Prelude to PATHWEIGH: pragmatic weight management in primary care

Abstract

Objective

Treatment of obesity-related diseases, rather than obesity itself, remains the mainstay of medical care. The current study examined a novel approach that prioritizes weight management in primary care to shift this paradigm.

Methods

PATHWEIGH is a weight management approach consisting of staff team training, workflow system management, and data capture from tools built into the electronic medical record (EPIC). PATHWEIGH was compared to standard of care (SOC) using two family medicine clinics in the same US healthcare system. Descriptive statistics compared patient-, provider-, and clinic-level factors between the groups among those with at least one weight-prioritized visit (WPV) and one follow-up weight over 14 months.

Results

Groups were similar in terms of total patient visits (7,353 vs. 7,984) and patients eligible for a WPV (i.e. >18 years + body mass index >25 kg/m²; 3,746 vs. 3,008, PATHWEIGH vs. SOC, respectively). However, more PATHWEIGH clinic patients (15.9% vs. 8.4%; P < 0.001) received at least one WPV. Although no difference was observed for average patient weight loss over 14 months (P = 0.991), the number of WPVs per patient was higher in PATHWEIGH (P < 0.001) and significantly associated with weight loss (P = 0.001), with an average decrease in weight of 0.55 kg per additional visit.

Conclusions

Results from the current study demonstrate early success in changing the paradigm from treating weight-related comorbidities to treating weight in primary care.

Background

Obesity—with its many comorbid conditions—has now surpassed smoking as the leading cause of preventable death in the United States.¹ Despite the fact that obesity is both treatable and preventable, treating the comorbidities, rather than obesity itself remains the mainstay of therapy. For example, glucose-lowering rather than weight-lowering is central in treating type 2 diabetes² leaving the root cause of the obesity unaddressed. Importantly, obesity is being increasingly recognized not only as a risk factor for other diseases, but as a disease unto itself.³ Despite the fact that <1% of people with any degree of being overweight or obese are offered anything other than lifestyle advice.⁴ Reasons for lack of weight management prioritization are extensive and complex, but include lack of clinician education on effective obesity management and at lack of processes incorporated as part of standard patient care that systematically address weight loss and weight loss maintenance long-term.^4–6 Suboptimal reimbursement is also a major factor and perceived barrier. Thus, there is a critical need to systematically address these diverse barriers with pragmatic approaches and evidence that facilitates the practice of weight management.

To address this gap, a community-based primary care family medicine clinic associated with the University of Colorado developed PATHWEIGH. PATHWEIGH prioritizes weight management in primary care in two main ways. First, by using a designated, time-efficient flowsheet built into EPIC (a widely utilized electronic medical record [EMR] system in the United States) that captures weight gain history and weight loss recommendations, as well as attending to practical issues regarding diagnosis and billing. Second, by providing clinician and team with (a) training on use of PATHWEIGH as a tool and (b) education on current effective practices for weight management. We had formerly pilot-tested the feasibility of using PATHWEIGH for a limited number of weight-prioritized visits (WPVs) conducted by one endocrinologist and one family medicine physician doctor in a primary care setting.⁷ The objective of the current investigation is to examine patient-, provider-, and clinic-level factors for WPVs when conducted exclusively by family medicine physicians where PATHWEIGH was available clinic-wide vs. a nearby, similar clinic where it was not. This observational cohort study serves as the prelude to a healthcare system-wide stepped wedge cluster randomized trial design that will be used within an effectiveness-implementation hybrid study.⁸

Materials and methods

Participants

Participants in this study were adults (age >18 years) with a body mass index (BMI) >25 kg/m² seen in one of two family medicine clinics in the Denver, Colorado metro area (~15 miles apart) between 1 August 2019 and 1 October 2020 and who received at least one WPV and one follow-up encounter with weight collected. WPVs were defined as a chief complaint or reason for the visit as “weight”, “overweight”, or “obesity” or ICD-10 codes E66–E66.9 and Z76.89. This start date corresponds to the date of deployment of PATHWEIGH across the entire family medicine clinic where it was developed. The standard of care (SOC) clinic was chosen as the comparison (control) clinic due to its similar patient volume and demographics to the PATHWEIGH clinic, with no specific intervention occurring in the SOC clinic.

All data were de-identified and devoid of personal health information on extraction using a proprietary process developed by the Health Data COMPASS Warehouse at the University. The EPIC EMR (Verona, WI) was the source of the data extraction. This study was declared exempt by the Colorado Multiple Institutional Review Board.

Interventions

Standard of care

Current SOC according to the Obesity Society recommends lifestyle counselling for patients with a BMI >25 kg/m² focusing on increasing physical activity, caloric restriction, and the avoidance of trans and saturated fats.⁹ Instruction on behavioural modification to achieve these goals is an essential, highly personalized aspect of weight management. Anti-obesity medication may be considered for those with a BMI >30 kg/m² or >27 kg/m² with weight-related comorbidities. Bariatric surgery may be considered for those with a BMI >40 kg/m² or >35 kg/m² with weight-related comorbidities.⁹

PATHWEIGH

PATHWEIGH incorporates two general features that make it unique from other approaches to weight management: the workflow and the tool built into EPIC.

Workflow was not materially altered, but rather enhanced, leaving maximal time for the clinician–patient conversation. The medical assistant was made aware the patient had a WPV ahead of time, then, at the time of the visit, (1) acquired vital signs, including height and weight (EPIC then calculates BMI), (2) listed obesity or weight management as the diagnosis, chief complaint, and reason for visit, and (3) took a brief weight history in a pre-built weight management questionnaire. This process took approximately 10 min. Providers were educated on how to use patient answers in the pre-built weight management questionnaire reviewing before the onset of the visit, to direct the patient visit, and also on best practices for weight management. The patient weight management questionnaire has formerly been published in full.⁷ The use of the PATHWEIGH tool was optional and conventional note formats were always available.

The tools that collectively constitute PATHWEIGH were built to capture a broad array of weight-related data in discrete fields in the EMR (i.e. EPIC). The flowsheet prompts the medical assistant and/or clinician to ask about a history of weight gain and loss, patient-centred weight goals, and the impact of weight on their health and quality of life. The sequence of questions was designed to direct the conversation and treatment plan.⁷ Laboratory tests of interest (e.g. A1c, liver function tests, lipids, and thyroid-stimulating hormone [TSH]) performed within the year before the visit are automatically imported into the tool. The weight changes are graphed in the tool for patient and provider viewing. Medication chronology can be placed under the weight graph. PATHWEIGH was not prescriptive to allow for its use in a variety of practice patterns, which will eventually be examined as predictors of patient weight loss. Generally, advice was aligned with the current Obesity Society guidelines and SOC (see above).⁹ Visits were 20–30 min in duration and weight-related follow-up was strongly encouraged.

Assessing PATHWEIGH uptake

Clinic-, provider-, and patient-level data were compared for WPVs as an assessment of the uptake of PATHWEIGH. Clinic-level factors include patient volume and flow. Provider-level factors include demographic information. Patient-level factors include demographic information, treatment patterns, health metrics, health behaviours and goals, and change in body weight. Selected metrics were examined as predictors of patient weight loss; the ultimate measure of effectiveness.

Statistical analysis

A stepwise exclusion scheme was developed a priori to focus on patients who had at least one WPV and one follow-up visit with weight collected. Specifically, patients were excluded from the analysis sequentially for (1) missing weight, (2) age <18 years, (3) BMI <25 kg/m², (4) no recorded BMI (due to missing height), (5) no WPVs, and (6) no follow-up visit with weight collected.

A test of proportions was used to compare patient volume and flow between the PATHWEIGH and SOC clinics at key stages of the exclusion criteria process. Descriptive statistics were used to compare provider- and patient-level factors (above), as well as patient weight loss over the study period. All descriptive results are presented as mean (SD) or median (first quartile, third quartile) for continuous variables, as appropriate, and frequency (percentage) for categorical variables, unless otherwise indicated. P-values for continuous variables were calculated using a two-sample t-test when presenting the mean or using two-sample Mann–Whitney U test when presenting the median, while P-values for categorical variables were calculated using Fisher’s exact test. P-values were calculated for all variables deemed to be of particular clinical significance a priori. Multiple linear regression was used to examine weight loss over time while sequentially adding variables for clinic, follow-up time in the study, and the number of WPVs. Variance inflation factors were calculated to evaluate for multicollinearity. Following extraction from EPIC, data were managed in R version 4.0.2 (The R Foundation for Statistical Computing, Vienna, Austria).

Results

Clinic-level factors

Overall data handling is shown in the study design flow chart (Fig. 1). Specifically, the total numbers of patients seen in the PATHWEIGH and SOC clinics between 1 August 2019 and 1 October 2020 are used as the starting denominator and are roughly similar between clinics (7,353 for PATHWEIGH and 7,984 for SOC). The stepwise exclusion scheme was applied as described above. Altogether, very similar numbers of total patients and patients eligible for a WPV were seen at the two clinics. However, a higher number, 597 patients out of 3,746 eligible patients (15.9%) had at least one WPV at the PATHWEIGH clinic vs. 252 patients out of 3,008 (8.4%) at the SOC clinic (P < 0.001), of those, similar proportions (69.7% of PATHWEIGH [416/597] vs. 65.5% of SOC [165/252]; P = 0.261) had at least one follow-up visit with a measured weight.

After applying the full exclusion criteria, 416 of the 3,746 patients at the PATHWEIGH clinic (11.1%) and 165 of the 3,008 patients at the SOC clinic (5.5%; P < 0.001; Fig. 1) remained for the comparison of patient demographics, health metrics, weight, and weight loss, as well as for the description of diet and exercise in a subset of the patients at the PATHWEIGH clinic.

Provider-level factors

Information on provider demographics is shown in Table 1. All included providers were physicians. There were 10 (7 female/3 male) at the PATHWEIGH clinic and 7 providers (5 female/2 male) at the SOC clinic with no difference in provider age (41.2 (10.7) years at PATHWEIGH and 47.7 (8.0) years at SOC; P = 0.193) or percentage clinical time (70% time for both; P = 0.834). Median year that providers began practicing was 2011 for PATHWEIGH and 2002 for SOC. All providers at both clinics performed at least one WPV during this data collection period.

Patient-level factors

For the 416 patients in the PATHWEIGH group and 165 patients in the SOC group, baseline demographics were comparable between groups with respect to baseline age (PATHWEIGH 49.9 (15.9) years vs. SOC 48.9 (15.8) years; P = 0.482), but did differ by race/ethnicity with more Hispanic patients in the PATHWEIGH clinic (17.5% vs. 6.7%; P = 0.001), and insurance status (largely driven by more Tricare and Medicaid at the PATHWEIGH clinic; P = 0.038; Table 2). In contrast, the SOC clinic had a higher proportion of male patients (66.1% [109/165] vs. 40.9% [170/416] for PATHWEIGH; P < 0.001). Baseline health metrics differed for the average numbers of weight-related comorbidities (0.7 (0.8) vs. 0.4 (0.6); P < 0.001) and depressive symptoms (PHQ-8 score 1.8 (4.4) vs. 0.5 (1.6); P < 0.001), for PATHWEIGH and SOC, respectively.

Table 2 also outlines baseline weight and weight loss over the data collection period. Baseline BMI was significantly higher in PATHWEIGH (36.0 (7.7) kg/m²) vs. SOC (32.0 (7.0) kg/m²; P < 0.001). The absolute weight loss between groups (1.2 (5.8) kg vs. 1.2 (5.1) kg, PATHWEIGH vs. SOC, P = 0.991), the rate of weight loss (0.2 (2.7) kg/month vs. 0.5 (1.8) kg/month, PATHWEIGH vs. SOC; P = 0.212), and the proportion of people losing >5, 10, or 15% body weight (P > 0.485 for all) were not significantly different between clinics. The average number of days between the baseline WPV and the last recorded visit with weight collected within the study period was 152.8 (99.7) days vs. 155.4 (96.0) days for PATHWEIGH vs. SOC, respectively (P = 0.779). Although the median number of WPVs was the same between clinics (median (Q1, Q3) of 1 (1, 2) for PATHWEIGH and 1 (1, 1) for SOC), there was a significant difference in this distribution between clinics, with PATHWEIGH patients having more WPVs (P < 0.001).

The use of the patient weight management questionnaire enabled the capture of additional information about the patients who presented for a WPV in the PATHWEIGH clinic (Table 3). The majority of patients with responses (59.7%; 40/67) described their eating style as omnivorous with 20.8% (16/77) estimating their daily caloric intake as 1,000–1,500 kcal, 44.2% (34/77) as 1,500–2,000 kcal, and 24.7% (19/77) as 2,000–2,500 kcal. The most common reported weight loss strategies were eating smaller portions (73.8%, 48/65), not eating when not hungry (52.3%, 34/65), not eating out (49.2%, 32/65), and prioritizing exercise (47.7%, 31/65).

Merging factors for weight loss

Three linear regression models were fit: (1) an unadjusted model with clinic as the sole covariate, (2) a model adding follow-up time as a covariate, and (3) a model further adding the number of WPVs as a covariate (Table 4). Clinic was not found to have significant associations with weight loss across all three models (P > 0.50 for all). Follow-up time did have a significant association in Model 2 when not including the number of WPVs as a predictor with 0.01 kg greater weight loss associated with each additional day of follow-up (P = 0.024, 95% CI: 0.00, 0.01), but was not significant after adjusting for a number of WPVs in Model 3 (P = 0.209). The number of WPVs had a significant effect on weight loss with 0.55 kg of weight loss per additional WPV (P = 0.001, 95% CI: 0.29, 0.80 kg). Follow-up time and number of WPVs were examined for possible multicollinearity using variance inflation factor and were not found to be a significant issue in the final model.

Discussion

Reclaiming weight management into mainstream medicine, particularly as part of primary care, has great potential to curtail the obesity epidemic. Findings from the current study serve as preliminary evidence this is possible. For example, despite similar numbers of patients coming to the clinics who were >18 years with a BMI >25 kg/m², twice as many patients at the PATHWEIGH clinic had at least one WPV. Although no difference was observed for average patient weight loss over the brief follow-up period, the number of WPVs per patient was higher in the PATHWEIGH clinic and significantly associated with greater weight loss (P = 0.001), with an average decrease in weight of 0.55 kg per additional visit. This work also builds upon the previous pilot work that was restricted to a single endocrinologist and single family medicine physician (board-certified in obesity medicine) who had PATHWEIGH to an entire family medicine clinic with PATHWEIGH to better reflect how the tools are used in primary care when the healthcare practitioners may not be experts in weight management.⁷

Prioritizing the treatment of obesity over the treatment of weight-related comorbidities represents a major shift in the practice of medicine. A recent review of modalities proven effective to alter physician practice included: audit and feedback, computerized decision support systems, continuing medical education, financial incentives, local opinion leaders, marketing, passive dissemination of information, patient-mediated interventions, reminders, and multifaceted interventions.¹⁰ Among the best known are the tactics employed by the Centers for Medicare and Medicaid Services (CMS)¹¹ and the Healthcare Effectiveness Data and Information Set (HEDIS)¹² dictating physician reimbursement. Nevertheless, when carefully reviewed, positive change in physician behaviour is most readily evidenced with their exposure to active educational methods and multifaceted interventions, including team-based care, group visits, and resources for patients (e.g. to gyms, commercial weight loss programmes, or help with prior authorization for medications).^13–15 PATHWEIGH is one such multifaceted intervention improving the efficiency of clinical workflow and showcasing tools for weight management on the required-to-use electronic health record. Active education was also provided on how to use the tools built into EPIC and how to provide weight management to patients. Despite the fact that no financial incentives were offered to providers or patients and the COVID-19 pandemic emerged during the study period, deployment of PATHWEIGH resulted in more WPVs vs. SOC and this proportion was sustained when looking at follow-up visits where weight was recorded. Together, these findings support the notion that prioritizing weight in primary care is possible.

Both provider- and patient-level factors have formerly been linked to embracing new models in primary care.¹⁵ For example, better outcomes for patients have been reported when they are treated by younger¹⁶ or female providers.¹⁷ Likewise, indicators suggest that younger, white and commercially insured patients may be more likely to directly schedule a specific type of visit in primary care.¹⁸ A similar demographic seeks weight loss, specifically, in primary care.¹⁹ The current analysis noted no difference between the clinics for provider age and sex but yielded more surprising results with respect to patient characteristics. In contrast to reports in the literature, patients at the PATHWEIGH clinic were more likely to be of Hispanic race/ethnicity, insured by Tricare or Medicaid, and possibly sicker (as evidenced by more self-reported comorbidities and higher scores of screening questionnaires for anxiety and depression). These observations imply that more patients may seek help with weight loss in primary care once such help is routinely available.

Merging selected factors into regression models revealed an increase in weight loss of 0.55 kg per additional WPV. Given the larger distribution of patients receiving WPVs in the PATHWEIGH group, greater weight loss would then be expected. Former pilot work revealed significantly more weight loss in the PATHWEIGH vs. control group (7.2% vs. 2.1%) over 18 months⁷; hence, the current result may be affected by the shorter follow-up (i.e. 14 months—2 of which were during COVID-19 lockdown—vs. 18 months before COVID), especially since follow-up between WPVs averaged 4–5 months. Nevertheless, longer-term follow-up will be required to evaluate this hypothesis, which is currently being enrolled through a pragmatic stepped wedge cluster randomized trial.⁸

While patient response rates to the questionnaire were generally low, the patient responses available were informative and may be used in the future to provide further insight into the implementation of weight management approaches in primary care. Dietary restriction is requisite for weight loss, but dietary self-report is notoriously inaccurate with approximately 30–40% under-reporting in daily calories consumed.¹⁶ The fact that most responding patients self-reported as omnivorous eaters, however, could present an opportunity to challenge patients with macronutrient restricted eating styles for which clinical trials evidence supports their health benefits.^17,18 Low self-reported physical activity is an attractive focus for weight loss efforts, but patients should be cautioned that physical activity alone has not been found to contribute more to weight loss maintenance than to active weight loss.^19,20 Behaviour change should focus on addressing stressors, cravings, binge eating, and sedentary behaviour, particularly in mid-adulthood, via behavioural health specialists or well-established theories guiding behaviour health change, such as assessing stage of change, invoking motivational interviewing, or using cognitive behavioural change.²¹ The combination of diet changes, increased physical activity and behaviour modification may be inadequate to achieve the patient’s goal for weight loss and weight loss maintenance. The opportunity for anti-obesity medication and bariatric procedures remains grossly under-utilized.⁴

Several limitations are worth considering when interpreting these results. First, clinics were not randomized and although selected based on similar overall characteristics, underlying differences exist in the final data. This included the higher-than-expected proportion of males in the SOC clinic. Additionally, because data were limited to those >18 years with a BMI >25 kg/m² who had a WPV, only patients meeting those criteria could be compared. It is not known how the demographics of these patients compare to those of the overall patient population seen in each clinic. Second, much of the data were collected during the COVID-19 pandemic, which could have had a significant impact on the data in ways that are unclear (e.g. the census in the two clinics fell 70–80% from March to May 2020). In conjunction with the COVID-19 pandemic, the routine care in the SOC clinic resulted in very small numbers of WPVs and a correspondingly smaller sample size, affecting the power to discern group differences. However, this may also be an indication that current standard practices are not leading to the number of WPVs needed to address and support patient concerns and challenges related to obesity. Anecdotally, many people reported weight gain during some portion of the pandemic. Lastly, the number of patients with responses to the survey questions varied considerably, potentially affecting the generalizability of these results.

In conclusion, obesity remains one of the greatest health challenges of our time. Reclaiming weight management into primary care has potential to improve the obesity epidemic. Major findings from the current study demonstrate deployment of PATHWEIGH resulted in more WPVs, supporting the notion that prioritizing weight in primary care is possible. This relatively small study was not able to identify provider- or patient-level predictors of patient weight loss, but did find that more WPVs significantly relate to more weight loss. These encouraging findings lay the groundwork for our larger, funded study examining these factors across primary care in an entire healthcare system.⁸

Acknowledgments

This study was declared exempt by the Colorado Multiple Institutional Review Board.

Funding

This study was supported by the National Institute of Diabetes and Digestive and Kidney Diseases [grant No. 1R18DK127003].

Conflict of interest

LP has received personal fees for consulting and/or speaking from Novo Nordisk, Sanofi, Eli Lilly, AstraZeneca, Boehringer Ingelheim, Medscape, WebMD, and UpToDate. JW, AK, EW, and ESK have no conflicts to disclose.

Data availability

Data are available upon request to the corresponding author ([email protected]).

References