Search

Background: Hospital-acquired infections are commonly spread through the movement of healthcare professionals (HCPs). Computational simulations provide a powerful tool for understanding how HCP behavior contributes to these infections, but how well they reflect the real world rests on a number of critical parameters. Our goal is to provide accurate, fine-grained estimates of real HCP movement and interaction parameters suitable for simulating the potential spread of pathogens over different types of inpatient facilities. Methods: We obtained a commercial data set with 44 million deidentified elements compiled from >27,000 HCPs from >30 job types. The data were collected over 27 months from >20 facilities of varying size using a proprietary electronic sensor system. Each observation recorded an HCP visiting 1 of 12,000 rooms (38% being patient rooms) and consisted of the entry and exit time stamps, hand hygiene behavior, and for many rooms, their (x, y) geometric coordinates within the facility. From these data, we can reconstruct the behavior (including location and hand-hygiene adherence) of each instrumented HCP across multiple shifts. Results: Distributions describing various aspects of HCP behavior (eg, arrival rates and dwell times) were derived using HCP job function, department or unit assignment, type of shift (day vs night), time of day, facility size, and staffing of facility. In a similar fashion, we constructed HCP cross-table transition probabilities using job type, room type, department type, unit type, and facility type. These distributions were used to generate reasonable HCP movement and behavior patterns in a simulation environment. Distributions of dwell time were, for the most part, heavy tailed, but they varied by type of job and facility: dwell times over all facilities, job types, and room types averaged ∼339 seconds (SD, 495 seconds), with a mean of maximums by job type of ∼37,168 seconds. However, these distributions differ within job type but across facilities (ie, nurses in 1 facility averaged 397 seconds, but 277 seconds in another) and within facility but across job type. For example, physicians averaged 292 seconds, whereas nurses averaged 397 seconds and physical therapists averaged 861 seconds. Conclusions: Our results provide a unique resource for disease modelers who wish to build meaningful simulations of the transmission of hospital-acquired infections. The scale and diversity of the data gave us the unique capability to provide, with confidence, distinct parameter sets for different types and sizes of healthcare facilities across a wide range of situations.Funding: NoneDisclosures: None

Background: Simulations based on models of healthcare worker (HCW) mobility and contact patterns with patients provide a key tool for understanding spread of healthcare-acquired infections (HAIs). However, simulations suffer from lack of accurate model parameters. This research uses Microsoft Kinect cameras placed in a patient room in the medical intensive care unit (MICU) at the University of Iowa Hospitals and Clinics (UIHC) to obtain reliable distributions of HCW visit length and time spent by HCWs near a patient. These data can inform modeling efforts for understanding HAI spread. Methods: Three Kinect cameras (left, right, and door cameras) were placed in a patient room to track the human body (ie, left/right hands and head) at 30 frames per second. The results reported here are based on 7 randomly selected days from a total of 308 observation days. Each tracked body may have multiple raw segments over the 2 camera regions, which we “stitch” up by matching features (eg, direction, velocity, etc), to obtain complete trajectories. Due to camera noise, in a substantial fraction of the frames bodies display unnatural characteristics including frequent and rapid directional and velocity change. We use unsupervised learning techniques to identify such “ghost” frames and we remove from our analysis bodies that have 20% or more “ghost” frames. Results: The heat map of hand positions (Fig. 1) shows that high-frequency locations are clustered around the bed and more to the patient’s right in accordance with the general medical practice of performing patient exams from their right. HCW visit frequency per hour (mean, 6.952; SD, 2.855) has 2 peaks, 1 during morning shift and 1 during the afternoon shift, with a distinct decrease after midnight. Figure 2 shows visit length (in minutes) distribution (mean, 1.570; SD, 2.679) being dominated by “check in visits” of Conclusions: Using fine-grained data, this research extracts distributions of these critical parameters of HCW–patient interactions: (1) HCW visit length, (2) HCW visit frequency as a function of time of day, and (3) time spent by HCW within touching distance of patient as a function of visit length. To the best of our knowledge, we provide the first reliable estimates of these parameters.Funding: NoneDisclosures: None

Background Sedentary work is hazardous. Over 80% of all US jobs are predominantly sedentary, placing full‐time office workers at increased risk for cardiovascular and metabolic morbidity and mortality. Thus, there is a critical need for effective workplace physical activity interventions. MapTrek is a mobile health platform that gamifies Fitbit use for the purpose of promoting physical activity. The purpose of this study was to test the efficacy of MapTrek for increasing daily steps and moderate‐intensity steps over 10 weeks in a sample of sedentary office workers. Methods and Results Participants included 146 full‐time sedentary office workers aged 21 to 65 who reported sitting at least 75% of their workday. Each participant received a Fitbit Zip to wear daily throughout the intervention. Participants were randomized to either a: (1) Fitbit‐only group or 2) Fitbit + MapTrek group. Physical activity outcomes and intervention compliance were measured with the Fitbit activity monitor. The Fitbit + MapTrek group significantly increased daily steps (+2092 steps per day) and active minutes (+11.2 min/day) compared to the Fitbit‐only arm, but, on average, participants’ steps declined during the study period. Conclusions MapTrek is an effective approach for increasing physical activity at a clinically meaningful level in sedentary office workers, but as with accelerometer use alone, the effect decreases over time. Clinical Trial Registration URL : https://www.clinicaltrials.gov . Unique identifier: NCT 03109535.

To diagnose hypertension, multiple blood pressure (BP) measurements are recommended. We randomized patients into three groups: EMR–only (patients recorded BP measurements in an electronic medical record [EMR] web portal), EMR + reminders (patients were sent text message reminders to record their BP measurements in the EMR), and bi–directional text messaging (patients were sent a text message asking them to respond with their current BP). Subjects were asked to complete 14 measurements. Automated messages were sent to each patient in the bi–directional text messaging and EMR + reminder groups twice daily. Among 121 patients, those in the bi–directional text messaging group reported the full 14 measurements more often than both the EMR–only group (P

Traditional disease surveillance systems suffer from several disadvantages, including reporting lags and antiquated technology, that have caused a movement towards internet-based disease surveillance systems. Internet systems are particularly attractive for disease outbreaks because they can provide data in near real-time and can be verified by individuals around the globe. However, most existing systems have focused on disease monitoring and do not provide a data repository for policy makers or researchers. In order to fill this gap, we analyzed Wikipedia article content. We demonstrate how a named-entity recognizer can be trained to tag case counts, death counts, and hospitalization counts in the article narrative that achieves an F1 score of 0.753. We also show, using the 2014 West African Ebola virus disease epidemic article as a case study, that there are detailed time series data that are consistently updated that closely align with ground truth data. We argue that Wikipedia can be used to create the first community-driven open-source emerging disease detection, monitoring, and repository system.

This large-scale study, consisting of 24.5 million hand hygiene opportunities spanning 19 distinct facilities in 10 different states, uses linear predictive models to expose factors that may affect hand hygiene compliance. We examine the use of features such as temperature, relative humidity, influenza severity, day/night shift, federal holidays and the presence of new residents in predicting daily hand hygiene compliance. The results suggest that colder temperatures and federal holidays have an adverse effect on hand hygiene compliance rates, and that individual cultures and attitudes regarding hand hygiene seem to exist among facilities., Comment: Accepted to ICHI 2017. 8 pages

Background Understanding how C. diffcile infection (CDI) is acquired in healthcare settings is key to designing interventions to mitigate CDI. The goal of this research is apply statistical methods, typically used to investigate regional outbreaks, to study spatio-temporal clustering of in-hospital CDI incidence. Methods We analyzed 1,804 CDI cases (out of 241,248 in-patient visits) at the University of Iowa Hospitals and Clinics (UIHC) during January 2005–December 2011. Letting T and D be time and space parameters, we constructed an observed CDI cluster graph by connecting pairs of CDI cases whose positive CDI tests occur within T days and D distance units of each other. In Experiment 1, for each CDI case, we replaced its actual time stamp by one picked uniformly at random from the time interval [January 2005, December 2011] and constructed a random CDI cluster graph. We tested the UIHC CDI case counts for seasonality and observed none, but did observe that the CDI counts increased significantly (weekly mean: 4.12–>8.11) starting in December 2009, when the C. diffcile Toxin A&B test was replaced by the C. diffcile Toxin PCR. So, we performed an Experiment 2 in which we sampled time stamps from a mixture of two uniform distributions, representing the periods of the two tests. Results We report sizes of connected components in the table below, for 10,000 trials of Experiments 1 and 2, for T = 14 days and varying D, a one setting in which D is set to the unit in which the CDI case occurs. The plots show the distribution of the mean and maximum component size (blue curves) for Experiment 2, for D = 2. D ObservedMean Comp. Size Experiment 1 Expected Mean Comp. Size (P-value) Experiment 2 Expected Mean Comp. Size (P-value) Observed Max. Comp. Size Experiment 1 Expected Max. Comp. Size (P-value) Experiment 2 Expected Max. Comp Size (P-value) 2 1.12 1.07 (0) 1.06 (0) 6 3.72 (0.01) 4.07 (0.05) 3 1.19 1.11 (0) 1.11 (0) 7 4.61 (0.03) 5.17 (0.1) 5 1.29 1.18 (0) 1.20 (0) 11 6.6 (0.01) 8.10 (0.1) 10 1.93 1.68 (0) 1.52 (0) 29 17.13 (0.02) 28.08 (0.4) Unit 1.63 1.38 (0) 1.01 (0) 23 9.74 (0) 12.87 (0.01) Conclusion Our analysis of the UIHC CDI cases shows significant spatio-temporal clustering in the observed CDI cluster graph. These results suggest that direct or environmental transmission may play a significant role in CDI acquisition at the UIHC. Funded by the CDC MInD-Healthcare. Disclosures All authors: No reported disclosures.

In the United States, livers for transplantation are distributed within donation service areas (DSAs). In DSAs with multiple transplant centers, competition among centers for organs and recipients may affect recipient selection and outcomes in comparison with DSAs with only 1 center. The objective of this study was to determine whether competition within a DSA is associated with posttransplant outcomes and variations in patients wait-listed within the DSA. United Network for Organ Sharing data for 38,385 adult cadaveric liver transplant recipients undergoing transplantation between January 1, 2003 and December 31, 2009 were analyzed to assess differences in liver recipients and donors and in posttransplant survival by competition among centers. The main outcome measures that were studied were patient characteristics, actual and risk-adjusted graft and patient survival rates after transplantation, organ quality as quantified by the donor risk index (DRI), wait-listed patients per million population by DSA, and competition as quantified by the Hirschman-Herfindahl index (HHI). Centers were stratified by HHI levels as no competition or as low, medium (or mid), or high competition. In comparison with DSAs without competition, the low-, mid-, and high-competition DSAs (1) performed transplantation for patients with a higher risk of graft failure [hazard ratio (HR) = 1.24, HR = 1.26, and HR = 1.34 (P < 0.001 for each)] and a higher risk of death [HR = 1.21, HR = 1.23, and HR = 1.34 (P < 0.001 for each)] and for a higher proportion of sicker patients as quantified by the Model for End-Stage Liver Disease (MELD) score [10.0% versus 14.8%, 20.1%, and 28.2% with a match MELD score of 31-40 (P < 0.001 for each comparison)], (2) were more likely to use organs in the highest risk quartile as quantified by the DRI [18.3% versus 27.6%, 20.4%, and 31.7% (P ≤ 0.001 for each)], and (3) listed more patients per million population [18 (median) versus 34 (P = not significant), 37 (P = 0.005), and 45 (P = 0.0075)]. Significant variability in patient selection for transplantation is associated with market variables characterizing competition among centers. These findings suggest both positive and negative effects of competition among health care providers.

Hand hygiene is important for patient safety; increasing hand hygiene compliance may reduce the frequency of healthcare-associated infections. This paper describes a distributed system that uses instrumented product dispensers and doorway monitors to systematically measure hand hygiene compliance as an alternative to compliance measurements by human observers, which is the current standard. The paper describes two experiments. The first experiment monitored 4,266 doorway crossings and 858 hand hygiene dispenser events for 4 patient rooms over 80 consecutive hours. The second experiment was part of a larger effort that included a direct comparison of a human observer with the automatically recorded observations. The results of the two experiments suggest that large quantities of data could be readily acquired, but the data was sensitive to several limitations not suffered by human observers including: distinguishing between single versus closely spaced multiple threshold crossings and distinguishing staff ...

Objective.To use social network analysis to design more effective strategies for vaccinating healthcare workers against influenza.Design.An agent-based simulation.Setting.A simulation based on a 700-bed hospital.Methods.We first observed human contacts (defined as approach within approximately 0.9 m) performed by 15 categories of healthcare workers (eg, floor nurses, intensive care unit nurses, staff physicians, phlebotomists, and respiratory therapists). We then constructed a series of contact graphs to represent the social network of the hospital and used these graphs to run agent-based simulations to model the spread of influenza. A targeted vaccination strategy that preferentially vaccinated more “connected” healthcare workers was compared with other vaccination strategies during simulations with various base vaccination rates, vaccine effectiveness, probability of transmission, duration of infection, and patient length of stay.Results.We recorded 6,654 contacts by 148 workers during 606 hours of observations from January through December 2006. Unit clerks, X-ray technicians, residents and fellows, transporters, and physical and occupational therapists had the most contacts. When repeated contacts with the same individual were excluded, transporters, unit clerks, X-ray technicians, physical and occupational therapists, and social workers had the most contacts. Preferentially vaccinating healthcare workers in more connected job categories yielded a substantially lower attack rate and fewer infections than a random vaccination strategy for all simulation parameters.Conclusions.Social network models can be used to derive more effective vaccination policies, which are crucial during vaccine shortages or in facilities with low vaccination rates. Local vaccination priorities can be determined in any healthcare facility with only a modest investment in collection of observational data on different types of healthcare workers. Our findings and methods (ie, social network analysis and computational simulation) have implications for the design of effective interventions to control a broad range of healthcare-associated infections.

Background Time and motion studies have been used to investigate how much time various health care professionals spend with patients as opposed to performing other tasks. However, the majority of such studies are done in outpatient settings, and rely on surveys (which are subject to recall bias) or human observers (which are subject to observation bias). Our goal was to accurately measure the time physicians, nurses, and critical support staff in a medical intensive care unit spend in direct patient contact, using a novel method that does not rely on self-report or human observers. Methods We used a network of stationary and wearable mote-based sensors to electronically record location and contacts among health care workers and patients under their care in a 20-bed intensive care unit for a 10-day period covering both day and night shifts. Location and contact data were used to classify the type of task being performed by health care workers. Results For physicians, 14.73% (17.96%) of their time in the unit during the day shift (night shift) was spent in patient rooms, compared with 40.63% (30.09%) spent in the physician work room; the remaining 44.64% (51.95%) of their time was spent elsewhere. For nurses, 32.97% (32.85%) of their time on unit was spent in patient rooms, with an additional 11.34% (11.79%) spent just outside patient rooms. They spent 11.58% (13.16%) of their time at the nurses' station and 23.89% (24.34%) elsewhere in the unit. From a patient's perspective, we found that care times, defined as time with at least one health care worker of a designated type in their intensive care unit room, were distributed as follows: 13.11% (9.90%) with physicians, 86.14% (88.15%) with nurses, and 8.14% (7.52%) with critical support staff (eg, respiratory therapists, pharmacists). Conclusions Physicians, nurses, and critical support staff spend very little of their time in direct patient contact in an intensive care unit setting, similar to reported observations in both outpatient and inpatient settings. Not surprisingly, nurses spend far more time with patients than physicians. Additionally, physicians spend more than twice as much time in the physician work room (where electronic medical record review and documentation occurs) than the time they spend with all of their patients combined.

This prospective study of cellulitis identified β-hemolytic streptococci as the dominating cause in all investigated subgroups. Group C/G streptococci were more frequently detected than group A streptococci. No single clinical feature substantially increased the probability of confirmed streptococcal etiology., Background. We determined the feasibility of using computer vision and depth sensing to detect healthcare worker (HCW)-patient contacts to estimate both hand hygiene (HH) opportunities and personal protective equipment (PPE) adherence. Methods. We used multiple Microsoft Kinects to track the 3-dimensional movement of HCWs and their hands within hospital rooms. We applied computer vision techniques to recognize and determine the position of fiducial markers attached to the patient's bed to determine the location of the HCW's hands with respect to the bed. To measure our system's ability to detect HCW-patient contacts, we counted each time a HCW's hands entered a virtual rectangular box aligned with a patient bed. To measure PPE adherence, we identified the hands, torso, and face of each HCW on room entry, determined the color of each body area, and compared it with the color of gloves, gowns, and face masks. We independently examined a ground truth video recording and compared it with our system's results. Results. Overall, for touch detection, the sensitivity was 99.7%, with a positive predictive value of 98.7%. For gowned entrances, sensitivity was 100.0% and specificity was 98.15%. For masked entrances, sensitivity was 100.0% and specificity was 98.75%; for gloved entrances, the sensitivity was 86.21% and specificity was 98.28%. Conclusions. Using computer vision and depth sensing, we can estimate potential HH opportunities at the bedside and also estimate adherence to PPE. Our fine-grained estimates of how and how often HCWs interact directly with patients can inform a wide range of patient-safety research.

This paper describes nagging, a technique for parallelizing search in a heterogeneous distributed computing environment. Nagging exploits the speedup anomaly often observed when parallelizing problems by playing multiple reformulations of the problem or portions of the problem against each other. Nagging is both fault tolerant and robust to long message latencies. In this paper, we show how nagging can be used to parallelize several different algorithms drawn from the artificial intelligence literature, and describe how nagging can be combined with partitioning, the more traditional search parallelization strategy. We present a theoretical analysis of the advantage of nagging with respect to partitioning, and give empirical results obtained on a cluster of 64 processors that demonstrate nagging's effectiveness and scalability as applied to A∗ search, αβ minimax game tree search, and the Davis–Putnam algorithm.

We present a new distributed algorithm for computing the minimum description length (MDL) in learning Bayesian inference networks from data. Our learning algorithm exploits both properties of the MDL-based score metric and a distributed, asynchronous, adaptive search technique called nagging. Nagging is intrinsically fault-tolerant, has dynamic load balancing features, and scales well. We demonstrate the viability, effectiveness, and scalability of our approach empirically with several experiments using networked machines. More specifically, we show that our distributed algorithm can provide optimal solutions for larger problems as well as good solutions for Bayesian networks of up to 150 variables.

Objective.To determine whether hand hygiene adherence is influenced by peer effects and, specifically, whether the presence and proximity of other healthcare workers has a positive effect on hand hygiene adherenceDesign.An observational study using a sensor network.Setting.A 20-bed medical intensive care unit at a large university hospital.Participants.Hospital staff assigned to the medical intensive care unit.Methods.We deployed a custom-built, automated, hand hygiene monitoring system that can (1) detect whether a healthcare worker has practiced hand hygiene on entering and exiting a patient’s room and (2) estimate the location of other healthcare workers with respect to each healthcare worker exiting or entering a room.Results.We identified a total of 47,694 in-room and out-of-room hand hygiene opportunities during the 10-day study period. When a worker was alone (no recent healthcare worker contacts), the observed adherence rate was 20.85% (95% confidence interval [CI], 19.78%–21.92%). In contrast, when other healthcare workers were present, observed adherence was 27.90% (95% CI, 27.48%–28.33%). This absolute increase was statistically significant (P < .01). We also found that adherence increased with the number of nearby healthcare workers but at a decreasing rate. These results were consistent at different times of day, for different measures of social context, and after controlling for possible confounding factors.Conclusions.The presence and proximity of other healthcare workers is associated with higher hand hygiene rates. Furthermore, our results also indicate that rates increase as the social environment becomes more crowded, but with diminishing marginal returns.Infect Control Hosp Epidemiol 2014;35(10):1277–1285

The OAMulator is a Web-based resource to support the teaching of instruction set architecture, assembly languages, memory, addressing, high-level programming, and compilation. The tool is based on a simple, virtual CPU architecture, called the One Address Machine. A compiler allows us to take programs written in a special programming language, called OAMPL, and transform them into OAM assembly. An OAM assembler/emulator interprets and executes OAM assembly code produced by the compiler or written directly by students. The OAMulator is targeted at non-CS students who take introductory courses in information technology or information systems. Such students are normally exposed to concepts of computer hardware and software, but it is difficult for them to make the connection between the two. The OAMulator takes the mystery out of CPU architecture by letting students gain confidence with the concepts of compilers and binary execution. The Web-based deployment allows students to work on problems in convenient locations, at their own pace, and with rewarding interaction.

This article introduces a parallel search-pruning technique called nagging. Nagging is sufficiently general to be effective in a number of domains; here we focus on an implementation for first-order theorem proving, a domain both responsive to a very simple nagging model and amenable to many refinements of this model. Nagging’s scalability and intrinsic fault tolerance make it particularly suitable for application in commonly available, low-bandwidth, high-latency distributed environments. We present several nagging models of increasing sophistication, demonstrate their effectiveness empirically, and compare nagging with related work in parallel search.

Objective This paper constructs an authorship-linked collection or corpus of anonymous, sex-seeking ads found on the classifieds website Craigslist. This corpus is then used to validate an authorship attribution approach based on identifying near duplicate text in ad clusters, providing insight into how often anonymous individuals post sex-seeking ads and where they meet for encounters. Introduction The increasing use of the Internet to arrange sexual encounters presents challenges to public health agencies formulating STD interventions, particularly in the context of anonymous encounters. These encounters complicate or break traditional interventions. In previous work [1], we examined a corpus of anonymous personal ads seeking sexual encounters from the classifieds website Craigslist and presented a way of linking multiple ads posted across time to a single author. The key observation of our approach is that some ads are simply reposts of older ads, often updated with only minor textual changes. Under the presumption that these ads, when not spam, originate from the same author, we can use efficient near-duplicate detection techniques to cluster ads within some threshold similarity. Linking ads in this way allows us to preserve the anonymity of authors while still extracting useful information on the frequency with which authors post ads, as well as the geographic regions in which they seek encounters. While this process detects many clusters, the lack of a true corpus of authorship-linked ads makes it difficult to validate and tune the parameters of our system. Fortunately, many ad authors provide an obfuscated telephone number in ad text (e.g., 867–5309 becomes 8sixseven5three oh nine) to bypass Craigslist filters, which prohibit including phone numbers in personal ads. By matching phone numbers of this type across all ads, we can create a corpus of ad clusters known to be written by a single author. This authorship corpus can then be used to evaluate and tune our existing near-duplicate detection system, and in the future identify features for more robust authorship attribution techniques. Methods From 7-1-2009 until 7-1-2011, RSS feeds were collected daily for 8 personal ad categories from 414 sites across the United States, for a total of 67 million ads. To create an anonymous, author-linked corpus, we used a regular expression to identify obfuscated phone numbers in ad text. We measure the ability of near-duplicate detection to link clusters in two ways: 1) detecting all ads in a cluster; and 2) correctly detecting a subset of ads within a single cluster. Ads incorrectly assigned to more than 1 cluster are considered false positives. All results are reported in terms of precision, recall, and F-scores (common information retrieval metrics) across cluster size, expressed as number of ads. Results 652,014 ads contained phone numbers, producing a total of 46,079 authorship-linked ad clusters. For detecting all ads within a cluster, precision ranged from 0.05 to 0.0 and recall from 0.02 to 0.0 for all cluster sizes. For detecting partial clusters, see Figure 1. Conclusions We find that near-duplicate detection alone is insufficient to detect all ads within a cluster. However, we do find that the process can, with high precision and low recall, detect a subset of ads associated with a single author. This follows the intuition that an author’s total set of ads is itself comprised of multiple self-similar subsets. While a near-duplicate detection approach can correctly identify subsets of ads linked to a single author, this process alone cannot attribute multiple clusters to a single author. Future work will explore leveraging additional linguistic features to improve author attribution. (Top) Evaluations for partial cluster detection using the near-duplicate identification approach to linking anonymous authorship in Craigslist ads and (bottom) the distribution of ad cluster sizes.

We performed whole-exome sequencing of a family with autosomal dominant Dandy-Walker malformation and occipital cephaloceles (ADDWOC) and detected a mutation in the extracellular matrix protein encoding gene NID1. In a second family, protein interaction network analysis identified a mutation in LAMC1, which encodes a NID1 binding partner. Structural modeling the NID1-LAMC1 complex demonstrated that each mutation disrupts the interaction. These findings implicate the extracellular matrix in the pathogenesis of Dandy-Walker spectrum disorders.

Background. Super-spreading events, in which an individual with measurably high connectivity is responsible for infecting a large number of people, have been observed. Our goal is to determine the impact of hand hygiene noncompliance among peripatetic (eg, highly mobile or highly connected) healthcare workers compared with less-connected workers. Methods. We used a mote-based sensor network to record contacts among healthcare workers and patients in a 20-bed intensive care unit. The data collected from this network form the basis for an agent-based simulation to model the spread of nosocomial pathogens with various transmission probabilities. We identified the most- and least-connected healthcare workers. We then compared the effects of hand hygiene noncompliance as a function of connectedness. Results. The data confirm the presence of peripatetic healthcare workers. Also, agent-based simulations using our real contact network data confirm that the average number of infected patients was significantly higher when the most connected healthcare worker did not practice hand hygiene and significantly lower when the least connected healthcare workers were noncompliant. Conclusions. Heterogeneity in healthcare worker contact patterns dramatically affects disease diffusion. Our findings should inform future infection control interventions and encourage the application of social network analysis to study disease transmission in healthcare settings.

This paper describes a method of using Craig list personal ads to better understand the movement behavior of anonymous, casual sex-seeking individuals within the men-who-have-sex-with-men community. Given recent dramatic increases in HIV and sexually transmitted disease within this community, gaining insight into how sexual networks connect neighborhoods and cities is important for formulating public health interventions. Due to the high degree of similarity exhibited by subsets of Craig list ads, and the presumption that a set of near-identical ads, when not spam, originate from the same author, we can apply techniques for efficient near-duplicate detection to identify clusters of near-identical ads. By examining each of these clusters and identifying differences in user-supplied location tags, we can then reconstruct an approximation of an anonymous individual's movement footprint over time, as well as estimate the rate at which ad authors seek sexual encounters. For the state of California, we find that 86% of all encounter requests for a given set occur within a 50 mile area, with only less that 4% of messages reflecting long-distance travel over 250 miles. 60% of all detected clusters reposted ads within 2 weeks of the first detected post. We show that even in the relatively noisy, unstructured data environment of anonymous personal ads, it is still possible to extract meaningful signal and identify useful social network properties for analysis.

Disease spread in hospital settings is a common and important problem in health care. Knowing the network of contacts between health care workers and patients can be very helpful in mitigating disease spread. In this work, we address the problem of inferring the contact network of health care workers at the University of Iowa Hospital and Clinics facilities by integrating two sources of data: hospital-wide computer login data and proximity data obtained from direct measurement in the Medical Intensive Care Unit using a wireless sensor network. We treat this problem as a variant of the network completion problem, where one small portion of the network is well known while the rest is sparingly sampled, and we want to complete the network. In this case, we want to transform the login network, where an edge connects two people who logged into computers within some time and distance, of the hospital into a contact network. We solve this problem by borrowing techniques from link prediction. We train and evaluate these techniques on synthetic login networks and contact networks obtained from the sensor data. Our results are promising in that we can predict contact networks from login networks with accuracies mostly between 70% and 90%.

We present a comprehensive approach to using electronic medical records (EMR) for constructing contact networks of healthcare workers in a hospital. This approach is applied at the University of Iowa Hospitals and Clinics (UIHC) – a 3.2 million square foot facility with 700 beds and about 8,000 healthcare workers – by obtaining 19.8 million EMR data points, spread over more than 21 months. We use these data to construct 9,000 different healthcare worker contact networks, which serve as proxies for patterns of actual healthcare worker contacts. Unlike earlier approaches, our methods are based on large-scale data and do not make any a priori assumptions about edges (contacts) between healthcare workers, degree distributions of healthcare workers, their assignment to wards, etc. Preliminary validation using data gathered from a 10-day long deployment of a wireless sensor network in the Medical Intensive Care Unit suggests that EMR logins can serve as realistic proxies for hospital-wide healthcare worker movement and contact patterns. Despite spatial and job-related constraints on healthcare worker movement and interactions, analysis reveals a strong structural similarity between the healthcare worker contact networks we generate and social networks that arise in other (e.g., online) settings. Furthermore, our analysis shows that disease can spread much more rapidly within the constructed contact networks as compared to random networks of similar size and density. Using the generated contact networks, we evaluate several alternate vaccination policies and conclude that a simple policy that vaccinates the most mobile healthcare workers first, is robust and quite effective relative to a random vaccination policy.

Objective.To explore how hand hygiene observer scheduling influences the number of events and unique individuals observed.Design.We deployed a mobile sensor network to capture detailed movement data for 6 categories of healthcare workers over a 2-week period.Setting.University of Iowa Hospital and Clinic medical intensive care unit (ICU).Methods.We recorded 33,721 time-stamped healthcare worker entries to and exits from patient rooms and considered each entry or exit to be an opportunity for hand hygiene. Architectural drawings were used to derive 4 optimal line-of-sight placements for observers. We ran simulations for different observer movement schedules, all with a budget of 1 hour of total observation time. We considered observation times of 1–15, 15–30, 30, and 60 minutes per station. We stochastically generated healthcare worker hand hygiene compliance on the basis of all data and recorded the total unit compliance as it would be reported by each simulated observer.Results.Considering a 60-minute total observation period, aggregate simulated observers captured 1.7% of the average total number of opportunities per day at best and 0.5% at worst. The 1–15-minute schedule captures, on average, 16% fewer events than does the 60-minute (ie, static) schedule, but it samples 17% more unique individuals. The 1–15-minute schedule also provides the best estimator of compliance for the duration of the shift, with a mean standard deviation of 17%, compared with 23% for the 60-minute schedule.Conclusions.Our results show that observations are sensitive to different observers' schedules and suggest the importance of using data-driven approaches to schedule hand hygiene audits.

We describe the design, implementation and evaluation of a novel multi-agent scenario generation framework for interactive virtual reality simulations towards preventive medicine education. Our scenario generation framework is based on recordings of human movements from a distributed sensor networks deployed in a real-world physical setting. The components of our framework include the generation of unique virtual agent behaviors from the sensor data, and algorithms for the generation of low level or gross movement behaviors such as path determination, directional traffic flows, collision avoidance and overtaking. The framework also includes the generation of high level fine actions for multi-agents such as techniques for interactive activities in pedagogical scenarios based on environment and temporal triggers. We applied our multi-agent scenario generation framework in an interactive simulation for hand hygiene education, and conduct an initial usability study to assess the educational benefits of the simulation to nursing students and evaluated the performance characteristics of our framework. Results of our quantitative and qualitative evaluations suggest that our framework was robust in creating engaging, compelling, and realistic interactive training scenarios with multiple virtual agents in simulated hospital situations.

The main contribution of this paper is a new domain-independent explanation-based learning (EBL) algorithm. The new EBL∗DI algorithm significantly outperforms traditional EBL algorithms both by learning in situations where traditional algorithms cannot learn as well as by providing greater problem-solving performance improvement in general. The superiority of the EBL∗DI algorithm is demonstrated with experiments in three different application domains. The EBL∗DI algorithm is developed using a novel formal framework in which traditional EBL techniques are reconstructed as the structured application of three explanation-transformation operators. We extend this basic framework by introducing two additional operators that, when combined with the first three operators, allow us to prove a completeness result: in the formal framework, every EBL algorithm is equivalent to the application of the five transformation operators according to some control strategy. The EBL∗DI algorithm employs all five proof-transformation operators guided by five domain-independent control heuristics.

Theory revision is the process of making corrections to a flawed or incomplete knowledge base on the basis of examples that expose those problems. The PTR algorithm is a theory revision algorithm that makes use of explicit bias to guide the detection of flawed knowledge base elements. In this paper, we examine the effectiveness of PTR's bias scheme in identifying flawed knowledge base elements, and we propose extensions to the PTR algorithm that support the use of additional bias to guide the process of correcting a flawed element once it has been located.

This paper describes the software package KELVIN, which supports the PPL (posterior probability of linkage) framework for the measurement of statistical evidence in human (or more generally, diploid) genetic studies. In terms of scope, KELVIN supports two-point (trait-marker or marker-marker) and multipoint linkage analysis, based on either sex-averaged or sex-specific genetic maps, with an option to allow for imprinting; trait-marker linkage disequilibrium (LD), or association analysis, in case-control data, trio data, and/or multiplex family data, with options for joint linkage and trait-marker LD or conditional LD given linkage; dichotomous trait, quantitative trait and quantitative trait threshold models; and certain types of gene-gene interactions and covariate effects. Features and data (pedigree) structures can be freely mixed and matched within analyses. The statistical framework is specifically tailored to accumulate evidence in a mathematically rigorous way across multiple data sets or data subsets while allowing for multiple sources of heterogeneity, and KELVIN itself utilizes sophisticated software engineering to provide a powerful and robust platform for studying the genetics of complex disorders.

International Society for Disease Surveillance Annual Conference 2011(ISDS) 1220673 Changes in alcohol-related travel behavior due to an under 21 ordinance the drinking culture. Of course, since the arresting officer cannot generally detect residency prior to arrest, arrest data still represent a geographically unbiased sample of bar patrons and can be used to explore changes in the mix of patrons. We hypothesize that the changes detected in the proportion of arrests of in town and out of town patrons reflect a more homogeneous student clientele, where town-gown tensions are less likely to arise. Of course, any reduction in out of town patrons also corresponds to a reduction in the risk of DUI-related fatalities, since students walk to the bars. There are several shortcomings to this study. First, our data are incomplete as data from other enforcement agencies was not available. Second, we were unable to directly confirm the link with violence or sexual assault, as additional data would be required to do so: these are our next steps. Alcohol abuse is one of the major leading causes of preventable mortality in the United States (1). Binge drinking or excessive alcohol consumption, categorized as a pattern of drinking that brings a person’s blood alcohol concentration (BAC) to 0.08 (2), has become a major cause for concern, especially in the 18to 20 year-old population. Iowa City is home to the University of Iowa, a large public university of 30,000 students. On June 1, 2010, the city council enacted a new ordinance prohibiting persons under 21 from entering or remaining in bars (establishments after 10:00 PM whose primary purpose is the sale of alcoholic beverages) after 10:00 PM (3). Prior to the ordinance, Iowa City was the only municipality in the region where underage patrons were allowed on premises. The new ordinance was enacted largely in response to public safety concerns, Arrest records were obtained from the University of Iowa Police Department containing all alcohol-related citations from January 1, 2004 to June 26, 2011. As the University of Iowa Police Department is one of the 4 local law enforcement agencies (Iowa City Police, Coralville Police and Johnson County Sheriff), these 7002 records represent a sample of alcohol-related arrests, albeit one focused on the downtown bar area frequented by college students. Each record contains the date of the arrest as well as the age and home address of the offender, allowing us to compare ‘in town’ offenders (i.e., from within Iowa City, Coralville and transients) with ‘out of town’ offenders. Records corresponding to football Saturdays, where some 50,000 people come to Iowa City to tailgate and attend the Big10 football game, were excluded from the analysis as not representative of the usual bar culture. A total of 1490 alcohol-related Background To study alcohol-related arrests during the time surrounding the introduction of an alcohol-related ordinance in the Iowa City, IA, area.

Introduction ILI data are collected via an Influenza Sentinel Provider Surveillance Network at the state level. Because participation is voluntary, locations of the sentinel providers may not reflect optimal geographic placement. This study analyzes two different geographic placement schemes*a maximal coverage model (MCM) and a K-median model, two location-allocation models commonly used in geographic information systems (GIS) (1). The MCM chooses sites in areas with the densest population. The K-median model chooses sites, which minimize the average distance traveled by individuals to their nearest site. We have previously shown how a placement model can be used to improve population coverage for ILI surveillance in Iowa when considering the sites recruited by the Iowa Department of Public Health (IDPH) (2). We extend this work by evaluating different surveillance placement algorithms with respect to outbreak intensity and timing (i.e., being able to capture the start, peak and end of the influenza season).

Introduction Public health officials and epidemiologists have been attempting to eradicate syphilis for decades, but national incidence rates are again on the rise. It has been suggested that the syphilis epidemic in the United States is a ‘rare example of unforced, endogenous oscillations in disease incidence, with an 8 11-year period that is predicted by the natural dynamics of syphilis infection, to which there is partially protective immunity’ (1). While the time series of aggregate case counts seems to support this claim, between 1990 and 2010, there seems to have been a significant change in the spatial distribution of the syphilis epidemic. It is unclear if this change can also be attributed to ‘endogenous’ factors or whether it is due to exogenous factors such as behavioral changes (e.g., the widespread use of the internet for anonymous sexual encounters). For example, it is pointed out that levels of syphilis in 1989 were abnormally high in counties in North Carolina (NC) immediately adjacent to highways (2). The hypothesis was that this may be due to truck drivers and prostitution and/or the emerging cocaine market (1). Our results indicate that syphilis distribution in NC has changed since 1989, diffusing away from highway counties (see Fig. 1).

Introduction Emerging event detection is the process of automatically identifying novel and emerging ideas from text with minimal human intervention. With the rise of social networks like Twitter, topic detection has begun leveraging measures of user influence to identify emerging events. Twitter’s highly skewed follower/followee structure lends itself to an intuitive model of influence; yet, in a context like the Emerging Infections Network (EIN), a sentinel surveillance listserv of over 1400 infectious disease experts, developing a useful model of authority becomes less clear. Who should we listen to on the EIN? To explore this, we annotated a body of important EIN discussions and tested how well 3 models of user authority performed in identifying those discussions. In previous work, we proposed a process by which only posts that are based on specific ‘important’ topics are read, thus drastically reducing the amount of posts that need to be read. The process works by finding a set of ‘bellwether’ users that act as indicators for ‘important’ topics and only posts relating to these topics are then read. This approach does not consider the text of messages, only the patterns of user participation. Our text analysis approach follows that of Cataldi et al. (1), using the idea of semantic ‘energy’ to identify emerging topics within Twitter posts. Authority is calculated via PageRank and used to weight each author’s contribution to the semantic energy of all terms occurring in within some interval ti. A decay parameter d defines the impact of prior time steps on the current interval.

Introduction The spread of infectious diseases is facilitated by human travel. Disease is often introduced by travelers and then spread among susceptible individuals. Likewise, uninfected susceptible travelers can move into populations sustaining the spread of an infectious disease. Several disease-modeling efforts have incorporated travel and census data in an effort to better understand the spread of disease. Unfortunately, most travel data are not fine grained enough to capture individual movements over long periods and large spaces. Alternative methods (e.g., tracking currency movements or cell phone signals) have been suggested to measure how people move with higher resolution but these are often sparse, expensive and not readily available to researchers. FourSquare is a social media application that permits users to ‘check-in’ (i.e., record their currentlocation at stores, restaurants, etc.) via their mobile telephones in exchange for incentives (e.g., location-specific coupons). FourSquare and similar applications (Gowalla, Yelp, etc.) generally broadcast each check-in via Twitter or Facebook; in addition, some GPS-enabled mobile Twitter clients add explicit geocodes to individual tweets. Here, we propose the use of geocoded social media data as a real-time fine-grained proxy for human travel.

Objective.To assess healthcare personnel (HCP) perceptions regarding implementation of sensor-based electronic systems for automated hand hygiene adherence monitoring.Design.Using a mixed-methods approach, structured focus groups were designed to elicit quantitative and qualitative responses on familiarity, comfort level, and perceived impact of sensor-based hand hygiene adherence monitoringSetting.A university hospital, a Veterans Affairs hospital, and a community hospital in the Midwest.Participants.Focus groups were homogenous by HCP type, with separate groups held for leadership, midlevel management, and frontline personnel at each hospital.Results.Overall, 89 HCP participated in 10 focus groups. Levels of familiarity and comfort with electronic oversight technology varied by HCP type; when compared with frontline HCP, those in leadership positions were significantly more familiar with (PPConclusions.While substantial concerns were expressed by all types of HCP, participants' recommendations for effective implementation of electronic oversight technologies for hand hygiene monitoring included addressing accuracy issues before implementation and transparent communication with frontline HCP about the intended use of the data.

Previous researchers have modelled the decision to accept a donor organ for transplantation as a Markov decision problem, the solution to which is often a control-limit optimal policy: accept any organ whose match quality exceeds some health-dependent threshold; otherwise, wait for another. When competing transplant centers vie for the same organs, the decision rule changes relative to no competition; the relative size of competing centers affects the decision rules as well. Using center-specific graft and patient survival-rate data for cadaveric adult livers in the United States, we have found empirical evidence supporting these predictions.

In this paper we describe the design of an effective caching mechanism for resource-limited, definite-clause theorem-proving systems. Previous work in adapting caches for theorem proving relies on the use of unlimited-size caches. We show how unlimited-size caches are unsuitable in application contexts where resource-limited theorem provers are used to solve multiple problems from a single problem distribution. We introduce bounded-overhead caches, that is, those caches that contain at most a fixed number of entries and entail a fixed amount of overhead per lookup, and we examine cache design issues for bounded-overhead caches. Finally, we present an empirical evaluation of bounded-overhead cache performance, relying on a specially designed experimental methodology that separates hardware-dependent, implementation-dependent, and domain-dependent effects.

This paper describes a spatial model for healthcare workers' location in a large hospital facility. Such models have many applications in healthcare, such as supporting time-and- motion efficiency studies to improve healthcare delivery, or modeling the spread of hospital-acquired infections. We use our model to estimate spatial distributions for healthcare workers in The University of Iowa Hospitals and Clinics (UIHC), a 700-bed comprehensive academic medical center spanning a total of 3.2 million square feet and employing about 8,000 healthcare workers. We model the UIHC as a metric space induced by walking distance between pairs of rooms, and with each room having a level of attractiveness representing the activity level in that room. We combine this with a model in which each healthcare worker has a center of activity and a probability density function that decays polynomially as we move away from the center. Using 12 million Electronic Medical Record (EMR)logins collected over 22 months, we solve for the model parameters for each room and each healthcare worker using heuristic techniques to make the problem computationally tractable. We then validate the model parameters obtained by comparing realworld expectations of healthcare worker behavior for several job categories to our model predictions (e.g., we verify that Unit Clerks are much more stationary than Respiratory Therapists). Finally we present solutions to two important applications. First, we use healthcare worker spatial distributions to generate random walks representing their movement through the hospital. We use these random walks to This paper describes a spatial model for healthcare workers' location in a large hospital facility. Such models have many applications in healthcare, such as supporting timeand- motion efficiency studies to improve healthcare delivery, or modeling the spread of hospital-acquired infections. We use our model to estimate spatial distributions for healthcare workers in The University of Iowa Hospitals and Clinics (UIHC), a 700-bed comprehensive academic medical center spanning a total of 3.2 million square feet and employing about 8,000 healthcare workers. We model the UIHC as a metric space induced by walking distance between pairs of rooms, and with each room having a level of attractiveness representing the activity level in that room. We combine this with a model in which each healthcare worker has a center of activity and a probability density function that decays polynomially as we move away from the center. Using 12 million Electronic Medical Record (EMR) logins collected over 22 months, we solve for the model parameters for each room and each healthcare worker using heuristic techniques to make the problem computationally tractable. We then validate the model parameters obtained by comparing realworld expectations of healthcare worker behavior for several job categories to our model predictions (e.g., we verify that Unit Clerks are much more stationary than Respiratory Therapists). Finally we present solutions to two important applications. First, we use healthcare worker spatial distributions to generate random walks representing their movement through the hospital. We use these random walks to access that resource.

Many efforts to automatically measure hand hygiene activity depend on radio-frequency identification equipment or similar technology that can be expensive to install. We have developed a method for automatically tracking the use of hand hygiene dispensers before healthcare workers enter (or after they exit) patient rooms that is easily and quickly deployed without permanent hardware.

Twitter is a free social networking and micro-blogging service that enables its millions of users to send and read each other's “tweets,” or short, 140-character messages. The service has more than 190 million registered users and processes about 55 million tweets per day. Useful information about news and geopolitical events lies embedded in the Twitter stream, which embodies, in the aggregate, Twitter users' perspectives and reactions to current events. By virtue of sheer volume, content embedded in the Twitter stream may be useful for tracking or even forecasting behavior if it can be extracted in an efficient manner. In this study, we examine the use of information embedded in the Twitter stream to (1) track rapidly-evolving public sentiment with respect to H1N1 or swine flu, and (2) track and measure actual disease activity. We also show that Twitter can be used as a measure of public interest or concern about health-related events. Our results show that estimates of influenza-like illness derived from Twitter chatter accurately track reported disease levels.

The goal of our work is the design and implementation of a virtual agents based interactive simulation for teaching and training healthcare workers in hand hygiene protocols. The interactive training simulation features a virtual instructor who teaches the trainee the Five Moments of hand hygiene, recommended by the Centers of Disease Control and the World Health Organization, via instructions and demonstrations in a tutorial phase. In an interactive training phase, a virtual health care worker demonstrates by interacting with a virtual patient and the patient's environment in ten randomly generated virtual scenarios. After watching each scenario the trainee evaluates if the virtual healthcare worker's actions are in accordance with the Five Moments of Hand Hygiene procedure. In a feedback phase, the trainee receives feedback on their performance in the training phase after which the trainee can either exit or return to any phase of the interactive simulation. We describe the design and development of the hospital environment, simulated virtual instructor, health care worker and patient, and the interactive simulation components towards teaching and training in healthcare best practices associated with hand hygiene.

Influenza-like illness data are collected via an Influenza Sentinel Provider Surveillance Network at the state level. Because participation is voluntary, locations of the sentinel providers may not reflect optimal geographic placement. The purpose of this study was to determine the "best" locations for sentinel providers in Iowa by using a maximal coverage model (MCM) and to compare the population coverage obtained with that of the current sentinel network. The authors used an MCM to maximize the Iowa population located within 20 miles (32.2 km) of 1-143 candidate sites and calculated the coverage provided by each additional site. The first MCM location covered 15% of the population; adding a second increased coverage to 25%. Additional locations provided more coverage but with diminishing marginal returns. In contrast, the existing 22 Iowa sentinel locations covered 56% of the population, the same coverage achieved with just 10 MCM sites. Using 22 MCM sites covered more than 75% of the population, an improvement over the current site placement, adding nearly 600,000 Iowa residents. Given scarce public health resources, MCMs can help surveillance efforts by prioritizing recruitment of sentinel locations.

Nosocomial (i.e., hospital-acquired) infections are a major cause of morbidity and mortality in the United States and throughout the world. Therefore, understanding, mediating, and limiting contagious infections are important problems, even in clinical settings. Contact networks of healthcare workers and patients provides a vehicle for modeling the spread of infection, enabling analytical and simulation-based studies. The contact network models are based on geographic maps of hospitals and records of social contact between health care workers and patients. Wireless technology can help in both, geolocating healthcare workers and patients in a hospital and in capturing a record of physical proximity among these agents. As a step towards this goal we have implemented a low-cost wireless system to instrument hand-hygiene events; this system can track the use of hand hygiene dispensers before healthcare workers enter or exit patient rooms.

We study the public goods problem associated with vaccinations. The externality created by an infection is composed of two parts, the probability of infection and the marginal infections generated if infected. We argue that the key component in a successful vaccination strategy is the second of these items but that current public policy focuses on the first. We use a newly collected data set coupled with agent-based simulations to study the spread of influenza and other infectious diseases in hospitals. We estimate the marginal infections created by various worker groups in a hospital in order to prioritize vaccine allocations across different healthcare worker groups in times of vaccine shortages. One primary focus of this paper is identifying the individual hospital workers who are most important to vaccinate. Surprisingly, we find that many groups with patient care responsibilities, such as physicians, play a small role in spreading influenza while others, such as unit clerks, play a much larger role.

A combinatorial auction (CA) is an auction that permits bidders to bid on bundles of goods rather than just a single item. Unfortunately, winner determination for CAs is known to be NP-hard. In this paper, we propose a distributed algorithm to compute optimal solutions to this problem. The algorithm uses nagging, a technique for parallelizing search in heterogeneous distributed computing environments. Here, we show how nagging can be used to parallelize a branch-and-bound algorithm for this problem, and provide empirical results supporting both the performance advantage of nagging over more traditional partitioning methods as well as the superior scalability of nagging to larger numbers of processors more traditional partitioning methods as well as the superior scalability of nagging to larger numbers of processors

The construction of intelligent autonomous systems is a core goal of artificial intelligence. Research in machine learning yields some insight in how autonomous systems could learn to plan more effectively. This paper describes an application of explanation-based learning, or EBL, to a robot manufacturing domain. The ARMS system (for Acquiring Robotic Manufacturing Schemata) learns how to assemble new mechanisms by observing and analyzing another agent's solution to sample assembly problems. While the ARMS system's major contribution is in the area of machine learning, the lessons learned from this work are readily applicable to autonomous agent research.

A number of experimental evaluations of explanation-based learning (EBL) have been reported in the literature on machine learning. A close examination of the design of these experiments reveals certain methodological problems that could affect the conclusions drawn from the experiments. This article analyzes some of the more common methodological difficulties, and illustrates them using selected previous studies.