Chapter 4 | Phase 0: Study Design and Sampling

A. Understanding the Study Design 

What is a study design and why does it matter? 

The study design is the plan for answering the critical monitoring and evaluation questions underlying the pharmaceutical system or more specifically, the  impact of DOH-PD policies and their accompanying interventions. Some of these questions include: How accessible and available are pharmaceuticals? How affordable are they? Are current prescribing and dispensing practices promoting rational use? In other words, the study design is like a roadmap that guides you throughout every step of the data collection process and tells you: 

1. What information to collect

2. Where to collect it from 

3. When to collect it 

4. How often to collect it  

5. The kinds of comparisons that will be made

6. How the comparisons will be made 

Like everything else in this guide, a strong appreciation of and adherence to this study design ensures the rigor of the data collection process and in turn, the quality of collected data. Again, your role in the field as the eyes and ears of the DOH-PD allows you to capture relevant information that can be made actionable at the policy-level. 

What study design am I using? 

This study will use two designs: (1) cross-sectional and (2) longitudinal (true longitudinal and repeated cross-sectional). We selected these designs because of the diversity of indicators being measured. That is, each indicator requires different information coming from different sources with varying frequency of collection, and so on. A diagram of the study design is provided below (Figure 4.1):

In simple terms, indicators requiring one-time measurement will use a cross-sectional design as this is useful for establishing a baseline or marking the completion of input indicators. On the other hand indicators requiring more frequent data collection– namely, monthly, quarterly, or annual, will use either a true longitudinal or repeated cross-sectional (also known as pseudo-longitudinal) design. This study design allows you, as data collectors, to capture both the static (cross-sectional) and changing (longitudinal) dynamics of the PD policies and interventions across the variety of the indicators in the Monitoring & Evaluation Framework (MEF). Key characteristics of these study designs are summarized below (Table 4.1): 

What is my role as a data collector?

As a data collector, your role is to implement the study design we developed. You will follow our guidelines on which design to use for each indicator, which not only ensures consistency but also allows for valid conclusions to be made from the data. You can look at the Data Collection Frequency (When to collect) column in the Detailed Description of Indicators (see Appendix H) for the specific indicator you’re interested to determine the study design:

1. Cross-sectional if the indicator requires a one time measurement. You will collect data only once and do not need to revisit the same sampling units multiple times.

2. Longitudinal if the indicator requires monthly, quarterly, or annual measurement.

   1. Longitudinal: You will visit the same sampling unit multiple times

   2. Repeated Cross-sectional: You will measure the same indicators in different sampling units multiple times 

B. Selecting the Sample 

Refer to the Sampling Design Worksheet completed during the Training of Trainers Workshop. The template is available here or in Appendix A. You should adhere to this during the entire data collection process. 

What is sampling and why does it matter? 

Sampling is the process of selecting a subset of sampling units from a defined population. In the context of this study, sampling allows us to gather a nationally representative sample without having to collect data from every health facility, pharmacy, patient, or pharmacist, or healthcare provider in the country. 

What sampling design and sample size am I using? 

We developed a sampling procedure following a multi-stage sampling design which is appropriate for collecting data from a large, geographically dispersed population. This basically means that sampling happens in stages, narrowing down from larger units (regions) to smaller ones (province, city, health facility, individual respondents). A detailed overview of this is available in Appendix A. 

We have also computed the sample size for this M&E study using a sample size for a frequency in a population calculator, available through OpenEpi version 3.01. After applying assumptions(1) and a 20% buffer to account for refusals and other causes for replacement, the final computed sample size is 460. To get the sample size for your pilot data collection, we used 30% of 460 (n = 138). These sample sizes were then allocated proportionately across regions using population size as proxy, as seen in the table below. The final column, Sample size of Pilot Study, is what you should refer to in order to determine your region’s sample size: 

How do I implement the Sampling Strategy?

Below is a step-by-step guide on how to implement the sampling strategy: 

1. Develop a Sampling Frame: Obtain a complete and up-to-date list of health facilities and pharmacies in your region. This will serve as your sampling frame, comprised of the following sampling units: 

   1. Health Facilities (hospitals, health centers, rural health units, barangay health units, etc.)

   2. Pharmacies, Botika ng Bayan sites  

   3. Individuals (patients, healthcare providers, pharmacists) - since a sampling frame for individuals will be difficult to assemble quickly, we will resort to purposive or convenience sampling for the pilot study.  

2. Stratify the Sampling Frame: Stratify the sampling frame of your region by province and city.  Each province or city will act as a separate strata.

3. Select Municipalities and Cities Using Random Sampling: All provinces will be selected. Within each province or city, randomly select one municipality or city respectively. To do this, use the random number function on Google Sheets following the steps outlined in Table 4.3. 

4. Select Health Facilities and Pharmacies within the Selected Municipalities and Cities: Automatically select the regional hospital and provincial hospital. Within the selected municipality or city, randomly select one health center. Randomly select pharmacies up to the sample size allocated to the region. To do this, use the random number function on Google Sheets following the steps outlined in Table 4.3. 

5. If Applicable, Select the Individual Respondents according to the sample size allocated for your region (Table 4.2). To do this, use the random number function on Google Sheets following the steps outlined in Table 4.3. However, for the pilot study, just select whichever physician, pharmacist, or patient is available. 

What are some considerations on the field? 

Although the sampling should have already been done prior to data collection, it can still be subjected to issues you may encounter on the field. Below are potential issues that you may face and what actions you can take to address them.