IADR Abstract Archives
Assessing a Data Quality Framework in BigMouth Dental Repository
Objectives: Create an infrastructure of data quality assessment around BigMouth (BM): an electronic data repository containing EHR data from eight dental institutions for research and improvement of care.
Methods: Tests were conducted around a chosen 5-dimension data quality framework: (1) Completeness – measured through the number of undefined diagnoses from 2012-2018, and replication via BM of a chosen study that used EHR data. Furthermore, data elements in EHR literature were reviewed for availability in BM. (2) Correctness – looked at links between diagnoses and correct usage of procedures codes. (3) Concordance & (4) Plausibility – measured externally by comparison of BM patient population, subdivided into age/gender, to the 2018 U.S Census. Internally, diagnoses trends for patient population were compared across BM institutions. (5) Currency was not directly measured as BM data is updated every 6 months (limitation).
Results: 25% of patients received only "undefined" diagnoses in BM per year during 2012-2018 for a chosen institution. 64 out of 69 cited data elements were present in BM. Data from BM was successfully used to replicate the study as a measurement for completeness. Diagnosis/procedure trends were found to be similar across 2016-2018. Age/gender population data was found to be consistent compared to the 2018 U.S Census. Similarly, caries and periodontal disease diagnoses data was found to be consistent within an institution, but differences were observed when comparing across sites.
Conclusions: Through assessments in the quality dimensions, we successfully established that BM contains the necessary data elements for data quality assessment. Many of these initial assessments demonstrated quality data, while some displayed room for improvement. Future, more in-depth, and continual analysis will be needed to fully ascertain data quality in the BM Database.
Methods: Tests were conducted around a chosen 5-dimension data quality framework: (1) Completeness – measured through the number of undefined diagnoses from 2012-2018, and replication via BM of a chosen study that used EHR data. Furthermore, data elements in EHR literature were reviewed for availability in BM. (2) Correctness – looked at links between diagnoses and correct usage of procedures codes. (3) Concordance & (4) Plausibility – measured externally by comparison of BM patient population, subdivided into age/gender, to the 2018 U.S Census. Internally, diagnoses trends for patient population were compared across BM institutions. (5) Currency was not directly measured as BM data is updated every 6 months (limitation).
Results: 25% of patients received only "undefined" diagnoses in BM per year during 2012-2018 for a chosen institution. 64 out of 69 cited data elements were present in BM. Data from BM was successfully used to replicate the study as a measurement for completeness. Diagnosis/procedure trends were found to be similar across 2016-2018. Age/gender population data was found to be consistent compared to the 2018 U.S Census. Similarly, caries and periodontal disease diagnoses data was found to be consistent within an institution, but differences were observed when comparing across sites.
Conclusions: Through assessments in the quality dimensions, we successfully established that BM contains the necessary data elements for data quality assessment. Many of these initial assessments demonstrated quality data, while some displayed room for improvement. Future, more in-depth, and continual analysis will be needed to fully ascertain data quality in the BM Database.