Context
The Netherlands Heart Registration (NHR) covers over 80,000 cardiac procedures per year in the Netherlands, including all common cardiac (surgical) procedures 19. Within the NHR, a voluntary VBHC program is organized. Patient-relevant outcomes corrected for case-mix are presented annually in a publicly accessible report. In physician-directed registration committees, physicians discuss the variation of outcomes and share good practices in a noncompetitive and safe setting. To date, all of 29 hospitals eligible to participate, take part in this VBHC program. In addition to transparent reporting and discussion of outcomes, innovative projects are also conducted as part of the VBHC program.
With the aim of rewarding improvement of quality of care more than the current Dutch payment model does (fee-for-service), 10 hospitals, healthcare insurance company Menzis and the NHR designed, implemented, and evaluated an outcome-based payment model for patients suffering CAD treated with a percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG). Additionally, the new model sought to improve the synergy between hospital management and physicians. Earlier research by Van Veghel et al found that not having good synergy is one of the main experienced barriers to improving cardiac care 20.
The VBP model
An outcome-based payment model was established with patient-relevant outcomes at the core. The model was designed for CAD patients treated with PCI or CABG. These treatments were selected because they are performed frequently and because the available data regarding patients’ characteristics and outcome measures as registered within the NHR are complete and of high quality 19. These outcome measures are aligned with the International Consortium for Health Outcomes Measurement (ICHOM) sets 21,22. The involved patient-relevant clinical outcomes were selected by expert groups following a fixed stepwise approach 19,23 and are presented in Table 1. The set of selected outcome measures was generic for all hospitals, with room for the hospitals and health insurers to jointly make additional specific agreements, for example regarding outcomes that are not part of the generic VBP model or specific incentives for the improvement of a specific outcome.
Table 1. Outcome measures included in the payment model for CABG and PCI, following Porter’s outcome hierarchy 24
Outcome hierarchy
|
CABG
|
PCI
|
Tier 1
|
Survival
|
120-day mortality
|
30-day mortality
|
1-year mortality
|
1-year mortality
|
Degree of health/recovery
|
Quality of life1
Physical health
Mental health
|
Quality of life1
Physical health
Mental health
|
Tier 2
|
Time to recovery and return to normal activities
|
|
|
Disutility of the care or treatment process
|
Cerebrovascular accident (CVA) (≤ 72 hours)
|
Urgent CABG (≤ 24 hours)
|
Deep sternal wound infection (≤ 30 days)
|
|
Surgical reexploration (≤ 30 days)
|
|
Tier 3
|
Sustainability of health/recovery and nature of recurrences
|
Myocardial infarction (≤ 1 year) 1
|
Myocardial infarction (≤ 30 days)
|
Coronary artery reintervention (≤ 1 year)
|
Occurrence of target vessel revascularization (≤ 1 year)
|
Long-term consequences of therapy
|
|
|
Abbreviations: CABG = coronary artery bypass grafting; PCI = percutaneous coronary intervention
1. No calculations were performed during the pilot for quality of life outcomes and myocardial infarction ≤ 1 year because of insufficient data.
For each hospital, outcomes were predicted based on the patients’ case-mix. Prediction was performed by means of a national prediction model based on all 29 centers, i.e., a mathematical formula in which all relevant patient characteristics were weighted. Weights were based on logistic regression analyses of a historical nationwide cohort of 5 subsequent years. The predicted outcomes were compared to the real outcomes as registered within the NHR. In case of significantly improved performance than predicted (p < 0.05), the hospital received a financial reward from the insurance company; in case of performance decline, financial penalties were given. Cost prices per outcome were determined based on the current payment system, Porter’s outcome hierarchy 3,24, and expert opinion. To create a noncompetitive culture in which the model could be safely tested, each hospital’s maximum financial impact was limited. During the pilot phase (2018-2020) the outcome-based model was applied annually; i.e. per calendar year, predicted outcomes were compared to real outcomes and financial consequences were calculated. Experiences with the application formed the basis for further fine-tuning of the VBP. Sharing a good practice about processes of care within the community was a prerequisite for receiving financial rewards.
The outcome-based payment model was designed and established by representatives from 10 Dutch hospitals (i.e. cardiologists, cardiothoracic surgeons, managers), and representatives of healthcare insurance company Menzis (i.e., healthcare purchaser, medical advisor and innovation manager) and of the NHR (i.e., expert in VBHC, data analyst and project manager).
Study design and population
To measure intended and unintended effects of the introduction of the VBP model, a mixed-method approach was applied. Data were collected by means of online questionnaires and semi-structured interviews with representatives of the participating hospitals. In addition, case-mix analyses were performed in patients treated with PCI or CABG between January 2016 and December 2020.
Questionnaire
A 17-item online questionnaire in Dutch was developed, containing questions regarding the perceptions of the impact of the model on treatment decisions, the organization of care, attention to clinical outcomes within the hospital, and physicians’ perceptions of the model (translated questionnaire in Additional file 1). Questions were rated on a 5-point Likert scale, ranging from strongly agree to strongly disagree. The questionnaire was sent at three different time points to 170 medical specialists (cardiologists and cardiothoracic surgeons) of eight hospitals: shortly before the introduction of the payment model to perform a baseline measurement, and one and two years afterwards to measure potential changes in perceptions over time. The questions were inspired by the questionnaire used by Fernandez et al., which aimed to measure how modifications in publicly reported outcome data have influenced the practice of medical specialists and to measure physicians’ attitudes toward public reporting 25. Questions were modified for the goal of the present study by two experts in health economics.
Interviews
Semi-structured group interviews were conducted between October and December 2020. With each participating hospital, a closing group meeting was organized to discuss the final version of the developed outcome-based payment model, and to further clarify the results of the questionnaire. All interviews were framed by a topic list, which was based on the 17-item questionnaire, containing questions about the impact of the model on attention to clinical outcomes and quality improvement within the hospital, and the impact on physicians’ behavior, such as risk-averse behavior. For each group meeting, a cardiologist, cardiothoracic surgeon, and a manager were invited. The group interviews were online, audiotaped and lasted approximately 45 minutes. For each interview, detailed notes were taken by the researcher.
Case-mix analyses
To identify potential unintended consequences of the implementation of the VBP model regarding risk-averse behavior, case-mix analyses were performed with data collected within the NHR. In the Netherlands, all hospitals are obliged to register baseline, procedural and outcome data regarding all performed PCIs and CABGs, and therefore, nationwide coverage of procedures is ensured.
Yearly analyses were performed on the current trend in the risk profile of patients treated with PCI or CABG, starting with the year 2016 (before the introduction of the VBP model) until 2020. Trends in the risk profile were monitored for several patient characteristics, which were selected by cardiologists and cardiothoracic surgeons based on the following criteria: 1) impact on the outcomes that are central in the VBP model; 2) completeness of available data; and 3) quality of available data. Examples of selected patient characteristics are renal insufficiency, cardiogenic shock and left ventricular ejection fraction. In addition, trends in risk scores, in which different patient characteristics were combined and weighted (i.e., EuroSCORE I and the NHR risk score), were monitored. The EuroSCORE is the current gold standard to predict mortality after cardiac surgery 26. The NHR risk score is based on all patient characteristics and their corresponding predictive value, as included in the logistic regression models frequently applied within the VBHC program of the NHR. These patient characteristics were previously selected by an expert panel following a fixed stepwise approach 19. All individual patient characteristics and the combined risk scores are presented in Table 2.
Table 2. Set of patient characteristics and weighted risk scores as selected for case-mix analyses
CABG
|
PCI
|
Individual patient characteristics
|
Age (mean)
|
Age (mean)
|
Multivessel disease (yes/no)
|
Multivessel disease (yes/no)
|
Renal insufficiency (eGFR<60 ml/min)
|
Renal insufficiency (eGFR<60 ml/min)
|
Previous cardiac surgery (yes/no)
|
Previous CABG (yes/no)
|
Urgency of the procedure (emergency+salvage)
|
Indication PCI (STEMI)
|
Critical preoperative status (yes/no)
|
Cardiogenic shock (yes/no)
|
Left ventricular ejection fraction (<30%)
|
Out of hospital cardiac arrest (yes/no)
|
|
|
Weighted risk scores
|
|
EuroSCORE I
|
|
NHR risk score1
|
NHR risk score2
|
Abbreviations: CABG=coronary artery bypass grafting; PCI=percutaneous coronary intervention; eGRF=estimated glomerular filtration rate; STEMI=ST elevation myocardial infarction; NHR=Netherlands Heart Registration
- The NHR risk score for CABG is based on the predictive values of the following patient characteristics: age, sex, chronic lung disease, diabetes, previous cardiac surgery, renal insufficiency, left ventricular ejection fraction and urgency of the procedure.
- The NHR risk score for PCI is based on the predictive values of the following patient characteristics: age, sex, diabetes, renal insufficiency, indication for PCI (STEMI/non-STEMI/elective), cardiogenic shock, out-of-hospital cardiac arrest, multivessel disease, chronic total occlusion, previous CABG, and previous myocardial infarction.
Starting in 2016, the risk profile of all patients who were treated with a PCI or CABG in the 10 participating hospitals was monitored yearly. For the dichotomous individual patient characteristics, percentages with corresponding 95% confidence intervals (CIs) were followed over time. For the continuous patient characteristics with normal distribution, the mean with corresponding 95% CI were followed. In the case of a nonnormal distribution, median values with corresponding interquartile ranges were monitored. The NHR risk scores were analyzed in two different ways: the trend over time regarding the mean weighted risk score (0-1) with corresponding 95% CI was monitored, and in addition the highest decile of the risk score was monitored separately. Moreover, for all variables as presented in Table 2, trends were monitored in patients treated within the hospitals that did not implement the VBP model, serving as a control group. The results of all analyses were discussed yearly by the cardiologists and cardiothoracic surgeons who participated in the pilot.