A Data Science-based approach to Identify Social Determinants of Health Impacting Access to Pediatric Radiology

doi:10.21203/rs.3.rs-4674294/v1

Download PDF

Research Article

A Data Science-based approach to Identify Social Determinants of Health Impacting Access to Pediatric Radiology

https://doi.org/10.21203/rs.3.rs-4674294/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background

Research on healthcare disparities in pediatric radiology is limited, leading to the persistence of missed care opportunities (MCO).

Objective

Evaluate the social determinants of health and sociodemographic factors related to pediatric radiology MCO before, during, and after COVID-19 pandemic.

Materials and Methods

The study examined all outpatient pediatric radiology exams at a pediatric medical center and its affiliate centers from 03/08/19 to 06/07/21, to identify missed care opportunities. Logistic regression with LASSO and Classification and Regression Tree (CART) analysis were used to explore factors and visualize relationships between social determinants and missed care opportunities.

Results

A total of 62,009 orders were analyzed, 30,567 pre-pandemic, 3,205 pandemic, and 28,237 post-pandemic. Median age was 11.34 (IQR: 5.24–15.02), with 50.8% females. MCO increased during the pandemic (33.5%) compared to pre-pandemic (17.1%) and post-pandemic (16.5%). Logistic regression revealed higher odds of MCO in the pre-pandemic period for orders involving fluoroscopy (OR:1.675), MRI (OR:1.991), nuclear medicine studies (OR:1.505), and ultrasound (OR:1.211), along with patients residing outside the state (OR:1.658) and across all age groups compared to adolescents. During the pandemic, increased distance from the examination site (OR:1.1), residing outside the state (OR:1.571), Hispanic (OR:1.492), lower household income ($25,000–50,000 [OR:3.660] and $50,000–75,000 [OR:1.866]), orders for infants (OR:1.43), and fluoroscopy (OR:2.303) had higher odds. In the post-pandemic period, factors such as living outside the state (OR:1.189), orders for children (OR:0.787), and being Hispanic (OR:1.148) correlate with higher odds of MCO.

Conclusion

Applying basic data science-based techniques is a helpful approach to understanding complex relationships between sociodemographic characteristics and disparities.

social determinants of health

pediatric

radiology

data science

One of the silver linings of the COVID-19 pandemic is that it underscored the negative direct impact of inequalities and social factors on access to imaging services and illustrated that health crises can have long-term effects on healthcare access beyond the acute setting. Initially, post-pandemic analyses focused on radiology exam-volume recovery. However, despite imaging centers resuming their pre-pandemic workflows, many historically underserved communities still faced differential access to healthcare services ^1–3. For example, a study conducted by Lacson et al. found that disparities in access to diagnostic radiology services worsened during the initial phases of reopening after the COVID-19 pandemic. Previous studies have revealed that individuals residing in underserved communities with high poverty and unemployment levels and experiencing chronic illnesses had lower odds of undergoing radiological examinations after the pandemic². These trends illustrate the need to develop new strategies for comprehensively identifying individuals or communities at higher risk of diminished access to healthcare resources using methods that incorporate sociodemographic and medical factors.

In this context, novel statistical strategies and healthcare metrics beyond conventional volume-based analyses have been proposed. For example, the term missed care opportunities has gained more attention recently as a dynamic outcome that encompasses several steps of the healthcare delivery process. It indicates any instance where the delivery of patient care is delayed or not provided at all for any reason⁴. It is by nature a multifactorial phenomenon involving complex interactions between social-, system-, and patient-related factors, which are amplified among pediatric radiology patients who also experience caregiver-related factors ⁴.

Similarly, in an effort to understand the complex interaction of social risk factors and health determinants associated with complex healthcare access metrics, data science and machine learning-based analytical approaches have been proposed due to their ability to incorporate the complex interaction of demographic, geographical, medical, and social information contained in medical records and provide more robust statistical models to explain differential access to healthcare services^5,6. Among these methods, visual representations and schematizations of complex interactions among variables, such as Classification and Regression Trees (CART), have provided predictive models to delineate social "risk profiles" and phenotypes ^7,8. Despite these advances in data science-based analytics, studies using these novel data science approaches to understanding health disparities in radiology are lacking.

Therefore, the purpose of this study is to use a data science-based approach to analyze complex interactions and effect modifiers between social determinants and health disparities among pediatric radiology patients experiencing missed care opportunities. We hypothesize that the COVID-19 pandemic exacerbated existing health disparities, having long-lasting effects beyond the return to care phases.

Study Setting and Population

In this retrospective, Institutional Review Board-approved and Health Insurance Portability and Accountability Act-compliant study, the imaging database of a quaternary mixed practice (children and adults) care center located in (BLINDED FOR REVIEW), with five affiliated outpatient facilities distributed all around the state was systematically assessed to identify all outpatient pediatric (< 18 years) radiology exam orders before (03/08/19 to 03/07/20), during (03/08/20 to 06/07/20), and after the COVID-19 local government-mandated shutdown (06/08/20 − 06/07/21). The sampling frame for this study was based on the statewide COVID-19 mandate by which all nonurgent healthcare services were deferred between March 9, 2020, and June 6, 2020, in the home state of the authors’ institution. Data collection included one year before and after the shutdown to account for seasonal variations in the volume of exams and characteristics of the patients. Participants with incomplete records or aged 18 years or older were excluded. Interventional radiology cases, studies performed for research purposes, or duplicated cases were also excluded. Emergency cases and rescheduled examinations were omitted to maintain consistency and minimize bias.

Data variables and measures

All pediatric radiology exam orders and radiology reports for patients in the sampling time frames were identified and retrieved from the institutional research data repository and electronic health records. The analysis included patient-specific sociodemographic factors (i.e., age, sex, race, ethnicity, language, and zip code.), exam site, imaging modality (Magnetic Resonance Imaging [MRI], Ultrasound [US], Computed Tomography [CT], Fluoroscopy [FL], or X-Ray [XR]), and exam characteristics such as indication, anesthesia/sedation requirement, or intravenous contrast use. Ethnicity and spoken language were self-reported and extracted from patients’ charts. Patients' zip codes were used to estimate median household income by pairing them with Internal Revenue Service publicly available data and were broken down into the following brackets ($25,000 - $50,000, $50,000 - $75,000, $75,000 - $100,000, $100,000 - $200,000 and $200,000 or more) ⁹. The distance between the patient's residence and the exam site was calculated as a straight line between the residential and hospital zip codes. Zip codes’ latitudes and longitudes were extracted from Census Geocoding Services¹⁰. This study's primary outcome was the number of radiology missed care opportunities before, during, and after the shutdown. Missed care opportunities were defined as any exam scheduled and programmed in the institutional electronic health records which did not occur for any reason. In contrast, no-shows occur when appointments are called off by the patient¹¹.

Statistical Analysis

Categorical variables were reported as frequencies and percentages, while continuous were reported as median and interquartile ranges (IQR). We compared sociodemographic and clinical variables between missed care opportunities and completed examinations for each period, using nonparametric Mann-Whitney U tests for all continuous variables (assuming heterogeneous variance) and X²-tests for categorical variables. All tests were two-sided, and the significance level was set to 0.05. Correction for multiple comparisons was performed as necessary using the Bonferroni method.

To study the association between missed care opportunities and candidate variables, we modeled a multivariate logistic regression analysis. This analysis was performed for each study period. To minimize the number of covariates included in the model, we used the Least Absolute Shrinkage and Selection Operator (LASSO) method to select the most relevant variables from the pool of candidate predictors. LASSO was selected as it consistently identifies the most relevant components for modeling and helps select an optimal subset of predictors ¹².

Subsequently, a Classification and Regression Tree (CART) was fitted for each period to understand and graphically depict the relationship between social determinants and missed care opportunities. The CART algorithm quantifies each variable's weight, builds "risk profiles," and displays them graphically, simplifying the interpretation of the interactions ^13,14. Additionally, the tree-based method exemplifies how an explanatory variable's effects are modified by other covariables, modeling the interactions between variables in a nonlinear manner. This methodology contrasts with classical regression models, as it can help delineate effect modifiers and complex interactions, whereas, in the classic approach, these interactions between covariables must be deduced and inferred based on a prior hypothesis. Covariates included for CART modeling were selected based on previously reported literature and their statistical significance in the multivariate analysis. The tree results were manually trimmed and pruned to improve the branches' interpretability. Analyses were performed in R Version 4.0.2 (Free Software Foundation's GNU Public License) and SPSS v26 (Armonk, NY).

General characteristics: A total of 62,009 unique orders were included for analysis, including 30,567 from the pre-pandemic, 3,205 from the pandemic, and 28,237 from the post-pandemic period. The median patient age of the overall population was 11.34, IQR: 5.24–15.02; female patients represented 50.8% (31,513/62,009) of the orders. Missed care opportunities significantly increased during the pandemic (n = 1,075/3,205; 33.5%) compared to the pre-pandemic period (n = 5,235/30,567; 17.1%), with a return to the baseline after the end of the shutdown (n = 4,664/28,237; 16.5%) (p < 0.05).

Multivariate analysis

Results of the multivariate analysis for each period are summarized in Table 1.

Pre-pandemic period: On multivariable analysis, during the baseline period, household income categories between $75,000-100,000 [OR: 0.832; CI, 0.749–0.924] and $100,000-200,000 [OR: 0.903; CI, 0.823–0.992] were also associated with lower odds of missed care opportunities. Higher odds were demonstrated for studies ordered for patients living out of the state [OR: 1.658; CI, 1.468–1.873). Likewise, there were higher odds in all age groups (children [OR: 1.218; CI, 1.138–1.304), infants [OR: 1.945; CI, 1.770–2.138], and neonates [OR: 2.094; CI, 1.608–2.728], compared to adolescents (reference group). Compared to CT, several differences between modalities were also found, including higher odds of missed care opportunities for fluoroscopy (OR: 1.675; CI, 1.387–2.023), MRI (OR: 1.991; CI, 1.669–2.374), nuclear medicine studies (OR: 1.505; CI, 1.058–2.142) and Ultrasound (OR: 1.211; CI, 1.020–1.437), and lower for radiographs (OR: 0.547; CI, 0.464–0.646). Orders for neuroimaging studies (OR: 0.548; CI, 0.486–0.618) were associated with lower odds of missed care opportunities.

Pandemic: During the COVID-19 pandemic, the distance between the patient's residence and the examination site (distance [OR: 1.1, CI: 1.01–1.19] and living out of the state [OR: 1.571, CI: 1.150–2.147]), self-identifying as Hispanic (OR:1.492; CI: 1.200-1.854), and having a lower household income ($25,000–50,000 [OR:3.660; CI, 2.736–4.894] and $50,000–75,000 [OR:1.866; CI, 1.427–2.439]) were associated with higher odds of missed care opportunities. Similarly, orders for infants (OR: 1.43, CI: 1.140–1.793) and fluoroscopy (OR: 2.303; CI, 1.428–3.715) had higher odds of missed care opportunities. Neuroimaging studies (OR: 0.544; CI, 0.392–0.754) were associated with lower odds of missed care opportunities.

Post-pandemic: In the post-pandemic period, being Hispanic (OR: 1.148; CI, 1.038–1.269) and living out of the state (OR:1.189, CI: 1.054–1.341), were again associated with higher odds of missed care opportunities. Asian race (OR: 0.835; CI, 0.724–0.962), orders for children (OR:0.787; CI, 0.732–0.847), not having English as a first language (OR: 0.743; CI, 0.662–0.835) and neuroimaging (OR: 0.570; CI 0.501–0.648), were associated with lower odds. Similar trends to those of the pre-pandemic period were seen for the odds of missed care opportunities between modalities.

Table 1

Results of the multivariate analysis
		PRE-PANDEMIC								PANDEMIC								POST-PANDEMIC
		OR		95% CI				P-Value		OR		95% CI				P-Value		OR		95% CI				P-Value
				Lower		Upper						Lower		Upper						Lower		Upper
Gender (male)		0.946			0.890		1.007		0.080		0.941		0.805		1.101		0.448		0.927		0.870		1.289		0.062
Age Category*	Adolescents		REFERENCE								REFERENCE								REFERENCE
	Children	1.218		1.138		1.304		0.000		0.988		0.831		1.173		0.887		0.787		0.732		0.847		0.000
	Infants	1.945		1.770		2.138		0.000		1.430		1.140		1.793		0.002		0.944		0.861		1.035		0.221
	Neonates	2.094		1.608		2.728		0.000		1.709		0.838		3.487		0.140		0.970		0.744		1.264		0.822
Modality	CT	REFERENCE								REFERENCE								REFERENCE
	FL	1.675		1.387		2.023		0.000		2.303		1.428		3.715		0.001		2.170		1.751		2.690		0.000
	MR	1.991		1.669		2.374		0.000		1.336		0.857		2.082		0.200		2.036		1.667		2.488		0.000
	NM	1.505		1.058		2.142		0.023		0.891		0.385		2.063		0.788		1.090		0.707		1.679		0.697
	US	1.211		1.020		1.437		0.028		1.056		0.695		1.605		0.798		1.444		1.189		1.755		0.000
	XR	0.547		0.464		0.646		0.000		0.704		0.471		1.053		0.088		0.795		0.659		0.960		0.017
Neurological indication		0.548		0.486		0.618		0.000		0.544		0.392		0.754		0.000		0.570		0.501		0.648		0.000
Anesthesia		1.033		0.870		1.227		0.708		0.829		0.540		1.272		0.391		1.026		0.863		1.220		0.773
Out of state		1.658		1.468		1.873		0.000		1.571		1.150		2.147		0.005		1.189		1.054		1.341		0.005
Distance (Km)		1.000		1.000		1.000		0.108		1.100		1.001		1.190		0.000		1.000		0.999		1.000		0.073
Race	White/ Caucasian	REFERENCE								REFERENCE								REFERENCE
	Asian	1.000		0.872		1.146		0.995		0.736		0.505		1.074		0.112		0.835		0.724		0.962		0.013
	Black	1.111		0.979		1.261		0.101		1.002		0.742		1.352		0.992		0.991		0.866		1.134		0.895
	American Indian/ Native Alaska	0.903		0.821		0.993		0.035		0.943		0.770		1.156		0.574		1.019		0.926		1.121		0.698
	Native Hawaiian	1.138		0.640		2.022		0.660		0.225		0.073		0.693		0.009		0.534		0.266		1.071		0.077
	Other/ Unavailable	0.549		0.161		1.871		0.338		0.000		0.000				0.999		0.982		0.369		2.612		0.971
Hispanic ethnicity		1.176		0.966		1.071		0.060		1.492		1.200		1.854		0.000		1.148		1.038		1.269		0.007
No English speaker		1.090		0.978		1.215		0.120		1.228		0.956		1.577		0.108		0.743		0.662		0.835		0.000
Household income category	$25,000 - $50,000	0.950		0.836		1.079		0.430		3.660		2.736		4.894		0.000		1.087		0.950		1.242		0.224
	$50,000 - $75,000	0.984		0.888		1.091		0.761		1.866		1.427		2.439		0.000		1.059		0.952		1.179		0.290
	$75,000 - $100,000	0.832		0.749		0.924		0.001		1.221		0.917		1.627		0.172		0.947		0.848		1.057		0.333
	$100,000 - $200,000	0.903		0.823		0.992		0.034		0.863		0.661		1.127		0.281		0.957		0.867		1.057		0.383
	$200,000 or more	REFERENCE								REFERENCE								REFERENCE

* The ages used in the text were defined using World Health Organization (WHO) definitions.

Data Science

CART: The CART allowed us to understand complex relationships between social determinants and missed care opportunities during the pre-pandemic (Fig. 1), pandemic (Fig. 2), and post-pandemic (Fig. 3) periods. For example, during the pre-pandemic period, missed care opportunities were mainly related to exam-specific characteristics (modality and neurological exams) and the patient's age. On the other side, during the COVID-19 pandemic, missed care opportunities were driven by social determinants of health such as household income, the distance between the patient's residence and the examination site, and ethnicity (Hispanic vs. Non-Hispanic). During the post-pandemic period, exam-specific characteristics and patients’ age re-gained relevance, displaying a similar behavior to the baseline (pre-pandemic). Similar to the pre-pandemic period, certain modalities (MR, US, and FL) and non-neurologic studies were associated with a higher rate of missed care opportunities during the recovery period. Nevertheless, contrary to the pre-pandemic period, ethnicity remained as a relevant predictor of the outcome in a specific subset of exam orders. Specifically, MRI, US, and FL studies ordered for neurological indication showed increased rates of missed care opportunities in patients who self-identified as Hispanics (Fig. 4).

Health disparities and inequities are frequently multifactorial and the result of various factors working in a complex combination rather than a single cause¹⁵. Historically underserved populations are often affected by more than one social determinant of health, and there is a complex relationship between social factors leading to less favorable outcomes among different communities¹⁶. In this study, we used data science analyses to comprehensively assess social determinants of health and their impact on access to diagnostic resources. Through our research, we were able to identify and delineate "risk phenotypes" that represent a combination of social determinants of health and exam-specific factors that synergically interact to increase the overall risk of missed care opportunities. This study pioneers the use of visual regression techniques to explore the complex relationship between various social determinants and health disparities among pediatric radiology patients. Unlike classical regression models, this approach allows for the identification of effect modifiers and intricate interactions, providing valuable insights without the need for predefined hypotheses.

As demonstrated by our result, the COVID-19 pandemic aggravated previously existing disparities in access to radiology services, contributing to differential and unequal healthcare outcomes for specific racial and ethnic groups. While this issue has been extensively characterized, previous explicative models and studies were limited to assessing individual numerical risks of subjects with specific characteristics, lacking a comprehensive delineation of the complex interactions leading to real-life risk phenotypes^15,17,18. After the COVID-19 pandemic, the healthcare system gradually returned to normality as center imaging volumes rebounded. However, it became increasingly clear that the sheer volume of imaging exams conducted was not the most significant factor in addressing the overarching healthcare challenges. Instead, the critical focus shifted towards recognizing the pivotal role of social determinants in each individual's health. It became evident that addressing the unique circumstances, socioeconomic factors, and access to care for each patient was paramount in achieving comprehensive and equitable healthcare outcomes.

Using the CART, we created risk profiles based on specific demographic factors to identify individuals who may be at a higher risk of experiencing health disparities. Specifically, we found that patients from high-poverty zip codes and those facing long distances to imaging centers had significantly higher odds of missed care opportunities. Even though we saw some progress in these patterns in the first year after the pandemic, Hispanic patients still had greater odds of missing their care opportunities. These comprehensive risk phenotypes can be used to allocate resources more effectively in times of limited availability and aid in overcoming financial and geographical barriers for communities with limited access to critical imaging exams.

The comparison between the pre-pandemic and post-pandemic periods has revealed some interesting findings. Even though the total of orders and proportional number of missed care opportunities returned to the baseline in the first year of reopening, several changes in factors related to missed care opportunities were evident. In the pre-pandemic phase, certain socioeconomic factors, such as higher household incomes, were linked to lower odds of missed care opportunities, while patients living out of state and across various age groups demonstrated higher odds. Additionally, there were differences across imaging modalities, with some, such as CT and X-rays, exhibiting lower odds, while others, such as MRI and fluoroscopy, had higher odds of missed care opportunities. However, in the post-pandemic period, there was a notable shift in the landscape. While the patterns across different imaging modalities largely returned to baseline, ethnicity became a significant factor, with Hispanic individuals exhibiting higher odds of missed care opportunities (Fig. 4). These results suggest that simple volume-based analyses can be insufficient to understand and elucidate the deleterious impact of inequities and disparities in healthcare access.

The demographic shifts in missed pediatric radiology opportunities across pre-pandemic, pandemic, and post-pandemic periods can be hypothesized as being driven by the systemic strain on healthcare resources, altered healthcare-seeking behaviors due to pandemic fears, and amplified economic hardships, all exacerbated by the pandemic. Mobility restrictions may have presented additional barriers to accessing care, particularly for underserved communities. These challenges, coupled with increased caregiver burdens from school closures and potential loss of health insurance due to unemployment, likely contributed to the higher rates of missed care opportunities during the pandemic. Even as conditions began to improve in the post-pandemic phase, the enduring economic effects and persisting disparities in health literacy and access continued to influence which demographic groups were more likely to miss scheduled healthcare services.

Despite radiology exam volumes being at a historically high level across all healthcare practices, disparities in access to outpatient diagnostic imaging services persist among historically underserved patient populations. This differential access to healthcare is, again, multifactorial and depicts an intricate relationship between patient, community, and system factors. Therefore, it is crucial to look beyond just numerical risk assessment and examine how these attributes interact to create an individual's risk profile. This issue is particularly prevalent among the pediatric population, as healthcare decisions are strictly made by parents or caregivers, and the threshold for consulting is different based on health literacy ¹⁹. Studies by Marin JR et al. and Shah SN et al. found that non-Hispanic Black and Hispanic children were less likely to receive imaging studies and more likely to miss care opportunities than white children, even when controlling for other factors that could be related to the outcome ^19,20. These discrepancies in imaging times can lead to late diagnoses, delayed treatment, and poor health outcomes²¹.

This study is limited by the single-center design, and the retrospective nature of the study may limit the generalizability of the results to different geographic areas and healthcare settings. The reliance on historical data introduces potential issues related to data accuracy, completeness, and biases in data collection. Moreover, the study's focus on a specific urban population may not fully capture the complexities of health disparities in more rural or diverse regions. While the study effectively identifies associations between health disparities and missed pediatric radiology opportunities during the COVID-19 pandemic, the broader applicability of these findings to other contexts warrants cautious consideration. Further research is necessary to address some of the limitations present in this particular setting.

In conclusion, the COVID-19 pandemic has exposed and exacerbated existing health disparities, revealing the intricate interplay of social determinants and their enduring impact on healthcare access. This newfound emphasis on understanding and mitigating social determinants of health underscored the need for a holistic and patient-centered approach to medical care, ensuring that individuals receive attention and support. Despite imaging centers returning to pre-pandemic workflows and higher diagnostic imaging volumes, disparities in imaging access persist among historically underserved communities. Applying basic data science-based techniques is a helpful approach to understanding complex relationships between sociodemographic characteristics and disparities and opens a new era for evaluating social determinants and informing the design of programs to improve access to radiology care.

Author Contribution

S.G.B: wrote the main manuscript, prepared figures, helped with data analysis, and reviewed the manuscript.V.P.T: wrote specific parts of the manuscript, prepared figures, and reviewed the manuscript.F.M.R: wrote specific parts of the manuscript and reviewed the manuscript.O.l.P: data collection and analysis.D.B: data collection and analysis.E.F: significantly enhanced the manuscript's intellectual content through his meticulous revision.M.G: Oversaw all steps of the manuscript preparation, conceived the presented idea, and reviewed the final manuscript version.

Campbell T, Galvani AP, Friedman G, Fitzpatrick MC. Exacerbation of COVID-19 mortality by the fragmented United States healthcare system: A retrospective observational study. The Lancet Regional Health - Americas. 2022;12:100264. doi:10.1016/j.lana.2022.100264
Lacson R, Shi J, Kapoor N, Eappen S, Boland GW, Khorasani R. Exacerbation of Inequities in Use of Diagnostic Radiology During the Early Stages of Reopening After COVID-19. Journal of the American College of Radiology. 2021;18(5):696-703. doi:10.1016/j.jacr.2020.12.009
Lacson R, Shi J, Kapoor N, Eappen S, Boland GW, Khorasani R. Exacerbation of Inequities in Use of Diagnostic Radiology During the Early Stages of Reopening After COVID-19. Journal of the American College of Radiology. 2021;18(5):696-703. doi:10.1016/j.jacr.2020.12.009
Gustafsson N, Leino-Kilpi H, Prga I, Suhonen R, Stolt M. Missed Care from the Patient’s Perspective – A Scoping Review. Patient Prefer Adherence. 2020;Volume 14:383-400. doi:10.2147/PPA.S238024
Zhang X, Pérez-Stable EJ, Bourne PE, et al. Big Data Science: Opportunities and Challenges to Address Minority Health and Health Disparities in the 21st Century. Ethn Dis. 2017;27(2):95-106. doi:10.18865/ed.27.2.95
Gumustop S, Gallo-Bernal S, McPeake F, Briggs D, Gee MS, Pianykh OS. Predicting health crises from early warning signs in patient medical records. Sci Rep. 2022;12(1):19267. doi:10.1038/s41598-022-23900-8
Dong W, Bensken WP, Kim U, Rose J, Berger NA, Koroukian SM. Phenotype Discovery and Geographic Disparities of Late-Stage Breast Cancer Diagnosis across U.S. Counties: A Machine Learning Approach. Cancer Epidemiology, Biomarkers & Prevention. 2022;31(1):66-76. doi:10.1158/1055-9965.EPI-21-0838
Breiman L, Friedman JH, Olshen RA, Stone CJ. Classification And Regression Trees. Routledge; 2017. doi:10.1201/9781315139470
Internal Revenue Service. SOI Tax Stats - Individual Income Tax Statistics - ZIP Code Data (SOI).
United States Census Bureau. Census Geocoding Services.
Flores EJ, Daye D, Peña MA, Lopez DB, Jaimes C, Glover M. Analysis of socioeconomic and demographic factors and imaging exam characteristics associated with missed appointments in pediatric radiology. Pediatr Radiol. 2021;51(11):2083-2092. doi:10.1007/s00247-021-05111-x
Zhang Y, Li R, Tsai CL. Regularization Parameter Selections via Generalized Information Criterion. J Am Stat Assoc. 2010;105(489):312-323. doi:10.1198/jasa.2009.tm08013
Breiman L, Friedman JH, Olshen RA, Stone CJ. Classification And Regression Trees. Routledge; 2017. doi:10.1201/9781315139470
Krzywinski M, Altman N. Classification and regression trees. Nat Methods. 2017;14(8):757-758. doi:10.1038/nmeth.4370
Abraham P, Bishay AE, Farah I, Williams E, Tamayo-Murillo D, Newton IG. Reducing Health Disparities in Radiology Through Social Determinants of Health: Lessons From the COVID-19 Pandemic. Acad Radiol. 2021;28(7):903-910. doi:10.1016/j.acra.2021.04.006
Stronks K, Kunst AE. The complex interrelationship between ethnic and socio-economic inequalities in health. J Public Health (Bangkok). 2009;31(3):324-325. doi:10.1093/pubmed/fdp070
Tai DBG, Shah A, Doubeni CA, Sia IG, Wieland ML. The Disproportionate Impact of COVID-19 on Racial and Ethnic Minorities in the United States. Clinical Infectious Diseases. 2021;72(4):703-706. doi:10.1093/cid/ciaa815
Lacson R, Shi J, Kapoor N, Eappen S, Boland GW, Khorasani R. Exacerbation of Inequities in Use of Diagnostic Radiology During the Early Stages of Reopening After COVID-19. Journal of the American College of Radiology. 2021;18(5):696-703. doi:10.1016/j.jacr.2020.12.009
Marin JR, Rodean J, Hall M, et al. Racial and Ethnic Differences in Emergency Department Diagnostic Imaging at US Children’s Hospitals, 2016-2019. JAMA Netw Open. 2021;4(1):e2033710. doi:10.1001/jamanetworkopen.2020.33710
Shah SN, Melvin P, Tennermann NW, Ward VL. Racial and ethnic disparities in pediatric magnetic resonance imaging missed care opportunities. Pediatr Radiol. 2022;52(9):1765-1775. doi:10.1007/s00247-022-05460-1
Waite S, Scott J, Colombo D. Narrowing the Gap: Imaging Disparities in Radiology. Radiology. 2021;299(1):27-35. doi:10.1148/radiol.2021203742

No competing interests reported.

Download PDF

Editorial decision: Revision requested
06 Aug, 2024
Reviews received at journal
02 Aug, 2024
Reviews received at journal
28 Jul, 2024
Reviews received at journal
22 Jul, 2024
Reviewers agreed at journal
11 Jul, 2024
Reviewers agreed at journal
10 Jul, 2024
Reviewers agreed at journal
08 Jul, 2024
Reviewers invited by journal
08 Jul, 2024
Editor assigned by journal
05 Jul, 2024
Submission checks completed at journal
05 Jul, 2024
First submitted to journal
02 Jul, 2024

You are reading this latest preprint version

A Data Science-based approach to Identify Social Determinants of Health Impacting Access to Pediatric Radiology

Status:

Version 1

Abstract

Background

Objective

Materials and Methods

Results

Conclusion

Figures

INTRODUCTION

MATERIALS AND METHODS

RESULTS

Data Science

DISCUSSION

CONCLUSION

Declarations

Author Contribution

References

Additional Declarations

Status:

Version 1