Indirect measures of TB incidence

1
PHASES OF TUBERCULOSIS CONTROL
140
120
100
80
case notification rate
60
40
20
0
0
5
10
15
20
25
30
years since DOTS introduction
surveys
surveillance
MEASUREMENT
burden
incidence
TARGETS
70 detection 85cure
50 reduction prevalence, deaths
50 reduction incidence rate
2
PHASES OF TUBERCULOSIS CONTROL
140
to elimination
incidence
burden
120
100
80
case notification rate
60
40
Iran Egypt Morocco
Pakistan Somalia Afghanistan
20
Gulf States
0
0
5
10
15
20
25
30
years since DOTS introduction
3
Four indirect measures of TB incidence
Incidence, prevalence, deaths derived by
rearranging 4 equations
4
From infection prevalence
From disease prevalence
300
300
ESTIMATING TB INCIDENCE
incidence
250
250
falling
200
200
b gt 50
incidence
ss/100,000
ss/100,000
incidence
incidence
150
150
steady
b 50
100
100
weighted duration
illness 2y
50
50
ss/- HIV/- DOTS/-
0
0
0
1
2
3
4
5
6
0
200
400
600
ARI (/year)
prevalence ss/100,000
From case notifications
From HIV prevalence
300
300
250
250
200
200
ss/100,000
ss/100,000
incidence
incidence
150
150
100
100
no HIV
50
proportion detected 0.8
50
0
0
0
50
100
150
200
250
0
0.1
0.2
0.3
0.4
notification ss/100,000
prevalence HIV/100,000
5

• Proportion detected guess from quality of
  surveillance system e.g. USA probably detects
  about 95 of cases
• More objective e.g. health units reporting in
  any year
• Beware circular arguments
• Method 1 weak with poor surveillance, but strong
  MS is the ultimate goal

6

• Disease prevalence from population surveys e.g.
  Philippines, China, Cambodia
• Duration time span of the condition measured in
  prevalence survey e.g. ss disease
• Duration from e.g. patients and physicians asked
  about reporting and treatment delays (often
  underestimated)

7
Estimating mean duration Egypt

• Proportion ss cases treated
• DOTS 0.4
• non-DOTS 0.5
• untreated 0.1
• Estimated ss durations (years)
• DOTS 1.0
• non-DOTS 1.5
• untreated 2.0

8
Prevalence vs incidence Korean civil
servantsTubercle and Lung Disease 76, 534 (1995)

• Prevalence PTB 1990 241/100K
• Incidence PTB 1989-90 84/100K/yr
• Estimated duration 241/84 2.9 years (bigger
  ratio for older age groups)

9
(No Transcript)
10
Estimating incidence from prevalence Cambodia

• prevalence ss 270/100K in 2002
• incidence DOTS (case notification rate) 141/100K
  in 2002
• duration DOTS (questionnaire survey) 1 year
• duration nonDOTS (no treatment) 2 years
• Therefore
• incidence ss nonDOTS 64/100K
• total incidence (DOTS nonDOTS) 205/100K in
  2002
• NB usually wide range on estimates

11

• Styblo ratio 1 ARI to 50 ss/100,000 population
  (range 40-60)
• Accuracy of ARI from tuberculin surveys?
• 150 breaks down when TB incidence not stable
  (gets bigger in decline), and in presence of HIV

12
Origin of Styblos ruleBull IUAT vol 60, 1985
13
(No Transcript)
14

• Accuracy of deaths from vital registration?
• Case fatality more accurately measured from
  observed cohorts (but fate of defaulters,
  transfers?)
• CFR less accurately from unseen patients, whether
  treated or untreated

15
Estimating case fatality rate Egypt

• Proportion ss cases treated
• DOTS 0.4
• non-DOTS 0.5
• untreated 0.1
• Estimated case fatality ss (years)
• DOTS 0.1
• non-DOTS 0.3
• untreated 0.7

16
TB incidence closely correlated with HIV
prevalence in Africa
1000
800
600
Estimated TB incidence
(per 100,000 population)
400
200
0
0
10
20
30
40
HIV prevalence, adults 15-49y
17
TB incidence weakly related to social and
economic variables infant mortality
6.5
6.0
5.5
AF
AS
5.0
CA
4.5
EE
Ln (estimated TB incidence)
SA
4.0
US
WP
3.5
ME
3.0
WE
Series10
2.5
2.0
1.5
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
5.5
Ln (Infant mortality rate 1991)
18
Association between the rate of TB among
country-of-birth-specific groups in Australia and
the rate of TB in the country of birth. From
Watkins Plant 2003.
19
Uses and abuses of burden

• Burden advocacy on scale of the problem (world,
  regions, countries)
• Trends direction of epidemic, impact of control
• Control progress to international targets
• Local burden and local progress to targets ??

20
Whats wrong with the estimation process?

• Cannot survey the whole world (infection or
  disease)
• Many estimates are based on guesses about case
  detection
• Estimates often too inaccurate or too biased to
  judge progress to case detection targets
• National estimates do not apply sub-nationally
• ? value of exploiting surveillance data