Disability-Adjusted Life Years (DALYs)
A high-level overview
DALYs
What is a DALY?
Some may refer to DALYs as a “gap” measure
- What does this mean?
DALYs seek to quantify the “gap” between a population’s current health & a normative “ideal” of health/life expectancy (gap between current health & full health)
What is a DALY?
More DALYs = worse health
“0” = perfect health; “1” = death
Goal: Minimize DALYs (minimize health loss)
Note
Often used in LMICs since disability weights are freely & publicly available
Comparison with “QALY”
Could be referred to as an “expectancy” measure; i.e., they measure healthy life years gained from interventions (combining quantity + quality of life in a single metric)
QALYs seek to quantify the expected quality-adjusted life years from an intervention (how many healthy life years will you gain)
Comparison with “QALY”
More QALYs = better health
“0” = death; “1” = perfect health
Goal: Maximize QALYs (maximize health gains)
Note
More likely to be used in high-income settings & even in these settings, utility’s can be hard to get
Comparison with “QALY”
- QALY: Uses preference-based “utility” weights about health states
- DALY: Uses condition/disease specific disability weights
Note
QALYs & DALYs are conceptually & numerically distinct
Why use summary measures of health?
- Allows for comparison across diseases/therapeutic areas
- Provides a common metric to inform resource allocation decision making
- Offers complementary information alongside condition-specific outcomes (e.g., HIV cases averted; hospitalizations avoided)
What is a DALY?
DALYs = YLL + YLD
- YLL (Years of Life Lost): The # of life years a person could have expected to live had they not died
- YLD (Years Lived with Disability)
Years of Life Lost to Disease
For a given condition c,
YLD(c) = D_c \cdot L_c
- D_c is the condition’s disability weight
- L_c is the time lived with the disease.
Years of Life Lost to Premature Mortality
- YLL are defined by by a “loss function.”
- Drawn from a reference life table, indicating remaining life expectancy at age a. YLL(a)= Ex(a)
Years of Life Lost to Premature Mortality
Different approaches to identifying the time lost due to premature mortality:
Exogenous: Maximum length of life observed in modern world, i.e., “synthetic life table”; irrespective of country and socioeconomic characterstics/etc. (GBD 2021 used a LE of 86.6 years at birth; roughly similar to Japan + Monaco)
Endogenous: From a local/national life table - specific to a population’s mortality risks and health states (e.g., India/Zambia)
Simulation-based: Estimated within a disease-specific model using time-to-death (e.g., CVD; life expectancy might be different than whole population)
Years of Life Lost to Premature Mortality
Approach depends on the purpose of the study
- Synthetic life table: Used to quantify disease burden across countries relative to a normative benchmark - helpful for global justice comparisons & resource allocations for LMICs
- Country-specific life table: Useful when evaluating interventions wtihin a specific setting (e.g., Uganda), where the goal is to quantify expected years of life lost within local mortality conditions.
- Synthetic life table: Used to quantify disease burden across countries relative to a normative benchmark - helpful for global justice comparisons & resource allocations for LMICs
Source: Anand & Reddy LSE 2019
DALYs
DALY(c,a) = YLD(c) + YLL(a)
- c is the condition’s disability weight
- a is the remaining life expectancy at age a
DALYs = YLL + YLD
Simple DALYs Example
- Scenario: Person born with HIV
- Without treatment: Dies at age 30; lives entire life with untreated HIV
- With treatment: Survives to age 50; lives entire life with treated HIV
DALYs = YLL + YLD
- Years of Life Lost (YLL): Measures the gap between age at death and expected life expectancy, based on a synthetic reference life table
- YLLs are based on the age at death
- YLL example: Providing HIV treatment delays death from age 30 to age 50
| Age | Life Expectancy | Age | Life Expectancy |
|---|---|---|---|
| 0 | 88.9 | 50 | 39.6 |
| 1 | 88.0 | 55 | 34.9 |
| 5 | 84.0 | 60 | 30.3 |
| 10 | 79.0 | 65 | 25.7 |
| 15 | 74.1 | 70 | 21.3 |
| 20 | 69.1 | 75 | 17.1 |
| 25 | 64.1 | 80 | 13.2 |
| 30 | 59.2 | 85 | 10.0 |
| 35 | 54.3 | 90 | 7.6 |
| 40 | 49.3 | 95 | 5.9 |
| 45 | 44.4 |
Source: http://ghdx.healthdata.org/record/ihme-data/global-burden-disease-study-2019-gbd-2019-reference-life-table
DALYs = YLL + YLD
- Life years (LYs) gained: 20 years
- Without treatment (death at age 30); YLL = LE(30) = 59.2. years
- With treatment (death at age 50); YLL = LE(50) = 39.6 years
- Change in YLL (no treatment) - YLL (with treatment) = 59.2-39.6 = 19.6 total YLLs averted
Important
YLL (measured as DALYs averted) \neq LYs gained!
DALYs = YLL + YLD
Now let’s look at the disability weight component of the DALY
YLD = DW x duration of disease (need to know age of onset; don’t need age of onset for YLL)
- The disability weight for HIV without treatment = 0.582 (source: GBD)
- The disability weight for HIV with treatment = 0.078 (source: GBD)
- Reflects the impact on health and daily life due to symptoms like weight loss, fatigue, frequent infections, etc.
- Age of onsent: Born with HIV (age of onset = 0)
DALYs = YLL + YLD
In our HIV example,
YLD = disability weight * duration with condition
- Without treatment = 0.582 x 30 years (if age of onset was at 25 years, duration would be shorter; this would be 0.582 x 5; but in our example, person born with HIV, so duration = entire lifespan)
- Without treatment, YLD = 17.46
- With treatment = 0.078 x 50 years (if age of onset = 25 years & treatment began at dx, this would be 0.078 x 25)
- With treatment, YLD = 3.9
- Change in YLD = 17.46 - 3.9 = 13.56 total YLDs averted
DALYs = YLL + YLD
- Without treatment:
- YLL = 59.2
- YLD = 0.582 × 30 years = 17.46
- Total DALYs = 76.66
- With treatment:
- YLL = 39.6
- YLD = 0.078 × 50 years = 3.9
- Total DALYs = 43.5
- DALYs averted = 76.66 − 43.5 = 33.16 DALYs averted
Disability Weights
- Common values for small set of named health conditions (e.g. early/late HIV, HIV/ART)
- First iteration: expert opinion
- Second iteration: Pop-based HH surveys in several world regions (13,902 respondents)
Paired comparison of two health state descriptions which worse
Probit regression to calculate disability weights
235 unique health states
Applying DALYs in your models
Applying DALYs in your models
Points relevant to this workshop
Commonly used DALY shortcut methods (like QALY-like proxies or death state accumulation) yield biased DALY and ICER estimates, especially when YLLs play a large role.
In this method, payoffs are determined endogenously within the model. Payoff value is applied to the absorbing death state, so YLLs will continue to accumulate even after all cohort members have died (because unlike QALYs, dead is given a disability weight of 1)
- Correct method: Counts YLL once at death using a life table (fixed penatly)
- Incorrect method: Keeps counting YLL every year someone is in the “dead” state; inflated penalty that depends on model length
Applying DALYs in your models
After the break: We will walk you through how to correctly model DALYs in the Amua software
- First, from the decision tree model exercise that you completed earlier
- Second, in the Markov model that you finalized