Incentives for Danish healthcare management based on a pilot outcome-based, patient-centric management model in psoriasis and psoriatic arthritis: the non-interventional IMPROVE study
Background Psoriasis (PsO) and psoriatic arthritis (PsA) are chronic diseases that affect patients’ quality of life. The purpose of the present study was to develop a pilot outcome-based, patient-centric management model for PsO and PsA.
Methods The non-interventional IMPROVE ( I ncentives for healthcare m anagement based on p atient- r elated o utcomes and v alu e ) study being conducted in Denmark consists of 5 phases: 1) collecting real-world evidence to estimate treatment patterns and disease burden to the healthcare sector and patients; 2) identifying disease aspects which matter most to patients by use of concept mapping; 3) conducting interviews with healthcare professionals and patient organization involved in a typical PsO or PsA patient journey in order to determine relevant measures to quantify patient-identified outcomes; 4) developing a value-based remuneration model based on outcomes from phases 1–3; and 5) testing the outcome-based model in pre-selected hospitals in Denmark.
Results Both PsO and PsA are associated with multiple co-morbidities, increased healthcare costs, and loss of earnings. Seven important ‘clusters’ of disease aspects were identified for both PsO and PsA, including uncertainty about disease progression and treatments, as well as inter-personal relations with healthcare providers. Hospital-based treatment was associated with high treatment costs. Although the outcome-based model could result in strategic behavior by doctors, those involved in defining the best outcome goals consider it unlikely.
Conclusion The new patient-centric outcome-based management model is expected to support optimal treatment and secure best possible outcomes for patients suffering from PsO or PsA. The practical implication of the present study are that the models developed are expected to increase focus on patient-centered healthcare, and help eliminate some of the inappropriate incentives that exist in activity-based remuneration systems.
Figure 1
Figure 2
Figure 3
Phase 1
In total, 13,025 patients with PsO and 10,525 patients with PsA were identified from the patient registries in Denmark. Both PsO and PsA were associated with increased healthcare costs and loss of earnings for patients suffering from the disease [30] . There was a significant increase in the mean annual treatment costs post-diagnosis of PsO and PsA, and there were inequalities in income and employment rates compared with matched controls. A number of different comorbidities were associated with both PsO and PsA [23]. CV disease and associated risk factors were more prevalent in patients with PsO and PsA than in matched controls. Patients with PsO had a particularly increased risk of mortality and death at a younger mean age compared with those with PsA[23].
Phase 2
Seven important ‘clusters’ of disease aspects were identified for PsO within the three superclusters of ‘Having psoriasis,’ ‘Treatment,’ and ‘Surroundings/treatment’ (Table 2). All clusters and sub-clusters within the supercluster of ‘Having psoriasis’ were considered important by patients (Table 2). Under the supercluster of treatment, patients considered biological therapy to be “miraculous”, and were “concerned” about what would happen if the medication stops working or if they are not given the medication anymore (Table 2).
Similarly, for PsA, seven important clusters of disease aspects were identified within the three superclusters of ‘Living with the condition,’ ‘Treatment,’ and ‘Surroundings/treatment’ (Table 3). Not having information about what they can do for themselves, pain, psychological disturbance, worries about treatment and side-effects, and the feeling that there was no understanding for/faith in them were the most important sub-clusters of disease aspects for patients with PsA (Table 3).
Phase 3
Interviews with different healthcare professionals and patient organization showed that underreporting was common, and that hospital treatment was associated with high treatment costs. Topical treatment at the level of the general practitioner was considered easy and associated with low treatment costs (Figure 1). The stakeholders’ response showed that while the short-term results of treatment depend primarily on doctor’s efforts, long-term results were dependent on those of the patient (Figure 1).
For the prediction model for treatment of PsO and PsA, the inputs were obtained from patients on background factors such as comorbidities, education level, age, place of residence, stress, smoking, and alcohol consumption; and from doctors on number of visits, number of attempted treatments; and treatments used by ATC codes (Figure 2). Treatment results were evaluated in terms of patient-reported outcomes (phase 2 results: based on importance and weightage for the different outcomes); observed conditions included QoL, income levels; and the labor market in terms of social benefit(s) and labor market activity (Figure 2).
Phase 4
The two prediction models developed (Model A and Model B) are described below.
In Model A, a prediction model was constructed using assessment (1), as shown below, in which the anticipated PASI after treatment (PSForv.) was predicted based on the individual patient (‘p’s’) PASI before the treatment (PS Før) and based on obesity (K) and smoking (R) of KRAM (diet, smoking, alcohol, exercise) factors, comorbidities (KM) and other background variables, such as age and gender (BG). The β values are based on existing literature estimates [31-33].
𝑃𝑆𝑝,𝐸𝑥𝑝. = 𝛽0 + 𝛽1𝑃𝑆𝑝,. + 𝛽2𝐾𝑅𝑝 + 𝛽3𝐾𝑀𝑝 + 𝛽4𝐵𝐺𝑝 (1)
Remuneration for the treatment of the individual patient ‘p’ was then determined based on the difference between the anticipated and the actual PASI (PSFakt.) after treatment using assessment (2) shown below:
𝐴𝑓𝑙.𝑝 = (𝑃𝑆𝑝,𝐸𝑥𝑝. − 𝑃𝑆𝑝,𝐴𝑐𝑡.) ∗ 𝛼𝑝 ∗ 𝑘 (2)
In this equation, α was a factor that meant that the remuneration was corrected to reflect the extent to which the treatment takes account of the patients' prioritization of a number of outcome goals as described in phase 3. If the treatment took into account those outcome goals that were important for the patient, the remuneration increased (α>1), and if not, the factor fell (α<1). Assessment (3) given below is an example of how α can be calculated based on two outcome goals:
[Due to technical limitations, the formula could not be displayed here. Please see the supplementary files section to access the formulas.] (3)
O was outcome goals and ω was the patient's weighting of a given outcome goal. If the patient weighted all outcome goals equally or the treatment produced the same change in all outcome goals, there was no correction in α (α=1) and the remuneration was based only on the PASI. k in assessment (2) above was a factor that determined the size of the reward for a more than expected reduction in the PASI. k was calibrated such that the maximum payment was at a level that was acceptable from a budgetary perspective. The final determination of α and k in assessment (2) were designed to ensure budgetary safety or to optimize the doctors' incentive, irrespective of which was assigned the highest priority.
Model B was also based on literature estimates [31-33]. In this model, the placing in the spread of anticipated PASI after treatment was ascertained for each individual patient. The placing for an individual patient ‘ρ’ was calculated based on the deviation from the mean anticipated PASI expressed in standard deviations (SDs as shown in assessment (4) below:
[See supp. files] (4)
where μ indicated the mean. The anticipated value of each outcome goal was calculated for the individual patient by assuming that they deviated from the mean in all patients in the same way for anticipated PASI. If the patient's anticipated PASI was one SD less than the mean, the patient's anticipated result was thus determined for all outcome goals as also being one SD less than the mean for the actual outcomes. The calculation for sub-outcome goal 1 was calculated as shown in assessment (5) below:
[See supp. files] (5)
The assumption underlying assessment (5) is illustrated in Figure 3.
The remuneration for treatment of the individual patient ‘p’ was calculated as the change in the total weighted outcome goals consisting of i sub-goals as shown in assessment (6) below, where Oi was sub-outcome goal i and wi was the individual patient's weighting of the sub-outcome goal i.
[See supp. files] (6)
k was, as in Model A, a factor that determined the size of the remuneration in order to reduce the PASI more than expected, and was calibrated such that the maximum payment was at a level that was acceptable from a budgetary perspective.
Comparison of the two prediction models
The main difference between the two models was in whether the model rewarded an improvement in the PASI or an improvement in outcome goals. Model A rewarded improvements in PASI and corrected for whether the treatment took into account the patient's wishes calculated by outcome goals. Model A was based on the doctor and patient working together to define the value of the treatment based on both clinical goals (PASI) and the patient's prioritization of outcome goals. Model B rewarded improvements in outcome goals and used only the PASI to allow for the importance of KRAM factors, comorbidities, and background variables for ease of creating improvements in the outcome goal for the individual patient. Hence, Model B was based more on patients’ prioritization of outcome goals; relevant clinical goals were those which the patient found important in relation to living the life they want.
Once an actual and an anticipated outcome of the treatment were calculated for each sub-outcome goal, as described above, the total outcome of the treatment was calculated by weighting the individual sub-outcome goals with the weighting attributed to them by the individual patient, and then scaling them using a factor, k, as described in equation (6) above. Establishing the factor k depended on which of the desired budgetary and incentive aspects were taken into account: 1) Compliance with an overall financial limit, and a limit to the reward, regardless of how large an improvement was created for the patient; 2) An incentive to achieve the best possible result for the individual patient, irrespective of the results for other patients; 3) Lack of possibilities to act strategically and exploit the remuneration system.
Because it may be difficult to satisfy all of these aspects at the same time, and in some cases, they may be mutually contradictory, the remuneration model ensured aspects 1 and 2 were met. With this model, although there is a theoretical risk of doctors acting strategically, it was believed that doctors’ focus on consideration for their patients will inhibit the risk of such behavior. This remuneration model proposes that the involved hospital department will receive the same amount of funds allotted under the de facto goal and limit control, but without the requirement for the now discontinued annual 2% improvement in productivity. An alternative to the 2% increase will, instead, be provided based on the patients' outcome goals. This 2% of the hospital department's budget then constitutes the financial limit which the value-based remuneration must adhere to. However, consideration for remuneration will only be given to those outcomes beyond the anticipated outcome, which are forecast in assessment (1) for Model A and assessment (5) for Model B.
An example of the application of the remuneration model
For a given hospital department, if the overall improvement (which can be either PASI or outcome goals) for all patients was 10 units more than expected, and the result for the patients collectively was 3 units more than expected, then the department will receive 30 per cent of the total limit as a reward for value. For Model B, factor k is calculated as in assessment (6), which determines remuneration of the individual hospital department as in assessment (8), where X is the total budget for value-based remuneration, n is all patients, and I is all sub-outcome goals.
[See supp. files] (8)
This approach complies with aspect 1, as the amount paid out can never exceed the total limit X. At the same time, substantial consideration is given to aspect 2, as the individual hospital department will usually be able to achieve a greater proportion of that financial limit by creating an even better outcome for the patients. The only scenario in which this would not be the case is where a hospital department is the only one to create a better than expected outcome for the patients. In this case, this department would receive the full financial limit, regardless of how much better than expected the created outcome is.
In theory, this proposed approach does not conform to aspect 3 as described previously. This is because the doctors could, in principle, not agree to organize their treatment in order to create the greatest possible value for the patient. As long as the doctors create the same low value, they will achieve the same remuneration as they would achieve if they all created high value for their patients. This is the downside of making remuneration for the treatment of individual patients dependent on the total value created for all patients. It is, therefore, a 'side effect' of ensuring compliance with the overall budget. However, we do not expect this to be a major problem in practice, as strategic behavior by doctors is possible, albeit unlikely. Strategic behavior amongst doctors can be a serious concern only in cases doctors consider PASI or the defined outcome goals in Models A and B as irrelevant treatment goals. The doctors being involved in the work of defining the best outcome goals should counteract this.
This is a list of supplementary files associated with this preprint. Click to download.
Posted 06 Jan, 2020
Incentives for Danish healthcare management based on a pilot outcome-based, patient-centric management model in psoriasis and psoriatic arthritis: the non-interventional IMPROVE study
Posted 06 Jan, 2020
Background Psoriasis (PsO) and psoriatic arthritis (PsA) are chronic diseases that affect patients’ quality of life. The purpose of the present study was to develop a pilot outcome-based, patient-centric management model for PsO and PsA.
Methods The non-interventional IMPROVE ( I ncentives for healthcare m anagement based on p atient- r elated o utcomes and v alu e ) study being conducted in Denmark consists of 5 phases: 1) collecting real-world evidence to estimate treatment patterns and disease burden to the healthcare sector and patients; 2) identifying disease aspects which matter most to patients by use of concept mapping; 3) conducting interviews with healthcare professionals and patient organization involved in a typical PsO or PsA patient journey in order to determine relevant measures to quantify patient-identified outcomes; 4) developing a value-based remuneration model based on outcomes from phases 1–3; and 5) testing the outcome-based model in pre-selected hospitals in Denmark.
Results Both PsO and PsA are associated with multiple co-morbidities, increased healthcare costs, and loss of earnings. Seven important ‘clusters’ of disease aspects were identified for both PsO and PsA, including uncertainty about disease progression and treatments, as well as inter-personal relations with healthcare providers. Hospital-based treatment was associated with high treatment costs. Although the outcome-based model could result in strategic behavior by doctors, those involved in defining the best outcome goals consider it unlikely.
Conclusion The new patient-centric outcome-based management model is expected to support optimal treatment and secure best possible outcomes for patients suffering from PsO or PsA. The practical implication of the present study are that the models developed are expected to increase focus on patient-centered healthcare, and help eliminate some of the inappropriate incentives that exist in activity-based remuneration systems.
Figure 1
Figure 2
Figure 3
Phase 1
In total, 13,025 patients with PsO and 10,525 patients with PsA were identified from the patient registries in Denmark. Both PsO and PsA were associated with increased healthcare costs and loss of earnings for patients suffering from the disease [30] . There was a significant increase in the mean annual treatment costs post-diagnosis of PsO and PsA, and there were inequalities in income and employment rates compared with matched controls. A number of different comorbidities were associated with both PsO and PsA [23]. CV disease and associated risk factors were more prevalent in patients with PsO and PsA than in matched controls. Patients with PsO had a particularly increased risk of mortality and death at a younger mean age compared with those with PsA[23].
Phase 2
Seven important ‘clusters’ of disease aspects were identified for PsO within the three superclusters of ‘Having psoriasis,’ ‘Treatment,’ and ‘Surroundings/treatment’ (Table 2). All clusters and sub-clusters within the supercluster of ‘Having psoriasis’ were considered important by patients (Table 2). Under the supercluster of treatment, patients considered biological therapy to be “miraculous”, and were “concerned” about what would happen if the medication stops working or if they are not given the medication anymore (Table 2).
Similarly, for PsA, seven important clusters of disease aspects were identified within the three superclusters of ‘Living with the condition,’ ‘Treatment,’ and ‘Surroundings/treatment’ (Table 3). Not having information about what they can do for themselves, pain, psychological disturbance, worries about treatment and side-effects, and the feeling that there was no understanding for/faith in them were the most important sub-clusters of disease aspects for patients with PsA (Table 3).
Phase 3
Interviews with different healthcare professionals and patient organization showed that underreporting was common, and that hospital treatment was associated with high treatment costs. Topical treatment at the level of the general practitioner was considered easy and associated with low treatment costs (Figure 1). The stakeholders’ response showed that while the short-term results of treatment depend primarily on doctor’s efforts, long-term results were dependent on those of the patient (Figure 1).
For the prediction model for treatment of PsO and PsA, the inputs were obtained from patients on background factors such as comorbidities, education level, age, place of residence, stress, smoking, and alcohol consumption; and from doctors on number of visits, number of attempted treatments; and treatments used by ATC codes (Figure 2). Treatment results were evaluated in terms of patient-reported outcomes (phase 2 results: based on importance and weightage for the different outcomes); observed conditions included QoL, income levels; and the labor market in terms of social benefit(s) and labor market activity (Figure 2).
Phase 4
The two prediction models developed (Model A and Model B) are described below.
In Model A, a prediction model was constructed using assessment (1), as shown below, in which the anticipated PASI after treatment (PSForv.) was predicted based on the individual patient (‘p’s’) PASI before the treatment (PS Før) and based on obesity (K) and smoking (R) of KRAM (diet, smoking, alcohol, exercise) factors, comorbidities (KM) and other background variables, such as age and gender (BG). The β values are based on existing literature estimates [31-33].
𝑃𝑆𝑝,𝐸𝑥𝑝. = 𝛽0 + 𝛽1𝑃𝑆𝑝,. + 𝛽2𝐾𝑅𝑝 + 𝛽3𝐾𝑀𝑝 + 𝛽4𝐵𝐺𝑝 (1)
Remuneration for the treatment of the individual patient ‘p’ was then determined based on the difference between the anticipated and the actual PASI (PSFakt.) after treatment using assessment (2) shown below:
𝐴𝑓𝑙.𝑝 = (𝑃𝑆𝑝,𝐸𝑥𝑝. − 𝑃𝑆𝑝,𝐴𝑐𝑡.) ∗ 𝛼𝑝 ∗ 𝑘 (2)
In this equation, α was a factor that meant that the remuneration was corrected to reflect the extent to which the treatment takes account of the patients' prioritization of a number of outcome goals as described in phase 3. If the treatment took into account those outcome goals that were important for the patient, the remuneration increased (α>1), and if not, the factor fell (α<1). Assessment (3) given below is an example of how α can be calculated based on two outcome goals:
[Due to technical limitations, the formula could not be displayed here. Please see the supplementary files section to access the formulas.] (3)
O was outcome goals and ω was the patient's weighting of a given outcome goal. If the patient weighted all outcome goals equally or the treatment produced the same change in all outcome goals, there was no correction in α (α=1) and the remuneration was based only on the PASI. k in assessment (2) above was a factor that determined the size of the reward for a more than expected reduction in the PASI. k was calibrated such that the maximum payment was at a level that was acceptable from a budgetary perspective. The final determination of α and k in assessment (2) were designed to ensure budgetary safety or to optimize the doctors' incentive, irrespective of which was assigned the highest priority.
Model B was also based on literature estimates [31-33]. In this model, the placing in the spread of anticipated PASI after treatment was ascertained for each individual patient. The placing for an individual patient ‘ρ’ was calculated based on the deviation from the mean anticipated PASI expressed in standard deviations (SDs as shown in assessment (4) below:
[See supp. files] (4)
where μ indicated the mean. The anticipated value of each outcome goal was calculated for the individual patient by assuming that they deviated from the mean in all patients in the same way for anticipated PASI. If the patient's anticipated PASI was one SD less than the mean, the patient's anticipated result was thus determined for all outcome goals as also being one SD less than the mean for the actual outcomes. The calculation for sub-outcome goal 1 was calculated as shown in assessment (5) below:
[See supp. files] (5)
The assumption underlying assessment (5) is illustrated in Figure 3.
The remuneration for treatment of the individual patient ‘p’ was calculated as the change in the total weighted outcome goals consisting of i sub-goals as shown in assessment (6) below, where Oi was sub-outcome goal i and wi was the individual patient's weighting of the sub-outcome goal i.
[See supp. files] (6)
k was, as in Model A, a factor that determined the size of the remuneration in order to reduce the PASI more than expected, and was calibrated such that the maximum payment was at a level that was acceptable from a budgetary perspective.
Comparison of the two prediction models
The main difference between the two models was in whether the model rewarded an improvement in the PASI or an improvement in outcome goals. Model A rewarded improvements in PASI and corrected for whether the treatment took into account the patient's wishes calculated by outcome goals. Model A was based on the doctor and patient working together to define the value of the treatment based on both clinical goals (PASI) and the patient's prioritization of outcome goals. Model B rewarded improvements in outcome goals and used only the PASI to allow for the importance of KRAM factors, comorbidities, and background variables for ease of creating improvements in the outcome goal for the individual patient. Hence, Model B was based more on patients’ prioritization of outcome goals; relevant clinical goals were those which the patient found important in relation to living the life they want.
Once an actual and an anticipated outcome of the treatment were calculated for each sub-outcome goal, as described above, the total outcome of the treatment was calculated by weighting the individual sub-outcome goals with the weighting attributed to them by the individual patient, and then scaling them using a factor, k, as described in equation (6) above. Establishing the factor k depended on which of the desired budgetary and incentive aspects were taken into account: 1) Compliance with an overall financial limit, and a limit to the reward, regardless of how large an improvement was created for the patient; 2) An incentive to achieve the best possible result for the individual patient, irrespective of the results for other patients; 3) Lack of possibilities to act strategically and exploit the remuneration system.
Because it may be difficult to satisfy all of these aspects at the same time, and in some cases, they may be mutually contradictory, the remuneration model ensured aspects 1 and 2 were met. With this model, although there is a theoretical risk of doctors acting strategically, it was believed that doctors’ focus on consideration for their patients will inhibit the risk of such behavior. This remuneration model proposes that the involved hospital department will receive the same amount of funds allotted under the de facto goal and limit control, but without the requirement for the now discontinued annual 2% improvement in productivity. An alternative to the 2% increase will, instead, be provided based on the patients' outcome goals. This 2% of the hospital department's budget then constitutes the financial limit which the value-based remuneration must adhere to. However, consideration for remuneration will only be given to those outcomes beyond the anticipated outcome, which are forecast in assessment (1) for Model A and assessment (5) for Model B.
An example of the application of the remuneration model
For a given hospital department, if the overall improvement (which can be either PASI or outcome goals) for all patients was 10 units more than expected, and the result for the patients collectively was 3 units more than expected, then the department will receive 30 per cent of the total limit as a reward for value. For Model B, factor k is calculated as in assessment (6), which determines remuneration of the individual hospital department as in assessment (8), where X is the total budget for value-based remuneration, n is all patients, and I is all sub-outcome goals.
[See supp. files] (8)
This approach complies with aspect 1, as the amount paid out can never exceed the total limit X. At the same time, substantial consideration is given to aspect 2, as the individual hospital department will usually be able to achieve a greater proportion of that financial limit by creating an even better outcome for the patients. The only scenario in which this would not be the case is where a hospital department is the only one to create a better than expected outcome for the patients. In this case, this department would receive the full financial limit, regardless of how much better than expected the created outcome is.
In theory, this proposed approach does not conform to aspect 3 as described previously. This is because the doctors could, in principle, not agree to organize their treatment in order to create the greatest possible value for the patient. As long as the doctors create the same low value, they will achieve the same remuneration as they would achieve if they all created high value for their patients. This is the downside of making remuneration for the treatment of individual patients dependent on the total value created for all patients. It is, therefore, a 'side effect' of ensuring compliance with the overall budget. However, we do not expect this to be a major problem in practice, as strategic behavior by doctors is possible, albeit unlikely. Strategic behavior amongst doctors can be a serious concern only in cases doctors consider PASI or the defined outcome goals in Models A and B as irrelevant treatment goals. The doctors being involved in the work of defining the best outcome goals should counteract this.