DALYs: Estimating healthy time lost to selected diseases and injuries

1
DALYs Estimating healthy time lost to selected
diseases and injuries

• John Powles
• MSt in Public Health

2
The steps in principle

• For a given condition (disease or type of injury)
  and reference (calendar) period, estimate the
  flows of healthy life lost
• 1. Completely, by death (YLL)
• 2. Partially, by onsets of impaired health
  expressed as equivalent years completely lost
  (YLD).
• Eg 10 years lost to a state half as bad as death
  (ie with a disability weight of 0.5)
  counts as equivalent to 5 years completely lost

3
Analytic choices

• Reference year
• Age groups
• GBD 2000
• 0-4
• 5-14
• 15-29
• 30-44
• 45-59
• 60-69
• 70-79
• 80

4
Disease and injury categories

• Mutually exclusive and exhaustive
• 4 levels of disaggregation
• total over 100
• each lowest level group includes a residual
  category
• Choose diseases and injuries of most importance
  in population of interest
• should reduce sum of residual categories to lt12
  of deaths

5
GBD standard list 1st level of disaggregation

• Gp I Communicable, maternal, perinatal and
  nutritional
• Gp II Non-communicable diseases
• Gp III Injuries
• (nb some NCDs have infective causes eg
  cancers of stomach, liver and cervix, valvular
  disease of the heart)

6
2nd level of disaggregationExample Gp I
Communicable, maternal etc

• Infectious and parasitic diseases
• Respiratory infections
• Maternal conditions
• Conditions arising during the perinatal period
• Nutritional deficiencies

7
3rd level of disaggregationExample A.
Infectious and parasitic diseases

• Tuberculosis
• Sexually transmitted (excl HIV)
• HIV/AIDS
• Diarrhoeal diseases
• Childhood cluster diseases
• Meningitis
• Hepatitis B and C
• Malaria
• Tropical cluster diseases
• Leprosy
• Dengue
• Japanese encephalitis
• Trachoma
• Intestinal nematode infections
• Other infectious diseases

8
4th level of disaggregationExample 5.
Childhood cluster diseases

• Pertussis
• Poliomyelitis
• Diphtheria
• Measles
• Tetanus

9
2nd and 3rd levels of disaggregation for injuries

• Unintentional injuries
• Road traffic accidents
• Poisonings
• Falls
• Fires
• Drownings
• Other unintentional injuries
• Intentional injuries
• Self-inflicted injuries
• Violence
• War
• Other intentional injuries

10
Problems in summing to a plausible total

• Dealing with the residuals (other )
• eg Take YLL for deaths due to other . (ie
  group total - those in the chosen list)
• Assume same ratio of YLD to YLL as calculated for
  the mean for chosen causes in that group
• Avoiding double counting
• eg HIV TB
• -gt precedence rules

11
Precedence rules
12
(No Transcript)
13
(No Transcript)
14
Basic disease model
15
Elaborated model for an infectionexample
Hepatitis B
16
Elaborated model for an infectionexample
Hepatitis B
17
Elaborated model for an infectionexample
Hepatitis B
18
Elaborated disease modelexample breast cancer
gt5cm at diagnosis
Source Australian Burden of Disease Study
19
Information sources for modelling the
epidemiology of selected conditions

• Usually unhelpful
• Routine health surveys (eg Health Survey for
  England)
• Information on services provided (eg hospital
  discharges, physician consultations)buthospital
  data can be useful for eg perinatal and maternal
  conditions, meningitis, stroke, MI, serious
  injuries

• Potentially helpful
• Disease registers
• Disease-specific surveys with validated
  instruments eg for depression
• Epidemiological studies (eg cohort/natural
  history)
• Expert panels

20
Why is self-report information from health
surveys usually unhelpful?

• Mismatch of lay and professional definitions
• eg Asthma
• Prevalence estimated from self-reports of
  wheezing or doctors diagnosis
• 2-3 times higher than
• that from surveys using objective test of airways
  hyper-responsiveness

21
Why is self-report information from health
surveys usually unhelpful?

• Subjects unaware of their condition
• eg diabetes, hypertension
• Biased perception
• eg 2cm on height, - 2 Kg on weight reduced
  prevalence of obesity by 40 in Australia

22
Commonly need to rely on

• Consulting disease experts
• for estimates of occurrence parameters
  (incidence, remission, case-fatality, duration)
• GBD1990 established 100 expert panels!
• Modelling software (DISMOD II)
• to force expert assessments into internal
  consistency (experience showed them to be often
  internally inconsistent by large amounts)

23
Dementia example of inconsistency in
meta-analysis findings (Australian BOD study)

• Meta-analyses of the prevalence of dementia in
  (mainly) EME countries (Jorm et al. 1987, Hofman
  et al. 1991 and Ritchie et al. 1992)
• One meta-analysis of incidence (Jorm and Jolley
  1998 -far fewer studies).
• Reasonable similarity in the reported prevalence
  of dementia across countries with the exception
  of prevalence estimates for the very old (85
  years).
• Not possible to create a DisMod model that was
  consistent between the incidence and prevalence
  figures of the meta-analyses
• Decided to model incidence from the meta-analysis
  prevalence figures reported by Jorm et al.

24
Dismod II