Supplier-Induced Demand

Supplier-Induced Demand in Health Care

Supplier-induced demand is the health economics hypothesis that doctors and other providers can raise the use of their own services by advising patients towards more care than those patients would choose if they knew what the provider knows. The provider acts as the patient's adviser and as the seller of the treatment, so the amount of care is set partly on the supply side and does not simply reflect patients' preferences and clinical need. The hypothesis matters because it weakens the usual reading of utilisation as revealed demand and affects how resource use observed under one payment system should be carried into an economic evaluation for another. Some of the most persuasive evidence compares the care given to medically informed patients, such as doctors, with the care given to comparable patients who lack that knowledge.

Evans and the origin of the hypothesis

Johnson's review for the Encyclopedia of Health Economics attributes the concept to Robert Evans, writing in 1974. Evans argued that the large information differential between physician and patient lets the physician exert direct, non-price influence on demand for the physician's own services. If the physician's economic position affects how far that influence is used, he concluded, models of the demand for care that leave out supplier behaviour are incompletely specified.

Evans noted that such behaviour could produce perverse results, such as prices rising when supply increases, and reviewed evidence from Canada and the United States that he read as consistent with substantial demand influence, with "responses of generated output to physician stock around 80 per cent". His policy conclusion was that limiting price inflation, correcting shortages or restraining unnecessary utilisation could not rest on conventional supply and demand models. Physician-induced demand is the usual name when the provider is a doctor, and provider-induced demand is also used.

Inducement measured against the patient's own interests

Johnson's review adopts McGuire's definition, under which physician-induced demand exists when the physician influences a patient's demand for care against the physician's interpretation of the patient's best interests, a form of the principal-agent problem in healthcare rooted in information asymmetry. The models of an inducing provider, the effects of a fee cut and the early tests based on physician numbers per head are on the demand inducement page.

Informed patients as the test group

If providers can induce demand because patients lack information, inducement should be weaker among medically informed patients, a test that compares patients rather than areas.

Surveys of doctors as patients

Bunker and Brown compared surgery rates in the families of lawyers, business people and ministers with those in physicians' families. Johnson reports that, contrary to the prediction, self-reported rates were equal or higher in physician families, and that Hay and Leahy, with controls for income, insurance and self-reported health, also found higher use among physicians.

Domenighetti and colleagues reached the opposite result in Switzerland. Measuring age- and sex-standardised use of seven common surgical procedures by questionnaire, they found use significantly higher in the general population than among physician-patients for every procedure except appendectomy. Johnson summarises the result as the average person's probability of receiving one of the seven operations being one-third above that of a physician or a physician's family member.

Audit studies and physician-mothers

Survey comparisons cannot rule out differences in health, prices or preferences between the groups. Currie, Lin and Zhang ran an audit study in China in which pairs of simulated patients with identical flu-like symptoms visited the same physician and one of each pair signalled knowledge of appropriate antibiotic use. Johnson reports that prescription rates were 25% higher for the uninformed patient. The authors read antibiotic overuse in China as largely supply-side, though the physician may have read the signal as a preference.

Johnson and Rehavi combined patient information with variation in financial incentives, using linked birth and physician licensing records from California, with a second dataset from Texas. Compared with other highly educated mothers, physician-mothers were 7% less likely to have a caesarean section in California and 8% less likely in Texas, almost entirely through fewer caesareans after labour had started. In hospitals owned by a large health maintenance organisation, where physicians are salaried and neither physician nor hospital gains financially from a caesarean, non-physician mothers had risk-adjusted caesarean rates about 5 percentage points lower than elsewhere. Physician-mothers had the same risk-adjusted rates in both settings, so the gap between the groups largely closed where the financial incentive was absent. Physician-mothers and their infants also had less morbidity, which the authors read as a sign that overuse of caesarean section harms patients.

Ownership, liability and payment levels

Inducement can work through ownership as well as fees. Baker studied orthopaedic surgeons and neurologists in the United States who began billing for magnetic resonance imaging between 1999 and 2005, and found that once orthopaedic surgeons began billing, MRI procedures within thirty days of a first visit rose by about 38%, with spending on other care for their patients also rising. Johnson adds that care given to reduce the physician's own liability risk also fits the definition.

Clemens and Gottlieb, using a 1997 consolidation of Medicare's payment areas that changed fees by area, found a long-run elasticity of Medicare care with respect to payment rates of around 1.5, with economically small health effects estimated with limited precision. More care at higher fees fits ordinary profit-seeking supply as well as some inducement models, and the wider evidence is on the fee-for-service page.

What the concept leaves out

Labelle, Stoddart and Rice observed that twenty years of work on supplier-induced demand had said little about how inducement affects patients' health. Their framework adds the clinical effectiveness of the services used to how well the agency relationship works, and separates types of utilisation that the literature had mixed together. They concluded that a continued focus on inducement alone may be too narrow a basis for policy on health service use.

Care recommended by a good agent may still be ineffective, and care pushed by a self-interested agent may still do the patient good.

Only the bottom row is supplier-induced demand in McGuire's sense, and the right-hand column is a problem of clinical effectiveness in either row.

Supplier-induced demand in economic evaluation and costing

Inducement affects evaluation through resource use, value and forecasting. A cost-effectiveness analysis that draws utilisation from fee-for-service claims data should report which payment system generated them.

Demand curves from observed use are read in welfare economics as the value of care to patients, and Johnson notes that inducement affects the interpretation of parameters central to the theory of optimal insurance. Where use is partly induced, willingness to pay inferred from utilisation overstates the value of care at the margin. Because uptake can depend on who owns a technology and how its use is paid, a budget impact analysis needs an explicit assumption about the provider response.

Worked example of an informed-patient benchmark

All numbers in this example are illustrative. A health system records 10,000 births a year to mothers without medical training in hospitals that pay obstetricians more for a caesarean than for a vaginal delivery. After risk adjustment, 30.0% of these births are by caesarean section. Medically informed mothers in the same health system have a rate 8% lower, or 30.0 × 0.92 = 27.6%. A caesarean costs GBP 1,800 more than a vaginal delivery.

1. Excess procedures. Treating the informed group's rate as the benchmark, the excess number of caesareans is

$$E = (r_U - r_I) \times N_U$$

where $E$ is the excess number of procedures a year, $r_U$ is the risk-adjusted rate among uninformed patients (0.300), $r_I$ is the rate among informed patients (0.276) and $N_U$ is the number of uninformed patients (10,000). The rate gap is 0.300 minus 0.276, which is 0.024, and 0.024 × 10,000 = 240 caesareans.

2. Cost of the excess. The extra cost is

$$C = E \times \Delta c$$

where $C$ is the annual extra cost in GBP and $\Delta c$ is the cost difference per procedure (GBP 1,800). The excess costs 240 × 1,800 = GBP 432,000 a year.

3. Allowing for sorting. Johnson and Rehavi attributed about a quarter of their gap to differences in the hospitals and obstetricians patients used, and the example assumes the same share. The gap within the same setting is then 0.024 × 0.75 = 0.018, which means 0.018 × 10,000 = 180 excess caesareans costing 180 × 1,800 = GBP 324,000 a year.

4. Interpretation. The calculation treats the informed group's rate as the right rate for everyone, which assumes informed patients are neither over-cautious nor different in preferences once risk is adjusted. The GBP 324,000 is a resource cost, and some of the 180 extra caesareans may still benefit mother or infant, so the figure marks the scale a payment or information change might address and is no promise of a saving.

Supplier-induced demand beside nearby concepts

Several terms overlap with supplier-induced demand, and the differences decide which policy response fits.

Moral hazard and inducement can both raise use in the same market, which makes them hard to separate in data. Clinical practice variation is often blamed on inducement, though Johnson notes the link has not been rigorously tested, and the credence good page covers overtreatment and overcharging.

Limitations and common misreadings

Informed-patient comparisons have a weakness, since doctors and their families may differ from other patients in health, income and taste for treatment, as the conflicting survey results show. Induced care is not necessarily worthless. Most of the evidence comes from the United States and from fee-for-service payment, and Johnson expects the size of the effect to differ across incentive environments, specialties, patient groups and treatments, so a single estimate should not be applied to a tax-funded system with salaried clinicians.

Sources

• Baker LC. Acquisition of MRI equipment by doctors drives up imaging use and spending. Health Affairs. 2010;29(12):2252-2259. doi:10.1377/hlthaff.2009.1099
• Clemens J, Gottlieb JD. Do physicians' financial incentives affect medical treatment and patient health? American Economic Review. 2014;104(4):1320-1349. doi:10.1257/aer.104.4.1320. https://pmc.ncbi.nlm.nih.gov/articles/PMC4144420/
• Currie J, Lin W, Zhang W. Patient knowledge and antibiotic abuse: evidence from an audit study in China. Journal of Health Economics. 2011;30(5):933-949. doi:10.1016/j.jhealeco.2011.05.009
• Domenighetti G, Casabianca A, Gutzwiller F, Martinoli S. Revisiting the most informed consumer of surgical services: the physician-patient. International Journal of Technology Assessment in Health Care. 1993;9(4):505-513. doi:10.1017/s0266462300005420
• Evans RG. Supplier-induced demand: some empirical evidence and implications. In: Perlman M, ed. The Economics of Health and Medical Care. London: Palgrave Macmillan; 1974. p. 162-173. doi:10.1007/978-1-349-63660-0_10
• Johnson EM. Physician-induced demand. In: Culyer AJ, ed. Encyclopedia of Health Economics, Volume 3. Amsterdam: Elsevier; 2014. p. 77-82. doi:10.1016/B978-0-12-375678-7.00805-1
• Johnson EM, Rehavi MM. Physicians treating physicians: information and incentives in childbirth. NBER Working Paper 19242. Cambridge, MA: National Bureau of Economic Research; 2013. https://www.nber.org/papers/w19242
• Labelle R, Stoddart G, Rice T. A re-examination of the meaning and importance of supplier-induced demand. Journal of Health Economics. 1994;13(3):347-368. doi:10.1016/0167-6296(94)90036-1