Title: Rapid Assessment
1Rapid Assessment for Avoidable
Blindness Dr. Hans Limburg
1 of 72
2Why do we need surveys?
- We need baseline data on avoidable blindness and
poor vision in a defined area to make adequate
planning and management of intervention
strategies possible - We need follow-up data to monitor progress of
blindness intervention
2 of 72
3What do you want to know?Why do you want to
know this?
3 of 72
4Indicators used
- prevalence of all blindness, severe visual
impairment (SVI) and visual impairment (VI) - main causes of blindness, SVI and VI
- prevalence of cataract blindness
- (incidence of (cataract) blindness)
- prevalence of (pseudo)aphakia
- Cataract Surgical Coverage
- prevalence of low vision
- visual outcome after cataract surgery
- cause of poor visual outcome
- barriers to cataract surgery
- age at time of surgery, place of surgery, type of
surgery, costs, satisfaction
4 of 72
5Definitions (WHO)
- Blindness
- VA lt3/60 with best correction in the better eye
or visual field of 10 degrees or less around
visual axis - Severe visual impairment
- VA lt6/60 3/60 with best correction in the
better eye or visual field of 20 degrees or less,
but more than 10 degrees - Visual impairment
- VA lt6/18 6/60 with best correction in the
better eye or visual field of 30 degrees or less,
but more than 20 degrees
5 of 72
6Definitions (WHO)
- .... with best correction ....
- this means that refractive errors are corrected
at time of survey - problem of refractive error is not acknowledged
- .... with available correction ....
- this measures vision with the correction
available with the patient - allows estimation of magnitude of refractive
error problem - RAAB measures first with available correction,
followed by pinhole vision
6 of 72
7The survey should
- provide baseline indicators that are needed for
adequate planning and for monitoring over time - be a simple and cheap procedure
- be carried out by local staff
- be used again after 5-8 years to measure how much
has changed
7 of 72
8Survey Design
persons of all ages in survey area
21,000
persons 50
3,000
persons 50 with operable cataract or
(pseudo)aphakia
150
Blind SVI VI male female
100
50
(pseudo)aphakia
cataract
Cataract Surgical Coverage
Outcome
Barriers
8 of 72
9Examination procedures in RAAB
- Measure VA with tumbling E-chart and with
available correction and pinhole correction - Assess lens status in each eye with distant
direct ophthalmoscopy and/or portable slitlamp - If VAlt6/18 and not due to cataract, corneal scar
or refractive error, then dilate pupil and
examine with direct ophthalmoscope and/or
slitlamp - Assess main cause of VAlt6/18 in each eye and for
the person - Poor vision due to cataract ask for barriers
- If operated for cataract ask details of surgery
9 of 72
10Survey form
10 of 72
11Avoidable blindness
- Main causes
- Cataract
- Trachoma
- Onchocerciasis
- Childhood blindness
- Refractive errors and low vision
- Glaucoma
- Diabetic retinopathy
- ARMD
- Population at risk
- People 50
- Special surveys
- Special surveys
- Special surveys
- Children and people 40-50
- Mainly people 50
- Mainly people 50
- People 50
11 of 72
12Focus on people 50
- 85 of all blindness in people 50
- Nearly all cataracts in people 50
- Prevalence high in people 50, hence sample size
can be small - Elderly people often not far away from the house
- Generally good compliance
12 of 72
13Why survey people aged 50 and not people of all
ages?(Madan Mohan, Survey of Blindness India.
Summary results, New Delhi 1989)
13 of 72
14Why survey people aged 50 and not people aged
40?(Madan Mohan, Survey of Blindness India.
Summary results, New Delhi 1989)
14 of 72
15Select survey area
- Total population in survey area between 0.5 and 5
million people - Area has management structure for eye care
- Population data by sex and by 5-year age groups
available - Population by sub-unit (village, town,
neighbourhood, enumeration area, polling station,
etc.) are available - Entire area is accessible for survey teams
- No problems with security
15 of 72
16Calculate sample size (S)
- For Simple random sampling
- Estimated prevalence of condition (P)
- (cataract) blindness (VAlt3/60) e.g. 4
- Acceptable variation around estimate (D)
- Typically 20 of P e.g. 3.2 - 4.8
- Confidence in estimate (Z)
- Typically 95
- Non-compliance (absence, refusals)
- e.g. lt10
- Sinfinite population ZZ(P(1-P))/DD
- Sfinite population Sinf./(1(Sinf./population))
16 of 72
17Sampling methods
- Simple Random Sampling
- Same chance of selection for every individual
- List of all persons aged 50 in survey area
- Too much traveling
- Cluster Random sampling
- Same chance of selection for every group of
people (cluster) - List of all population units in survey area
- Less traveling
- Within cluster people share similar conditions,
leading to less variation in results. To correct
that loss of variation, a correction factor has
to be applied Design Effect (DEFF)
17 of 72
18Enumeration areas
18 of 72
19Sample size cluster sampling
- Sample size is determined also by size of
clusters and variance between and within clusters
Design Effect (DEFF) - Cluster size 40 gt DEFF 1.3
- Cluster size 50 gt DEFF 1.4
- Cluster size 60 gt DEFF 1.5
- Multiply sample size for SRS with DEFF
- DEFF is close to 1.0 when condition is evenly
spread in community and 2.0 or higher when it
clusters in families or geographical areas
19 of 72
20Calculate sample size
20 of 72
21Multistage sampling
- Select population area by systematic sampling
from sampling frame - census enumeration areas with population, or
- list of settlements with population, or
- other list of geographic distribution of total
population - Sub-divide selected population unit in segments
with equal population, enough to provide required
number of people aged 50 - Randomly select one segment
- Visit all houses in selected segment
- Examine all eligible people in these houses,
until required number has been examined
21 of 72
22Sampling frame
- census enumeration areas with population, or -
list of settlements with population, or - other
list of geographic distribution of total
population
22 of 72
23Systematic sampling
1
- Suppose we want to test a sample of size 100 from
a population of 2000, we test every 2000/100
every 20th member (sampling interval) - The starting point is determined by multiplying
the sampling interval by a random number between
0 and 1 (e.g. 0.45103 x 20 9.0206) - The first member to be tested is 9. The next
member will be 9 20 29, then 49, 69, 89, etc. - In this way 100 members are selected, evenly
spread over the population
9
29
49
69
1929
1949
1969
1989
2000
23 of 72
24Selection of clusters from sampling frame
1
- Sampling frame is list of all population units
with their population in survey area - Create cumulative column of population
- This cumulative list represents all persons in
survey area and in each geographical sub-area - Divide the total population by required number of
clusters to calculate sampling interval - determine starting point by multiplying sampling
interval by a random number between 0 and 1 - Add sampling interval to starting point to
identify second person, etc. - Make a list of the population units where the
selected persons are living. These are the
locations of the clusters. - This procedure ensures random selection of
clusters with a probability proportional to the
population size
1,000,000
24 of 72
25How to select the required number of people aged
50 in the cluster location
- 1. Random walk
- 2. Compact segment sampling
25 of 72
26Start at centre of cluster area
Select direction by spinning a bottle
At every crossing, select direction by spinning
bottle
Continue until you have examined the required
number in the cluster
26of 72
27Compact segment sampling
- If the area from which the cluster has to be
selected has a large population, bias may be
introduced through the selection process which
houses are to be visited. This can be reduced by
dividing the area in equal segments with each
enough population to provide required number of
eligible people (aged 50) - Population in segment
- Cluster size / population 50 (e.g. 50 / 21.1
237) - Number of segments
- Population area / population in segment (1482 /
237 6.3)
27 of 72
28Total population 1482
21.1 of population is aged 50
50 / 0.211 237 on average, 50 people aged 50
in every 240 people
2
divide area in 6 segments of around 240 people
3
1
6
4
5
Select one segment at random and examine all
houses until you have completed 50 people aged
50
28 of 72
29Compact segment sampling
- Advantages
- Less bias, because
- No subjective decisions to be taken by staff on
which direction to continue - All households in selected segment are visited
- Higher compliance
- All people aged 50 or higher are sure to be
visited - Less people have to stay home for examination
29 of 72
30Compact segment sampling
- Advance team
- will visit selected cluster area 3-5 days before
survey team - will divide the population unit in equal segments
on the census map - if no map is available, they will make a sketch
map of the population unit and divide this in
segments of similar size - will locate next nearest population unit in case
original unit has not enough eligible people - will contact local leaders to announce purpose
and date of survey - will contact local health worker/social worker to
ensure they accompany the survey team
30 of 72
31Materials needed per team
- 1 direct ophthalmoscope (spare batteries)
- 1 portable slit lamp and mydriatic (optional)
- 1 pinhole, preferable with multiple holes
- 1 torch with spare batteries
- 2 E charts (see attachment for sizes)
- 2 pencils with eraser and sharpener
- 1 rope-measure for 6 and 3 meter
- 1 clipboard to hold the forms
- as many survey forms as cluster size
- map of population unit divided in segments
- referral slips and basic drugs for treatment
- identity card
- shoulder bag to carry all materials
31 of 72
3232 of 72
3333 of 72
34Inter-observer variation
- Before the teams start the actual field work it
has to be checked whether they can make an
adequate diagnosis - Each team examines 40-50 people aged 50
- The teams should not know the findings of the
other teams - Findings of most experienced team are considered
correct - Findings of other teams are compared with those
of most experienced team (Gold Standard) - Agreement is measured by Kappa statistic.
34of 72
35Inter-observer variation
Each team examines 40-50 people aged 50, e.g.
patients and their company from the outpatient
department
35 of 72
36Inter-observer variation
36 of 72
37Report on inter-observer variation
37 of 72
3838 of 72
3939 of 72
4040 of 72
4141 of 72
4242of 72
4343 of 72
44Cleaning of data
- During data entry
- In case of missing or inconsistent entry
- field turns red
- when curser is placed on a red field, a message
appears indicating the error - when saving this record the same message appears.
- Checking all survey and IOV records
- Enter data immediately after a cluster was
completed and use the menu Reports Consistency
check to check consistency of all records
entered so far - Print out errors and ask team leader to correct
- Validation of double data entry
- Enter all records twice by different operators.
- Compare data files. File is considered clean if
there are no differences
44 of 72
45Consistency checks
45 of 72
46Consistency checks
46 of 72
47Consistency checks
- Enter survey records immediately after the teams
return from the field and run the consistency
check. - In case of any errors, the field staff may still
remember the patient and the error can be
corrected. - Fragment from consistency report.
47 of 72
48Validate double entry
- Survey records are entered by two different data
entry operators - Use Records Validation through double data
entry menu to compare survey data files - When completed, both data files are compared by
linking the patient ID - When all records are exactly the same, both
records are considered to be entered correctly - When records differ, both records have to be
compared with the survey form and be corrected. - Run validate again until both data files show no
differences.
48 of 72
49Data analysis
49 of 72
50Prevalence reports
- Sample prevalence
- Prevalence as calculated from findings in the
survey - Age and sex adjusted prevalence
- Compare age and sex composition of sample differs
with actual population in the survey area. - Population file gives population by sex and
5-year age group in survey area. - If different, prevalence is adjusted by software
50 of 72
5151 of 72
5252 of 72
5353 of 72
54Cataract Surgical Coverage
- Objective
- - coverage indicator
- - measure accessibility utilisation of
services - - measures actual field situation
- - independent of reporting system
- CSC eyes
-
- Where a no. (pseudo)aphakic eyes
- b no. eyes with operable cataract
- CSC persons
- Where x persons with bilateral
(pseudo)aphakia - y persons with 1 (ps)aphakic and 1 operable
cataract eye - z persons with bilateral operable cataract
a a b
x y x y z
54 of 72
55Age and sex adjusted prevalence
55 of 72
56Barriers to cataract surgery
- Ask open question
- Why have you not been operated yet?
- List of pre-defined barriers
- Interviewer selects 1-2 barriers from list that
reflect best the patients reply - Analysis of barriers by
- Males, females
- Bilateral and unilateral blindness and SVI due to
cataract (VAlt3/60 VAlt6/60)
56 of 72
57Barriers
57 of 72
58Barriers to cataract surgery
58 of 72
59Visual outcome after cataract surgery
59 of 72
60Long-term visual outcome after cataract surgery
- with IOL non-IOL couching
- with available correction best corrected or
pinhole vision - during last 5 years more than 5 years ago
- outcome by place of surgery
- major causes of poor outcome
- age at time of surgery
- satisfaction
- causes of poor outcome after surgery
- type of surgery by sex
60 of 72
61WHO guidelines on Visual Outcome of Cataract
Surgery
61 of 72
62Best corrected vision after 1 year in clinical
trials
62 of 72
63Average visual outcome in population based
studies
63 of 72
64Visual outcome after cataract surgery in
population based surveysPlease note variation in
- Post-operative period varies from weeks to
decades - Quality of surgical facilities (basic to
excellent) - Experience and skills of surgeons (couchers)
- Supply and replacement of spectacles
- Initial good outcome may go down due to other eye
disorders, reducing vision with age - Outcome data from surveys may not do justice to
recent advancements in IOL surgery, but may very
well reflect what the public sees and what
determines their expectations and trust to regain
sight after surgery
64 of 72
65Also . . . .
- Good outcome will motivate other patients to come
forward for surgery - Poor outcome will deter other cases
- In most surveys fear of losing sight was major
reason not to come for surgery - When causes of poor outcome are known, it will be
possible to address these causes and thereby
improve results of cataract surgery
65 of 72
6666 of 72
67Service indicators
- Age at time of surgery
- Place of surgery, by sex
- Costs of services provided, by sex
- Use of spectacles, by sex
- Type of surgery (IOL non-IOL)
- Cause of poor outcome
- Satisfied with results of operation
67 of 72
6868 of 72
69Tables and graphs
69 of 72
70Sampling error and Design effect
70 of 72
71RACSS / RAAB conducted in
- India, 1995 onwards
- Paraguay, 1999
- Turkmenistan, 2000
- Vietnam, 2000
- Pakistan, 2000
- Myanmar, 2001
- Cambodia, 2001
- Mauretania, 2001
- Mali, 2001
- Guadalajara, Mexico, 2001
- Peru, 2002
- BA, Argentina, 2003
- Campinas, Brazil, 2003
- Nigeria,
- Venezuela, 2004
- Guatemala, 2004
- Lombok, Indonesia, 2004
- Cuba, 2004
- Nakuru, Kenya, 2005
- Philippines, 2005, 2006
- Colombia, 2005
- Bangladesh, 2005
- Botswana, 2006
- Nuevo Leon, Mexico, 2005
- Rwanda, 2006
- Bio Bio, Chile, 2006
- Kunming, China 2006
- Laos, 2006
- Kericho, Kenya, 2007
- Vietnam, 2007
- Zanzibar, 2007
- Jiangxi, China, 2007
- Cambodia, 2007
71 of 72
72Thank you!
72 of 72