Hidden Bedside Rationing in the Netherlands: a Cross-sectional Survey Among Internists

Background Healthcare rationing can be dened as withholding benecial care for cost reasons. One form in particular, hidden bedside rationing, is problematic because it may result in conicting loyalties for physicians, unfair inequality among patients and illegitimate distribution of resources. Our aim is to establish whether bedside rationing occurs in the Netherlands, whether it qualies as hidden and what physician characteristics are associated with its practice. Methods Cross-sectional online questionnaire on knowledge of -, experience with -, and opinion on rationing among physicians in internal medicine within the Dutch healthcare system. Multivariable ordinal logistic regression was used to explore relations between hidden bedside rationing and physician characteristics. Results The survey was distributed among 1139 physicians across 11 hospitals with a response rate of 18% (n=203). Most participants 64%) had prescribing cheaper of treatment while a more but expensive alternative bedside rationing.

and Goold have proposed three conditions by which to recognise bedside rationing: "a physician must (1) withhold, withdraw, or fail to recommend a service that, in the physician's best clinical judgment, is in the patient's best medical interests; (2) act primarily to promote the nancial interests of someone other than the patient (including an organisation, society at large, and the physician himself or herself); and (3) have control over the use of the medically bene cial service". (5,7) When bedside rationing is done without revealing the decision to ration or its rationale to the patient, it quali es as hidden bedside rationing. (2,8) Hidden bedside rationing can result in unfair inequality and illegitimate distribution of resources. (9)(10)(11)(12) Furthermore, it opens up the possibility for distribution or restriction of resources based on clinically irrelevant characteristics such as ethnicity, gender, age or skin colour. (13) Hidden bedside rationing inherently violates informed consent because a patient consents to a course of treatment while being insu ciently informed about an existing, more expensive, alternative. (8,14) Most empirical research into healthcare rationing or, more speci cally, bedside rationing, focuses on the willingness to ration or on factors affecting rationing decisions. (15) A European cross-sectional study estimated that 56.3% of physicians engage in bedside rationing. (16) In the United Kingdom, the NHS rations care on different levels but many decisions are still left to the bedside. (17) Yet little evidence is available regarding the practice of hidden bedside rationing in the Netherlands with its regulated competitive healthcare market. Our objective was to establish whether bedside rationing occurs in the Netherlands, whether it quali es as hidden and which physician characteristics are associated with its everyday practice.

Study Design and Participants
All (seven) Dutch academic hospitals were invited to participate, of which one declined. We subsequently invited one large general hospital within the region of each participating academic hospital, of which also one declined. Thus, six academic and ve general hospitals participated. We conducted a cross-sectional survey among internists (i.e. consultants) and internal medicine residents (i.e. specialist registrars) employed in these hospitals. All eligible physicians were approached through e-mail with a standardised open invitation to participate. [Additional File 1] Participation was voluntary and no remuneration was offered.

Setting
The Netherlands has a regulated competitive universal health insurance system as of 2006.(18) Private health insurance is mandatory for all citizens while health insurers are required to cover a statutory bene t package. The government de nes the content of this bene t package upon advice of the National Healthcare Institute.(18) Government regulates and subsidises insurance and insurers are not allowed to reject applicants. The vast majority of hospital care is included in the statutory bene t package, but all adults are required to pay an annual maximised deductible (EUR 385 / USD 465). In 2015 0,2% of the population was uninsured. (18) The Dutch medical profession includes approximately 20,000 specialists (i.e. consultants) of whom 50% are salaried, 37.5% are in private practice (under fee-for-service or lump-sum) and 12.5% are combining the two. (19) The majority of hospitals are non-pro t organisations. (18) Specialists are paid by hospitals, while hospital payment is negotiated between insurers and individual hospitals and is often case-based. (18) Identifying opportunities for bedside rationing within a healthcare system can be complicated. (5,7) The conditions proposed by Ubel and Goold entail that opportunities for bedside rationing are potentially abundant in the Dutch healthcare system, as physicians maintain a high degree of control over the use of medically bene cial services. (18,20) Salaried physicians have negligible nancial self-interest in the Dutch context and because balance-billing is not allowed this is also limited for physicians in private practice. (18,20) However, promoting the nancial interest of the employing organisation or society at large remain possible arguments in the Dutch setting. (18,20) De nition of Variables Bedside rationing was de ned as by Ubel and Goold.(5) Bedside rationing was subsequently considered 'hidden' when it complied with two additional conditions: both the decision to set limits as well as the rationale for that decision were not disclosed to patients. (12) Four physician characteristics were hypothesised to in uence (hidden) bedside rationing: 'year of graduation', 'residency', 'type of employing hospital' and 'mode of employment'. Year of graduation and residency were considered because of the possible difference in medical education regarding costconsciousness as well as the difference in clinical experience. The type of employing hospital (academic or general) was considered, not only because of the differences in caseload but also because of a possible difference in employed physician characteristics. Finally, the mode of compensation has been shown to in uence physician behaviour and therefore mode of employment (salaried, private practice or other) was considered as well. (21,22) Survey The survey was constructed and revised in cooperation with medical professionals. [Additional File 2] A pilot was tested among 20 physicians and the survey was adapted accordingly.(23) Both the content and phrasing of the survey built upon previous research. (6,15) The choice for a web-survey was made given the ease of distribution and data-collection as well as the fact that it was less prone to socially desirable answers. (24) As the concept of 'rationing' has different connotations, the survey speci cally addressed the study de nitions of both healthcare rationing and bedside rationing. (25) The survey consisted of four questions regarding everyday practice and four statements regarding physician opinion. Questions used a ve-point Likert scale, with a sixth option to give an alternative answer. (26,27) Because the dimension in question was frequency, answers ranged from never to always.(28) Statements used a seven-point Likert scale with answers ranging from completely disagree (1) to completely agree (7). (26,27) Data Collection and Analysis All data were collected through Explora, a secure online research tool developed by the Dutch Hospital Association. Data were analysed using descriptive statistics. Likert scales 1-3 were grouped as 'disagreed to some extent' and 5-7 as 'agreed to some extent'. Missing physician characteristics for completed surveys were imputed using multiple imputation (n=5). Due to the small sample size and unknown distribution of relevant characteristics over the total target population, survey data were not weighted.
The relations between the independent variables and bedside rationing as well as disclosure were explored with univariable and multivariable proportional odds logistic regression for ordinal outcomes.
Bedside rationing was operationalised as a dependent variable as a response of 'often' or 'always' to item C in the survey. [Additional File 2] Hidden bedside rationing was operationalised as a dependent variable as a response of 'not often' or 'never' to item D in the survey. [Additional File 2] The effect of the independent variables was expressed as odds ratios and 95% con dence intervals. Overall variance explained by the models was expressed as Pseudo R2. Subgroup analyses were not planned due to the number of invited participants and limited number of expected participants. We used the STROBE cross sectional checklist when writing our report.(29) All statistical tests were performed in IBM SPSS Statistics 24.

Participants
An invitation to participate was sent to 1139 internists and residents. A total of 240 responded (21%) of which 34 respondents did not complete the survey and three were junior doctors. After excluding these, a total of 203 (18%) participants remained of which 86 (42%) were residents and 109 (54%) were employed at an academic hospital. [ Table 1] With regard to employment mode, 105 (52%) participants indicated to be in training yet only 86 (42%) reported to be a resident. This difference was most likely explained by nal year residents working within a fellowship in their respective sub-specialisation and were therefore considered as being in training.  (7) Intensive Care 5 (3) Nephrology 18 (9) Oncology 15 (7) Vascular Disease 15 (7) Resident 86 (42) Not willing to answer 5 (3)

Bedside Rationing
A total of 88 (43%) participants could envision a physician denying a patient a course of treatment because of cost consideration. When asked about their everyday practice, 129 (64%) participants had prescribed a cheaper course of treatment while a more effective, but more expensive, alternative had been available. Of these physicians 13 (6%) reported doing so often. Other answers (n=7; 3%) included physicians who disclosed they did not feel informed enough about costs in order to consider it. All physicians with experience prescribing cheaper treatment were subsequently asked how often they disclose this to their patients. Most (n=97; 71%) disclosed this on some occasions. Sixteen participants (12%) reported to always disclose prescribing less effective but cheaper treatment while thirty-two (24%) indicated to never disclose such decisions. [ Figure 1] Figure 1 Histograms of Survey Outcome per ItemItems A-D: Questions regarding everyday practice on a ve-point scale ranging from never to always with a sixth option to give an alternative answer. Item D was answered by a subset of participants (n=136) who gave a response to item C other than 'Never'. Items E-H: Statements regarding physician opinion on a seven-point Likert scale ranging from completely disagree (1) to completely agree (7).

* The Y-axis depicts number of participants † Numbers above bars indicate total number of responses per answer
Nearly three quarter of participants (n=148; 73%) agreed to some extent that a physician ought to explain their cost considerations when withholding a course of treatment because this is considered too expensive. Fifty-eight participants (29%) were in complete agreement and only four (2%) completely disagreed. Among the additional clari cations, participants agreed that explanation should be provided, but the decision to ration ought not to be made by physicians in the consulting room but by policymakers in an overarching mechanism. Treatment Cost Some physicians reported lacking knowledge of treatment cost to in order to consider it. Half of participants (n=100; 49%) reported they rarely feel su ciently informed about treatment cost to discuss it, while two per cent (n=4) always feels equipped to do so. None reported to always bring up treatment cost, compared to sixty participants that reported never to do so (30%). Most indicated they do not discuss it often (n=93; 46%) opposed to three (2%) that bring up treatment cost regularly.

Physician Characteristics Affecting Bedside Rationing
Physicians working in an academic hospital reported more bedside rationing behaviour compared to physicians working in a general hospital, OR=21 (95% CI 15-27, p<0.001). Residents reported more bedside rationing behaviour compared to internists, OR=11 (95% CI 8.1-14, p<0.001). In the multivariable analysis only 'type of hospital' had a statistically signi cant effect. Physicians working in an academic hospital reported a higher frequency of bedside rationing behaviour compared to physicians working in a general hospital, OR=17 (95% CI 6.1-48, p<0.001). Pseudo R2 for this model was 0.43. [ Table 2] Residents reported more disclosure compared to internists, OR=24 (95% CI 17-34, p<0.001). An increase in years since graduation (per 10 years) was associated with an increase in the odds of disclosure, OR=1.7 (95% CI 1.5-1.9, p<0.001). In the multivariate analysis 'residency', 'mode of employment', 'mode of payment and 'years since graduation (per 10 years)' had a statistically signi cant effect on disclosure. An increase in years since graduation (per 10 years) was associated with an increase in the odds of disclosure, with an odds ratio of 1.5 (95% CI 1.

Discussion
The objective of our study was to establish whether bedside rationing occurs in the Netherlands, whether it quali es as hidden and whether there are associations with speci c physician characteristics. To our knowledge, this is the rst empirical study that explores physicians' knowledge of -, experience with -, and opinion on bedside rationing in the Netherlands. First, we found that physicians occasionally limit patient choice for reasons of cost, indicating that bedside rationing indeed occurs. Second, we found that the decision to set these limits and its rationale are not always disclosed to the patient, qualifying it as hidden. Nearly two thirds of participating physicians indicate they have at some point engaged in bedside rationing. Although the occurrence of bedside rationing is in line with previous evidence, the extent is larger than expected. (16,30) Residency and working at an academic hospital were associated with more bedside rationing. Furthermore, residents are more likely to disclose bedside rationing decisions to their patients than internists while salaried physicians are less likely to do so than physicians who are in private practice. Hidden bedside rationing in the Netherlands therefore seems to be more prevalent among salaried internists.
The consequences of hidden bedside rationing are potentially severe.(9,10,13) Most physicians acknowledged this and commented they do not consider the bedside the appropriate setting for rationing decisions. The wish for such decisions to be made away from the bedside is not new. For example, 70% of randomly selected US neurologists agreed that if rationing decisions have to be made, they should be made away from the bedside and speci cally not while practitioners are caring for patients.(31) Our ndings can therefore be taken as an argument in favour of a more explicit alternative to bedside rationing. One such alternative is administrative gatekeeping: a concept which is an intermediate between unrestricted advocacy of patients and bedside rationing. (12) Physicians are still to be involved in rationing, but at higher organisational levels. It allocates the con icting duties to different locations: unrestricted patient advocacy at the bedside and rationing scarce resources when constructing (clinical) guidelines and policy. The latter does entail that physicians at times will have to provide less than optimal treatment to individual patients when this results from the agreed upon policy. In a study of randomly selected GPs 69% agreed that physicians should adhere to guidelines that discourage the use of expensive interventions with a small advantage over (cheaper) standard interventions. (16) Although the potential for unfair inequality and illegitimate distribution of resources as a result of hidden bedside rationing warrants a more explicit way of rationing, some degree of bedside rationing has been deemed unavoidable. (10,12,32) Priority setting is and will remain an important part of healthcare policy. Additionally, a more explicit way of rationing has its own drawbacks. Decisions at higher organizational levels are imprecise and aimed at large groups of patients rather than tailored to individuals. (33) Moreover, it may add to an already much debated-upon administrative burden for physicians, both in the Netherlands and abroad. (34,35) One aim of a more explicit system would be to improve transparency in decision-making in order to increase the willingness of both the public and other stakeholders to accept rationing decisions. (36,37) Yet, studies have shown it is the perception of a transparent decision-making process rather than actual transparency that increases acceptance of decisions. (36,37) Research into more explicit alternatives therefore needs to address its potential impact on bureaucracy, inappropriate care and stakeholder perception as these could be barriers to implementation.
Another notable nding is that three quarter of participants rarely discuss treatment cost with patients in their everyday practice, of which around a third even reports never to do so. However, when asked for their opinion, nearly three quarters believes a physician ought to explain this decision and its (cost) reasons.
The latter is in line with international evidence. (33,38,39) For example, 62% of randomly selected US physicians agreed it is important to discuss treatment cost with patients. (39) The discrepancy between everyday practice and physicians' views might be explained in part by the fact that only four participants (2%) feel su ciently informed about treatment cost. These ndings tie into research where 59% of physicians indicated that better knowledge of costs would change their ordering of investigations. (40) Physicians seem to make rationing decisions at the bedside while insu ciently informed about treatment cost, which in turn poses questions as to how well these decisions are being made. (33) Other contributing factors to not discussing treatment cost are likely health system related. In the Netherlands, healthcare included in the statutory bene t package poses no direct nancial risk to patients other than the maximised deductible and some speci c out-of-pocket payments. Therefore, physicians in the Netherlands might not prioritise informing patients about treatment cost which might be different in other healthcare systems. (18,41) As previous research also emphasised, there is a need for a more extensive public debate on hidden bedside rationing as well as on its possible (more explicit) alternatives. (11,42,43) This debate should include patients and the general public (i.e. future patients), as well as physicians and policymakers. Recently, a citizen panel conducted in the Netherlands systematically asked citizens from various backgrounds about the importance of different criteria to set limits. (44,45) Building on this and other international studies, future research might focus on rational democratic deliberation as legitimisation of overarching rationing policies. (42,46,47) Limitations As with previous studies regarding this topic, physicians might have been hesitant to report engaging in bedside rationing due to the sensitivity of the topic. Moreover, self-reporting is limited only to bedside rationing that physicians are aware of. Therefore, our survey may underestimate the occurrence of bedside rationing.
The response rate, although in line with previous surveys among medical personnel, is low. (48) Respondents not-missing-at-random may have led to both an over-and an underestimation. It is possible that physicians who are more cost-conscious are more willing to participate and vice versa. Due to the small sample, survey weights were not used and could therefore not account for the non-response. However, participant characteristics are distributed evenly across type of hospital, residency, gender and employment mode.
Our results are limited to one medical specialisation in eleven hospitals in the Dutch healthcare system. The choice for internal medicine is in line with previous research on bedside rationing. (15,16,49) Although there is no reason to assume this sample differs greatly from the whole Dutch population of internists, extrapolations to other medical specialisations or healthcare systems ought to be done with caution.
Finally, we cannot conclude cost-reduction is the sole reason for choice limitation due to the construction of the survey. In the questions pertaining to prevalence of bedside rationing the reasoning for choice limitation is not phrased explicitly. We can therefore only establish an association rather than causality. Moreover, no degree of 'more effective' or 'more expensive' is given. This is notably di cult yet important: very expensive treatment which is only slightly more effective could be easier to forego than when it would be tremendously more effective.

Conclusion
Our study indicates that hidden bedside rationing occurs in the Netherlands. Residency and working at an academic hospital are associated with more bedside rationing. Residents are more likely to disclose bedside rationing decisions to their patients than internists while salaried physicians are less likely to do so than physicians who are in private practice. The question of how distribution of healthcare ought to take place and to what degree that should include bedside rationing remains topical and warrants a more public debate. Embedding informed public and expert opinion into policies at higher organisational levels can decrease the need for bedside rationing.

Declarations
Ethics approval and consent to participate Because our study concerned a single survey without excessive time-consumption the need for approval of our study was waived prior to data collection under the Dutch Medical Research Act by the Institutional Review Board.(50) All participants provided informed consent.

Consent for publication
Not applicable.

Availability of data and materials
The datasets generated and/or analysed during the current study including all individual deidenti ed participant data are available from the corresponding author upon reasonable request. An English  Histograms of Survey Outcome per Item Items A-D: Questions regarding everyday practice on a ve-point scale ranging from never to always with a sixth option to give an alternative answer. Item D was answered by a subset of participants (n=136) who gave a response to item C other than 'Never'. Items E-H: Statements regarding physician opinion on a seven-point Likert scale ranging from completely disagree (1) to completely agree (7). * The Y-axis depicts number of participants