Development of goal-sensitive health-related utility assessment procedures.

PURPOSE: To measure the degree to which people express willingness to trade life or health for nonmedical goals. METHOD: In 3 studies, outpatients provided important life goals. In study 1, patients performed time-tradeoff between life-years and goal achievement and chose between states that varied in goal achievement, life expectancy, and disability; in study 2, patients made choices that traded off health state and goal achievement; in study 3, patients performed time-tradeoff assessments in 3 different goal achievement contexts. RESULTS: In study 1 (n = 58), participants were eager to trade life-years for goal achievement, trading, on average, 71% of their remaining life for certain achievement v. certain nonachievement or 54% of their remaining life for their expected likelihood of achievement v. nonachievement. Life expectancy, disability status, and goal achievement each had a significant main effect on utility. In study 2 (n = 54), participants equally preferred a moderately impaired health state with goal achievement to perfect health without goal achievement and more strongly preferred the moderately impaired state with goal achievement than other less impaired states without goal achievement. Study 3 (n = 62) demonstrated that the mere discussion of goals and goal achievement or nonachievement in the context of a standard time-tradeoff assessment (without trading off goals) did not impact the assessment. CONCLUSIONS: Nonmedical life goals are important determinants of quality of life. People express willingness to trade off life and health in pursuit of these goals, which are extrinsic to the standard quality-adjusted life-year model.

A randomized controlled trial to increase physical activity and reduce obesity in a predominantly African American group of women with mobility disabilities and severe obesity.

OBJECTIVE: This randomized controlled trial tested a tailored, telephone-based physical activity coaching intervention for a predominantly African American group of women with severe obesity and mobility disability. METHODS: We recruited 92 clinic patients from the University of Illinois at Chicago Medical Center referred by their physicians during 2004-2007 and randomized participants to one of three groups--awareness(informational brochure, no coaching), lower support (phone coaching only) and higher support (phone coaching plus monthly exercise support group)--to determine the efficacy of a tailored coaching intervention on key health outcomes, which included body weight and body mass index, blood pressure, cholesterol, physical activity (barriers and self-reported activity), movement and mobility, general health, and social support. RESULTS: The higher support group had the greatest reduction in Body Mass Index (BMI) (7.4%) compared with a 0.2% and 1.6% increase in BMI for the lower support and awareness groups, respectively (pb.01). Both the higher and lower support groups had a greater increase in physical activity scores (39% and 30%, respectively)compared with a decline of 13% in the awareness group (pb.05). CONCLUSION: Providing phone-based coaching and monthly in-person exercise support group sessions appear to be an effective approach for reducing body weight and increasing physical activity among severely obese, disabled adults residing in difficult social environments.

Life goals and health decisions: what will people live (or die) for?

OBJECTIVE: Quality of life may represent not just quality of health but also the degree to which an individual achieves personally meaningful extrinsic goals unrelated to life duration that are not incorporated in the standard quality-adjusted life year model. The objectives of this study are to develop a typology of life goals and explore whether goal type is related to willingness to consider trading life years or health for goals. DESIGN: . Surveys of 50 Chicago-area residents and 101 inpatients. Outcomes. Participants provided up to 5 goals. For each, they reported 1) how long the goal might take to achieve, 2) whether they would prefer a shorter lifetime with certain goal achievement to their full lifetime without goal achievement, and 3) whether they would prefer lower quality of health with certain goal achievement to their full health without goal achievement. RESULTS: Participant goals were classified by 2 investigators into 7 broad categories: family, wealth, job, education, health/fitness, travel, and personal fulfillment. Respondents in both samples were more likely to be willing to trade life years (community odds ratio [OR] = 7.39, P=0.0004; patient OR=1.82, P=0.008) or health (community OR= 5.11, P = 0.0042; patient OR = 1.83, P = 0.0498) to achieve family goals than other types of goals. CONCLUSIONS: The authors derive a manageable typology of goals that may affect medical decisions and demonstrate interrater reliability. Because willingness to trade life years varies by type of goal, typical time-tradeoff assessments may be systematically influenced by respondents' goals.

Predictors of urinary tract infection based on artificial neural networks and genetic algorithms.

BACKGROUND: Among women who present with urinary complaints, only 50% are found to have urinary tract infection. Individual urinary symptoms and urinalysis are not sufficiently accurate to discriminate those with and without the diagnosis. METHODS: We used artificial neural networks (ANN) coupled with genetic algorithms to evolve combinations of clinical variables optimized for predicting urinary tract infection. The ANN were applied to 212 women ages 19-84 who presented to an ambulatory clinic with urinary complaints. Urinary tract infection was defined in separate models as uropathogen counts of > or =10(5) colony-forming units (CFU) per milliliter, and counts of > or =10(2) CFU per milliliter. RESULTS: Five-variable sets were evolved that classified cases of urinary tract infection and non-infection with receiver-operating characteristic (ROC) curve areas that ranged from 0.853 (for uropathogen counts of > or =10(5) CFU per milliliter) to 0.792 (for uropathogen counts of > or =10(2) CFU per milliliter). Predictor variables (which included urinary frequency, dysuria, foul urine odor, symptom duration, history of diabetes, leukocyte esterase on urine dipstick, and red blood cells, epithelial cells, and bacteria on urinalysis) differed depending on the pathogen count that defined urinary tract infection. Network influence analyses showed that some variables predicted urine infection in unexpected ways, and interacted with other variables in making predictions. CONCLUSIONS: ANN and genetic algorithms can reveal parsimonious variable sets accurate for predicting urinary tract infection, and novel relationships between symptoms, urinalysis findings, and infection.

Placebo surgery research: a blinding imperative.

In placebo surgery research, intervention group subjects receive a real surgical procedure, and control subjects receive a sham surgical procedure. The sham procedure is designed to recapitulate enough of the real procedure to blind participants to group assignment. Placebo surgery designs are used to control for placebo effects, and to eliminate the bias that might result if participants and investigators were aware of group assignment when outcomes are measured. In placebo surgery studies, in which participants are placed at increased risk to achieve the scientific validity provided by blinding, it is imperative that the blind be maintained, and maintainable, to justify putting subjects at risk to achieve it. In circumstances in which the outcome assessors are not fully blinded, or in which the intervention is not amenable to rigorous blinding, placebo surgery studies are not ethical. Investigators doing placebo surgery trials should obtain data on the integrity of the blind from both subjects and outcome assessors, to inform the validity of the results and to provide input for "blindability" decisions involving future trials.

Implementation and evaluation of a low-literacy diabetes education computer multimedia application.

OBJECTIVE: To evaluate a clinic-based multimedia intervention for diabetes education targeting individuals with low health literacy levels in a diverse population. RESEARCH DESIGN AND METHODS: Five public clinics in Chicago, Illinois, participated in the study with computer kiosks installed in waiting room areas. Two hundred forty-four subjects with diabetes were randomized to receive either supplemental computer multimedia use (intervention) or standard of care only (control). The intervention includes audio/video sequences to communicate information, provide psychological support, and promote diabetes self-management skills without extensive text or complex navigation. HbA(1c) (A1C), BMI, blood pressure, diabetes knowledge, self-efficacy, self-reported medical care, and perceived susceptibility of complications were evaluated at baseline and 1 year. Computer usage patterns and implementation barriers were also examined. RESULTS: Complete 1-year data were available for 183 subjects (75%). Overall, there were no significant differences in change in A1C, weight, blood pressure, knowledge, self-efficacy, or self-reported medical care between intervention and control groups. However, there was an increase in perceived susceptibility to diabetes complications in the intervention group. This effect was greatest among subjects with lower health literacy. Within the intervention group, time spent on the computer was greater for subjects with higher health literacy. CONCLUSIONS: Access to multimedia lessons resulted in an increase in perceived susceptibility to diabetes complications, particularly in subjects with lower health literacy. Despite measures to improve informational access for individuals with lower health literacy, there was relatively less use of the computer among these participants.

Do physicians know when their diagnoses are correct? Implications for decision support and error reduction.

OBJECTIVE: This study explores the alignment between physicians' confidence in their diagnoses and the "correctness" of these diagnoses, as a function of clinical experience, and whether subjects were prone to over-or underconfidence. DESIGN: Prospective, counterbalanced experimental design. SETTING: Laboratory study conducted under controlled conditions at three academic medical centers. PARTICIPANTS: Seventy-two senior medical students, 72 senior medical residents, and 72 faculty internists. INTERVENTION: We created highly detailed, 2-to 4-page synopses of 36 diagnostically challenging medical cases, each with a definitive correct diagnosis. Subjects generated a differential diagnosis for each of 9 assigned cases, and indicated their level of confidence in each diagnosis. MEASUREMENTS AND MAIN RESULTS: A differential was considered "correct" if the clinically true diagnosis was listed in that subject's hypothesis list. To assess confidence, subjects rated the likelihood that they would, at the time they generated the differential, seek assistance in reaching a diagnosis. Subjects' confidence and correctness were "mildly" aligned (kappa=.314 for all subjects, .285 for faculty, .227 for residents, and .349 for students). Residents were overconfident in 41% of cases where their confidence and correctness were not aligned, whereas faculty were overconfident in 36% of such cases and students in 25%. CONCLUSIONS: Even experienced clinicians may be unaware of the correctness of their diagnoses at the time they make them. Medical decision support systems, and other interventions designed to reduce medical errors, cannot rely exclusively on clinicians' perceptions of their needs for such support.

Use of genetic algorithms for neural networks to predict community-acquired pneumonia.

BACKGROUND: Genetic algorithms have been used to solve optimization problems for artificial neural networks (ANN) in several domains. We used genetic algorithms to search for optimal hidden-layer architectures, connectivity, and training parameters for ANN for predicting community-acquired pneumonia among patients with respiratory complaints. METHODS: Feed-forward back-propagation ANN were trained on sociodemographic, symptom, sign, comorbidity, and radiographic outcome data among 1044 patients from the University of Illinois (the training cohort), and were applied to 116 patients from the University of Nebraska (the testing cohort). Binary chromosomes with genes representing network attributes, including the number of nodes in the hidden layers, learning rate and momentum parameters, and the presence or absence of implicit within-layer connectivity using a competition algorithm, were operated on by various combinations of crossover, mutation, and probabilistic selection based on network mean-square error (MSE), and separately on average cross entropy (ENT). Predictive accuracy was measured as the area under a receiver-operating characteristic (ROC) curve. RESULTS: Over 50 generations, the baseline genetic algorithm evolved an optimized ANN with nine nodes in the first hidden layer, zero nodes in the second hidden layer, learning rate and momentum parameters of 0.5, and no within-layer competition connectivity. This ANN had an ROC area in the training cohort of 0.872 and in the testing cohort of 0.934 (P-value for difference, 0.181). Algorithms based on cross-generational selection, Gray coding of genes prior to mutation, and crossover recombination at different genetic levels, evolved optimized ANN identical to the baseline genetic strategy. Algorithms based on other strategies, including elite selection within generations (training ROC area 0.819), and inversions of genetic material during recombination (training ROC area 0.812), evolved less accurate ANN. CONCLUSION: ANN optimized by genetic algorithms accurately discriminated pneumonia within a training cohort, and within a testing cohort consisting of cases on which the networks had not been trained. Genetic algorithms can be used to implement efficient search strategies for optimal ANN to predict pneumonia.

By-person factor analysis in clinical ethical decision making: Q methodology in end-of-life care decisions.

OBJECTIVE: To determine the usefulness of Q methodology to locate and describe shared subjective influences on clinical decision making among participant physicians using hypothetical cases containing common ethical issues. DESIGN: Qualitative study using by-person factor analysis of subjective Q sort data matrix. SETTING: University medical center. PARTICIPANTS: Convenience sample of internal medicine attending physicians and house staff (n = 35) at one midwestern academic health sciences center. INTERVENTIONS: Presented with four hypothetical cases involving urgent decision making near the end of life, participants selected one of three specific clinical actions offered for each case. Immediately afterward and while considering their decision, each respondent sorted twenty-five subjective self-referent items in terms of the influence of each statement on their decision-making process. By-person factor analysis, where participants are defined as variates, yielded information about the attitudinal background the physicians brought to their consideration of each hypothetical case. We performed a second-order factor analysis on all of the subjective viewpoints to determine if a smaller core of shared attitudes existed across some or all of the four case vignettes. Factor scores for each item and post-sort comments from interviews conducted individually with each respondent guided the interpretation of ethical perspective used by these respondents in making clinical decisions about the cases. MEASUREMENTS AND MAIN RESULTS: Second-order factor analysis on seventeen viewpoints used by physicians in the four hypothetical urgent decision cases revealed three moderately correlated (r2 < 40%) subjective core attitudinal guides used broadly among all the cases and among sixteen of the seventeen original factors. Across all the cases, our participants were guided in general by: (1) patient-focused beneficence, (2) a patient- and surrogate-focused perspective that includes risk avoidance, and (3) best interest of the patient guided by ethical values. Economic impact on the physician, expediency in resolution of the situation, and the expense of medical treatment were not found to be influential determinants in this study. CONCLUSIONS: Q sorting and by-person factor analysis are useful qualitative methodological tools to study the complex structure of subjective attitudes that influence physicians in making medical decisions. This study revealed the subjective viewpoints used by our physician participants as they made ethically challenging treatment decisions. The three second-order factors identified here are grounded in current bioethical values as well as the personal traits of physicians. The participants' decision methods appear to resemble casuistry more than principle-based decision making. Generalizability of results will require further studies.

Prediction of community-acquired pneumonia using artificial neural networks.

BACKGROUND: Artificial neural networks (ANN) have been used in the prediction of several medical conditions but have not been previously used to predict pneumonia. The authors used ANN to predict the presence or absence of pneumonia among patients presenting to the emergency department with acute respiratory complaints and compared the results with those obtained using logistic regression modeling. METHODS: Feed-forward back-propagation ANN were trained on sociodemographic, symptom, sign, comorbidity, and radiographic outcome data among 1,044 patients from the University of Illinois (the training cohort) and were applied to 116 patients from the University of Nebraska (the testing cohort). ANN trained using different strategies were compared to each other and to main-effects logistic regression. Calibration accuracy was measured as mean square error and discrimination accuracy as the area under a receiver operating characteristic (ROC) curve. RESULTS: A 1 hidden-layer ANN trained using oversampling of pneumonia cases had an ROC area in the training cohort of 0.895, which was greater than the area of 0.840 for logistic regression (P = 0.026). This ANN had an ROC area in the testing cohort of 0.872, not significantly different from its area in the training cohort (P = 0.597). Operating at a threshold of 0.25, the ANN would have detected 94% to 95% of patients with pneumonia in the 2 cohorts while correctly excluding 39% to 50% of patients with other conditions. ANN trained using other strategies discriminated equally in the 2 cohorts but no better than did logistic regression. CONCLUSIONS: Among adults presenting with acute respiratory illness, ANN accurately discriminated patients with and without pneumonia and, under some circumstances, improved on the accuracy of logistic regression.

Token swap test revisited.

The token swap test measures the association between row and column variables of a 2 x 2 table in sample misclassification space, and makes no assumptions about repeated, random sampling from a source population. Despite its conceptual usefulness, the token swap test is not implemented by standard statistical software packages. Here the author describes 'tokenSwaps', a Mathematica program that performs a token swap test. The 'tokenSwaps' program also performs a one-tailed Fisher exact test, allowing results of the two methods to be compared. The program uses recursive functional programming and local rewrite rules to achieve substantial coding economy. Examples of the operation of the program are given, and its limitations are discussed.

Exploring the boundaries of plausibility: empirical study of a key problem in the design of computer-based clinical simulations.

All clinical simulation designers face the problem of identifying the plausible diagnostic and management options to include in their simulation models. This study explores the number of plausible diagnoses that exist for a given case, and how many subjects must work up a case before all plausible diagnoses are identified. Data derive from 144 residents and faculty physicians from 3 medical centers, each of whom worked 9 diagnostically challenging cases selected from a set of 36. Each subject generated up to 6 diagnostic hypotheses for each case, and each hypothesis was rated for plausibility by a clinician panel. Of the 2091 diagnoses generated, 399 (19.1%), an average of 11 per case, were considered plausible by study criteria. The distribution of plausibility ratings was found to be statistically case dependent. Averaged across cases, the final plausible diagnosis was generated by the 28th clinician (sd = 8) who worked the case. The results illustrate the richness and diversity of human cognition and the challenges these pose for creation of realistic simulations in biomedical domains.

Parametric receiver operating characteristic curve analysis using mathematica.

Several computer programs have been written to perform receiver operating characteristic (ROC) curve analysis, and are available in the public domain. Here, the author provides the theory and description for 'rocMath', a Mathematica program that performs parametric ROC curve analysis. The 'rocMath' program has some advantages over other ROC curve programs, including the ability to provide, through optional arguments: (a) user-specified pointwise confidence limits, as well as default 95% limits, on ROC curve area and on true-positive rates; (b) ROC curve plots with data points, a fitted curve, and user-specified pointwise confidence bands; and (c) ROC curve areas, tables, and plots based on a logistic distribution as well as on a standard normal distribution. In addition, the code of 'rocMath' can be modified to address additional ROC curve applications. The program uses Mathematica's ability to operate on purely symbolic as well as numeric data to achieve substantial coding efficiency. Limitations of the 'rocMath' program are also discussed.