This post is a bit disjointed as I am trying to put together a couple of different ideas from a set of Twitter conversations on network directories, network adequacy and general attractiveness of plans to members who have real needs.
— Daniel Polsky (@healthecon_dan) November 13, 2015
My reply to the question about creating better network tools was that we need tools that allow members who are shopping for plans to get a reliable and accurate (enough) estimate of provider availability:
At the same time, there was an interesting tweet on the accessibility of all classes of AIDS/HIV medications on Exchange.
— Caroline Pearson (@CPearsonAvalere) November 11, 2015
This led me to think back to one of my favorite posts on formularies and Gresham’s Law:
Insurers are required to accept and cover HIV patients. They don’t want to. So they are trying to avoid them by being fugly.
Insurance companies still want to tilt their risk pools to be as healthy as possible while letting their competitors eat the costs of covering the known sick….
Let’s assume that any given insurance company wants to minimize their HIV treatment costs and that they are also required to cover any HIV patient who signs up with them. The goal then for the insurance company is to make themselves as unattractive as legally possible to HIV patients. Futzing around with networks is possible, but since most HIV treating docs and facilities are common providers for lots of much healthier members, this will not be too effective. Additionally plans are required to contract with Ryan White AIDS clinics. The simplest legal way to target unattractiveness to HIV patients is to make the drugs as expensive as possible.
In my previous position at Mayhew Insurance, I was the directory provider data nerd among my many other responsibilities. We redesigned our web directory to accommodate the massive changes in our provider network perspective for the Exchanges. Before the Exchanges went live, Mayhew Insurance had a single imperative for the networks — Bigger is better. With the Exchanges we chopped our commercial networks into half a dozen sashimi thin slices. Each network was aimed at a different price point and a different membership disease burden mixture.
As part of the directory redesign process, I learned a whole lot about the web analytics side of the business.
We tracked what searches were used from both the public side and the password protected member side (that data was linked to member information which then meant it was linkable to claim information), what specialties were searched for most often and when, what searches were deep and which searches were shallow or quickly backed out of. We looked where clicks occurred, we looked at where people hovered. The data was large, it was robust and if there was a data miner available, it would have been extremely valuable. The data was mainly being collected at that time for human computer interaction purposes as it informed us that the way we categorized and displayed urgent care clinics was very unfriendly while our links to external provider directories and prescription drug searches was confusing. People liked the individual provider search functionality, so we templated the redesign off of that module.
That type of data can be collected from any proprietary network directory. For current members, it would be useful data for user interaction purposes as well as seeking correlations with current and near future care. This could be a good thing as a person who is looking for a rheumatologist could be flagged so that they get sent a water exercise class brochure at the community center three neighborhoods over from them. Or it could push people preemptively into clinical management before crisis hits so the insurer saves money by averting a crisis which helps the individual out as well as hospital stays are not fun.
There is a downside towards this type of data if it is in the insurer’s hands during a shopping period. It would allow insurers to tweak their networks during the open enrollment period to make them less attractive to sick people. If a network had two available rheumatologists who have appointments in the next sixty days and are located forty miles away from a prospective member who needs a rheumatologist, they are less likely to go into that network than a network that has eleven available rheumatologists including three at the medical building the next town over. Those types of network decisions are made during the construction and filing process for plan/policy approval already. However during the enrollment process, tweaks are allowed to occur. Insurers could ask providers to change their panel status (availability for new patients) or confirm if an office that is open two days a month should be listed in the directory. More likely, insurers would look at the parameterized directory searches as both an early warning system for the contract year and as a critical design input for the following plan year.
If a health plan is seeing a large proportion of new to them members searching for rheumatologists before they buy, health plans that are seeking to shift as much risk off their books won’t actively recruit rheumatologists and won’t push back against rheumatologists who close their panels or reduce office hours for the next plan year. The goal would be to make their plans ugly to people with arthritis so that they can be someone else’s problem.
Good tools are useful in the first period for member and prospective member knowledge. However, if the tools are based at the health insurer level, the information gained during the first period (open enrollment to plan design deadline for the next plan year) will create massive health selection information that the insurance industry will aggressively game to the detriment of the member. If the tools are based at either the Exchange level or a third party platform where all insurers submit machine readable directories for display and the process analytics can not be released to the insurers, this could be a way to provide better information without as much selection seeking behavior.