The present study evaluated the incidence of hypoparathyroidism among 4848 eligible patients who underwent thyroid surgery at two tertiary centers. Among these patients, 1370 (28.26%) experienced transient and 251 (5.18%) experienced hypoparathyroidism during the study period. Because these two centers utilized a common CDM protocol, the results from these two centers could be analyzed together. Univariate analysis identified meaningful baseline factors associated with hypoparathyrodism, whereas multivariate analysis failed to identify statistically significant factors.
Our CDM protocol was based on the Observational Health Data Sciences and Informatics (OHDSI) database, which was designed for utilization in large scale observational studies.8,9 These CDMs were developed to manage large amounts of data in the medical field, including claims and clinical data. The advantages of using a standardized CDM include the compatibility of data among different institutions, allowing the use of standard analytical tools.10,11 In South Korea, efforts have begun to share EMR data in the form of CDMs among more than 40 tertiary medical centers throughout the country. Since 2018, these institutions have converted their EMR data to CDM format using OHDSI open-source resources12.
In accordance with this trend, CDMs have been used for medical research in several areas, including an overview of traumatic brain injury models 13, mitochondrion-related diseases based on neurologic examinations14, and evaluation of hospitalization and mortality rates of patients with atrial fibrillation 15. A recent CDM-based study from South Korea on patients with inflammatory bowel disease found that prognosis was less favorable in patients with early onset than late onset disease12. CDM based studies related to thyroid operations or their sequelae have not yet been published.
A nation-wide insurance database study among South Korean patients who underwent total thyroidectomy in 2012 indicated that the prevalence of persistent post-operative hypoparathyroidism was 10.4%, comparable to our results16. Differences between studies were likely due to differences in the operational definitions of persistent hypoparathyroidism, the study periods, and the institutions. Persistent hypoparathyroidism in our study was defined as a serum PTH concentration < 10 pg/mL at least once within the first 6 months after thyroidectomy and at least once more 6 to 18 months after thyroidectomy, whereas the other study, which was based on claims data, had a more complicated operational definition based on active vitamin D prescription. A CDM analysis based on serum PTH level could objectively adjust the operational definition of hypoparathyroidism, whereas studies based on insurance claims do not provide such flexibility and rely on subjective determinations. For example, a prescription for vitamin D and calcium prescription may have resulted from concurrent osteoporosis, a situation that cannot be detected by the operational definition based on claim data. We were able to exclude such patients because we could check the serum PTH level of patients at both institutions. Finally, our analysis included patients who underwent surgery by expert high-volume thyroid surgeons, whereas the national insurance claim database contains results in patients who underwent surgery by both expert and novice surgeons, which may result in a higher incidence of hypoparathyroidism.
This study had several limitations. First, because of the retrospective nature of this cross sectional multi-institutional study, the results may have been influenced by selection or indication biases. This is more pronounced by the inclusion only of patients who underwent surgery at tertiary medical centers with high volume surgeons. Furthermore, only the parameters that were coded in accordance with the CDM protocol could be analyzed. For example, the pathology reports are not the same in the two institutions; therefore, only factors common to both could be incorporated into the database. This would lead to a loss of data and may explain why our multivariate analysis did not yield any significant results.
In conclusion, our CDM study revealed that the incidences of persistent and transient hypoparathyroidism were 5.18% and 28.3%, respectively, among patients at the two tertiary centers. Furthermore, when compared with total bilateral thyroidectomy, patients who underwent radical tumor resection were 50% and 97% more likely to experience transient and persistent hypoparathyroidism, respectively. This study not only estimated the incidence of and risk factors for post-operative hypoparathyroidism at two institutions, but, by integrating and analyzing the databases at the two institutions, suggested that this analysis could be expanded to include other large databases that share the same Observational Health Data Sciences and Informatics protocol. These findings also suggested that, by integrating large institutional databases without relying on subjective factors, CDMs can produce large data results in endocrine surgery. Further CDM studies involving multiple tertiary hospitals may provide more representative and comprehensive analyses in endocrine surgery.