has an adverse effect on analysis performed with the data, but it can also affect the management of a clinical trial itself. We propose two graphical visualization schemes to aid in managing the completeness of a clinical research dataset: the binary completeness grid (BCG) for single patient observation, and the gradient completeness grid (GCG) for an entire dataset. We use these tools to manage three clinical trials. Two are ongoing observational trials, while the other is a cohort study that is complete. The completeness grids revealed unexpected patterns in our data and enabled us to identify records that should have been purged and identify missing follow-up data from sets of observations thought to be complete. Binary and gradient completeness grids provide a rapid, convenient way to visualize missing data in clinical datasets.

Real-Time Enrollment Dashboard For Multisite Clinical Trials.

OBJECTIVE: Achieving patient recruitment goals are critical for the successful completion of a clinical trial. We designed and developed a web-based dashboard for assisting in the management of clinical trial screening and enrollment. MATERIALS AND METHODS: We use the dashboard to assist in the management of two observational studies of community-acquired pneumonia. Clinical research associates and managers using the dashboard were surveyed to determine its effectiveness as compared with traditional direct communication. RESULTS: The dashboard has been in use since it was first introduced in May of 2014. Of the 23 staff responding to the survey, 77% felt that it was easier or much easier to use the dashboard for communication than to use direct communication. CONCLUSION: We have designed and implemented a visualization dashboard for managing multi-site clinical trial enrollment in two community acquired pneumonia studies. Information dashboards are a useful tool for clinical trial management. They can be used as a standalone trial information tool or included into a larger management system.

The upper respiratory tract microbiome of hospitalised patients with community-acquired pneumonia of unknown aetiology: a pilot study.

The composition of the upper respiratory tract microbiome may play an important role in the development of lower respiratory tract infections. Here, we characterised the microbiome of the nasopharynx and oropharynx of hospitalised patients with community-acquired pneumonia (CAP) with unknown aetiology in an attempt to obtain insight into the aetiology of CAP. A random sample of 10 patients hospitalised with CAP previously enrolled in a separate clinical trial (ClinicalTrials.gov registry, Study ID: NCT01248715) in which a complete microbiological workup was not able to define an aetiology were analysed in this pilot study. This larger trial (n = 1,221) enrolled patients from 9 adult hospitals in Louisville, Kentucky, USA. Nasopharyngeal and oropharyngeal swabs were obtained for metagenomic analysis. Polymerase chain reaction (PCR) for Streptococcus pneumoniae was performed in all patients. One patient had a distinct nasophararyngeal microbiome consisting largely of Haemophilus influenzae. This was the only patient with a negative PCR for S. pneumoniae in both nasophararyngeal and oropharyngeal specimens. Overall, substantial differences were found between nasophararyngeal and oropharyngeal microbiomes. The upper respiratory tract microbiome of only one patient suggested H. influenzae as a probable aetiology of CAP. Although this was a pilot study of only 10 patients, the presence of S. pneumoniae in the upper respiratory tract of the other 9 patients warrants further investigation.

Evaluation of the effectiveness of improved hydrogen peroxide in the operating room.

The aim of our study was to evaluate the effectiveness of a 1-step, ready-to-use improved hydrogen peroxide (IHP) cleaner and disinfectant for between-case use in the operating room. We found high cleaning efficacy (84%-96%) and a high compliance rate with the IHP (84%). With good surface compatibility, low toxicity, rapid dwell time, ease of use, and excellent cleaning efficacy, the IHP may be considered an option for between-case cleaning and disinfection in the operating room.

A single genotype of Acinetobacter baumannii expresses multiple antibiotic susceptibility phenotypes.

It is common practice to define a bacterium with unequal antibiotic susceptibility phenotypes as a different genotype. Here, we describe the utility of this practice for Acinetobacter baumannii. We determined antibiotic susceptibility for 19 clinical isolates of A baumannii. Repetitive sequence polymerase chain reaction method was used to determine genotypes. Although all isolates were of the same genotype, 79% were predicted to be different based on susceptibility patterns. The antibiotic susceptibility phenotype does not accurately predict the genotype of A baumannii.

The value of ready-to-use disinfectant wipes: compliance, employee time, and costs.

Cleaning and disinfection practices of environmental surfaces are critical interventions for reducing health care-associated infections. We studied the value of ready-to-use cleaning and disinfection wipes compared with the traditional towel and bucket method. When using ready-to-use wipes, we found compliance to be significantly higher, a more rapid cleaning and disinfection process, and potential cost savings. Facilities should consider these products when making environmental services product selections.

The effectiveness of the polysaccharide pneumococcal vaccine for the prevention of hospitalizations due to Streptococcus pneumoniae community-acquired pneumonia in the elderly differs between the sexes: results from the Community-Acquired Pneumonia Organization (CAPO) international cohort study.

BACKGROUND: The effectiveness of the 23-valent pneumococcal polysaccharide vaccine (PPV23) to prevent hospitalizations due to Streptococcus pneumoniae community-acquired pneumonia (SpCAP) is controversial. Recent literature suggests that vaccine effectiveness may be influenced by sex. In this study, we define the effectiveness of prior PPV23 vaccination for the prevention of hospitalizations due to SpCAP, and evaluate the impact of sex on this effectiveness. METHODS: This was a nested case-control study from the CAPO international cohort study database. SpCAP was defined as CAP plus S. pneumoniae identified in blood, bronchoalveolar lavage, sputum, or urinary antigen. Vaccination with PPV23 prior to hospitalization was defined as documented in the medical record. A propensity score-weighted logistic regression model was used to calculate odds ratios. The adjusted vaccine effectiveness (aVE) was calculated as 1-adjusted odds ratio. RESULTS: From a total of 2688 elderly adult hospitalized patients with CAP, SpCAP was identified in 279 (10%). The overall aVE was 37% (95% CI: 10.1-55.4%, P=0.01). For males, the aVE was 34% (95% CI:-1.0% to 57.3%, P=0.06). For females the aVE was 68% (95% CI: 40.3-83.0%, P=0.001). CONCLUSIONS: PPV23 protects elderly patients from hospitalization due to SpCAP, but female sex drives the effectiveness. Future analysis of vaccine trials should consider the importance of sex as a stratification factor.

The role of infection prevention conferences to build and maintain knowledge-sharing networks: a longitudinal evaluation.

Well-connected knowledge-sharing networks (KSNs) of infection preventionists are vital to the profession. Face-to-face networking during conferences is considered critical to build and maintain KSNs; however, this has never been formally studied. We used a pre-post survey design to evaluate the effect of a regional infection prevention meeting on the KSNs. We found that the meeting did not alter the KSNs of infection preventionists. Current meeting structures should be re-evaluated with a goal to improve KSNs.