Detecting substance-related problems in narrative investigation summaries of child abuse and neglect using text mining and machine learning

Natural language processing-driven state machines to extract social factors from unstructured clinical documentation

Objective

This study sought to create natural language processing algorithms to extract the presence of social factors from clinical text in 3 areas: (1) housing, (2) financial, and (3) unemployment. For generalizability, finalized models were validated on data from a separate health system for generalizability.

Materials and Methods

Notes from 2 healthcare systems, representing a variety of note types, were utilized. To train models, the study utilized n-grams to identify keywords and implemented natural language processing (NLP) state machines across all note types. Manual review was conducted to determine performance. Sampling was based on a set percentage of notes, based on the prevalence of social need. Models were optimized over multiple training and evaluation cycles. Performance metrics were calculated using positive predictive value (PPV), negative predictive value, sensitivity, and specificity.

Results

PPV for housing rose from 0.71 to 0.95 over 3 training runs. PPV for financial rose from 0.83 to 0.89 over 2 training iterations, while PPV for unemployment rose from 0.78 to 0.88 over 3 iterations. The test data resulted in PPVs of 0.94, 0.97, and 0.95 for housing, financial, and unemployment, respectively. Final specificity scores were 0.95, 0.97, and 0.95 for housing, financial, and unemployment, respectively.

Discussion

We developed 3 rule-based NLP algorithms, trained across health systems. While this is a less sophisticated approach, the algorithms demonstrated a high degree of generalizability, maintaining >0.85 across all predictive performance metrics.

Conclusion

The rule-based NLP algorithms demonstrated consistent performance in identifying 3 social factors within clinical text. These methods may be a part of a strategy to measure social factors within an institution.

Harnessing Machine Learning in Tackling Domestic Violence-An Integrative Review

Domestic violence (DV) is a public health crisis that threatens both the mental and physical health of people. With the unprecedented surge in data available on the internet and electronic health record systems, leveraging machine learning (ML) to detect obscure changes and predict the likelihood of DV from digital text data is a promising area health science research. However, there is a paucity of research discussing and reviewing ML applications in DV research.

METHODS

We extracted 3588 articles from four databases. Twenty-two articles met the inclusion criteria.

RESULTS

Twelve articles used the supervised ML method, seven articles used the unsupervised ML method, and three articles applied both. Most studies were published in Australia (n = 6) and the United States (n = 4). Data sources included social media, professional notes, national databases, surveys, and newspapers. Random forest (n = 9), support vector machine (n = 8), and naïve Bayes (n = 7) were the top three algorithms, while the most used automatic algorithm for unsupervised ML in DV research was latent Dirichlet allocation (LDA) for topic modeling (n = 2). Eight types of outcomes were identified, while three purposes of ML and challenges were delineated and are discussed.

CONCLUSIONS

Leveraging the ML method to tackle DV holds unprecedented potential, especially in classification, prediction, and exploration tasks, and particularly when using social media data. However, adoption challenges, data source issues, and lengthy data preparation times are the main bottlenecks in this context. To overcome those challenges, early ML algorithms have been developed and evaluated on DV clinical data.

A text-based approach to measuring opioid-related risk among families involved in the child welfare system

BACKGROUND

The public health significance of the opioid epidemic is well-established. However, few states collect data on opioid problems among families involved in child welfare services. The absence of data creates significant barriers to understanding the impact of opioids on the service system and the needs of families being served.

OBJECTIVE

This study sought to validate binary and count-based indicators of opioid-related maltreatment risk based on mentions of opioid use in written child welfare summaries.

DATA AND PROCEDURES

We developed a comprehensive list of terms referring to opioid street drugs and pharmaceuticals. This terminology list was used to scan and flag investigator summaries from an extensive collection of investigations (N = 362,754) obtained from a state-based child welfare system in the United States. Associations between mentions of opioid use and investigators' decisions to substantiate maltreatment and remove a child from home were tested within a framework of a priori hypotheses.

RESULTS

Approximately 6.3% of all investigations contained one or more opioid use mentions. Opioid mentions exhibited practically signficant associations with investigator decisions. One in ten summaries that were substantiated had an opioid mention. One in five investigations that led to the out-of-home placement of a child contained an opioid mention.

CONCLUSION

This study demonstrates the feasibility of using simple text mining procedures to extract information from unstructured text documents. These methods provide novel opportunities to build insights into opioid-related problems among families involved in a child welfare system when structured data are not available.

Developing Prediction Model for Children’s Social Competence Using Machine Learning

Objectives: This study aims to identify the types of latent classes of children’s social competence, and to develop a model using machine learning to predict the type and identify relatively important variables.Methods: Data were collected from 466 children aged three to five years and their mothers. Children’s social competence was classified by level. Latent class analysis, machine learning model construction, and performance evaluation were performed using R 3.6.1 and R-Studio 1.2.5033. The machine learning algorithms used were logistic regression, lasso logistic regression, random forest, and gradient-boosted decision tree models.Results: First, according to the characteristics of the latent class of children’s social competence, it was classified into two types: ‘high level’ and ‘low level’. Second, a machine learning algorithm was applied according to the latent class. The best performing model was the random forest model. Third, the most important variable in predicting the social competence type was identified as ‘harm avoidance’ in the children’s temperament. Fourth, another major variable was a ‘shift’ in the children’s executive functions.Conclusion: This study is meaningful as it suggests the possibility of predicting and discriminating children’s social competence and various developmental aspects by applying machine learning, the latest technique, to predict the types of children’s social competence.

Extracting social determinants of health from electronic health records using natural language processing: a systematic review

OBJECTIVE

Social determinants of health (SDoH) are nonclinical dispositions that impact patient health risks and clinical outcomes. Leveraging SDoH in clinical decision-making can potentially improve diagnosis, treatment planning, and patient outcomes. Despite increased interest in capturing SDoH in electronic health records (EHRs), such information is typically locked in unstructured clinical notes. Natural language processing (NLP) is the key technology to extract SDoH information from clinical text and expand its utility in patient care and research. This article presents a systematic review of the state-of-the-art NLP approaches and tools that focus on identifying and extracting SDoH data from unstructured clinical text in EHRs.

MATERIALS AND METHODS

A broad literature search was conducted in February 2021 using 3 scholarly databases (ACL Anthology, PubMed, and Scopus) following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A total of 6402 publications were initially identified, and after applying the study inclusion criteria, 82 publications were selected for the final review.

RESULTS

Smoking status (n = 27), substance use (n = 21), homelessness (n = 20), and alcohol use (n = 15) are the most frequently studied SDoH categories. Homelessness (n = 7) and other less-studied SDoH (eg, education, financial problems, social isolation and support, family problems) are mostly identified using rule-based approaches. In contrast, machine learning approaches are popular for identifying smoking status (n = 13), substance use (n = 9), and alcohol use (n = 9).

CONCLUSION

NLP offers significant potential to extract SDoH data from narrative clinical notes, which in turn can aid in the development of screening tools, risk prediction models, and clinical decision support systems.

Enabling remote learning system for virtual personalized preferences during COVID-19 pandemic

The education system worldwide has been affected by the Corona Virus Diseases 2019 (COVID-19) pandemic, resulting in the interruption of all educational institutions. Moreover, as a precautionary measure, the lockdown has been imposed that has severely affected the learning processes, especially assessment activities, including exams and viva. In such challenging situations, E-learning platforms could play a vital role in conducting seamless academic activities. In spite of all the advantages of remote learning systems, many hurdles and obstacles, like a selection of suitable learning resources/material encounter by individual users based on their interests or requirements. Especially those who are not well familiar with the internet technology in developing countries and are in need of a platform that could help them in resolving the issues related to the online virtual environment. Therefore, in this work, we have proposed a mechanism that intelligently and correctly predicts the appropriate preferences for the selection of resources relevant to a specific user by considering the capabilities of diverse perspectives users to provide quality online education and to make work from home policy more effective and progressive during the pandemic. The proposed system helps teachers in providing quality online education, familiarizing them with advanced technology in the online environment. It also semantically predicts the preferences for virtual assistance of those users who are in need of learning the new tools and technologies in short time as per their institutional requirements in order to meet the quality standards of online education. The experimental and statistical results have demonstrated that the proposed virtual personalized preferences system has improved overall academic activities as compared to the current method. The proposed system enhanced user's learning abilities and facilitated them in selecting short courses while using different online education tools adopted/suggested by the institutions to conduct online classes/seminars/webinars etc., as compared to the conventional classes/activities.

Predicting youth at high risk of aging out of foster care using machine learning methods

BACKGROUND

Youth who exit the nation's foster care system without permanency are at high risk of experiencing difficulties during the transition to adulthood.

OBJECTIVE

To present an illustrative test of whether an algorithmic decision aid could be used to identify youth at risk of existing foster care without permanency.

METHODS

For youth placed in foster care between ages 12 and 14, we assessed the risk of exiting care without permanency by age 18 based on their child welfare service involvement history. To develop predictive risk models, 28 years (1991-2018) of child welfare service records from California were used. Performances were evaluated using F1, AUC, and precision and recall scores at k %. Algorithmic racial bias and fairness was also examined.

RESULTS

The gradient boosting decision tree and random forest showed the best performance (F1 score = .54-.55, precision score = .62, recall score = .49). Among the top 30 % of youth the model identified as high risk, half of all youth who exited care without permanency were accurately identified four to six years prior to their exit, with a 39 % error rate. Although racial disparities between Black and White youth were observed in imbalanced error rates, calibration and predictive parity were satisfied.

CONCLUSIONS

Our study illustrates the manner in which potential applications of predictive analytics, including those designed to achieve universal goals of permanency through more targeted allocations of resources, can be tested. It also assesses the model using metrics of fairness.

Do Aftercare Services Take Young People's Health Behaviour into Consideration? A Retrospective Document Analysis from Finland

BACKGROUND

Health disparities among children and young people predict health disparities in adulthood and cause a long-term detrimental impact on a person's quality of life as well as increased costs to society.

PURPOSE

The purpose of the study was to describe the health behaviour of Finnish young people (n = 600) who had left aftercare services by the end of April 2015 based on their electronic patient records.

DESIGN AND METHODS

A retrospective document analysis. Data were collected from the register by a structured worksheet designed for this study and analysed using descriptive statistical methods.

RESULTS

The entries made in the aftercare participants' records concerning their weight, sleep and rest, smoking and sexual health were insufficient. The documents contained clearly more detailed reports of their substance use. Women were affected more than men by lifestyles other than substance abuse endangering their health. An increase in the number of out-of-home care placements produces a spike in risky behaviour.

CONCLUSIONS

The young people in aftercare are in an unequal position compared to the mainstream population in the area of health promotion. Their risky behaviour is not identified or taken into consideration.

PRACTICE IMPLICATIONS

Aftercare and the related multiprofessional collaboration must be developed to prevent risks related to the young persons` health and to take these comprehensively into account.

Psychiatric Disorders Among Older Black Americans: Within- and Between-Group Differences

Psychiatric disorders impose significant personal, social, and financial costs for individuals, families, and the nation. Despite a large amount of research and several journals focused on psychiatric conditions, there is a paucity of research on psychiatric disorders among Black Americans (i.e., African Americans and Black Caribbeans), particularly older Black Americans. The present literature review examines research on psychiatric disorders among older Black Americans and provides a broad overview of research findings that are based on nationally representative studies. Collectively, this research finds: (1) older African Americans have lower rates of psychiatric disorders than younger African Americans; (2) family support is not protective of psychiatric disorders, whereas negative interaction with family members is a risk factor; (3) everyday discrimination is a risk factor for psychiatric disorders; (4) both older African Americans and African American across the adult age range have lower prevalence rates of psychiatric disorders than non-Latino whites; (5) Black Caribbean men have particularly high rates of depression, posttraumatic stress disorder, and suicide attempts; and (6) a significant proportion of African American older adults with mental health disorders do not receive professional help. This literature review also discusses the “Race Paradox” in mental health, the Environmental Affordances Model, and the importance of investigating ethnicity differences among Black Americans. Future research directions address issues that are directly relevant to the Black American population and include the following: (1) understanding the impact of mass incarceration on the psychiatric disorders of prisoners’ family members, (2) assessing the impact of immigration from African countries for ethnic diversity within the Black American population, (3) examining the impact of racial identity and racial socialization as potential protective factors for psychiatric morbidities, and (4) assessing racial diversity in life-course events and their impact on mental health.