Years Lived with Disability

Years Lived with Disability and the Non-Fatal Burden of Disease

Years lived with disability (YLDs) are the part of the disability-adjusted life year (DALY) that records illness people live with rather than die from, which is why low back pain and depression rank high in YLD tables while causing few deaths. The counting method has changed more than the short formula suggests. The first Global Burden of Disease (GBD) study counted the future health loss of new cases, discounted and age-weighted, while every GBD round since 2010 has counted the health loss lived in the reporting year. Incidence-based and prevalence-based counts part company when a disease is declining, as an illustrative condition below shows, and YLD figures are reported, combined with years of life lost (YLLs) and used in priority setting and cost per DALY averted. The severity weights, and the simulation that combines them when people have several conditions, are covered on the disability weight page.

Two ways to count non-fatal health loss

Every YLD calculation multiplies time spent in a health state by a weight for its severity, but the time can be counted from two starting points. An incidence-based count follows the people whose condition began in the reference year and adds up all the health loss they will have until remission or death. A prevalence-based count takes everyone living with the condition in the reference year and records only the health loss of that year.

The GBD 1990 study and the World Health Organization (WHO) updates that followed it used the incidence perspective.

$$YLD_{inc} = \sum_{s} I_s \times DW_s \times L_s$$

where $I_s$ is the number of incident cases of sequela $s$ (a health consequence of a disease or injury) in the reference year, $DW_s$ is its disability weight and $L_s$ is the average duration of a case until remission or death, in years. WHO's methods document explains the choice as keeping YLDs consistent with YLLs, which take an inherently incidence perspective because each death in the year carries all of its future lost life.

From GBD 2010 onwards, and in WHO's Global Health Estimates, YLDs are calculated from prevalence.

$$YLD_{prev} = \sum_{s} P_s \times DW_s$$

where $P_s$ is the number of people living with sequela $s$ in the reference year, measured in GBD as point prevalence, and $DW_s$ is as defined above. Duration no longer appears as a separate term, because a long-lasting condition builds up more prevalent cases. When incidence and duration have been stable for a long time, prevalence is approximately incidence multiplied by average duration, $P \approx I \times L$, and the two counts agree.

Why GBD 2010 moved to prevalence

The reasons given for the switch were mainly about data and about usefulness to policy. WHO's methods document lists the drawbacks of the incidence approach that led both GBD 2010 and WHO to change.

• It does not reflect the current burden of a condition whose incidence has fallen substantially.
• It needs estimates of both incidence and average duration, whereas for many conditions the data collected are mainly on prevalence.
• It assigns all of a condition's YLDs to the ages at onset, whereas policy-makers are often more interested in the ages at which the health loss is experienced.
• Adjusting for comorbidity is more straightforward with prevalence.

The most visible effect was on the age distribution of YLDs. In WHO's example of congenital hearing loss, all YLDs fall at age 0 in the incidence perspective but are spread fairly evenly across age groups in the prevalence perspective. Prevalence-based YLDs were already used for other purposes, such as period healthy life expectancy, which GBD 2021 still calculates from age-specific mortality rates and YLDs per capita.

Discounting and age weighting before 2010

The original DALY added two value choices to the YLD calculation, a 3% discount rate and age weights, and both applied to YLDs as well as YLLs. In an incidence-based count, discounting shrinks the health loss of a long-lasting condition, since most of it lies in the future. With continuous discounting and no age weighting, the YLDs for one new case are as follows.

$$YLD_{case} = DW \times \frac{1 - e^{-rL}}{r}$$

where $DW$ is the disability weight, $L$ is the duration in years and $r$ is the annual discount rate. At a rate of 0.03, ten years with a condition count as about 8.64 years before weighting, and forty years as about 23.3.

Age weighting also multiplied each year lived with the condition by a weight that was highest in young adulthood and lower in childhood and old age.

GBD 2010 calculated neither YLLs nor YLDs with age weighting or discounting, and WHO later did the same. WHO's methods document records a problem specific to YLDs. Discounting started at the year of incidence for YLDs but at the year of death for YLLs, so the two components were discounted from different points. It also reports that dropping both conventions substantially increased the absolute number of DALYs and raised the share of DALYs at younger and older ages.

Comorbidity and adding YLDs across causes

YLDs for separate causes cannot simply be added, because many people, especially at older ages, have more than one condition at once; the simulation GBD uses to combine weights, and its effect on global totals, are set out on the disability weight page. For users of YLD tables, the consequence is that a cause's YLDs are net of its share of the overlap, so they are smaller than its prevalence times its stand-alone weight.

Worked example of a declining condition counted both ways

The figures are illustrative and do not describe a real disease. A condition begins at exactly age 20, lasts exactly 10 years and then resolves completely, causes no deaths, and has a disability weight of 0.20. For decades 100 people a year developed it, and each annual cohort is assumed to contribute one full person-year in each of its 10 years.

1. Steady state, no discounting. The incidence-based count follows this year's 100 new cases over their 10 years.

$$YLD_{inc} = 100 \times 0.20 \times 10 = 200$$

where $YLD_{inc}$ is the incidence-based YLD for the reference year. The prevalence-based count takes the 10 cohorts living with the condition this year, 1,000 person-years in all.

$$YLD_{prev} = 1{,}000 \times 0.20 = 200$$

where $YLD_{prev}$ is the prevalence-based YLD for the same year. In a steady state the two methods agree.

2. Adding 3% discounting. Under the pre-2010 discounting convention (age weighting left aside) each new case's 10 years are discounted.

$$YLD_{inc} = 100 \times 0.20 \times 8.639 = 172.8$$

where 8.639 is the discounted value of 10 years at 3% from the formula above. The prevalence count has no future stream to discount, so the gap of 27.2 YLDs comes from the convention alone.

3. Incidence halves. A prevention programme has halved new cases to 50 a year; it has run for five years, this year included. This year's incidence-based count is as follows.

$$YLD_{inc} = 50 \times 0.20 \times 10 = 100$$

where the inputs are as in step 1 apart from the 50 new cases. The 10 cohorts now living with the condition are five of 100 people (onset 5 to 9 years ago) and five of 50 (onset 0 to 4 years ago), so prevalence is 750 person-years.

$$YLD_{prev} = (5 \times 100 + 5 \times 50) \times 0.20 = 750 \times 0.20 = 150$$

where the first bracket counts the person-years lived by the older and newer cohorts.

4. Ages at which the loss falls. The incidence count places all 100 YLDs at age 20. The prevalence count spreads its 150 across ages 20 to 29 as follows. The 250 person-years lived by people aged 20 to 24 give 50 YLDs, and the 500 person-years lived by people aged 25 to 29, who became ill before the programme, give 100.

5. What the numbers show. The incidence-based figure halves as soon as incidence halves, crediting the programme at once with all the future health loss it prevents. The prevalence-based figure has fallen by only a quarter, because people who became ill before the programme are still living with the condition, and it reaches 100 only after another five years, when the last of the larger cohorts has recovered. Neither figure is the programme's effect in an economic evaluation. A model would credit each year of the programme with 50 cases prevented, or 100 YLDs averted over the 10 years each case would have lasted (86.4 with continuous discounting at 3%, as in the formula above), whatever the annual burden figure shows.

Leading causes and the YLD share of the DALY

GBD reports YLDs as counts, all-age rates and age-standardised rates, each with an uncertainty interval, by cause, age, sex, location and year. Counts show where non-fatal health loss is concentrated, and age-standardised rates show whether it is changing once population growth and ageing are taken out.

GBD 2010 estimated 777 million YLDs worldwide in 2010, up from 583 million in 1990. YLDs per person rose with age, from 5.4% at age 5 to 28.6% at age 80, and on average each person in the world had an 11% reduction in overall health. Set against the 2.490 billion DALYs in the companion GBD 2010 analysis, YLDs made up about 31% of global DALYs. The split varied widely. YLLs made up about half of DALYs in high-income regions but over 80% in sub-Saharan Africa, so the YLD share ranged from about half to under a fifth.

GBD 2021 estimated 907 million YLDs in 2021, 31.3% of global DALYs, against 1.98 billion YLLs. Low back pain (70.2 million YLDs), depressive disorders (56.3 million) and headache disorders (48.0 million) were the three leading causes, and COVID-19, the leading cause of DALYs that year, ranked 19th for YLDs. Global YLDs rose from 738 million in 2010, while the age-standardised YLD rate, which removes the effects of population growth and ageing, increased by only 2.6%.

The two figures for 2010, 777 million from GBD 2010 and 738 million from GBD 2021, show that each round re-estimates the whole time series; GBD 2010 stated that its results superseded all previously published GBD results, and WHO warns that its estimates for 2000 to 2019 are not directly comparable with its earlier ones. Trends should therefore be read within one round.

How YLDs relate to YLLs, DALYs and healthy life expectancy

YLDs and YLLs are added to give DALYs, but since 2010 they count time differently. YLLs assign each death in the year all of its future lost life, an incidence-type count, while YLDs record only the non-fatal health loss lived in the year. A fall in deaths therefore moves the DALY at once, whereas a fall in the incidence of a long-lasting non-fatal condition reaches it gradually, as in the worked example. The table sets out the boundaries between the measures.

YLDs are a loss measure on the DALY scale and are not the same as QALYs lost; the differences between disability weights and the utility weights behind the quality-adjusted life year are covered on the disability weight page.

YLDs in priority setting and cost per DALY averted

Burden estimates show where non-fatal health loss is large and growing. The GBD 2010 authors concluded that health systems will need to respond to rising numbers of people with disorders that largely cause disability rather than death. A large YLD total does not by itself show that effective and affordable interventions exist, so burden figures inform priorities without setting them.

In economic evaluation, YLDs enter the effect measure through DALYs averted. An intervention averts YLDs by preventing new cases, shortening their duration or reducing their severity, and a model follows the people affected over a time horizon. WHO-CHOICE now uses a lifetime horizon and reports health benefits both undiscounted and discounted at 3%, with costs discounted at 3%. The discounting dropped from GBD burden estimates therefore remains an open choice in cost-effectiveness analysis with DALYs. Fox-Rushby and Hanson found that many such analyses could not be compared because their assumptions differed or were unclear, and proposed minimum reporting criteria; Larson later set out a discrete-time calculation that matches the discounting used for costs.

Limits of YLD estimates

YLDs are only as good as the prevalence, severity and weight data behind them. GBD 2021 sets out several limits in its own estimates.

• Severity data. The proportion of cases at each severity level for most non-fatal causes comes from the US Medical Expenditure Panel Survey, and severity distributions do not capture variation in access to treatment, so better care in a country may not show up in its YLDs.
• Comorbidity. The correction assumes independent comorbidity, which underestimates overlap for causes such as mental disorders, where the pattern depends on the range of disorders a person has, so YLDs for those causes are overstated.
• Uncertainty. The 48.0 million YLDs from headache disorders in 2021 had a 95% uncertainty interval of 9.80 million to 101 million.

Every YLD also inherits the weaknesses of its disability weight, including the equity critiques discussed on the disability weight page. A common misreading is to treat YLDs per person as a share of people. The 28.6% at age 80 in GBD 2010 is average health loss per person, not the proportion of people with a disability.

Sources

• Bertram MY, Lauer JA, Stenberg K, Tan Torres Edejer T. Methods for the economic evaluation of health care interventions for priority setting in the health system: an update from WHO CHOICE. International Journal of Health Policy and Management. 2021;10(11):673-677. doi:10.34172/ijhpm.2020.244
• Fox-Rushby JA, Hanson K. Calculating and presenting disability adjusted life years (DALYs) in cost-effectiveness analysis. Health Policy and Planning. 2001;16(3):326-331. doi:10.1093/heapol/16.3.326
• GBD 2021 Diseases and Injuries Collaborators. Global incidence, prevalence, years lived with disability (YLDs), disability-adjusted life-years (DALYs), and healthy life expectancy (HALE) for 371 diseases and injuries in 204 countries and territories and 811 subnational locations, 1990-2021: a systematic analysis for the Global Burden of Disease Study 2021. The Lancet. 2024;403(10440):2133-2161. doi:10.1016/S0140-6736(24)00757-8
• Larson BA. Calculating disability-adjusted-life-years lost (DALYs) in discrete-time. Cost Effectiveness and Resource Allocation. 2013;11:18. doi:10.1186/1478-7547-11-18
• Murray CJL. Quantifying the burden of disease: the technical basis for disability-adjusted life years. Bulletin of the World Health Organization. 1994;72(3):429-445.
• Murray CJL, Vos T, Lozano R, Naghavi M, Flaxman AD, Michaud C, et al. Disability-adjusted life years (DALYs) for 291 diseases and injuries in 21 regions, 1990-2010: a systematic analysis for the Global Burden of Disease Study 2010. The Lancet. 2012;380(9859):2197-2223. doi:10.1016/S0140-6736(12)61689-4
• Rothman KJ, Greenland S, Lash TL. Modern Epidemiology. 3rd ed. Philadelphia: Lippincott Williams and Wilkins; 2008. Chapter 3.
• Vos T, Flaxman AD, Naghavi M, Lozano R, Michaud C, Ezzati M, et al. Years lived with disability (YLDs) for 1160 sequelae of 289 diseases and injuries 1990-2010: a systematic analysis for the Global Burden of Disease Study 2010. The Lancet. 2012;380(9859):2163-2196. doi:10.1016/S0140-6736(12)61729-2
• World Health Organization. WHO methods and data sources for global burden of disease estimates 2000-2019. Global Health Estimates Technical Paper WHO/DDI/DNA/GHE/2020.3. Geneva: WHO; 2020. Sections 1.4, 2 and 2.1 to 2.5.