1
|
Yu Z, Peng C, Yang X, Dang C, Adekkanattu P, Gopal Patra B, Peng Y, Pathak J, Wilson DL, Chang CY, Lo-Ciganic WH, George TJ, Hogan WR, Guo Y, Bian J, Wu Y. Identifying social determinants of health from clinical narratives: A study of performance, documentation ratio, and potential bias. J Biomed Inform 2024; 153:104642. [PMID: 38621641 PMCID: PMC11141428 DOI: 10.1016/j.jbi.2024.104642] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 10/25/2023] [Revised: 04/09/2024] [Accepted: 04/12/2024] [Indexed: 04/17/2024]
Abstract
OBJECTIVE To develop a natural language processing (NLP) package to extract social determinants of health (SDoH) from clinical narratives, examine the bias among race and gender groups, test the generalizability of extracting SDoH for different disease groups, and examine population-level extraction ratio. METHODS We developed SDoH corpora using clinical notes identified at the University of Florida (UF) Health. We systematically compared 7 transformer-based large language models (LLMs) and developed an open-source package - SODA (i.e., SOcial DeterminAnts) to facilitate SDoH extraction from clinical narratives. We examined the performance and potential bias of SODA for different race and gender groups, tested the generalizability of SODA using two disease domains including cancer and opioid use, and explored strategies for improvement. We applied SODA to extract 19 categories of SDoH from the breast (n = 7,971), lung (n = 11,804), and colorectal cancer (n = 6,240) cohorts to assess patient-level extraction ratio and examine the differences among race and gender groups. RESULTS We developed an SDoH corpus using 629 clinical notes of cancer patients with annotations of 13,193 SDoH concepts/attributes from 19 categories of SDoH, and another cross-disease validation corpus using 200 notes from opioid use patients with 4,342 SDoH concepts/attributes. We compared 7 transformer models and the GatorTron model achieved the best mean average strict/lenient F1 scores of 0.9122 and 0.9367 for SDoH concept extraction and 0.9584 and 0.9593 for linking attributes to SDoH concepts. There is a small performance gap (∼4%) between Males and Females, but a large performance gap (>16 %) among race groups. The performance dropped when we applied the cancer SDoH model to the opioid cohort; fine-tuning using a smaller opioid SDoH corpus improved the performance. The extraction ratio varied in the three cancer cohorts, in which 10 SDoH could be extracted from over 70 % of cancer patients, but 9 SDoH could be extracted from less than 70 % of cancer patients. Individuals from the White and Black groups have a higher extraction ratio than other minority race groups. CONCLUSIONS Our SODA package achieved good performance in extracting 19 categories of SDoH from clinical narratives. The SODA package with pre-trained transformer models is available at https://github.com/uf-hobi-informatics-lab/SODA_Docker.
Collapse
Affiliation(s)
- Zehao Yu
- Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, FL, USA
| | - Cheng Peng
- Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, FL, USA; Cancer Informatics Shared Resource, University of Florida Health Cancer Center, Gainesville, FL, USA
| | - Xi Yang
- Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, FL, USA; Cancer Informatics Shared Resource, University of Florida Health Cancer Center, Gainesville, FL, USA
| | - Chong Dang
- Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, FL, USA
| | - Prakash Adekkanattu
- Information Technologies and Services, Weill Cornell Medicine, New York, NY, USA
| | - Braja Gopal Patra
- Department of Population Health Sciences, Weill Cornell Medicine, New York, NY, USA
| | - Yifan Peng
- Department of Population Health Sciences, Weill Cornell Medicine, New York, NY, USA
| | - Jyotishman Pathak
- Department of Population Health Sciences, Weill Cornell Medicine, New York, NY, USA
| | - Debbie L Wilson
- Department of Pharmaceutical Outcomes & Policy, College of Pharmacy, University of Florida, Gainesville, FL 32611, USA
| | - Ching-Yuan Chang
- Department of Pharmaceutical Outcomes & Policy, College of Pharmacy, University of Florida, Gainesville, FL 32611, USA
| | - Wei-Hsuan Lo-Ciganic
- Department of Pharmaceutical Outcomes & Policy, College of Pharmacy, University of Florida, Gainesville, FL 32611, USA
| | - Thomas J George
- Division of Hematology & Oncology, Department of Medicine, College of Medicine, University of Florida, Gainesville, FL, USA
| | - William R Hogan
- Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, FL, USA
| | - Yi Guo
- Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, FL, USA; Cancer Informatics Shared Resource, University of Florida Health Cancer Center, Gainesville, FL, USA
| | - Jiang Bian
- Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, FL, USA; Cancer Informatics Shared Resource, University of Florida Health Cancer Center, Gainesville, FL, USA
| | - Yonghui Wu
- Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, FL, USA; Cancer Informatics Shared Resource, University of Florida Health Cancer Center, Gainesville, FL, USA.
| |
Collapse
|
2
|
Hancock DW, Burnes DPR, Pillemer KA, Czaja SJ, Lachs MS. Psychometric Properties of The Five-Item Victimization of Exploitation (FIVE) Scale: A Measure of Financial Abuse of Older Adults. THE GERONTOLOGIST 2023; 63:993-999. [PMID: 35395679 PMCID: PMC10653197 DOI: 10.1093/geront/gnac048] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 02/04/2022] [Indexed: 11/12/2022] Open
Abstract
BACKGROUND AND OBJECTIVES Elder mistreatment affects at least 1 in 10 older adults. Financial abuse, or exploitation, of older adults is among the most commonly reported forms of abuse. Few validated measures exist to measure this construct. We aim to present a new psychometrically validated measure of financial abuse of older adults. RESEARCH DESIGN AND METHODS Classical test theory and item response theory (IRT) methodologies were used to examine a five-item measure of financial abuse of older adults, administered as part of the New York State Elder Mistreatment Survey. RESULTS Factor analysis revealed a single factor best fits the data, which we labeled as financial abuse. Moreover, IRT analyses revealed that these items discriminated well between abused and nonabused persons and provided information at high levels of the latent trait θ, as is expected in cases of abuse. DISCUSSION AND IMPLICATIONS The Five-Item Victimization of Exploitation Scale has acceptable psychometric properties and has been used successfully in large-scale survey research. We recommend this measure as an indicator of financial abuse in elder abuse, or mistreatment prevalence research studies.
Collapse
Affiliation(s)
- David W Hancock
- Department of Geriatrics and Palliative Medicine, Weill Cornell Medical College, New York, New York, USA
| | - David P R Burnes
- Factor-Inwentash School of Social Work, University of Toronto, Toronto, Canada
| | - Karl A Pillemer
- College of Human Ecology, Cornell University, Ithaca, New York, USA
| | - Sara J Czaja
- Department of Geriatrics and Palliative Medicine, Weill Cornell Medical College, New York, New York, USA
| | - Mark S Lachs
- Department of Geriatrics and Palliative Medicine, Weill Cornell Medical College, New York, New York, USA
| |
Collapse
|
3
|
Burnes D, Pillemer K, Rosen T, Lachs MS, McDonald L. Elder abuse prevalence and risk factors: findings from the Canadian Longitudinal Study on Aging. NATURE AGING 2022; 2:784-795. [PMID: 37118505 PMCID: PMC10154033 DOI: 10.1038/s43587-022-00280-2] [Citation(s) in RCA: 1] [Impact Index Per Article: 0.5] [Reference Citation Analysis] [Abstract] [Key Words] [MESH Headings] [Grants] [Track Full Text] [Subscribe] [Scholar Register] [Received: 03/11/2022] [Accepted: 08/04/2022] [Indexed: 04/30/2023]
Abstract
Elder abuse (EA) is a pervasive problem with serious consequences. Previous population-based EA risk factor research has largely used cross-sectional designs that limit causal inferences, or agency records to identify victims, which threatens external validity. Based on a national, prospective, population-based cohort sample of older adults (n = 23,468) over a 3-year period from the Canadian Longitudinal Study on Aging, the current study sought to estimate the prevalence of EA and identify risk and protective factors. Past-year prevalence of any EA was 10.0%. Older adults with greater vulnerability related to physical, cognitive and mental health, childhood maltreatment and shared living were at higher EA risk, while social support was protective against EA. Older adults identifying as Black or reporting financial need were at heightened EA risk. This longitudinal, population-based study advances our understanding of EA risk/protective factors across several domains and informs the development of EA prevention strategies.
Collapse
Affiliation(s)
- David Burnes
- University of Toronto, Factor-Inwentash Faculty of Social Work, Toronto, Ontario, Canada.
| | - Karl Pillemer
- Cornell University, College of Human Ecology, Ithaca, NY, USA
| | - Tony Rosen
- Division of Geriatric Emergency Medicine, Department of Emergency Medicine, Weill Cornell Medicine, New York, NY, USA
| | - Mark S Lachs
- Geriatrics and Palliative Medicine, Weill Cornell Medicine, New York, NY, USA
| | - Lynn McDonald
- University of Toronto, Factor-Inwentash Faculty of Social Work, Toronto, Ontario, Canada
| |
Collapse
|
4
|
Yip SH, O'Connor D, Shakeri Shemirani F, Brown A, Hyman J. Emergency assistance in situations of abuse, neglect, and self-neglect: exploring the complexity and challenges. J Elder Abuse Negl 2022; 34:124-151. [PMID: 35470776 DOI: 10.1080/08946566.2022.2070317] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 10/18/2022]
Abstract
Health and social care professionals are often called upon to provide emergency intervention without the adult's consent in situations of abuse, neglect, and self-neglect. Little is known about this process despite implications related to health care costs and individual rights. In this qualitative study, 17 health care professionals with experience enacting emergency legislation in BC were interviewed to better understand what leads to an emergency response and how these professionals carry out their role. Five components emerged in the provision of emergency assistance: assessing intolerable risks, assessing incapability, balancing ethical values, exploring resources, and consulting/collaborating. Attention is drawn to the significance of social location, including age, socio-economic, and Indigenous background for influencing the process, sometimes in unintended ways. The challenges in providing emergency interventions may result in social inequities and delay in providing care at times. The importance of bringing a reflexive and intersectional lens to providing interventions is highlighted.
Collapse
Affiliation(s)
- So Han Yip
- Manager at the ReAct Adult Protection Program, Vancouver Coastal Health, West Vancouver, British Columbia, Canada
| | - Deborah O'Connor
- Professor at the School of Social Work, University of British Columbia and Co-Director at the Centre for Research on Personhood in Dementia, Vancouver, BC, Canada
| | - Farimah Shakeri Shemirani
- Social worker at Vancouver Coastal Health, North Shore Palliative Care Program, Vancouver, BC, Canada
| | - Amanda Brown
- Director at the ReAct Adult Protection Program, Vancouver Coastal Health, West Vancouver, BC, Canada
| | - Jenny Hyman
- Social Work Site Leader at Providence Health Care, Mount St. Joseph Hospital, Vancouver, BC, Canada
| |
Collapse
|
5
|
Pillemer K, Burnes D, Hancock D, Eckenrode J, Rosen T, MacNeil A, Lachs MS. Lack of Association of Elder Mistreatment With Mortality. J Gerontol A Biol Sci Med Sci 2021; 77:1699-1705. [PMID: 34939085 PMCID: PMC9373958 DOI: 10.1093/gerona/glab348] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.7] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 07/23/2021] [Indexed: 11/14/2022] Open
Abstract
BACKGROUND Prior research is limited and inconsistent on the degree to which elder mistreatment (EM) is associated with mortality. This study uses data from a 10-year, prospective, population-based study of EM to determine the adjusted effects of EM on older adult mortality, after controlling for other health and socioeconomic covariates. METHODS The New York State Elder Mistreatment Prevalence Study conducted a random-sample telephone survey of older adults (n = 4 156) in 2009 (Wave 1). The current study employs EM and covariate data from Wave 1 and data on mortality status through Wave 2 (2019). EM was operationalized both as experiencing EM and as severity of EM. The survey measured overall EM and separate subtypes (emotional, physical, and financial abuse, and neglect). RESULTS The hypothesis was not supported that abused and neglected older people would have higher rates of death over the study. Individuals who were victims of EM were no more likely to die over the following 10 years, compared with those who were not mistreated, after controlling for covariates. Furthermore, the severity of EM, as measured by the frequency of mistreatment behaviors, also was not associated with mortality risk. CONCLUSIONS The finding that self-reported EM did not raise the risk of earlier death in this sample is encouraging. Future research should work to identify factors that may moderate the relationship between EM and mortality, such as social support/isolation, quality of family relationships, or involvement with formal support service systems.
Collapse
Affiliation(s)
- Karl Pillemer
- Department of Human Development, Cornell University, Ithaca, New York, USA.,Weill Cornell Medicine, Division of Geriatrics and Palliative Medicine, New York, New York, USA
| | - David Burnes
- Factor-Inwentash Faculty of Social Work, University of Toronto, Toronto, Ontario, Canada
| | - David Hancock
- Weill Cornell Medicine, Division of Geriatrics and Palliative Medicine, New York, New York, USA
| | - John Eckenrode
- Department of Human Development, Cornell University, Ithaca, New York, USA
| | - Tony Rosen
- Department of Emergency Medicine, Weill Cornell Medicine/NewYork-Presbyterian Hospital, New York, New York, USA
| | - Andie MacNeil
- Factor-Inwentash Faculty of Social Work, University of Toronto, Toronto, Ontario, Canada
| | - Mark S Lachs
- Weill Cornell Medicine, Division of Geriatrics and Palliative Medicine, New York, New York, USA
| |
Collapse
|