Medical ethics: enhanced or undermined by modes of payment?

Abstract

Background

In the medical literature [1, 2, 7], the view prevails that any change away from fee-for-service (FFS) jeopardizes medical ethics, defined as motivational preference in this article. The objective of this contribution is to test this hypothesis by first developing two theoretical models of behavior, building on the pioneering works of Ellis and McGuire [4] and Pauly and Redisch [11]. Medical ethics is reflected by a parameter α, which indicates how much importance the physician attributes to patient well-being relative to his or her own income. Accordingly, a weakening of ethical orientation amounts to a fall in the value of α. While traditional economic theory takes preferences as predetermined, more recent contributions view them as endogenous (see, e.g., Frey and Oberholzer-Gee [5]).

Methods

The model variant based on Ellis and McGuire [4] depicts the behavior of a physician in private practice, while the one based on Pauly and Redisch [11] applies to providers who share resources such as in hospital or group practice. Two changes in the mode of payment are analyzed, one from FFS to prospective payment (PP), the other to pay-for-performance (P4P). One set of predictions relates physician effort to a change in the mode of payment; another, physician effort to a change in α, the parameter reflecting ethics. Using these two relationships, a change in ethics can observationally be related to a change in the mode of payment. The predictions derived from the models are pitted against several case studies from diverse countries.

Results

A shift from FFS to PP is predicted to give rise to a negative observed relationship between the medical ethics of physicians in private practice under a wide variety of circumstances, more so than a shift to P4P, which can even be seen as enhancing medical ethics, provided physician effort has a sufficiently high marginal effectiveness in terms of patient well-being. This prediction is confirmed to a considerable degree by circumstantial evidence coming from the case studies. As to physicians working in hospital or group practice, the prediction is again that a transition in hospital payment from FFS to PP weakens their ethical orientation. However, this prediction could not be tested because the one hospital study found relates to a transition to P4P, suggesting that this mode of payment may actually enhance medical ethics of healthcare providers working in a hospital or group practice.

Conclusion

The claim that moving away from FFS undermines medical ethics is far too sweeping. It can only in part be justified by observed relationships, which even may suggest that a transition to P4P strengthens medical ethics.

Appendix

This Appendix contains the steps leading to Eqs. (2), (3), (6), (7), (10), (11), (14), and (15) of the text.

Physician in private practice subject to a shift from FFS to PP: Differentiation of Eq. (1) yields the first-order condition for an (interior) optimum,

$$ \frac{dU}{de} = \pi \cdot \upsilon '\left[ P \right]( - c' + \alpha \cdot w') + (1 - \pi ) \cdot \upsilon '\left[ F \right](r'\; - c'\; + \alpha \cdot w') = 0 $$

(17)

For an interior solution to be obtained, marginal cost $c^{\prime}$ must evidently exceed marginal patient well-being weighted by ethics, $\alpha \cdot w'$. This renders the first term negative (mirroring the disincentive effect of fixed prospective payment), while the second is positive as long as the FFS margin is nonnegative.

Now consider an exogenous change dπ > 0, i.e., an increased share of prospective payment. Since the first derivative of the utility function must have the value zero before and after the change, the right-hand side of Eq. (19) below is zero. Its left-hand side shows the two components of the change in marginal utility induced by dπ and the adjustment of effort de,

$$ \frac{{\partial^{2} U}}{{\partial e^{2} }} \cdot de + \frac{{\partial^{2} U}}{\partial e\partial \pi } \cdot d\pi = 0 $$

(18)

This can be solved to obtain

$$ \frac{de}{d\pi } = - \frac{{\partial^{2} U/\partial e\partial \pi }}{{\partial^{2} U/\partial e^{2} }} $$

(19)

Therefore, assuming that the second-order condition for a maximum $\partial^{2} U/\partial e^{2}\, <\, 0$ is satisfied, the sign of $de/d\pi$ is determined by the sign of $\partial^{2} U/\partial e\partial \pi$ (≈symbolizes proportionality),

$$ de/d\pi \approx \partial^{2} U/\partial e\partial \pi = \upsilon '\left[ P \right]( - c'\; + \alpha \cdot w') - \upsilon '\left[ F \right](r'\; - c'\; + \alpha \cdot w') $$

(20)

Since the first-order condition is always satisfied by assumption, Eq. (20) can be simplified, using an expression derived from Eq. (17), $\upsilon '\left[ F \right](r'\; - c'\; + \alpha \cdot w') = - \pi /(1 - \pi ) \cdot \upsilon '\left[ P \right](\; - c'\; + \alpha \cdot w')$. Substitution into Eq. (20) yields Eq. (2) of the text,

$$ \begin{aligned} de/d\pi \approx \upsilon '\left[ P \right]( - \;c' + \;\alpha \cdot w') + \pi /(1 - \pi ) \cdot \upsilon '\left[ P \right]( - c'\; + \alpha \cdot w') \hfill \\ \quad \;\quad \;\; \approx 1/(1 - \pi ) \cdot \upsilon '\left[ P \right]( - c'\; + \alpha \cdot w')\; \hfill \\ \end{aligned} $$

(21)

Now consider another shock, $d\alpha > 0$, representing an exogenous change causing medical ethics to be strengthened or more pronounced. For instance, some physician may simply be a more ethical type than another, which raises the question of whether this has an impact on his or her effort. By analogy, the sign of $de/d\alpha$ is the same as the sign of $\partial^{2} U/\partial e\partial \alpha$, which is approximated by

$$ de/d\alpha \approx \partial^{2} U/\partial e\partial \alpha = \pi \cdot \upsilon '\left[ P \right] \cdot w'\; + \;(1 - \pi ) \cdot \upsilon '\left[ F \right] \cdot w' $$

(22)

This is Eq. (3) of the text; it is an approximation because in principle there is an additional term [see Eqs. (1) and (17)], amounting to $+ \pi \cdot \upsilon ''\left[ P \right]( - \;c'\; + \alpha \cdot w')^{2} + (1 - \pi )\upsilon ''\left[ F \right]( - \;r'\; - c'\; + \alpha \cdot w') \cdot w'\; < 0$ given the usual assumption of decreasing marginal utility, i.e., $\upsilon ''\left[ \cdot \right] < 0$. To reflect risk aversion in the context of a risk prospect $\upsilon ''\left[ \cdot \right] \,< \,0$ would be required; however, given that risk aversion is neglected, $\upsilon ''\left[ \cdot \right] = 0$ is a justifiable simplification since this term is of second order relative to $\upsilon '\left[ \cdot \right]$.

Physician in private practice subject to a shift from FFS to P4P The change dπ has to be replaced by a change dω > 0, indicating an increase in the share of activity financed by P4P. From Eq. (5), one has the modified first-order condition,

$$ \frac{dU}{de} = \omega \cdot \upsilon '\left[ O \right](p' \cdot w'\; - c'\; + \alpha \cdot w') + (1 - \omega )\upsilon '\left[ F \right](r'\; - \;c'\; + \alpha \cdot w') = 0 $$

(23)

From this, one obtains

$$ \begin{aligned} de/d\omega &\approx \partial^{2} U/\partial e\partial \omega = \upsilon '\left[ O \right](p'w'\; - \;c'\; + \alpha \cdot w') - \upsilon '\left[ F \right]( - c'\; + \alpha \cdot w') \hfill \\ &= 1/(1 - \omega ) \cdot \upsilon '\left[ O \right]\left\{ {( - c'\; + \;(p'\; + \alpha )w'} \right\} \hfill \\ \end{aligned} $$

(24)

after substitution of the expression $\upsilon '\left[ F \right](r'\; - c'\; + \alpha \cdot w') = - \omega /(1 - \omega ) \cdot \upsilon '\left[ O \right]\left\{ { - c'\; + \;(p'\; + \alpha )w'} \right\}$derived from the first-order condition (23). This is Eq. (6) of the text. Next, differentiation of Eq. (23) w.r.t. α yields

$$ de/d\alpha \approx \partial^{2} U/\partial e\partial \alpha = \omega \cdot \upsilon '\left[ O \right] \cdot w'\; + \;(1 - \omega ) \cdot \upsilon '\left[ F \right] \cdot w' $$

(25)

In full analogy with Eq. (22) above. This is Eq. (7) of the text.

Hospital and group practice physician subject to a shift from FFS to PP One obtains this by differentiating Eq. (9) of the text,

$$ \frac{dU}{de} = \pi \cdot \upsilon '\left[ P \right]( - C'/n + \alpha \cdot w') + (1 - \pi ) \cdot \upsilon '\left[ F \right]\left\{ {R'(e)/M - c'\; + \alpha \cdot w'} \right\} = 0 $$

(26)

For an interior solution to be obtained, either the marginal cost per patient $C^{\prime}/n$ must exceed marginal patient well-being weighted by ethics $\alpha \cdot w'$, or marginal revenue per physician $R^{\prime}/M$ must exceed the physician’s own marginal cost of effort $c'$.

Let the importance of PP increase. Assuming again that the second-order condition, $\partial^{2} U/\partial e^{2} < 0$ is satisfied, the sign of $de/d\pi$ is determined by the sign of $\partial^{2} U/\partial e\partial \pi$,

$$ de/d\pi \approx \partial^{2} U/\partial e\partial \pi = \upsilon '\left[ P \right]( - C'/n + \,\alpha \cdot w') - \upsilon '\left[ F \right](R'/M - c'\; + \alpha \cdot w') $$

(27)

By Eq. (27), one has $\upsilon '\left[ F \right](R'/M - c'\; + \;\alpha \cdot w') = - \pi /(1 - \pi ) \cdot \upsilon '\left[ P \right]( - C'/n + \;\alpha \cdot w')$.

Substituting this into Eq. (29), one obtains Eq. (10) of the text,

$$ \begin{aligned} de/d\pi &\approx \upsilon '\left[ P \right]( - C'/n + \alpha \cdot w')\\&\quad - \pi /(1 - \pi ) \cdot \upsilon '\left[ P \right]( - C'/n + \alpha \cdot w') \hfill \\ &\approx 1/(1 - \pi ) \cdot \upsilon '\left[ P \right]( - C'/n + \alpha \cdot w') \hfill \\ \end{aligned} $$

(28)

As to $de/d\alpha$, one approximately obtains from Eq. (29), again neglecting terms involving second derivatives of utility as below Eq. (25),

$$ de/d\alpha \approx \partial^{2} U/\partial e\partial \alpha = \;w'\left\{ {\pi \cdot \upsilon '\left[ P \right]\, + \,(1 - \pi )\upsilon '\left[ F \right]} \right\}, $$

(29)

which is identical to Eq. (22). This is Eq. (11) of the text.

Hospital and group physician subject to a shift from FFS to P4P Finally, consider an increase in the share of P4P-financed activity, dω > 0. Differentiation of Eq. (12) yields

$$ \frac{dU}{de} = \omega \cdot \upsilon '\left[ O \right](R' \cdot w'/M - c'\; + \alpha \cdot w') + (1 - \omega )\upsilon '\left[ F \right](R'/M - c'\; + \alpha \cdot w') = 0 $$

(30)

which implies $\upsilon '\left[ F \right](R'/M - c'\; + \alpha \cdot w') = - \omega /(1 - \omega ) \cdot$ $\upsilon '\left[ O \right](R' \cdot w'/M - c'\; + \alpha \cdot w')$.

Applying Eqs. (19) and (30) in analogous manner, and substituting the one obtained in Eq. (14) of the text,

$$ \begin{aligned} de/d\omega &\approx \partial^{2} U/\partial e\partial \omega\\ &= \upsilon '\left[ O \right]\left\{ {(R' \cdot w'/M - c'\; + \alpha \cdot w') - \upsilon '\left[ F \right](R'/M - c'\; + \alpha \cdot w')} \right\} \hfill \\ &= \upsilon '\left[ O \right]\left\{ {(R' \cdot w'/M - c'\; + \alpha \cdot w')\; + 1/1 - \omega ) \cdot \;(R'/M - c'\; + \alpha \cdot w')} \right\} \hfill \\ &= 1/(1 - \omega ) \cdot \upsilon '\left[ O \right]\left\{ {R'/M(w'\; - 1) - c'\; + \alpha \cdot w')} \right\} \hfill \\ \end{aligned} $$

(31)

As to $de/d\alpha$, it is again analogous to Eq. (22) above (again neglecting the second derivatives of utility),

$$ de/d\omega \approx \,\;1/(1 - \omega ) \cdot \upsilon '\left[ O \right]\left\{ {R'/M \cdot (w'\; - 1) - c'\; + \alpha \cdot w')} \right\} $$

(32)

This is Eq. (15) of the text.