Predicting malnutrition from longitudinal patient trajectories with deep learning.

Malnutrition is common, morbid, and often correctable, but subject to missed and delayed diagnosis. Better screening and prediction could improve clinical, functional, and economic outcomes. This study aimed to assess the predictability of malnutrition from longitudinal patient records, and the external generalizability of a predictive model. Predictive models were developed and validated on statewide emergency department (ED) and hospital admission databases for California, Florida and New York, including visits from October 1, 2015 to December 31, 2018. Visit features included patient demographics, diagnosis codes, and procedure categories. Models included long short-term memory (LSTM) recurrent neural networks trained on longitudinal trajectories, and gradient-boosted tree and logistic regression models trained on cross-sectional patient data. The dataset used for model training and internal validation (California and Florida) included 62,811 patient trajectories (266,951 visits). Test sets included 63,997 (California), 63,112 (Florida), and 62,472 (New York) trajectories, such that each cohort's composition was proportional to the prevalence of malnutrition in that state. Trajectories contained seven patient characteristics and up to 2,008 diagnosis categories. Area under the receiver-operating characteristic (AUROC) and precision-recall curves (AUPRC) were used to characterize prediction of first malnutrition diagnoses in the test sets. Data analysis was performed from September 2020 to May 2021. Between 4.0% (New York) and 6.2% (California) of patients received malnutrition diagnoses. The longitudinal LSTM model produced the most accurate predictions of malnutrition, with comparable predictive performance in California (AUROC 0.854, AUPRC 0.258), Florida (AUROC 0.869, AUPRC 0.234), and New York (AUROC 0.869, AUPRC 0.190). Deep learning models can reliably predict malnutrition from existing longitudinal patient records, with better predictive performance and lower data-collection requirements than existing instruments. This approach may facilitate early nutritional intervention via automated screening at the point of care.

A body bag can save your life: a novel method of cold water immersion for heat stroke treatment.

Non-exertional heat stroke is a life-threatening condition characterized by passive exposure to high ambient heat, a core body temperature of 40°C (104°F) or greater, and central nervous system dysfunction. Rapid cooling is imperative to minimize mortality and morbidity. Although evaporative and convective measures are often used for cooling heat stroke patients, cold water immersion produces the fastest cooling. However, logistical difficulties make cold water immersion challenging to implement in the emergency department. To our knowledge, there is no documented case utilizing a body bag (ie, human remains pouch) as a cold water immersion tank for rapid resuscitation of heat stroke. During a regional heat wave an elderly woman was found unconscious in a parking lot with an oral temperature of 40°C (104°F) and altered mental status. She was cooled to 38.4°C (101.1°F) in 10 minutes by immersion in an ice- and water-filled body bag. The patient rapidly regained normal mentation and was discharged home from the ED. This case highlights a novel method for efficient and convenient cold water immersion for heat stroke treatment in the emergency department.

Social Determinants of Hallway Bed Use.

INTRODUCTION: Hallway beds in the emergency department (ED) produce lower patient satisfaction and inferior care. We sought to determine whether socioeconomic factors influence which visits are assigned to hallway beds, independent of clinical characteristics at triage. METHODS: We studied 332,919 visits, across 189,326 patients, to two academic EDs from 2013-2016. We estimated a logistic model of hallway bed assignment, conditioning on payor, demographics, triage acuity, chief complaint, patient visit frequency, and ED volume. Because payor is not generally known at the time of triage, we interpreted it as a proxy for other observable characteristics that may influence bed assignment. We estimated a Cox proportional hazards model of hallway bed assignment on length of stay. RESULTS: Median patient age was 53. 54.0% of visits were by women. 42.1% of visits were paid primarily by private payors, 37.1% by Medicare, and 20.7% by Medicaid. A total of 16.2% of visits were assigned to hallway beds. Hallway bed assignment was more likely for frequent ED visitors, for lower acuity presentations, and for psychiatric, substance use, and musculoskeletal chief complaints, which were more common among visits paid primarily by Medicaid. In a logistic model controlling for these factors, as well as for other patient demographics and for the volume of recent ED arrivals, Medicaid status was nevertheless associated with 22% greater odds of assignment to a hallway bed (odds ratio 1.22, [95% confidence interval, CI, 1.18-1.26]), compared to private insurance. Visits assigned to hallway beds had longer lengths of stay than roomed visits of comparable acuity (hazard ratio for departure 0.91 [95% CI, 0.90-0.92]). CONCLUSION: We find evidence of social determinants of hallway bed use, likely involving epidemiologic, clinical, and operational factors. Even after accounting for different distributions of chief complaints and for more frequent ED use by the Medicaid population, as well as for other visit characteristics known at the time of triage, visits paid primarily by Medicaid retain a disproportionate association with hallway bed assignment. Further research is needed to eliminate potential bias in the use of hallway beds. [West J Emerg Med. 2020;21(4)949-958.].

Exposure, hazard, and survival analysis of diffusion on social networks.

Sociologists, economists, epidemiologists, and others recognize the importance of social networks in the diffusion of ideas and behaviors through human societies. To measure the flow of information on real-world networks, researchers often conduct comprehensive sociometric mapping of social links between individuals and then follow the spread of an "innovation" from reports of adoption or change in behavior over time. The innovation is introduced to a small number of individuals who may also be encouraged to spread it to their network contacts. In conjunction with the known social network, the pattern of adoptions gives researchers insight into the spread of the innovation in the population and factors associated with successful diffusion. Researchers have used widely varying statistical tools to estimate these quantities, and there is disagreement about how to analyze diffusion on fully observed networks. Here, we describe a framework for measuring features of diffusion processes on social networks using the epidemiological concepts of exposure and competing risks. Given a realization of a diffusion process on a fully observed network, we show that classical survival regression models can be adapted to estimate the rate of diffusion, and actor/edge attributes associated with successful transmission or adoption, while accounting for the topology of the social network. We illustrate these tools by applying them to a randomized network intervention trial conducted in Honduras to estimate the rate of adoption of 2 health-related interventions-multivitamins and chlorine bleach for water purification-and determine factors associated with successful social transmission.

Social connectedness is associated with fibrinogen level in a human social network.

Socially isolated individuals face elevated rates of illness and death. Conventional measures of social connectedness reflect an individual's perceived network and can be subject to bias and variation in reporting. In this study of a large human social network, we find that greater indegree, a sociocentric measure of friendship and familial ties identified by a subject's social connections rather than by the subject, predicts significantly lower concentrations of fibrinogen (a biomarker of inflammation and cardiac risk), after adjusting for demographics, education, medical history and known predictors of cardiac risk. The association between fibrinogen and social isolation, as measured by low indegree, is comparable to the effect of smoking, and greater than that of low education, a conventional measure of socioeconomic disadvantage. By contrast, outdegree, which reflects an individual's perceived connectedness, displays a significantly weaker association with fibrinogen concentrations.

Social network targeting to maximise population behaviour change: a cluster randomised controlled trial.

BACKGROUND: Information and behaviour can spread through interpersonal ties. By targeting influential individuals, health interventions that harness the distributive properties of social networks could be made more effective and efficient than those that do not. Our aim was to assess which targeting methods produce the greatest cascades or spillover effects and hence maximise population-level behaviour change. METHODS: In this cluster randomised trial, participants were recruited from villages of the Department of Lempira, Honduras. We blocked villages on the basis of network size, socioeconomic status, and baseline rates of water purification, for delivery of two public health interventions: chlorine for water purification and multivitamins for micronutrient deficiencies. We then randomised villages, separately for each intervention, to one of three targeting methods, introducing the interventions to 5% samples composed of either: randomly selected villagers (n=9 villages for each intervention); villagers with the most social ties (n=9); or nominated friends of random villagers (n=9; the last strategy exploiting the so-called friendship paradox of social networks). Participants and data collectors were not aware of the targeting methods. Primary endpoints were the proportions of available products redeemed by the entire population under each targeting method. This trial is registered with ClinicalTrials.gov, number NCT01672580. FINDINGS: Between Aug 4, and Aug 14, 2012, 32 villages in rural Honduras (25-541 participants each; total study population of 5773) received public health interventions. For each intervention, nine villages (each with 1-20 initial target individuals) were randomised, using a blocked design, to each of the three targeting methods. In nomination-targeted villages, 951 (74·3%) of 1280 available multivitamin tickets were redeemed compared with 940 (66·2%) of 1420 in randomly targeted villages and 744 (61·0%) of 1220 in indegree-targeted villages. All pairwise differences in redemption rates were significant (p<0·01) after correction for multiple comparisons. Targeting nominated friends increased adoption of the nutritional intervention by 12·2% compared with random targeting (95% CI 6·9-17·9). Targeting the most highly connected individuals, by contrast, produced no greater adoption of either intervention, compared with random targeting. INTERPRETATION: Introduction of a health intervention to the nominated friends of random individuals can enhance that intervention's diffusion by exploiting intrinsic properties of human social networks. This method has the additional advantage of scalability because it can be implemented without mapping the network. Deployment of certain types of health interventions via network targeting, without increasing the number of individuals targeted or the resources used, could enhance the adoption and efficiency of those interventions, thereby improving population health. FUNDING: National Institutes of Health, The Bill & Melinda Gates Foundation, Star Family Foundation, and the Canadian Institutes of Health Research.