J. Michael Hardin

Association Rules and Data Mining in Hospital Infection Control and Public Health Surveillance

Objectives: The authors consider the problem of identifying new, unexpected, and interesting patterns in hospital infection control and public health surveillance data and present a new data analysis process and system based on association rules to address this problem. Design: The authors first illustrate the need for automated pattern discovery and data mining in hospital infection control and public health surveillance. Next, they define association rules, explain how those rules can be used in surveillance, and present a novel process and system—the Data Mining Surveillance System (DMSS)—that utilize association rules to identify new and interesting patterns in surveillance data. Results: Experimental results were obtained using DMSS to analyze Pseudomonas aeruginosa infection control data collected over one year (1996) at University of Alabama at Birmingham Hospital. Experiments using one-, three-, and six-month time partitions yielded 34, 57, and 28 statistically significant events, respectively. Although not all statistically significant events are clinically significant, a subset of events generated in each analysis indicated potentially significant shifts in the occurrence of infection or antimicrobial resistance patterns of P. aeruginosa . Conclusion: The new process and system are efficient and effective in identifying new, unexpected, and interesting patterns in surveillance data. The clinical relevance and utility of this process await the results of prospective studies currently in progress.

Agitation in nursing home residents: The role of gender and social context

We investigated the relationship among gender of resident, staff social interaction, and agitation in 46 (31 male and 15 female) nursing home residents with clinically significant agitation. Direct observations were conducted of resident behaviors and environmental contextual events using a computer-assisted, real-time observational system. The system recorded frequency, duration, and temporal sequencing of events. Results show that female residents displayed almost three times the amount of agitation as male residents (35% vs. 13% of total observation time, respectively), although men in the study were more likely to receive psychoactive drugs for their agitation. Staff spent similar amounts of time verbally interacting and touching male and female residents. Sequential analyses were conducted to examine the likelihood of staff verbal and touch interactions both preceding and following resident agitation using Bakeman and Queras (1995) SDIS-GSEQ program. Results suggest that staff touch and verbal interaction elicit agitation in a significant proportion of residents. Once agitation occurs, staff were likely to respond by interacting verbally, but not physically, with the resident.

Obtaining Distribution Functions by Numerical Inversion of Characteristic Functions with Applications

Abstract We review and discuss numerical inversion of the characteristic function as a tool for obtaining cumulative distribution functions. With the availability of high-speed computing and symbolic computation software, the method is ideally suited for instructional purposes, particularly in the illustration of the inversion theorems covered in graduate probability courses. The method is also available as an alternative to asymptotic approximations, Monte Carlo, or bootstrap techniques when analytic expressions for the distribution function are not available. We illustrate the method with several examples, including one which is concerned with the detection of possible clusters of disease in an epidemiologic study.

Compliance With Vision-Screening Guidelines Among a National Sample of Pediatricians

OBJECTIVE The American Academy of Pediatrics (AAP) recommends vision screening from birth through adolescence, with visual acuity testing and binocular screening to begin at age 3 years. The 1996 AAP guidelines advised referral for visual acuity worse than 20/40 for children aged 3 to 5 years and worse than 20/30 for children aged 6 years and older. Our objective was to describe vision-screening and referral practices in a national sample of primary care pediatricians. METHODS We mailed a survey to a random sample of US pediatricians. Initial nonresponders were mailed up to 3 additional surveys. All mailings occurred between May and October 1998. Analyses focused on primary care pediatricians and consisted of descriptive statistics and regression analyses. The main outcome measure was compliance with 1996 AAP recommendations for vision screening. RESULTS Of the 1491 surveys mailed, 888 (60%) were returned, including 576 (65%) from primary care pediatricians. Vision-screening methods included visual acuity testing (92%), cover test (64%), red reflex test (95%), fundoscopic examinations (65%), and stereopsis testing (32%). Respondents routinely performed visual acuity testing at 3 years (37%), 4 years (79%), 5 years (91%), 6 years (80%), 7-12 years (82%), and 13-18 years (80%). Visual acuity thresholds for referring 3- and 4-year-olds were 20/40 (47%, 51%), 20/50 (36%, 32%), or worse than 20/50 (14%, 12%). The majority of pediatricians referred children aged 5 years and older at 20/40, although thresholds worse than 20/40 were reported commonly (18%-33%). Logistic regressions were done to identify factors associated with higher likelihood of performing specific screening tests. Although no factor was consistently associated with use of all screening tests, size of the practice was significant in several regression models. CONCLUSIONS Many pediatricians do not follow AAP guidelines for vision screening and referral, especially in younger children. Two thirds of pediatricians do not begin visual acuity testing at age 3 years as recommended, and about one fifth do not test until age 5 years. In addition, one fourth do not perform cover tests or stereopsis testing at any age.

Excess Disability During Morning Care in Nursing Home Residents With Dementia

Excess disability was examined in 17 nursing home residents with dementia by comparing their performance of morning care tasks under two activities of daily living (ADL) caregiving approaches-a dependence-supportive one under usual care and an independence-supportive one under functional rehabilitation. The results suggest that excess disability in severely cognitively impaired and functionally disabled residents can be reduced by increasing opportunities for independent activity, and substituting nondirective and directive verbal assists for physical assists. Further, the findings indicate that increased independence in ADL can be achieved without increasing disruptive behaviors and can foster appropriate requests for task-related help during caregiving. Functional rehabilitation, however, requires more time than usual care.

Predictors of need-driven behaviors in nursing home residents with dementia and associated certified nursing assistant burden.

Objective: We examined predictors of staff-reported need-driven behaviors and resistiveness to care in nursing home residents with dementia and predictors of certified nursing assistant (CNA) burden related to both constructs. Background and proximal factors from the need-driven dementia-compromised behavior model [Algase, D.L., Beck, C., Kolanowski, A., Whall, A., Berent, S., Richards, K., et al. (1996). Need-driven dementia-compromised behavior: An alternative view of disruptive behavior. American Journal of Alzheimers Disease, 5, 10–19] were examined as potential predictors of need-driven behaviors (NDBs) and resistiveness to care and CNA burden. Method: We used secondary data analysis of prospective data from 10 nursing homes in Birmingham, Alabama. One-hundred and sixty-one residents (83.43 ± 8.56 years) with mini mental state examination (MMSE) score = 6.41 (± 6.66) were assessed via chart review, resident surveys, and surveys of CNAs. Results: Multiple regression models revealed that cognitive functioning, activities of daily living functioning, race, gender, and CNA-reported weekly resident pain intensity were associated with resident NDBs. Regression models also revealed that weekly pain intensity and medical comorbidity were associated with CNA burden associated with the resident NDBs. However, we were unable to explain a significant amount of variance in the resistiveness to care or CNA burden associated with resistiveness to care. Discussion: Results underscore the role of pain in both resident NDBs and associated CNA burden. Future research should focus on predictors of resident resistiveness to care and the relation of pain assessment and management practices to CNA burden. Moreover, interventions to improve resident care should seek to include CNAs in institutional pain assessment and management processes.

Data Mining and Clinical Decision Support Systems

Data mining is a process of pattern and relationship discovery within large sets of data. Because of the large volume of data generated in healthcare settings, it is not surprising that healthcare organizations have been interested in data mining to enhance physician practices, disease management, and resource utilization. This chapter discusses a variety of data mining techniques that have been used to develop clinical decision support systems, including decision trees, neural networks, logistic regression, nearest neighbor classifiers. In addition, genetic algorithms, biologic and quantum computing, and big data analytics as well as methods of evaluating and comparing the different approaches are also discussed.

Lack of Independent Relationships Between Left Ventricular Mass and Cardiovascular Reactivity to Physical and Psychological Stress in the Coronary Artery Risk Development in Young Adults (CARDIA) Study

The objective of this study was to determine whether exaggerated blood pressure (BP) reactivity to stress and psychosocial characteristics are related to left ventricular mass (LVM) in a large cohort of young adults. Analyses were conducted with 3,742 participants of the CARDIA study (945 white men, 1,024 white women, 781 black men, and 992 black women), evaluated in 1990 to 1091 with echocardiographic measurement of LVM. Analyses were stratified by gender and race. The relationships of LVM/height2.7 and cardiovascular reactivity to physical and psychological stressors (treadmill exercise, cold pressor, video game, and star-tracing tasks), were examined in both univariate and multivariate analyses adjusting for baseline BP, weight, and other relevant biobehavioral variables. The relationships between LVM and several psychosocial characteristics (hostility, anger suppression, anxiety, depressive symptoms, and education) were also assessed. Systolic blood pressure (SBP) reactivity to exercise was significantly related to LVM in black and white men; LVM was 10% greater among white men with exaggerated (upper quintile) peak exercise SBP than among other white men. SBP reactivity to the cold pressor test was related to LVM in all race/gender groups, although the relationship remained significant only among white men and women in the multivariate analysis. Diastolic blood pressure (DBP) reactivity to the video game was related to LVM only among black men in adjusted analyses. After adjusting for resting BP, weight, and other covariates in linear multiple regression models, SBP reactivity to exercise explained only 3% of the variance in LVM among white men. Otherwise, reactivity to other stressors or psychosocial variables accounted for no more than 1% of the variance in LVM. It was concluded that among a cohort of young adults, blood pressure reactivity to physical and mental stressors did not add substantially to the prediction of LVM when resting BP, weight, and other covariates were taken into account.

Neurocognitive and Emotional Functioning in Lung Transplant Candidates: A Preliminary Study

This study reports descriptive data on the neurocognitive and emotional functioning of a clinical series of 39 lung transplant candidates. Results found this group to be of average intelligence, with average simple attention and concentration. Candidates displayed mild to moderate psychomotor slowing, moderate dysnomia, and mild verbal memory and learning deficits. Relative to available normative data, 39% of the patients demonstrated performance deficits at or below three standard deviations from the mean on two or more neurocognitive test variables. Cluster analysis of MMPIs identified three groups: Cluster 1 consisted of 26% of the sample, with significant elevations on scales 1, 2, 3, 4, 7, and 8 suggestive of clinically significant levels of emotional distress; Cluster 2 consisted of 39% of the sample, with significant elevations on scales 1, 2, and 3 suggestive of moderate somatic concerns and mild emotional disruption; Cluster 3 consisted of 35% of the sample, with no significant MMPI scale elevations. Nearly one-third of the sample obtained MMPI F-K scores suggestive of a “fake-good” response set. Neurocognitive test performance and MMPI scales were only minimally related, suggesting that emotional disruption cannot be used as an explanation for neurocognitive deficits in this sample.

Validity of pure-tone hearing screening at well-child visits.

OBJECTIVE To estimate the sensitivity and specificity of pure-tone audiometry hearing screening in the primary care setting. DESIGN Prospective cohort study. SETTING Eight academic and private pediatric practices. PARTICIPANTS A subset of children from a convenience sample of 1061 children between 3 and 19 years of age were screened for hearing loss using pure-tone audiometry. Intervention Formal audiologic evaluations (gold standard) for those children referred by their primary care physician (28 children) and for a random sample of children not referred (102 children). Main Exposure Pure-tone audiometry screening. MAIN OUTCOME MEASURES Audiologic evaluations. RESULTS A total of 28 children were referred to an audiologist for formal hearing testing after pure-tone audiometry screening during a well-child visit, at which 25 children did not pass the initial screening and 3 could not complete the screening. Of the 25 children, only 7 were evaluated by an audiologist, for a follow-up rate of 25%. One child was diagnosed as having hearing loss. Formal audiologic assessment was also performed on a random sample of 102 children who were not referred to the audiologist. For the random sample, hearing loss was identified in 2 of 76 (3%) children who passed and 1 of 16 (6%) children who did not pass pure tone audiometry screening. The sensitivity and specificity of pure-tone audiometry were 50% and 78%, respectively. CONCLUSION In light of the increasing burden on physicians to provide preventive care, this study calls into question the value of hearing screening using pure-tone audiometry during well-child visits given the lack of follow-up after referral and the poor sensitivity.