Qualitative Data Analysis
Qualitative data analysis is resource intensive and time-consuming. For interviewing and focus group discussions, qualitative analysis often includes identifying categories, patterns, themes, meaning, and concepts that emerge from responses and linking the concepts to substantive theories. Qualitative data analysis is an iterative process, and it is important to include someone who has experience conducting qualitative data analysis in your assessment team if you will be asking a lot of open-ended questions or conducting focus group discussions.
Quantitative Data Analysis
In theory, the same analytic techniques can be applied to primary and secondary data. Practically, however, secondary data are typically far more numerous than primary data, thus certain analyses may only be possible on secondary data (for instance, longitudinal analyses). It is also particularly important in secondary data analyses to make special efforts to understand the strengths, limitations, and context of the data set since those analyzing the data are not the same individuals who designed and conducted the data collection.
Descriptive Statistics: The indicators defined during the planning stage can be presented across the whole dataset, and stratified by variables such as geography, age, or disease group. Results can be presented as tables, maps, and figures, which should include time trends if longitudinal data are available. Case data should be presented as both counts and either rates (incidence) or proportions (prevalence). Estimates of uncertainty (e.g., confidence intervals) should be included in all cases. Analyses are typically conducted using statistical software such as R, SAS, Stata, or Python.
Hypothesis Testing: While most surveillance assessments will use only descriptive statistics, it may be of interest to test hypotheses around drivers of surveillance system performance. For instance, is timeliness associated with human resources for surveillance? Was there a significant effect of a given policy change on a surveillance system outcome? Hypotheses should be pre-specified at the design phase, explicitly incorporated into the data analysis plan, and incorporated into sample size calculations.
Modeling: Depending on the goals of the assessment, it may be necessary to move beyond descriptive statistics. For instance, there may be an interest in using modeling techniques to smooth indicator estimates over space or time, to impute missing cases of disease to surveillance system gaps, or conduct “nowcasting,” which uses statistical models to address lags in disease reporting. However, these methods are more commonly used to analyze surveillance data for epidemiologic purposes, rather than assess the surveillance system itself.
Stakeholder Engagement in Data Interpretation
Once data are analyzed, it is critical to engage stakeholders in data interpretation to enhance the quality and relevance and to ensure that stakeholders are part of the process including development of next steps or application of findings to surveillance activities, promote ownership of the findings and next steps, allocate resources to programming, and support evidence-based decision making. Consider waiting to finalize reports until a stakeholder workshop has been held and their feedback is integrated into the final report.