Capítulo 2 - Anatom ía, Patología y Examinación
2.1 Órganos del sistema visual
2.1.2 Contenido del globo ocular
A common approach was taken to data analysis across all three sets of interviews. For the patient and carer interviews, three researchers familiarised themselves with the data by reading and rereading the
transcripts and identified key issues, concepts and themes. These served as a basis for developing a coding
framework of initial categories and themes, which was tested on a 10% sample of transcripts before being used to code all transcripts line by line. Two researchers coded the same 20 transcripts, and a third
researcher coded a further 10. Any discrepancies were discussed and, if agreed, the minor modifications
were made to the coding framework.
Each major theme was allocated to a chart in Microsoft Excel (Microsoft Corporation, Redmond, WA, USA), with a row for every case and a column for each category. Data were entered into the relevant cells and summaries were prepared at the base of every column. The data were compared within and across cases, searching for commonalities and differences. This enabled the research team to identify patterns and main
issues arising from the interviews. The original transcripts andfield notes were referred to in order to ensure
that analysis took the context of interviews into account.
In the case of the professional interviews, the thematic framework brought together key themes emerging from those transcripts with a priori issues and questions derived from the aims and objectives of the study
and concepts raised by respondents in thefirst two phases. One researcher, in discussion with two others,
tested and adapted the framework on the basis of a 25% sample of interviews transcribed. All transcripts were subsequently coded line by line by one researcher, using NVivo software (version 9.2, QSR International,
Warrington, UK) to manage the data. A second researcher double codedfive transcripts to check the validity
of the framework. The NVivo programme’s framework matrix function was used to allocate each theme to a
chart, with a row for every source and a column for each category. Data were mapped and interpreted by searching for associations between themes and identifying the commonalities and differences in
interviewees’ attitudes.
The researchers involved discussed and refined their analysis throughout the process, drawing on their
different backgrounds (medicine and public health, nursing, psychology, sociology and social policy) and their different relationships with the data (study design, interviewers or data analysts) to reach agreement. Findings were discussed with an external group of research partners, comprising people who had
Chapter 4 Analysis of linked Hospital Episode
Statistics and mortality data
T
he aim of this section of the project was to analyse data on care received by older adult patients with specific causes of death, in the year before death. Hospital Episode Statistics (HES) records details of each episode of inpatient care provided by a consultant. Total length of each admission (made up of all consultant episodes), dates of admission and discharge, destinations, procedures undertaken and specialty of treatment are recorded.Linked HES and mortality data were obtained for patients aged≥ 75 years who had been admitted to
hospital with a diagnosis of heart failure or lung cancer in the 12 months before their death. At the time of the study, it was not possible to select a group of individuals using mortality data and then extract HES
data relating to their hospital use. The study data set was, therefore, defined by diagnoses recorded by
HES and not by cause of death. Hence, patients who had received hospital treatment for specified
conditions but did not have any of these recorded as a cause of death would still be included in the study.
As the starting point for the study data set was hospital admissions, the data provide a reflection of
resource use, but are liable to inaccuracies within HES. Deficiencies in the quality of HES data are well
recognised, including many mistakes of coding and omission.79,80People who died without any inpatient
episodes in the 12 months before death would not be included in this study. However, we know that
78% of all deaths recorded by the Office for National Statistics (ONS) are associated with one admission
in the last year of life; the same is true for 88% of all age cancer deaths and 66% of deaths from cardiovascular disease.81
Pseudonymised HES containing records of admitted patient care in England, linked to ONS mortality data,
were selected for years 2000–1 to 2010–11. The data were restricted to people aged > 74 years with
diagnoses of either lung cancer or heart failure [International Statistical Classification of Diseases and
Related Health Problems, Tenth Revision (ICD-10) codes I50–I50.9, I11.0. I42.0, I25.5, I42.9, C33.0–34.9].
HES data variables were selected relating to diagnosis, length of stay and discharge destination. No variables that would identify geographical locations for patients or institutions or dates of birth were used in this analysis. The Index of Multiple Deprivation (IMD) was provided with the data, calculated from the postcode of residence provided on hospital admission. Mortality data were obtained only for patients who had been in hospital in the year before their death and, therefore, had a record in the HES. Cause of death and whether this took place at home or in a communal establishment were extracted. Data were reported to be cleaned prior to being released for analysis.