Introduction

Healthcare is increasingly expensive. Switzerland has one of the highest health care expenditures per capita in the world and devotes about 11 % of the gross national product (GNP) to health care(1). As a large percentage of health care expenditures are the results of doctors‘ decisions, a majority of Swiss doctors in a cross-sectional study agreed that trying to contain costs was their responsibility and that they should worry about the costs of tests and procedures they order (2).

Decisions and choices in health care inescapably involve value judgments. Intrinsic to any value judgment is the consideration of what you get for your money and hence all costs are weighed against the expected benefits and related to alternative, less or more expensive choices (3). Unfortunately, physicians rarely know the charges of the services, tests, and procedures they order or perform (4, 5). Enhancing cost knowledge and cost awareness is potentially cost-saving: A recent systematic review of charge transparency interventions (6) found, that having real-time access to charges changed ordering and prescribing behavior in the majority of studies. Of the clinically based interventions looking at laboratory and radiology ordering, seven of the nine studies reported statistically significant cost reduction when charges were displayed.
The aim of the present study was to describe the level of cost knowledge of cardiological tests and procedures among medical students, residents and doctors in Switzerland and discuss trends in cost perception in health expenditures.

Methods

Study Setting

Health insurance is compulsory in Switzerland (under constitutional law since 1994). Around 60 authorized non-profit insurers offer compulsory health insurance (basic insurance) and optional daily allowance insurance. Compulsory health insurance provides cover for illness, maternity and accidents and offers the same range of services and benefits to all insured people. Compulsory health insurance is financed by policyholders’ contributions (premiums) and co-payments (deductible, retention fee, contribution to the costs of a hospital stay) and federal and cantonal funding (premium subsidies).

Ambulatory care is provided by doctors in private practice or institutional (e.g. hospital-based outpatient clinic, community practice/ambulatory health care center). The service providers are reimbursed mostly “fee-for-service” (in contrast to case-based payments in inpatient setting) based on a nation-wide uniform tariff catalogue (in the present study period: TARMED 01.09.00_BR_KVG).

Data set

We compiled a set of 13 common cardiological services, tests and procedures and calculated the costs applying a tax point value of 0.89 (cantonal reference Zurich) to the corresponding codes of the TARMED-tariff catalogue. Costs of implanted materials were excluded, since they are very variable and not part of a fixed tariff code. Detailed codes are listed in Appendix 1.
The study was reviewed and approved by the ethics committee of the Canton of Zurich, Switzerland (BASEC-Nr. Req-20147-00296).

Data acquisition

Participants were randomly recruited by mailing lists, messaging app or via direct contact. A mailing list has been used in particular regarding the physicians group. A systematic email has been sent by the national professional association of cardiologists (Swiss Society of Cardiology, SSC) to all members (including cardiologists from different region and with different cardiologic background/work field in Switzerland).
Also, known general practitioners and internists working in different regional hospitals in Switzerland/Zurich were encouraged either via email or via direct contact to send the link containing the questionnaire to their fellow physicians.

Medical students were recruited via WhatsApp Messenger App, containing a link to get to the questionnaire. As there exist group chats including almost all of the students of equal academic year and in the same region, sending there has been considered the most reasonable option for recruiting. Most students were recruited from the German-speaking part of Switzerland, including Zurich, Basel and Bern.
Primary data relevant to the subsequent statistical analysis was acquired by means of an online questionnaire tool. Participation was voluntary. In order to enhance the response rate, a lottery drawing for participants was initiated.

In the questionnaire, personal data for each participant was recorded (age, gender, professional setting). Subsequently, participants had to estimate the costs of the 13 predefined services, tests and procedures in Swiss Francs (CHF). The three general services included consultation fees (15 minutes) for a cardiologist and a family doctor, respectively; and a medical report of 1 page. The seven diagnostic tests listed were an electrocardiogram (ECG), a bicycle stress test (exercise test), a 24 hours Holter ECG monitoring, three different types of heart ultrasound tests (transthoracic echocardiography (TTE), transesophageal echocardiography (TEE) as well as pharmacological stress echo) and a pacemaker (PM) control of a 2 lead device. The three interventions consisted of a dual chamber PM implantation, a coronary angiogram as well as a simple percutaneous coronary intervention (PCI) using 1 stent.

Short technical descriptions of the procedures and tests were provided. Accurate cost perception was defined as an estimate within a ±25 %-range of the effective reimbursement amount.

Statistics / Data interpretation

Continuous data are expressed as medians and interquartile ranges (IQR) or as mean ± standard deviation (SD) as appropriate, and categorical data as number and percentage (%).
A p value of < 0.05 was considered statistically significant. Statistical analyses were performed on R Studio (V 1.1.463).
To interpret the data, we used a behavioral economics approach proposed by nobel prize winners Kahnemann and Smith on human judgment and decision-making: Humans are unreliable decision makers, their judgments are strongly influenced by several factors. “Noise” describes the chance variability of judgments, whereas “bias” states if estimates are generally either too high or too low (7).
The accuracy of an individual’s (i) estimation was represented by a percentage over or under estimation of every position (c).

An individual’s bias is given by the average of all accuracies across the 13 positions.

An individual’s noise is represented by the standard deviation across the 13 positions.

Two regression models (M1 & M2) were used to test the impact of independent variables on the dependent variables as follows (students aren´t included, as they´re mutually exclusive with the other independent variables):

M1: Bias β0+ β1 × Noise + β2 × Age + β3 × Gender + β4 × Practitioner + β5 × Resident + β6 × Hospital.Physician + ε

M2: Noise = β0 + β1 × Age + β2 × Gender + β3 × Practitioner + β4 × Resident + β5 × Hospital.Physician + ε

Results

Study population

A total of 939 participants, who completed personal data entry and estimated at least one test or procedure, were enrolled (172 physicians and 767 medical students).
Medical students had a mean age of 22.7 years (SD ±2.4 years) and 70 % were women. They were recruited from universities all over Switzerland. All academic years were represented. (see Table 1), with a medical degree obtainable after six years of study. (Table 1)
Physicians had a mean age of 43.3 years (SD ±10.3 years) and 31 % were women. They were grouped in residents, hospital-based physicians (of different hierarchic levels / functions) and practitioners (physicians in private practice). (see Table 2). The majority of included physicians were specialized in internal medicine and/or cardiology.

Accuracy of cost estimation

Accurate cost perception was defined as an estimate within a ±25 %-range of the effective reimbursement amount. Figure 1 shows the percentage of accurate cost estimates ordered by subgroups (students, residents, hospital-based physicians and practitioners) and gestures.

Furthermore, we calculated the overall proportion of medical gestures estimated within ±25 % of the reimbursement rate (see Figure 2) and found substantial differences between the subgroups: Whereas in the student group, only 19.3 % (SD ±8.7 %) of estimates were within the defined range, practitioners indicated the costs accurately in 55.4 % (SD ±23.5 %) overall (ranging from 14.7 % in pacemaker-implantation to 82.3 % in stressechocardiography). Residents (26.2 %, SD ±9.2 %) and hospital-based physicians (38.0 %, SD ±14.4 %) performed intermediately.

Invasive (and costly) procedures (PCI, PM-implantation, coronary angiogram) seem to be most difficult to estimate for all subgroups (e.g. for PM-implantation: 17.0 % correct estimates, SD ±5.9 %) (see Figure 3) by procedures.

Trends in cost perception

To assess over- or underestimations, we calculated the mean differences between effective reimbursement and estimated costs per procedure and groups (see Table 3). Table 4 shows the accuracy of an individual’s estimation, represented by a percentage over or under estimation of every position.

The previously described lack of accuracy is well reflected by the considerable high standard deviations (noise): They were highest in students, intermediate in residents/hospital-based physicians and lowest in practitioners. Indeed, in the regression model (M2) the only statistically significant predictor of variability (noise) was age (less variability with advanced age, p<0.05). Interestingly, variability (noise) was gender-independent (p=ns). Overall, M2 was statistically significant (p<0.001) with an R2 of 0.025. (see Appendix 2)
In general, overestimation was the most prevalent perception bias. Nevertheless, practitioners tended to underestimate several procedures (namely consultation, report, 24h-ECG, TTE and PM-implantation). In the regression model (M1), bias was mainly influenced by variability/noise (Effect size 0.61). The more uncertainty was present, the more biased were the estimates. Interestingly, bias was gender-dependent: Women tended to be more overestimating than men (effect size 0.13, p<0.05). Overall, M1 was statistically significant (p<0.001) with an R2 of 0.82 (see Appendix 3).

Discussion

Patients in the Swiss health care system incur substantial out-of-pocket costs: one third of health care spending comes from copayments and other private payments (1). Unlike countries with a long tradition of a national health service or comprehensive social insurance, Switzerland faces no historically based societal expectation that the state or taxpayers will systematically cover all health care expenses (9). In such a setting, shared decision-making in choosing diagnostic or therapeutic procedures should also elucidate economic cost considerations.

The level of cost knowledge of cardiological tests and procedures among medical students, residents and doctors in Switzerland is modest: The overall proportion of medi-
cal gestures estimated correctly within ±25 % of the reimbursement rate ranged from 10 % (students) to 55 % in practitioners. Similar previously published analyses from other medical subspecialties were comparable: Swedish emergency department physicians had a mean deviation to the real cost of 52 % with a correct estimation of an average of 28 % (10). In members of surgical teams, only 18.6 % of estimates were considered correct (11). Postgraduate physician trainees across all disciplines demonstrate limited awareness of the costs of commonly ordered imaging examinations, only 5.7 % of responses were within the correct ±25 % range (12). Another study showed that family doctors underestimated costs of expensive drugs and laboratory investigations and overestimated costs of inexpensive drugs (13).

In the present study, accuracy showed substantial differences between the subgroups and type of gesture. Lack of accuracy is reflected by the variability (statistical noise) of estimates. The only statistically significant predictor of variability was age (less variability with advanced age). Increasing (professional or general) experience seems to sharpen the accuracy of cost estimation. This effect was gender-independent.

In general, overestimation was the most prevalent cost perception bias. Interestingly, this bias was gender-dependent: Women tended to be more overestimating than men. Nevertheless, bias was mainly influenced by variability (statistical noise) – the more uncertainty was present, the more biased were the estimates. This is an interesting finding, as Kahneman suggested Bias and Noise to be independent (8). Either in a system of legally limited health expenditures (governmental defined global budget, actually discussed in Switzerland) or systems with substantial out-of-pocket costs for patients, overestimating cost perception by the health care provider is problematic: Overestimation will result in more restrictive ordering than it would be appropriate and affordable for the individual patient.

More profound and proactive education of physicians about costs, reimbursement, and charges associated with the care they deliver, would improve decision making applying proper value judgments in economic consideration of cost related to the differential benefits to be derived from alternative (less or more expensive) choices.

Limitations

Despite an overall representative population sample size, several potentially influencing factors could not be analyzed because of small subgroups (e.g. hierarchic position within the hospital, regional differences). Since the participation was voluntary, we cannot exclude a certain population selection bias. Additionally, the methodological standard of assessing accuracy within a ±25 %-range is questionable, larger absolute estimation ranges (e.g. for more expensive gestures) have been suggested (14).

Conclusion

The level of cost knowledge of cardiological tests and procedures among medical students, residents and doctors in Switzerland is modest, correctly estimated costs ranged from 10 % in students to 55 % in practitioners. In general, the costs were overestimated. Increasing experience seems to sharpen the accuracy of cost estimation. Either in systems of governmental defined global budget or systems with substantial out-of-pocket costs for patients, overestimated costs will potentially result in more restrictive ordering than it would be appropriate and affordable for the individual patient.

History
Manuscript submitted: 19.02.2024
Accepted after revision: 23.04.2024

Acknowledgments
We would like to thank Verena Reichl for assistance for data management and administration.

Author contributions
AM and CW contributed to the conception of the work and to the acquisition of data. AM, CW and RM contributed to the analysis and interpretation of data for the work and drafted the manuscript. PB and LO critically revised the manuscript. All gave final approval and agree to be accountable for all aspects of work ensuring integrity and accuracy.