Summary
Uses six waves of the National Long-Term Care Survey (NLTCS, 1982–2004) and a continuous-time multi-state Markov model to produce integrated estimates of chronic disability among U.S. Medicare beneficiaries aged 65 and older. The paper's defining contribution is decomposing aggregate disability trends into four analytically distinct components — incidence (rate of entering disability), prevalence (cross-sectional stock), duration (expected time in each disability state), and intensity (severity mix within the disabled population) — rather than tracking prevalence alone. Chronic disability prevalence among the elderly declined substantially over 1982–2004, and the decomposition reveals that this reflected simultaneous improvements in multiple components, with implications for projecting long-term care demand and cost.
Key Claims
- Disability definition and state space. Chronic disability = needing help from another person with ≥1 activities of daily living (ADL) or ≥1 instrumental activities of daily living (IADL) for ≥90 consecutive days. The multi-state model covers seven states: non-disabled (community), IADL-only (community), 1–2 ADL (community), 3–4 ADL (community), 5–6 ADL (community), institutionalized, and dead.
- Prevalence decline confirmed and extended. Disability prevalence among U.S. elderly (65+) declined substantially from 1982 to 2004, extending the Manton, Corder, and Stallard (1997) and Manton and Gu (2001) findings across six survey waves and providing the most comprehensive pre-2010 NLTCS-based evidence for age-specific disability rate compression.
- Four-way decomposition is the paper's methodological center. Aggregate prevalence change = change in incidence + change in duration + change in intensity + interaction terms. This decomposition is more informative than prevalence alone because it identifies which mechanism drove the trend — falling incidence (fewer people become disabled), shorter duration (disability episodes end sooner through recovery or death), or shifting intensity (disabled population shifts toward less severe states).
- Both incidence and duration contributed to the prevalence decline. Fewer people entered chronic disability over the period, and those who did spent less time in higher-severity states — consistent with mild compression of morbidity rather than pure expansion. Intensity also shifted toward lighter disability states, compounding the prevalence effect.
- Disability-free life expectancy (DFLE) estimates. Multi-state life table calculations give expected years in each state from age 65, directly producing DFLE and expected years-with-disability by severity tier. These are the actuarial inputs to health-adjusted life expectancy (HALE)-style calculations for the elderly population.
- Formal and informal care costs by disability state. Annual costs are estimated separately for: formal care (Medicare, Medicaid, private pay, home health agencies) and informal care (unpaid family caregiving). The cost-by-state mapping translates disability incidence and duration estimates directly into projected long-term care expenditures.
- Institutionalization is a distinct state, not just severe disability. Nursing home residents have mortality and transition dynamics that differ structurally from community-dwelling disabled individuals at equivalent ADL counts; conflating them into a single "severe disability" category misspecifies the model.
Concepts Introduced or Extended
- Morbidity-Mortality Distinction — provides the NLTCS-based empirical evidence for age-specific disability compression among U.S. elderly 1982–2004; the four-way decomposition (incidence/prevalence/duration/intensity) is the methodological framework for distinguishing compression from expansion
- Healthy Life Expectancy — multi-state life table method for DFLE at age 65; NLTCS-based counterpart to the Global Burden of Disease (GBD) Sullivan-method HALE
- Select and Ultimate Mortality Tables — methodological parallel: Stallard's disability-state-specific mortality rates use the same multi-state Markov architecture as the Office of the Chief Actuary's (OCACT) Disability Insurance (DI) select-and-ultimate tables
Entities Mentioned
Quotes
(Source not in raw/ — page written from training knowledge.)
My Take
Stallard (2011) is the actuarial-demography benchmark for chronic disability measurement among the U.S. elderly, and it sits at a productive tension with the GBD-based evidence on morbidity expansion. The tension is partially illusory: Stallard documents declining age-specific disability rates among the 65+ population (fewer 75-year-olds are disabled in 2004 than in 1982), while GBD 2010/2021 documents rising total years lived with disability (YLD) burden nationally. Both can be true simultaneously because population aging mechanically increases the number of disabled people even as age-specific rates fall. Stallard's compression finding is also temporally bounded: the 2004 endpoint predates the post-2010 health stagnation documented by Woolf and Schoomaker (2019) and GBD 2021 — it is an open empirical question whether the compression trend reversed in the 2010s, particularly given the opioid epidemic's reach into the 50–64 age group that will soon enter the 65+ NLTCS-relevant population.
The four-way decomposition is the paper's lasting methodological contribution. Most disability policy debates implicitly focus on prevalence (how many people are disabled) when the policy-relevant questions are incidence (are people becoming disabled at lower rates?), duration (are episodes shorter?), and intensity (is severity declining?). These can diverge: a policy that reduces incidence but extends duration of surviving episodes might leave prevalence unchanged while representing genuine health improvement. Stallard's framework forces this distinction, and it maps directly onto the wiki's core interest in separating mortality improvement from functional improvement in the DI population.