Search

Objectives.To evaluate the positive predictive value of machine learning algorithms for early assessment of adverse birth risk among pregnant women as a means of improving the allocation of social services. Methods. We used administrative data for 6457 women collected by the Illinois Department of Human Services from July 2014 to May 2015 to develop a machine learning model for adverse birth prediction and improve upon the existing paper-based risk assessment. We compared different models and determined the strongest predictors of adverse birth outcomes using positive predictive value as the metric for selection. Results. Machine learning algorithms performed similarly, outperforming the current paper-based risk assessment by up to 36%; a refined paper-based assessment outperformed the current assessment by up to 22%. We estimate that these improvements will allow 100 to 170 additional high-risk pregnant women screened for program eligibility each year to receive services that would have otherwise been unobtainable. Conclusions. Our analysis exhibits the potential for machine learning to move government agencies toward a more data-informed approach to evaluating risk and providing social services. Overall, such efforts will improve the efficiency of allocating resource-intensive interventions. [ABSTRACT FROM AUTHOR]

Abstract: This paper provides an overview of problems in multivariate modeling of epidemiologic data, and examines some proposed solutions. Special attention is given to the task of model selection, which involves selection of the model form, selection of the variables to enter the model, and selection of the form of these variables in the model. Several conclusions are drawn, among them: a) model and variable forms should be selected based on regression diagnostic procedures, in addition to goodness-of-fit tests; b) variable-selection algorithms in current packaged programs, such as conventional stepwise regression, can easily lead to invalid estimates and tests of effect; and c) variable selection is better approached by direct estimation of the ° of confounding produced by each variable than by significance-testing algorithms. As a general rule. before using a model to estimate effects, one should evaluate the assumptions implied by the model against both the data and prior information. [ABSTRACT FROM AUTHOR]

Objectives. We evaluated the public health benefits of traffic laws targeting speeding and drunk drivers (British Columbia, Canada, September 2010). Methods. We studied fatal crashes and ambulance dispatches and hospital admissions for road trauma, using interrupted time series with multiple nonequivalent comparison series. We determined estimates of effect using linear regression models incorporating an autoregressive integrated moving average error term. We used neighboring jurisdictions (Alberta, Saskatchewan, Washington State) as external controls. Results. In the 2 years after implementation of the new laws, significant decreases occurred in fatal crashes (21.0%; 95% confidence interval [CI] = 15.3, 26.4) and in hospital admissions (8.0%; 95% CI = 0.6, 14.9) and ambulance calls (7.2%; 95% CI = 1.1, 13.0) for road trauma. We found a very large reduction in alcohol-related fatal crashes (52.0%; 95% CI = 34.5, 69.5), and the benefits of the new laws are likely primarily the result of a reduction in drinking and driving. Conclusions. These findings suggest that laws calling for immediate sanctions for dangerous drivers can reduce road trauma and should be supported. [ABSTRACT FROM AUTHOR]

We have proposed needed information management capabilities for future US health departments predicated on trends in health care reform and health information technology. Regardless of whether health departments provide direct clinical services (and many will), they will manage unprecedented quantities of sensitive information for the public health core functions of assurance and assessment, including population- level health surveillance and metrics. Absent improved capabilities, health departments risk vestigial status, with consequences for vulnerable populations. Developments in electronic health records, interoperability and information exchange, public information sharing, decision support, and cloud technologies can support information management if health departments have appropriate capabilities. The need for national engagement in and consensus on these capabilities and their importance to health department sustainability make them appropriate for consideration in the context of accreditation. [ABSTRACT FROM AUTHOR]

Objectives. We identified communities of interconnected people that might serve as normative reference groups for individual-level behavior related to latrine adoption. Methods. We applied an algorithmic social network method to determine the network community from respondent-reported social ties of 16 403 individuals in 75 villages in rural Karnataka, India; data were collected from 2006 to 2008. We used multilevel modeling to test the association between latrine ownership and community-level and village-level latrine ownership. We also investigated the degree to which network cohesion affected individual latrine ownership. Results. Three levels of social contacts (direct friends, social network community, and village) significantly predicted individual latrine ownership, but the strongest effect was found at the level of social network communities. In communities with high levels of network cohesion, the likelihood was decreased that any individual would own a latrine; this effect was significant only at lower levels of latrine ownership, suggesting a role for network cohesion in facilitating the nonownership norm. Conclusions. Although many international health and development interventions target village units, these results raise the possibility that the optimal target for public health interventions may not be determined through geography but through social network interactions. [ABSTRACT FROM AUTHOR]

Objectives. We examined depression within a multidimensional framework consisting of genetic, environmental, and sociobehavioral factors and, using machine learning algorithms, explored interactions among these factors that might better explain the etiology of depressive symptoms. Methods. We measured current depressive symptoms using the Center for Epidemiologic Studies Depression Scale (n = 6378 participants in the Wisconsin Longitudinal Study). Genetic factors were 78 single nucleotide polymorphisms (SNPs); environmental factors-13 stressful life events (SLEs), plus a composite proportion of SLEs index; and sociobehavioral factors-18 personality, intelligence, and other health or behavioral measures. We performed traditional SNP associations via logistic regression likelihood ratio testing and explored interactions with support vector machines and Bayesian networks. Results. After correction for multiple testing, we found no significant single genotypic associations with depressive symptoms. Machine learning algorithms showed no evidence of interactions. Naïve Bayes produced the best models in both subsets and included only environmental and sociobehavioral factors. Conclusions. We found no single or interactive associations with genetic factors and depressive symptoms. Various environmental and sociobehavioral factors were more predictive of depressive symptoms, yet their impacts were independent of one another. A genome-wide analysis of genetic alterations using machine learning methodologies will provide a framework for identifying genetic-environmental-sociobehavioral interactions in depressive symptoms. [ABSTRACT FROM AUTHOR]

Objectives. We examined rates of and risk factors for repeat syphilis infection among men who have sex with men (MSM) in California. Methods. We analyzed 2002 to 2006 California syphilis surveillance system data. Results. During the study period, a mean of 5.9% (range: 4.9%-7.1% per year) of MSM had a repeat primary or secondary (PS) syphilis infection within 2 years of an initial infection. There was no significant increase in the annual proportion of MSM with a repeat syphilis infection (P=.42). In a multivariable model, factors associated with repeat syphilis infection were HIV infection (odds ratio [OR]= 1.65; 95% confidence interval [CI]=1.14, 2.37), Black race (OR=1.84; 95% CI=1.12, 3.04), and 10 or more recent sex partners (OR=1.99; 95% CI=1.12, 3.50). Conclusions. Approximately 6% of MSM in California have a repeat PS syphilis infection within 2 years of an initial infection. HIV infection, Black race, and having multiple sex partners are associated with increased odds of repeat infection. Syphilis elimination efforts should include messages about the risk for repeat infection and the importance of follow-up testing. Public health attention to individuals repeatedly infected with syphilis may help reduce local disease burdens. [ABSTRACT FROM AUTHOR]

Abstract: We used the Health Services Commission data from Manitoba, Canada to identity complications resulting from hysterectomy, cholecystectomy, and prostatectomy which led to hospital readmissions. For each procedure, two specialists independently judged whether the readmissions were for surgery-related complications on the basis of liberally interpreted literature guidelines. Then, each pair of physicians met to resolve differences; only complications agreed upon by physicians were retained in our computer-based analysis. The analysis was done in Three steps: algorithms were developed using guidelines from the literature, physician input, and 1974 hospital claims; these were then modified using 1975 data: finally, the algorithms were tested with 1976 data. The computerized algorithms developed were compared with the clinical decisions of physician panels. The results showed high specificity, sensitivity, and predictive value. Given the increasing availability of routinely collected data bases, the possibilities for inexpensively monitoring the outcomes of different providers and institutions are appealing. More extensive validation and application of the methodology to a greater number of procedures are necessary to implement such a program. (Am J Public Health 1985: 75:1288-1295.) [ABSTRACT FROM AUTHOR]

A previous study showed that a clinical algorithm for respiratory illnesses consisting of a checklist, a set of instructions (logic), and computer audit/feedback, could reduce costs significantly while maintaining a high quality of care. The results of this study show that the algorithm system, developed and validated at one primary care clinic can be successfully imported to another primary care clinic. In the present study, the algorithm system significantly improved the completeness of the medical records reduced the use of medical tests by 20 per cent-75 per cent and reduced non-provider costs by 36 per cent per patient visit. This study also shows that all three components of the algorithm system appear to he necessary to achieve these improvements and maintain a high quality of medical care. These results suggest that a wider use of the algorithm system for minor acute medical problems is both feasible and useful in providing high-quality cost-effective care that is audilable. [ABSTRACT FROM AUTHOR]

Abstract: A clinical algorithm was developed for staging of Hodgkin disease, This represented the treatment logic employed in a specified time period and was used to categorize charges incurred in the staging of 50 consecutive patients who underwent this process at a university hospital. Charges were categorized as necessary (in concordance with the algorithm), justifiable (a result of care for a separate medical disorder or for consideration of distance), and perceived excess (neither necessary nor justifiable). Thirty per ¢ of the total charges were identified as perceived excess. Hospitalization accounted for 68 per ¢ and laboratory charges 22 per ¢ of the perceived excess charges. Fifty-eight per ¢ of the patients had a percentage of perceived excess bed charges exceeding 30 per ¢ of the total. Perceived excess medical hospitalization correlated positively with time from diagnosis to treatment decision. It is recommended that attempts to reduce the expense of Hodgkin lymphoma staging be directed at decreasing hospitalization. (Am J Public Health 1981: 71:376-380.) [ABSTRACT FROM AUTHOR]

We performed an exploratory evaluation of the Environmental Protection Agency's algorithm for assessment of risk of exposure to asbestos in buildings. Observers scored five sites using the algorithm. One group (N = 23) conducted their assessments relying only on written instructions, while another (N = 20) was trained in the use of the algorithm. Five professional industrial hygienists, experienced in risk assessment, also used the algorithm on the same sites. A wide range of highly variable results, for both untrained and trained observers coupled with poor comparability with expert's scores raise concern as to the reliability of this assessment tool. The development of a valid and reliable "lay" index of risk is still needed. (Am J Public Health 1983; 73: 1179-1181.) [ABSTRACT FROM AUTHOR]

The author reflects on the various uses of algorithms in medicine. He notes that algorithms have evolved to be an essential tool in patient care, particularly minor surgery, minor illnesses, acute medical emergencies, and pediatric telephone triage. He is impressed by the increasing popularity of algorithms as educational tools for physicians and patients. He argues that algorithms can help healthcare professionals shape their decisions and clarify their knowledge.