Using EMR data to evaluate a physician-developed lifestyle plan for obese patients in primary care

Summary of results

The objective of this analysis was to determine whether patients who received a physician-developed lifestyle intervention lost weight compared with patients who received usual care. Overall, the intervention group did not lose more weight than the control group, with the exception of older men, in whom a substantial weight reduction was recorded. Another encouraging outcome was an improvement in systolic blood pressures. It is interesting to note that patients aged 18 to 34 years gained weight in both the control and intervention groups (an average of 5.15 kg and 4.38 kg, respectively). Young adults, in general, tend to see a greater rate of weight gain than other age groups,^15,16 and those living in rural communities are more at risk of becoming obese.¹⁷ Other factors include differences in the approach that might have been taken by the index physician (a middle-aged man) in implementing the plan for patients of different ages and sexes. Paradoxically, the inclusion of a “no smoking” recommendation might have promoted weight gain for those attempting to quit and lessened the effect of the intervention. However, the intervention physician had given smoking cessation advice on a persistent basis and it was recognized by both patients and provider that weight gain could occur. Confounding attributable to these limitations is difficult to avoid or adjust for given the variable quality and quantity of data in the EMR. A larger sample of physicians and perhaps a supplementary patient survey would have been necessary to account for these factors.

Implications for primary care practices and research

The physician who implemented the intervention believed he had observed many patients achieving substantial weight loss by following the lifestyle recommendations, but sought empirical evidence to confirm that perception. A trade-off exists between the intensity and extensiveness of an intervention and its probable effect size. Studies of high-intensity interventions are at risk of limited external validity and generalizability if the intervention is enhanced beyond a level feasible in real-world circumstances.³ Simpler interventions are feasible in such settings for patients struggling with complex medical issues, but these are less likely to measure desired outcomes. Using routinely collected EMR data for purposes of evaluation and research might mitigate these biases by providing objective perspectives on clinical activities and patient outcomes for large samples of patients of appropriate heterogeneity and homogeneity, over substantial periods of time, at reasonable cost.

The use of primary care EMR clinical data (apart from billing data) in Canada for the purposes of clinical evaluation and research is still in its early stages, despite having been a routine source of data in many countries for some time.¹⁸ To drive forward clinical improvement and innovation, primary care research needs to be pragmatic, timely, accessible to clinicians, and not cost prohibitive. This analysis conforms to these requirements.

Limitations

Our research design was not randomized and did not involve close matching of control patients to individuals in the intervention group, but it is reflective both of typical clinical contexts and of patient behaviour in community-based primary care settings. Although EMR data present obvious advantages for clinical evaluation and primary care research, they are not without challenges. Behavioural data relevant to risk factors are frequently inconsistently recorded and would benefit from standardization. Because EMR use is a relatively recent activity in Canadian primary care, and owing to the complexities of EMR implementation, data recorded in the “early years” are often inconsistent and incomplete. For instance, only 68 obese patients were identified in the intervention physician’s panel within the initial time period of the research. Currently, this physician cares for more than 300 obese patients. As one of the inclusion criteria was having at least one weight measurement in the year before the intervention, many obese patients were excluded from the study at the onset owing to missing weight data. This might have introduced bias into the results.

Several other limitations exist relating to the research design. Sample sizes in both groups were small, as only one clinic acted as the test site and only one physician’s patients within it made up the intervention group. Because this analysis was created in response to the interests of a single physician, the findings are meaningful to the index clinic, but are not directly generalizable to a wider population. In addition, there is a possibility of contamination between groups, as over the course of the intervention and the period after the intervention, physicians treating the control group might have learned about the lifestyle plan and adopted some or all recommendations for counseling their patients. However, this would facilitate a conservative estimate of the effectiveness of the intervention, which is arguably preferable. Finally, we assumed physicians in both groups provided a similar standard of care; this might not be the case and could have influenced patient outcomes. Ultimately, these limitations are the reality of pragmatic, real-world primary care research and are possibly unavoidable.

Conclusion

A simple lifestyle and dietary intervention distributed to obese patients by a family physician in community practice as part of routine clinical care was associated with significant weight loss in older men and a significant reduction in systolic blood pressure for all participants. These findings support the physician’s observation that the intervention appeared to be beneficial, although the scale of advantage was less than the physician proposed. Health care innovation in routine primary care clinical practice is feasible and should be encouraged, but rigorous evaluation is important to understand the scale of the effects observed.