Group insurance and scream minimization

The majority of insured Americans are insured through an employer based benefit arrangement. Commercial health insurance tends to pay clinicians significantly more per service provided than Medicare. This leads to very expensive insurance premiums that come out of employee’s cash compensation.

Group insurance is purchased through two channels. There are fully insured plans where a group sends an insurer a big check every year to take on all cost risk after cost-sharing is paid. This fully insured model is how the ACA individual market works. The other model is the self-insured model where an employer group contracts with an insurer to run the logistics of paying for health care for their employees, but the group holds financial risk of really expensive claims. Self-insurance is very common for large employers (500+ covered lives) and not uncommon for medium sized groups (100-500 covered lives). Small employers can provide insurance through either route.

Medicare and Medicaid keep their per-unit costs relatively low by engaging in administrative pricing with take it or leave it offers. Almost every doc and hospital will take Medicare pricing and payment. Many docs and hospitals will take Medicaid pricing and payment policies (although new research has done a nice job of modeling the hassle costs of Medicaid payment policy as a deterrent to broad networks!). ACA plans use narrow networks to sometimes extract low provider pricing from clinicians.

So why does commercial group insurance cost so much?

I call it the scream minimization function of HR acting as a poor agent for the principal.

We can think of the purchasing decision for health insurance of a large insurer to be done by the HR department acting as agents for the company as a whole. The HR department has a strong budget constraint. The can’t spend more than X for healthcare for a given group of employees. They engage with brokers/agents/consultants to search for bids that satisfy this budget constraint. Once they get a menu of plausible options that satisfy the budget constraint, we can think of HR engaging in locally motivated behavior. One of these behaviors may be a strong desire to not get screamed at (literally, figuratively or bureaucratically) by employees. And while most people would prefer not to piss off other people, there is probably a strong preference to avoid getting into conflicts with high prestige and power individuals while individuals/groups with low power/prestige can be ignored or minimized. This means making sure that the network is attractive to individuals with high power and prestige. Individuals with high power and prestige may earn well above median income in the group. Individuals with higher income are less deterred by cost-sharing and are likely to have a higher willingness to pay.

From this perspective, lets think about two plans that cost the company the same. The first choice has a narrow network and relatively low cost-sharing while the second choice has a broad network and twice the cost-sharing. There is a trade-off if we hold premiums constant — cost-sharing vs. network. If we think that there is a strong likelihood that the second, broad network with more cost sharing minimizes the power weighted push-back on the health plan choice, we should expect HR to choose the big network and pay a high per unit price for the privilege.

There is a recent working paper by Nicholas Tilipman at UIC that maths this up a lot:

In this paper, I study the determinants of health plan offerings among large employers and whether these plan choices reflect preferences that are aligned with that of employees. This is an important market to study this issue: employer-sponsored insurance (ESI) is a significant part of the healthcare landscape (ESI), representing approximately 30% of health expenditures. Moreover, costs in the employer market have been rising rapidly in recent years. Per-enrollee expenditures in the private ESI market have increased about 46% between 2008 and 2018, compared with an approximate 21% increase in Medicare per-enrollee spending over the same period.2 This paper sheds light on the whether a portion of these rising costs can be attributed to mismatch between employer and employee preferences.
I focus my analysis on employer decisions over health plan provider networks and, in particular, the decision of whether to offer “narrow-network” benefit designs as part of their plan menus. Health insurers and employers have increasingly started offering these insurance plans as a means of containing spending and offering consumers low-cost options.3 Despite the increasing popularity of narrow-network plans on the health insurance Exchanges, however, employers have been slower to adopt, design, and offer such products. In 2016, only 7% of employers nationally offered a narrow network as part of their plan menu (Hall and Fronstin, 2016) …

This leaves the question of why employers exhibit this behavior in menu choice. I explore several candidate possibilities: (a) inability to alter co-premiums or benefits (by virtue of the employer I study being a public rather than private employer); (b) employer mistakes or misperceptions; and (c) heterogeneity in the types of employees the employer values most when designing benefits. While I am not able to fully separate these channels, I find substantial evidence for (c). Specifically, about 30% of the estimated employer-employee mismatch can be attributed to employers placing a higher weight on the network references of the oldest workers in the distribution, while as much as 80% of the mismatch can be explained by employers emphasizing the preferences of employees in certain geographic regions. These regions tend to be ones that are less dense, have fewer competition among health care providers, and are situated close to the state border. As such, households in these regions stand to lose the most utility from a loss of a notable provider….

What he finds is that employers choose networks, that, on average, are bigger than the networks individual employees would prefer. And these big networks are expensive relative to optimal networks.

He comes to a different conclusion than my toy mental model but the finding is that employers buy bigger than optimal networks.

One of the few intriguing and potentially enduring aspects of the Trump administration health policy legacy is the creation of Individual Contribution Health Reimbursement Accounts (ICHRA). These accounts allow for employers to give a defined business group of employees an age adjusted sum of money to spend on ACA individual market plans. These ICHRA plans are likely attractive to individuals with locally below average expected medical costs and system interactions. These individuals are more likely to either buy a narrow network with high actuarial value and spend the entire employer subsidy, or buy a narrow network, low actuarial value plan and use some of the employer subsidy to pay some cost-sharing. Individuals with expensive prospective healthcare spending are likely to be worse off as the broader network ACA plans are likely significantly more expensive than the skinniest network plans on the marketplace.

These are two very long ways to get to the point that most of the time, most people are mostly indifferent to their health insurance, but the problem is the not knowing when indifference transitions to top of mind prioritization.