Table 3 A summary of the PMTCT-ACT process

Part 1

• Create a new time variable (days on ART for each visit date relative to the date of ART initiation).

• Expand the data set to a full, balanced panel that covers the time period needed for the analysis (e.g. from the earliest date of ART initiation through the latest date of delivery for all patients in the analysis).

Part 2

• Reshape the panel dataset (a “long” dataset) to a “wide” dataset.

Part 3

• Use the pill count calculator to count the number of days of ARVs on hand for each day in the wide dataset.

• Reshape the dataset from wide to long (back to a panel dataset).

Part 4

• Creates a new time variable for each date that is the number of days from that date to the date of delivery.

• For each date, create a new indicator variable (hasarvs) equal to 1 if the patient has 1 or more days on ART on hand that day (that day is potentially covered with ART).

Part 5

• Collapse the panel dataset over a specified period of time.

• The example provided is the final 24 weeks of the pregnancy (e.g., days to delivery −168 to 0).

• Create a final outcome measure (coverage with ART during the final 24 weeks of pregnancy).