Is anyone working with dental data?
We have a few questions about interpretation of codes.
NW
1 Like
(I am from the same team as Nick, btw.)
Are there non-US OMOP data sites, that would have data on dental procedures. If you see tooth extraction, do you know which tooth was extracted? (same for fillings)
for example, NHIMC cdm dataset has a lot of concepts and RCs of tooth procedures like filling, extraction and so on. but there isnât any information of specific anatomic sites.
Iâm curently thinking about this. In France we will have a code for the tooth extraction and, in another fields, we will have the teeth identified. I donât know how to store this information in the CDM⌠either the source information and the standard information.
I have seen @Vojtech_Huser post about snomed post coordination but, well, I donât know what to do with itâŚ
We continue to explore dental data and dental questions. We submitted one full paper to fall AMIA conference and one abstract to SNOMED CT Expo. We also tried to engage AMIA Dental WG. @Rosnyni , please email me such that I can provide more.
Here is the text of our SNOMED CT Expo abstract where we try to have international target codes for dental procedures. There is also a github link.
Using SNOMED CT procedural concepts for dental research: US Medicaid use case
Vojtech Huser, Nick Williams
ABSTRACT: Observational research using dental billing data can provide real world evidence about dental procedures. US Medicaid data are unique in including large number of dental care procedural events. We mapped dental procedures (encoded in Common Dental Terminology; CDT) to SNOMED CT procedures. Using SNOMED CT as target terminology for dental procedures allows formulation of studies that can be executed in international scope. We mapped 30 most frequent dental procedures from US claims (covering 90% of billed events). In some cases, granularity offered by SNOMED CT leads to mapping to a broader term. This broader mapping is still acceptable and allows for meaningful research analysis. We also similarly mapped dental procedures from the Czech republic to pilot the international research scope. We submitted and further plan to contribute possible improvements to dental procedures to appropriate SNOMED CT extension or international edition.
TEXT: Observational research using dental billing data can provide real world evidence about dental procedures. US Medicaid plan is unique in including significant dental coverage. Medicaid data (unlike other US claims databases) thus include dental care data. The study is part of a larger project that converts Medicaid data into OMOP (Observational Medical Outcomes Partnership) Common Data Model (CDM). We utilize Virtual Research Data Center (VRDC) provided by the Center for Medicare and Medicaid Services (CMS). SNOMED CT concepts are used in OMOP model to achieve semantic data harmonization (SNOMED CT concept are designated as âstandard conceptsâ).
We mapped dental procedures (encoded in Common Dental Terminology; CDT) to SNOMED CT procedures. Using SNOMED CT as target terminology for dental procedures allows formulation of studies that can be executed in international scope. International agreement on a target terminology for a given clinical domain (lab results, diagnoses, procedures) is an ongoing informatics challenge. For dental procedures, this challenge is also present. For international analysis, a country-wide license or no copyright on target terminology is desirable (copyright by American Dental Association on CDT does not satisfy this desideratum).
First, we analyzed existing SNOMED CT dental procedures content by inspecting 1127 procedures identified using queries. The queries expressed in SNOMED CT Expression Constraint Language used the following elements â<< 118817003 |Procedure on oral cavity (procedure)|â, â<< 118818008 |Procedure on tooth (procedure)|â, â<< 81733005 |Dental surgical procedure (procedure)|â. Besides manual review, we used R language scripts to analyze relationships of those concepts and adherence to current editorial guide.
Second, we identified most frequent dental procedural codes that represent 90% of billed dental events for Medicaid (using only data from the state of Washington; total of 30 codes) and mapped those to SNOMED CT procedures. 6 codes were exact matches and 24 were ânarrow (source code) to broad (target SNOMED CT code).â Billing dental codes included concept features that are out of scope of SNOMED CT. We documented aspects omitted in the mapping. For example (omitted content shown in quotes): âperiodicâ oral evaluation â âestablished patientâ, âcomprehensiveâ oral evaluation â ânew or established patientâ, resin-based composite â âtwo surfacesâ, comprehensive orthodontic treatment âof the adolescent dentitionâ). For each non-exact mapping we classified it as acceptable or not for dental research.
For research analysis, ânarrow to broadâ mapping was acceptable to still allow for meaningful descriptive studies (note that the direction is from billing source code to target SNOMED CT code; contrary to the majority of maps maintained by SNOMED International). It is very common that research phenotype definitions group many related billing codes into higher level groups. This common practice theoretically lessens the impact of possibly information-losing narrow to broad mapping. In other words, researchers often do not utilize all possible semantic granularity of billing codes. For our scope of international descriptive analysis of dental procedures, loss of granularity and careful mapping considerations are inevitable because of multiple country-specific billing frameworks. To demonstrate international comparison, we similarly mapped frequent dental procedural codes from the Czech republic national procedural terminology. For some mappings, multi paragraph full description and coding guidance for each code (in addition to just long concept name translated to English) provided additional details that facilitated a more precise mapping.
We also analyzed SNOMED CT codes for tooth and tooth surface to be used in conversion of Medicaid data into OMOP model. Accurate data specifying on which tooth the procedure was performed are sometimes of greater research importance than perfectly matching granularity of mapping. Our mappings and additional results are available at project repository at CRI/VRDC/project/dental at master ¡ lhncbc/CRI ¡ GitHub.
The OMOP model uses SNOMED CT as target standard terminology for diagnoses, but there is lack of consensus about analogous harmonizing target terminology for procedural history data. The reason for this is difference in granularity (and scope) of SNOMED CT procedures in comparison with billing procedural terminologies and possibly hesitancy to use narrow to broad mappings. Our work explores whether mapping acceptable for research purposes is possible and how to document and maintain such narrow to broad mapping (possibly improve over time; shifting it into exact mapping). Billing data is readily accessible to researchers and it is unlikely that data coded natively in SNOMED CT from dental Electronic Health Record (EHR) would be available in the near future (due to limited direct adoption of SNOMED CT by dental EHRs).
As future work, we plan to map additional codes that would cover more volume Medicaid dental billing events. A total of 72 codes would cover 99% of billed events (42 extra codes would have to be mapped). We submitted and further plan to contribute possible improvements to dental procedures to US SNOMED CT extension or international edition. Use of SNOMED CT as target terminology common terminology can facilitate analysis portability across countries.