Page 135 - AIH-1-2

Artificial Intelligence in Health SDoH in clinical narratives

leading to biased estimates of the effect size. Despite this Acknowledgments
risk, the extensive inclusion of variables in our model was
a deliberate choice, reflective of the exploratory nature None.
of our research. This project aimed to uncover existing Funding
relationships and identify factors potentially associated
with SDoH mentions in the literature. To mitigate the risk This work has been funded by John Snow Labs Inc.
of arbitrary variable selection, we employed a stepwise Conflict of interest
approach, including only variables with p-values of 0.001
or less, ensuring that each variable included in the model The authors declare that they have no competing interests.
contributed significantly to the explanatory power of our
analysis. Author contributions
However, we acknowledge that understanding Conceptualization: Julio Bonis, Veysel Kocaman
the causality behind these associations requires more Formal Analysis: Julio Bonis
sophisticated modeling techniques. Our findings provide Investigation: Julio Bonis, Veysel Kocaman
the foundation for future research endeavors and in-depth Methodology: Julio Bonis, David Talby
studies that can employ more advanced statistical models Writing – Original Draft: Julio Bonis
to unravel the causal pathways linking SDoH to health Writing – Review & Editing: Veysel Kocaman, David Talby
outcomes. These studies will be crucial for developing
targeted interventions and policies aimed at addressing Ethics approval and consent to participate
SDoH more effectively within health-care practices and Not applicable.
research.
Consent for publication
5. Conclusion Not applicable.

The limited mentions of SDoH in clinical case reports
underscore the necessity for better SDoH integration into Availability of data
medical documentation. To mitigate biases in statistical The dataset utilized for the logistic regression analysis will
analyses using clinical notes or medical journal content, be made available upon publication. Interested parties can
consistent recording and reporting of SDoH are essential. obtain access for academic purposes by directly contacting
Spark NLP offers promising avenues for enhancing the the authors (julio@johnsnowlabs.com or veysel@
extraction and analysis of SDoH from EHRs, highlighting johnsnowlabs.com) and signing a data access agreement.
the importance of AI model development to prevent biases
that could negatively affect health-care fairness and delivery. References
For future research, conducting a similar analysis on 1. McGinnis JM, Williams-Russo P, Knickman JR. The case for
the factors associated with SDoH mentions in the full more active policy attention to health promotion. Health Aff
texts of clinical case reports could yield deeper insights. (Millwood). 2002;21:78-93.
In addition, analyzing actual EHR notes to compare the doi: 10.1377/hlthaff.21.2.78
prevalence and representation of SDoH across different 2. Galea S, Tracy M, Hoggatt KJ, DiMaggio C, Karpati A.
specialties or health-care centers could provide valuable Estimated deaths attributable to social factors in the United
information. Such comparative studies could elucidate States. Am J Public Health. 2011;101:1456-1465.
the representation and documentation of SDoH across
various health-care settings, potentially guiding targeted doi: 10.2105/AJPH.2010.300086
interventions and policy changes to promote equitable 3. Hatef E, Kharrazi H, Nelson K, et al. The association between
health-care outcomes. neighborhood socioeconomic and housing characteristics
with hospitalization: Results of a national study of Veterans.
In conclusion, enhancing the documentation and J Am Board Fam Med. 2019;32:890-903.
representation of SDoH in the medical literature is critical
for advancing toward more informed, equitable, and doi: 10.3122/jabfm.2019.06.190138
effective health-care practices and policies. Future studies 4. Hood CM, Gennuso KP, Swain GR, Catlin BB. County
focused on expanding the scope of analysis to full texts and health rankings: Relationships between determinant factors
EHRs could significantly contribute to our understanding and health outcomes. Am J Prev Med. 2016;50:129-135.
and implementation of SDoH in clinical care. doi: 10.1016/j.amepre.2015.08.024

Volume 1 Issue 2 (2024) 129 doi: 10.36922/aih.2737