Consideration of oral health in rare disease expertise centres: a retrospective study on 39 rare diseases using text mining extraction method.

BACKGROUND: Around 8000 rare diseases are currently defined. In the context of individual vulnerability and more specifically the one induced by rare diseases, ensuring oral health is a particularly important issue. The objective of the study is to evaluate the pattern of oral health care course for patients with any rare genetic disease. Description of oral phenotypic signs-which predict a theoretical dental health care course-and effective orientation into an oral healthcare were evaluated.
MATERIALS AND METHODS: We set up a retrospective cohort study to describe the consideration of patient oral health and potential orientation to an oral health care course who have at least been seen once between 1 January 2017 and 1 January 2020 in Necker Enfants Malades Hospital. We recruited patients from this study using the data warehouse, Dr Warehouse® (DrWH), from Necker-Enfants Malades Hospital.
RESULTS: The study sample included 39 rare diseases, 2712 patients, with 54.7% girls and 45.3% boys. In the sample studied, 27.9% of patients had an acquisition delay or a pervasive developmental disorder. Among the patient files studied, oral and dental phenotypic signs were described for 18.40% of the patients, and an orientation in an oral healthcare was made in 15.60% of patients. The overall "network" effect was significantly associated with description of phenotypic signs (corrected p = 1.44e-77) and orientation to an oral healthcare (corrected p = 23.58e-44). Taking the Defiscience network (rare diseases of cerebral development and intellectual disability) as a reference for the odd ratio analysis, OSCAR, TETECOU, FILNEMUS, FIMARAD, MHEMO networks stand out from the other networks for their significantly higher consideration of oral phenotypic signs and orientation in an oral healthcare.
CONCLUSION: To our knowledge, no study has explored the management of oral health in so many rare diseases. The expected benefits of this study are, among others, a better understanding, and a better knowledge of the oral care, or at least of the consideration of oral care, in patients with rare diseases. Moreover, with the will to improve the knowledge on genetic diseases, oral heath must have a major place in the deep patient phenotyping. Therefore, interdisciplinary consultations with health professionals from different fields are crucial.

Background

For the European Union, a disease is considered “rare” if it affects fewer than one in 2000 people. Based on this definition, 27–36 million people affected by a rare disease are estimated in the European Union. Around 8000 rare diseases are currently defined [1, 2].

In France and Europe, territorial university-hospital networks have been created for the management of rare diseases. These Networks are composed of health professionals, researchers, and patient associations. They contribute to the improvement of care, support, training, information, research and innovation. In France, the networking has been set up by groups of pathologies and/or affected systems in the framework of National Plans for Rare Diseases (PNMR). 23 Health Networks for Rare Diseases have been created by the Ministry of Solidarity and Health in agreement with the Ministry of Higher Education, Research and Innovation. They include centres of expertise organized in specialized care by groups of pathologies, covering the whole territory of France. At the European level, French networks are part of ERN Networks; each-one being dedicated to a set of rare diseases that share common aspects.

In the context of individual vulnerability and more specifically the one induced by rare diseases, ensuring oral health is a particularly important issue [3]. Indeed, oral microenvironment is a crossroads of vital functions such as mastication, swallowing, breathing and phonation. Pathogenic oral microbiota induces dental caries, periodontal diseases requiring tooth extractions and treatments with consecutive discomfort, economical costs, and pain. Several oral pathogenic bacteria impact distant organs such as heart or kidney and may worsen several general diseases such as metabolic disorders or ectopic mineralization. Finally, a healthy oral-dental status is part of the person self-esteem and -confidence. Indeed, oral-facial aesthetic handicap, pain or altered facial-buccal expression due to oral diseases impacts the emotional and social life of the individuals. Indeed, the identification of rare diseases with dental and/or oral manifestations [1, 2] has been established disease-by-disease. Individual reports are registered in database of clinical description and/or gene mutation [1, 4] suggesting that 15% rare diseases would affect the oral status and requires dental treatment. This knowledge on oral health in rare diseases is cumulative but not conceived as a global issue.

At the Necker-Enfants Malades pediatric hospital (Paris), all the 23 French specific rare disease networks are represented with the presence of more than forty distinct rare disease centres of expertise. All reports of consultations, hospitalizations, complementary examinations, and hospital stays are automatically copied in the Necker Hospital data warehouse (Dr Warehouse®) developed by the Imagine institute. They are processed by using natural language methods to allow physician to handle large sets of data, including narrative reports, particularly for complex rare but also chronic diseases, and to describe the healthcare trajectory of patients. This tool was used here to conduct an unbiased study covering the rare disease networks on oral health and treatment. The objective of the study is to evaluate the pattern of oral health care course for patients with any rare genetic disease. To do so, description of oral phenotypic signs—which predict a theoretical dental health care course—and effective orientation into an oral healthcare were evaluated. This network-by-network analysis will guide efforts to effectively implement dental and oral phenotyping and improve the understanding of genetic diseases.

Materials and methods

Study setting and design

We set up a retrospective cohort study to describe the consideration of patients' oral health and their potential orientation to an oral health care course who have at least been seen once between 1 January 2017 and 1 January 2020 in Necker Enfants Malades Hospital.

Participants

To explore the research question through rare diseases with mostly different clinical signs and being part of distinct rare diseases networks and centres of expertise within the studied hospital, we established a list of representative diseases.

We included patients from 3-year-old with a rare disease (confirmed diagnosis) in the list of selected diseases (Table 1) who consulted between 1.1.2017 and 1.1.2020 at Necker Hospital and whom have been seen at least once in the medical genetics department of Necker Hospital.

Table 1

Rare diseases network and diseases selected due the number of included patients

Rare disease	Rare disease French network	Type of diseases
Prader Willi syndrome	AnDDI-Rares	Rare diseases with somatic and cognitive developmental abnormalities
Di George syndrome
Friedreich ataxia	BRAIN-TEAM	Rare diseases with motor or cognitive expression of the central nervous system
Fallot tetralogy	CARDIOGEN	Hereditary cardiac diseases
Hypertrophic cardiomyopathy
Dilated cardiomyopathy
Bourneville tuberous sclerosis	Defiscience	Rare diseases of brain development and intellectual disability
WEST syndrome
Polymicrogyria
Rett Sydrome
X fragile syndrome
Alagille syndrome	FILFOIE	Rare liver diseases
Sclerosing cholangitis
Budd Chiari syndrome
Spinal amyotrophy	FILNEMUS	Rare neuromuscular diseases
Ectodermal dysplasia	FIMARAD	Rare dermatological diseases
Bullous autoimmune dermatoses
Epidermolysis bullosa
Neurofibromatosis
Hirschsprung disease	FIMATHO	Rare abdominal-thoracic diseases
Oesophageal atresia
Short Bowel Syndrome
Thyroid ectopia	FIRENDO	Rare endocrine and gynaecological diseases
Turner syndrome
Cushing's disease
PFPA syndrome	Fai2r	Rare autoimmune and auto-inflammatory diseases
Congenital hyperinsulinism	G2M	Rare hereditary metabolic diseases
Phenylketonuria
Willebrand disease	MHEMO	Rare constitutional haemorrhagic diseases
Cystic fibrosis	Muco CFTR	Cystic fibrosis
Alport syndrome	ORKID	Rare renal diseases
Multicystic dysplastic kidney
Haemolytic uremic syndrome
Ehlers Danlos syndrome	OSCAR	Rare bone diseases
Osteogenesis imperfecta
Bardet Biedl syndrome	SENSGENE	Rare sensory diseases
Pierre Robin syndrome	TETECOU	Rare diseases of the head, neck and teeth
Moebius syndrome
Palate and cleft lip

The list of rare diseases was determined to perform an analysis based on sufficient data. We firstly considered the three most frequent diseases in each centre of expertise of Necker Hospital. To do so, we screened this list including 123 rare diseases into the National Data Bank on Rare Diseases, which collates all the data filled in by the centres of expertise. To select the conditions to be included in the study, we grossly reviewed all medical records of patients meeting the inclusion criteria for all conditions. Conditions for which the data in the medical records were poorly populated were excluded, ending with 39 identified rare diseases. We recruited patients from this study using the data warehouse, Dr Warehouse® (DrWH), from Necker-Enfants Malades Hospital. DrWH is a document-based open-source data warehouse oriented toward narrative clinical reports from the Electronic Health Records (EHRs). It contains more than 7.5 million clinical free-text documents produced at Necker Hospital from 2009, for more than 700,000 individuals and more than 20 medical departments. It allows searching for patients from structured data (e.g. biology) and free text (e.g. hospital reports) [5-7]. Dr Warehouse is a hospital data warehouse automatically fed by copying the electronic health records and other sources of data produced during the care of the patients. We used Dr Warehouse search engine to extract features localized in the free text (narrative reports) and almost never in the coded data (ICD10). All our questions and hypotheses are based on the way in which doctors fill in patients' medical records in their own words and not through specific coding.

The study population was constructed on each pathology studied using the Dr Warehouse® search engine by entering the inclusion criteria and the name of the concerned pathology. The sample selection was done in Dr Warehouse by a keyword search: “disease name” was used. Each patient was manually reviewed to exclude patients with an unconfirmed diagnosis.

Data collection

Dr Warehouse provides the functionality (called eCRF) to set up a clinical search form directly in the user interface. This tool was designed to facilitate the extraction of variables of interest from the EHR. For each variable, it proposes a value extracted from the EHR and displays the text area containing the items sought in the reports. The user must validate or modify the automatically extracted values.

We defined the variables (1) semi automatically extracted from textual data: main diagnosis (based on the list of rare diseases), oral status, phenotypic signs description (yes or no), orientation in an oral health care course, (2) automatically extracted from structured data: gender, age, postal code of residence, number of consultations in the hospital, number of hospitalizations in the hospital, number of different medical departments consulted, (3) and manually filled: Fdep (French DEPrivation index), index of social deprivation according to the postal code of residence), presence of a pervasive developmental disorder, and inclusion in a rare disease network.

To determine oral status, we searched for French words beginning with "dent" to find all clinical reports containing words in the field of dentistry (i.e., equivalent in English: tooth, teeth, dental, dentist, dentition etc.). We added in the research the following words to extend the semantic field: “carie (decay), gencive (gum), oral (oral), bouche (mouth). With this method, we were able to retrieve all patients with records mentioning these keywords.

Statistical analysis

Descriptive statistics

A descriptive analysis of the database generated was carried out. Categorical variables were summarized by frequency and percentage. Continuous variables were summarized by mean (1st; 3rd quartile), minimum and maximum.

Total patient data cross-tabulated with Social Deprivation Factor data, and information about the pathways caring for the patients. The FDEP was cross-referenced to the original table via zip code and a zip code to commune code conversion table. The original match data did not include a Paris commune code. They were replaced by the INSEE 2018 census data. The FDEP (2008) per postal code was calculated by averaging the FDEP of the corresponding communes weighted by their respective populations (2018).

The care course data were cross-referenced to the patient data via the actual principal diagnosis (i.e., the first value in the 'Principal diagnosis' column that contains more than one).

Analytical statistics

The model applied was a logistic regression where the output (dependent) variable was either the orientation in an oral healthcare or the description of oral phenotypic signs.

The importance of each predictor variable of interest was assessed using a likelihood ratio test comparing a full model containing said variable to a restricted model void of it. When deemed necessary, additional covariates were added to both models to control for potential confounding effects.

We were mainly interested in the effect of the services visited on oral health care. We started by testing the variables that will be used to fit the variables describing the services with models including only these variables.

Then we tested the variables describing the services visited globally (i.e., 1 variable = all the services visited) by including in the comparison model the variables to be adjusted. All the statistical tests were bilateral with a significance level of 5%, which means that a p value < 5% was considered significant.

All the variables tested up to this stage were corrected by the Bonferroni procedure applied to the number of variables tested * nb of outcomes. In total, we had 2 outcomes (phenotypic signs/orientation, 8 variables tested: Age, Sex, number of hospitalizations, number of consultations, number of different medical services, oral status, FDEP).

Results

Descriptive data

The sample studied included 2712 patients, children, and adults with 39 different pathologies, in 17 different rare disease networks.

The rare diseases studied and the networks they belong to were described in Table 1. We made our extraction on 39 rare diseases representing, almost in an exhaustive way, all the rare disease networks created in France and present at the Necker-Enfants-Malades hospital.

The study sample included 2712 patients, with 54.7% girls and 45.3% boys (Table 2). The population was adult and paediatric with an average age of 19.1 years and a median age of 14.5 years (Table 3). In the sample studied, 27.9% of patients had an acquisition delay or a pervasive developmental disorder. Among the patient files studied, oral and dental phenotypic signs were described for 18.40% of the patients, and an orientation in an oral healthcare was made in 15.60% of patients (Table 2).

Table 2

Description of the study population: categorical variables

Variables		n = 2712	%
Sex	Female	1494	55.1
	Male	1218	44.9
Age repartition	< 18 YO	1577	58.1
	≥ 18 YO	1135	41.9
Description of phenotypic signs	Yes	500	18.4
	No	2212	81.6
Orientation in an oral health care	Yes	422	15.6
	No	2290	84.4
Oral condition	Very bad	77	2.8
	Bad	189	7.0
	Good	211	7.8
	Not filled	2235	82.4
Presence of a pervasive developmental disorder	Yes	758	27.9
	No	1954	72.1

Table 3

Description of the study population: continuous variables

Variables	n	Mean	Std	25%	50%	75%	Min;max
Number of hospitalizations	2712	2.24	4.25	0	1	3	0;51
Number of consultations	2712	26.7	39.5	6	15	31	0;752
Median length of stay (days)	2712	3.38	6.85	0	1.54	4.13	0;101
Number of different medical services attended	2712	5.14	3.62	2	4	7	0;23
Fdep	2712	− 0.73	1.28	− 1.81	− 0.557	0.305	− 4.22;2.35
Age	2712	19.1	14.5	9	15	23	3;93

We also looked, adjusted on age, at the average number of consultations per patient (26.7), the average number of hospitalizations (2.24), the median length of hospital stays in days (3.38), and the average number of different hospital medical services that patients attended (5.14). We also calculated the average FDEP (− 0.73) (Table 3).

We studied the distribution of our variables of interest (description of phenotypic signs (Fig. 1) and orientation in an oral healthcare (Fig. 2) according to the patients' rare disease network and disease.

Fig. 1

Description of phenotypic signs according to the patients' rare disease network and disease. Counts were ordered by proportion of phenotype description. Count of patients are indicated in parentheses

Fig. 2

Description of orientation in an oral healthcare according to the patients' rare disease network and disease. Counts were ordered by proportion of orientation. Count of patients are indicated in parentheses

Networks whose practitioners most frequently noted oral phenotypic signs and referred the most patients for oral care were the TETECOU (rare diseases of the head, neck, and teeth), FIMARAD (rare dermatological diseases), OSCAR (rare bone diseases), and the MHEMO (haemophilia and other constitutional bleeding diseases) networks. In terms of pathologies, the oral phenotypic signs were most noted in patients with ectodermal dysplasia, osteogenesis imperfecta, and Pierre Robin syndrome (Fig. 1).

Analytic data

About orientation in an oral care course, we show a significant statistical association about the number of hospitalizations (corrected p = 2.81e−10), the number of consultations (corrected p = 7.16e−122), and the number of medical services attended (corrected p = 5.35e−22).

Regarding the description of oral phenotypic signs, we show a significant statistical association with respect to the number of hospitalizations (corrected p = 9.74e−07), the number of consultations (corrected p = 1.71e−13), and the number of medical services attended (corrected p = 8.15e−25), and the FDEP index (corrected p = 0.02) (Table 4).

Table 4

Association between the variables studied and orientation in an oral healthcare/description of oral phenotypic signs

Variables		Orientation*				Description of phenotypic signs**
		p value	p value cor	OR	95% IC	p value	p value cor	OR	95% IC
Age		0.11*	1	0.99	0.99;1.00	0.02*	0.29	0.99	0.98;1
Gender [ref = F]	M	0.26	1	1.13	0.88;1.33	0.10*	1	1.18	0.95;1.39
Number of hospitalizations (age)		1.75e−11*	2.81e−10	1.07	1.05;1.09	6.09e−08*	9.74e−07	1.06	1.03;1.08
Number of consultations (age)		4.47e−13*	7.16e−12	1.01	1.006;1.01	1.07e−14*	1.71e−13	1.01	1.007;1.01
Number of different medical services attended (age)		3.35e−23*	5.35e−22	1.14	1.12;1.18	4.06e−25*	6.5e−24	1.14	1.12;1.17
Fdep		0.00264*	0.04	1.13	1.04;1.23	0.00*	0.02	1.14	1.05;1.22
Presence of a pervasive developmental disorder [ref = no]	yes	0.799	1	1.03	0.84;1.35	0.23	1	0.87	0.73;1.13

Bolded numbers represent significant values

When covariates adjustment is included the names of the covariates are indicated within parentheses. When the variable is categorical the reference value used is indicated in square brackets.

is the uncorrected p value,

the corrected value (using a Bonferroni procedure with n test = 16), a p value with an asterisk * indicates inclusion as a covariate in further models based on a threshold of 0.2 on the nominal uncorrected p value

*Covariates selected for the outcome “Description of phenotypic signs”, are number of consultations from beginning, number of hospitalizations from beginning, FDEP, number of different medical services attended, age, gender

**Covariates selected for the outcome “orientation” are: number of consultations from beginning, number of hospitalizations from beginning, FDEP, number of different medical services attended, age

When covariates adjustment is included the names of the covariates are indicated within parentheses. When the variable is categorical the reference value used is indicated in square brackets.