Chernew et al. 2016 — Understanding the Improvement in Disability Free Life Expectancy in the U.S. Elderly Population

Summary

Using the Medicare Current Beneficiary Survey (MCBS, 1991–2009, $N = 251{,}872$ person-year observations) merged with National Center for Health Statistics (NCHS) mortality data, Chernew et al. document that disability-free life expectancy (DFLE) at age 65 rose $1.8$ years from 1992 to 2008 while disabled life expectancy fell $0.5$ years — total life expectancy (LE) rose only $1.3$ years, so healthy years grew faster than total years (strong compression). An Oaxaca decomposition across 15 disease categories identifies time-until-death effects ($-4.9$ percentage points (pp)), cardiovascular disease (CVD) ($-2.5$ pp), and vision problems ($-1.7$ pp) as the three largest contributors to the $7.4$ percentage-point disability rate decline; the CVD and vision findings together account for $63\%$ of the DFLE gain. The paper then uses the validated IMPACT cardiovascular mortality model and a within-person ordered probit panel design for cataract surgery to attribute a portion of the improvement to specific medical advances, though the causal claims are described as speculative.

Key Claims

• Main result: DFLE at age 65 increased $1.8$ years from 1992 to 2008; disabled LE fell $0.5$ years; total LE rose $1.3$ years — strong compression confirmed. Overall disability rate fell from $\approx 49.5\%$ to $\approx 42.5\%$ ($-7.4$ pp).
• Oaxaca decomposition (Table 3): The three largest contributors to the $7.4$ pp disability decline: time-until-death effects ($-4.9$ pp, the largest single component), cardiovascular disease ($-2.5$ pp), and vision problems ($-1.7$ pp). Partially offset by diabetes ($+0.9$ pp) and central nervous system (CNS) conditions ($+0.5$ pp, predominantly Alzheimer's disease).
• CVD mechanism: The $-2.5$ pp CVD contribution is $68\%$ coefficient effect (conditions becoming less disabling per prevalence unit) and $32\%$ prevalence effect. Ischemic heart disease (IHD) prevalence fell $29.5\% \to 22.9\%$; its disability coefficient fell from $0.11$ to $0.06$. This is disease-disability decoupling.
• Vision mechanism: The $-1.7$ pp vision contribution is entirely a prevalence effect — the disability coefficient for vision problems was unchanged ($0.13$ in both periods), but prevalence fell from $38.4\%$ to $24.7\%$. Fewer people with vision problems drove this improvement, not better management of existing cases.
• Time-until-death framing: Disability is heavily concentrated near death. People $>36$ months from death saw disability decline $-0.46$ pp/year (contributing $-4.9$ pp over the full period). People $<12$ months from death remained $\approx 80\%$ disabled throughout — unchanged from 1991–1994 to 2006–2009. Compression occurred among those with time to spare.
• IMPACT model for CVD (Table 5): $228{,}910$ fewer IHD deaths in 2009 than expected under 1992 rates. $51\%$ attributable to treatment: statins for secondary prevention ($\approx 20$k lives), ACE inhibitors ($\approx 13$k), beta-blockers ($\approx 25$k), statins for lipid reduction ($\approx 18$k), antihypertensives ($\approx 9$k). $44\%$ attributable to risk factor changes: lower systolic blood pressure ($22\%$ of total), lower total cholesterol ($30\%$), though rising body mass index (BMI) and diabetes prevalence offset some gains ($-13\%$). CVD treatment translates to $+0.26$ DFLE years at age 65 ($\approx 15\%$ of total DFLE gain).
• Cataract surgery panel regression (Table 7): Vision problem prevalence fell $40\% \to 25\%$ over the period; cataract surgery receipt rose from $20\% \to 33\%$ of beneficiaries. Ordered probit with lagged health status: cataract surgery receipt (lagged one period) significantly reduces subsequent vision problems (coefficient $-0.092$, robust standard error (SE) $0.024$) and disability ($-0.126$, robust SE $0.021$), both $p < 0.001$. Increased cataract surgery explains $\approx 27\%$ of the vision-related disability decline, translating to $\approx 0.08$ DFLE years ($\approx 5\%$ of total).
• Disease-level DFLE contributions (Figure 7): Cardiovascular disease $+0.98$ yr DFLE (IHD $+0.73$ yr + stroke $+0.25$ yr); vision $+0.28$ yr; cancer group $+0.23$ yr; unexplained (primarily the $>36$-months-from-death structural shift) $+0.65$ yr. CNS conditions (Alzheimer's, Parkinson's) contributed negatively to DFLE; diabetes added $+0.2$ yr to disabled LE.
• Exceptions: Alzheimer's disease and Parkinson's disease showed increasing disability coefficients (Alzheimer's: $0.25 \to 0.28$; Parkinson's: $0.18 \to 0.24$) — worsening disablement per case. These are exceptions to the overall compression trend.
• Data gap: The MCBS does not capture mental illness or musculoskeletal conditions (back pain), which Global Burden of Disease (GBD) 2010 identifies as the #1 and #2 sources of years lived with disability (YLDs) nationally. The compression finding applies to the conditions the MCBS measures.

Concepts Introduced or Extended

• Compression of Morbidity — provides the most complete disease-level decomposition of U.S. elderly compression, 1992–2008; extends Cutler, Ghosh, and Landrum (2013) (1992–2005) by adding years, disease-level DFLE attribution, treatment mechanism analysis
• Morbidity-Mortality Distinction — shows that morbidity and mortality can improve simultaneously through distinct mechanisms; confirms compression at 65+ while explicitly citing Case and Deaton (2015) and Chetty et al. (2016) for the opposite near-elderly trend
• Healthy Life Expectancy — DFLE as an activities of daily living / instrumental activities of daily living (ADL/IADL) binary indicator alternative to GBD health-adjusted life expectancy (HALE); methodologically distinct but directionally consistent; confined to age 65+

Entities Mentioned

• Michael Chernew
• David Cutler
• Kaushik Ghosh
• Mary Beth Landrum

Quotes

"Years of healthy life expectancy at age 65 increased by 1.8 years over that time period, while disabled life expectancy fell by 0.5 years."

"The largest improvements in healthy life expectancy come from reduced incidence and improved functioning for those with cardiovascular disease and vision problems. Together, these conditions account for 63 percent of the improvement in disability-free life expectancy."

"There are some conditions that our data do not ask about — mental illness and musculoskeletal issues (back pain, for example) — that have been shown to be major contributors to disability in other studies." [citing US Burden of Disease Collaborators 2013]

"Recent work has documented a slowdown or even reversal of improvements in morbidity and mortality in more recent periods, particularly in the near elderly." [citing Case and Deaton 2015, Chetty et al. 2016, Martin et al. 2010, Chen and Sloan 2015]

My Take

The disease-level DFLE decomposition is the paper's real contribution — quantifying which conditions drive compression, how much, and by which mechanism (prevalence vs. disabling-impact) is more informative than the headline DFLE number. The CVD attribution is well-supported via the validated IMPACT model; the cataract surgery finding is novel and clinically plausible, though endogeneity (healthier patients may elect surgery earlier) cannot be fully ruled out. The authors acknowledge this.

A critical observation from Table 3: the time-until-death component ($-4.9$ pp) is larger than CVD ($-2.5$ pp) and vision ($-1.7$ pp) combined. Much of the compression is structural — people dying later means more of the population is in the "far from death, likely healthy" segment — rather than disease-specific treatment gains. This limits how much of the documented compression can be attributed to identifiable medical interventions.

Critical limitation for this wiki: the paper is restricted to age 65+, uses MCBS (ADL/IADL indicators only, no mental health or musculoskeletal conditions), and explicitly acknowledges opposite trends at younger ages. The Disability Insurance (DI)-relevant population is working-age (20–64), where Case-Deaton (2015) and Chetty et al. (2016) document declining morbidity and mortality trends. This paper supports compression for the elderly but cannot be read as evidence that DI applicants — working-age, with disproportionate musculoskeletal and mental health impairments — are becoming functionally healthier.