Stephenie C. Lemon

Classification and regression tree analysis in public health: Methodological review and comparison with logistic regression

Background: Audience segmentation strategies are of increasing interest to public health professionals who wish to identify easily defined, mutually exclusive population subgroups whose members share similar characteristics that help determine participation in a health-related behavior as a basis for targeted interventions. Classification and regression tree (C&RT) analysis is a nonparametric decision tree methodology that has the ability to efficiently segment populations into meaningful subgroups. However, it is not commonly used in public health.Purpose: This study provides a methodological overview of C&RT analysis for persons unfamiliar with the procedure.Methods and Results: An example of a C&RT analysis is provided and interpretation of results is discussed. Results are validated with those obtained from a logistic regression model that was created to replicate the C&RT findings. Results obtained from the example C&RT analysis are also compared to those obtained from a common approach to logistic regression, the stepwise selection procedure. Issues to consider when deciding whether to use C&RT are discussed, and situations in which C&RT may and may not be beneficial are described.Conclusions: C&RT is a promising research tool for the identification of at-risk populations in public health research and outreach.

Association between adult attention deficit/hyperactivity disorder and obesity in the US population.

Attention deficit hyperactivity disorder (ADHD) is a neurobehavioral disorder that affects ∼2.9–4.7% of US adults. Studies have revealed high rates of ADHD (26–61%) in patients seeking weight loss treatment suggesting an association between ADHD and obesity. The objective of the present study was to test the association between ADHD and overweight and obesity in the US population. Cross‐sectional data from the Collaborative Psychiatric Epidemiology Surveys were used. Participants were 6,735 US residents (63.9% white; 51.6% female) aged 18–44 years. A retrospective assessment of childhood ADHD and a self‐report assessment of adult ADHD were administered. Diagnosis was defined by three categories: never met diagnostic criteria, met full childhood criteria with no current symptoms, and met full childhood criteria with current symptoms. The prevalence of overweight and obesity was 33.9 and 29.4%, respectively, among adults with ADHD, and 28.8 and 21.6%, respectively, among persons with no history of ADHD. Adult ADHD was associated with greater likelihood of overweight, (odds ratio (OR) = 1.58; 95% confidence interval (CI) = 1.05, 2.38) and obesity (OR = 1.81; 95% CI = 1.14, 2.64). Results were similar when adjusting for demographic characteristics and depression. Mediation analyses suggest that binge eating disorder (BED), but not depression, partially mediates the associations between ADHD and both overweight and obesity. Results suggest that adult ADHD is associated with overweight and obesity.

Contributions of weight perceptions to weight loss attempts: differences by body mass index and gender

Previous studies have consistently observed that women are more likely to perceive themselves as overweight compared to men. Similarly, women are more likely than men to report trying to lose weight. Less is known about the impact that self-perceived weight has on weight loss behaviors of adults and whether this association differs by gender. We conducted a cross-sectional analysis among an employee sample (n=899) to determine the association of self-perceived weight on evidence-based weight loss behaviors across genders, accounting for body mass index (BMI) and demographic characteristics. Women were more likely than men to consider themselves to be overweight across each BMI category, and were more likely to report attempting to lose weight. However, perceiving oneself to be overweight was a strong correlate for weight loss attempts across both genders. The effect of targeting accuracy of self-perceived weight status in weight loss interventions deserves research attention.

Interventions for patients, providers, and health care organizations

Clinicians and the organizations within which they practice play a major role in enabling patient participation in cancer screening and ensuring quality services. Guided by an ecologic framework, the authors summarize previous literature reviews and exemplary studies of breast, cervical, and colorectal cancer screening intervention studies conducted in health care settings. Lessons learned regarding interventions to maximize the potential of cancer screening are distilled. Four broad lessons learned emphasize that multiple levels of factors—public policy, organizational systems and practice settings, clinicians, and patients—influence cancer screening; that a diverse set of intervention strategies targeted at each of these levels can improve cancer screening rates; that the synergistic effects of multiple strategies often are most effective; and that targeting all components of the screening continuum is important. Recommendations are made for future research and practice, including priorities for intervention research specific to health care settings, the need to take research phases into consideration, the need for studies of health services delivery trends, and methods and measurement issues. Cancer 2004.

Design characteristics of worksite environmental interventions for obesity prevention.

Objective: This paper describes the design characteristics of the National Heart, Lung, and Blood Institute (NHLBI)‐funded studies that are testing innovative environmental interventions for weight control and obesity prevention at worksites.

Preventive healthcare use, smoking, and alcohol use among Rhode Island women experiencing intimate partner violence

OBJECTIVE Intimate partner violence (IPV) poses major health threats to women, including increased risk for several chronic health conditions. The impact of IPV on use of preventive health services is not well understood. Although several studies indicate that female victims of IPV have higher rates of alcohol abuse, this has not been replicated in population-based studies. The association of IPV with smoking has not been a major research focus. The purpose of this study was to examine the association between physical and psychological IPV in the past 12 months and preventive healthcare use, smoking, and alcohol use among women. METHODS Data on 1643 women aged 18-54 from the 1999 Rhode Island Behavioral Risk Factor Surveillance System were analyzed. Logistic regression, controlling for age, race, marital status, education, insurance status, and functional disability, was used to model the associations of IPV with (1) checkups, (2) clinical breast examinations (CBEs), (3) Pap smear screening, (4) cigarette smoking, and (5) high-risk alcohol use. RESULTS Prevalence of physical IPV was 4.1%. The prevalence of psychological IPV, in the absence of physical IPV was 4.5%. Physical IPV was associated with receiving regular Pap smears odds ratio ([OR] = 2.39, 95% confidence interval [CI] 1.01-5.70), current smoking (OR = 2.07, 95% CI 1.03-4.18), and high-risk alcohol use (OR = 4.85, 95% CI 2.02-11.60). Psychological IPV was associated with high-risk alcohol use (OR = 3.22, 95% CI 1.46-7.09). CONCLUSIONS Women experiencing IPV regularly access preventive healthcare, providing healthcare providers with opportunities to assess and counsel women for IPV in addition to smoking and high-risk alcohol use.

Association of Post‐Traumatic Stress Disorder and Obesity in a Nationally Representative Sample

Recent studies suggest a possible link between post‐traumatic stress disorder (PTSD) and obesity risk, which would have implications for the development of obesity‐related diseases in this population. The present study examined the association between PTSD and obesity and whether this association differed by sex in a representative sample of the US population. A secondary objective was to determine whether the association between PTSD and obesity was mediated by binge eating disorder (BED). Data were from the Collaborative Psychiatric Epidemiology Surveys (CPES), which comprises three nationally representative cross‐sectional surveys that were conducted between 2001 and 2003. Logistic regression analyses weighted to represent the general US adult population were performed. In the total sample of 20,013 participants, rates of obesity were 24.1% for persons without a lifetime history of PTSD and 32.6% among persons with PTSD in the past year. Adjusting for socio‐demographic characteristics, depression, substance and alcohol abuse/dependence, and psychotropic medication status, past year PTSD was associated with greater likelihood of obesity (odds ratio (OR) = 1.51; 95% confidence interval (CI) = 1.18, 1.95), with no differences by gender. BED did not statistically mediate the relationship between PTSD and obesity. The present study provides support for a link between PTSD and obesity. Findings further existing literature by indicating that the association is consistent across sexes and is not statistically mediated by BED.

Lifestyle behaviours and weight among hospital-based nurses

AIMS The purpose of this study was to (i) describe the weight, weight-related perceptions and lifestyle behaviours of hospital-based nurses, and (ii) explore the relationship of demographic, health, weight and job characteristics with lifestyle behaviours. BACKGROUND The obesity epidemic is widely documented. Worksite initiatives have been advocated. Nurses represent an important part of the hospital workforce and serve as role models when caring for patients. METHODS A sample of 194 nurses from six hospitals participated in anthropometric measurements and self-administered surveys. RESULTS The majority of nurses were overweight and obese, and some were not actively involved in weight management behaviours. Self-reported health, diet and physical activity behaviours were low, although variable by gender, age and shift. Reports of co-worker norms supported low levels of healthy behaviours. CONCLUSIONS Findings reinforce the need to address the hospital environment and culture as well as individual behaviours for obesity control. IMPLICATIONS FOR NURSING MANAGEMENT Nurse managers have an opportunity to consider interventions that promote a climate favourable to improved health habits by facilitating and supporting healthy lifestyle choices (nutrition and physical activity) and environmental changes. Such efforts have the potential to increase productivity and morale and decrease work-related disabilities and improve quality of life.

Prediction of dengue disease severity among pediatric Thai patients using early clinical laboratory indicators.

Background Dengue virus is endemic in tropical and sub-tropical resource-poor countries. Dengue illness can range from a nonspecific febrile illness to a severe disease, Dengue Shock Syndrome (DSS), in which patients develop circulatory failure. Earlier diagnosis of severe dengue illnesses would have a substantial impact on the allocation of health resources in endemic countries. Methods and Findings We compared clinical laboratory findings collected within 72 hours of fever onset from a prospective cohort children presenting to one of two hospitals (one urban and one rural) in Thailand. Classification and regression tree analysis was used to develop diagnostic algorithms using different categories of dengue disease severity to distinguish between patients at elevated risk of developing a severe dengue illness and those at low risk. A diagnostic algorithm using WBC count, percent monocytes, platelet count, and hematocrit achieved 97% sensitivity to identify patients who went on to develop DSS while correctly excluding 48% of non-severe cases. Addition of an indicator of severe plasma leakage to the WHO definition led to 99% sensitivity using WBC count, percent neutrophils, AST, platelet count, and age. Conclusions This study identified two easily applicable diagnostic algorithms using early clinical indicators obtained within the first 72 hours of illness onset. The algorithms have high sensitivity to distinguish patients at elevated risk of developing severe dengue illness from patients at low risk, which included patients with mild dengue and other non-dengue febrile illnesses. Although these algorithms need to be validated in other populations, this study highlights the potential usefulness of specific clinical indicators early in illness.

Step Ahead A Worksite Obesity Prevention Trial Among Hospital Employees

BACKGROUND The worksite represents a promising venue in which to address the issue of obesity. DESIGN Pair-matched, cluster-RCT. Data were collected from 2005 to 2008 and analyzed in 2008. SETTING/PARTICIPANTS A random sample of 806 employees was selected to represent the workforce of six hospitals in central Massachusetts. INTERVENTION The 2-year ecologic intervention sought to prevent weight gain through changes in worksite weight-related norms using strategies targeted at the organization, interpersonal environment, and employees. MAIN OUTCOME MEASURES The primary outcome was change in BMI at the 12- and 24-month follow-ups. Change in perceptions of organizational commitment to employee health and normative coworker behaviors were secondary outcomes. RESULTS There was no impact of the intervention on change in BMI from baseline to 12 (beta=0.272; 95% CI=-0.271, 0.782) or 24 months (beta=0.276; 95% CI=-0.338, 0.890) in intention-to-treat analysis. When intervention exposure (scale=0 to 100) was used as the independent variable, there was a decrease of 0.012 BMI units (95% CI=-0.025, 0.001) for each unit increase in intervention participation at the 24-month follow-up. Employees in intervention sites reported significantly greater improvements in perceptions of organizational commitment to employee health at 12 and 24 months compared to control sites, but there was no impact on perceptions of normative coworker behaviors. CONCLUSIONS The intervention had a dose-response relationship with BMI, with positive effects proportional to extent of participation. Although the intervention was able to change organizational perceptions, successfully improving changes in actual and perceived social norms may be needed to achieve population-level impact in complex worksite organizations.