J Cheng1,2, S E Gregorich1,2,3, S A Gansky1,2,4,5, S A Fisher-Owens2,6, A M Kottek2,4, J M White1,2, E A Mertz1,2,4. 1. Department of Preventive and Restorative Dental Sciences, School of Dentistry, University of California-San Francisco, San Francisco, CA, USA. 2. Center to Address Disparities in Children's Oral Health, School of Dentistry, University of California-San Francisco, San Francisco, CA, USA. 3. Department of Medicine, School of Medicine, University of California-San Francisco, San Francisco, CA, USA. 4. Philip R. Lee Institute for Health Policy Studies, University of California-San Francisco, San Francisco, CA, USA. 5. Bakar Computational Health Sciences Institute, University of California-San Francisco, San Francisco, CA, USA. 6. Department of Pediatrics, School of Medicine, University of California-San Francisco, San Francisco, CA, USA.
Abstract
INTRODUCTION: Electronic health record (EHR) systems provide investigators with rich data from which to examine actual impacts of care delivery in real-world settings. However, confounding is a major concern when comparison groups are not randomized. OBJECTIVES: This article introduced a step-by-step strategy to construct comparable matched groups in a dental study based on the EHR of the Willamette Dental Group. This strategy was employed in preparation for a longitudinal study evaluating the impact of a standardized risk-based caries prevention and management program across patients with public versus private dental insurance in Oregon. METHODS: This study constructed comparable dental patient groups through a process of 1) evaluating the need for and feasibility of matching, 2) considering different matching methods, and 3) evaluating matching quality. The matched groups were then compared for their average ratio in the number of decayed, missing, and filled tooth surfaces (DMFS + dmfs) at baseline. RESULTS: This systematic process resulted in comparably matched groups in baseline covariates but with a clear baseline disparity in caries experience between them. The weighted average ratio in our study showed that, at baseline, publicly insured patients had 1.21-times (95% CI: 1.08 to 1.32) and 1.21-times (95% CI: 1.08 to 1.37) greater number of DMFS + dmfs and number of decayed tooth surfaces (DS + ds) than privately insured patients, respectively. CONCLUSION: Matching is a useful tool to create comparable groups with EHR data to resemble randomized studies, as demonstrated by our study where even with similar demographics, neighborhood and clinic characteristics, publicly insured pediatric patients had greater numbers of DMFS + dmfs and DS + ds than privately insured pediatric patients. KNOWLEDGE TRANSFER STATEMENT: This article provides a systematic, step-by-step strategy for investigators to follow when matching groups in a study-in this case, a study based on electronic health record data. The results from this study will provide patients, clinicians, and policy makers with information to better understand the disparities in oral health between comparable publicly and privately insured pediatric patients who have similar values in individual, clinic, and community covariates. Such understanding will help clinicians and policy makers modify oral health care and relevant policies to improve oral health and reduce disparities between publicly and privately insured patients.
INTRODUCTION: Electronic health record (EHR) systems provide investigators with rich data from which to examine actual impacts of care delivery in real-world settings. However, confounding is a major concern when comparison groups are not randomized. OBJECTIVES: This article introduced a step-by-step strategy to construct comparable matched groups in a dental study based on the EHR of the Willamette Dental Group. This strategy was employed in preparation for a longitudinal study evaluating the impact of a standardized risk-based caries prevention and management program across patients with public versus private dental insurance in Oregon. METHODS: This study constructed comparable dental patient groups through a process of 1) evaluating the need for and feasibility of matching, 2) considering different matching methods, and 3) evaluating matching quality. The matched groups were then compared for their average ratio in the number of decayed, missing, and filled tooth surfaces (DMFS + dmfs) at baseline. RESULTS: This systematic process resulted in comparably matched groups in baseline covariates but with a clear baseline disparity in caries experience between them. The weighted average ratio in our study showed that, at baseline, publicly insured patients had 1.21-times (95% CI: 1.08 to 1.32) and 1.21-times (95% CI: 1.08 to 1.37) greater number of DMFS + dmfs and number of decayed tooth surfaces (DS + ds) than privately insured patients, respectively. CONCLUSION: Matching is a useful tool to create comparable groups with EHR data to resemble randomized studies, as demonstrated by our study where even with similar demographics, neighborhood and clinic characteristics, publicly insured pediatric patients had greater numbers of DMFS + dmfs and DS + ds than privately insured pediatric patients. KNOWLEDGE TRANSFER STATEMENT: This article provides a systematic, step-by-step strategy for investigators to follow when matching groups in a study-in this case, a study based on electronic health record data. The results from this study will provide patients, clinicians, and policy makers with information to better understand the disparities in oral health between comparable publicly and privately insured pediatric patients who have similar values in individual, clinic, and community covariates. Such understanding will help clinicians and policy makers modify oral health care and relevant policies to improve oral health and reduce disparities between publicly and privately insured patients.
Entities:
Keywords:
caries management; electronic health record; health equity; matching; observational cohort study; propensity score
Authors: Susan A Fisher-Owens; Stuart A Gansky; Larry J Platt; Jane A Weintraub; Mah-J Soobader; Matthew D Bramlett; Paul W Newacheck Journal: Pediatrics Date: 2007-09 Impact factor: 9.703
Authors: Suhasini Bangar; Ana Neumann; Joel M White; Alfa Yansane; Todd R Johnson; Gregory W Olson; Shwetha V Kumar; Krishna K Kookal; Aram Kim; Enihomo Obadan-Udoh; Elizabeth Mertz; Kristen Simmons; Joanna Mullins; Ryan Brandon; Muhammad F Walji; Elsbeth Kalenderian Journal: Appl Clin Inform Date: 2022-01-19 Impact factor: 2.342