We found that 30% of overweight children attending the five-year PCHE were assessed to be normal weight by the GPs. The presence of previous notes on abnormal weight was associated with an increased sensitivity of the GPs’ assessment. Moreover, a high frequency of visits to practice and use of BMI tended to be associated with a higher sensitivity. Specificity has traditionally been regarded as a factor of paramount importance in the diagnosis of childhood overweight [18
]. The specificity was high in our study.
The PPV in our study was 60.2%, implying that only 60.2% of children assessed to be overweight by their GPs were overweight according to DS. The contrast between the relatively high specificity and the low PPV can be explained by the relatively low prevalence of overweight in the population. The NPV of the GPs’ assessment implies that most of the children who were assessed to be normal weight by their GPs were also normal weight according to the DS. PPV and NPV relate to the studied population and thereby account for the problems of misclassification.
The GPs’ assessments fell short of correctly identifying all cases of overweight, yet they remain a valuable tool in detecting childhood overweight.
The GPs’ identification of overweight was higher than in medical record review studies of paediatric clinicians, documenting overweight in 28% [10
] and 31% [19
] of overweight children and obesity in 53% of obese children [2
]. The shared design of these studies was a retrospective review of medical record diagnoses. Hence, the physicians may not have paid particular attention to weight.
We used DS as gold standard for defining overweight and obesity in our cohort. Although BMI works fairly well at the population level, this measure cannot determine the composition of body mass, and thus cannot discriminate between lean body mass and fat-containing body mass [20
] and caution is warranted when predicting an individual's content/distribution of body fat [21
]. Hence, use of BMI carries the risk that someone with a low level of body fat and very large muscle mass may be misclassified as overweight due to a high BMI. This may to some extent explain the discrepancies between the GPs’ assessment and BMI defined overweight.
Analyses were made using IS [16
] and DS [15
] in order to be able to compare with international studies. Differences in prevalence due to the two standards may be explained by differences in overweight prevalence in the populations founding the basis for these standards.
Recall bias may be reduced because of the survey's relation to a concrete PCHE. Second, the study captured the GPs’ clinical assessment of actual children attending their clinics as opposed to previous studies using body images to evaluate physicians’ identification of overweight [8
]. Hence, we believe that this study provides a high-quality evaluation of GPs’ clinical assessment of childhood overweight.
The participating GPs were younger than non-participating GPs and female GPs were over-represented compared with non-participating GPs. Further, it is likely that the GPs in our study had a particular interest in the subject and we may therefore expect our results to overestimate identification of overweight in children. Selection bias due to overweight may be suspected because of the lower prevalence of overweight in the cohort compared with the background population (15.9% and 11.6% for six-year-old girls and boys, respectively) [22
]. This selection may partly be explained by the absence of socioeconomically less privileged families from the PCHEs [13
], for instance bilingual children [23
Notably, the lacking validation of the GPs’ weight-for-stature assessment scale prompts limitations. We aimed at targeting all children with a weight-for-stature somewhat above normal. We therefore provided the GPs with a five-point assessment scale. It is, however, uncertain whether the GPs would have categorized children with a weight-for-stature above normal as overweight in the absence of such a scale.
There seemed to be some statistical association between the surveyed variables “Attendance frequency” and “Familiarity”. However, both variables were included in the analysis in order to deal with situations where the PCHE was conducted by a doctor less familiar with the child and its family, e.g. GP trainees.
The variables “Familiarity” and “Previous observations of abnormal weight” seemed to be associated. Whether familiarity in itself gives rise to more observations of abnormal weight is not known.
The GPs’ assessment of overweight may be influenced by the socioeconomic status of the children. Unfortunately, our data did not permit analysis of such associations.
As GPs measure weight and height in most children until the age of five at the PCHEs [14
], this clinical setting provides an opportunity for early identification of overweight. Our study shows a gap between GPs’ assessment of overweight and BMI-defined overweight. More explanations may be likely. As childhood overweight is a relatively new phenomenon in Denmark, there may remain doubt about overweight definitions. The widely used growth charts for height and weight may be insufficient for diagnosing overweight. Earlier studies emphasized perceived lack of professional success and management opportunities as barriers for health professionals [24
Overweight in childhood is associated with overweight in adulthood [26
] and with multiple ensuing health problems [28–32
]. Failure to identify childhood overweight may impair prevention of adult overweight and disease. A qualitative study of obese adults revealed that restricted attention to obesity could lead to neglect of patients’ problems [33
]. Further research is needed to form the basis for better tools for identification of children at high risk of developing permanent overweight and ensuing health problems.