Long-term care insurance (LTCI) is almost 50 years old. In the 1990s, more than 100 companies were selling policies to individuals and to groups. In 2003, the pattern of annual increases in sales came to an abrupt end and LTCI policy sales began to decline rapidly. Between 2003 and 2009, individual policy sales declined by 9% per year. Fewer than 15 carriers remain in the market.¹

As the U.S. population ages, the market for hybrid life and LTCI policies has been growing. In addition, we see new standalone LTCI products filed in 2025 from insurers that had previously ceased most sales.

We show in Figure 1 the expected benefits (paid claims) that long-term care (LTC) insurers will provide over the remaining life of the policies as of year-end 2023. We did not model new sales of LTCI or hybrid life and LTCI.

The National Association of Insurance Commissioners (NAIC) reported LTCI incurred claims of approximately $16 billion in 2023. We forecast an increase in LTCI paid claims through about 2041, with a peak of paid claims of $42 billion. Since we did not include new sales estimates, the projection of paid claims declines over the remaining decades, reflecting the aging of the in-force population. We also assumed that historical paid claims were similar to reported incurred claims for the purpose of our dynamic validation.

We used a first-principles actuarial projection model. We combined our understanding of the current nationwide LTCI policy base with our expectations of mortality, voluntary lapse, and morbidity (claim incidence and severity) to project paid claims for this exercise. The sections below provide further details of the model and assumptions. The actual future mix of policyholders (by policy type and demographics) will almost certainly be different from what we’ve assumed here.

Assumptions and Methodology

Projection model

We built a first-principles model in Milliman’s proprietary Integrate® software (powered by the MG-ALFA® calculation engine and formula database). The software’s standard calculation logic allows for separate modeling of claim incidence, utilization, claim recovery rates, and mortality (including separate modeling of active and disabled life mortality). All assumptions were modeled on a first situs of care basis.

We assumed a starting claim reserve of approximately $47 billion at December 31, 2023, and projected paid claims covered by this reserve. To smooth the initial claims in our projection we interpolated paid claims between 2023 and 2034. This was important given our methodology of bucketing attained ages at the valuation date and that applying the mix of new incurrals and existing paids produced discontinuity in the projection. The smoothing did not materially change our incidence rate projection.

Distribution of business

We assumed the following mix of business, for roughly 6 million LTCI policyholders at December 31, 2023. These assumptions are based on NAIC reported results and our experience. We used this as the in-force input for our projection.

• Average attained age at year-end 2023 of 71
• Male (36.6%) / female (63.4%)
• Issue year (buckets): 1995 (3.0%), 2000 (22.1%), 2005 (29.8%), 2010 (31.5%), 2015 (6.1%), 2020 (7.5%)
• Reimbursement (85%) / indemnity (15%) / cash (none)
• 5% compound for life inflation (50%) / no inflation (50%)
• Married at issue (60%) / single at issue (40%)
• Benefit period: 3 year (70%) / lifetime (30%)
• Elimination period: all 90-day
• Policy type: comprehensive
• Tax-qualified status for all policies
• Original daily benefit amount: $225 (non-inflation) / $150 (inflation)
• Assumed issue age and underwriting distribution that produced a projected incidence rate consistent with historical incidence rates from Form 1.

Morbidity Assumptions

We modeled incidence, utilization, and claim continuance assumptions.

Morbidity assumptions are for the most part based on the 2023 Milliman LTC Guidelines. The 2023 Guidelines are based primarily on claims from 2010 through 2019 to include most recent, relevant experience. We gave consideration to the experience during the COVID-19 pandemic in 2020 and 2021, but did not directly use this experience for assumption setting. The primary 2010–2019 study period for the 2023 Guidelines includes roughly 460,000 claims and 30 million life years of exposure and represents experience from 18 of the top 25 LTC insurers (based on lives in-force). The inception-to-date dataset includes 1.1 million claims and 82 million life years of exposure. This significant amount of data makes the Guidelines a credible baseline for developing morbidity assumptions. The general approach to updating the Guidelines is to analyze how recent experience aligns to the prior set of Guidelines. The 2023 Guidelines used predictive analytics to update the assumptions, including the development of incidence, claim termination (disabled mortality and recovery), and transfer rates.

• Incidence: We used incidence rates from the Milliman 2023 LTC Guidelines.
• Utilization: We assumed a starting utilization for skilled nursing facilities (SNFs) and assisted living facilities (ALFs) of 95% and a starting utilization for home healthcare (HHC) of 67% for reimbursement policies.
• Cost of care: For facility care, we used a 4% cost of care inflation and 100% days utilization. For home care, we used a 3% cost of care inflation and 70.5% days utilization. For all three situses, we used 95% dollars utilization, and floored future utilization at 10% less than starting utilization.
• Continuance: This is related to how long a claim will persist. Typically, this varies by lifetime or non lifetime benefit period and gender. We used continuance rates that vary by starting care situs from the Milliman 2023 Guidelines. The Guidelines’ continuance curves are also split into disabled life mortality and recovery rates used in our model.
• Underwriting selection factors: Selection factors were applied to the morbidity basis to reflect the effect of underwriting and the method of issue. The Guidelines’ selection factors vary by underwriting level (e.g., tight, moderate, or loose), the method of issue (e.g., full underwriting or guaranteed issue), issue age, gender, and marital status. Our research has shown that there is significant difference in morbidity by marital status due to a partner being able to help administer care at home. This difference wears off over time as divorces and deaths of spouses occur.

Policy termination rates

The mortality and lapse assumptions are based on industry experience. Our projection model reflects that a policyholder can terminate their policy under three scenarios: death, voluntary lapse, or benefit exhaustion.

• Deaths: We selected the 2012 individual annuity mortality (IAM) table based on industry trend. Since we are using a first-principles model, we used scalars to convert from total mortality to active life mortality. These scalars are based on internal Milliman research. Disabled life mortality is consistent with the Milliman 2023 Guidelines.
• Lapse: Most companies in the industry see an ultimate voluntary lapse rate of around 0.5% to 1.5%. We chose an ultimate lapse rate of 1.0% starting in duration 5.
• Benefit exhaust: This is based on when a given claimant with non-lifetime benefits is expected to use up their benefits based on their policy design and utilization assumptions. The projection model tracks accumulated claim payments / service days to calculate benefit exhausts within the model.

Ancillary benefits or paid-up policies

We did not explicitly model ancillary benefits directly (we assume they are reflected in the final true up of paid claims). We did not model any paid-up policies or nonforfeiture benefits.

Caveats and Limitations

The attached results are based on Milliman research and on our experience in working with LTCI carriers. Actual experience will vary from our estimates for many reasons, including differences in policy design such as benefit periods, elimination periods, reimbursement levels, policyholder demographics, geography, and the delivery of healthcare services, as well as other non-random and random factors. It is important that actual experience be monitored and adjustments made as appropriate. Our estimates are not predictions of the future; they are estimates based on the assumptions. If the underlying data or other listings are inaccurate or incomplete, this analysis may also be inaccurate or incomplete. Emerging results should be carefully monitored with assumptions adjusted as appropriate. Our analysis is based on historical practice patterns and treatments that may change over time.

Milliman has developed certain models to estimate the values included in this report. The intent of the models is to project future LTC paid claims. We have reviewed the models, including their inputs, calculations, and outputs for consistency, reasonableness, and appropriateness to the intended purpose and in compliance with generally accepted actuarial practice and relevant actuarial standards of practice (ASOP).

Milliman does not intend to legally benefit any third-party recipient of its work product. Even though Milliman has consented to the release of its work product to a third party, any third-party recipient of this report should not rely upon Milliman's report, but should engage qualified professionals for advice appropriate to its own specific needs. The statements contained in the report are those of the authors and do not necessarily represent the views of Milliman or its other consultants.

We, Rachel Marsiglio, Juliet Spector, and Robert Eaton, are members of the American Academy of Actuaries and meet its Qualification Standards to provide this statement of actuarial opinion.

¹ Cohen, M. A., Kaur, R., & Darnell, B. (2013). Exiting the market: Understanding the factors behind carriers' decision to leave the long-term care insurance market. Office of the Assistant Secretary for Planning and Evaluation. Retrieved February 27, 2025, from https://aspe.hhs.gov/reports/exiting-market-understanding-factors-behind-carriers-decision-leave-long-term-care-insurance-market-1.