Mapping Benefit Triggers to Model States

Why Contract Triggers Must Be Mapped to States

A long-term care model describes how a person’s condition changes over time. A contract, however, pays according to its own wording. Receiving care does not necessarily mean that a policyholder qualifies for benefits, and a change in care status does not always produce the same change in contractual eligibility. The model therefore needs a precise mapping from its states to the contract’s trigger.

The key question is not simply, “Which care level is the person in?” It is, “Given the model state at this time, does the contract’s condition hold, and what benefit category follows?” A usable mapping makes that answer unambiguous for every state.

Definition: A model state is a category containing the information used to describe the policyholder at a time. A benefit trigger is the contractual condition that must hold for a benefit to be payable. A state is qualifying if the trigger holds whenever the policyholder is in that state. A trigger-to-state mapping specifies which states qualify and, if relevant, which benefit category applies in each state.

A care-level state and a qualifying state need not be identical. For example, two people may receive the same general level of support, while only one satisfies a contract’s additional certification condition. If the model records only care level, it cannot distinguish their contractual outcomes. It must also represent the relevant certification status, or otherwise split the care-level state.

Representing Qualification and Cessation

For a model with states $1,\ldots,n$, define an eligibility indicator $e_j$ for each state $j$: $e_j=1$ if the trigger is met in that state and $e_j=0$ otherwise. If the contract also has different payable categories, define a benefit amount $b_j$ for each state. A zero amount can represent a state in which nothing is payable.

e_j = \begin{cases} 1, & \text{if the contractual trigger holds in state } j,\\ 0, & \text{otherwise,} \end{cases} \qquad b_j = \begin{cases} \text{applicable amount in state }j, & e_j=1,\\ 0, & e_j=0. \end{cases}

The state transitions then describe how qualification starts or ends. A transition from a nonqualifying state to a qualifying state represents entry into eligibility. A transition in the opposite direction represents cessation, provided the contract’s rule is based on the current condition. These descriptions concern eligibility status; whether a payment is made immediately or only after another contractual requirement is satisfied is a separate question.

The mapping should follow the contract’s condition, not an informal label such as “in care.” In particular, list the conditions that must hold together, the conditions that are alternatives, and any condition that causes eligibility to stop. A condition that is phrased as “and” is not equivalent to one phrased as “or.”

Worked Example: Separating Care Receipt From Qualification

A hypothetical policy pays $100 per day only while the policyholder both receives covered residential support and has an active certification from the contract’s assessor. Use four states: $N$ (no covered support), $R$ (covered support but no active certification), $Q$ (covered support and active certification), and $D$ (death). Map the trigger to states and identify how qualification can begin and cease.

Assumptions and model data. The effective annual interest rate is $i=0.04$, but no present value is being calculated, so interest does not affect the mapping. Assume a continuous-time Markov chain with state order $N,R,Q,D$, initial state $N$ with probability 1, and $D$ absorbing. For illustration, the transition intensities per year are $N\to R:0.10$, $N\to D:0.02$, $R\to N:0.25$, $R\to Q:0.20$, $R\to D:0.04$, $Q\to R:0.15$, and $Q\to D:0.08$. Death is represented by entry to $D$; no separate mortality table is needed for this state-mapping exercise.

Step 1: Translate the wording into a logical condition. Both covered support and active certification are required. State $R$ has support but lacks certification, so it fails the trigger. State $Q$ has both conditions, so it satisfies the trigger. State $N$ has no covered support, and $D$ is not a payable living state.

Step 2: Assign eligibility and benefit amounts. In state order $N,R,Q,D$, the eligibility indicators are $(0,0,1,0)$, and the daily benefit amounts are $(\$0,\$0,\$100,\$0)$. The transition $R\to Q$ starts qualification. The transition $Q\to R$ ends qualification because certification is no longer active, even though covered support continues.

Step 3: Check that the transitions are consistent. The total exit intensity from $N$ is $0.10+0.02=0.12$ per year. From $R$, it is $0.25+0.20+0.04=0.49$; from $Q$, it is $0.15+0.08=0.23$. These sums confirm that the listed exits are accounted for in each living state. The state mapping itself does not depend on the sizes of those intensities.

Common mistake avoided: Paying in both $R$ and $Q$ would treat receipt of support as sufficient. That contradicts the stated requirement for active certification. Only $Q$ qualifies.

When a State Must Be Split

A model state can be retained as a single state only when it gives enough information to determine the trigger and to represent the future transitions relevant to the model. If a single care category includes both qualifying and nonqualifying cases, the category generally needs to be split. Otherwise the model cannot assign a single eligibility indicator to that state.

Even when two states have the same current eligibility and benefit, they should not automatically be merged. If they have different possible transitions, or different transition intensities, combining them may discard information needed to describe future state probabilities. Conversely, states with different internal descriptions may be combined for the benefit mapping if they have the same contractual outcome, but only if the model remains adequate for its other purposes.

This is a modeling decision, not a change to the contract. The contract defines the trigger; the model’s state definitions must preserve enough information to apply it.

Worked Example: Mapping a Tiered Trigger With an Additional Condition

A hypothetical contract pays $70 per day in a lower qualifying category and $130 per day in a higher qualifying category. The trigger is met in the lower care tier only if an assessor’s current severity score is at least 2. In the higher tier, the trigger is met only if certification is active. Consider states $L_0$ (lower tier, score below 2), $L_1$ (lower tier, score at least 2), $H_0$ (higher tier, no active certification), $H_1$ (higher tier, active certification), and $D$ (death). Determine the eligibility and benefit mapping.

Assumptions and model data. This is a hypothetical continuous-time model with initial state $L_0$ with probability 1; death is an absorbing state, and the death intensity from each living state is $0.03$ per year. The effective annual interest rate is $i=0.04$; it is stated for completeness but is not used because this example does not discount payments. Nondeath transition intensities are not needed to classify the states.

Step 1: Test the lower-tier condition. In $L_0$, the score is below 2, so the lower-tier trigger fails. In $L_1$, the score is at least 2, so the trigger holds and the lower benefit applies.

Step 2: Test the higher-tier condition. In $H_0$, certification is inactive, so the trigger fails despite the higher care tier. In $H_1$, certification is active, so the trigger holds and the higher benefit applies.

Step 3: Record the mapping. In state order $L_0,L_1,H_0,H_1,D$, the eligibility indicators are $(0,1,0,1,0)$, and the daily amounts are $(\$0,\$70,\$0,\$130,\$0)$. A transition $L_0\to L_1$ would start eligibility; a transition $L_1\to L_0$ would stop it. Similarly, $H_0\to H_1$ starts eligibility, while $H_1\to H_0$ stops it. A move from $L_1$ to $H_1$ changes the applicable amount from $70 to $130, while a move from $L_1$ to $H_0$ changes the amount to zero because the higher-tier certification condition is not met.

Check. The qualifying states are exactly those satisfying their tier-specific condition: $L_1$ and $H_1$. The nonqualifying states $L_0$, $H_0$, and $D$ all have zero payable amount. The mapping does not infer eligibility merely from the care tier.

Common mistake avoided: Assigning a benefit to every higher-tier state would incorrectly include $H_0$. The tier alone does not satisfy the stated trigger.

Why the Markov State Must Contain Trigger-Relevant Information

A continuous-time Markov model is useful when the future distribution of states, conditional on the present state, does not require additional knowledge of the past. For a finite-state model with generator $Q$, the row vector of state probabilities $p(t)$ satisfies $p'(t)=p(t)Q$. A contractual eligibility indicator can be attached to each state when the state records the information needed to decide whether the trigger holds now.

The reason for this requirement can be seen directly. Suppose two policyholders are both labeled as being in the same model state, but one currently satisfies the trigger and the other does not. The state label cannot determine one eligibility indicator $e_j$ for both people. The state definition is therefore too coarse for the contract mapping and should be refined.

A second issue arises when the trigger depends on history, rather than only on the current condition. For instance, if a contract condition refers to how long a qualifying condition has persisted, two people with the same current care and certification status may have different eligibility because their histories differ. A model that keeps only the current condition cannot apply the rule consistently. The general remedy is to augment the state with the relevant history summary, such as a duration category, if the model is intended to represent that rule. The detailed treatment of any waiting or elimination period is a separate contractual calculation.

Write the trigger precisely.
List each required condition and identify whether the wording requires all conditions or allows alternatives.

Evaluate the trigger state by state.
Assign qualifying or nonqualifying status from the contractual facts represented in each state.

Refine states where necessary.
Split a state if its members can have different current eligibility or if combining them loses relevant future-transition information.

Assign the applicable category.
For each state, record the eligibility indicator and, where the contract distinguishes categories, the corresponding benefit amount or label.

Check changes in both directions.
Identify transitions that start qualification, end qualification, or change the applicable category.

Once the mapping is defined, state probabilities can be aggregated to describe the chance of qualifying at a particular time and the expected current payment rate. This is an application of the state mapping, not a present-value calculation.

P(\text{qualifying at }t)=\sum_{j=1}^{n}p_j(t)e_j, \qquad E[\text{current payment rate at }t]=\sum_{j=1}^{n}p_j(t)b_j.

These equations follow by conditioning on the state at time $t$. The mutually exclusive state probabilities sum over all possible states. In each state, eligibility is fixed by $e_j$, and the payment rate is fixed by $b_j$; multiplying each state’s value by its probability and adding gives the corresponding expectation. If the state definitions do not determine eligibility or payment, these sums cannot correctly represent the contract.

Worked Example: Aggregating Eligibility Across Model States

At a particular time, a hypothetical policyholder has the following state probabilities: $N$, no qualifying care, 0.45; $C_{1,0}$, lower care category without the required certification, 0.20; $C_{1,1}$, lower category with certification, 0.15; $C_2$, higher category meeting its trigger, 0.12; and $D$, death, 0.08. The contract pays $90 per day in $C_{1,1}$, $150 per day in $C_2$, and nothing elsewhere. Find the probability of current eligibility and the expected current daily payment rate.

Assumptions and data. The probabilities are a supplied distribution over the five mutually exclusive states at the stated time and sum to $0.45+0.20+0.15+0.12+0.08=1.00$. The effective annual interest rate is $i=0.04$, but no discounting is involved. Death is absorbing; the supplied probabilities already include the probability of being in $D$, so no separate mortality calculation is required.

Step 1: Identify qualifying states. The trigger is met in $C_{1,1}$ and $C_2$, and in no other state. Therefore the current eligibility probability is $0.15+0.12=0.27$.

Step 2: Multiply each payable amount by its state probability. The lower-category contribution is $0.15\times \$90=\$13.50$ per day. The higher-category contribution is $0.12\times \$150=\$18.00$ per day.

Step 3: Add the contributions. The expected current payment rate is $\$13.50+\$18.00=\$31.50$ per day.

Check. The eligibility probability $0.27$ is no greater than 1, and the expected payment rate lies between $0 and the maximum state payment of $150 per day. No payment is assigned to the nonqualifying states or to death.

Common mistake avoided: Multiplying the total probability of being in either care category by one benefit amount would ignore both the certification condition and the different amounts in the two qualifying states.

Common Mapping Errors

Equating care receipt with eligibility. A contract may impose an additional condition. Apply every part of the trigger before marking a state as qualifying.
Checking only entry into eligibility. Also identify how qualification ceases. A model that allows recovery or loss of certification must represent the resulting change in status.
Using a state that is too broad. If people within one state can have different trigger outcomes, the state does not support a consistent benefit mapping.
Forgetting that categories can change. A transition can leave eligibility intact but change the applicable benefit category, or it can move directly from payable to nonpayable status.
Ignoring history dependence. If a rule depends on duration or a prior event, current care status alone may not be sufficient to determine eligibility.

Key takeaway: Translate the contract’s trigger into a state-by-state rule. Every state should have a definite eligibility status and, when needed, a definite benefit category. Split states when the model does not retain enough information to make that assignment consistently.

Check Your Understanding

For each problem, map the stated trigger to the model states and check both qualification and cessation where relevant.

A hypothetical contract pays only when both a covered home-support service is being received and an assessor’s approval is current. The model states are $A$ (no service), $B$ (service, no current approval), $C$ (service and current approval), and $D$ (death). Which state qualifies? (Answer: $C$ only.)
A benefit is payable in states $L_1$ and $H_1$, but not $L_0$, $H_0$, or $D$. A transition $H_1\to H_0$ occurs when certification lapses. What happens to eligibility on that transition? (Answer: Eligibility ceases; the transition is from qualifying to nonqualifying.)
A model has one state called “moderate care,” but the contract requires both moderate care and active certification. Some people in that state have certification and some do not. What modeling change is required before assigning an eligibility indicator? (Answer: Split the state into at least certified and uncertified states, or otherwise record certification in the state.)
At a time point, the probabilities of states $N,Q_1,Q_2,D$ are $0.50,0.18,0.22,0.10$. Only $Q_1$ and $Q_2$ qualify. Find the probability of current eligibility. (Answer: $0.18+0.22=0.40$.)
A contract pays $80 per day in $Q_1$, $140 per day in $Q_2$, and zero elsewhere. At a given time, $P(Q_1)=0.25$, $P(Q_2)=0.10$, and all remaining probability is in nonpayable states. Find the expected current payment rate. (Answer: $0.25(\$80)+0.10(\$140)=\$20+\$14=\$34$ per day.)

What You'll Learn