Big data and predictive modelling for the opioid crisis: existing research and future potential

Looking into the black mirror of the overdose crisis: Assessing the harms of collaborative surveillance technologies in the United States response

Drug overdose is a leading cause of death among adults in the United States, prompting calls for more surveillance data and data sharing across public health and law enforcement to address the crisis. This paper integrates Black feminist science and technology studies (STS) into an anthropological analysis of the collision of public health, policing, and technology as embedded in the US National Overdose Response Strategy and its technological innovation, the Overdose Detection Mapping Application Program (ODMAP). The dystopian Netflix series "Black Mirror," which explores the seemingly useful but quietly destructive potential of technology, offers a lens through which to speculate upon and anticipate the harms of collaborative surveillance projects. Ultimately, I ask: are such technological interventions a benevolent approach to a public health crisis or are we looking into a black mirror of racialized surveillance and criminalization of overdose in the United States?

Predictability of buprenorphine-naloxone treatment retention: A multi-site analysis combining electronic health records and machine learning

BACKGROUND AND AIMS

Opioid use disorder (OUD) and opioid dependence lead to significant morbidity and mortality, yet treatment retention, crucial for the effectiveness of medications like buprenorphine-naloxone, remains unpredictable. Our objective was to determine the predictability of 6-month retention in buprenorphine-naloxone treatment using electronic health record (EHR) data from diverse clinical settings and to identify key predictors.

DESIGN

This retrospective observational study developed and validated machine learning-based clinical risk prediction models using EHR data.

SETTING AND CASES

Data were sourced from Stanford University's healthcare system and Holmusk's NeuroBlu database, reflecting a wide range of healthcare settings. The study analyzed 1800 Stanford and 7957 NeuroBlu treatment encounters from 2008 to 2023 and from 2003 to 2023, respectively.

MEASUREMENTS

Predict continuous prescription of buprenorphine-naloxone for at least 6 months, without a gap of more than 30 days. The performance of machine learning prediction models was assessed by area under receiver operating characteristic (ROC-AUC) analysis as well as precision, recall and calibration. To further validate our approach's clinical applicability, we conducted two secondary analyses: a time-to-event analysis on a single site to estimate the duration of buprenorphine-naloxone treatment continuity evaluated by the C-index and a comparative evaluation against predictions made by three human clinical experts.

FINDINGS

Attrition rates at 6 months were 58% (NeuroBlu) and 61% (Stanford). Prediction models trained and internally validated on NeuroBlu data achieved ROC-AUCs up to 75.8 (95% confidence interval [CI] = 73.6-78.0). Addiction medicine specialists' predictions show a ROC-AUC of 67.8 (95% CI = 50.4-85.2). Time-to-event analysis on Stanford data indicated a median treatment retention time of 65 days, with random survival forest model achieving an average C-index of 65.9. The top predictor of treatment retention identified included the diagnosis of opioid dependence.

CONCLUSIONS

US patients with opioid use disorder or opioid dependence treated with buprenorphine-naloxone prescriptions appear to have a high (∼60%) treatment attrition by 6 months. Machine learning models trained on diverse electronic health record datasets appear to be able to predict treatment continuity with accuracy comparable to that of clinical experts.

The Performance of ChatGPT on the American Society for Surgery of the Hand Self-Assessment Examination

BACKGROUND

This study aims to compare the performance of ChatGPT-3.5 (GPT-3.5) and ChatGPT-4 (GPT-4) on the American Society for Surgery of the Hand (ASSH) Self-Assessment Examination (SAE) to determine their potential as educational tools.

METHODS

This study assessed the proportion of correct answers to text-based questions on the 2021 and 2022 ASSH SAE between untrained ChatGPT versions. Secondary analyses assessed the performance of ChatGPT based on question difficulty and question category. The outcomes of ChatGPT were compared with the performance of actual examinees on the ASSH SAE.

RESULTS

A total of 238 questions were included in the analysis. Compared with GPT-3.5, GPT-4 provided significantly more correct answers overall (58.0% versus 68.9%, respectively; P = 0.013), on the 2022 SAE (55.9% versus 72.9%; P = 0.007), and more difficult questions (48.8% versus 63.6%; P = 0.02). In a multivariable logistic regression analysis, correct answers were predicted by GPT-4 (odds ratio [OR], 1.66; P = 0.011), increased question difficulty (OR, 0.59; P = 0.009), Bone and Joint questions (OR, 0.18; P < 0.001), and Soft Tissue questions (OR, 0.30; P = 0.013). Actual examinees scored a mean of 21.6% above GPT-3.5 and 10.7% above GPT-4. The mean percentage of correct answers by actual examinees was significantly higher for correct (versus incorrect) ChatGPT answers.

CONCLUSIONS

GPT-4 demonstrated improved performance over GPT-3.5 on the ASSH SAE, especially on more difficult questions. Actual examinees scored higher than both versions of ChatGPT, but the margin was cut in half by GPT-4.

Generative Artificial Intelligence Performs at a Second-Year Orthopedic Resident Level

Introduction Artificial intelligence (AI) models using large language models (LLMs) and non-specific domains have gained attention for their innovative information processing. As AI advances, it's essential to regularly evaluate these tools' competency to maintain high standards, prevent errors or biases, and avoid flawed reasoning or misinformation that could harm patients or spread inaccuracies. Our study aimed to determine the performance of Chat Generative Pre-trained Transformer (ChatGPT) by OpenAI and Google BARD (BARD) in orthopedic surgery, assess performance based on question types, contrast performance between different AIs and compare AI performance to orthopedic residents. Methods We administered ChatGPT and BARD 757 Orthopedic In-Training Examination (OITE) questions. After excluding image-related questions, the AIs answered 390 multiple choice questions, all categorized within 10 sub-specialties (basic science, trauma, sports medicine, spine, hip and knee, pediatrics, oncology, shoulder and elbow, hand, and food and ankle) and three taxonomy classes (recall, interpretation, and application of knowledge). Statistical analysis was performed to analyze the number of questions answered correctly by each AI model, the performance returned by each AI model within the categorized question sub-specialty designation, and the performance of each AI model in comparison to the results returned by orthopedic residents classified by their respective post-graduate year (PGY) level. Results BARD answered more overall questions correctly (58% vs 54%, p<0.001). ChatGPT performed better in sports medicine and basic science and worse in hand surgery, while BARD performed better in basic science (p<0.05). The AIs performed better in recall questions compared to the application of knowledge (p<0.05). Based on previous data, it ranked in the 42nd-96th percentile for post-graduate year ones (PGY1s), 27th-58th for PGY2s, 3rd-29th for PGY3s, 1st-21st for PGY4s, and 1st-17th for PGY5s. Discussion ChatGPT excelled in sports medicine but fell short in hand surgery, while both AIs performed well in the basic science sub-specialty but performed poorly in the application of knowledge-based taxonomy questions. BARD performed better than ChatGPT overall. Although the AI reached the second-year PGY orthopedic resident level, it fell short of passing the American Board of Orthopedic Surgery (ABOS). Its strengths in recall-based inquiries highlight its potential as an orthopedic learning and educational tool.

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.

The Role of Artificial Intelligence in Anterior Cruciate Ligament Injuries: Current Concepts and Future Perspectives

The remarkable progress in data aggregation and deep learning algorithms has positioned artificial intelligence (AI) and machine learning (ML) to revolutionize the field of medicine. AI is becoming more and more prevalent in the healthcare sector, and its impact on orthopedic surgery is already evident in several fields. This review aims to examine the literature that explores the comprehensive clinical relevance of AI-based tools utilized before, during, and after anterior cruciate ligament (ACL) reconstruction. The review focuses on current clinical applications and future prospects in preoperative management, encompassing risk prediction and diagnostics; intraoperative tools, specifically navigation, identifying complex anatomic landmarks during surgery; and postoperative applications in terms of postoperative care and rehabilitation. Additionally, AI tools in educational and training settings are presented. Orthopedic surgeons are showing a growing interest in AI, as evidenced by the applications discussed in this review, particularly those related to ACL injury. The exponential increase in studies on AI tools applicable to the management of ACL tears promises a significant future impact in its clinical application, with growing attention from orthopedic surgeons.

Visual Analytics for Data-Driven Understanding of the Substance Use Disorder Epidemic

The substance use disorder epidemic has emerged as a serious public health crisis, presenting complex challenges. Visual analytics offers a unique approach to address this complexity and facilitate effective interventions. This paper details the development of an innovative visual analytics dashboard, aimed at enhancing our understanding of the substance use disorder epidemic. By employing record linkage techniques, we integrate diverse data sources to provide a comprehensive view of the epidemic. Adherence to responsive, open, and user-centered design principles ensures the dashboard's usefulness and usability. Our approach to data and design encourages collaboration among various stakeholders, including researchers, politicians, and healthcare practitioners. Through illustrative outputs, we demonstrate how the dashboard can deepen our understanding of the epidemic, support intervention strategies, and evaluate the effectiveness of implemented measures. The paper concludes with a discussion of the dashboard's use cases and limitations.

Update of a Multivariable Opioid Overdose Risk Prediction Model to Enhance Clinical Care for Long-term Opioid Therapy Patients

BACKGROUND

Clinical opioid overdose risk prediction models can be useful tools to reduce the risk of overdose in patients prescribed long-term opioid therapy (LTOT). However, evolving overdose risk environments and clinical practices in addition to potential harmful model misapplications require careful assessment prior to widespread implementation into clinical care. Models may need to be tailored to meet local clinical operational needs and intended applications in practice.

OBJECTIVE

To update and validate an existing opioid overdose risk model, the Kaiser Permanente Colorado Opioid Overdose (KPCOOR) Model, in patients prescribed LTOT for implementation in clinical care.

DESIGN, SETTING, AND PARTICIPANTS

The retrospective cohort study consisted of 33, 625 patients prescribed LTOT between January 2015 and June 2019 at Kaiser Permanente Colorado, with follow-up through June 2021.

MAIN MEASURES

The outcome consisted of fatal opioid overdoses identified from vital records and non-fatal opioid overdoses from emergency department and inpatient settings. Predictors included demographics, medication dispensings, substance use disorder history, mental health history, and medical diagnoses. Cox proportional hazards regressions were used to model 2-year overdose risk.

KEY RESULTS

During follow-up, 65 incident opioid overdoses were observed (111.4 overdoses per 100,000 person-years) in the study cohort, of which 11 were fatal. The optimal risk model needed to risk-stratify patients and to be easily interpreted by clinicians. The original 5-variable model re-validated on the new study cohort had a bootstrap-corrected C-statistic of 0.73 (95% CI, 0.64-0.85) compared to a C-statistic of 0.80 (95% CI, 0.70-0.88) in the updated model and 0.77 (95% CI, 0.66-0.87) in the final adapted 7-variable model, which was also well-calibrated.

CONCLUSIONS

Updating and adapting predictors for opioid overdose in the KPCOOR Model with input from clinical partners resulted in a parsimonious and clinically relevant model that was poised for integration in clinical care.

Big data evidence of the impact of COVID-19 hospitalizations on mortality rates of non-COVID-19 critically ill patients

The COVID-19 virus caused a global pandemic leading to a swift policy response. While this response was designed to prevent the spread of the virus and support those with COVID-19, there is growing evidence regarding measurable impacts on non-COVID-19 patients. The paper uses a large dataset from administrative records of the Brazilian public health system (SUS) to estimate pandemic spillover effects in critically ill health care delivery, i.e. the additional mortality risk that COVID-19 ICU hospitalizations generate on non-COVID-19 patients receiving intensive care. The data contain the universe of ICU hospitalizations in SUS from February 26, 2020 to December 31, 2021. Spillover estimates are obtained from high-dimensional fixed effects regression models that control for a number of unobservable confounders. Our findings indicate that, on average, the pandemic increased the mortality risk of non-COVID-19 ICU patients by 1.296 percentage points, 95% CI 1.145-1.448. The spillover mortality risk is larger for non-COVID patients receiving intensive care due to diseases of the respiratory system, diseases of the skin and subcutaneous tissue, and infectious and parasitic diseases. As of July 2023, the WHO reports more than 6.9 million global deaths due to COVID-19 infection. However, our estimates of spillover effects suggest that the pandemic's total death toll is much higher.

Can Artificial Intelligence Pass the American Board of Orthopaedic Surgery Examination? Orthopaedic Residents Versus ChatGPT

BACKGROUND

Advances in neural networks, deep learning, and artificial intelligence (AI) have progressed recently. Previous deep learning AI has been structured around domain-specific areas that are trained on dataset-specific areas of interest that yield high accuracy and precision. A new AI model using large language models (LLM) and nonspecific domain areas, ChatGPT (OpenAI), has gained attention. Although AI has demonstrated proficiency in managing vast amounts of data, implementation of that knowledge remains a challenge.

QUESTIONS/PURPOSES

(1) What percentage of Orthopaedic In-Training Examination questions can a generative, pretrained transformer chatbot (ChatGPT) answer correctly? (2) How does that percentage compare with results achieved by orthopaedic residents of different levels, and if scoring lower than the 10th percentile relative to 5th-year residents is likely to correspond to a failing American Board of Orthopaedic Surgery score, is this LLM likely to pass the orthopaedic surgery written boards? (3) Does increasing question taxonomy affect the LLM's ability to select the correct answer choices?

METHODS

This study randomly selected 400 of 3840 publicly available questions based on the Orthopaedic In-Training Examination and compared the mean score with that of residents who took the test over a 5-year period. Questions with figures, diagrams, or charts were excluded, including five questions the LLM could not provide an answer for, resulting in 207 questions administered with raw score recorded. The LLM's answer results were compared with the Orthopaedic In-Training Examination ranking of orthopaedic surgery residents. Based on the findings of an earlier study, a pass-fail cutoff was set at the 10th percentile. Questions answered were then categorized based on the Buckwalter taxonomy of recall, which deals with increasingly complex levels of interpretation and application of knowledge; comparison was made of the LLM's performance across taxonomic levels and was analyzed using a chi-square test.

RESULTS

ChatGPT selected the correct answer 47% (97 of 207) of the time, and 53% (110 of 207) of the time it answered incorrectly. Based on prior Orthopaedic In-Training Examination testing, the LLM scored in the 40th percentile for postgraduate year (PGY) 1s, the eighth percentile for PGY2s, and the first percentile for PGY3s, PGY4s, and PGY5s; based on the latter finding (and using a predefined cutoff of the 10th percentile of PGY5s as the threshold for a passing score), it seems unlikely that the LLM would pass the written board examination. The LLM's performance decreased as question taxonomy level increased (it answered 54% [54 of 101] of Tax 1 questions correctly, 51% [18 of 35] of Tax 2 questions correctly, and 34% [24 of 71] of Tax 3 questions correctly; p = 0.034).

CONCLUSION

Although this general-domain LLM has a low likelihood of passing the orthopaedic surgery board examination, testing performance and knowledge are comparable to that of a first-year orthopaedic surgery resident. The LLM's ability to provide accurate answers declines with increasing question taxonomy and complexity, indicating a deficiency in implementing knowledge.

CLINICAL RELEVANCE

Current AI appears to perform better at knowledge and interpretation-based inquires, and based on this study and other areas of opportunity, it may become an additional tool for orthopaedic learning and education.

Early warnings and slow deaths: A sociology of outbreak and overdose

In this paper, we offer a sociological analysis of early warning and outbreak in the field of drug policy, focusing on opioid overdose. We trace how 'outbreak' is enacted as a rupturing event which enables rapid reflex responses of precautionary control, based largely on short-term and proximal early warning indicators. We make the case for an alternative view of early warning and outbreak. We argue that practices of detection and projection that help to materialise drug-related outbreaks are too focused on the proximal and short-term. Engaging with epidemiological and sociological work investigating epidemics of opioid overdose, we show how the short-termism and rapid reflex response of outbreak fails to appreciate the slow violent pasts of epidemics indicative of an ongoing need and care for structural and societal change. Accordingly, we gather together ideas of 'slow emergency' (Ben Anderson), 'slow death' (Lauren Berlant) and 'slow violence' (Rob Nixon), to re-assemble outbreaks in 'long view'. This locates opioid overdose in long-term attritional processes of deindustrialisation, pharmaceuticalisation, and other forms of structural violence, including the criminalisation and problematisation of people who use drugs. Outbreaks evolve in relation to their slow violent pasts. To ignore this can perpetuate harm. Attending to the social conditions that create the possibilities for outbreak invites early warning that goes 'beyond outbreak' and 'beyond epidemic' as generally configured.

Development and evaluation of a risk algorithm predicting alcohol dependence after early onset of regular alcohol use

AIMS

Likelihood of alcohol dependence (AD) is increased among people who transition to greater levels of alcohol involvement at a younger age. Indicated interventions delivered early may be effective in reducing risk, but could be costly. One way to increase cost-effectiveness would be to develop a prediction model that targeted interventions to the subset of youth with early alcohol use who are at highest risk of subsequent AD.

DESIGN

A prediction model was developed for DSM-IV AD onset by age 25 years using an ensemble machine-learning algorithm known as 'Super Learner'. Shapley additive explanations (SHAP) assessed variable importance.

SETTING AND PARTICIPANTS

Respondents reporting early onset of regular alcohol use (i.e. by 17 years of age) who were aged 25 years or older at interview from 14 representative community surveys conducted in 13 countries as part of WHO's World Mental Health Surveys.

MEASUREMENTS

The primary outcome to be predicted was onset of life-time DSM-IV AD by age 25 as measured using the Composite International Diagnostic Interview, a fully structured diagnostic interview.

FINDINGS

AD prevalence by age 25 was 5.1% among the 10 687 individuals who reported drinking alcohol regularly by age 17. The prediction model achieved an external area under the curve [0.78; 95% confidence interval (CI) = 0.74-0.81] higher than any individual candidate risk model (0.73-0.77) and an area under the precision-recall curve of 0.22. Overall calibration was good [integrated calibration index (ICI) = 1.05%]; however, miscalibration was observed at the extreme ends of the distribution of predicted probabilities. Interventions provided to the 20% of people with highest risk would identify 49% of AD cases and require treating four people without AD to reach one with AD. Important predictors of increased risk included younger onset of alcohol use, males, higher cohort alcohol use and more mental disorders.

CONCLUSIONS

A risk algorithm can be created using data collected at the onset of regular alcohol use to target youth at highest risk of alcohol dependence by early adulthood. Important considerations remain for advancing the development and practical implementation of such models.

Rapid opioid overdose response system technologies

PURPOSE OF REVIEW

Opioid overdose events are a time sensitive medical emergency, which is often reversible with naloxone administration if detected in time. Many countries are facing rising opioid overdose deaths and have been implementing rapid opioid overdose response Systems (ROORS). We describe how technology is increasingly being used in ROORS design, implementation and delivery.

RECENT FINDINGS

Technology can contribute in significant ways to ROORS design, implementation, and delivery. Artificial intelligence-based modelling and simulations alongside wastewater-based epidemiology can be used to inform policy decisions around naloxone access laws and effective naloxone distribution strategies. Data linkage and machine learning projects can support service delivery organizations to mobilize and distribute community resources in support of ROORS. Digital phenotyping is an advancement in data linkage and machine learning projects, potentially leading to precision overdose responses. At the coalface, opioid overdose detection devices through fixed location or wearable sensors, improved connectivity, smartphone applications and drone-based emergency naloxone delivery all have a role in improving outcomes from opioid overdose. Data driven technologies also have an important role in empowering community responses to opioid overdose.

SUMMARY

This review highlights the importance of technology applied to every aspect of ROORS. Key areas of development include the need to protect marginalized groups from algorithmic bias, a better understanding of individual overdose trajectories and new reversal agents and improved drug delivery methods.

Between Care and Control: Examining Surveillance Practices in Harm Reduction

As harm reduction programs and services proliferate, people who use drugs (PWUD) are increasingly subjected to surveillance through the collection of their personal information, systematic observation, and other means. The data generated from these practices are frequently repurposed across various institutional sites for clinical, evaluative, epidemiological, and administrative uses. Rationales provided for increased surveillance include the more effective provision of care, service optimization, risk stratification, and efficiency in resource allocation. With this in mind, our reflective essay draws on empirical analysis of work within harm reduction services and movements to reflect critically on the impacts and implications of surveillance expansion. While we argue that many surveillance practices are not inherently problematic or harmful, the unchecked expansion of surveillance under a banner of health and harm reduction may contribute to decreased uptake of services, rationing and conditionalities tied to service access, the potential deepening of health disparities amongst some PWUD, and an overlay of health and criminal-legal systems. In this context, surveillance relies on the enlistment of a range of therapeutic actors and reflects the permeable boundary between care and control. We thus call for a broader critical dialogue within harm reduction on the problems and potential impacts posed by surveillance in service settings, the end to data sharing of health information with law enforcement and other criminal legal actors, and deference to the stated need among PWUD for meaningful anonymity when accessing harm reduction and health services.

Estimating County-Level Overdose Rates Using Opioid-Related Twitter Data: Interdisciplinary Infodemiology Study

BACKGROUND

There were an estimated 100,306 drug overdose deaths between April 2020 and April 2021, a three-quarter increase from the prior 12-month period. There is an approximate 6-month reporting lag for provisional counts of drug overdose deaths from the National Vital Statistics System, and the highest level of geospatial resolution is at the state level. By contrast, public social media data are available close to real-time and are often accessible with precise coordinates.

OBJECTIVE

The purpose of this study is to assess whether county-level overdose mortality burden could be estimated using opioid-related Twitter data.

METHODS

International Classification of Diseases (ICD) codes for poisoning or exposure to overdose at the county level were obtained from CDC WONDER. Demographics were collected from the American Community Survey. The Twitter Application Programming Interface was used to obtain tweets that contained any of the 36 terms with drug names. An unsupervised classification approach was used for clustering tweets. Population-normalized variables and polynomial population-normalized variables were produced. Furthermore, z scores of the Getis Ord Gi clustering statistic were produced, and both these scores and their polynomial counterparts were explored in regression modeling of county-level overdose mortality burden. A series of linear regression models were used for predictive modeling to explore the interpretability of the analytical output.

RESULTS

Modeling overdose mortality with normalized demographic variables alone explained only 7.4% of the variability in county-level overdose mortality, whereas this was approximately doubled by the use of specific demographic and Twitter data covariates based on a backward selection approach. The highest adjusted R² and lowest AIC (Akaike Info Criterion) were obtained for the model with normalized demographic variables, normalized z scores from geospatial analyses, and normalized topic counts (adjusted R²=0.133, AIC=8546.8). The z scores of the Getis Ord Gi statistic appeared to have improved utility over population-normalization alone. In this model, median age, female population, and tweets about web-based drug sales were positively associated with opioid mortality. Asian race and Hispanic ethnicity were significantly negatively associated with county-level burdens of overdose mortality.

CONCLUSIONS

Social media data, when transformed using certain statistical approaches, may add utility to the goal of producing closer to real-time county-level estimates of overdose mortality. Prediction of opioid-related outcomes can be advanced to inform prevention and treatment decisions. This interdisciplinary approach can facilitate evidence-based funding decisions for various substance use disorder prevention and treatment programs.

Using Machine Learning to Predict Treatment Adherence in Patients on Medication for Opioid Use Disorder

OBJECTIVE

Patients receiving medication for opioid use disorder (MOUD) may continue using nonprescribed drugs or have trouble with medication adherence, and it is difficult to predict which patients will continue to do so. In this study, we develop and validate an automated risk-modeling framework to predict opioid abstinence and medication adherence at a patient's next attended appointment and evaluate the predictive performance of machine-learning algorithms versus logistic regression.

METHODS

Urine drug screen and attendance records from 40,005 appointments drawn from 2742 patients at a multilocation office-based MOUD program were used to train logistic regression, logistic ridge regression, and XGBoost models to predict a composite indicator of treatment adherence (opioid-negative and norbuprenorphine-positive urine, no evidence of urine adulteration) at next attended appointment.

RESULTS

The XGBoost model had similar accuracy and discriminative ability (accuracy, 88%; area under the receiver operating curve, 0.87) to the two logistic regression models (accuracy, 88%; area under the receiver operating curve, 0.87). The XGBoost model had nearly perfect calibration in independent validation data; the logistic and ridge regression models slightly overestimated adherence likelihood. Historical treatment adherence, attendance rate, and fentanyl-positive urine at current appointment were the strongest contributors to treatment adherence at next attended appointment.

DISCUSSION

There is a need for risk prediction tools to improve delivery of MOUD. This study presents an automated and portable risk-modeling framework to predict treatment adherence at each patient's next attended appointment. The XGBoost algorithm appears to provide similar classification accuracy to logistic regression models; however, XGBoost may offer improved calibration of risk estimates compared with logistic regression.

Machine learning nonresponse adjustment of patient-reported opioid consumption data to enable consumption-informed postoperative opioid prescribing guidelines

Background

Post-discharge opioid consumption is a crucial patient-reported outcome informing opioid prescribing guidelines, but its collection is resource-intensive and vulnerable to inaccuracy due to nonresponse bias.

Methods

We developed a post-discharge text message-to-web survey system for efficient collection of patient-reported pain outcomes. We prospectively recruited surgical patients at Beth Israel Deaconess Medical Center in Boston, Massachusetts from March 2019 through October 2020, sending an SMS link to a secure web survey to quantify opioids consumed after discharge from hospitalization. Patient factors extracted from the electronic health record were tested for nonresponse bias and observable confounding. Following targeted learning-based nonresponse adjustment, procedure-specific opioid consumption quantiles (medians and 75th percentiles) were estimated and compared to a previous telephone-based reference survey.

Results

6553 patients were included. Opioid consumption was measured in 44% of patients (2868), including 21% (1342) through survey response. Characteristics associated with inability to measure opioid consumption included age, tobacco use, and prescribed opioid dose. Among the 10 most common procedures, median consumption was only 36% of the median prescription size; 64% of prescribed opioids were not consumed. Among those procedures, nonresponse adjustment corrected the median opioid consumption by an average of 37% (IQR: 7, 65%) compared to unadjusted estimates, and corrected the 75th percentile by an average of 5% (IQR: 0, 12%). This brought median estimates for 5/10 procedures closer to telephone survey-based consumption estimates, and 75th percentile estimates for 2/10 procedures closer to telephone survey-based estimates.

Conclusions

SMS-recruited online surveying can generate reliable opioid consumption estimates after nonresponse adjustment using patient factors recorded in the electronic health record, protecting patients from the risk of inaccurate prescription guidelines.

Education as drug policy: A realist synthesis of continuing professional development for opioid agonist therapy

BACKGROUND

Continuing professional development (CPD) for opioid agonist therapy (OAT) has been identified as a key health policy strategy to improve care for people living with opioid use disorder (OUD) and to address rising opioid-related harms. To design and deliver effective CPD programs, there is a need to clarify how they work within complex health system and policy contexts. This review synthesizes the literature on OAT CPD programs and educational theory to clarify which interventions work, for whom, and in what contexts.

METHODS

A systematic review and realist synthesis of evaluations of CPD programs focused on OAT was conducted. This included record identification and screening, theory familiarization, data collection, analysis, expert consultation, and iterative context-intervention-mechanism-outcome (CIMO) configuration development.

RESULTS

Twenty-four reports comprising 21 evaluation studies from 5 countries for 3373 providers were reviewed. Through iterative testing of included studies with relevant theory, five CIMO configurations were developed. The programs were categorized by who drove the learning outcomes (i.e., system/policy, instructor, learner) and their spheres of influence (i.e., micro, meso, macro). There was a predominance of instructor-driven programs driving change at the micro level, with few policy-driven macro-influential programs, inconsistent with the promotion of CPD as a clear opioid crisis policy-level intervention.

CONCLUSION

OAT CPD is challenged by mismatches in program justifications, objectives, activities, and outcomes. Depending on how these program factors interact, OAT CPD can operate as a barrier or facilitator to OUD care. With more deliberate planning and consideration of program theory, programs more directly addressing diverse learner and system needs may be developed and delivered. OAT CPD as drug policy does not operate in isolation; programs may feed into each other and intercalate with other policy initiatives to have micro, meso, and macro impacts on educational and population health outcomes.

Oxycodone-related deaths in Sweden 2006-2018

AIM

To identify and characterize oxycodone related deaths in Sweden from 2006 to 2018 and to compare them to other opioid-related deaths.

METHODS

To assess the factors contributing to the deaths, we used multinomial logistic regression to compare oxycodone-related deaths extracted from all forensic autopsy examinations and toxicology cases in the age groups 15-34 (reference group), 35-54 and 55-74 with regard to sex, presence of benzodiazepines and alcohol at the time of death, prescription of oxycodone, benzodiazepines and antidepressants, previous substance use-related (SUD) treatment, and manner of death. The oxycodone related deaths were compared with deaths with presence of other opioids.

RESULT

We identified 575 oxycodone-related deaths, and the rate increased during the study period from 0.10 to 1.12 per 100,000 in parallel with an increase of oxycodone prescriptions from 3.17 to 30.33 per 1000. Oxycodone-related deaths amounted to 10.0% of all opioid-related deaths. The deaths occurred mainly in older patients previously being prescribed oxycodone. Benzodiazepines were present at the time of death in 403 (70%) and alcohol in 259 (45%). Prescriptions of any opioid for pain (61%), oxycodone (50%), benzodiazepines (67%) and antidepressants (55%) were common. Only 15% had received treatment for SUD during the last year.

CONCLUSION

Oxycodone-related deaths increased in Sweden between 2006 and 2018 in parallel to an increase in oxycodone prescriptions. The increase occurred mainly in older patients being prescribed oxycodone for pain. There might be specific interventions needed to avoid oxycodone-related deaths compared to other opioid-related deaths associated with illicit opioid use.

Machine Learning for Predicting Risk of Early Dropout in a Recovery Program for Opioid Use Disorder

Background: An increase in opioid use has led to an opioid crisis during the last decade, leading to declarations of a public health emergency. In response to this call, the Houston Emergency Opioid Engagement System (HEROES) was established and created an emergency access pathway into long-term recovery for individuals with an opioid use disorder. A major contributor to the success of the program is retention of the enrolled individuals in the program. Methods: We have identified an increase in dropout from the program after 90 and 120 days. Based on more than 700 program participants, we developed a machine learning approach to predict the individualized risk for dropping out of the program. Results: Our model achieved sensitivity of 0.81 and specificity of 0.65 for dropout at 90 days and improved the performance to sensitivity of 0.86 and specificity of 0.66 for 120 days. Additionally, we identified individual risk factors for dropout, including previous overdose and relapse and improvement in reported quality of life. Conclusions: Our informatics approach provides insight into an area where programs may allocate additional resources in order to retain high-risk individuals and increase the chances of success in recovery.

Using administrative data to predict cessation risk and identify novel predictors among new entrants to opioid agonist treatment

BACKGROUND

Longer retention in opioid agonist treatment (OAT) is associated with improved treatment outcomes but 12-month retention rates are often low. Innovative approaches are needed to strengthen retention in OAT. We develop and compare traditional and deep learning-extensions of Cox regression to examine the potential for predicting time in OAT at individuals' first episode entry.

METHODS

Retrospective cohort study in New South Wales, Australia including 16,576 people entering OAT for the first time between January 2006 and December 2017. We develop 12-month OAT cessation prediction models using traditional and deep learning-extensions of the Cox regression algorithm with predictors evaluated from linked administrative datasets. Proportion of explained variation, calibration, and discrimination are compared using 5 × 2 cross-validation.

RESULTS

Twelve-month cessation rate was 58.4%. The largest hazard ratios for earlier cessation from the deep learning model were observed for treatment factors, including private dosing points (HR=1.54, 95% CI=1.49-1.60) and buprenorphine medication (HR=1.43, 95% CI=1.39-1.46). Diagnostic codes for homelessness (HR=1.09, 95% CI=1.04-1.13), outpatient treatment for drug use disorders (HR=1.10, 95% CI=1.06-1.15), and occupant of vehicle accident (HR=1.04, 95% CI=1.01-1.07) from past-year health service presentations were identified as significant predictors of retention. We observed no improvement in performance of the deep learning model over traditional Cox regression.

CONCLUSIONS

Deep learning may be more useful in identifying novel risk factors of OAT retention from administrative data than evaluating individual-level risk. An increased focus on addressing structural issues at the population level and considering alternate models of care may be more effective at improving retention than delivering fully personalised OAT.

Mapping drug epidemiology futures

Epidemiology is a core discipline generating evidence to inform and drive drug policy. In this essay, we speculate on what the future of drug epidemiology might become. We highlight for attention two areas shaping the future of drug epidemiology: nesting epidemiology within a 'syndemic' and 'relational' approach; and innovating in relation to causal inference in the face of complexity. We argue that shifts towards a more relational approach emphasise contingency, including in relation to how drugs might constitute benefit or harm. This leads us to speculate on a 'positive epidemiology'; one that is configured not merely in relation to harm but also in relation to the potential benefits of drugs in relation to well-being. In responding to the complex challenges of delineating contingent causalities, we emphasise the potential of carefully conducted observational study designs that go beyond statistical associations to test causal inference. We acknowledge that each of these developments we describe - a shift towards more relational approaches which emphasise contingent causation, and methodological innovations in relation to establishing causal inference - can be at odds with the other in how they imagine drug epidemiology futures.

Futures-oriented drugs policy research: Events, trends, and speculating on what might become

One concern in the field of drugs policy is how to make research more futures-oriented. Tracing trends and events with the potential to alter drug futures are seen as ways of becoming more prepared. This challenge is made complex in fast evolving drug markets which entangle with shifting social and material relations at global scale. In this analysis, we argue that drugs policy research orientates to detection and discovery based on the recent past. This narrows future-oriented analyses to the predictable and probable, imagined as extensions of the immediate and local present. We call for a more speculative approach; one which extends beyond the proximal, and one which orientates to possibilities rather than probabilities. Drawing on ideas on speculation from science and technology and futures studies, we argue that speculative research holds potential for more radical alterations in drugs policy. We encourage research approaches which not only valorise knowing in relation to what might happen but which conduct experiments on what could be. Accordingly, we trace how speculative research makes a difference by altering the present through making deliberative interventions on alternative policy options, including policy scenarios which make a radical break with the present. We look specifically at the 'Big Event' and 'Mega Trend' as devices of speculative intervention in futures-oriented drugs policy research. We illustrate how the device of Mega Trend helps to trace as well as to speculate on some of the entangling elements affecting drug futures, including in relation to climate, environment, development, population, drug production, digitalisation, biotechnology, policy and discourse.