When researching whether smoking is bad for people’s health, it is vital to consider health outcomes in light of the number and quality of cigarettes smoked over a particular time period. Health outcomes of those who smoke one high quality cigarette a month are expected to be different from the health outcomes of those who smoke 20 poor quality cigarettes a day. Similarly, when considering whether and how QoL changed, it is critical to consider the quantity and quality of the CLIQ intervention that participants engaged with.
Level of individual participation is a proxy for the quantity of CLIQ received by each
participant. The relative success of implementation in each area is a proxy for the quality of CLIQ delivered.
As discussed in section 5.3, inclusion in the impact sample was determined by
participation in the QLAs. Data on the number of scheduled CLIQ activities (both QLAs and training sessions) that participants took part in, also represents an outcome of the
intervention. The discussion below builds on section 5.3 (p134) and draws on the analysis of implementation presented in this chapter, to establish participation and implementation variables. These variables are both findings on participation and implementation (as outcomes of the CLIQ intervention) and variables for the analysis of findings, thus bridging my discussion of methodology (chapters 5 and 6) and findings (chapters 7, 8 and 9). This section ends by reflecting on the definition of the impact sample and establishing a sub- sample (the core sample) as a basis for empowerment analysis.
6.4.1 Participation and Implementation Analysis Variables
The level of individual participation of those included in the impact sample has been classified as good, average or poor, as defined in Table 6-2, based on individual attendance data from QLAs and computer training sessions. Overall, participation by almost three fifths of the sample was good, with one fifth displaying average participation and another one fifth displaying poor participation. On an area basis, eMpumalanga stands out with almost three quarters of participants falling into the good participation category, whereas this proportion was around half for the other areas. In eNingizimu, the proportion of participants with poor participation was lower than eNtshonalanga and eNyakatho, and about the same as
eMpumalanga, resulting in eNingizimu being second best. eNtshonalanga displays the worst
scenario with poor participation at 43% (followed by eNyakatho at 26%). The areas can therefore be ranked in terms of participation from the best (eMpumalanga) to the worst (eNtshonalanga) as shown in Table 6-2.
Table 6-2: Nature of individual participation per area
Nature of individual participation Total eMpuma-langa eNingi-
zimu eNya-
katho eNtshona -langa Good: Attended all CLIQ activities or skipped
only one activity 57% 73% 50% 52% 49%
Average: Skipped more than one activity but
did attend some training 21% 21% 45% 22% 8%
Poor: Did not attend any training but did attend
sufficient QLAs to establish QoL change. 22% 6% 5% 26% 43%
Impact sample 113 33 20 23 37
Participation ranking 1st 2nd 3rd 4th (where 1st = best and 4th = worst)
By adapting to the local context, implementation differed across the four areas and these differences together with the context meant that the quality and nature of the CLIQ intervention varied between areas. Detail on the differences in the implementation between the four areas, together with a timeline of implementation is presented in Appendix E (p320). In order for the analysis of individual outcomes to take account of the quality of the CLIQ process in the relevant area, a single variable representing the success of implementation was needed to reduce the complex experience and findings regarding implementation, into a usable form for cross analysis. Six indicators of implementation were identified by reflecting on important aspects of the par process. Each area was assessed on these indicators and a summary of the results are presented in Table 6-3.
When computer training was delayed, this often affected the project negatively. Rows 1, 2 and 3 (Table 6-3) show that eMpumalanga enjoyed the best computer training schedule overall. The long delay between modules 1 and 2 in eNingizimu had a detrimental effect on participants’ learning processes, despite timing between later activities being good. The delivery of computer training was the worst in eNtshonalanga, with participants only receiving phase 1 training (which was severely delayed).
Table 6-3: Indicators of success of CLIQ implementation
Process Indicator Manje Sites Maduzane Sites
eMpumalanga eNingizimu eNyakatho eNtshonalanga
1] Months between initial- QLA and phase 1 training (desired:
1-2 months for manje, and 1 month after 2nd initial-QLA for maduzane)
Acceptable-Good:
3 months, slight delay due to software problem on computers.
Problematic:
10 months till completion of phase 1 module 2 training: mod1 and mod2 training were separated by 8 months due to lack of TC connectivity.
Problematic:
3 months after 2nd initial-QLA (and 10 months after 1st initial-QLA) due to scheduling problems with trainer.
Problematic:
13 months after 2nd initial-QLA (and 20 months after 1st initial- QLA) due to non- functional TC
2] Months between phase 1 and phase 2 training (desired: 6 months)
Acceptable:
11 months, delayed by problems
accessing trainer
Good:
4 months Acceptable:
8 months, but delayed due to TC non-functionality
Problematic:
No phase 2 training due to delayed phase 1.
3] Months between phase 2 training and final- QLA (desired: 5 to 6 months)
Acceptable:
2 and 3 months Good:
4 months Problematic:
2 weeks, due to knock-on effect of initial delays and subsequent TC non-functionality
Problematic:
2 weeks between phase 1 training and final-QLA
4] Access to free computer hours at local TC
Good:
access to TC was reasonable regarding supply side issues
Acceptable: some problems relating to non-connectivity, opening hours of TC and business of TC
Problematic:
hours of use per visit were restricted and TC often too full to use, CLIQ free use was discouraged
Problematic:
During 2 weeks between training and final-QLA, access to TC was severely restricted.
5] TC facilitator on hand and
generally willing to help participants
Good:
TC facilitators kept good participant- use records and assisted them with computer use most of the time
Problematic:
high turnover of TC facilitators and lack of motivation and interest from some facilitators
Problematic:
disinterested TC facilitator
Problematic:
No TC facilitator with access to TC
6] Relationship between
fieldworkers and participants
Good:
excellent rapport and many friendships made
Acceptable: Acceptable-Good:
fieldworkers were comfortable and made some friends
Acceptable:
Implementation Ranking
1st 2nd 3rd 4th
(where 1st =best & 4th =worst)
Area differences in the functionality of computers and the internet reduced comparability across sites (see row 4), because participants’ experiences of attempted computer use was affected by whether the computers and internet were working when they reached the telecentre. The attitude of the telecentre facilitator also affected participants’
experience of using the telecentre in terms of whether or not participants felt welcomed and entitled to use the computers, and whether they were assisted when needed. Access to the telecentre to use free hours and assistance from the telecentre facilitator placed
eMpumalanga as the best, and eNtshonalanga and eNyakatho, both as the worst (see rows 4 and 5, Table 6-3). The personal relationships established between fieldworkers and
participants from each area, was an indicator of the level of rapport established between the two groups, influencing the quality of information (see row 6, Table 6-3).
Reflecting on the entire process within the four areas, my subjective ranking of implementation places eMpumalanga first, followed by eNingizimu and then eNyakatho, with eNtshonalanga last. The implementation ranking and participation ranking therefore placed the areas in the same order. To cross check the ranking of areas, participation statistics were analysed on an activity basis by area (rather than on the total number of activities per participant). The result of this analysis confirms the participation and
implementation rankings shown above in Table 6-2 and Table 6-3 (see Appendix I-Table 1, p327). CLIQ participation and implementation can therefore be viewed as a continuum, as illustrated in Figure 6-1.
Figure 6-1: Continuum of CLIQ implementation and participation
Note: The position of the four areas on this continuum is a guesstimate, guided by the area’s rank and my subjective perception of implementation in each area.
The level of implementation in the area, together with an individuals’ level of
participation can be construed as varying levels of a CLIQ dose, representing the intensity of the intervention.77 For example, those from eMpumalanga who participated in all activities got more of a better CLIQ, while those from eMpumalanga who participated in half the activities got less of a better CLIQ. Those from eNyakatho who participated in all activities, got more but of a lower quality CLIQ, and so on.
With a sample of over 100 participants, some quantitative analyses of the results were possible. Individual levels of participation (as good-average-poor) from Table 6-2 and area- based implementation ranking from Table 6-3 were used to quantitatively analyse results with respect to QoL change and CLIQ impact. A quantitative analysis of reasons for QoL change and the nature of CLIQ impact was based on codes that resulted from an intensive process of coding 113 participants’ IIDIs. I used Grounded Theory (GT) techniques to explore the outputs from fieldwork, attempting as far as possible to create categories for analysis which emerged from the IIDI notes. GT is regarded as suitable for use within participatory research:
[GT] sets out to find what theory accounts for the research situation as it is. In this respect, it is like action research: the aim is to understand the research situation.
The aim ... is to discover the theory implicit in the data. (Dick, 2005:3)
GT was also suitable because it does not seek to impose theory on findings in order to make sense of them: “Grounded theory is a qualitative approach that generates theory from observation. It provides the structure often lacking in other qualitative approaches without sacrificing flexibility or rigor” (Calloway and Knapp, n.d.:1). Following the quantitative analysis of results in chapter 7, a qualitative analysis is presented based on participants’
experiences in conjunction with my insight and analysis of project implementation, as presented in this chapter.
6.4.2 The Core Sample
Iterative and cyclical processes are most useful. Similar to the action reflection cycle (see subsection 4.3.5, p83), I engaged in an unplanned, non-linear and iterative process of
77 The analogy of medicinal dosage was suggested by Julian May (principle investigator for CLIQ), who at the time was also my supervisor.
“If I had changed (my well- being) I would not be attending at CLIQ project” (JikileF23).
analysis that has included stages of fieldwork, reading, reflection, writing and debate. One benefit of this is that assumptions and decisions can be revisited based on new insights arising from intermediate findings. This has occurred a few times within the process of analysing CLIQ outcomes, which was possible due to the length of time I have been engaged with the analysis and write up of different aspects of the project.
In early 2011, I had to decide which participants’ data would be included in the analysis of impact and which would be excluded. Selecting those from whom we had data regarding their perceived level of QoL in mid 2008 and again at another point up to mid 2010 (i.e. from the second initial-QLA, the mid-QLA or the final-QLA) allowed for the identification of change in QoL and seemed a logical, implementable and defendable rule. Implementing this rule led to the demarcation of an impact sample of participants.
With the benefit of greater insight into individual experiences and a better understanding of how different area processes affected local outcomes, I reviewed this
decision in early 2013 for the purposes of analysing empowerment outcomes for this thesis.
With eNtshonalanga participants not having the experience of the mid-QLA, phase 2 training or computer use (beyond a week) and with the extensive delay in their phase 1 training, this rule no longer seemed appropriate. Of the 37 eNtshonalanga participants in the impact sample, 14 (or 38%) did not participate beyond the second initial-QLA in early 2009. QoL change data for these 14 participants and another seven from other areas, was limited to the direction of change and did not include any detail on reasons for change or CLIQ impact.
In order to determine extent of impact, it seemed more reasonable to base
calculations on a sample for whom we had information on CLIQ impact, which I refer to as the core sample. Table 6-4 shows the spread of the core sample of 92 participants across the four areas. The process at eNtshonalanga was substantially different to the processes in the other three areas. A delay of about one year between the second initial-QLA and phase 1 training (April 2010) contributed to attrition which was 30%, compared to 11% at the other rural area (eMpumalanga). The impact of CLIQ is discussed with respect to either the
impact sample (113 participants) or the core sample (92 participants) in chapter 7, as appropriate.
Table 6-4: CLIQ’s core sample