Child vaccination · BMJ Glob Health · 2021
“Phone calls got newborns vaccinated on time.” Restore a missing minus sign: maybe they did nothing.
The calls’ effect on vaccination on time at birth. Dot: score. Bar: range.
The paper says
In rural India, says the summary, families who heard at least half of a phone service’s vaccine calls more often had newborns vaccinated on time at birth: an effect score of 0.08 (0 = none), with a range of 0.08 to 0.24, all above zero.
Its own data show
Its own table gives the same 0.08 with a range of −0.08 to 0.24, which includes zero. Such ranges sit evenly around the score: 0.08 is 0.16 from each end. The summary’s range lacks the minus, leaving 0.08 at one end: impossible. Only the table’s printed p = 0.03 disagrees: its own range implies 0.33.
A range spans the effects the data can’t rule out; one that includes 0 means the calls may have done nothing. A p-value is how often luck alone gives a result this big, and the range fixes it: 0.33 is 1 time in 3, 0.03 is 1 in 33. This is one of the calls’ only two reported benefits.