[Reliability of anthropometric data obtained in children seen at the Primary Public Healthcare Service Network in Alagoas, Brazil].

INTRODUCTION Monitoring child growth is a very important tool not only to build useful indicators for the evaluation and planning of public policies, but also for appropriate care to child health. In order to accomplish that anthropometric data from services need to present satisfactory reliability. OBJECTIVE To investigate the reliability of weight and height variables obtained from children at public healthcare services (PHS) of Alagoas. METHODS To build the sample, 20 municipalities were drawn among the 102 of the state. Then, two PHS were drawn per municipality, and in these, about ten children under five years. The final sample consisted of 40 PHS and 347 children. Anthropometric measures (weight and height) were taken (1) during routine service (S); (2) by the researcher using standard equipment (R), and (3) by the researcher using service equipment (RS). RESULTS The Kappa test indicated a "substantial" level of agreement (K=0.69) in the classifications of weight-for-age (WA) between S and R and "almost perfect" (K = 0.83) in those between RS and R. Regarding height-for-age (HA), the level of agreement between S and R was "fair" (K = 0.27), to "moderate" (K = 0.56) between RS and R. A Technical Measurement Error according to age groups (< or > 24 months) indicated problems in the technique and equipment. Measurements of body weight obtained in PHS routine were systematically overestimated (p<0.05; sign test), leading to low sensitivity, high specificity and a high rate of false negatives (38.5% for WA and 57.1% for HA). CONCLUSION The low reliability observed for the height variable cautions against its utilization to build health indicators, until the scenario can be improved. The weight variable may be used, but its quality must be improved by training anthropometrists, and through the acquisition and maintenance of equipment.