timing and design of repeat surveys in asia: outcomes of ... · 8 | myanmar – sample size...
TRANSCRIPT
1 |
Timing and design of repeat surveys in Asia: outcomes of
February 2012 workshop
5th meeting of the WHO Global Task Force on TB Impact Measurement
Geneva, 9-10 May
Babis Sismanidis
2 |
Overview
Background to the February 2012 workshop
Bangladesh
Myanmar
the Philippines
Viet Nam
3 |
Background to workshop
Four Asian countries are planning repeat TB prevalence surveys around 2015: Bangladesh, Myanmar, the Philippines and Viet Nam
Repeat surveys are more methodologically demanding than stand alone surveys (Chapter 9, the Lime Book)
Workshop to support these four Asian countries
– Learn from recent experience in Cambodia and China
– Discuss internationally recommended design and optimal timing for repeat surveys
– Group work to explore plausible sample size scenarios
4 |
Bangladesh
National TB prevalence survey completed in 2008
52,098 survey participants (15 years of age), participation rate 80%
Prevalence under-estimated due to survey design
Another prevalence survey is planned for 2013/2014
It will be designed as a first (and not a repeat) survey
Rates per 100,000
(95% CI) 2010
Smear positive 63
(44-89)
Design effect = 2.7
5 |
Myanmar – Background
National TB prevalence survey completed in 2010
51,367 survey participants (15 years of age), participation rate 89%
Large variability in between-cluster prevalence (k)
Repeat survey planned for 2014 (4 years later)
Rates per 100,000
(95% CI) 2010
Bacteriologically-
confirmed
613
(502-748)
k=0.62
Smear positive 242
(186-315)
k=0.81
6 |
Myanmar – Factors driving prevalence
2010 – 2014
DOTS expansion, low coverage of MDR
Low level HIV
PPM expansion
Faster economic expansion expected
Limited health insurance
7 |
Case studies: Cambodia, China
ANNUAL CHANGE Cambodia China
Time period 2002-2011 2005-2010
GDP per capita +10 +20
HIV prevalence among TB -5.6 NA
% 0-14 over total population -2.6 -1.2
TB expenditure +5.7 +9
PPM NA NA
ACSM NA NA
Ratio of suspect:notification NA -5.8
Notifications
Smear positive
All forms
-4.6
4.7
-3.4
-0.7
PREVALENCE
Smear positive
Bacteriologically-confirmed
-5.6
-4.9
-9.0
-5.8
8 |
Myanmar – Sample size calculation
Annual decrease in TB prevalence (both outcomes): 5% Smear positive outcome: p1=242 → 194
Bacteriologically-confirmed outcome: p1=613 → 490
1Perform sample size calculation as a stand-alone survey, based on precision. 2Perform sample size calculation assuming 2010 prevalence estimate is the true level of prevalence. 3Perform sample size calculation assuming 2010 prevalence level and allowing for uncertainty around this level 4Perform sample size calculation using prior information from 2010 survey to inform anticipated values of prevalence in 2014
20% precision, 20% reduction compared to 2010 prevalence levels, same cluster
size and k as with the 2010 survey
Method of calculation Smear positive outcome Bacteriologically-confirmed outcome
Single survey1 110,844 54,750
Repeat surveys2 145,270 71,540
Repeat surveys3 Undefined Undefined
Bayesian4 90,520 Not yet calculated
9 |
Myanmar – Conclusions
Smear positive outcome: prohibitively large sample sizes required
– either due to the level of prevalence is not very high
– or due to short duration between the two surveys
Bacteriologically-confirmed outcome: single survey approach feasible
Consider delaying the survey to 2016/2017 in order to allow for large reductions in prevalence levels to be reached
10 |
the Philippines – Background
National TB prevalence surveys conducted in 1997 and 2007
Repeat survey planned for 2014 (7 years later)
Rates per
100,000
(95% CI)
1997 2007 Annual
decrease (%)
Bacteriologically-
confirmed
960
(750-1160)
660
(510-810)
k=0.55
3.7%
Smear positive 360
(280-450)
260
(170-360)
k=0.79
3.3%
11 |
the Philippines – Factors driving prevalence
1997 – 2007 – DOTS expansion, low coverage of MDR
– Low level HIV
– PPM expansion
– Slow economic expansion
– Limited health insurance
2007 – 2014 (sustained levels of decrease in TB expected) – Some more PPM (PPM coverage currently stable)
– Universal health insurance?
– Faster economic growth?
– MDR control (target 30% DST in retreatment)
– Negative factors: diabetes rising fast, aging population
12 |
the Philippines – Sample size calculation
Annual decrease in TB prevalence (smear positive): 3.3% Annual decrease in TB prevalence (bacteriologically confirmed): 3.7% Smear positive outcome: p1=260 → 200
Bacteriologically-confirmed outcome: p1=660 → 488
1Perform sample size calculation assuming 2007 prevalence estimate is the true level of prevalence. 2Perform sample size calculation assuming 2007 prevalence level and allowing for uncertainty around this level 3Perform sample size calculation using prior information from 2007 survey to inform anticipated values of prevalence in 2014
20% precision, 23% and 26% reduction respectively for smear positive and
bacteriologically confirmed compared to 2007, same cluster size and k as with the
2007 survey
Method of calculation Smear positive outcome Bacteriologically-confirmed outcome
Repeat surveys1 88,557 38,054
Repeat surveys2 Undefined Undefined
Bayesian3 47,517 18,855
13 |
the Philippines – Conclusions
Decline 2007–2014 should not be assumed faster than decline 1997–2007
Smear positive outcome: – use Bayesian approach, but still almost 50,000 (too expensive)
– do the survey in 2017 or later instead of 2014
Bacteriologically-confirmed outcome: the survey may show an overall decline since 1997
Consider delaying the survey to 2017 – Too late for an MDG assessment
– But evidence from 3 surveys indicating the country is on track
14 |
Viet Nam – Background
National TB prevalence survey conducted in 2007
94,179 survey participants (15 years of age), participation rate 91%
Large variability in between-cluster prevalence (k)
Repeat survey planned for 2014 (7 years later)
Rates per 100,000
(95% CI) 2007
Bacteriologically-
confirmed
307
(249-366)
k=0.59
Smear positive 197
(150-244)
k=0.75
15 |
Viet Nam – Factors driving prevalence
2007 – 2014
DOTS expansion, low coverage of MDR
Low level of HIV: Antenatal clinic surveillance 0.37% in 2005, 0.25% in 2010, an annual decrease of -9.3%
PPM expansion: currently only covers about 5% of private health providers
Fast economic expansion. Total increase between 2007 and 2010 in GDP per capita is about 10%
Aging of the population
Suspect:Notified smear positive ratio increased from 10.4 in 2007 to 11.4 in 2010
16 |
Viet Nam – Sample size calculation
Annual decrease in TB prevalence (both outcomes): 4.3% Smear positive outcome: p1=197 → 138
Bacteriologically-confirmed outcome: p1=307 → 215
1Perform sample size calculation assuming 2007 prevalence estimate is the true level of prevalence. 2Perform sample size calculation assuming 2007 prevalence level and allowing for uncertainty around this level 3Perform sample size calculation using prior information from 2007 survey to inform anticipated values of prevalence in 2014
20% precision, 30% reduction compared to 2007 prevalence levels, same cluster
size and k as with the 2007 survey
Method of calculation Smear positive outcome Bacteriologically-confirmed outcome
Repeat surveys1 95, 524 81,517
Repeat surveys2 Undefined Undefined
Bayesian3 Not yet calculated 133,592
17 |
Viet Nam – Conclusions
There is political commitment in the country to measure the level of prevalence in 2015, in order to assess if the West Pacific Region target of 50 % reduction in prevalence, compared with 2000 levels, has been reached.
2007 survey only one culture per survey participant. The repeat survey should culture two specimens instead.
To end up with less than 100,000 the repeat survey must be: – conducted when the reduction in prevalence is more than 30% compared
with the 2007 survey, which could be around 2015/2016,
– consider using bacteriologically-confirmed as the primary outcome,
– consider using the Bayesian approach to sample size calculation.
18 |
Next steps
Produce and disseminate workshop report
Follow up with all four countries and continue TA
Communicate with funding agencies for the possibility of ensuring funds even is surveys are delayed
19 |
Acknowledgements
Contributing to the background document: Maridel Borja, Muhammad Farid, Katherine Floyd, Sian Floyd, Philippe Glaziou, Cornelia Hennig, Thandar Lwinn, Fulvia Mecatti, Sukyi Myo, Binh Hoa Ngueyn, Viet Nhung Nguyen, Nobu Nishikori, Ikushi Onozaki, Nemia Sucaldito, Yinyin Xia, Norio Yamada
Contributing to the repeat prevalence survey work: Cambodia, China
20 |
Question to the Task Force
Do you agree with the provisional conclusions from the February 2012 workshop?