PROVIDENT: Development and Validation of a Machine Learning Model to Predict Neighborhood-level Overdose Risk in Rhode Island.

BACKGROUND: Drug overdose persists as a leading cause of death in the United States, but resources to address it remain limited. As a result, health authorities must consider where to allocate scarce resources within their jurisdictions. Machine learning offers a strategy to identify areas with increased future overdose risk to proactively allocate overdose prevention resources. This modeling study is embedded in a randomized trial to measure the effect of proactive resource allocation on statewide overdose rates in Rhode Island (RI). METHODS: We used statewide data from RI from 2016 to 2020 to develop an ensemble machine learning model predicting neighborhood-level fatal overdose risk. Our ensemble model integrated gradient boosting machine and super learner base models in a moving window framework to make predictions in 6-month intervals. Our performance target, developed a priori with the RI Department of Health, was to identify the 20% of RI neighborhoods containing at least 40% of statewide overdose deaths, including at least one neighborhood per municipality. The model was validated after trial launch. RESULTS: Our model selected priority neighborhoods capturing 40.2% of statewide overdose deaths during the test periods and 44.1% of statewide overdose deaths during validation periods. Our ensemble outperformed the base models during the test periods and performed comparably to the best-performing base model during the validation periods. CONCLUSIONS: We demonstrated the capacity for machine learning models to predict neighborhood-level fatal overdose risk to a degree of accuracy suitable for practitioners. Jurisdictions may consider predictive modeling as a tool to guide allocation of scarce resources.

Translating Predictive Analytics for Public Health Practice: A Case Study of Overdose Prevention in Rhode Island.

Prior applications of machine learning to population health have relied on conventional model assessment criteria, limiting the utility of models as decision support tools for public health practitioners. To facilitate practitioners' use of machine learning as a decision support tool for area-level intervention, we developed and applied 4 practice-based predictive model evaluation criteria (implementation capacity, preventive potential, health equity, and jurisdictional practicalities). We used a case study of overdose prevention in Rhode Island to illustrate how these criteria could inform public health practice and health equity promotion. We used Rhode Island overdose mortality records from January 2016-June 2020 (n = 1,408) and neighborhood-level US Census data. We employed 2 disparate machine learning models, Gaussian process and random forest, to illustrate the comparative utility of our criteria to guide interventions. Our models predicted 7.5%-36.4% of overdose deaths during the test period, illustrating the preventive potential of overdose interventions assuming 5%-20% statewide implementation capacities for neighborhood-level resource deployment. We describe the health equity implications of use of predictive modeling to guide interventions along the lines of urbanicity, racial/ethnic composition, and poverty. We then discuss considerations to complement predictive model evaluation criteria and inform the prevention and mitigation of spatially dynamic public health problems across the breadth of practice. This article is part of a Special Collection on Mental Health.

Does educational attainment modify the causal relationship between adiposity and cardiovascular disease? A Mendelian randomization study.

A greater risk of cardiovascular disease is associated with low educational attainment and high adiposity. Despite the correlation between low educational attainment and high adiposity, whether educational attainment modifies the risk of CVD caused by high adiposity remains poorly understood. We investigated the effect of adiposity (body mass index [BMI] and waist-to-hip ratio adjusted for BMI [WHRadjBMI]) on incident CVD among individuals with varying education levels, using associational and one-sample Mendelian randomization (MR) survival analyses. Data were collected from 2006 to 2021, and sample sizes were 254,281 (27,511 CVD cases) for BMI and 253,968 (27,458 CVD cases) for WHRadjBMI. In the associational model, a standard deviation (SD) higher BMI was associated with 19.81 (95% CI: 18.55-21.06) additional cases of incident CVD per 10,000 person-years for individuals with a secondary education, versus 32.96 (95% CI: 28.75-37.17) for those without. When university degree served as the education variable, education group differences attenuated, with 18.26 (95% CI: 16.37-20.15) cases from a one SD higher BMI for those with a university degree versus 23.18 [95% CI: 21.56-24.72] for those without. For the MR model, an SD higher BMI resulted in 11.75 (95% CI: -0.84-24.38) and 29.79 (95% CI: 17.20-42.44) additional cases of incident CVD per 10,000 person-years for individuals with versus without a university degree. WHRadjBMI exhibited no effect differences by education. While the associational model showed evidence of educational attainment modifying the relationship between adiposity and incident CVD, it does not modify the association between adiposity and incident CVD in the MR models. This suggests either less education does not cause greater risk of incident CVD from high adiposity, or MR models cannot detect the effect difference. The associational point estimates exist within the MR models' confidence intervals in all BMI analyses, so we cannot rule out the effect sizes in the associational models.

Respiratory pandemic preparedness learnings from the June 2020 COVID-19 outbreak at San Quentin California State Prison.

PURPOSE: This study aims to characterize the June 2020 COVID-19 outbreak at San Quentin California State Prison and to describe what made San Quentin so vulnerable to uncontrolled transmission. DESIGN/METHODOLOGY/APPROACH: Since its onset, the COVID-19 pandemic has exposed and exacerbated the profound health harms of carceral settings, such that nearly half of state prisons reported COVID-19 infection rates that were four or more times (and up to 15 times) the rate found in the state's general population. Thus, addressing the public health crises and inequities of carceral settings during a respiratory pandemic requires analyzing the myriad factors shaping them. In this study, we reported observations and findings from environmental risk assessments during visits to San Quentin California State Prison. We complemented our assessments with analyses of administrative data. FINDINGS: For future respiratory pathogens that cannot be prevented with effective vaccines, this study argues that outbreaks will no doubt occur again without robust implementation of additional levels of preparedness - improved ventilation, air filtration, decarceration with emergency evacuation planning - alongside addressing the vulnerabilities of carceral settings themselves. ORIGINALITY/VALUE: This study addresses two critical aspects that are insufficiently covered in the literature: how to prepare processes to safely implement emergency epidemic measures when needed, such as potential evacuation, and how to address unique challenges throughout an evolving pandemic for each carceral setting.

Identifying Predictors of Opioid Overdose Death at a Neighborhood Level With Machine Learning.

Predictors of opioid overdose death in neighborhoods are important to identify, both to understand characteristics of high-risk areas and to prioritize limited prevention and intervention resources. Machine learning methods could serve as a valuable tool for identifying neighborhood-level predictors. We examined statewide data on opioid overdose death from Rhode Island (log-transformed rates for 2016-2019) and 203 covariates from the American Community Survey for 742 US Census block groups. The analysis included a least absolute shrinkage and selection operator (LASSO) algorithm followed by variable importance rankings from a random forest algorithm. We employed double cross-validation, with 10 folds in the inner loop to train the model and 4 outer folds to assess predictive performance. The ranked variables included a range of dimensions of socioeconomic status, including education, income and wealth, residential stability, race/ethnicity, social isolation, and occupational status. The R2 value of the model on testing data was 0.17. While many predictors of overdose death were in established domains (education, income, occupation), we also identified novel domains (residential stability, racial/ethnic distribution, and social isolation). Predictive modeling with machine learning can identify new neighborhood-level predictors of overdose in the continually evolving opioid epidemic and anticipate the neighborhoods at high risk of overdose mortality.

Preventing Overdose Using Information and Data from the Environment (PROVIDENT): protocol for a randomized, population-based, community intervention trial.

BACKGROUND AND AIMS: In light of the accelerating drug overdose epidemic in North America, new strategies are needed to identify communities most at risk to prioritize geographically the existing public health resources (e.g. street outreach, naloxone distribution efforts). We aimed to develop PROVIDENT (Preventing Overdose using Information and Data from the Environment), a machine learning-based forecasting tool to predict future overdose deaths at the census block group (i.e. neighbourhood) level. DESIGN: Randomized, population-based, community intervention trial. SETTING: Rhode Island, USA. PARTICIPANTS: All people who reside in Rhode Island during the study period may contribute data to either the model or the trial outcomes. INTERVENTION: Each of the state's 39 municipalities will be randomized to the intervention (PROVIDENT) or comparator condition. An interactive, web-based tool will be developed to visualize the PROVIDENT model predictions. Municipalities assigned to the treatment arm will receive neighbourhood risk predictions from the PROVIDENT model, and state agencies and community-based organizations will direct resources to neighbourhoods identified as high risk. Municipalities assigned to the control arm will continue to receive surveillance information and overdose prevention resources, but they will not receive neighbourhood risk predictions. MEASUREMENTS: The primary outcome is the municipal-level rate of fatal and non-fatal drug overdoses. Fatal overdoses will be defined as unintentional drug-related death; non-fatal overdoses will be defined as an emergency department visit for a suspected overdose reported through the state's syndromic surveillance system. Intervention efficacy will be assessed using Poisson or negative binomial regression to estimate incidence rate ratios comparing fatal and non-fatal overdose rates in treatment vs. control municipalities. COMMENTS: The findings will inform the utility of predictive modelling as a tool to improve public health decision-making and inform resource allocation to communities that should be prioritized for prevention, treatment, recovery and overdose rescue services.

How does public perception of antibiotic use on dairy farms contribute to self-reported purchasing of organic?

The United States regulates the use of antibiotics in agricultural settings to address the global antibiotic resistance problem. Conventional dairy cows treated with antibiotics are kept in the herd and after the withholding period milk is harvested. On organic farms, the US organic standard on antibiotic use requires sick dairy cows to be treated, but treated cows must be removed from the herd and their milk can never again be sold as certified organic. This study investigated the US public's perceptions of the organic dairy farming, antibiotic use on dairy farms, and whether these perceptions affect consumer's self-reported purchasing behavior for organic. We used a nationally representative phone-based survey of 1000 US adults and characterized participants' self-reported (i) knowledge of the legality of antibiotic use on dairy farms (conventional and organic) and (ii) frequency of purchasing organic instead of conventional dairy products, as well as several demographic and other variables. The results indicated that participants' knowledge about antibiotic use practices in dairy farming have no effect on their self-reported purchasing behavior for organic or conventional dairy products. However, respondents who were familiar with the regulations of antibiotic use on dairy farms were more likely to oppose the US organic standard on antibiotic use in dairy farming and thought that past antibiotic use should not permanently remove a cow's organic status. These findings contribute to understanding of public perceptions that shape the US dairy organic market. PRACTICAL APPLICATION: Income, employment, health and political values, but not consumers' knowledge about antibiotic use in dairy farming, affect self-reported purchasing behavior for organic dairy products. However, consumers who are familiar with the regulations of antibiotic use on US dairy farms disagree with the US organic standard on antibiotic use mandating loss of organic status for any cattle treated with antibiotics. These findings may be useful to organic markets.

Methodological Challenges in Estimating the Lifetime Medical Care Cost Externality of Obesity.

There is a great deal of variability in estimates of the lifetime medical care cost externality of obesity, partly due to a lack of transparency in the methodology behind these cost models. Several important factors must be considered in producing the best possible estimate, including age-related weight gain, differential life expectancy, identifiability, and cost model selection. In particular, age-related weight gain represents an important new component to recent cost estimates. Without accounting for age-related weight gain, a study relies on the untenable assumption that people remain the same weight throughout their lives, leading to a fundamental misunderstanding of the evolution and development of the obesity crisis. This study seeks to inform future researchers on the best methods and data available both to estimate age-related weight gain and to accurately and consistently estimate obesity's lifetime external medical care costs. This should help both to create a more standardized approach to cost estimation as well as encourage more transparency between all parties interested in the question of obesity's lifetime cost and, ultimately, evaluating the benefits and costs of interventions targeting obesity at various points in the life course.

Predicted Lifetime Third-Party Costs of Obesity for Black and White Adolescents with Race-Specific Age-Related Weight Gain.

OBJECTIVE: There exists enormous variation in estimates of the lifetime cost of adolescent obesity by race. To justify policy measures to reduce obesity rates nationally in this demographic, the costs of obesity in late adolescence must first be discerned. Although several researchers have sought to quantify obesity's true cost, none has accounted for race-specific age-related weight gain, a vital component in producing an accurate estimate. METHODS: This paper employs a Markov model of BMI category state changes separately for black and white males and females from age 18 to 75 applied to updated estimates of obesity's costs and effect on mortality to quantify the median lifetime cost of obesity at age 18. RESULTS: This study found lower lifetime costs than previously, largely because of the dramatic gain in weight among normal-weight individuals, particularly black males, that occurs in early adulthood. CONCLUSIONS: A substantial portion of obesity's prevalence, and therefore cost, for black males and females comes from age-related weight gain in early adulthood. This speaks to the persistent threat of obesity beyond adolescence for this demographic, and further research should focus on whether policy can modify the behaviors and environment through which and in which this sharp increase in weight occurs.