Organizing the hospital for high volume

1
Rapid Assessment of Avoidable Blindness (RAAB
5) Dr. Hans Limburg
1 of 102
2
What is RAAB 5?
2 of 102
3
What is RAAB?

• Rapid Assessment of Avoidable Blindness
• Population-based survey on blindness and visual
  impairment

3 of 102
4
Why do we need surveys?

• Baseline data for planning
• Follow-up data to monitor progress of ongoing
  blindness intervention

4 of 102
5

Population based surveys

• Conventional population based blindness surveys
  expensive
• Conventional blindness surveys give national or
  province level data
• If used in large populations (gt50 million) less
  useful for district planning

5 of 102
6
Causes of blindness incl. URE
37 million blind
6 of 102
Foster A. et al. Changing patterns in global
blindness. Community Eye Health Journal.
20082137-39
7
What is the aim of RAAB?

• provide baseline indicators for planning and
  monitoring over time
• sound epidemiological methodology
• simple, cheap and quick procedure
• basic ophthalmic examination
• carried out by local staff
• used again after 8-12 years to measure how much
  has changed (interventions!)
• At district level (ideally 0.5-5 million pop.)

7 of 102
8
Indicators used

• prevalence of all blindness, severe visual
  impairment (SVI) and moderate visual impairment
  (MVI)
• main causes of blindness, SVI and VI
• prevalence 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
• prevalence of diabetic retinopathy (optional)
• uncorrected refractive error
• age at time of surgery, place of surgery, type of
  surgery, costs, cause of poor outcome

8 of 102
9
Why is it Rapid?

• Restrict to people aged 50
• High prevalence ? low sample size
• Standard methodology
• Enumeration and examination in one visit
• Basic eye examination
• Special software
• Calculate sample size
• Random selection of clusters
• Inter-observer variation assessment
• Simple data entry
• In-built error checks
• Automatic data analysis comparable results

9 of 102
10
Focus 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 cooperation

10 of 102
11
The Gambia Blindness by age(Faal H, Minassian
DC, Dolin PJ, et al. Evaluation of a national eye
careprogramme re-survey after 10 years. Br J
Ophthalmol. 200084948951)
85 of blindness in people aged 50
11 of 102
12
The Gambia Causes of blindness
Dineen B, Foster A, Faal H.A Proposed Rapid
Methodology to Assess the Prevalence and Causes
of Blindness and Visual Impairment. Ophthalmic
Epidemiology. 20061331-34 Causes in people 50
reflect causes in total population
Cause of blindness Total population Population 50
Cataract 46 48
Aphakia 13 15
Trachoma/CO 22 17
Glaucoma 9 11
Other 11 9
12 of 102
13
Sample size calculation (Madan Mohan, Survey of
Blindness India. Summary results, New Delhi
1989)
Sample size increased 70 to gain 5 more cases
Age group Population (million) Prevalence blindness No. blind Sample size
40-49 50-59 60-69 70 70.9 48.8 29.6 13.7 0.3 2.0 5.9 10.4 222,000 952,000 1,767,000 1,290,000

Total 40 Total 50 163.3 92.3 2.6 4.3 4,231,000 4,010,000 5,733 3,371
13 of 102
14
RA versus conventional Survey

• Blindness survey
• Focus gt1 risk group
• lower prevalence
• Sample size gt15,000
• Detailed examination
• Disease intervention
• Expert staff
• Large survey population (10-100 mln.)
• Custom data analysis
• Takes long (years)
• Expensive (0.5-10 mln)

• Rapid assessment
• Focus on 1 risk group
• higher prevalence
• Sample size 2500-5000
• Basic examination
• Planning and follow-up
• Local staff
• Smaller survey population (0.5-5 mln.)
• Automatic data analysis
• Rapid (months)
• Cheap (20-40,000 US)

14 of 102
15
Avoidable 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

15 of 102
16
Where surveys were done
RAAB
RACSS
Custom survey
16 of 102
17
History

• 1994 District Rapid Assessments developed in
  India decentralised eye care services
• 2000 Modified into Rapid Assessment of Cataract
  Surgical Services (RACSS) for WHO
• 2005 Modified to Rapid Assessment of Avoidable
  Blindness (RAAB) with focus on VISION 2020
  district level planning
• 2013 RAAB 5 with reports on URE and optional DR
  module

17 of 102
18
Where to conduct RAAB?

• Total population ideally 0.5 - 5 million
• Management structure for eye care
• Population composition by gender and by 5-year
  age groups available
• Population by sub-unit (enumeration area,
  village, town, neighbourhood, polling station,
  etc.) available
• Detailed maps available
• Entire area is accessible for survey teams
• No problems with security

18 of 102
19
Sampling
19 of 102
20
Sampling methods

• Simple Random Sampling
• Same chance of selection for every individual
• List of all individuals 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)

20 of 102
21
Simple random sampling

• Principle
• Equal chance for each unit (person 50) to be
  selected
• Procedure
• Make a list of all people aged 50 in the survey
  area and number them
• Randomly select individuals aged 50
• No stratification possible (e.g. male/female,
  urban/rural, etc.)

21 of 102
22
How can we choose persons aged 50 in this
district through simple random sampling?
Population list may not be available
People scattered inefficient!
22 of 102
23
Cluster sampling

• Principle
• Equal chance for each cluster to be selected
• Procedure
• Make list (sampling frame) of smallest
  population units (census enumeration units,
  election rolls, GP practices, etc.)
• Randomly select population units
• In selected population unit, 50 residents aged
  50 are selected at random for examination

23 of 102
24
Multi-stage sampling of clusters
3
2
1
Do not need population list
People together efficient!
24 of 102
25
Sampling villages from a list
Village A B C D E F G H I J Pop size 1,600 200 3,200 400 800 200 1,200 200 1,600 600
Total population 10,000 Total population 10,000 Total population 10,000
With random sampling the small villages will be
over-represented
25 of 102
26
Cluster sampling
Compute cumulative population
Village A B C D E F G H I J Pop size 1,600 200 3,200 400 800 200 1,200 200 1,600 600 Cumulative 1,600 1,800 5,000 5,400 6,200 6,400 7,600 7,800 9,400 10,000
Total population 10,000 Total population 10,000 Total population 10,000
26 of 102
27
Cluster sampling

• Example
• Total cumulative population 10,000
• Need to sample 20 clusters of 50 people
• Sampling interval 10,000/20 500
• Select random number between 1 and 500
• say 300
• Start with village containing individual 300 and
  select 50 people from that village
• Next cluster to select
• starting point sampling interval
• 300 500 800
• Write down 20 clusters to be chosen

27 of 102
28
Cluster sampling
Village A B C D E F G H I J Pop size 1,600 200 3,200 400 800 200 1,200 200 1,600 600 Cumulative 1,600 1,800 5,000 5,400 6,200 6,400 7,600 7,800 9,400 10,000
Clusters 1,2,3 4 5,6,7,8,9,10 11 12 13 14,15 16 17,18,19 20
1 - 300 2 - 800 3 - 1,300 4 - 1,800 5 - 2,300 6 - 2,800 7 - 3,300 8 - 3,800 9 - 4,300 10- 4,800 11 - 5,300 12 - 5,800 13 - 6,300 14 - 6,800 15 - 7,300 16 - 7,800 17 - 8,300 18 - 8,800 19 - 9,300 20 - 9,800

Sampling with probability proportionate to size
28 of 102
29
Cluster sampling

• Requires a sampling frame of very recent
  population data
• Preferably census data (detailed maps)
• Alternatives, e.g. electoral role, list of GP
  practices.
• Table with composition of population in survey
  area by 5-year age group and gender

29 of 102
30
Calculate sample size (S)

• For Simple random sampling
• Estimated prevalence of condition (P)
• 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))

30 of 102
31
Sample 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 1.4
• Cluster size 50 gt DEFF 1.5
• Cluster size 60 gt DEFF 1.6
• 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

31 of 102
32
Calculate sample size
32 of 102
33
Sampling frame

• - census enumeration areas with population, or
• - list of settlements with population, or
• other list of geographic distribution of total
  population
• format Excel 97/2003

34 of 102
34
Random selection of clusters with probability
proportional to size of population in RAAB
software
35 of 102
35
Enumeration areas
36 of 102
36
37 of 102
37
38 of 102
38
Google Earth!
39 of 102
39
How to select the required number of people aged
50 in the population unit

• 1. Random walk
• 2. Compact segment sampling

40 of 102
40
Start 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
41 of 102
41
Compact segment sampling

• Divide total population in cluster 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)
• e.g. for Sierra Leone
• Cluster size / population 50 (50 / 10 500)

42 of 102
42
Total 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
43 of 102
43
Compact 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

44 of 102
44
Compact 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

45 of 102
45
Multistage 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

33 of 102
46
46 of 102
47
47 of 102
48
48 of 102
49
49 of 102
50
Examination pocedure
50 of 102
51
Examination procedures

• 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

51 of 102
52
Definitions (WHO)

• Blindness
• VA lt3/60 with available 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 available correction in the
  better eye or visual field of 20 degrees or less,
  but more than 10 degrees
• Moderate visual impairment
• VA lt6/18 6/60 with available correction in the
  better eye or visual field of 30 degrees or less,
  but more than 20 degrees

52 of 102
53
Survey form
53 of 102
54
Equipment 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

54 of 102
55
Comparing findings of RAAB in 2005 and the 2010
study on posterior segment eye diseases
(Mathenge W, Bastawrous A, Foster A, Kuper H.
The Nakuru Posterior Segment Eye Disease Study -
Methods and Prevalence of Blindness and Visual
Impairment in Nakuru, Kenya. Ophthalmology
201211920332039)Conclusions This survey
provides reliable estimates of blindness and VI
prevalence in Nakuru.... This survey validates
the use of RAAB as a method of estimating
blindness and VI prevalence.It is also strongly
suggestive that the RAAB methodology being used
throughout Africa and worldwide is a robust and
reliable methodology.
55 of 102
56
56 of 102
57
Inter-observer variation
57 of 102
58
Inter-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 50 people
• 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.

58 of 102
59
Inter-observer variation
Each team examines 40-50 people aged 50, e.g.
patients and their company from the outpatient
department
59 of 102
60
Inter-observer variation
60 of 102
61
61 of 102
62
62 of 102
63
Report on inter-observer variation
63 of 102
64
Field work
64 of 102
65
65 of 102
66
66 of 102
67
67 of 102
68
68 of 102
69
69 of 102
70
70 of 102
71
71 of 102
72
Data entry
72 of 102
73
73 of 102
74
Cleaning 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

74 of 102
75
Consistency checks
75 of 88
76
Consistency 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.

76 of 102
77
Validate double data 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.

77 of 102
78
Data analysis
78 of 102
79
Data analysis
79 of 102
80
Survey Design
persons of all ages in survey area
21,000
persons 50
3,000
persons 50 with operable cataract or
(pseudo)aphakia
100
Blind SVI MVI male female
50
50
(pseudo)aphakia
cataract
Cataract Surgical Coverage
Outcome
Barriers
80 of 102
81
Prevalence reports

• Sample prevalence
• Prevalence as calculated from findings in the
  survey
• Age and sex adjusted prevalence
• Compare age and sex composition of sample 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

81 of 102
82
82 of 102
83
83 of 102
84
84 of 102
85
Cataract 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
85 of 102
86
Age and sex adjusted prevalence
86 of 102
87
Barriers 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)

87 of 102
88
Barriers
88 of 102
89
Barriers to cataract surgery
89 of 102
90
Visual outcome after cataract surgery
90 of 102
91
Long-term visual outcome after cataract surgery

• with IOL non-IOL
• with available correction best corrected or
  pinhole vision
• during last 3 years 4-6 years gt6 years
• outcome by place of surgery
• major causes of poor outcome
• age at time of surgery
• type of surgery by sex

91 of 102
92
WHO guidelines on Visual Outcome of Cataract
Surgery
WHO Working Group - 1998 - Informal consultation
on analysis of prevention of blindness
outcomes. Geneva, 1998. WHO/PBL/98.68
Post-operative acuity Available correction Best correction
Good 6/18 gt80 gt90
Borderline lt6/18 6/60 lt15 lt5
Poor lt6/60 lt5 lt5
92 of 102
93
Best corrected vision after 1 year in clinical
trials
93 of 102
94
Average visual outcome in population based
studies
94 of 102
95
Visual outcome after cataract surgery in
population based surveys.Please 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

95 of 102
96
Also . . . .

• 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

96 of 102
97
97 of 102
98
Service indicators

• Age at time of surgery
• Place of surgery, by sex
• Costs of services provided, by sex
• Type of surgery (IOL non-IOL)
• Cause of poor outcome

98 of 102
99
99 of 102
100
Tables and graphs
100 of 102
101
Sampling error and Design effect
101 of 102
102
RACSS and RAAB 1999 - 2009
RACSS RAAB
1995 India 50 2000 Vietnam 8 2000 Cambodia
3 2002 Pakistan 2004 Indonesia 2 2004 Myanmar
3 2005 Philippines 2005 Bangladesh 2006 China
2007 Vietnam 16 2007 Cambodia 2007 Laos 2007
China 2008 China 2008 Nepal 3 2008
Russia 2007 Iraq 2008 Quatar 2008 Yemen 2010
Myanmar 2010 China

• 1999 Paraguay
• 2002 Peru
• 2003 Argentina
• 2003 Brazil
• 2004 Venezuela
• 2004 Guatemala
• 2004 Cuba
• 2006 Mexico
• 2007 Chile
• 2008 Colombia
• 2008 Dominican Rep.
• 2008 Ecuador
• 2008 Paraguay

2001 Turkmenistan 2001 Mauretania 2001
Mali 2003 Nigeria 2005 Kenya 2006
Rwanda 2007 Kenya 2007 Kenya 2007 The Gambia
2007 Tanzania 2007 Zanzibar 2007 Uganda 2008
Palestine 2008 Mali 2008 Eritrea 2008
Sudan 2010 Zambia 2010 Sierra Leone
101 of 102
103
Thank you!
102 of 102