# An Inherent Weakness in the Completion Factor Approach to Calculating an Incurred But Not Reported (IBNR) Reserve

Apr 27, 2023 | IBNR

Incurred But Not Reported (IBNR) Claim reserves for a fiscal period are healthcare related costs that individuals have incurred due to an event such as Dr. appointments, hospital visits, etc. – that have or will generate invoices that haven’t yet been processed or adjudicated and, as such, have not shown up in the cash flow of a healthcare sponsoring organization. Because the events that gave rise to those invoices have taken place, they must be accounted for as a ‘reserve’ on the financial pages of the sponsoring entity for the period in which they took place.

While there are different methods used to calculate reserves, most approaches involve determining a completion factor from a claim experience that is known.  For example, if prior claim patterns are such that on average 30% of all claims for a given month are processed and known in the month in which they occur, then for the final month of a reporting period, if \$30,000 of claims for that month are known and reported, one might assume fully incurred claims for that month of \$100,000 (i.e. \$30,000 / 0.30) will ultimately be reported, including \$70,000 (\$100,000 – \$30,000) in the IBNR for that reporting period.

Suppose that in our example above, only \$20,000 of claims for that final month of the reporting period are known and reported. Using the completion factor approach would suggest the total for the month would ultimately be \$67,000 (i.e., \$20,000 / .30) – of which \$47,000 is IBNR. This is about \$23,000 less than the first reserve calculation, or about 33%  – a meaningful difference. The second, lower IBNR, might be warranted if, in that situation, the lower reported claims were due to differences in claim experience for those two groups. However, what if the difference was due to some processing delays and not due to lower overall claim experience? The result would be an understatement of the IBNR. In addition to suggesting an IBNR \$23,000 lower than the first estimate, it would also miss the \$10,000 (\$30,000 less \$20,000) yet to be processed under normal circumstances.

One step to take to avoid such a mistake would include calculating an average claims cost per member per month (pmpm) for prior months to make sure the last month’s results aren’t unusual. If the number of covered lives remains steady during the coverage period and significant deviations from the average pmpm cost are observed, then a moving average of the last six, or twelve months, might be used as a minimum basis for the final month’s total claim experience. In addition, an analysis could include looking at past years to note whether seasonal patterns emerge so no or little adjustment might be warranted.

Another situation that could lead to incorrect results is the case where a very large, non-recurring claim event is in the data. This can likely cause a large overstatement. For example, if we take the illustrative data from above and suppose a large claim of \$75,000 is included, in addition to the \$30,000, then using the completion factor calculation approach, the estimated IBNR for that month would be \$245,000 (i.e. \$75,000 + \$30,000) / .30 = \$350,000, for an IBNR of \$350,000 – \$105,000 = \$245,000. If the estimate for incurred claims (without the large claim) would normally be about \$100,000, then it would make sense to look at the pmpm alternative for the month in question.

Another potential distortion can occur if the number of check cycles to providers is greater than the average. For example, if an administrator reimburses providers bi-weekly, or 26 times per year, then in any given year, two months will have one cycle more than the other ten months. Thus, if the data happens to contain an additional cycle, the normal completion factor approach would – all other things the same – likely overstate the estimated fully incurred claims and the estimated IBNR.

In some of these cases, a helpful step would be a call to the plan sponsor. Perhaps there has been a change in the third party-administrator and processing patterns under the new administrator are different. There could also be an interruption in processing services because the change in administrators has not gone smoothly. This could possibly suppress payments in the initial months and then inflate payments in later months as claims inventories are reduced. In both cases, claims patterns would not be reliable until administration of the plan ‘settled’. In cases like these, it’s prudent to consult with the administrator to learn of issues affecting the reported claims paid. Other interruptions in claims payments could be the result of employee strikes or, as in the case of COVID, employee absences and alternative working arrangements.

Additionally, be mindful of the location and time of year for which the claim reporting is taking place. An IBNR calculation for a southwest Florida group in 2022 would have had to take a Hurricane into account as many businesses and facilities had to be shut down for several weeks – affecting both administrators and service providers.

It has been said that IBNR calculations are a combination of art and science so care must be taken to have an open mind when performing and reviewing these reserve calculations. While it’s always desirable to have an ‘iron-clad’ documented formula, in many cases supportable reasoning and experienced judgment must be used.

Looking for an experienced actuarial team to complete your IBNR calculations? Connect with our highly-qualified consultants at Watkins Ross to learn more about our health plan services.

Christian Veenstra, ASA, MAAA, FCA
President, Watkins Ross