Recording of weight in electronic health records: an observational study in general practice

In this observational study, data from the EHRs of general practices in the Netherlands that participated in NIVEL Primary Care Database (NIVEL-PCD) were combined with data from the Occupational and Environmental Health Cohort Study (AMIGO study). Both cross-sectional and longitudinal analyses were applied.

The NIVEL-PCD comprises anonymised data from the EHRs of a representative sample (~ 10%) of all general practices in the Netherlands [14]. In general practice, EHRs are used by GPs and practice nurses to record information on consultations, diagnostic measurements, drug prescriptions, referrals, and morbidity according to the International Classification of Primary Care version 1 (ICPC-1).

The AMIGO study is a longitudinal study on the occupational and environmental determinants of disease and well-being. Participants for this study were recruited through 99 general practices that participated in the NIVEL-PCD in 2011 and 2012. All patients born between 1945 and 1981 who were registered at one of the 99 general practices were invited by their GP to participate in the AMIGO study. In total, 14,829 patients filled in the informed consent form and completed the baseline questionnaire between April 2011 and July 2012. The design of the AMIGO study has been described in more detail by Slottje et al. [15].

Information on the patients’ sex, year of birth, height, weight, level of education, smoking status, and alcohol consumption was obtained from the baseline questionnaire of the AMIGO study. Patient age was calculated as 2012 minus their year of birth. Information on GP consultations, diagnostic measurements, and morbidity over the period from 2012 to 2015 was obtained from the NIVEL-PCD.

Figure 1 shows the selection process for the study population included in the cross-sectional analyses, which used data from 2012. Eligibility criteria were applied at both the general practice and patient levels. Due to a failure in data extraction, data from 13 general practices that participated in the NIVEL-PCD were not available. Other practices (n = 35) were excluded due to poor data quality (i.e., data recording < 46 weeks/year, or < 70% of the recorded disease episodes labelled with the relevant ICPC code).

From the selected general practices, patients were excluded if they met the following criteria: (1) incomplete registration in general practice, (2) missing data on height and/or weight in the baseline questionnaire of the AMIGO study, or (3) no consultation with their GP in 2012 (because GPs needed to have the opportunity to record the weight of their patients). In total, 6141 patients from 51 general practices fulfilled these criteria, which appeared to be a representative sample of the total AMIGO study population (see Additional file 1). Subsequently, BMI was calculated using self-reported height and weight from the AMIGO study. A total of 3446 patients were classified as being overweight (i.e. BMI ≥25 kg/m²), who were included in the present study.

For the longitudinal analyses, data from 2012 to 2015 was used, and the eligibility criteria at the general practice level were also applied for the years 2013, 2014, and 2015. Furthermore, only patients who attended at least one annual GP consultation between 2012 and 2015 were selected. The final study population for the longitudinal analyses included 1516 patients.

For the cross-sectional analyses, a binary variable termed “weight recording” was generated, which indicated whether there was at least one BMI or weight measurement recorded in the patient’s EHR in 2012. For the longitudinal analyses, additional binary variables for “weight recording” were generated for the years 2013, 2014, and 2015.

The following variables were generated from data recorded in the AMIGO study: BMI category (BMI ≥25 and < 30 kg/m², BMI ≥30 kg/m²), sex (male, female), age (31–40 years, 41–50 years, 51–60 years, 61–67 years), highest achieved level of education (low, vocational education/community college; intermediate, vocational/high school; high, college/university or higher), smoking status (never, former, current), alcohol consumption (never, ≤1 day/week, 2–3 days/week, 4–5 days/week, 6–7 days/week). These variables were similarly categorised as presented in the design article for the AMIGO study [15].

Four (clusters of) chronic conditions known to be associated with overweight were selected based on morbidity data from the 2012 NIVEL-PCD: cardiovascular disorders (ICPC K74-K76, K86-K87, K89-K92, K99, T93), osteoarthritis (ICPC L89-L91), diabetes mellitus (ICPC T90), and COPD (ICPC R91, R95). We created an additional binary variable indicating whether a patient had none or at least one of the four selected (clusters of) chronic conditions related to being overweight. Furthermore, the mean BMI of each patient was calculated from all available recorded BMI (or height and weight) measures recorded in 2012.

This study evaluated the extent of weight recording in general practices in the Netherlands among an adult population who self-reported as being overweight. Our findings show that greater age, lower education level, and higher self-reported BMI were positively related to weight recording. Furthermore, in accordance with our hypothesis, higher rates of weight recording were found for patients with diabetes mellitus, COPD, or cardiovascular disorders. These are all chronic conditions for which regular weight measurement is recommended in the clinical guidelines for GPs, and for which weight or BMI recording represents a performance indicator within a bundled payment system [12].

The presence of diabetes mellitus was found to be the variable most strongly associated with weight recording, consistent with the findings of other studies [6, 9, 10, 16]. Furthermore, in line with a recent review of similar studies of the UK primary healthcare system, our study indicates that some patients are less likely to be identified as being overweight by their GP, including younger patients and patients without a chronic condition [17]. These findings are also supported by an Australian study which showed a positive association between age and weight recording [10], and a study of Dutch GPs’ weight management policy, which showed that weight was less often discussed with patients without weight-related comorbidities [18].

In contrast to other studies that indicate a higher frequency of weight recording in females, we found no difference between male and female patients. Furthermore, we found that a higher education level was associated with lower rates of weight recording, which differs from the results of a previous study which showed no association between education level and weight recording [16]. The discrepancy in findings related to socio-demographic characteristics may be due to differences in the selection criteria for study populations and the time-frames of studies, in addition to differences in healthcare systems between countries.

A strength of this study is the linkage of the cohort from the AMIGO study with routinely recorded data from general practices, which enabled us to combine information on self-reported socio-demographic and lifestyle determinants with health outcomes recorded in EHRs. Furthermore, to our knowledge, this is the first study to assess weight recording in Dutch general practices for patients who self-reported as being overweight.

A limitation of this study is the generalisability to the total population, as the study population consisted only of adults aged 31–67 years. Furthermore, we selected overweight patients based on self-reported height and weight, meaning that patients who did not identify themselves as being overweight were not included in the study. However, we do not believe that this had a large effect on the external validity of the study, as the proportion of overweight individuals (56%) in our study population is comparable to that of the general Dutch population of adults aged ≥20 years [19]. Additionally, our study showed good concordance between self-reported BMI and mean recorded BMI for patients with available data, so weight status does not seem to have been underestimated.

A representative sample size of the AMIGO study population was selected for the current study, even though it included only a subset of the cohort members. A large proportion of the initial study population had to be excluded due to insufficient data quality of the general practices, which might have resulted in a selection bias. The included general practices, which had higher levels of data quality, could potentially be systematically different to general practices with lower levels of data quality. However, we suspect that most of the variation in data quality among the general practices is unrelated to clinical performance, but instead due to software issues, as suggested by van der Bij and colleagues [4].

In the present study, we only included patients who had attended at least one consultation per year with their GP. Previous studies also only included ‘active’ patients, that is those who underwent a minimal number of consultations during a certain period [5, 9, 10]. Patients who do not consult their GP regularly are probably more healthy, and would therefore have their weight recorded by a GP less often. Thus, weight recording in our study population presumably occurred more frequently when compared to the total overweight population.

Routine weight recording in EHRs could help GPs identify overweight patients and monitor and support them in weight management programs, such as prevention programs that are embedded in primary healthcare [20, 21]. This study showed reasonable completeness of weight recording for overweight patients with a chronic condition, for whom regular weight evaluation is recommended in the clinical guidelines for GPs, and for which weight recording is incorporated as a performance indicator within a bundled payment system. In (relatively younger) overweight patients without a chronic condition related to being overweight, for whom weight measurement is not specifically required, we found that weight was recorded in only a third of these patients over a 4-year time frame. However, in the group of relatively young adults aged 31–40 years, the presence of overweight was already considerably high, with about 40% of patients classified as being overweight. To prevent weight-related health problems and their associated healthcare costs, discussing weight at an early stage should be recommended for this patient group.

Overweight management has been shown to be more frequent among patients with a documented weight status [22]. The importance of discussing overweight was illustrated by a large study conducted in the US, which found that patients were more likely to perceive themselves as being overweight, accompanied by an increased desire to lose weight, if they had been told by their healthcare professional they were overweight [23]. However, research from the UK suggests that GPs can feel uncomfortable talking about overweight, and may not always feel responsible for discussing weight management with their patients [24]. In the Netherlands, most GPs consider weight management to be part of their responsibility of providing care, but they face other barriers such as time constraints [18]. A solution may be for GPs to delegate some weight management tasks to practice nurses, who already play an important role in lifestyle counselling in Dutch primary healthcare [25]. In addition, providing feedback to general practices on their recording habits could help GPs become more aware of these habits, and would probably enhance weight recording [3, 4]. Furthermore, a performance indicator payment for weight recording in all patients could possibly improve recording and support early interventions in overweight individuals.