With rising healthcare costs and an aging population, the demand for long-term services and supports is expected to grow significantly in the coming years.[1] Although sales of standalone long-term care (LTC) insurance have slowed, combination products (also called “combo” or “hybrid” products) that bundle life insurance or an annuity with LTC protection have gained popularity over the last decade.[2] As carriers’ and reinsurers’ balance sheets have expanded with combo LTC business, the potential benefits of enhancing actuarial modeling capabilities has grown as well.
Oliver Wyman regularly publishes an industry survey on actuarial software and modeling practices across the US insurance industry. Among survey participants that write combo LTC business, one standout finding is that combo LTC products rank among the most complex to model. This finding prompted a follow-up survey dedicated entirely to combo LTC modeling across pricing, valuation and projection use cases. With 16 participants representing over 90% of combo LTC business written, the survey provides comprehensive market coverage and insights on actuarial software, model architecture, assumptions and reserves.
This article summarizes some of the key insights from the survey.
Actuarial Software
AXIS, Prophet and MG-ALFA were cited as the most commonly used vendor-developed actuarial platforms for combo LTC modeling, with usage varying by use case. Custom solutions typically use a mix of R, Python and/or VBA and are more common for pricing compared to reserving and projection use cases.
As shown in Figure 1, more than half of the survey participants said they plan to make material upgrades to their existing models or build entirely new models in the near term.[3] The authors interpret this as evidence that companies may still be evaluating approaches that balance desired sophistication and capabilities with compatibility across their broader modeling ecosystems.
Figure 1
Model Status
Model Architecture and Assumptions
Over the years, companies have enhanced their standalone LTC modeling methodologies, migrating from more opaque techniques (e.g., claims cost methods) to first-principles approaches.[4] These models require more detailed assumptions but generally provide greater insight into the drivers of profitability by separately modeling the underlying components of LTC experience and allowing their individual and combined effects on financial results to be evaluated. The survey found that first-principles modeling for combo LTC products, which generally reflects the interaction between the life insurance component and the LTC benefits, is more common for pricing and projections than for statutory reserves.
Modeling combo LTC products can be complex because a policy can transition through numerous “states” (e.g., active with no prior disability, active with a prior disability, disabled). Stochastic simulation models (e.g., Monte Carlo) explicitly track the potential paths a policyholder may follow, often running many simulations until a desired level of convergence is achieved, whereas deterministic models produce probability-weighted outcomes. As shown in Figure 2, the survey found that deterministic models are the norm for combo LTC among participants; however, participants also indicated the use of stochastic approaches for pricing and occasionally for longer-horizon projection modeling.[5]
Figure 2
Model Type (Pricing)
Finally, while standalone LTC is often modeled using assumptions that vary by care settings (e.g., nursing home, assisted living, home care), the survey also showed that combo LTC products are typically modeled using a single blended aggregate care setting to keep the model more manageable (i.e., simpler assumptions and fewer modeled states).
Conservation of total mortality can be an important consideration for combo LTC modeling, primarily because life and annuity mortality experience is typically measured on a total life basis, while first-principles LTC models require separate mortality assumptions for the active and disabled populations. To maintain conservation of total mortality, most survey participants dynamically solve for active life mortality within their model using total and disabled life mortality inputs.
Additionally, the survey identified benefit recovery modeling as a challenge. Survey results showed that explicit state tracking of recoveries is uncommon in deterministic models due to the inherent complexity, although it is common in stochastic simulation models. As a simplification, deterministic models typically assume recoveries return to the active population with either full or averaged/prorated benefits.
US Statutory Reserves
Survey responses indicated that reserve modeling practices differ significantly by issue era—particularly between pre-PBR (principles-based reserves) and PBR-era business (pre-PBR business was generally issued before January 1, 2020). Survey participants indicated that they typically calculate the life and LTC acceleration components of pre-PBR statutory reserves separately, often using a blending mechanism to reduce reserve redundancy. In contrast, PBR-era statutory reserves are often determined as an integrated reserve that reflects the interaction between the life and LTC acceleration components, as illustrated in Figure 3.[6]
Figure 3
Life/LTC interaction (Statutory Reserves)
Combo LTC products can include various linked-benefit riders, such as inflation protection and extension of benefits riders. Survey responses indicated that these riders are typically modeled on a first-principles basis within the same model as the underlying life insurance component, particularly for PBR-era business.
Looking Ahead
Combo LTC modeling techniques and methodologies are still evolving across the industry. As combo LTC blocks grow and more policyholders go on claim, organizations may place greater importance on developing sound actuarial models to support product pricing, financial reporting and strategic decision-making. Investing in these modeling capabilities now can help position organizations to better understand their risks and capitalize on emerging product development opportunities.
This article is provided for informational and educational purposes only. Neither the Society of Actuaries nor the respective authors’ employers make any endorsement, representation or guarantee with regard to any content, and disclaim any liability in connection with the use or misuse of any information provided herein. This article should not be construed as professional or financial advice. Statements of fact and opinions expressed herein are those of the individual authors and are not necessarily those of the Society of Actuaries or the respective authors’ employers.
Dustin Plotkin, FSA, MAAA, ACIA, is a senior principal at Oliver Wyman. He can be reached at dustin.plotkin@oliverwyman.com.
Michael Rustemeyer, FSA, MAAA, is a senior manager at Oliver Wyman. He can be reached at michael.rustemeyer@oliverwyman.com.
Gabriel Alboukrek, FSA, is a senior manager at Oliver Wyman. He can be reached at gabriel.alboukrek@oliverwyman.com.
Endnotes
[1] “The State of Long-Term Care Insurance—2025,” Issue Brief, American Academy of Actuaries, February 2025, https://www.actuary.org/wp-content/uploads/2025/03/health-brief-2025-stateofltc.pdf.
[2] “Is Life Insurance the Answer to the Growing Long-Term Care Need in the U.S.?” LIMRA, August 28, 2025, https://www.limra.com/en/newsroom/industry-trends/2025/is-life-insurance-the-answer-to-the-growing-long-term-care-need-in-the-u.s/.
[3] Split of developed model versus not developed model is specifically for US statutory valuation models; count of carriers/reinsurers with planned refinements is based on all model use cases (i.e., pricing, valuation, projection).
[4] First-principles models deploy component-level assumptions, which typically include lapse, incidence, mortality (varying by active and disabled populations), claim recovery and benefit utilization.
[5] Count of survey participants in Figure 2 is for pricing use cases only.
[6] The count of survey participants in Figure 3 is based on statutory reserve model use cases only. Seven survey participants have PBR business and are non-PBR exempt.