Problem
Medicare Advantage (MA) is the private insurance version of Medicare, the US government health programme for people aged 65 and over. Private insurers run MA plans and receive a fixed monthly payment from the government for each person they enrol. That fixed payment creates a temptation: attract healthier, cheaper patients and avoid sicker, more expensive ones. This practice is known as 'cream skimming'.
In 2018, the US Congress passed the CHRONIC Care Act, which took effect in 2020. It gave private MA insurers new freedom to offer non-medical extras alongside standard healthcare such as meal deliveries, home safety equipment, transport to shops, even park passes. The idea was that giving insurers greater flexibility could help them better manage the health of people with complex needs and reduce healthcare costs. We set out to test whether the law might also make cream skimming worse.
Science
We built a game theoretic model to test what private insurers would do under the new law. Our model maps the decisions of three groups: the Centers for Medicare and Medicaid Services (CMS), which sets payment rules; the private MA insurer; and the patients themselves.
To capture the new flexibility in our model, we distil the range of services an insurer could offer into two stylised options. The first is preventive: services that target a patient's condition and reduce the chance of a serious health episode but cost more to deliver to sicker patients. The second is general wellbeing: services such as transport or home help that improve daily life equally for everyone. The key insight built into our model is that patients with the same medical diagnosis can vary enormously in how vulnerable they are, because of factors such as diet, housing and income, known as social determinants of health (SDOH). Because the government's payment system does not fully account for these differences, two patients who look identical for payment purposes can cost very different amounts to treat.
Outcome
We find that what insurers choose to do depends critically on how much the government pays them for each patient. When payments are high, insurers choose preventive services, enrol more patients and deliver better health outcomes. When payments are lower, insurers may prefer to offer general wellbeing services and set their charges to attract less vulnerable, cheaper patients.
Most importantly, at intermediate payment levels, this can lead insurers to enrol fewer people than they would have before the law. In other words, under some conditions the additional flexibility created by the law can make cream skimming worse rather than better. Our analysis shows that these conditions can plausibly arise in practice, particularly when patients differ substantially in their vulnerability, general wellbeing services are attractive relative to their cost, and preventive services have only a modest effect on the risk of serious health episodes.
We also find that insurers who are efficient at delivering medical care but find preventive services expensive are the most likely to cream skim. Being good at keeping medical treatment costs low does not necessarily make an insurer more likely to offer preventive services or avoid cream skimming.
Change
This is the first formal analysis of the CHRONIC Care Act's potential effects on insurer behaviour and patient enrolment. Our research shows why CMS may need to think carefully about the payments it makes to MA insurers following the change in policy. The payment system was not changed when the new flexibility was introduced, yet our results show that the level of these payments can determine whether the new flexibility improves health or instead creates opportunities for cream skimming.
One possible response is to adjust payments so that insurers have a stronger incentive to provide preventive services. But our model shows that this may require relatively high payments, and the additional cost can be large enough to outweigh the resulting health gains. The government may therefore face a genuine choice between better health outcomes at higher cost and broader patient welfare at lower cost.
Our analysis also points to other possibilities policymakers could consider. The freedom to offer a wider range of services could, for example, depend on an insurer's bids or past behaviour. And accounting for social determinants of health when setting payments could reduce opportunities for cream skimming, although doing so may be difficult in practice.
Woonam Hwang, KAIST; Jónas Oddur Jónasson, ImperiaBusiness School; Heikki Peura, Aalto University, 'Favorable risk selection in Medicare Advantage: the effect of allowing non-medical services', Manufacturing & Service Operations Management (forthcoming)