Open AccessSpatial Linkage of Electronic Health Records and Census Data in A Children’s Hospital
Author(s) -
Jonathan M. Tan,
Vicky Tam,
Jorge A. Gálvez,
Grace Hsu,
William Quarshie,
Jack O. Wasey,
Olivia Nelson,
Allan F. Simpao
Publication year - 2020
Publication title -
international journal of population data science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.602
H-Index - 7
ISSN - 2399-4908
DOI - 10.23889/ijpds.v5i5.1644
Subject(s) - medicine , neighbourhood (mathematics) , socioeconomic status , demography , logistic regression , unemployment , retrospective cohort study , census , medical record , emergency medicine , pediatrics , environmental health , surgery , population , mathematical analysis , mathematics , sociology , economics , economic growth
IntroductionSocial determinants of health (SDOH) has a significant impact on access to health. Risk stratification of patients who have difficulty accessing care could allow for triaging of peri-operative resources. Unfortunately, there is a limited amount of SES factors available to study in electronic health records (EHR).
Objectives and ApproachOur objective was to understand the SES and location risk factors that are associated with paediatric patients arriving late to the hospital for elective surgery. We conducted a retrospective study of paediatric patients requiring elective surgery from 2015-2019. Spatial linkage of EHR data with US Census–ACS 2017 data was conducted. Analysis was at patient and neighbourhood block group levels. Statistical analysis was conducted utilizing SAS, Python and ArcGIS.
ResultsOur study had 40,943 patients with 7,453 patients (18.2%) who arrived ≥15 minutes late from their scheduled arrival. Patient level risk factors for arriving late included younger age, Black and Indian patients, non-English speaking, government insurance, increased co-morbidities and earlier appointments. The median time of arrival for patients arriving late was 23.0 minutes (18.0-33.0 minutes IQR), versus the on-time group of 7 min (4-22 minutes IQR) early. Median drive time and distance, using network analysis was not a significant factor. Statistically significant neighbourhood risk factors for arriving late included block groups with high unemployment, households receiving public assistance, low income households, higher number of high-school drop outs, female-headed households, more renter-occupied houses, and areas with high turnover. Logistic regression demonstrated neighbourhoods with the lowest quintile SES were 30% more likely to be late than the areas with the highest SES (p<0.001).
Conclusion / ImplicationsWe successfully identified patient and neighbourhood socioeconomic risk factors for arriving late to the hospital leveraging geospatial methods and EHR data. Leveraging EHR data with geospatial analytics can augment our understanding of the SDOH that may impact the delivery of care.