Analysis and interpretation of data

1
Analysis and interpretation of data

• IDSP training module for state and district
  surveillance officers
• Module 9

2
Learning objectives

• Identify the role, importance and techniques of
  data analysis
• Sources and management of data for valid
  conclusions
• Choose appropriate descriptive and analytical
  methods
• List outcome measures for feedback
• Generate reports with tables and graphs

3
All levels must analyze surveillance data

• Health workers
• Increase of cases
• Medical officers in primary health centres
• Outbreak detection
• Seasonal trends
• District surveillance officers
• All of the above
• Advanced analyses

4
Selected outcomes of data analysis

• Identification of outbreaks / potential outbreaks
• Identification of appropriate and timely control
  measures
• Prediction of changes in disease trends over time
  
• Identification of problems in health systems
• Improvement of the surveillance system through
• Identification of regional differences
• Identification of differences between the private
  and the public sectors
• Identification of high-risk population groups

5
Sources of data

• Sub-Centre
• Primary health centre
• Community health centre
• District
• Private practitioners
• Private nursing homes
• Identified laboratories
• Medical colleges
• Police departments
• State

6
Types of data

• Syndromic case data
• Presumptive case data
• Confirmed case data
• Sentinel case data
• Regular surveillance data
• Urban data
• Rural data

7
Periodicity of data collection

• Weekly
• High priority (Acute flaccid paralysis)
• As soon as a case is detected
• Data on outbreaks are collected and analyzed
  separately

8
Analysis of data at the district surveillance unit

• Computer software provides ready outputs
• District surveillance officer prepares a report
• Technical committee reviews and needs to bear in
  mind
• The strength and weakness of data collection
  methods
• Reliability and validity of data
• The separate disease profiles
• The user-friendliness of graphs
• The need to calculate rates before comparisons

9
What computers cannot do

• Skills
• Contact reporting units for missing information
• Interpret laboratory tests
• Make judgment about
• Epidemiologic linkage
• Duplicate records
• Data entry errors
• Declare a state of outbreak

• Attitudes
• Looking
• Thinking
• Discussing
• Taking action

10
Expressed concerns versus reality

• Concerns commonly expressed
• Statistics are difficult
• Multivariate analysis is complex
• Presentation of data is challenging

• Mistake commonly observed
• Data are not looked at

11
Basic surveillance data analysis

• Count, divide and compare
• Direct comparisons between number of cases are
  not possible in the absence of the calculation of
  the incidence rate
• Descriptive epidemiology
• Time
• Place
• Person

12
1. Count, Divide and Compare (CDC)

• Count
• Count cases that meet the case definition
• Divide
• Divide cases by the population denominator
• Compare
• Compare rates across
• Age groups
• Districts
• Etc.

13
2. Time, place and person descriptive analysis

• Time
• Graph over time
• Place
• Map
• Person
• Breakdown by age, sex or personal characteristics

14
A. Analysis over time

• Absolute number of cases
• Does not allow comparisons
• Analysis by week, month or year
• Incidence
• Allows comparisons
• Analysis by week, month or year

15
Acute hepatitis (E) by week, Hyderabad, AP,
India, March-June 2005
Absolute number of cases per week
120
100
80
Number of cases
60
40
20
0
1
8
15
22
29
4
12
19
26
3
10
17
24
31
7
14
21
28
March April
May
June
First day of week of onset
Interpretation The source of infection is
persisting and continues to cause cases
16
Reported varicella and typhoid cases, Darjeeling
district, West Bengal, India, 2000-4
Incidence by year
Interpretation The parallel increase between
varicella (that should be constant) and typhoid
suggests that increasing rates of typhoid are
secondary to improved reporting
17
2. Analysis by place

• Number of cases by village or district
• Does not control for population size
• Spot map
• Incidence of cases by village or district
• Controls for population size
• Incidence map

18
Reported cases of measles, Cuddalore district,
Tamil Nadu, Dec 2004 Jan 2005
Spot map of absolute number of cases
Annagraman
Interpretation Cases were reported from tsunami
affected non-affected areas, thus the cluster was
not a consequence of the tsunami
Cuddalore
Panruti
Parangipattai
Kurinjipadi
Vridha-chalam
Kamma-puram
Bhuvanagiri
Mangalore
Keerapalayam
Nallur
Kumaratchi
Kattumannar Kail
19
Incidence of acute hepatitis (E) by block,
Hyderabad, AP, India, March-June 2005
Incidence by area
Attack rate per100,000 population
0
1-19
20-49
50-99
100
Open drain
Interpretation Blocks with hepatitis are those
supplied by pipelines crossing open sewage drains
Pipeline crossing open sewage drain
20
3. Analysis per person

• Distribution of cases by
• Age
• Sex
• Other characteristics(e.g., Ethnic group,
  vaccination status)
• Incidence by
• Age
• Sex
• Other characteristics

21
Distribution of cases according to a
characteristic
Immunization status of probable measles cases,
Nai, Uttaranchal, India, 2004
19
81
Immunized
Unimmunized
Interpretation The outbreak is probably caused
by a failure to vaccinate
22
Probable cases of cholera by age and sex,
Parbatia, Orissa, India, 2003
Incidence according to a characteristic
Nu
m
b
e
r
of
c
a
s
es
Po
pu
l
a
t
i
on
I
nc
i
d
e
nc
e
0
t
o4
6
1
1
3
5
.
3

A
g
e g
r
o
up
(
In
y
e
ar
s
)
5
t
o1
4
4
1
9
0
2
.
1

1
5
to
2
4
5
1
2
8
3
.
9

2
5
to
3
4
5
1
4
4
3
.
5

3
5
to
4
4
6
1
2
9
4
.
7

4
5
to
5
4
4
8
8
4
.
5

5
5
to
6
4
8
6
7
1
1
.
9

gt
6
5
3
8
7
3
.
4

M
a
l
e
1
7
4
8
1
3
.
5

S
ex
F
e
m
a
l
e
2
4
4
6
5
5
.
2

Tot
al
T
ot
a
l
4
1
9
4
6
4
.
3

Interpretation Older adults and women are at
increased risk of cholera
23
Seven reports to be generated

• Timeliness/completeness
• Description by time, place and person
• Trends over time
• Threshold levels
• Compare reporting units
• Compare private / public
• Compare providers with laboratory

24
Report 1 Completeness and timeliness

• A report is said to be on time if it reaches the
  designated level within the prescribed time
  period
• Reflects alertness
• A report is said to be complete if all the
  reporting units within its catchment area
  submitted the reports on time
• Reflects reliability

25
Interpretation of timeliness and completeness
26
Report 2 Weekly/ monthly summary report

• Based upon compiled data of all the reporting
  units
• Presented as tables, graphs and maps
• Takes into account the count, divide and compare
  principle
• Absolute numbers of cases and deaths are
  sufficient for a single reporting unit level
• Incidence rates are required to compare reporting
  units

27
Epidemiological indicators to use in weekly /
monthly summary report

• Cases
• Deaths
• Incidence rate
• Case fatality ratio

28
Report 3 Comparison with previous weeks/ months/
years

• Help detect trend of diseases over time
• Weekly analysis compare the current week with
  data from the last three weeks
• Alerts authorities for immediate action
• Monthly and yearly analysis examine
• Long term trends
• Cyclic pattern
• Seasonal patterns

29
Acute hepatitis by week of onset in 3 villages,
Bhimtal block, Uttaranchal, India, July 2005
Example of weekly analysis
90
80
70
60
50
Number of cases
40
30
20
10
0
1st week
3rd week
1st week
1st week
1st week
1st week
3rd week
4th week
3rd week
4th week
3rd week
4th week
2nd week
4th week
2nd week
2nd week
2nd week
May
June
July
August
September
Week of onset
Interpretation The second week of July has a
clear excess in the number of cases, providing an
early warning signal for the outbreak
30
Malaria in Kurseong block, Darjeeling District,
West Bengal, India, 2000-2004
Example of monthly and yearly analysis
45
40
Incidence of malaria
35
Incidence of Pf malaria
30
25
Incidence of malaria per 10,000
20
15
10
5
0
July
July
July
July
May
July
April
May
April
May
April
May
April
May
April
June
June
June
June
June
March
March
March
March
March
August
August
August
August
August
January
October
January
October
January
October
January
October
January
October
February
February
February
February
February
November
December
November
December
November
December
November
December
November
December
September
September
September
September
September
2000
2001
2002
2003
2004
Months
Interpretation There is a seasonality in the end
of the year and a trend towards increasing
incidence year after year
31
Report 4 Crossing threshold values

• Comparison of rates with thresholds
• Thresholds that may be used
• Pre-existing national/international thresholds
• Thresholds based on local historic data
• Monthly average in the last three years
  (excluding epidemic periods)
• Increasing trends over a short duration of time
  (e.g., Weeks)

32
Report 5 Comparison between reporting units

• Compares
• Incidence rates
• Case fatality ratios
• Reference period
• Current month
• Sites concerned
• Block level and above

33
Interpretation of the comparison between
reporting units
34
Report 6 Comparison between public and private
sectors

• Compare trends in incidence of new cases/deaths
• Incidences are not available for private provider
  since no population denominators are available
• Good correlation may imply
• The quality of information is good
• Events in the community are well represented
• Poor correlation may suggest
• One of the data source is less reliable

35
Report 7 Comparison of reports between the
public health system and the laboratory
36
Frequency of reports and analysis
37
Review of analysis results by the technical
committee

• Meeting on a fixed day of every week
• Review of a minimum of
• 4 reports weekly
• 7 reports monthly
• Review by disease wise
• Search for missing values
• Check the validity
• Interpret
• Prepare summary reports and share
• Take action

38
Limitations in analysis of surveillance data

• The quality of data may be problematic
• Poor use of case definition
• Under-reporting
• There may be a time lag between detection,
  reporting and analysis
• Under-reporting occurs
• However, if the level of under-reporting is
  constant, trends may still be analyzed and
  outbreaks may still be detected
• The representativeness may be poor
• Engage the private sector to diversify reporting
  sources

39
Conclusion

• Analysis is a major component of surveillance
  links data collection and program implementation
• While it is important to analyze data, its also
  important that analyzed reports are sent to the
  appropriate authorities
• Higher level
• Lower level

40
Points to remember

• Surveillance data identifies outbreaks and
  describe conditions by time, place and person
• Surveillance helps monitor disease control and
  assess the impact of services
• Data analysis must occur at each level
• Analyzed data is presented in tables, graphs with
  comparisons with previous data