Understanding Primary Care Provider Attitudes and Behaviors Regarding Cardiovascular Disease Risk and Diabetes Prevention in the Northern Midwest Open Access

The purpose of this research was threefold: 1) assess PCP knowledge, attitudes, beliefs, shared decision-making, and experience in the prevention and management of CV risk factors in patients with prediabetes; 2) determine PCP opinions on currently available clinic resources, prevention strategies, and EMR-driven CDS related to CV risk and/or diabetes prevention; and 3) test the hypothesis that differences in PCP attitudes and opinions exist based on level of clinic rurality.

We invited all 299 physician, nurse practitioner, and physician assistant PCPs practicing in 36 Essentia Health primary care clinics in Minnesota, North Dakota, and Wisconsin to take part in a cross-sectional survey. These 36 clinics are included in a randomized control trial of an EMR-based CV CDS tool, which will be reported on separately. No compensation was provided for taking part. The survey reported here was administered pre-intervention. Exclusion criteria included survey non-response and PCPs who reported seeing patients in these clinics 0 days per week or were missing data on this question.

Essentia Health is a nonprofit integrated health care system with 75 clinics and 15 hospitals spanning four northern Midwestern states (Minnesota, Wisconsin, North Dakota, and Idaho). Essentia Health employs >900 physicians and 1,000 advanced practitioners in primary and specialty care serving a geographic service area classified as 83.8% rural, with a population base of 2 million people, of whom 56.2% live in a rural area and 45.2% live in an area with a health professional shortage (21). This practice setting presents a unique opportunity to gauge PCP attitudes toward diabetes and CV disease prevention, as well as make comparisons between rural and non-rural practitioners regarding these beliefs.

Eligible PCPs were emailed notifications about taking part in the electronic survey by Essentia Health’s primary care leadership. An initial email survey request was then made, followed by as many as seven reminder notifications, including a second email from the primary care leader with the fifth reminder notification. Survey completion implied PCP consent. The survey was administered through REDCap (Research Electronic Data Capture) (22) by the Survey Research Center at HealthPartners Institute. This study was approved by Essentia Health’s institutional review board.

The survey contained measures addressing PCP demographics, experience, prediabetes care, CV risk factor management, use of CV risk calculations, shared decision-making behaviors, and EMR decision support usability, including items adapted from two previously validated scales, the Shared Decision Making Questionnaire–physician version (SDM-Q-Doc) (23) and the System Usability Scale (SUS) (24). The SDM-Q-Doc was selected because it is designed specifically for providers and required little modification for focusing on CV risk; the SUS was chosen because of its adaptability, with only the system name needing modification. This article focuses on PCP attitudes about prediabetes, CV risk, shared decision-making, available clinic resources, prevention strategies, and current EMR decision support.

The SDM-Q-Doc gives a score of 0 to 100, representing the minimum (0) to maximum (100) levels of shared decision-making (23). We adapted the six response categories on the SDM-Q-Doc into five categories (“strongly disagree,” “disagree,” “neither agree nor disagree,” “agree,” and “strongly agree”) and addressed this change by normalizing the scale to range from 0 to 100. Questions on the current EMR adapted from the SUS included the following responses: “strongly disagree,” “somewhat disagree,” “neither agree nor disagree,” “somewhat agree,” and “strongly agree” (24). SUS scoring ranges from 0 to 100, with 100 representing the highest degree of system usability (24). The SUS does not have specific cutoffs; rather, it can be used to compare one system’s usability with another, under the assumption that higher scores equate to greater usability.

Data were analyzed using descriptive statistics and nonparametric tests of association. Two-tailed χ² cross tabulations and Wilcoxon rank sum independent group t tests for non-normally distributed continuous data were used to assess significant subgroup differences in PCP responses based on level of clinic rurality. Specifically, numeric clinic rural-urban commuting area (RUCA) codes were initially combined into four categories: metro, micro, rural, and small (25). We dichotomized these categories as “metro/micro” (referred to in this article as “non-rural”) and “rural/small” (referred to as “rural”) for subgroup analysis. Five-category Likert responses were also condensed for cross-tab analyses, such that “strongly disagree/(somewhat) disagree/neither agree nor disagree” = 0 and “strongly agree/(somewhat) agree” = 1, because there were less than five responding PCPs in some categories of clinic rurality. We used a 95% CI and corresponding P <0.05. Analyses were conducted in SPSS 23 (IBM Corp., Armonk, N.Y.) (26).

Of the 299 PCPs emailed an invitation to take part in the survey, 183 responded, giving a 61% response rate. PCPs took, on average, 12 minutes (SD 7.6) to complete the survey. Two of the 183 PCPs had since left the study clinics, and 7 were missing data on number of days practiced in the clinic, leaving a full sample of 174 PCPs for this study. Most respondent PCPs (60%) were women, and 41% practiced in rural or small-town clinics. Additional PCP demographics, including for non-respondents, are presented in Table 1.

As shown in Table 2, on a scale from 0 to 10, PCP respondents reported discussing diabetes prevention with their patients much of the time (mean 6.72, SD 2.17). Most also felt prepared to discuss metformin and glucose medications (mean 7.13, SD 2.45) and dietary and physical activity recommendations (mean 8.43, SD 1.47) for prevention of diabetes and reduction of CV risk. PCP respondents overwhelmingly endorsed the importance of screening adult patients at risk for prediabetes (mean 8.97, SD 1.38).

Non-rural PCP respondents (mean 7.00, SD 2.13) had a higher frequency of discussing diabetes prevention with patients than rural PCPs (mean 6.31, SD 2.18) (Z = –2.03, P = 0.042) (Table 2). Although no statistically significant difference was seen between the two groups on preparedness to discuss metformin or glucose medications for diabetes prevention or CV risk reduction, non-rural PCPs (mean 8.57, SD 1.58) felt slightly more prepared to discuss dietary and physical activity recommendations than rural PCPs (mean 8.23, SD 1.29) (Z = –2.32, P = 0.020). Furthermore, non-rural PCPs (mean 9.20, SD 1.22) gave more importance to screening adult patients for prediabetes than did rural PCPs (mean 8.62, SD 1.52) (Z = –2.79, P = 0.005).

PCP respondents tended to report that they discussed CV risk reduction with their high-risk patients at a typical clinic visit for non-acute illnesses (mean 7.33, SD 1.71) (Table 2). A variety of CV risk calculation methods were reported by all 174 PCP respondents, including estimating patients’ CV risk themselves (6%) or using a smartphone-driven calculator (19%), a link within the EMR (56%), a Web-driven calculator (8%), another method (1%), or none (10%) (data not compared between rural and non-rural providers). However, the ease with which providers followed the U.S. Preventive Services Task Force aspirin guideline recommendations in determining patient benefit was mixed (mean 5.94, SD 2.60) (Table 2).

With regard to shared decision-making with high-CV risk patients, PCPs agreed or strongly agreed with the SDM-Q-Doc measures, ranging from 70% for thoroughly weighing different treatment options with patients to 89% for telling the patient about different options for reducing CV risk (Table 3). Overall, PCPs had a mean SDM-Q-Doc score of 75.56 (SD 14.43).

When compared with rural PCP respondents (mean 6.85, SD 1.68), non-rural PCP respondents (mean 7.66, SD 1.65) had a significantly higher level of agreement with the statement that they discussed CV risk reduction with their high-risk patients at typical clinic visits for non-acute illnesses (Z = –3.12, P = 0.002) (Table 2). No other statistically significant difference was seen on other CV risk or shared decision-making attitudes between rural and non-rural PCPs (Table 3). Furthermore, no meaningful difference in SDM-Q-Doc scores was found between rural (mean 75.63, SD 13.15) and non-rural (mean 75.51, SD 15.31) PCPs (Z = –0.16, P = 0.869).

Most PCPs (68%) reported that clinic visit times were often too short to discuss CV risk factor care with their patients at high risk for CV disease and diabetes (Table 4). Opinions varied on whether resources were too tightly limited to improve CV risk factor care. PCP responses regarding current EMR decision support for CV risk assessment and management were also variable. For example, opinion was split on whether the current EMR decision support was too cumbersome/awkward to use to help manage a patient’s CV risk. Although most PCPs (78%) agreed that they would like to use the EMR decision support more often to help patients better manage CV risk, only 41% somewhat agreed or strongly agreed that they felt confident using it in this manner. On a scale from 0 to 100, the mean SUS score was 53.27 (SD 17.10), which suggests that the usability of the current EMR decision support related to CV risk management could be enhanced for PCPs. Of note, when the effect of level of rurality was examined for these dichotomized items, no significant difference was seen between groups on any item. Furthermore, no significant difference was seen in SUS scores between rural (mean 53.11, SD 16.19) and non-rural (mean 53.38, SD 17.80) PCPs (Z = –0.24, P = 0.807).

As shown in Table 5, most PCPs did not perceive their clinic as using the listed diabetes prevention and CV risk factor strategies, although there were exceptions. The most commonly used CV risk factor care strategies included the following: reported measurements of provider or clinic performance for CV risk factor outcomes (74%, with 45% stating it worked well and 29% stating that it did not); guideline-driven reminders for services the patient should receive that appear when seeing the patient (62%, with 34% stating it worked well and 28% stating that it did not); and a systematic approach to identify and remind patients at high risk for diabetes or CV disease who are due for health services (56%, with 38% stating it worked well and 18% stating that it did not). Regarding the components of care management routinely provided to patients at high risk for developing diabetes or CV disease, most PCPs reported the following: reviewing and individualizing the care management plan with patients (52%, with 29% stating it worked well and 23% stating that it needed improvement); helping patients set individualized treatment goals (58%, with 28% stating it worked well and 30% stating that it needed improvement); reviewing the patient’s history of targeted clinical measurements over time (e.g., blood pressure, LDL cholesterol, A1C, weight) (86%, with 58% stating it worked well and 28% stating that it needed improvement); and providing individualized patient education and support (55%, with 30% stating it worked well and 25% stating that it needed improvement).

When examining differences between the three response options and level of rurality, rural and non-rural providers only differed significantly on routinely helping patients who are at high risk for diabetes or CV disease set individualized treatment goals: χ² = 6.339, (df = 2, n = 145), P = 0.042 (Table 5). Differences appear to exist between subgroups on opinions for individuals who do provide this assistance, where 36% of rural providers reported that doing so worked well, whereas only 21% of non-rural providers said the same (not shown). Similarly, 37% of non-rural providers responded that providing this assistance with treatment planning for these high-risk patients needed improvement, compared to 20% of rural providers. No other significant difference was seen between rural and non-rural PCPs on the items in Table 5.

Few studies have examined PCPs’ attitudes about CV disease and diabetes prevention in primary care practice, although some have addressed PCP attitudes and beliefs regarding CV risk (17) and diabetes (18–20). We expand this body of knowledge by assessing PCP attitudes and shared decision-making for managing CV risk and engaging in diabetes prevention with their patients, as well as evaluating PCP perceptions of current clinic resources, prevention strategies, and EMR-driven CDS in Essentia Health, a large, predominantly rural northern Midwestern integrated health care system. Our findings show that PCPs generally reported placing a high level of importance on diabetes prevention, screening for prediabetes, and assessing patients’ CV risk in a patient-centered manner. However, clinic visit time was generally described as too short to discuss CV risk factor care, similar to findings by Doroodchi et al. (17). Opinions varied on the availability of clinic resources related to CV risk factor care, the cumbersomeness of the current EMR decision support, and prevention strategies. Furthermore, although PCPs wanted to use EMR-based decision support when managing patients’ CV risk, few reported feeling confident in doing so. These findings suggest opportunities for improving and investing more in clinic resources and EMR-driven CDS for CV risk assessment and management within a large integrated primary care system.

We observed a few statistically significant, albeit clinically marginal, differences between rural and non-rural health care system PCPs. Non-rural PCPs were slightly more likely than rural PCPs to discuss CV risk reduction and diabetes prevention with their patients, be prepared to discuss dietary and physical activity recommendations, and give more importance to screening for prediabetes. The effect of level of rurality may be weak. In addition, differences between rural and non-rural practitioners, including across this integrated health care system, may be small or nonexistent in the studied topic areas. Moreover, state-level initiatives, such as MN Community Measurement, establish common quality standards for health care systems to meet in areas such as diabetes and vascular health (27), which may account for the clinically marginal differences in PCP beliefs and attitudes based on level of clinic rurality, particularly when these quality standards are extended throughout the multistate health care system. Essentia Health’s three regions spanning four northern Midwestern states have been rebranded as three markets under “One Mission, One Essentia,” with common quality measures and goals. Future research could make comparisons between the effectiveness of various statewide initiatives on prediabetes and CV risk management in primary care.

Although PCPs in this health care system report high interest in CV disease and diabetes prevention, barriers to providing desired care clearly exist. As PCPs attempt to improve care for patients, time constraints may continue to impede patient-centered care. CDS tools and systems that use algorithms to automate typically manual EMR actions have been shown to reduce time spent by providers gathering data from patients’ EMRs (28), as well as increase clinicians’ screening, prevention, treatment, and clinical testing of CV disease (29). CDS can also support the chronic care model by maximizing EMR utility in primary care, incorporating evidence-driven guidelines into care recommendations, and facilitating informed patient and provider shared decision-making (15,16,30). Of note, most PCP respondents in this study who used CV risk calculators reported doing so through the EMR. This result suggests the viability of expanding the capability of the EMR to provide more value through identification and prioritization of treatment recommendations for the prevention of diabetes and CV disease (31,32).

Limitations to this study include the potential effects of nonresponse bias and social desirability inherent in survey methods. However, our broad inclusion criteria (PCPs practicing in 36 Essentia Health primary care clinics who saw patients at least 1 day per week), high response rate, and confidential data collection helped mitigate these issues. An additional limitation is that our constructed dichotomous variable representing level of PCP clinic rurality (rural or non-rural), created based on clinic RUCA codes, could be constructed differently by other researchers. Moreover, opportunities to improve diabetes prevention and CV risk reduction in primary care exist aside from the EMR-driven decision support tools focused on in this survey. Lastly, generalizability is limited to other PCPs in the Essentia Health system, although a further limitation is that only PCPs practicing in the 36 study clinics were included in the survey. The research team will be conducting post-intervention PCP surveys in future years, as well as reviewing patient EMR data regarding CV risk assessment and diabetes prevention outcomes in relation to the EMR-based CDS CV risk assessment tool being tested in study intervention clinics, making comparisons with control clinics. Future articles will report these findings.

A considerable number of adults in the United States have prediabetes and high CV risk. However, many are unaware of either this health risk or the options for preventing future disease and complications. PCPs believe that diabetes and CV disease prevention is important, but are working within limited time constraints. They perceive a need for better EMR-driven CDS tools and resources to improve the quality of preventive care delivered within the primary care setting.

The authors thank HealthPartners Institute and Essentia Health staff and providers for their contributions to this study. They also thank Dr. A. Lauren Crain, who aided data management, and Mary Van Beusekom, who provided copyediting.

No potential conflicts of interest relevant to this article were reported.

The study titled Improving Prediabetes Cardiovascular Care With EHR-Based Decision Support was funded by the National Institutes of Health (NIH) National Heart, Lung, and Blood Institute (NHLBI) (Trial Registration NCT02759055).

M.L.H. wrote the manuscript and analyzed the data. M.L.H., D.M.S., and J.R.D. determined the analysis plan. D.M.S. and J.R.D. designed the study and contributed to the manuscript. C.I.A., K.A.O., J.M.S.-H., P.J.O., J.Y.Z., and S.P.D. reviewed and edited the manuscript. D.M.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.