The purpose of analysing data collected in a study is to describe the data in meaningful terms. In using a quantitative approach that uses numbers to organize, analyze and interpret data, the researcher had to make sense of a set of numbers and communicate them in some logical fashion to the reader. Data analysis allowed the researcher to make sense of the data that was collected, provide information that
120
is understandable and discover relationships and differences of the phenomenon being explored. Data presentation by means of tables and graphs helped to summarise the data, thus making it easier to understand (Brown and Saunders, 2008:18).
Data analysis comprises of four steps: identification of the study objectives, data preparation, implementation of analysis and data presentation (Myatt, 2007:7). These four steps, as applied in the research study, are discussed below.
Identification of the study objective
The objective of the study with regard to the data collection instrument was to collect data on the current nursing care practices related to endotracheal tube verification, endotracheal tube cuff pressures, endotracheal tube suctioning and mechanical ventilator settings. Professional nurses in the public and private adult critical care units in the Nelson Mandela Metropole were included for participation in the research study.
Data preparation
The process of data collection and preparation is critical to the confidence with which decisions can be made. The data needs to be put into a tabular format in order to characterize all the variables. Data also needs to be cleaned by resolving any ambiguities, errors and removing redundant and problematic data. Details concerning the steps taken to prepare data for analysis should be recorded. This not only provides documentation of the activities performed, but also provides a methodology to apply to a similar data set in the future (Myatt, 2007:17).
On collection of the completed questionnaires, the data analysis process started by ensuring that the data was reliable and represented the defined target population. All questionnaires were checked for comprehensiveness and completeness. As enrolled nurses are employed in the critical care units in both public and private sectors, the demographic section of all the questionnaires had to be checked to ensure that only professional nurses completed the questionnaires, especially since the researcher was not available on all the shifts to hand out the questionnaires.
121
Coding is the process of transforming data into numerical symbols that can be entered easily into the computer (Burns and Grove, 2009:432). Data was coded after the collection process was completed. Apart from the reference number on each questionnaire, a number was assigned to each questionnaire from the private and public divide. Numbers 1-40 were assigned to the questionnaires collected from professional nurses in critical care units in the public sector, while numbers 41-100 were assigned to questionnaires collected from professional nurses in the private sector. The coding process was necessary to assist in capturing the findings from the different health care sectors and the allocated numbers were thus only reflected in the data verification and frequency tables and not in the final presentation of data; however, these aided in differentiating between the nursing care practices of the professional nurses in the public and private sectors respectively.
Each variable for each question was then coded for capturing to the data capture sheet prepared by the statistician. Data were categorised by using constant, dichotomous, discrete and continuous variables. It was decided to categorise demographic variables to allow for cross tabulation, if required during the analysis process. The variables were categorised as follows:
Group: public=1 and private=2; Gender: male=1 and female=2;
Years‟ experience: <=5 years=1; 6-10 years=2; 11 and more years=3; Post-basic qualification category: Yes=1; no=2;
Position held: non-leadership (including all the permanent and agency worker)=1; leadership (unit managers, shift leader and clinical facilitator/mentor)=2.
The variables for the questions in Sections B, C, D and E of the questionnaire were coded from 1-8 depending on the responses to the questions posed. Data tables for the capturing of the data were created by a statistician.
Having performed the preliminary data characterization, the data cleaning process was then done prior translating it into a suitable form for data analysis. Questionnaires were re-checked for missing data. Variables measured on interval
122
and ratio scales were cleaned by checking for the correct responses. Variables that were not noted by the participants and missing data points were removed from the data tables.
Data transformation was done in order to make sense of the raw data. Data transformations included using normalization methods such as minimum-maximum formulas and decimal scaling. Factor analysis was done to assess the interrelationships among large numbers of variables and disentangles those relationships to identify clusters of variables that were most closely linked together. Similarity grouping measurements were done for each response in each section. The care practice items were weighted with the digits 0, 1 and 2 and 3 respectively. The higher weighting (2 or 3) constituted the best nursing care practice as recommended in the narrative literature review. The lower weighting (1) represented adherence to what is marginally accepted as constituting best practice. It is thus suitable but not necessarily best practice. The weighting of 0 represented non-adherence to either of the aforementioned.
Frequency tables for every response in each section were created by the statistician using Statistica version 9.0. These frequency tables indicated the responses for both the public and private sectors. Furthermore, in order to establish the quality of nursing care practices, a rating scale ranging from low, average and high was created. A rating in the high category constituted adherence to best recommended practices. A rating in the average category indicated practices that are marginally accepted but not necessarily best recommended practice. The rating in the low category indicated non-adherence to recommended best practice. The higher the nursing care practice score, the closer the adherence to best practice recommendations. Similarly, the lower the nursing care practice score, the less likely was the adherence to best practice recommendations.
Each section on the different nursing care practices was analysed using a percentage ranging from 20% to 99%. Frequency distribution tables were created for each of the nursing care practices, as well as a combined representation of all four nursing care practices. Correlation analysis between the public and private sectors
123
for each section, as well as the combination of the four nursing care practices was done.
Implementation of the data analysis
Implementation of data as analysed comprised two major tasks: summarizing the data and finding hidden relationships amongst variables. Data was summarized using tables and graphs. According to Myatt (2007:36), tables can be used to present both detailed and summary level information about a data set. It allows the reader to look at individual observation or summaries. Graphs visually communicate information about variables in data sets and the relationship between them. Graphs present the data by visually replacing numbers with graphical elements. It enables us to visually identify trends, ranges, frequency distributions and relationships and make comparisons. Summary tables were used to display the demographic data and complications related to over-inflation and under-inflation of the endotracheal tube cuff pressures. Various graphs can be used to display and summarise data. To ensure consistency, bar graphs were used throughout the study to display and summarise the data.
Descriptive, inferential and correlation statistics techniques were used to describe and summarize the data. According to Brown and Saunders (2008:2), descriptive statistics are concerned with quantitative data and the methods for describing them, whereas inferential statistics make inferences about populations by analysing data gathered from samples and deal with methods that enable a conclusion to be drawn from this data. Correlation statistics quantify relationships within data. The descriptive, inferential and correlation statistics that were used were the Chi-square and p value calculations and were applied in Chapter 4 of this study. The p value is frequently presented to illustrate the likelihood of the results of a topic, with its value always being between zero and one. A p value of 0.5 indicated that there is a 1 in 2 probability of the results being due to chance. In this case, it would be unlikely to accept the results of the study as being significant. Conventionally, a p value of 0.05 or below would be accepted as being statistically significant. That means a probability of 1 in 20 that the result is due to chance (Craig and Smyth, 2007:135).
124
Within this study, the Chi-square test of independence was used to explore if there were relationships between the responses from the professional nurses in the critical care units of the public and private health care sectors. A Chi-square p value less than 0.05 was considered statistically significant. For statistically significant results, Cramer‟s V was calculated to determine whether the results were practically significant, indicated by V values greater than 0.10.
Finding hidden relationships refers to the identification of important facts, relationships or trends in the data, which are not obvious from a summary alone (Myatt, 2007:3). Within this research study, the relationship between the public and private sectors was explored with specific reference to each nursing care practice section, as well as the combination of the four nursing care practices.
Data presentation
Data presentation involves setting up a plan to deliver the results of the analysis to the identified consumer, which, in this case, is the reader of the research project. The report of the data analysis is presented in Chapter 4 of this study.