1026 Int Diploma IA2 v2

Element IA2: Loss Causation and Incident Investigation


© Santia 2013 ® – restricted use only

Page 1 sc/1026/v2



Page 2 sc/1026/v2



Page 2 sc/1026/v2


Contents Theories of Loss Causation Denitions Accident Studies Studies Accident Causation Causation Models Models

Quantitative Analysis of Accident and Ill-health Data Accident and Ill-health Statistics Statistics Epidemiology Presentation and Interpretati Interpretation on of Data

Reporting and Recording of Accidents, Ill-Health and Dangerous Occurences Denitions

Loss Investigations

5 5 6 7 17 17 21 25

33 34

Procedures Procedure s for Investigating Accidents and Ill-Health

36 38

References

46


Page 3 sc/1026/v2



Page 4 sc/1026/v2


Theories of Loss Causation It is important to be able to understand the performance of an organisation in relation to historical events. Knowledge of undesirable events or circumstances which can lead to injury or ill-health in the workplace is of particular value to organisations when setting out their safety management systems and risk control measures to prevent further undesirable occurrences. The terminology used in dening such events and recording the information gained from accident and incident investigation should be clearly understood by all those involved in the organisation.

Denitions Accident A useful denition of an accident is ‘an unplanned, unwanted event which results in loss’. Loss is not conned to personal injury, other examples being loss of business revenue or opportunity. ILO accident denitions include: ▪

Occupational accident - an occurrence arising out of or in the course of work that results in a fatal or non-fatal occupational injury.

▪

Occupational injury - death, any personal injury or disease resulting from an occupati onal accident.

▪

Commuting (travel) accident - an accident resulting in occupational injury involving loss of working time occurring on the direct way between the place of work and: ▪

The worker’s principal or secondary residence;

▪

The place where the worker usually takes his / her meals; or

▪

The place where the worker usually receives his / her remuneration.

Incident The above denition of an accident does not include those events that could have caused loss. Such events are often called incidents or near misses. A useful denition of an incident is ‘an unplanned, unwanted event that has the potential to result in loss’. The ILO OSH 2001 standard describes an incident as; “An unsafe occurrence arising out of or in the course of work where no personal injury is caused.”


Page 5 sc/1026/v2


Accident Studies Whatever term is used it is purely chance that prevents an incident from becoming an accident. The severity of injuries or damage resulting from an accident is also often a matter of chance.

Accident Triangles There have been many studies carried out to try to establish a relationship between accidents and incidents. These are illustrated by accident triangles. The two most famous studies of accidents are by Heinrich 1931 and Bird 1969. Bird took a larger sample size than Heinrich and therefore his work is deemed to be the more reliable. Figure 1: The Heinrich Accident Triangle 1931

1

29

300

Figure 2: The Bird Accident Triangle 1969

1

10

30

600

From these accident triangles it can be concluded that incidents are the foundation of major injuries. Major accidents can be avoided and the severity of injury can be reduced through reporting and investigating near misses. The theory of loss control is built around these studies. As the studies also demonstrate that the majority are near misses or accidents resulting in damage, the greatest benet from a loss control programme can therefore be achieved by focussing on these areas.


Page 6 sc/1026/v2


If incidents are left uncontrolled they may well develop into major accidents; however none of these studies attempted to count the number of situations (actions or conditions) that had the potential to result in injury. Such situations are called hazards. There are always valuable lessons to be learned by investigating accidents, but if uncontrolled hazards are ignored, they could cause accidents which could have been prevented.

Accident Case Studies Previous disasters and quotations from their inquiry reports illustrate the importance of robust management systems in accident prevention. See Disaster Summaries.

Accident Causation Models Single Cause Approach Heinrich Domino Theory According to Heinrich, “ A preventable accident is one of ve factors in a sequence that results in an injury. The injury is invariably caused by an accident and the accident in turn is always the result of the factor that immediately preceded it”. Figure 3: The Heinrich Domino Sequence

A

B

C

D E

A B C D E

= = = = =

Ancestry and social environment Fault of the person Unsafe action or condition Accident Injury


Page 7 sc/1026/v2


Ancestry and Social Environment Undesirable character traits, such as recklessness, stubbornness, avariciousness, violent temper, nervousness and excitability may be inherited. A person’s social environment may reinforce existing, and promote new, undesirable traits or interfere with his / her education. Both inherited traits and environmental factors may cause faults of the person.

Fault of the Person A person’s inherited or acquired faults may lead to unsafe acts, such as horseplay, standing under danger areas, careless starting of machines and removal of safeguards. Individuals may also allow unsafe conditions, such as mechanical or physical hazards from operating machinery with guards removed, to exist and persist.

Unsafe Action or Condition Acts of the individual or conditions may arise which are unsafe e.g. operating a machine without a guard, oil spillage not cleaned up.

Accident Unsafe performance may develop into events that cause injury, e.g. falls from height, slips and trips, etc.

Injury This is the resultant loss from an accident, e.g. death, disablement, fractures, lacerations, etc. Heinrich believed that the rst two dominoes could be removed through a lengthy period of education aimed at changing attitudes. Changing an individual’s attitude towards health and safety is required in an accident prevention programme; however it may take time and will not produce immediate results. Removing dominoes 4 and 5 occurs too late in the domino sequence; therefore Heinrich suggested that the accident prevention programme should concentrate on domino 3, unsafe actions and conditions.


Page 8 sc/1026/v2


Bird and Loftus Domino Theory Bird and Loftus were concerned that the Heinrich domino theory placed all the blame upon an individual and his / her background without consideration of the inuence of management. They also suggested that by focussing only on unsafe acts and conditions in order to prevent accidents management action was limited. Bird and Loftus modied the domino sequence as illustrated in the following gure: Figure 4: The Bird and Loftus Domino Sequence

A

B

C

D E

A B C D E

= = = = =

Lack of management control Basic causes (personal and job factors) Immediate causes (unsafe acts and conditions) Accident Loss A and B = Underlying causes

The Bird and Loftus sequence suggests that a lack of management control permits basic causes (personal and job factors), leading to immediate causes (e.g. substandard practices, conditions or errors), which are the direct cause of the accident, which results in loss. Loss may be categorised as negligible, minor, serious or catastrophic.

Lack of Management Control Safety should be a paramount consideration throughout an organisation. Managers must manage and monitor safety in the same way as other corporate functions. Safety programmes must incorporate systematic consideration of technical, organisational and behavioural safeguards. Behavioural safeguards should not only relate to those persons whose errors may lead directly to accidents, but also to the culture of the organisation and the scope of decisions by senior members of staff.

Basic Causes - Personal & Job Factors Personal Factors These are factors particular to the individual which can cause them to behave in a certain way. These may include lack of knowledge, skill, information, instruction; physical unsuitability; misperception; poor motivation; perceived or actual pressure from fellow colleagues to behave in a certain way; mistaken priorities; or actions and lapses of attention from boredom, fatigue and distractions.


Page 9 sc/1026/v2


Job Factors These factors are associated with the job or task the person is carrying out and may include factors such as incorrect or damaged equipment supplied for the job; unsuitable timescales allocated for the job or pressure to complete the job quickly; poor ergonomic design of the equipment or workstation; poor working environment; lack of supervision or low priority placed on safety.

Immediate Causes – Unsafe Acts and Conditions These behaviours of the individual are violations or errors. The UK HSE Guidance publication ‘Reducing error and inuencing behaviour’ HSG 48 denes a violation as; “…..a deliberate deviation from rules, procedure, instructions or regulations”. And errors as; “…unintended actions or decisions involving a deviation from an accepted standard, including routine violations, which lead to an undesirable outcome…” The reasons for these violations or errors can be a combination of personal and / or job factors. As a result, the individual may act in a certain way, or conditions may arise, which are unsafe e.g. operating a machine without a guard, press ing the wrong button on a control panel believing it to the correct one, oil spillage not cleaned up.

Accident Unsafe performance may develop into events that cause injury, e.g. falls from height, slips and trips, etc.

Injury This is the resultant loss from an accident, e.g. death, disablement, fractures, lacerations, etc. The Bird and Loftus sequence is important to the loss control programme as it can be applied to all accidents. The model has limitations in that each single factor is considered in relationship to the cause preceding it and the event following it, whereas in practice many causes may occur at the same time and the accident may be a result of a combination of factors.


Page 10 sc/1026/v2


Multi-Causation Approach A serious accident normally has multiple causes with chains and combinations of events. Rather than using a single chain approach, the multi-causati on approach allows for consideration of the various immediate causes and underlying sub-causes of an accident. With this approach the numerous contributory factors can be identied and examined. Figure 5: The Multi-Cause Theory

Lack of Management Control Basic Causes 1 2

Unsafe Act

3 Accident

Injury

4 5

Unsafe Condition

6

Root Cause Analysis A more practical multi-causation approach to accident analysis is achieved by the development of root cause analysis. Root cause analysis is a simple ‘fault tree’ technique for identifying all of the direct causes and then analysing each of them in turn to nd out the underlying causes. More complex quantied fault trees are used in risk assessment. The ultimate aim is to nd out why the systems of management failed to prevent each of the underlying causes that contributed to the accident. This type of analysis allows a wide range of recommendations to be made to prevent recurrence. For example: Late for an appointment, Mr Smith slips on a patch of oil. The accident report form was completed as follows: “Oil cleaned up with granules, Mr Smith should take more care in future”. The cause of the injury has been identied; however this would not prevent recurrence.


Page 11 sc/1026/v2


Figure 6: Example of Root Cause Analysis The cause cause ofofthe accident The the injury

A man slips and strikes his head on the floor

The direct / immediate The of direct causes causes the accident

of the accident He was hurrying

There was oil on the floor

A1

B1

A2

A3

Lack of management control

B2

B3


The lighting was poor he did not see the oil

C1

C2

C3

Basic Causes


A thorough investigation might include questions about: ▪

Why Mr Smith was in such a hurry?

▪

Why were rules prohibiting running not known or not enforced?

▪

Was Mr Smith aware that running in the workplace is unsafe?

The source of the oil spillage was found to be a fork-lift truck; it would be useful to know: ▪

Why it was leaking?

▪

Why the leak had not been rectied?

▪

Why the spill had not been cleaned up before the accident?

Why was the lighting poor? ▪

Were there an adequate number of lights?

▪

Were they switched on?

▪

Were they properly maintained?

It can be seen that the questions examine the systems for maintenance, training, supervision and fault reporting, etc. Accident investigation should try to nd out why management systems failed to prevent the accident rather than apportion blame. The domino theory of accident causation although simplistic can assist in structuring accident investigations. It has limitations in that considering just one chain of events (the linear approach) relates more to immediate causes rather than errors in planning, design, etc. and it is reactive rather than proactive, therefore cannot be used to predict likelihood of events. Whereas contributory causes in the multi-causal theory of accident causation combine in complex and random fashion.


Page 12 sc/1026/v2


Each contributory cause has its own chain or chains of events, which can be graphically represented by root cause (simple fault tree) analysis. As such it can be more complex or difcult to understand and require more time or resources than the linear domino theory approach. The multi-causal theory of accident causation can, unlike the linear approach, be used to predict the likelihood of events. However, both can be used to reduce their likelihood.

Reason’s Model of Accident Causation James Reason distinguishes organisational accidents from individual accidents noting that organisational accidents have multiple causes involving many people operating at different levels within the organisation. Reason’s basic thesis suggests that all organisational accidents entail the breaching of barriers and safeguards intended to keep damaging and injurious hazards away from vulnerable people or assets (potential losses). Figure 7: The relationship between hazards, defences and losses

The next question to consider is – how are the defences breached? Reason identies three sets of factors:- human, technical and organisati onal which are governed by two universal processes common to all technological organisations:▪

Production processes (e.g. manufacture of goods, extraction of raw materials, delivery of service) which expose people and assets to danger; and

▪

Protection processes of various types whose purpose is to protect people and assets from the danger arising from production processes.

Reason considered that defences could be categorised by function or by means of achieving that function. The term “defences in depth” was coined to explain the concept of successive layers of protection guarding against the possible failure of the one in front.


Page 13 sc/1026/v2


Defences in depth can be implemented as follows: ▪

Creation of hazard awareness and understanding;

▪

Clear guidance on safe operation;

▪

Alarms and warning of imminent danger;

▪

Restoring systems to a safe state in off-normal situations;

▪

Positioning barriers between hazards and potential losses;

▪

Eliminate or control hazards that escape the barrier; and

▪

Means of escape and rescue should hazard containment fail.

In an ideal world, the layers of defences in depth prevent the hazard from adversely affecting people and assets. In the real world, each layer of defence will have gaps or weaknesses. These are conceptualised as holes in Swiss cheese, in Reason’s model. Figure 8: The Ideal and Reality for Defences in Depth


Page 14 sc/1026/v2


Reason’s Swiss cheese metaphor is best conceptualised as a moving picture with each defence coming into and out of play depending on local conditions. The holes in each defence are also moving, shrinking or growing in response to operator actions and local needs. The holes in the Swiss cheese are created by active failures and/or latent conditions. ▪

Active failures are unsafe acts that have a direct and immediate effect on system safety. Typically they are errors and violations at the sharp end of the system such as opening the wrong valve or bypassing guards or interlocks.

▪

Latent conditions include: poor design, gaps in supervision, undetected manufacturing defects, training gaps and maintenance failures. Latent conditions may be present for many years before they combine with active failures and local circumstances to penetrate layers of defences. Latent conditions typically arise from strategic, top level decisions.

The necessary conditions for an organisational accident require a rare conjunction of a set of holes in successive defences permitting the hazard to reach and damage people or other assets. Figure 9: Accident trajectory passing through corresponding holes in layers of defence

Reason’s conceptualisation of the causal sequence of an organisational accident shows that organisational factors lead to the creation of error provoking local working conditions resulting in an individual (or team) unsafe act. If the defences subsequently fail there is a resultant loss. In an accident investigation the sequence of inquiry is reversed. Working back from the loss, consideration is rst given to determining how the defences failed and what active failures and latent conditions were involved. For each unsafe act the investigation seeks to determine what local conditions shaped or provoked it, and ultimately what upstream organisational factors contributed to the development of each local condition.


Page 15 sc/1026/v2


Figure 10: Stages in the development and investigation of an organisational accident.


Page 16 sc/1026/v2


Quantitative Analysis of Accident and Ill-health Data Accident and Ill-health Statistics Monitoring and review are essential elements of a safety management system. Every organisation should have objectives and a plan for continually improving health and safety performance. A measure of past performance will not normally be found from one indicator, a number of indicators will always be necessary. Reactive measures allow a number of performance indicators over a period of time to indicate the effectiveness of accident prevention strategies and hence develop the continual improvement plan. When the occurrence of accidents and incidents vary, they inuence the identication of trends and a measure of those parameters to determine various rat es of events, rather than measuring simple accident gures, can be used. Reducing gures over a period of time generally indicates an improvement in performance. The use of statistics should be carefully monitored, as they can be useful in the analysis of performance, however they do have limitations. Conning data to reportable injuries will not give a true indication of performance, as it will not reect the severity of the injury. Also a low reportable injury incidence rate will not necessarily mean good safety performance as it does not cover non-reportable injuries or damage incidents. Variations in health and safety cultures may affect the recording of non-reportable accidents as employees may be reluctant to report them and supervisors, who are held responsible for accidents, may discourage such reports. These limitations should not discourage managers from recording accident statistics, which are necessary for gauging, with interpretation, health and safety performance.


Page 17 sc/1026/v2


Accident Incidence Rate This is probably the most widely used formula, and is the number of dened accidents in a period per thousand employees (10 3). Number of dened accidents x 1,000 Average number employed When comparing and benchmarking incidence rates across various departments, sites or sectors, States or countries, it is important to compare like with like, i.e.: ▪

That they relate to the same type of accident, as denitions may vary;

▪

That the working hours are the same, for example if large amounts of overtime are worked the incidence rate will be increased when compared to when it is not; and

▪

That the denition of those employed is comparable, for example if part time staff are included in one calculation the incidence rate will be decreased when compared to when they are not.

Fatal Accident Incidence Rate This indicates the number of fatalities per 1000 (10 3) employees. Similar limitations for working hours and the average number employed as for accident incidence rate (above) apply. Although the denition of a fatal accident is reliable and, as fatalities occurring at the time of the accident remain difcult to hide, the number of fatalities is generally deemed to be accurate, however a potential limitation is the reporting of fatalities occurring within one year of an accident. Number of Fatalities Average number employed

x 1,000

Accident Frequency Rate This is the number of dened accidents in a period per 100,000 (10 5) person hours worked. In some organisations the incidence rate may not be a reliable guide as there may be many part time workers, long shifts and overtime working. There also may be wide changes of activity, hence risks, from one period to another. In such cases the number of hours worked can be used. Number of dened accidents in period x 100,000 Total person hours worked in period


Page 18 sc/1026/v2


Accident Severity (or Gravity) Rate One way of improving comparisons would be to take into account the consequence of accidents in terms of lost time to an organisation rather than the denition of what constitutes the accident itself. Severity rate is the number of days lost in a period per 1000 (10 3) person hours worked. It is useful to consider the severity of the accidents in terms of time lost due to injury in addition to the number of accidents. The results can identify trends. However the results can be distorted by low morale, sick pay policies, incentives and “light duties” work. Additionally, long term injuries such as a back injury can also distort the rate. Total number of days lost in a period x 1000 Total person hours worked in a period

Ill-health Prevalence Rate The ill- health prevalence rate is calculated as: Number of people with a disease or condition at a specic time x 1000 Number of people in the population at risk at the specied time If the data has been collected at a single point in time it is referred to as a point prevalence rate. It is often more convenient to collect data over a specied period of time, in which case it is a period prevalence rate. In interpreting a prevalence rate the following factors should be considered: ▪

Severity of illness – if the mortality rate is high the prevalence rate is depressed;

▪

Duration of illness – short duration illnesses will have a lower prevalence than long duration illnesses; and

▪

Number of new cases – if many people develop a disease its prevalence is higher than if few people do so

The incidence rate is used to describe the number of new cases of an illness occurring in a specic time period.


Page 19 sc/1026/v2


Example For example an organisation which manufactures components for the automotive industry is based on a single site and employs 750 people. Table 1 provides recent accident data recorded for the company. Table 1: Example of the Numbers of Accidents with Time 2008

2009

2010

2011

10

12

12

15

Average hours worked

3520

3500

3500

3530

Days lost due to accidents

500

80

600

600

Fatal Accidents

2

0

2

2

No of accidents

The data in the table can be used to calculate for each year, different statistical measures of performance. Accident Incidence Rate = (No of accidents/No of employees) x 1000 2008 = 13.3 2009 = 16 2010 = 16 2011 = 20 This data demonstrates an increasing trend in the AIR. Accident Frequency Rate = (No of accidents/Total person hours worked) x 100 000 2008 = 284 2009 = 342 2010 = 342 2011 = 424 This shows an increasing trend although the data is stable between 2009/10. Further investigation may be used to establish reasons as to why the data is on an upward trend although stable between 2009/10 Accident severity/gravity rate = (No of days lost/Total person hours worked) x 1000 2008 = 140 2009 = 22 2010 = 170 2011 = 170 Again a increased number over time although the data is particularly affected by the number of fatalities Fatal Accident Incidence Rate = (No of fatalities / Average No employed) x 1000 2008 = 2.6 2009 = 0 2010 = 2.6 2011 = 2.6 Clearly this indicates a worryingly high level of fatalities in the organisation. It may be useful to use this data to compare to the national or industry average


Page 20 sc/1026/v2


Epidemiology Epidemiological analysis was rst applied to the study of disease epidemics. Epidemiology only identies patterns of data distribution; therefore subsequent investigations are necessary to identify the reasons for occurrence. An example is that of typhoid fever which was a major disease in cities. A trial and error study of those affected considered what they ate, the work they did and plotted their home locations. Investigations revealed that the only common thing used by the victims were the same wells for drinking water. Closing the wells in the locality resulted in a marked decline in occurrences and stopped the spread of the disease. The cause of the disease was unknown for many years; however the closure of the wells had a benecial effect. This example forms the basic principles of epidemiology: ▪

Identify the problem to eliminate or reduce through patterns of occurrence;

▪

Investigate, usually by trial and error, the information (data dimension) for possible causal factors for the patterns of occurrence; and

▪

Take remedial action.

Epidemiological analysis considers data gathered for a population. In statistical terms the word ‘population’ does not necessarily only refer to a body of people, it refers to all cases or situations. Analysis of data in a population seeks to identify variables, i.e. any distinguishing attribute or characteristic. Data variables may be a combination of: ▪

Nominal variable – named categories e.g. type of accident, injury site, etc;

▪

Ordinal variable – ranking or relative condition, e.g. fatality, major injury, minor injury etc; and

▪

Quantity variable – differing quantities, e.g. age, size, etc.

Each of the factors under study is termed a Data Dimension.


Page 21 sc/1026/v2


Health Data Dimensions Health data dimensions can include many subsets such as;

The Nature of the Ill-health Condition The diagnosed disease or ill-health condition and the organ(s) affected.

The Nature of Work A causal link may be established between the disease or ill-health condition, the type of work undertaken and substances exposed to, e.g. carpal tunnel syndrome and keyboard use.

Personal Details Age, gender, weight, height, etc. may contribute to the cause of the disease or ill-health condition.

Non-occupational Factors Non-occupational factors can directly affect the health of an individual when combined with occupational exposures. Non-occupational examples include: ▪

Smoking, including passive smoking;

▪

Poor diet;

▪

Obesity;

▪

Alcohol and drug abuse;

▪

Lack of, or unsuitable, exercise; and

▪

Hobbies leading to unhealthy exposures, e.g. recreational noise exposures, repetitive use of keyboards in personal computing or use of game consoles, etc.

Pre-employment Health Screening and Medical Examination Assists in the avoidance of employing people with health problems, however it can identify a predisposition to certain occupational ill-health conditions. Pre-employment health screening usually consists of completing a questionnaire and medical tests, such as lung function testing and audiometric testing, which can be useful to determine a baseline for comparison;

Health Surveillance Data from return to work interviews should supplement the results of health surveillance. Whilst health surveillance is designed to identify the onset of ill-health at an early stage it must be remembered that it does not prevent ill-health, however it is an indication of the effectiveness of control measures and can assist with epidemiological studies, for example into biological resistance; and,


Page 22 sc/1026/v2


External Reports on Occupational Ill-health Data for cases of occupational diseases is available from external sources and varies in incidence due to the different reporting and recording requirements and because of underreporting and bias. External sources include: ▪

Under The Health and Occupation Reporting network (THOR), SWORD / EPIDERM were sources relating to respiratory and skin diseases;

▪

EU / World Health Organisation (WHO);

▪

Death Certicates;

▪

Industrial sectors / Trades Associations;

▪

Insurance Companies; and

▪

International Labour Organisation (ILO).

Accident and Incident Data Dimensions Epidemiology can be applied to accidents and incidents where the same types of data dimension are available. Accident and incident data dimensions include, where appropriate:

The Category of Accident or Incident The types of accident or incident can reveal the effectiveness of existing, and the need for further, risk control measures. Types of accident include: ▪

Contact with moving machinery;

▪

Struck by moving object or vehicle;

▪

Struck against something xed;

▪

Manual handling;

▪

Slip, trip or fall;

▪

Contact with electricity;

▪

Exposure to harmful substance;

▪

Exposure to re / explosion.


Page 23 sc/1026/v2


The Time of Day Analysis of the times of accidents and incidents can reveal trends. Shift work and ‘unsociable hours’ working can affect the way in which people act at work. The natural rhyt hms of the human body may well result in individuals being more vulnerable at certain times of the day. What is more certain is that extended hours of work are likely to result in fatigue which is a contributory factor in many accidents. Accident and incident statistics may also identify seasonal trends.

The Part of the Body Injured Consideration of the part of the body injured can reveal weaknesses in control measures, e.g. back injuries in manual handling operations.

The Nature and Severity of the Injury Examples of types of injury include fatalities, amputations, fractures, dislocated joints, sprains, strains, lacerations, abrasions, puncture wounds and contusions. As the accident triangles demonstrated earlier there is usually a higher incidence of minor injuries in comparison to major injuries. Thus major accidents can be avoided and the severity of injury can be reduced through near miss reporting, investigation and control. Analysis of the nature and severity of injuries can then be used to determine the effectiveness of existing risk control measures and whether any further control measures are necessary.

The Location of the Accident or Incident Mapping the location and type of accidents and incident s onto a plan can assist in the identication of groups of incidence, which is not readily apparent from the review of a spreadsheet list; and,

The Category of Persons Affected; Consideration of the persons affected, for example into employees, temporary workers, contractors, visitors and members of the public can reveal trends. Results may identify those who repeatedly suffer accidents, either due to a lack of aptitude necessitating consideration of an individual’s suitability for a specic role or they may simply identify those who are risk takers, their accidents due to intentional violations, necessitating consideration of disciplinary action.


Page 24 sc/1026/v2


Presentation and Interpretation of Data Tables, Charts and Graphs Data is normally presented in a table, referred to as a frequency table as it identies the frequency of incidence, as demonstrated in Table 2. Table 2: Total Accidents by Category in the First Quarter 2012 2012

Jan

Feb

March

Machinery contact

2

0

1

Struck by

0

0

0

Struck against

2

2

2

Slip or trip

2

1

2

Fall

0

0

0

Manual handling

4

3

3

Electricity

0

0

0

Harmful substance

1

0

0

Fire / explosion

0

0

0

Other

0

0

1

In addition to frequency tables the data can be presented diagrammatically using, bar charts, pie charts and line graphs. Where data is grouped, as in Figure 11, the frequency distribution diagram is known as a histogram or barchart.


Page 25 sc/1026/v2


Figure 11: Total Accidents by Category in the First Quarter of 2012 5

4

3

2

1

0

Jan

Feb

March

Figure 11 represents the data from Table 2 diagrammatically in a bar chart. The data could also be represented in a line graph, as in Figure 12. Comparison of the two diagrams suggests that the better format of representation here is a bar chart. Figure 12: Total Accidents by Category in the First Quarter of 2012 5

4

3

2

1

0

Jan

Feb

March

Figures 11 and 12 represent data dimensions for a limited period of time, however they indicate control, with zero incidence, of struck by, fall, electricity and re / explosion risks. The occurrence of struck against, slip or trip and manual handling accidents are consistent each month.


Page 26 sc/1026/v2


A limitation of this data is that it does not indicate the severity of the injuries sustained, nor does it reect the population affected or the hours worked. Another way of representing data is in the form of a pie chart. A pie chart is preferable where a comparison of each category with the total is required. Figure 13 represents the accident numbers for March 2012 from Table 2. Figure 13: Accidents by Category March 2012

March Machinery contact Struck by Struck against Slip or trip Fall Manual handling Electricity Harmful substance Fire / explosion Other

The pie chart above shows that the highest category for accidents in March 2012 was “Slip or Trip”. Similar pie charts could be drawn for January and February and then compared against March. Trends or patterns could then be identied for further investigation.


Page 27 sc/1026/v2


Summary Statistics In a practical situation it is possible to summarise a large set of numerical data by using various summaries for comparison. These summaries include the range, central tendency and variance. The central tendency is used to indicate the mid-point of a set of numbers and the mean, median and mode are used. These can be demonstrated using the data in Table 3 to consider accident days lost Table 3: Lost Time Accident Data in April Number of days lost per accident (x)

Number of accidents (f)

Number of days lost (fx)

0

1

0

1

7

7

2

4

8

3

2

6

4

1

4

5

2

10

Total = 17

Total = 35

Table 3 is indicating that there was a total of 17 accidents in April giving a total of 35 days lost. Looking at the rst two columns in the table, it can also be determined that of the 17 accidents, 1 resulted in no lost time, 7 resulted in 1 day lost time each, 4 resulted in 2 days lost time each and so on. This accident data could also be represented in a line graph (Figure 14) with each accident plotted separately according to its days lost, starting with accident 1 with 0 days lost. Figure 14: Days Lost per Accident 6

5

t 4 s o L s 3 y a D 2

1

0 1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

Accidents Days lost per accident


Page 28 sc/1026/v2


Using the data from Table 3, the following analysis can be carried out:

Range The range gives an indication of the spread of numbers and is calculated by subtracting the lowest number from the largest. Therefore the range of days lost for the accidents (x) in Table 3 is 5 (5 minus 0). Range of days lost = 5

Central Tendency – Mean The mean is commonly termed the average. From Table 3 the mean number of days lost is calculated by dividing the total person days lost (fx) by the total number of accidents causing days lost (f). Mean = Total (fx) / Total (f) = 35 / 17 = 2.06 Mean days lost = 2.06

Central Tendency – Median The median is the middle value in a data series. Where there are an uneven number of data series then the median is simply the mid-data, however where there is an even number of data series then the median is taken as the mean of the two ‘middle’ data series. In Table 3 and Figure 14 there are 17 accidents (f) causing lost time, therefore the median lies on the 9 th accident (half way along the graph in gure 14). The 9 th accident resulted in 2 days lost time (x), therefore the median days lost is 2. Median days lost = 2

Central Tendency – Mode The mode is the most frequently occurring value in a set of numbers, so from Table 3 and Figure 14 the most frequent number of accidents (f) is 7, therefore the mode of days lost (x) is 1. Mode days lost = 1


Page 29 sc/1026/v2


Analysis From Table 3 and Figure 14, the range of days lost is 5, the mean days lost is 2.06, the median is 2 days lost and the mode is 1 day lost. These statistics can be compared against statistics from earlier months to identify lost time trends. The use of mean, median and mode for the central tendency will depend upon the manner of presentation. For the example above it would be unlikely that a mean value of 2.06 would be used, as whole or half day representation is more common. The mean value is also the most sensitive measure as it is affected more by extremes of data. The median may be more appropriate where fractional values are not realistic. The mode represents the value that is more likely to be repeated in future, as it represents the most frequently occurring value. Summarising sets of numbers using one or more measures of the range and central tendency can be adequate in some circumstances, but in others it is not. Consider the data in Table 4. Table 4: Accident Lost Days - First Quarter Number of days lost (x) Number of accidents (f)

0

1

2

3

4

Jan

2

0

16

0

2

Feb

10

0

0

0

10

4

4

4

4

4

March

In each month there were 20 accidents (total f) giving a total number of days lost each month of 40 days (total xf). Therefore for each month; Range of days lost = 4 (4 minus 0), Mean of days lost = Total (xf) = 40 = 2 Total (f) 20 It can be seen that although the spread of data in Table 4 is greatly varied, using the mean gives the same result. Therefore, alternative measures such as variance or standard deviation can be used to summarise the differences.


Page 30 sc/1026/v2


Figure 15 demonstrates the differences between the data from Table 4. Figure 15: Accident Lost Days First Quarter 18 16

s 14 t n e d i 12 c c A 10 f o 8 r e b 6 m u 4 N 2 0 0

1

2

3

4

Days Lost Jan

Feb

March

Variance and Standard Deviation Variance and standard deviation calculations accurately reect the spread of measurements within a range. Variance is the mean of the squares of the deviations from the mean. Where n is the number of data, m is the mean and the xi are the data values. Squaring the deviations causes large deviations to contribute even larger amounts to the variance, while small deviations contribute very small amounts - so large deviations have more weighting. Standard deviation is a common measure of statistical dispersion and is the square root of the variance. Its units are the same as the original data, so it is easier to handle as a weighted average of the data’s distance from the mean.


Page 31 sc/1026/v2


Analysing the Information The limitations of interpreting results are twofold; over generalisation of the results and misinterpretation. As statistical data is historic it does not necessarily indicate future performance. The data from any study only applies for the population studied under the conditions at the time of the study. One particular use of accident and incident rates is to enable trend analysis to be completed. This is useful when the organisation is facing a number of different changes, since the parameters within the organisation for calculating the rates remain the same and an indication of the movements of those rates, beyond signicant boundaries, can be determined.

Variability When analysing trends it should be clear that any change of direction of the trend is more than just a chance or random uctuation. For example, an organisation that manufactures 500 components per day experiences 100 recorded accidents in a year. Given that nothing changes it may be expected that approximately 100 accidents would occur. If the case arose whereby 101 or 99 accidents were recorded in a particular year then this could be down to random uctuation. However if gures were recorded of 80 or 120 accidents a year then the difference would be more signicant from the expected gure of 100. For long-term data analysis identifying upper limits and lower limits of signicance of deviation, the trend can be calculated and plotted. Deviation of the trend beyond these signicant boundaries would then indicate external inuences beyond that of random uctuation.

Validity A study only identies patterns of data distribution and assumptions must not be made about the correlation between data without considering the variables for a plausible causal link, this may require further trial and error investigation of existing or further data. For example in a study of accidents Powell et al. 1971 found that there was correlation between those deemed as sociable and outgoing (extroverts) and the number of recorded injuries. It may have been tempting to assume that extroverts had more accidents. A further study, however, demonstrated that extroverts were more likely to report injuries, demonstrating a link between personality traits and the reporting of injuries rather than their occurrence.

General Errors When using accident and incident rates as a measure or comparator of an organisation’s performance there must be clearly drawn guidelines as to denitions, interpretation and multipliers that are used. The level of risk between the organisations must also be comparable in order to successfully use these rates as comparators between organisations. For example an organisation that has an incidence rate of 100 may indicate that the high risk levels within that organisation are well managed. However comparison with another organisation, which also has an incident rate of 100, may indicate that that organisation is very poorly managing the low levels of risk contained within the organisation. It would be reasonable then to suggest that the types of comparison that can be made are between parts of the same organisation, between similar organisations, between organisations within a relevant industry or service sector, or comparisons between countries. All these comparisons of course would be subject to clearly dened terminology and boundaries in the scope of the comparison.


Page 32 sc/1026/v2


Reporting and Recording of Accidents, Ill-Health and Dangerous Occurences The International Labour Organisation Code of Practice entitled “Recording and Notication of Occupational Accidents and Diseases” 1996 sets out a framework for the reporting at Governmental and employer level, accidents and diseases that occur during work. The Code of Practice sets out both the general principle that any national system should be aimed at accident prevention and details the scope of national reporting requirements and includes for example commuting accidents associated with work and requirements for dangerous occurrences as well as accidents and diseases. This Code of Practice details specications at both national and employer level for: ▪

The legal requirements for setting up an accident reporting system;

▪

The detailed arrangements for reporting;

▪

Arrangements for recording;

▪

Arrangements for notication;

▪

Details of statistical analysis required; and

▪

Details of the investigation required.

The Annexes to this Code of Practice contains denitions of topics such as: ▪

Denition of reportable diseases;

▪

Denition of types of occupation;

▪

Classication of the type of accident.


Page 33 sc/1026/v2


Denitions Accident An accident can be dened as ‘ an unwanted, unplanned event which results in a loss of some kind ’ , e.g. a person tripping over an uneven surface and breaking their leg. ILO accident denitions include: ▪

Occupational accident - an occurrence arising out of or in the course of work that results in a fatal or non-fatal occupational injury.

▪

Occupational injury - death, any personal injury or disease resulting from an occupatio nal accident.

▪

Commuting (travel) accident - an accident resulting in occupational injury involving loss of working time occurring on the direct way between the place of work and: ▪

the worker’s principal or secondary residence;

▪

the place where the worker usually takes his / her meals; or

▪

the place where the worker usually receives his / her remuneration.

Trafc accidents in which workers are involved during working hours and which occur in the course of paid work are considered as occupational accidents. One of the Annex to the Code of Practice details 14 categories of injury which should be reported, e.g. dislocations and amputations.

Near-Miss A near-miss, or incident, can be dened as ‘ an unwanted, unplanned event that had the potential to result in a loss’ , e.g. dropping a tool box off a platform onto a walkway below, just missing an employee. A near miss is a warning of deciencies in an organisation’s systems that could result in actual harm. It is important that they are reported and acted upon to avoid actual loss in the future. The ILO denes an incident as an unsafe occurrence arising out of or in the course of work where no personal injury is caused.


Page 34 sc/1026/v2


Occupational Ill-health Occupational ill-health includes any acute or chronic ill-health caused by physical, chemical or biological agents as well as adverse affects on mental health. Schedule 1 of this Code of Practice species 29 occupational diseases that should be considered in a national reporting system, e.g. hearing impairment caused by noise, diseases caused by alcohol and glycol and occupational asthma caused by sensitising agents. The Code of Practice has a specic requirement for accident investigation at both national and employer levels. An employer level investigation should be undertaken by a ‘competent person’ with the purposes of: ▪

Establishing what happened.

▪

Determining the cause of what happened.

▪

Identifying measures to prevent a reoccurrence.


Page 35 sc/1026/v2


Loss Investigations Reasons for Investigation ▪

To ensure that the organisation is operating within national legislation.

▪

To gain an understanding of how and why things went wrong.

▪

To gain a true idea of what really happens and how the work is really done.

▪

To identify deciencies in risk control management.

Benets Arising from an Investigation ▪

Prevent a future similar adverse event.

▪

Prevent business loss due to disruption, stoppage, lost orders and the cost of legal actions.

▪

Improvement in employee morale and attitude towards health and safety. Employees will be more co-operative in implementing new safety precautions if they were involved in the decision and can see that a problem was resolved.

The UK Health and Safety Executive has published useful guidance on the practical investigati on of accidents and adverse events. This publication, HSG 245, entitled ‘Investigating Accidents and Incidents’ sets out a logical and structured 4 step approach to accident investigation; ▪

Gather information;

▪

Analyse information,

▪

Identify control measures; and

▪

Develop and implement an action plan.


Page 36 sc/1026/v2


Events to be Investigated Having been notied of an adverse event and been given basic information on what happened, a decision is needed as to whether it should be investigated and if so, to what depth. It is the potential consequences and the likelihood of the adverse event reoccurring that should determine the level of investigation not simply the injury or ill-health suffered on this occasion, e.g. is the harm likely to be serious or is this likely to happen often? For example, the potential consequences of a near miss may be great if the event were to happen again. Table 5 sets out criteria for determining the level of investigation which is appropriate for an adverse event. The worst possible consequences of the adverse event should be considered, e.g. a scaffold collapse may not have caused any injury but has the potential to cause major or fatal injuries. Table 5: Table of Events that Should be Investigated and in What Depth Likelihood of recurrence

Potential worst consequence of adverse event Minor

Serious

Major

Fatal

Certain Likely Possible Unlikely Rare Risk

Minimal

Low

Medium

High

Investigation Level

Minimal Level

Low Level

Medium Level

High Level

Investigation Team ▪

In a minimum level investigation the relevant supervisor will look into the circumstances of the event and try to learn any lessons which will prevent future occurrences.

▪

A low level investigation will involve a short investigation by the relevant supervisor or line manager into the circumstances and immediate underlying and root causes of the adverse event, to try and prevent a reoccurrence and to learn any general lessons.

▪

A medium level investigation would be a more detailed investigation by the relevant supervisor or line manager, health and safety adviser and employee representatives and will examine the immediate, underlying and root causes.

▪

A high-level investigation will involve a team based investigation, involving supervisors or line managers, health and safety advisers and employee representatives. It will be carried out under the supervision of a senior manager or director and will look for the immediate, underlying, and root causes.

Timeline The urgency of the investigation will depend on the magnitude and the nature of the risk involved. In general adverse events should be investigated and analysed as soon as possible, not simply for good practice but memory is best and motivation greatest immediately after such an incident.


Page 37 sc/1026/v2


Procedures for Investigating Accidents and Ill-Health It is unwise to depend only upon accident and ill-health investigations as a means of their prevention. Too often investigations fail to identify root causes because they: ▪

Only identify the immediate cause, not the underlying causes;

▪

Focus on the people or equipment that were at fault, ignoring the organisational controls that should have prevented them from being at fault; and

▪

Look for a single cause when there may be several direct causes, each of which resulted from some failure in the management system.

Immediate Action The immediate action after an accident should be to deal with the casualty by implementation of rst aid arrangements and seeking medical assistance where necessary. The health and safety of those dealing with the casualty must be assured by making the area safe. The scene of the incident should then be preserved for investigation purposes.

Selecting the Investigation Team Investigations may be carried out by individuals, provided they are competent to do so, although a team approach is better. A documented arrangement for accident investigation should be recorded. Often the number of persons investigating an accident is determined by the seriousness or potential seriousness of the outcome. The team might typically include: ▪

The Supervisor or Manager;

▪

The Health and Safety Adviser;

▪

Relevant specialists, e.g. Engineers;

▪

The Occupational Health Nurse; and

▪

The Safety Representative.


Page 38 sc/1026/v2


Investigation The investigation should take place immediately after the incident occurs. Its aim should be to establish the facts rather than to establish blame. Sources of information can be obtained from three main sources namely the scene, relevant documents and the people who were involved or witness to the event.

Scene The sooner the scene of the accident / incident is visited, the more useful the information will be. Transient conditions such as the presence of fumes or wet oors, and conditions that are correct and in place as well as those that are not, should be noted. For example, in cases such as slips / trips / falls or accidents involving vehicles where lighting and visibility, condition of oors were all satisfactory, it is useful to make a note of it. Photographs, videos, CCTV and dimensioned sketches can be extremely useful.

Documents Risk assessments and operating procedures are essential documents in any investigation. Other documents that might provide useful information include: ▪

Training records;

▪

Maintenance records;

▪

Inspection and audit records;

▪

Test certicates and maintenance logs; and

▪

Previous accident records.


Page 39 sc/1026/v2


People The accident should be discussed with the casualty as early as is reasonable. Witnesses should be interviewed one at a time by people who have been trained in interviewing techniques to identify the facts rather than opinion. Considerable skill and restraint is required when obtaining information from individuals whom may be suspicious or worried about the purpose of the interview or may put their own interpretation on events. A witness is anyone who has anything factual and useful to say about any matter that is relevant.

Interviewing Witnesses An interview can be described as “a conversation with a purpose”. That purpose is to establish the truth. It is the skills of the interviewer which will dictate how effectively the task is carried out. To this end, it is necessary that interviews are carried out impartially, with an open mind and that they are planned, carried out effectively and the information gathered is reviewed and analysed. Before the Interview Before carrying out an interview there are several aspects to be considered: ▪

Who should be present;

▪

Purpose of the interview;

▪

How will the interview is to be structured; and

▪

Where the interview is to be held.

The interviewee may request to be accompanied by a colleague In this instance. It is essential that the only person interviewed is the interviewee and that the person accompanying them remains a ‘silent support’. It is important to only interview one person at a time. If groups are interviewed, people may not come forward with critical information because the others are there. The purpose of any interview is to establish the truth. To enable this to be done effectively and efciently, the interview must be planned and a set of aims and objectives prepared. This will not only provide a structure for questions and thus improve the ow of the interview, but it will also give a greater degree of condence to the interviewee.


Page 40 sc/1026/v2


Factors to Consider ▪

The interviewer controls the interview and by doing so, the interview can full its aims and objectives.

▪

The interview location is comfortable, there is somewhere to sit, it can be held in a condential manner and that others cannot hear what is said.

▪

Plan to ensure there are no interruptions during the interview.

▪

Avoid interviewing people just before they are due to leave the workplace as people will want to get away.

▪

If it is urgent to speak with someone it may be necessary to hold the interview in a location more convenient to them.

▪

Time should be allowed by the interviewer before the interview to think about aims and objectives for the interview. What facts need to be established? What questions should be asked? What points raised in other interviews need verication?

▪

It is often helpful to prepare a list of questions to ask.

During the Interview Initial introductions are important and an explanation should be given that the purpose of the interview is not to apportion blame but to establish the facts of what happened so actions can be take the stop a reoccurrence. This introduction should also include details of how the interview will progress and what topics will be covered. Time spent building a good rapport with the interviewee is not wasted. It allows the interviewer to both relax the interviewee and themselves and also to encourage the interviewee to talk. The interviewer should always try to use appropriate language and speak in a manner which the interviewee understands. Clichés, abbreviations and jargon should be avoided as these may create misunderstandings. Judgmental statements should also be avoided. Always consider the welfare of the interviewee. If the interview is likely to be lengthy, consider arranging refreshments. Make sure that the interviewee is comfortable before the interview commences. Appropriate questions should be used to stimulate and encourage the interviewee to use their own powers of reasoning and to gather more information about what happened.


Page 41 sc/1026/v2


There are three types of questions that should be used: ▪

Open Questions These allow the interviewee to offer a lot of information, rather than conning them to yes-no answers. For example, instead of asking a closed question, such as ‘Did you see the fork truck?’, to which the answer is either ‘Yes’ or ‘No’, ask ‘Tell me what you saw when you entered the warehouse’, which allows the interviewee to open up in their reply. Open questions allow both the interviewer and interviewee to establish rapport, whilst closed questions limit communication

▪

Probing Questions These demonstrate an interest in the interviewee and encourages them to keep talking. They increase the quality and quantity of information disclosed and conrm the interviewer’s understanding of such information, e.g. how fast was the truck travelling?

▪

Closed Questions These make the pieces of information gathered specic and conrms that the interviewer has listened and understood the precise details of what has been said. However, closed questions, by their nature, can limit access to detailed responses, e.g. was the truck travelling too fast?

The most difcult questioning skill to accomplish is that of asking open questions, which is one to which the interviewee cannot simply give a short reply such as ‘yes’ or ‘no’. Because of this they are very useful for both gathering information and encouraging the speaker to talk more freely and deeply about a subject. Open questions have a tendency to use the same few words at the start. These are: ▪

What;

▪

When;

▪

Where;

▪

How;

▪

Who; and

▪

Why.

It is very difcult to ask a closed question if the sentence starts with one of these words. Closed questions are only of value if we need to check a detail, such as ‘Was the forktruck red?’, to which the answer can only be ‘Yes’ or ‘No’. However they do not normally provide us with signicant information, so are of limited value. Probing questions look like open questions, but are used to explore the topic that the open question has brought out into the open. Although facing the interviewee with a barrage of questions should be avoided at all costs it must be acknowledged that carefully framed questions are vitally important. Figure 16 suggests that open questions are used to seek information, probing questions to build upon that information and closed questions to test understanding of what has been said. Figure 16 also demonstrates how the questioning should go through three stages: seeking information, building information, and testing understanding.


Page 42 sc/1026/v2


Figure 16: Three Stages of Questioning Gathering Information

Open Questions

Building Information

Probing Questions

Testing Understanding

Closed Questions

Before the interview is nished, a review should be taken of all the information obtained from the interviewee, so that both are quite clear and understand what has been said. It may also be appropriate that the interviewee gives a written statement with regards to the accident / incident. This should be signed and dated by the interviewee and a copy given to them if requested. When the interview has nished, the dialogue should never be closed. It should always be explained that if the interviewee remembers anything else that may be useful, they can come back and see the interviewer. The interviewer should also explain that they may need to speak to them again should the need arise during the investigation. After the Interview After the interview, there needs to be a time of reect where the following is determined; ▪

Were the aims and objectives achieved?

▪

What new information was gained?

▪

How does all this t in with what is already known?

▪

How did the interviewer perform in the interview?

Consideration needs to be taken of the new knowledge obtained in order to decide what the next step will be e.g. look at other documentation evidence, revisit the scene, interview others etc.


Page 43 sc/1026/v2


Recording and Analysing the Results Initially the aim is to identify the possible direct causes or contributory factors, then to establish the underlying causes. It may be convenient to consider causal factors in the following categories, materials, equipment, environment and people (MEEP) using the root cause analysis technique such as fault trees to identify the direct and indirect causes. No standard report form exists for recording accidents, however the ILO publication; Recording and Notication of Occupational Accidents and Diseases identies that organisations should be required to report to the National level government as a minimum the following: (a)

Enterprise, establishment and employer;

(b)

Injured person;

(c)

Injury; and

(d)

Accident and its sequence:

For commuting accidents, the relevant necessary information to be notied should be specied. Information from the report can then be used within a database for statistical analysis, as described earlier. Improving accident reporting, including the reporting of minor injuries is essential to ensure representative data. In order to achieve this, it requires the implementation of ‘user friendly’ reporting systems, establishment of a ‘no blame’ culture and the provision of feedback on actions. Computer-based systems can help to identify common features and trends. The use of computer technology has facilitated the storage and analysis of accident data. This can be very useful both for measuring health and safety performance and for identifying essential areas for improvement. Investigators need to be aware of two potential problems with such statistics: ▪

Some organisations / departments / shifts might be more likely to report accidents than others; and

▪

If the data entered is not accurate then the analysis will be inaccurate too.


Page 44 sc/1026/v2


Conclusions and Recommendations Accident investigations are useful for the identication of root causes and management failures as discussed earlier in the Bird and Loftus domino sequence. By implementing further control measures similar accidents can be avoided, therefore it is essential that the record be reviewed to ensure completion of any recommended actions. Accident / incident information is used to measure health and safety performance and to identify weaknesses in management systems. There is no point in gathering information about an accident unless the information is used to prevent a recurrence. The following questions should be asked following an accident or incident; ▪

What were the direct or immediate causes?

▪

Were the appropriate physical control measures in place?

▪

Were the appropriate performance standards (procedures, laws, etc.) complied with, if not; why not?

▪

Why were these causes unrecognised and uncorrected before the event?

▪

If human error was a contributory factor, why did it occur and how could it be prevented or its consequences be reduced?

▪

Is there a need to review the risk assessments that were previously carried out (which presumably concluded that the risks were adequately controlled)?

▪

Were the appropriate organisational control measures (management systems) in place, if so why did they not prevent the accident?

▪

What prevented the worst from happening?

▪

Was the immediate response as effective as it might have been?


Page 45 sc/1026/v2

1026 Int Diploma IA2 v2

Recommend Documents