Before the unwinding began, Tricia Brooks highlighted why call center data would be an important piece of the puzzle in understanding how states’ returns to routine operations were going. Reporting call center data, along with other data reporting requirements enacted by the Consolidated Appropriations Act, 2023 (CAA) help CMS track how the unwinding is going in each state. However, since we examined the first batch of call center data provided by CMS from March, it has become increasingly clear that the data may not be as useful as we hoped it would be.
With the latest round of data released by CMS, we now have 9 months of call center metrics from 49 states and the District of Columbia (Note: South Dakota does not operate a call center and therefore was not included in our analysis.). Overall, we have seen total call volume drastically increasing as states have been resumed regular Medicaid renewals and terminating coverage. During this time, nearly 10 million individuals called state help lines, often waiting for near an hour or more to talk to a representative. These numbers are likely an underestimate of the actual number of calls and average call wait time.
It is important to note that these metrics are only an average for a range of callers’ experiences. For example, while call wait times went up on the national level, this data cannot show that Spanish-language in Florida callers had to wait four times longer (close to two and a half hours!) to speak to a live representative than English-language callers. Not only that, but an August 2023 report by UnidosUS shows that almost a third of all Spanish-language callers had their calls disconnected before reaching a live representative, compared to a tenth of English-language callers. A recent update by the advocacy group found that metric increased to 8 in 10 calls being automatically disconnected, leaving thousands without the help they need. These issues prompted lawmakers in the state to put more than $12 million in the budget to help improve Florida’s call center system. (The budget has yet to be signed by Governor Ron DeSantis.)
CMS describes some of the shortcomings in call center data through detailed footnotes in each data release. These notes identify issues like whether the total call volume metric contains calls for other benefit programs (such as SNAP or WIC) or state-based marketplace coverage, or if the state gives the option for callbacks. In some cases, the note simply states that the total number does include all calls received by all call centers. The list of caveats is lengthy and extends across most call center data, often with multiple notes per state. In fact, Oklahoma is the only state without clarifying notes on its call center data. That said, the data from CMS is the only publicly-reported call center statistics from all 50 states and the District of Columbia, and is worth examining for broad trends as the unwinding process continues. So what should we know when looking at the CMS call center metrics? Let’s review the some of the most common caveats in this data:
1. Does not include all calls received by call centers
This footnote is seemingly straightforward but could potentially have a huge impact on call center data. At the most basic level, it means reported call volume, abandonment rates, and wait times were lower than actual since calls were omitted. North Carolina, which seems to have some of the best call center statistics in the nation, exemplifies this effect. There, the average time before a call is answered is zero minutes and less than 1% of calls are dropped, either intentionally or unintentionally. Depending on which or how many calls were excluded and if this was used in combination with other data notes (i.e. NC data has (C3) Does not include all calls received by call centers; (C4) Does not include all calls received after business hours; and (C6) Includes only calls transferred to a live agent), the effect can vary.
2. Does not include all calls received after business hours
It is important to remember that the state reported metrics only represent the average for the reporting month and do not show the range of user experiences. This is especially relevant in cases of states that include wait times and abandonment rates from calls when the call center is not open. How can you compare the wait time of someone calling at a peak time, say midday during lunch break, with wait times after work hours? Alternatively, consider how the abandonment rate in states that do include calls after business hours (where someone may hang up because the call center is closed) may differ from states that do not include them (where people may abandon the call or be disconnected after waiting on hold). Considering the call volume metric, it is impossible to know what share of the calls did or did not come in after business hours, where all of these individuals and families are not getting connected to the help they need. Including or not including these numbers could artificially inflate or deflate any of the three metrics.
3. Includes calls for other benefit programs
More than half of states rely on the same small group of individuals to answer calls about multiple state assistance programs. This means that those who are looking to get help with their Medicaid or CHIP applications/renewals are waiting in the same queue with people looking for help on SNAP, child care assistance, or other benefits. In fact, the top 10 states with the highest average wait times (ID, RI, MT, FL, ME, NM, UT, MO, NV, AZ) are also states that include calls for other public benefits. Interestingly, seven states (CT, KY, MD, MN, NY, RI, VT) operate an integrated call center for Medicaid and the state-based marketplace (SBM); and call center data for three of those states (KY, MD, VT) also includes calls regarding other benefit programs.
4. Includes only calls transferred to a live agent
Forty states reported that their data includes only calls transferred to a live agent on at least one of their three metrics – meaning any calls that are abandoned due to long wait times or disconnected before the caller is connected to a live agent were not counted. Additionally, the effect of this can be amplified when states also have other caveats.
The increased data transparency through public reporting of call center metrics is something to be celebrated and we hope will continue beyond the reporting period required by the CAA which ends on June 30, 2024. But due to the limitations of how the data are reported, it should be used with caution, as it may not be fully representative of what beneficiaries are actually experiencing when seeking assistance.
