Integrated medical and dental electronic health records (EHRs) offer a number of advantages for clinical and research applications, such as identifying associations between periodontitis and multiple medical conditions, according to a new study in the Journal of Clinical Periodontology (February 5, 2013).
Proponents of EHRs have long touted data mining as a key benefit of this form of medical and dental record keeping. But this is one of the first studies to demonstrate how a combined medical and dental EHR system can be used to identify associations between oral disease and multiple medical conditions in a large patient population.
"Identifying the connections between periodontitis and systemic diseases can ensure a holistic approach to patient care management," wrote the study authors, from Columbia University. "Regrettably, information about dental conditions is often inaccessible to physicians, partly due to inadequate interdisciplinary collaboration between dentistry and medicine."
At Columbia University Medical Center, electronic records have been used to develop a dental data warehouse at the College of Dental Medicine and a clinical data warehouse at its teaching hospital, NewYork-Presbyterian Hospital. In fact, many patients have records in both databases, the researchers noted.
For this study, they analyzed the electronic records of patients who underwent dental treatment and medical treatment at the two facilities. They used a case-control design to compare the prevalence of medical conditions between patients who had received treatment for periodontitis (case) and patients without periodontal treatment but with general dental maintenance (control).
Pittsburgh School of Dental Medicine
"Using all of the available patient records in our medical center, we first identified 2,126 patients whose dental record contained at least one Common Dental Terminology (CDT) periodontal treatment code (D4000-D4999) between January 3, 2007, and June 17, 2011," the researchers wrote.
They then identified 2,429 patients with one of two common dental maintenance codes (D0120 or D0150) but no periodontal treatment codes within the same time frame. These dental records were then matched to the medical records using first name, last name, gender, and date of birth.
"If any one of these fields was not exactly matched, the patient was removed from further analysis," they wrote.
The resulting study sample (N = 2,475) contained 1,235 cases and 1,240 controls. For each of these patients, the study authors then extracted all International Classification of Diseases 9th Revision (ICD-9) codes, plus date of birth, gender, ethnicity, dental date of service, and medical diagnostic date. To do the association mining, they further extracted unique ICD-9 codes for all patients' diagnoses that occurred between January 1, 2004, and December 16, 2011, and grouped the codes into 17 categories according to American Medical Association ICD-9 guidelines.
"By extracting the unique ICD-9 codes from each patient, we initially arrived at 36,230 ICD-9 codes in the control group and 38,541 ICD-9 codes in the case group, with a total of 3,908 unique ICD-9 codes in the entire dataset," the researchers wrote.
The data then underwent four types of analysis:
- Chi-squared goodness of fit tests with multiple hypotheses correction for each unique ICD-9 code observed by at least one patient in both groups
- Multivariate logistic regression modeling of chi-square significant ICD-9 codes to identify association while adjusting for age, gender, ethnicity, and tobacco abuse
- Multivariate logistic regression for each ICD-9 category with the same adjustment factors
- Literature validation of identified associations
A surprising finding
These analyses yielded some interesting and important findings, the researchers noted. For example, they confirmed known associations between periodontitis and diabetes (type I and II), hypertension, hypercholesterolemia, hyperlipidemia, and conditions pertaining to pregnancy and childbirth.
But they also identified a previously unreported association between periodontitis and benign prostatic hyperplasia in male patients younger than 70 years of age that could have clinical significance given the role of both diseases in inflammation pathways.
"This study demonstrates that vast amounts of clinical data made available by EHRs for both medical and dental care are useful and usable for discovering hidden disease knowledge, such as disease associations," lead author Mary Regina Boland, the research staff officer in the department of biomedical informatics at Columbia, told DrBicuspid.com. "It highlights the importance of supporting interoperability among EHRs used across different disciplines so that we can better integrate such data to further research and clinical care, and it serves as a terrific case study showing the value of interdisciplinary collaboration in life science."
EHRs can also provide a standards-based method for sharing data between clinicians and dentists, which would facilitate collaborative care, she added.
"With the burgeoning adoption of EHR systems by clinicians and dentists throughout the U.S., we believe that other record systems can be linked to uncover disease knowledge that will inform clinicians to provide a better quality of care for their patients."
While the Columbia study is not the first nor the largest to investigate how linked medical and dental records can be used to identify associations between oral and systemic diseases, it does do something previous studies haven't, according to Mei Song, PhD, a research scientist in the University of Pittsburgh School of Dental Medicine who has conducted similar research.
"The importance of this paper is that it is the first to explore all possible associations at the same time," she said. "Other studies have looked at one medical and one dental condition, such as periodontitis and diabetes, or two medical and two dental conditions. Looking at all possible things is very good for new knowledge discovery."
This approach to mining linked medical and dental records offers a number of advantages for researchers, she added, including being able to study diseases in a large sample with more statistical power, identify patients with rare diseases, and track patients over an extended time period.
But it also has the potential to impact clinical care, Song noted.
"Currently, the studies that are done in this field are epidemiological, which look at risk factors," she said. "With this approach, we can look at outcomes, which can generate a lot of evidence and potentially guide clinical care. Right now, we are looking at how this condition is associated with that condition. But with this approach, we can look at how if a patient is being treated with treatment A compared to treatment B, which is better?"