Self-Reported Adherence Struggles to Match Pill Counts in Ethiopian Pregnancy Supplement Study
Pregnancy nutrition programs depend on a simple question with a complicated answer: did the patient actually take the supplement? In a baseline analysis from a quasi-experimental study in Ethiopia, investigators compared several self-reported adherence measures for multiple micronutrient supplementation against pill counts in pregnant women receiving antenatal care. The comparison matters because the entire value of a nutrition program can be distorted if adherence is measured poorly. A tool that looks efficient in the clinic but drifts away from actual consumption can make a successful intervention appear weak, or a weak intervention look stronger than it is.
The study was built around a real-world public health setting, not an idealized trial room. That matters in maternal nutrition, where supplements are distributed through routine services and adherence often depends on memory, counseling quality, pill burden, side effects, household dynamics, and whether tablets are shared, saved, or discarded. In that environment, a self-report instrument has to do more than sound practical; it has to track actual behavior closely enough to guide program decisions.
Why adherence measurement is such a fragile problem in maternal nutrition
Adherence to antenatal supplements is not a binary event lived in a sterile research environment. It is shaped by intermittent clinic attendance, competing household demands, and the everyday reality of pregnancy symptoms and appetite changes. When women are asked whether they took their tablets, the answer is filtered through recall and social desirability. When pill counts are used instead, the method depends on a very different assumption: the tablets that remain in the bottle reflect tablets that were not swallowed.
That assumption is useful, but it is not perfect. The study’s own framework acknowledged the major weaknesses of pill counts as a reference standard. Women may share tablets with family members, discard them, or overconsume them, and any of those behaviors can distort the count. Even so, pill counts remain a common comparator because they are concrete and easier to audit than recollection alone. The baseline analysis therefore treated pill counts as the reference while testing whether self-reported measures could approximate them with acceptable accuracy.
The investigators examined three distinct statistical approaches to judge validity: receiver operating characteristic analysis, inflation factor analysis, and Bland-Altman agreement testing. Each method asks a different question. ROC analysis evaluates how well a measure separates adherent from nonadherent participants. Inflation factor analysis examines whether self-report systematically exaggerates adherence. Bland-Altman testing looks for agreement between methods across the range of observations, not just at the average. Used together, these tools can expose whether a questionnaire is merely convenient or actually dependable.
A baseline snapshot designed to test measurement, not assumptions
This was baseline data from a quasi-experimental study, so the analysis focused on the starting point before the intervention component unfolded. That design choice is important because it isolates the measurement problem itself. If self-report is unreliable at baseline, the same weakness can follow the program through follow-up and blur any true effect of nutrition counseling or service delivery changes.
Pill-count constraints were built into the interpretation from the beginning. The investigators also flagged the possibility that participants taking more than the prescribed daily dose could skew adherence estimates based on remaining tablets. In other words, even the reference measure is partly vulnerable to behavior outside the bottle. That is a familiar problem in maternal supplementation studies, where the line between prescribed intake and real-world use is often less tidy than investigators would like.
The statistical handling reflected similar caution. Bland-Altman analysis depends on assumptions of normality and homoscedasticity. Formal normality tests failed, yet the data stayed within acceptable skewness and kurtosis ranges, allowing the investigators to proceed with interpretation. The result is not a simple yes-or-no verdict on self-report. It is a more clinically useful reminder that measurement validity can look acceptable by one test and shaky by another, depending on which kind of error is being examined.
What the comparison means for antenatal programs
For maternal nutrition programs, the practical issue is not academic elegance. It is whether frontline teams can trust the adherence numbers they collect. If self-reported adherence systematically overstates supplement use, then program coverage may look better than it is, and counseling strategies may be deployed on a false premise. If self-report underestimates use, effective services may be judged too harshly. Either error can lead to bad policy.
The Ethiopian setting gives this problem particular force. The country carries a heavy maternal nutrition burden, and multiple micronutrient supplementation is intended to support pregnancy outcomes through improved intake of key vitamins and minerals. The biologic rationale is straightforward: maternal micronutrient sufficiency supports fetal growth, maternal hematologic status, and broader pregnancy health. But none of that benefit can be confidently assessed unless adherence is measured in a way that tracks actual tablet use with reasonable fidelity.
This is where the comparison between methods becomes clinically consequential. ROC analysis can reveal whether a self-report item can distinguish adherent from nonadherent participants. Inflation factor analysis can show whether the measure consistently inflates adherence estimates. Bland-Altman plots can reveal whether disagreement widens as adherence changes. Each of those patterns would mean something different for a nutrition program manager deciding how much faith to place in routine data.
A one-size-fits-all adherence question is rarely enough. Recall bias, tablet sharing, and deliberate misreporting can each pull estimates in a different direction. In antenatal care, where supplement delivery often depends on the same staff who counsel, dispense, and document use, the temptation is to prefer the simplest measure available. Simplicity helps workflow, but it does not guarantee validity.
The measurement lesson for real-world maternal care
The study’s value lies in the discipline of its comparison. Rather than accepting self-report as a convenient stand-in for behavior, it tested whether different self-reported adherence measures could hold up against a more concrete reference. The answer carries weight for any setting where multiple micronutrient supplements are distributed at scale and program success is judged partly through patient report.
One sentence captures the problem plainly: pill counts are helpful, but they are not a perfect truth standard. When tablets can be shared, discarded, or taken in extra doses, even the best reference has blind spots.
For researchers, the baseline analysis reinforces the need to validate adherence tools in the same populations where they will be used. For maternal health services, it points toward a more cautious reading of self-reported supplement intake. And for policy teams designing nutrition programs, it argues for measurement strategies that do not rely on memory alone when the stakes include fetal development and maternal health.
The abstract leaves open the downstream comparison with follow-up data, but the baseline message is already clear: in pregnant women receiving multiple micronutrient supplements in Ethiopia, self-report has to earn its credibility against more objective measures, not assume it from convenience.