Is X-ray Exposure Related to Dentist Remuneration?

Background: In many market settings individuals are encouraged to switch health care providers as a means of ensuring more competition. Switching may have a potentially undesirable side effect of increasing unnecessary treatment. Focusing on the most common source of medical radiation (dental X-rays), the purpose of this study was to assess whether, upon switching dentist, X-ray exposure increases depending on the type of provider payment. Methods: The analysis used longitudinal data from 2005 to 2016 covering a 5% random sample of the Scottish adult population covered by the National Health Service (NHS). Multiple xed-effects panel regression analyses were employed to determine the correlation of provider remuneration with patients' likelihood of receiving an X-ray upon switching to a new dentist other things equal. A broad set of covariates including a patient’s copayment status was controlled for. Results: Upon switching to a dentist who was paid fee-for-service, patients had a by 9.6 %-points (95% CI: 7.4%-11.8%) higher probability of receiving an X-ray, compared to switching to a salaried dentist. Results were robust when accounting for patient exemption status, as well as unobserved patient and dentist characteristics. Conclusions: In comparison to staying with the same dentist, patients may be exposed to substantially more X-rays upon switching to a dentist who is paid fee-for-service. There may need to be better guidance and regulation to protect the health of those who have to switch provider due to moving and greater caution in advocating voluntary switching.


Contributions To The Literature
Switching between providers is encouraged in many health care markets as a means to increase competition.
We found evidence that patients visiting a dentist for the rst time receive more dental radiographs based on the dentist's remuneration status.
These ndings highlight the importance of addressing providers' nancial incentives when implementing health care market reforms.

Background
Choice of health care provider is often dictated by location. People who move over large distances have to switch provider. In addition health care systems have increasingly seen the introduction of market mechanisms (1,2) in which switching of providers is encouraged (3) as a means of improving competition. Contact with a new patient is an opportunity to offer treatment and so a potentially undesirable medical side effects of patient switching is excessive treatment. It is well-established that some types of provider remuneration such as fee-for-service (4), increase the propensity to treat leading for calls for it to be replaced or regulated. Where a treatment is potentially harmful this implies that there might be a risk to a patient who is either forced to or chooses to switch provider. To this end, we examine whether patients are exposed to additional ionizing radiation upon switching between dentists under different provider payment schemes. We exploit large-volume individual-level administrative data and the unique characteristics of the most frequently applied type of medical imaging: dental X-rays. We focus on variations in the use of dental radiographs when patients switch between providers who receive either fee-for-service or salary payments.
Medical radiography is associated with a radiation risk for patients. While X-rays serve a crucial role in diagnostics, radiation exposure is a well-known human carcinogen (5). Dental X-rays are among the most used medical radiographs and the most common source of arti cial radiation exposure (6). Current clinical guidelines in many countries con rm that dental radiographs should only be used if the patient's bene ts exceed the risks, that is only upon strict clinical indication (7). Nonetheless the procedure is frequently delivered as part of a routine examination, especially when visiting a dentist for the rst time.
There is little evidence on the exact carcinogenic effects of low dose radiation, such as caused by dental radiographs, because the sample sizes required to study them would need to be extremely large (8). The consensus held by radiological protection organizations is based on the so-called linear no threshold (LNT) model, which postulates that risk of radiation induced cancer increases linearly with exposure and that there is no safe threshold (9). Indeed, a number of studies have been able to show an association between dental X-rays and the risk for meningiomas and salivary tumours (6,10,11).
It has been established that health care providers adapt their behaviour to nancial incentives (12).
Ideally, nancial rewards can be used to improve providers' performance and the quality of services provided (13). Empirical evidence however remains weak that these goals are easily achieved (14,15).
Previous studies have found that the introduction of nancial incentives has increased treatment intensity and created supplier induced demand. Evidence indicates that this often constitutes ineffective care with little effects on patient health, rather than a reduction of existing rationing (16).
Most of the existing literature on the topic has focused on general medicine, however there is a growing body of evidence looking at dental care in particular (17,18). Studies show that nancial incentives have the potential to increase utilization of dental check-ups (19) and increase the number of individuals under dentist supervision without a reduction in service quality (20). However remuneration schemes used for dentists can enable the provision of additional care with limited or no marginal bene ts for patients (21) with potential adverse effects for patients (22). Notably, fee-for-service payments have been shown to increase the probability of dental radiographs being provided to patients (23). Ours is the rst study to focus on the impact of switching dentist on this potentially harmful treatment.
Based on existing studies of the impact of fee-for-service arrangements in general and for dentistry in particular, our hypotheses were that (i) patients receive signi cantly more dental radiographs upon switching to another dentist; and (ii) the effect on dental radiographs upon switching to another dentist is signi cantly larger under fee-for-service as compared to salary payment.

Institutional Background
Dental care in Scotland is provided through a public-private mix. Public dental care is primarily delivered by the NHS General Dental Service. For the most part GDS is provided by private dentists that have arrangements with the NHS. They are being remunerated based on a mix of capitation and fee-for-service payments. The exact arrangements of the GDS reimbursements are detailed in the Statement of Dental Remuneration (24). There are also a number of salaried dentists that provide GDS in areas where independent private dentists didn't meet local needs. Since 2014 these salaried dentists are part of the Public Dental Service (PDS), whose main responsibility is to provide GDS to patients that cannot access care from private dentists. In general patients receiving care under the GDS pay 80% of the treatment fees up to a certain limit out-of-pocket. However patients can be exempt from these charges for a number of different reasons (25). Private dentists are free to provide dental care to paying patients outside of the GDS. The fact that our data set does not include any information on these services should not cause problems for validity of the results. Our results would depict lower bound estimates, if X-rays are provided privately on a regular basis.
The guideline for oral health assessment published by the Scottish Dental Clinical Effectiveness Programme (SDCEP) determines the situations in which dental radiographs should be taken as a diagnostic tool (26). In particular the SDCEP determines that "routine 'screening' radiography (e.g. panoramic radiographs) must not be carried out and new radiographs must not be undertaken without rst examining existing lms or without clear justi cation". In cases where radiographs are necessary, dentists should undertake all necessary steps to minimize the need for further radiography.

Data
Our study is based on a data set originating from the GDS database, provided to us by the Information Services Division (ISD) of NHS Scotland. The data was extracted from the Management Information and Dental Accounting System (MIDAS) that contains all treatment claims made under the GDS. ISD drew a random sample of 5% all individuals registered in the GDS in 2016 and tracked them back until 2005. Our data set thus includes the full dental histories of all sampled individuals allowing us to conduct a panel analysis. In total our data set contains 4,369,961 claimed services provided to 146,239 (around 2.7% patients by 4,894 different dentists over the 11-year study period. Besides the detailed information on dental claims, our data set included variables on various patient and dentist characteristics. We collapsed the data onto the treatment band level for analysis. Treatment bands were de ned as all treatments that were initiated on the same starting date, disregarding actual date of delivery and all accompanying treatments. The reason for this choice is based on the GDS practice. When a patient visits a dentist, the latter develops a treatment plan (if necessary) that might include multiple follow-up visits. In the MIDAS database all these treatments are recorded as having the same start date. We only retained the treatment bands, where a patient visits a dentist for the rst time during the study period and having received a dental X-ray during a previous treatment band by a different provider. Figure 1 illustrates this exclusion process.
[ Figure 1 goes here] Outcomes and Covariates Table 1 summarizes the operationalization of the outcome variable and chosen covariates. The outcome variable for our analyses was the delivery of at least one dental X-ray (small dental lm) to a patient during any given treatment band. The main independent variable of interest is the dentists' remuneration status. We also take into account the dentists' experience. This is based on the date, the dentists rst registered with the GDS. While dentists theoretically could already have practiced privately beforehand, we assume that for the vast majority, this represents their professional experience accurately. The time elapsed in between treatment bands is also controlled for. We expect that the longer that duration, the higher the probability of severe changes in dental health status, which would render the provision of dental radiographs recommended in accordance with the SDECP guidelines.
Patient characteristics that we controlled for include patient age and deprivation of the area of residence. The latter is based on the Scottish Index of Multiple Deprivation that takes into account a number of socio-economic characteristics across seven different domains (27). The scores are applied to 6,976 geographical areas of equal population. Additionally, we controlled for patients being exempt from copayments, to check to the effect of patient nancial incentives.

Analytical Approach
This study adopts a previously established theoretical framework concerning the role of nancial incentives in relation to dental x-raying (23). The framework takes into account that X-ray provision depends on the interaction between patients and dentists. In general, the probability of a dental radiographic examination is dependent on the dentist and patient characteristics, the dentist's remuneration scheme, the patient's dental health status and the patient's own nancial contribution. If dentists act altruistically in the best interest of their patients, their reimbursement scheme should not be signi cantly associated with the probability of providing a dental radiograph.
This framework can be operationalized as a linear probability model (23). While the binary outcome variable opens the possibility of using non-linear regression models, we chose to follow a linear approach for the practical reasons of computation speed and ease of interpretation of the regression coe cients. As we are interested in analysing the correlation of dentist reimbursement with X-ray provision, we have to confront the fact that, relying on observational data, reimbursement status is not randomly assigned, thus creating a possible endogeneity bias. We address this issue by exploiting the longitudinal nature and panel characteristics of our data in a multiple xed effects model. Through repeated observations of the same dentists and estimating dentist xed effects to correct for unobserved heterogeneity, we can take into account the effect of changes in reimbursement status. Similarly, many patient characteristics are not contained in the data. To avoid biased results based on unobserved time-constant characteristics, we also estimate patient xed effects.
We estimated the regression model for the sample of treatment bands, where patient visited a dentist for the rst time and have previously received an X-ray during our study period. We are thereby able to estimate the effects of remuneration status on X-ray provision for new patients. As a control we also estimated the model for the overall sample of treatment bands. All calculations were carried out using Stata/SE 14.2 (28); the two-way xed effects estimations were carried out using the felsdvreg command. Table 2 summarizes the descriptive statistics of our sample. Among the 399,115 treatments sessions where people visited a new dentist for the rst time, 28.32% (113,010) included a dental X-ray. We observe a big difference between fee-for-service dentists 28.93% (106,567), compared to the 20.93% (6,443) of salaried dentists in terms of providing a dental X-ray during the rst visit. Our two-way xed effects regression model (Table 3) con rms the substantial difference between salaried and fee-for-service dentists observed on the descriptive level. The probability of a salaried dentist providing an X-ray for a new patient is 9.6 percentage points (95% CI [-0.118; -0.074]) lower than for their fee-for-service peers. We nd that a patient's deprivation ( is also negatively associated with X-rays, meaning that exempt patients are less likely to receive one. A longer duration in between treatment bands increased the likelihood of the patient receiving an X-ray (0.0039 [0.0037; 0.0039]). The effect of a dentist's experience approaches 0 in the analysis.

Discussion
Individuals move between health providers for a variety of reasons both non-discretionary and in response to encouragement to seek out better health care. This study shows that patients visiting a dentist for the rst time are more likely to be X-rayed if the dentist is remunerated on a fee-for-service scheme, leading to potential excess radiation exposure. We focused on the association between reimbursement mechanisms and dental radiograph frequency, which allows us to identify a difference in treatment that should not exist. This difference indicates potentially inequitable treatment based on provider remuneration leading to excess radiation exposure or risk of under-treatment.
Dental radiographs are the most common form of arti cial radiation exposure (6). Clinical guidelines limit their use to situations, where the bene ts outweigh the risks for the patients (26). We con rm and add to the previous results that fee-for-service payments are associated with more dental radiographs (23) and show that the difference in frequency of X-ray provision is even more substantial during rst time visits. Whilst dental radiographs are an indispensable diagnostic tool, limiting unnecessary X-rays should be among the goals of every health system. Based on the regression results presented in Table 3, and considering the care provided by salaried dentists as reference level, this would mean that switches to fee-for-service dentists during the observation period were correlated 38,315 dental X-rays in excess of switching to a salaried dentist. Accordingly, the estimated excess radiation exposure experienced by our study population, relative to if dentists were paid on a xed salary would amount to 11,495-827,604 µSv (29).
This study has a number of implications for the regulation and oversight of dental health care provision. It illustrates a continuing risk of fee-for-service arrangements in leading to excessive X-rays and identi es an at-risk group of patients -those that have to, or choose to, switch dentist and arrive at a fee-forservice provider. It also highlights a hitherto undocumented risk of encouraging patients to switch between providers. Whilst that encouragement might be bene cial in respect of improving the market for dental health care, it might also be placing patients at risk.

Conclusions
The provision of dental radiographs should be linked to clear bene ts for the patients and not carried out as a routine treatment. Our ndings indicate that patients visiting a dentist for the rst time are more likely to receive a dental radiograph, if that dentist is paid fee-for-service. This highlights that fee-forservice remuneration can potentially lead to adverse effects for patients. Furthermore, these adverse effects can be exacerbated through reforms that encourage patients to shop around for providers. When implementing such reforms to improve competition, the effects of nancial incentives need to be considered.

Declarations
Ethics approval and consent to participate Access to the data and approval for the research was provided by the Public Bene t and Privacy Panel for Health and Social Care-HSC-PBPP under the reference 1516-0557.

Consent for publication
No individual person's data is included in the analysis or manuscript.

Availabitlity of data and materials
The analysis exploits individual-level claims data from NHS Scotland. Due to data protection regulations, we're not allowed to publicly share the original research data and can only provide do-les of the analysis and aggregated information about the dataset if desired.

Competing Interest
All authors declare: no support from any organisation for the submitted work; no nancial relationships with any organisations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have in uenced the submitted work.

Funding
This research was conducted as part of the ADVOCATE project (Added Value for Oral Care). The ADVOCATE project has received funding from the European Commission's Horizon 2020 research and innovation program under grant agreement 635183: http://advocateoralhealth.com.
Authors' contributions MC and SL developed the analytical approach. OK analyzed the data. MC and SL contributed to the interpretation of the results. OK was the major contributor to the manuscript. MC and SL provided comments and read and approved the nal manuscript. Figure 1 Sample selection process.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. STROBEChecklist.pdf