Mortality (the probability of dying) and morbidity (the prevalence of disease or functional limitation) are related but conceptually distinct phenomena that can move in opposite directions. A third concept — ability to work — is distinct from both. Conflating any two of the three produces systematic policy errors.
The concrete manifestation of morbidity expansion in aging populations is Multimorbidity — the simultaneous co-occurrence of 2+ chronic diseases in the same individual. Marengoni et al. (2011) synthesized 41 studies to show that – of persons meet the multimorbidity threshold, making it the modal health state of older adults. The key epidemiological finding that connects multimorbidity to the morbidity-mortality distinction is the controversial mortality result: disease count alone does not reliably predict mortality — disability status mediates the relationship. Landi et al. (2010) found multimorbidity predicted 4-year mortality only when combined with disability. This mirrors the broader morbidity-mortality distinction: survival can improve (fewer multimorbid people die) while functional capacity stagnates (more people live in a multimorbid, disabled state). See Multimorbidity for the full evidence synthesis.
The U.S. Burden of Disease Collaborators (2013) provide the strongest cross-national empirical test of the compression vs. expansion hypotheses for the U.S., using the GBD 2010 framework across diseases, sequelae, and Organisation for Economic Co-operation and Development (OECD) comparators.
Core finding: expansion, not compression. Age-specific Years Lived with Disability (YLD) rates in the U.S. were essentially flat 1990–2010, while years of life lost (YLL) rates (premature mortality) fell substantially. Because morbidity held constant while mortality improved, YLDs grew from to of total U.S. Disability-Adjusted Life Years (DALYs) over the period. The paper's conclusion: "morbidity and chronic disability now account for nearly half of the health burden in the United States."
Top disabling conditions unchanged. The same eight conditions topped the YLD list in both 1990 and 2010: low back pain (#1), major depressive disorder (MDD) (#2), other musculoskeletal (#3), neck pain (#4), anxiety disorders (#5), chronic obstructive pulmonary disease (COPD) (#6), drug use disorders (#7), diabetes (#8). The stability of this ranking over 20 years is striking — the leading sources of disability did not respond to the same forces that reduced mortality.
GBD disability ≠ Social Security Administration (SSA) disability. This distinction is critical for this wiki. GBD disability weights (0–1 scale) capture any health loss across sequelae. SSA disability requires inability to engage in substantial gainful activity after considering age, education, and past work experience. Musculoskeletal pain and MDD are leading GBD disabling conditions and are also common DI/Supplemental Security Income (SSI) qualifying diagnoses — but the overlap is conceptual, not definitional. A person can have high GBD disability weight while retaining SSA-defined work capacity, and vice versa.
OECD rank decline. U.S. life expectancy (LE) rank fell th th and healthy life expectancy (HALE) rank fell th th among OECD nations 1990–2010, meaning the U.S. improved more slowly than peers in both total longevity and disability-free longevity.
Social determinants excluded. The GBD risk-factor analysis focused on behavioral, environmental, and metabolic risks and explicitly excluded income, education, and inequality because consistent effect-size estimates per specific disease could not be established under the required evidentiary criteria. The paper notes this exclusion "should not be taken as implying that they are less important."
GBD 2021 (US Burden of Disease Collaborators 2024) extends the 2013 analysis from national to state-level estimates and confirms that morbidity expansion continued — and in some states worsened dramatically — over 1990–2021.
US states surpass every country in YLD rates. In 2021, West Virginia, Kentucky, Oklahoma, Pennsylvania, New Mexico, Ohio, Tennessee, and Arizona had higher age-standardised YLD rates than any of the 204 countries in the GBD dataset. Afghanistan ranked th globally behind these states. The US as a whole ranked th globally (only countries had higher national YLD rates: Afghanistan, Lesotho, Liberia, Mozambique, South Africa, Central African Republic) — a far worse relative position than GBD 2013 had implied for the national average.
Within-state YLD extremes. Hawaii had the lowest age-standardised YLD rate ( per ); West Virginia had the highest ( per ). The WV/Hawaii YLD gap of per represents the full spectrum of subnational variation within a single country.
Drug use disorders now #3 YLD cause. Drug use disorder YLDs rose (–) 1990–2021 and became the #3 cause of age-standardised YLDs nationally by 2021, behind only low back pain (#1) and depressive disorders (#2). This stands alongside the mortality increase from the same conditions — rising YLDs and rising mortality simultaneously indicate incidence growth, not merely better survival with a fixed burden.
Depressive disorders . Depressive disorder YLD rate rose (–) 1990–2021. For females, mental health disorders had the highest age-standardised YLD rate. The co-dominance of musculoskeletal and mental health conditions in both GBD YLD rankings and SSA DI/SSI qualifying diagnoses — observed in 2013 and still true in 2021 — confirms the persistent overlap between measured disability burden and the DI allowance pattern.
HALE implication. US HALE ranked nd/nd (male/female) globally in 1990 but fell to th/th by 2021 — a larger decline than LE rank alone would suggest, because morbidity worsened faster than mortality. West Virginia HALE of years ranked st/th globally, meaning a WV male can expect fewer healthy life-years than the average male in other countries. This is the most concentrated measure of morbidity expansion's geographic severity within the US.
As chronic diseases replace infectious diseases as the primary cause of death, the relationship between mortality and morbidity inverts. If the probability of dying from a condition falls faster than the probability of contracting it, more people live longer with the disease — increasing chronic disease prevalence even as mortality falls.
Formally: with a constant recovery rate, if fatality risk declines more than incidence risk, prevalence rises. A population can simultaneously achieve rising life expectancy and a rising burden of chronic illness.
| Hypothesis | What happens to disability-free life expectancy |
|---|---|
| Morbidity compression | Grows faster than total LE — people die later and healthier |
| Morbidity expansion | Grows slower than total LE — people die later but spend more years sick |
| Dynamic equilibrium | Tracks total LE — the disabled fraction of life stays constant |
Evidence is mixed and depends on the population, period, disability measure, and methodology. U.S. data (Crimmins, Zhang, and Saito 2016) over 1970–2010 suggest dynamic equilibrium with some compression at ages 65+, but little evidence of health improvement during working ages (20–64). Critically, the compression finding is measure-dependent: Crimmins and Beltrán-Sánchez (2010) show that when morbidity is defined as disease prevalence or mobility functioning (rather than activities of daily living (ADL) disability), National Health Interview Survey (NHIS) data for 1998–2006 show expansion — disease prevalence increased across virtually all age groups, mobility functioning deteriorated, and years lived with disease and disability both rose under the Sullivan method. The ADL-disability compression and disease-expansion findings are compatible (disease spreads but is managed before it becomes observable ADL disability), but they carry opposite implications for healthcare costs. See Compression of Morbidity.
The strongest evidence for compression comes from two complementary sources using different elderly survey data. Stallard (2011) uses six waves of the National Long-Term Care Survey (NLTCS, 1982–2004) in a continuous-time multi-state Markov model to decompose aggregate disability trends into four analytically distinct components: incidence (rate of entering disability), duration (expected time in each disability state), intensity (severity mix within the disabled population), and prevalence as the integrated outcome. The 1982–2004 decline in elderly disability prevalence reflected simultaneous improvements in all three upstream components — fewer people entered chronic disability, episodes were shorter, and the disabled population shifted toward less severe states. Crucially, Stallard's four-way decomposition is more informative than prevalence alone: incidence compression (fewer people becoming disabled) and duration compression (shorter or less severe episodes) can diverge from prevalence, and Stallard documents that both occurred.
Cutler, Ghosh, and Landrum (2013) provide the foundational Medicare Current Beneficiary Survey (MCBS)-based test using observations from 1991–2009 linked to death records. Main result at age 65, 1992–2005: LE years, disability-free life expectancy (DFLE) years, disabled LE years — strong compression confirmed. An Oaxaca decomposition attributes of the disability rate decline to cardiovascular disease (CVD) ( pp) and vision problems ( pp), predominantly through conditions becoming less disabling rather than less prevalent (disease-disability decoupling). Disability is concentrated in the final year of life: are disabled within 12 months of death; compression occurred among those more than months from death, where disability rates fell substantially. See Compression of Morbidity.
Chernew et al. (2016) provide a complementary MCBS-based estimate extending the same data through 2008: total LE yr, DFLE yr, disabled LE yr. An Oaxaca decomposition (Table 3) shows the single largest contributor is the time-until-death effect ( pp): compression occurred among individuals months from death, whose disability rate fell pp/year; the end-of-life period ( months from death, disabled) was essentially unchanged. Disease contributions: CVD pp (mainly disease-disability decoupling, not prevalence reduction) and vision pp (entirely prevalence-driven). Diabetes and central nervous system (CNS) conditions partially offset these gains. This evidence is confined to age 65+ and does not extend to the working-age (20–64) population relevant to DI. The same paper explicitly notes that Case and Deaton (2015) and Chetty et al. (2016) document the opposite trend — worsening morbidity and mortality — in near-elderly and lower-socioeconomic status (SES) groups.
A precursor signal for working-age morbidity expansion appeared in Olshansky et al. (2005): rising disability rates and declining fitness levels at younger (working) ages were documented as early evidence that the morbidity-mortality divergence was no longer confined to the elderly. Disability rates had risen and fitness levels declined in working-age Americans even while cardiovascular treatment was extending old-age survival — consistent with morbidity expansion beginning to operate at working ages years before GBD 2010 would confirm it nationally. The paper attributed this signal to the obesity epidemic; in retrospect, deaths of despair (which began rising in the mid-1990s) contributed as well. See S. Jay Olshansky.
Obesity is a specific morbidity pathway linking the working-age morbidity expansion to DI enrollment. Burkhauser and Cawley (2004) provide the causal identification: using biological relatives' weight as an instrument for own body mass index (BMI) (Panel Study of Income Dynamics (PSID) and National Longitudinal Survey of Youth 1979 (NLSY79), ages –), they find that obesity raises disability income receipt by – pp (two-stage least squares (2SLS)), with ordinary least squares (OLS) estimates downward-biased by – due to BMI measurement error and reverse causation. This is a concrete illustration of the morbidity--mortality distinction: obesity generates substantial morbidity (musculoskeletal limitations, metabolic disease, work restrictions) that raises DI receipt without necessarily raising short-run mortality for working-age adults.
The obesity doubling between the early 1980s and 2002 — from to of U.S. adults — parallels the doubling of DI rolls over the same period. The causal mechanism runs through musculoskeletal conditions (low back pain, osteoarthritis) and metabolic conditions (diabetes, cardiovascular disease precursors), which are the leading morbidity categories in both the GBD YLD rankings and SSA allowance statistics. This connection between obesity-driven morbidity expansion and DI enrollment growth is one mechanism behind the composition shift from high-mortality (circulatory, cancer) toward lower-mortality (musculoskeletal) conditions documented by Liebman (2015). See DI Growth Decomposition.
These are three separate concepts that interact in complex, non-linear ways:
The shift in DI primary diagnoses in Germany over 1983–2008 is the clearest documented time-series case of the morbidity-mortality disconnect in the DI context. Over the same period that German life expectancy improved by ~7–8 years at age 60–65:
Mental illness and musculoskeletal conditions together account for ~50% of men and ~60% of women entering German DI — conditions that are non-lethal, unrelated to the cardiovascular and cancer causes that drive mortality trends. This diagnosis shift means aggregate mortality data provide almost no information about the disease burden underlying DI enrollment growth; the two phenomena are tracking largely orthogonal conditions. This is direct empirical evidence for the morbidity-mortality distinction operating in the DI institutional context, at a population level, over 25 years. See DI Uptake and Health Decoupling.
The shift in DI Beneficiary Mortality improvement from the 1990s onward is partly explained by a compositional shift toward musculoskeletal and mental impairments (less lethal, longer-duration conditions) and away from cancer and cardiovascular disease (more lethal). This lowers measured mortality without necessarily improving the beneficiaries' functional capacity or work ability.
Liebman (2015) provides the sharpest quantification of this shift: if musculoskeletal and mental health incidence had remained at 1985 levels (with all other conditions following their actual paths), the 2007 DI beneficiary ratio would have been approximately lower than it was. Because these conditions carry much lower mortality than the circulatory and cancer conditions they displaced, the incidence shift mechanically drives down measured beneficiary mortality even if no individual beneficiary's health improves. See DI Growth Decomposition.
Rutledge et al. (2018) found that the underlying health of DI and SSI claimants remained essentially unchanged from 1989–2013 — consistent with mortality improvement driven by composition, not by genuine health gains in the applicant pool.
Raut (2017) provides the most direct empirical test of the mortality--work-capacity claim for DI beneficiaries. Using a competing-risks model on 1981–2000 Continuous Work History Sample (CWHS) entrants, Raut compares death exit rates and recovery exit rates between 1981–1985 and 1991–1995 entrant cohorts. The result:
This is the sharpest available evidence that DI mortality improvement does not reflect improvement in work capacity. If beneficiaries were becoming genuinely healthier — able to return to work — recovery exit rates would have risen alongside (or ahead of) falling death exit rates. Instead, the two trends diverged: the same medical advances that reduced in-program mortality did not generate increased rates of return to the labor market.
The musculoskeletal/mental finding reinforces this: these impairment types have the lowest death exit rates ( and respectively for young entrants over 9 years) — the compositional shift toward these conditions drives measured mortality down — yet their recovery rates, while higher than other conditions, remain low in absolute terms ( and for ages – over 9 years). Two-thirds of young musculoskeletal entrants are neither recovering nor dying within 9 years; they remain on the rolls. See DI Growth Decomposition for the duration-accumulation mechanism.
A recurring methodological concern with cross-national morbidity comparisons is that self-reported illness rates may reflect differences in diagnosis, physician communication, or reporting norms rather than genuine differences in disease burden. Banks, Marmot, Oldfield, and Smith (2006) resolve this for the US–England comparison by pairing self-reported chronic disease rates (the Health and Retirement Study [HRS] and the English Longitudinal Study of Ageing [ELSA]) with matched biological markers (the National Health and Nutrition Examination Survey [NHANES] and Health Survey for England). The biological markers — glycated hemoglobin (HbA1c), C-reactive protein (CRP), fibrinogen, and high-density lipoprotein cholesterol (HDL-C) — show exactly the same country and SES patterns as the self-reports: Americans have higher inflammatory burden (CRP 20% higher, fibrinogen 17% higher) and worse metabolic profiles (HDL-C 14% lower) at every SES level. This finding validates self-reported morbidity as a genuine indicator of underlying disease burden and not a reporting artifact. It also confirms that the SES gradient in chronic disease prevalence — steeper in the US than in England — is real, not a differential propensity to report illness. See American Health Disadvantage.
A methodological threat to measuring true health trajectories in longitudinal panels is selective mortality: individuals in the worst health are most likely to die and exit the panel, so the surviving observed sample appears healthier than the true conditional trajectory of the original cohort. This is a distinct mechanism from the morbidity-mortality-work-capacity triangle, but it shapes how any of the three phenomena can be measured from data.
Heiss, Börsch-Supan, Hurd, and Wise (2007) estimate a latent-health model on all four Health and Retirement Study (HRS) cohorts (; 1992–2002) using an Ornstein-Uhlenbeck heterogeneity process to recover true health dynamics while accounting for:
Key results on selective mortality:
Implication for the morbidity-mortality distinction: Declining observed disability prevalence in an aging longitudinal panel may reflect selective mortality — the sickest exiting via death — rather than genuine recovery or health improvement. Any study that interprets improving average health in a surviving cohort as evidence of morbidity compression must first rule out this mechanism. The Heiss et al. finding that -year disability persistence is (not the naively lower Markov estimate) shows that without correcting for selective mortality, morbidity appears to compress more than it actually does.
Heiss (2011) further quantifies the decomposition of the cross-sectional age profile of poor/fair SRH in the HRS: at ages – the direct aging effect is pp but survivorship selection is pp, yielding a cross-sectional increase of only pp. At ages , survivorship selection ( pp) exceeds the direct aging effect ( pp), making health appear to improve cross-sectionally. Any study claiming morbidity compression at the oldest ages must first rule out selective mortality as the mechanical explanation. See Selective Mortality.
A potential objection to using measured consumption declines as welfare evidence is that disabled individuals may substitute home production for market goods — replacing purchased food with home-cooked food, for instance — so that market consumption falls even as true consumption is stable or improving. Aguiar and Hurst (2005) documented exactly this pattern for retirees (consumption falls on retirement but diet quality improves as home food production rises), and its absence for disability is not self-evident.
Meyer and Mok (2013) directly test this via PSID time-use data for Chronic-Severe disability. The result refutes the substitution hypothesis decisively:
This finding is consequential for the three-way distinction at the core of this concept page. Disability reduces well-being along all three dimensions simultaneously: it raises morbidity (health deterioration), does not raise measured mortality in the short run for Chronic-Severe (unlike acute illness), and does reduce consumption-based well-being — confirming that work capacity and financial well-being decline together, not merely as a measurement artifact.
The no-substitution finding also validates Meyer-Mok's use of food consumption as a welfare proxy and the Chetty (2006) optimal benefit calibration with . See Nonhealth Risk and DI Insurance Value.
Weaver (2020) provides a complementary argument from a different direction. If declining DI beneficiary mortality reflected a loosening disability standard (i.e., healthier people entering the rolls), then denied applicants — those who failed to qualify — should look substantially healthier than approved ones. Instead, the million denied applicants have health profiles nearly identical to approved beneficiaries: fair or poor health vs. ; difficulty standing vs. ; frequent depression/anxiety vs. . The standard is screening a pool that is uniformly severely impaired — not sorting a mixed-health pool into healthy-denied and sick-approved. See DI Denied Population.