Article type
Year
Abstract
Background:
In a qualitative evidence synthesis (QES), too much data due to a very large number of included studies can undermine thorough analysis. In this QES, we built on our previous experience of developing and applying a sampling framework to limit the number of included studies for data extraction. Reflecting on our application of sampling in QES can contribute to strengthening QES methodology.
Objectives:
To discuss the lessons learnt from developing a sampling framework for a World Health Organization (WHO) commissioned QES on targeted client mHealth communication to users, clients, patients and the community.
Methods:
First, we divided the included studies by population group. Second, we mapped eligible studies by extracting key information from each study. Third, we developed and applied a sampling frame within each of the population groups based on previous sampling frames. The sampling frame included the following steps:
1) Include studies set in low- and middle-income (LMIC) settings.
2) Include studies scoring a three or more on a scale of data richness.
3) Include studies from the widest variety of mHealth intervention topics.
Results:
Sampling for this review built on our experiences of developing a sampling framework. However, we had to think through how to alter it to meet a wide range of population groups, mHealth interventions and health topics. We will discuss lessons learnt from applying purposive sampling to this review. These included challenges such as sampling within population groups, studies falling into two different population groups and a number of studies with thin data. The sampling ensured that the few studies that do exist from LMICs were included. Purposive sampling was harder to apply in this QES due to multiple target populations, with some studies falling into more than one target population. A potential problem with this sampling approach is that we may have inadvertently excluded deviant cases.
Conclusions:
We acknowledge that there is a trade-off between including all studies and trying to ensure that volume of data does not undermine the analysis process. Future review teams may want to consider scanning included but not sampled studies for deviant cases once findings have been identified.
Patient or healthcare consumer involvement:
The sampling framework ensured that voices from all available settings were included.
In a qualitative evidence synthesis (QES), too much data due to a very large number of included studies can undermine thorough analysis. In this QES, we built on our previous experience of developing and applying a sampling framework to limit the number of included studies for data extraction. Reflecting on our application of sampling in QES can contribute to strengthening QES methodology.
Objectives:
To discuss the lessons learnt from developing a sampling framework for a World Health Organization (WHO) commissioned QES on targeted client mHealth communication to users, clients, patients and the community.
Methods:
First, we divided the included studies by population group. Second, we mapped eligible studies by extracting key information from each study. Third, we developed and applied a sampling frame within each of the population groups based on previous sampling frames. The sampling frame included the following steps:
1) Include studies set in low- and middle-income (LMIC) settings.
2) Include studies scoring a three or more on a scale of data richness.
3) Include studies from the widest variety of mHealth intervention topics.
Results:
Sampling for this review built on our experiences of developing a sampling framework. However, we had to think through how to alter it to meet a wide range of population groups, mHealth interventions and health topics. We will discuss lessons learnt from applying purposive sampling to this review. These included challenges such as sampling within population groups, studies falling into two different population groups and a number of studies with thin data. The sampling ensured that the few studies that do exist from LMICs were included. Purposive sampling was harder to apply in this QES due to multiple target populations, with some studies falling into more than one target population. A potential problem with this sampling approach is that we may have inadvertently excluded deviant cases.
Conclusions:
We acknowledge that there is a trade-off between including all studies and trying to ensure that volume of data does not undermine the analysis process. Future review teams may want to consider scanning included but not sampled studies for deviant cases once findings have been identified.
Patient or healthcare consumer involvement:
The sampling framework ensured that voices from all available settings were included.