• No results found

Analysis: Coding, comparison and memoing

Preliminary coding, comparison and memo writing

At this very preliminary stage of the study the techniques of grounded theory began to be applied. Case selection required the first cursory read through of not only the cases that would become this study sample but also the cases which were part of the broader cohort of those who choked – ‘the general public’. While analysis had not ‘‘formally’ begun, distinctions and similarities between the clinical group of those at-risk of choking being studied and other possible at-risk groups (for example, drug users) started to generate ideas which highlighted possible significant social phenomena such as setting, behaviour, and the presence of others. These impressions were notated for future reference. Preliminary coding had also begun within the data, initially via separating out ‘those with medical conditions’ and ‘those over 70 without medical conditions’.

It is also perhaps worth noting here that it was probably the most emotionally draining and distressing time of the whole project for me as a ‘researcher’; reading an overwhelming amount of data on over 900 deaths I could make

no claim to objectivity or distance from possible subjects, the circumstances of their deaths, or the distress their deaths caused others. Lives with all their complexity leapt from the pages, confirming the power of coronial reports to illuminate the social phenomenon of choking and bring the experiences of those involved ‘alive’. While the task of selection had objective parameters, the first reading through of these cases was done mostly in tears and lead to my personal reflection through memo writing, thinking and talking with others as to the breadth of an area which clinically I had previously considered ‘just’ one of the risks of dysphagia.

Organising the data and analysis First level of organisation and coding

The initial read through of cases for sample selection had left me with a number of impressions. At-risk adults either lived in private residences or assisted-care settings, and where they lived influenced who might be a part of the social context of their choking. Assisted-care settings were defined in this study as any living situation (temporary or permanent) which included the presence of paid personnel who had an overall support or management role in regard to the at-risk adult. Personnel could be personal care assistants, disability workers, registered nurses, doctors or facility administrators. The allocation of residential status for these two categories was not always straightforward. Where the person choked and who might be responsible for their care was used as a guide, reflecting informal support (family and

friends) and formal support (paid carers as indicated above). Below are examples of decisions that were made when grouping cases, where the person could have potentially been placed in either accommodation group:

If the person was living in their own home and went into hospital to have an operation and while there choked; they were counted as being in an assisted-care setting i.e. the hospital.

If the person was living at home (private residence) and choked and was subsequently taken to hospital where they died shortly after; the case was counted in the

private residence group. This was because any contributing factors and the initial stages of any management would have taken place in the home context.

If the person was living in their home and received some home care this was counted as private residence as the paid carer was only present for short periods of time and had limited influence.

It was also noted at this point of analysis that these two contexts – private residence and assisted-care setting – seemed to have an influence on the coronial investigation process. In the 256 cases studied, no deaths occurring in a private residence proceeded to inquest. It was suspected at this stage in the analysis that there may be a greater expectation on formal, paid carers to provide a greater level of care than that expected of informal carers. I

therefore decided to split the data into two, based on accommodation status. This splitting of the data not only allowed for greater ease in exploration of similarities and differences across the two types of environments (a dynamic contributing to constant comparative analysis) but also at a more

fundamental level given the size of the sample, provided two distinct data sets which would support the grounded theory goal of theoretical sufficiency (Dey, 1999); allowing for the emergence of a theory that would broadly encompass both environmental and support scenarios and those potentially involved in the circumstances around choking events.

It was suggested early on in my doctoral candidature that my sample could be limited to just those cases that went to inquest. These 14 cases were more likely to have the full set of coronial investigation reports and lengthy additional data from the inquest proceedings. This would have been a sensible thing to do given the magnitude of data involved in analysing 256 cases. I was however (having already immersed myself in the data in my first study) loath to lose the rich data available in non-inquest cases (both those where the person was in private or assisted-care setting), so I proceeded with the 256 cases. In hindsight I do not believe I would have achieved the richness of data available in the larger data set had I only focused on inquest cases, let alone approached theoretical sufficiency, thus inquest cases and

non-inquest cases provided two additional data slices that could be compared and contrasted.

Second level of organisation and coding

The case reports being studied were by their nature often long descriptive narratives that tried to piece together what had happened. This narrative form had a natural underlying temporality: what happened before the choking event (documented in the police description of circumstances leading up to the event or the autopsy finding that identified the person’s health prior to choking); what happened at the time of the event (witness reports or expert speculation); and what happened after the choking event (such as the at-risk person being placed on a ventilator or staff giving statements at an inquest). This temporal dimension was apparent across all cases (that is the original 900). It in effect equated to a ‘beginning, middle and end’ of an individual’s choking account and provided a way of structuring the volume of the sample data and aided analysis. The allocation of data to one of the temporal

sections drew out preliminary information and started to organise the data in a way that would shed light on the research question by highlighting the circumstances of each part of the choking story. Thus as all sample cases (256) and their available reports (police, autopsy, toxicology, inquest reports and coroner’s findings) were read through in detail, data for each case were condensed, organised and coded under 4 major categories:

1. History of dysphagia and management

The data items under this heading related temporally to the beginning and middle of a person’s choking story. Questions that emerged at this point included: Was there any indication of swallowing or eating problems prior to the fatal choking event? Had risks for choking been identified? These were questions that seemed to relate to the beginning of the story. Additional questions that seemed temporally relevant to the middle of the story included: Had a management plan or other strategies been put in place to address any risks identified? An overarching question was: Were there social circumstances that might have impacted on the answers to these questions?

This last question expanded the more traditional view of choking (where there are discrete risk factors that can/should be controlled) to consider the possibility of circumstances (environmental, behavioural, relational,

emotional) related directly or indirectly to choking that were a part of the social experience of choking and potentially not controllable.

The next 2 organising categories pertained temporally to the end of an individual’s story.

2. Physical signs (of obstruction) and identification

Questions which emerged from this category included: Why didn’t everyone who choked react the same way? Did those present (victim and witnesses) recognise that choking was occurring? Were there social circumstances that influenced how choking manifested and was recognised?

3. Emergency response

Questions which emerged from this category encompassed: Who was involved? How did witnesses respond? Did first aid go as planned, and if not why not? Were there social circumstances that influenced how people responded? And finally: Were there social circumstances that influenced a death prevention approach?

The final organising category could relate to any of the stages in an individual’s story.

4. Coroner’s recommendations

This category organised specific coroner’s comments and recommendations about an individual case. Questions which emerged from this category

included: What did coroners identify as risk factors for choking? Did coroners solely align themselves with the management of risk? How did coroners respond to cases where risks were identified but not controlled? How did coroners’ responses potentially influence the social context of choking? Table 1 illustrates how data were condensed and organised.

Table 3.1

Illustration of how information was organised and coded into four basic categories Comments code

Red – history of dysphagia and its management

Blue – physical signs/identification

Green – emergency response

Orange – coroner’s recommendations Black - context descriptions/comments

Column 1 2 3 4 5 6 Code*/ Available Reports M/ F Age Medical history Location/ Meal/ Food

Condensed, Summarised and Grouped Data** (D = deceased) E.g. 111* P [police] F [finding] A [autopsy] F 84 Dementia ACF [aged care facility] Lunch

• Vitamised lunch being fed

• D told not to eat any bread with food given to her [presumably bread in reach]

• D “having trouble breathing”, D “jolted in her chair”, D was “blue in the face”

• Heimlich attempted by RN but difficult to do as D a “dead weight”

• Suctioned “small piece of bread removed from the D throat”

• Placed on bed board advised by ambulance to “rub over the D ribs” no avail

• CPR commenced

• Ambulance arrived, airway contained fluid and a lump of ?meat in left posterior oro-pharynx

Researcher Notes

No specific indication why on pureed 112*

P, F

M 81 Dementia ACF • Finding - choked on food • No further info 113 P F 75 Parkinson’s Own home Snack peach

• Family report number of problems related to chewing food and restriction of airway, therefore D continually had trouble swallowing food

• Evening, eating peach fallen to the floor and began to choke and stopped breathing

• Family commenced CPR and called ambulance. On arrival amb. staff continued CPR. A second ambulance attended and removed a large piece of peach from D’s trachea.

• D then required defibrillation, was intubated, ventilated and transported to hospital.

• On life support for 2 days, did not regain consciousness was not

responding neurologically. Life support withdrawn with permission of family.

Illustrative only -some details changed for identity protection.

*Each case received a unique numerical code so that the original de-identified case narrative could be returned to in future steps of the analysis and to extract evidence quotes as required. This returning to the narrative at different stages in analysis was an important check for the researcher to offset the dangers of condensing and summarizing information which could lead to relevant surrounding context being lost.

** Information available on each case varied significantly from essentially no information (e.g. …choked on food) to lengthy reports. The above examples of data summaries are reflective of cases with limited or moderate amounts of information available. Some summaries of cases were several pages long and are not represented here. Coroners’ recommendations were only present in some cases.

General demographic information (gender, age, medical history, residential setting), time of choking, and type of obstructing foreign body were retained for each case. Previous research (Balandin et al., 2009; Berzlanovich et al., 2005; Bradway, 1996) had identified these factors may be relevant to the study of choking.

The allocating of data to the temporal sections, the questions that were generated as a result of this allocation and the sensitivity to the data which expanded as the questions were used to probe the data, all occurred

simultaneously. Categories started to form, and as cases were organised in the above way each new case began to be scrutinised as either containing data that added to the tentative categories forming or suggested new categories. Memo writing included impressions of concrete categories such as ‘behaviour’ but also more abstract ones such as ‘misconceptions’, in addition to how some of the categories were linked. The following articulates this process in more detail.

Coding and categorising

Line by line coding and categorisation

The condensing and coding of data under the four headings (1. History of dysphagia and management; 2. Physical signs and identification; 3.

Emergency response; and 4. Coroner’s recommendations) was possible due to each report undergoing line by line coding. Line by line coding was used to categorise data under the four headings but also to identify categories within the temporal sections (beginning, middle, and end) of the choking

experience.

The following example in Figure 3.1 shows how data extracts were identified and initially linked to possible issues; overall this data extract example

belonged to the middle temporal section. Identifying data extracts was done while always considering the context and what possible abstract meaning was present based on the relationship between context and substantive data. The initial starting point to coding was the question: Is there any

phenomenon (physical and/or social) indicated in this sentence/excerpt that might either contribute to choking or be protective against choking occurring? Phenomena could be something discrete like ‘the person gorging their food’ or relational ‘the carer providing supervision and doing a medicine round at the same time’. In this case the at-risk adult had cognitive problems, raising the question of whether such a factor may influence how accurately an at-risk adult can identify choking and their need for assistance.

Figure 3.1: Example of coding data extracts

The at-risk adult in this case died from this choking event in the toilet. Possible categories produced from this coding included: ‘identification of choking’, ‘first aid technique’, ‘victim awareness’, and ‘unexpected response’. As each new case was analysed these categories were either confirmed, expanded or collapsed into others.

Clarifying this analytic step further, and using the condensed examples given in Table 3.1, in Case 111: D told not to eat any bread with food given to heris a data extract, as is Family commenced CPR and called ambulance in Case 113 and choked on food in Case 112. Data extracts could be direct quotes from reports or a summary of a longer piece of narrative where non-relevant information had been excluded. A memo (context note) was attached to an

extract if the face value reading of the extract was distorted or diminished by the lack of contextual information.

Links and category formation started to happen simultaneously at this level of analysis. For example: D told not to eat any bread with food given to her I linked with the deceased’s diagnosis of dementia (did the person understand the instruction?), which then lead to tentative categories including: person’s behaviour (taking inappropriate food); faulty assumptions (carer assuming instruction would be understood and followed); training (carer not

understanding cognitive deficits in people with dementia); lack of supervision; quality of life (person preferred bread to puree). In this study therefore a data extract could be placed in one or more initial categories, in keeping with constant comparative analysis of data and possible categories generated. One possible flaw in this making of links and development of categories as illustrated in the bread example above was that I was speculating that the deceased person’s dementia might be sufficiently advanced to impact on their understanding. This may not have been the situation. This possible problem was offset by constantly comparing data extracts across cases, seeking to confirm whether there were other examples of people with likely cognitive problems being expected to have awareness and self-monitor or manage their own behaviour. The at-risk adult in Case 029 (Figure 3.1) who reported being ‘ok’ and the at-risk adult in Case 111(Table 3.1) who took bread against instruction therefore formed a possible pattern of

communication and cognitive difficulties impacting on their risk of choking.

Challenges to analysis

Coronial reports are secondary data that have already been filtered through the perceptions of both those who were directly involved in the choking deaths and those charged with documenting and retelling key elements after the fatal event. Accuracy, detail and interpretation therefore may all be limitations inherent in the data and potentially additionally influenced by my own filters.

In the example: “...PCA tapped deceased back lightly which seemed to give her some relief” (Case 029, Figure 3.1). ‘Relief’ here could have entailed the person’s colour returning to normal, or her breathing becoming less

constricted, or her facial expression relaxed or some other manifestation, but this information was not reported. Semantically, ‘tapped’ may equate with a ‘hit’ to some people. Lack of detailed information and different words could affect interpretation. Additionally, other than police investigations, which are the most immediate recording of events in coronial investigations, inquests and further questioning of witnesses may occur long after the actual event:

“I found the explanations given in court to be, on the whole, inconsistent with available independent evidence and unreliable. I accept on reflection this may be

accounted for by elapse of time and contamination of the witnesses’ recall by concern and extensive cross-

fertilisation and discussion brought about as part of the process of dealing with an unexpected death”

(009 Inquest – Coroner’s comment). This delay in questioning may result in lapses of memory and

reinterpretations of what occurred, on top of researcher interpretations. However, constant comparative analysis over so many reports I believe diluted the impact of such possible interpretation errors. At face value ‘tap’, ‘rub’, ‘blow’ and ‘hit’ used to describe the application of first aid could have all meant the same thing, but by linking environment and the age of some at-risk adults and witness comments across different cases, it became apparent that these words overall reflected a genuine degree of difference in the degree of force applied from a first aid perspective. Some carers expressed concern that they would hurt the choking person; all such data comparisons aided analysis.

As the simultaneous identification of extracts and their coding occurred, multiple categories evolved, such as ‘at-risk adult’s behaviour’, ‘supervision’, or ‘communication/documentation’. These categories emerged over the