We examined the current state of BMI documentation and documentation and control of associated risk factors by BMI category, based on EHR data from almost 80,000 adult patients seen in a 12-clinic primary care network during 18 months in 2005-2006. Our findings demonstrate variations in risk factor recognition and control for obese persons compared to those of normal weight, as well as for those with and without CVD or type 2 diabetes. We found that documentation of BMI varied widely by clinic site and was overall low. Among patients with a BMI recorded, documentation of most risk factors was higher in patients with obesity compared with normal weight patients; however, control of risk factors was poorer in obesity than normal weight. Patients without a documented history of CVD or diabetes had strikingly more dissimilar rates of documentation and control between weight categories than patients with CVD or diabetes. Overall, patients with obesity with or without CVD or diabetes had lower rates of risk factor documentation and control than may be ideal given their high absolute risk of adverse health outcomes.
Our study builds upon findings from previous studies. Lemay et al audited medical records from 465 adult patients seen at a community health center during one week in February of 1999 to evaluate height and weight documentation and obesity diagnosis by the practitioner over a prior six-month period, and found that only 63% of their patient cohort had a height and weight documented [27]. Similar to our study, Lemay looked a group of patients in normal clinical practice; however, we were able to expand upon this study in terms of a far larger study sample as well as a longer sampling frame to assess provision of services (18 months versus 6 months), potentially providing a more accurate picture of risk factor evaluation. Six years after this study and seven years after the establishment of the NHLBI guidelines, we found the percentage of BMI documentation to be essentially the same (63% versus 60.5%). Rifas-Shiman et al studied 5,025 members of the same insurance plan and group practice who were a subset of participants from a cohort study and were continuously enrolled since 1999, had a visit in 2000, had a BMI measurement between January 1, 2000 and December 31, 2000, and who did not have medical conditions related to weight loss or CVD. They found that higher BMI was an independent predictor of increased fasting glucose, triglyceride, LDL cholesterol, and HDL cholesterol screening [28]. Rifas-Shiman identified lipid and glucose abnormalities over a two-year period, comparable in time to our study. Neither we nor Rifas-Shiman could evaluate attempts at management, so reasons for increased documentation such as guideline adherence could not be assessed. Our analysis extends prior studies by evaluating whether those with documentation of risk factors had those risk factors in control according to Adult Treatment Panel III (ATP III) metabolic syndrome guidelines for a patient with average cardiovascular risk.
Molenaar et al studied rates of treatment and control of hypercholesterolemia, hypertension, and diabetes in normal weight, overweight, and obese subjects with a history of these conditions utilizing a CVD-free subset of the Framingham Heart Study. They found that subjects with hypercholesterolemia and hypertension who were obese were more likely to have these conditions treated than normal weight subjects. Rates of control of hypertension and hypercholesterolemia were uniformly poor and did not differ between weight groups. Rates of control of diabetes were poor among all three weight groups, but subjects who were obese were less likely to have control of fasting blood glucose than normal weight subjects. The goal of our study, however, was different from that of Molenaar. Molenaar studied rates of treatment and control of hypercholesterolemia, hypertension, and diabetes in normal weight, overweight, and obese subjects with a history of these conditions who were free of CVD, a subgroup of patients with a clear indication for BMI screening. Our goal was to evaluate in all patients, both with and without obesity-associated risk factors, the current state of BMI screening and subsequent screening for associated risk factors by BMI group, according to HEDIS and NHLBI guidelines. Molenaar et al examined a well-known, standardized study population, where only 196 subjects had missing BMI data. Our population was a non-standardized data source, and our subjects were a mixture of both those with and without CVD and its risk factors, with further subanalysis in patients with a history of CVD and diabetes. Despite these differences, findings from the structured Framingham population and from our analysis of usual clinical care are strikingly similar. This minimizes to a large degree the concern that high rates of missing BMI information could have distorted our findings, and may explain why our percentages of control were higher than those found by Molenaar, albeit still low overall [29].
The rapidly growing prevalence of obesity has pushed BMI assessment to appropriate prominence as a newly proposed reportable HEDIS measure. BMI assessment will be made by several means, including survey of EHRs in health care networks. Our results suggest that the HEDIS BMI Assessment measure has potential to provide a timely quality and safety foundation to improve care for patients with obesity. At least in our large primary care network, there clearly is substantial room for improvement in documentation of BMI and documentation and control of BMI-associated risk factors. While height and weight and BMI documentation may reflect individual physician practice styles, by speaking with medical directors we found that lack of height and weight appeared to be a clinic-level and not an individual PCP issue, with height and weight recording often performed or not performed before the PCP sees the patient. It is expected that introduction of the HEDIS BMI Assessment measure will improve this state of affairs, although the effectiveness of BMI documentation alone to improve care remains to be demonstrated.
According to our study, patients with already documented CVD or diabetes are more likely to have risk factor documentation and control regardless of BMI category. Those without a documented history of CVD or diabetes demonstrated more variation in risk factor documentation and control by BMI category. At least two recent studies corroborate these data. Melamed et al measured BMI in 289 patients in seven family practice clinics in Israel, and found that BMI was documented in only half of obese patients and 39% of overweight patients, and that patients with other chronic medical conditions were more likely to have BMI documented than those without documented comorbidities [30]. Waring et al looked at 2,330 overweight and obese patients included in the Cholesterol Education and Research Trial, and found increased odds of overweight or obesity management in relation to weight-related comorbidities for those with moderate or severe obesity [31].
Risk factor control appears to be related to a previously diagnosed risk factor and not to obesity. These findings become even more relevant in light of recent studies that demonstrate increased CVD risk in patients who are overweight and obese compared to normal weight patients, independent of hypertension and hypercholesterolemia [32]. This implies an important need to recognize overweight and obesity, ideally using a simple technique such as BMI, in order to enhance CVD and diabetes complications prevention. Our findings suggest that PCPs are aware of CVD, diabetes, and obesity as strongly tied risk factors, but that they may not recognize obesity as a risk factor for morbidity and mortality independent of these other comorbidities.
Limitations
Our results must be interpreted with some limitations in mind. This is clinical, not research, data, which naturally suffers from information that is missing and inconsistent in its recording. We carefully addressed missingness with multiple methods of ascertaining exposures, and addressed inconsistencies in data recording by removal of clinically illogical extreme outliers. Evaluation of BMI documentation rates, with the inherent missing data, was a goal of our study. Furthermore, despite its limitations, evaluation of clinical data is a strength of this study, in that it provides a glimpse into current obesity care and insights into improving this practice. Although our data were derived from a single academic health care network, the sites included a representative mixture of urban, suburban, and hospital-based practices, making our findings generalizable and potentially replicable. Another important strength of our study is the use of a long-standing, widely used EHR encompassing all aspects of patient care from a large network of diverse clinics and patient populations. EHRs, while not a perfect tool due to the potential for improperly entered or overlooked data, have great potential for research, with studies showing that EHRs have potential for increasing documentation and treatment of obese patients [33]. Finally while there were differences seen in the percentages of documentation of risk factors, this may be due mainly to test indication, whereby certain tests such as cholesterol levels are more likely to be ordered on most patients than fasting glucose. However, our overall documentation numbers were large enough to yield consistent results across BMI categories.