Lichtenberg 2019 — The Impact of Access to Prescription Drugs on Disability in Eleven European Countries

Summary

Lichtenberg (2019) uses two-way (country × disease) fixed-effects (FE) models on Survey of Health, Ageing and Retirement in Europe (SHARE) Wave 6 (2015) data for 45,592 respondents aged 50+ in 11 European countries and 31 diseases to test whether greater prescription drug access — measured by the number of new chemical entities (NCEs) launched 1982–2015 per country-disease cell — reduces disability. Drug launches significantly reduce severe limitation (−4.9 percentage points [pp], from 21.8% to 16.9%), any limitation (−7.7 pp, from 61.1% to 53.4%), and Activities of Daily Living (ADL) limitations (−29%). Obesity is identified as the likely offsetting factor that explains why no aggregate disability trend is observed in period data despite drug-driven improvements. Published in Disability and Health Journal 12(3): 375–386.

Key Claims

Drug launches reduce severe disability by −4.9 pp: The coefficient on LAUNCHES_1982_2015 in the log-odds of severe limitation model is negative and highly significant (p=.004). The aggregate counterfactual implies drug launches reduced the probability of severe limitation from 21.8% to 16.9% — a 22% relative reduction. This is the primary disability outcome.
Drug launches reduce any limitation by −7.7 pp: Significant at p=.033. Probability of any (severe or mild) limitation reduced from 61.1% to 53.4%.
ADL limitations reduced by 29%: Significant at p=.024. Mean ADL limitations reduced from ~0.477 (counterfactual) to 0.341 (actual), a reduction of 0.136.
No significant effect on Instrumental Activities of Daily Living (IADL) limitations (p=.415), suggesting drug effects are concentrated in physical/medical ADL rather than instrumental daily activities.
Small but marginally significant quality-of-life improvement (CASP index, p=.075): +1.6% increase in the CASP (Control, Autonomy, Self-Realization, Pleasure) index.
Market size drives drug launches: Coefficient on ln(PREV) in the launches model = 0.366 (p=.0237) — larger relative disease prevalence leads to more relative drug launches for that disease in that country, confirming the "misery loves company" market-size hypothesis.
Drug access varies substantially across countries: Top 3 countries (by NCEs launched 1982–2015) had 42% more launches than bottom 3 (709 vs. 501 NCEs). Even countries with similar totals differ in which diseases were covered — Italy had $\geq$ 2 more drugs for 4 diseases but $\geq$ 2 fewer for 3 others compared to Spain.
Obesity offsets drug gains: SHARE data show obesity (body mass index [BMI] $\geq$ 30) among the 50+ population rose from 17.1% (2004) to 22.2% (2015). This adverse behavioral trend likely offset drug-driven disability reductions in aggregate period analyses, explaining why Verropoulou and Tsimbos (2017) detected no significant disability trend in SHARE 2004–2013. Drug benefits are real but masked by concurrent obesity-driven deterioration.
Cost-effectiveness: Population-weighted pharmaceutical expenditure on post-1981 drugs ≈$611/person/year for those 45+. Expenditure of $611 reduced severe limitation probability by 4.9 pp — cost-effective if willingness to pay to avoid severe limitation exceeds $12,469 (before counting drug benefits for mortality reduction or hospitalization avoidance, which prior research confirms are large).
Identification strategy: Two-way fixed effects control for all country attributes that affect all diseases equally (income, health system, education) and all disease attributes that affect all countries equally (e.g., that Alzheimer's is more disabling than hypertension on average). The remaining variation — whether Italy had more breast cancer drugs than Spain — drives the identification.

Concepts Introduced or Extended

Entities Mentioned

Frank R. Lichtenberg

Quotes

"The estimates imply that drug launches during 1982–2015 reduced the probability of severe limitation in 2015 by 4.9 percentage points, from 21.8% to 16.9%; they reduced the probability of any limitation by 7.7 percentage points, from 61.1% to 53.4%; and they reduced the mean number of Activities of Daily Living limitations by about 29%."

"The reduction in disability attributable to increasing availability of prescription drugs may have been offset by adverse trends in behavioral risk factors, such as obesity."

My Take

This paper provides the first large-scale cross-national evidence that pharmaceutical access causally reduces disability (as opposed to mortality). It directly fills the gap flagged by Cutler (2004) and others: the standard value of a statistical life (VSL) framework for valuing medical innovation focuses entirely on mortality; the quality-of-life / disability dimension is typically ignored. Lichtenberg's −4.9 pp severe limitation finding is substantial — roughly comparable to major public health interventions. The identification through country-disease variation in drug launches is clever and plausible; the two-way FE design is a sensible response to the observational challenge. The main limitations are: (1) drug launch = access to, not actual use of, a drug; (2) attribution of disability to specific diseases is impossible when 60% of patients have multimorbidity (addressed by the (1/N) weighting but not fully resolved); (3) the obesity confounder discussion is presented descriptively rather than identified. The paper fits naturally alongside Cutler (2004) and extends the value-of-innovation framework from mortality to morbidity.