Characteristics of dentists and patients associated with appropriate antibiotic prescriptions by French dentists: a cross-sectional study using Health Insurance databases

We considered in the present study all general dental practitioners of the Grand Est region (5,550,000 inhabitants [19]) in France (67,000,000 inhabitants), who had seen more than 100 different patients and who had prescribed more than ten antibiotics in 2019. We chose to use 2019 data because dental activity was impacted in 2020 because of the COVID-19 pandemic.

We conducted a one-year (2019) retrospective cross-sectional study using the reimbursement databases of the French Health Insurance, that covers 99% of the population. In France, all antibiotics must be prescribed by health professionals and are then reimbursed by the Health Insurance. The Health Insurance reimbursement databases can link each antibiotic dispensed by community pharmacies to the prescribers’ and patients’ characteristics such as age, gender, and the presence of chronic disease. Information on the prescription is not available, so dispensation was used as a proxy for prescription.

We previously developed and validated four PIs estimating the appropriateness of antibiotic prescriptions at the dentist level, calculable without information on clinical diagnoses and using routine reimbursement databases [17]. These PIs, selected from the literature and based on national guidelines [20], cover the most frequent situations of inappropriate prescribing in dental care. They were associated with a clear target reflecting a high quality of care. Performance was defined as the percentage of dentists who reached the target. Additional file 1: Table S1 presents the definition (numerator/denominator) of the four PIs and their associated target [17].

An appropriateness score was calculated for each dentist as follows [18]:

A PI was considered applicable if it was based on at least ten situations per year for the dentist. A cluster analysis was then conducted to group dentists according to their appropriateness score, i.e., to the appropriateness of their antibiotic prescribing practices. The Ward method was used to identify clusters according to the degree of similarity between observations in a same cluster.

The association between the appropriateness of antibiotic prescribing practices and the following 2019 dentists’ and patients’ characteristics was investigated. For each included dentist, we considered: the geographic area of professional practice (Alsace, Champagne-Ardenne or Lorraine); the dentist’s gender; the practice setting (rural (< 5,000 inhabitants), suburban (5,000 to 20,000 inhabitants) or urban (> 20,000 inhabitants)); the number of years of practice; the main type of procedure performed in daily practice (prosthetic, prevention, operative dentistry, implantology, parodontology, or surgery); the number of different patients seen during the year; the number of procedures per patient (in France, the unit of reimbursement is the procedure, which may consist in several acts and/or several visits); the percentage of patients with chronic diseases among patients seen during the year [21]; the percentage of patients with low income [22]; the percentage of patients < 16 years old; and the percentage of patients > 65 years old.

The association between the appropriateness of antibiotic prescribing practices and the 2019 dentists’ drug prescribing profiles was explored considering the following variables: the total drugs expenses per patient; the number of units of all drugs per patient; the number of antibiotic prescriptions per 1000 patients; the number of systemic non-steroidal anti-inflammatory drugs per 1000 patients; and the number of systemic corticosteroids per 1000 patients.

The PIs’ results were calculated and presented as medians and interquartile ranges (IQR), and performance (i.e. the percentage of dentists who reached the target) as percentage.

A hierarchical clustering method was conducted to identify clusters of dentists according to their appropriateness score, i.e. the appropriateness of their antibiotic prescribing practices. For each identified cluster, the median appropriateness score, its IQR and the performance for each PI are presented.

The dentists’ and patients’ characteristics were compared between clusters using Chi-square tests for categorical variables (expressed as percentage) and Kruskal–Wallis tests for continuous variables (expressed as medians and IQR). Characteristics associated with appropriateness of practices with a p-value < 0.10 in these analyses were then included in a multivariable polytomous logistic regression model to identify characteristics independently associated with the clusters with less appropriate practices as compared with the one with the most appropriate practices. When two characteristics were highly collinear, the one with the lower p-value in bivariate analyses was considered in the multivariable model.

For dentists’ drug prescribing profiles, only bivariate analyses (Chi-square or Kruskal–Wallis tests) were used to compare the clusters, because of the high collinearity between all considered variables. However, if appropriate practices were significantly associated with ‘total drugs expenses per patient’ and /or ‘number of units of all drugs per patient’, we planned to adjust analyses on the number of procedures per patient and the percentage of patients with chronic diseases (variables reflecting patients’ health), in order to interpret these results (i.e., to identify if the volume of drug prescriptions was due to differences in patients’ health or in dentists’ practices).

In 2019, 3,372 general dental practitioners practiced in the Grand Est region of France, and 3,014 (89.4%) met the eligibility criteria for the present study (> 100 patients and > 10 antibiotics prescribed in 2019). Their mean age was 47.5 ± 12.6 years, with a mean number of years of practice of 17.7 ± 12.6, and 56.9% were men. On average, they took care of 804 ± 391 different patients in 2019 for a mean number of 2,439 ± 1,212 procedures. The 3,014 dentists prescribed 373,975 antibiotics in 2019, that were dispensed to 308,123 different patients. Among these patients, 16.7% were aged < 16 years and 17.3% > 65 years, 11.8% had a chronic disease, and 7% had a low income.

Additional file 1: Table S2 presents the results for the four proxy indicators [17]. Important variations between dentists were noted, with large interquartile ranges. Performance (percentage of dentists who reached the target) is presented in Table 1. The performance results were not optimal, particularly for the PI 4 (10.5% of dentists prescribed less than 5% of antibiotics rarely indicated in dental care) and PI 1 (39.6% of dentists prescribed at least 10 times more amoxicillin than amoxicillin-clavulanate).

Three clusters of dentists were identified. Cluster 1 grouped 686 dentists (22.8%) having better than average practices, Cluster 2 grouped 1,241 dentists (41.2%) having average practices and Cluster 3 grouped 1,087 dentists (36.1%) having worse than average practices. Table 1 presents the performances of the three dentists’ clusters for each PI. Substantial differences in performances were noted between Cluster 1 and Cluster 3 (from a 30.0% difference for PI 4 to a 70.6% difference for PI 1), highlighting a huge variability in dentists’ practices.

Table 2 presents the results of the bivariate analyses. All dentists’ characteristics were significantly associated with the appropriateness of their antibiotic prescriptions, except the practice setting. Regarding patients’ characteristics, the percentages of patients with chronic diseases and > 65 years old were significantly associated with the appropriateness of dentists’ antibiotic prescriptions.

Table 3 presents the results of the multivariable polytomous logistic analyses. The geographic area and the number of procedures per patient were significantly associated with the appropriateness of antibiotic prescriptions, as well as the dentist’s gender and the number of years of practice (only for Cluster 3). The number of patients was associated with the appropriateness of antibiotic prescriptions but not linearly: better practices were associated with an intermediate number of patients.

Overall, dentists who had the most appropriate antibiotic prescriptions were male with less years of practice, practicing in the Lorraine region, having a predominant surgery practice, with a lower percentage of patients aged > 65 years old, a lower percentage of patients with chronic diseases, and fewer procedures per patient.

All drug prescribing variables were significantly associated with the appropriateness of antibiotic prescriptions, except the number of systemic corticosteroids prescriptions per 1000 patients (Table 4). The associations between the appropriateness of antibiotic prescriptions and the two variables, the total drug expenses per patient and the number of units of all drugs per patient, were not modified when adjusted for patients’ health (i.e., the number of procedures per patient and the percentage of patients with chronic diseases).

Overall, dentists who had the most appropriate antibiotic prescriptions prescribed less drugs, less antibiotics, less systemic non-steroidal anti-inflammatory per patient, and had less drug expenses per patient.

The present cross-sectional study allowed us to identify three clusters of dentists according to the appropriateness of their antibiotic prescription practices: 23% of our sample of 3,014 dentists belonged to the “better than average practices” cluster, 41% to the “average practices” cluster and 36% to the “worse than average practices” one. Performance results showed that dentists’ practices were suboptimal and highly variable between dentists and between PIs.

We identified three types of characteristics that were associated with the appropriateness of antibiotic prescriptions. The first one was characteristics related to dentists. Male dentists in the Lorraine region, with less years of practice, and having a predominant surgery practice, had the best antibiotics’ prescription practices. As it has been shown for general practitioners [23], we can hypothesise that male dentists might have a better attendance to postgraduate training, which has a positive impact on knowledge [24, 25], and thus on compliance with guidelines. But this hypothesis would need further investigation. The better practices observed for the Lorraine area might be due to the differences in the training provided between the local universities (one for each area: Alsace, Champagne-Ardenne and Lorraine), and/or to the presence of the regional antimicrobial stewardship network AntibioEst [26] which started in Lorraine in 2003, before being extended to the whole Grand Est region in 2018. This regional network coordinates a multifaceted AMS strategy, including for example postgraduate training, guidelines, or infectious diseases advice freely available by phone. Previous studies in dental care have already shown that a more recent year of graduation is associated with a better level of knowledge and practice in antibiotic prescribing [24, 27]. Finally, dentists who have a predominant surgery practice could be more aware of guidelines because surgery might be at higher risk of complications than other procedures, and/or because they have less emergency as these procedures are scheduled, so their antibiotic prescriptions might be more anticipated and protocolised.

The second type of characteristics was characteristics related to patients. Dentists with more complex patients, i.e. with older patients, more patients with chronic diseases, and patients who received more procedures, were more likely to have worse than average prescription practices. As shown elsewhere [28], dentists might be more likely to prescribe antibiotics inappropriately, in particular an increased use of broad-spectrum antibiotics, to prevent a perceived increased risk of bacterial complications in these patients.

Finally, characteristics related to dentists’ drug prescribing profile were associated with antibiotics prescribing practices. Dentists with worse than average practices had higher drug expenses and prescribed more drugs, more antibiotics and more systemic non-steroidal anti-inflammatory drugs (non-steroidal anti-inflammatory drugs are usually not indicated in bacterial infections in dentistry [29]). We showed that this was not related to patients’ health but to dentists’ drug prescriptions habits. Overall, dentists who prescribed antibiotics more appropriately seemed to have an overall more prudent use of drugs, as already observed for general practitioners [18].

Few published studies aimed at identifying characteristics associated with the appropriateness of antibiotic prescriptions made by dentists. Kerr et al. conducted in 2020 an online questionnaire-based survey to investigate potential factors that may impact the appropriateness of antibiotic prescriptions to treat adults with acute dental pain in the United Kingdom (UK). The inappropriate prescribing practice was defined as a prescription of antibiotic(s) in a clinical situation that does not require it. The authors identified that the likelihood of inappropriate prescriptions was increased by the following factors: the qualification from a non-UK university, the low or no confidence in achieving adequate local anaesthesia, the duration of the appointment (< 20 min versus all other times), and the lack of a postgraduate qualification [25].

A scoping review conducted in 2018 identified dentist-related factors influencing the use and the misuse of antibiotics in dentistry [14]. Some of the characteristics we identified in the present study were already reported in the literature: practice location, practice type, years in practice, and dentist’s gender.

This study brings innovative and useful findings, as data on the topic is scarce. However, our work has some limitations. First, we aimed to explore the association between characteristics related to dentists or their patients and the appropriateness of their antibiotic prescriptions, but we cannot prove a causal relationship. Therefore, more studies should be conducted to further investigate our results. Second, the characteristics we investigated were limited to those available from the routine reimbursement databases. We acknowledge that they do not explain all the variability of practices, and that other characteristics—not explored here—could also influence this variability.

The present study showed that performance results of dentists were far from optimal, especially in the clusters of dentists with average and worse than average practices, and presented a wide variability between clusters, highlighting the urgent need for AMS interventions in dentistry. Therefore, this study could be useful to help guiding AMS interventions. First, AMS interventions can target dentists belonging to the “worse than average” cluster and focus on the situations covered by PIs that showed the worst performances (e.g., prescription of amoxicillin rather than amoxicillin-clavulanate, indications for prescription of pristinamycin, spiramycin-metronidazole or doxycycline). In addition, AMS interventions can target dentists presenting characteristics that were associated with inappropriate prescriptions, such as those located in specific geographic area or those practicing for a long time. Tailoring AMS strategies and interventions to these prescribers and to situations of highest concern (i.e., frequently source of antibiotic misuse) might help to reduce the number of antibiotic prescriptions by dentists and to improve their appropriateness [30]. Consequently, AMS in dentistry might lead to an optimisation of the clinical management of infections by dentists, with an improvement of clinical results and no increase of unintended consequences (i.e., toxicity and resistance emergence) [31].

The method of the present study might also be used in other countries, by adapting the PIs and their targets to the national prescribing guidelines and exploring relevant characteristics.