OBJECTIVE

To use the framework of the Health Belief Model (HBM) to explore factors associated with metformin use among adults with prediabetes.

RESEARCH DESIGN AND METHODS

We analyzed survey data from 200 metformin users and 1,277 nonmetformin users with prediabetes identified from a large, insured workforce. All subjects were offered the National Diabetes Prevention Program (DPP) at no out-of-pocket cost. We constructed bivariate and multivariate models to investigate how perceived threat, perceived benefits, self-efficacy, and cues to action impacted metformin use and how demographic, clinical, sociopsychological, and structural variables impacted the associations.

RESULTS

Adults with prediabetes who used metformin were younger and more likely to be women and to have worse self-rated health and higher BMIs than those with prediabetes who did not use metformin. Those who used metformin were also more likely to be aware of their prediabetes and to have a personal history of gestational diabetes mellitus or a family history of diabetes. After consideration of perceived threat, perceived benefits, self-efficacy, and cues to action, the only independent predictors of metformin use were younger age, female sex, higher BMI, and cues to action, most specifically, a doctor offering metformin therapy.

CONCLUSIONS

Demographic and clinical factors and cues to action impact the likelihood of metformin use for diabetes prevention. Perceived threat, perceived benefits, and self-efficacy were not independently associated with metformin use. These results highlight the importance of patient-centered primary care and shared decision-making in diabetes prevention. Clinicians should proactively offer metformin to patients with prediabetes to facilitate effective diabetes prevention.

Adults with prediabetes and overweight or obesity can reduce their risk of diabetes by participating in intensive lifestyle intervention programs such as the National Diabetes Prevention Program (DPP) and by using metformin pharmacotherapy (1,2). Unfortunately, weight loss is difficult to maintain with lifestyle interventions and there are barriers to National DPP participation including inconvenience, lack of time, and discomfort. In addition, some people are simply not willing to participate in intensive lifestyle interventions. Although the U.S. Food and Drug Administration has not approved metformin for diabetes prevention, it has been shown to be safe and effective for diabetes prevention (3) and is the only pharmacologic agent recommended by the American Diabetes Association to delay or prevent the development of type 2 diabetes in individuals at risk (4). Metformin is inexpensive, and in one study investigators found that most people with prediabetes in the U.S. meet American Diabetes Association criteria for consideration for treatment with metformin (5).

Despite the availability of metformin for diabetes prevention, previous research has shown that it is rarely prescribed in everyday clinical practice. From 2005 to 2012, the estimated age-adjusted prevalence of metformin use for diabetes prevention among U.S. adults with prediabetes was <1% (6). Between 2010 and 2012, the prevalence of metformin use in an insured population of adults with prediabetes was slightly higher but was still only 3.7% (7). In that study, metformin use was more common among women, those with obesity, and those with two or more comorbid conditions (7). In our analysis of metformin use in an insured population with prediabetes who were encouraged to participate in the National DPP, we found a higher rate of metformin use (9.9%), particularly among women and those with obesity (8).

Many clinicians perceive barriers to diabetes prevention including patients’ inability to modify their lifestyles, lack of economic resources, and poor medication adherence. Nevertheless, family physicians who hold a positive attitude toward prediabetes as a clinical construct are more likely to recommend use of metformin to their patients with prediabetes (36% vs. 21%) (9). Recent studies have been performed to improve shared decision-making between clinicians and adults with prediabetes to increase the use of metformin for diabetes prevention with promising results (10). One common reason clinicians give for not prescribing metformin is uncertainty about the appropriate time for prescribing metformin over the course of prediabetes and type 2 diabetes (11), despite the fact that research has shown clear benefits of metformin therapy for adults with prediabetes before the diagnosis of diabetes.

Prediabetes provides a window of opportunity for diabetes prevention. Accordingly, it is important to understand factors associated with metformin use as a protective health action to mitigate diabetes risk. The Health Belief Model (HBM) provides a framework to examine how a person’s beliefs about a health problem including perceived threat, perceived benefits, and self-efficacy explain engagement in a health-promoting behavior (12,13) and how demographic, clinical, sociopsychological, and structural characteristics influence these beliefs (14). The HBM also includes assessment of how cues to action may trigger the behavior. We postulated that the HBM would be a useful framework to identify effective shared decision-making strategies to improve uptake of metformin therapy for diabetes prevention.

The aims of this study were to use the framework of the HBM to explore factors associated with the use of metformin among people with prediabetes, to identify factors that influence an individual’s decision to use metformin for diabetes prevention, and to inform interventions to increase the uptake of metformin for the primary prevention of type 2 diabetes.

Study Population

The University of Michigan (U-M) is a large public research university in Ann Arbor, MI, with satellite campuses in Flint and Dearborn, MI. U-M Premier Care is the U-M’s employer-sponsored health insurance program. It is managed by Blue Care Network (BCN). Approximately 85,000 U-M employees, dependents, and retirees are insured by U-M Premier Care. In 2015, U-M launched a 3-year pilot study to cover the National DPP at no out-of-pocket cost for U-M Premier Care members who were ≥18 years of age, had prediabetes, and had overweight or obesity. As described previously, we evaluated the effectiveness of three proactive strategies to identify U-M Premier Care members with prediabetes and encourage them to enroll in the National DPP (15).

For this analysis, we began with a population of 64,131 employees, dependents, and retirees ≥18 years of age with U-M Premier Care insurance and without evidence of diabetes. We then identified 8,131 people who had prediabetes as previously described (8,16).

Metformin Use

Although treatment of prediabetes with metformin was not specifically encouraged (15), we were able to assess use of metformin with U-M Premier Care pharmacy claims data. We initially dichotomized individuals with prediabetes as 1) having used metformin in the interval 1 year before to 1 year after they were notified that they had prediabetes and were invited to enroll in the National DPP (N = 802) and 2) as having not used metformin within the interval 1 year before or after their notification (N = 7,329). A subset of metformin users (n = 269 [34%]) was identified as having initiated metformin therapy only after they were invited to participate in the National DPP. We conducted a sensitivity analysis to determine whether there were differences in the characteristics of the 802 individuals who used metformin before or after their invitation compared with those who used metformin only after their National DPP invitation and found no differences in demographic factors (age, sex, race), clinical factors (BMI, blood pressure, cholesterol, HbA1c), or claims diagnoses (overweight/obesity, hypertension, dyslipidemia, smoking, cardiovascular disease) between the groups (Supplementary Table 1). Therefore, we defined any metformin use among the 802 individuals with prediabetes as the primary outcome of interest.

Survey Participants

To assess the domains of the HBM (Fig. 1) and to describe associations between those domains and metformin use among individuals with prediabetes, we surveyed all U-M Premier Care members with prediabetes who enrolled in the National DPP, a random sample of those who did not enroll, and all those with prediabetes who were invited to enroll in the National DPP in 2017 and 2018 and had not enrolled by July 2019 but filled at least one prescription for metformin through 2018. We mailed surveys to 492 people who filled prescriptions for metformin 1 year before or after their National DPP invitation and received 200 completed surveys (41% crude response rate). We also mailed 2,957 surveys to people who did not use metformin and received 1,277 completed surveys (43% crude response rate). When we excluded those who had died, opted out, or could not be contacted because of incorrect addresses, the adjusted response rates were 48% and 51%, respectively (Supplementary Fig. 1).

Figure 1

Theoretical model to explain metformin use among employees, dependents, and retirees with prediabetes.

Figure 1

Theoretical model to explain metformin use among employees, dependents, and retirees with prediabetes.

Close modal

Assessment of the Domains of the HBM

We postulated that perceived threat, perceived benefits, and self-efficacy would explain an individual’s decision to use metformin (14) and that perceived severity or seriousness of diabetes, perceived susceptibility to diabetes, and worry about developing diabetes would contribute to perceived threat (17). To assess perceived threat, we combined items for assessment of perceived seriousness of diabetes (using the item, “I think that diabetes is a serious health problem,” with a 4-point Likert scale with 4 indicating strong agreement) and assessment of perceived susceptibility (using the item, “What do you think your risk is for developing diabetes over the next 5 years?” with respondents selecting “almost no chance” [1], “slight chance” [2], “moderate chance” [3], or “high chance” [4]) and a two-item validated subscale from the Risk Perception Survey for Developing Diabetes (RPS-DD) (18) for measurement of worry about developing diabetes. Higher scores indicate a higher perceived threat of diabetes. To assess perceived benefits, we used a validated two-item version of the Personal Control subscale from the RPS-DD (1820), with use of 4-point Likert scales with scores from 1 to 4, higher scores indicating higher perceived benefit of personal action to modify diabetes risk. To measure self-efficacy, we used two questions from the RPS-DD (18), with use of 4-point Likert scales, higher scores indicating a greater capacity to take health-protective action. To assess events that might serve as “cues to action,” we used practice-level data to determine whether the participant received a phone call, letter, or portal message about prediabetes from their primary care provider’s office or self-reported that a doctor recommended that they enroll in the National DPP. We also asked participants whether their doctor had ever offered them metformin for diabetes prevention.

The surveys were also used to assess demographic, clinical, sociopsychological, and structural factors (14). Demographic variables included age, sex, and race/ethnicity. Clinical factors included self-rated health, self-reported height and weight (used to calculate BMI), and history of high blood pressure. Sociopsychological variables included highest level of educational attainment, annual household income, and affiliation with U-M as an employee, retiree, or dependent. Structural factors included self-reported history of gestational diabetes mellitus, family history of diabetes, and awareness of prediabetes. Knowledge of the risk factors for developing type 2 diabetes was assessed with the 11-item Diabetes Risk Knowledge section of the RPS-DD (18).

Data Analysis

Initially, we compared survey respondents and nonrespondents who used metformin and survey respondents and nonrespondents who never used metformin using health plan administrative data alone (Supplementary Table 2). For those who used metformin, we found only small differences in age, sex, number of primary care physician and specialist visits in the past year, and triglyceride levels between respondents and nonrespondents. For those who never used metformin, we found small differences in the same factors and in BMI, diastolic blood pressure, HDL cholesterol, and smoking. Therefore, we focused our analyses on participants with prediabetes who completed the surveys and compared the domains of the HBM for those who used and did not use metformin.

We compared the characteristics of individuals with prediabetes who used and did not use metformin and assessed differences between the groups using t tests for continuous variables and χ2 tests for categorical variables (8). Using multivariate logistic regression models, we tested whether perceived threat, perceived benefits, self-efficacy, and cues to action were associated with metformin use. Associations between perceived threat and demographic, clinical, sociopsychological, and structural factors were also tested with use of independent linear regression models. Finally, we constructed fully adjusted multivariate models predicting metformin use including all HBM variables and important factors and used stepwise regression with P value <0.05 for entry and >0.05 for exit to identify the most parsimonious multivariate model. Missing variables were excluded from the analyses but in general accounted for <10% of data.

The study was reviewed and approved by the University of Michigan Institutional Review Board (HUM no. 00108065) and was granted a waiver for documented informed consent. Study participants were mailed informed consent documents with the surveys, and the return of a completed survey was considered to imply consent. All analyses were performed with SAS 9.4 (SAS Institute, Cary, NC).

Metformin Users Versus Nonusers of Metformin

The study population included 64,131 employees, dependents, and retirees ≥18 years of age with U-M Premier Care insurance (Supplementary Fig. 1). Of these, 8,131 (13%) were identified as having prediabetes and 802 of them (10%) filled at least one prescription for metformin in the interval from 1 year before to 1 year after their invitation to participate in the National DPP. Of these, 200 people with prediabetes who used metformin and 1,277 people with prediabetes who did not use metformin completed surveys (Supplementary Fig. 1). Survey response rates were adjusted to remove from the denominator those who opted out, died, or had invalid addresses, resulting in adjusted survey response rates of 48% and 51%, respectively.

Metformin users were younger and more likely to be women (Table 1). Metformin users also reported higher BMIs and worse self-rated heath than nonusers. Metformin users were more likely to have lower educational attainment and lower levels of income than nonusers and were more likely to be employees rather than retirees or dependents. Metformin users were also more likely to be aware of their prediabetes, have a personal history of gestational diabetes mellitus, and have a family history of diabetes than nonusers. Although we did not specifically ask about facilitators and barriers to metformin use, we did ask survey respondents to indicate reasons for not participating in the National DPP. In general, metformin users tended to indicate a greater number of reasons for not participating in the National DPP (mean ± SD 1.8 ± 1.3 reasons) compared with metformin nonusers (1.5 ± 1.1 reasons, P = 0.0622). Metformin users were more likely than nonusers to report that they were too busy at work to attend a lifestyle program (21% vs. 15%, respectively; P = 0.0625) and to report that the lifestyle programs were held on inconvenient days (13% vs. 6%, P = 0.0025) and inconvenient times (13% vs. 7%, P = 0.0370).

Table 1

Baseline characteristics of employees, dependents, and retirees ≥18 years of age with prediabetes who completed surveys stratified by metformin use and results of multivariate logistic regression models

Prediabetes and completed surveyAny metformin useNo metformin usePFully adjusted* multivariate model OR (95% CI)Most parsimonious multivariate model OR (95% CI)
N (%) 1,477 (100) 200 (14) 1,277 (86)    
Demographic factors       
 Age (years) 53 ± 11 49 ± 12 54 ± 10 <0.0001 0.973 (0.950–0.995) 0.970 (0.950–0.991) 
 Sex    <0.0001   
  Women 937 (63) 154 (77) 783 (61)  1.456 (0.828–2.560) 1.682 (1.023–2.767) 
  Men 540 (37) 46 (23) 494 (39)  Reference Reference 
 Race    0.2523   
  Asian 114 (9) 8 (5) 106 (10)    
  Black 78 (6) 11 (7) 67 (6)    
  White 1,027 (83) 129 (85) 898 (83)    
  Other 15 (1) 3 (2) 12 (1)    
Clinical factors       
 Self-rated health 3.4 ± 0.8 3.1 ± 0.9 3.4 ± 0.8 <0.0001   
 BMI (kg/m231 ± 7 35 ± 8 30 ± 7 <0.0001 1.041 (1.009–1.074) 1.050 (1.020–1.081) 
 High blood pressure 443 (30) 67 (34) 376 (29) 0.2444   
Sociopsychological factors       
 Education    0.0117   
  ≤Some college 517 (35) 89 (45) 428 (34)  Reference  
  College graduate 340 (23) 37 (19) 303 (24)  0.739 (0.390–1.399)  
  >4-year college graduate 600 (41) 73 (37) 527 (42)  1.109 (0.630–1.950)  
 Family income    0.0006   
  <$75,000/year 518 (37) 93 (45) 425 (35)  Reference  
  ≥$75,000/year 884 (63) 101 (52) 783 (65)  0.672 (0.400–1.131)  
 U-M workforce status    0.0281   
  Employee 857 (61) 131 (69) 726 (59)  Reference  
  Retiree 156 (11) 14 (7) 142 (12)  0.862 (0.347–2.142)  
  Dependent 398 (28) 44 (23) 354 (29)  0.755 (0.423–1.346)  
Structural factors       
 Prediabetes awareness 734 (50) 135 (68) 599 (47) <0.0001 0.814 (0.463–1.429)  
 History of gestational diabetes mellitus 137 (9) 26 (13) 111 (9) 0.0509 0.978 (0.451–2.119)  
 Family history of diabetes 757 (51) 125 (63) 632 (49) 0.0006 0.867 (0.536–1.403)  
 Knowledge of type 2 diabetes risk factors 6.5 ± 2.1 6.4 ± 2.1 6.5 ± 2.1 0.6726   
Perceived threat = serious + susceptibility + worry 8.6 ± 1.4 9.3 ± 1.3 8.5 ± 1.4 <0.0001 1.176 (0.979–1.413)  
Perceived benefit: personal control over diabetes risk 3.3 ± 0.6 3.2 ± 0.6 3.3 ± 0.6 0.0084 1.032 (0.678–1.570)  
Self-efficacy for diabetes prevention 1.7 ± 0.5 1.8 ± 0.6 1.7 ± 0.5 0.0074 1.127 (0.705–1.803)  
Cues to action       
 Clinic outreach or doctor encouragement 493 (33) 82 (41) 411 (32) 0.0140 0.719 (0.428–1.210)  
 Doctor offered metformin 225 (15) 157 (79) 68 (5) <0.0001 73.82 (43.27–125.93) 65.82 (41.49–104.42) 
Prediabetes and completed surveyAny metformin useNo metformin usePFully adjusted* multivariate model OR (95% CI)Most parsimonious multivariate model OR (95% CI)
N (%) 1,477 (100) 200 (14) 1,277 (86)    
Demographic factors       
 Age (years) 53 ± 11 49 ± 12 54 ± 10 <0.0001 0.973 (0.950–0.995) 0.970 (0.950–0.991) 
 Sex    <0.0001   
  Women 937 (63) 154 (77) 783 (61)  1.456 (0.828–2.560) 1.682 (1.023–2.767) 
  Men 540 (37) 46 (23) 494 (39)  Reference Reference 
 Race    0.2523   
  Asian 114 (9) 8 (5) 106 (10)    
  Black 78 (6) 11 (7) 67 (6)    
  White 1,027 (83) 129 (85) 898 (83)    
  Other 15 (1) 3 (2) 12 (1)    
Clinical factors       
 Self-rated health 3.4 ± 0.8 3.1 ± 0.9 3.4 ± 0.8 <0.0001   
 BMI (kg/m231 ± 7 35 ± 8 30 ± 7 <0.0001 1.041 (1.009–1.074) 1.050 (1.020–1.081) 
 High blood pressure 443 (30) 67 (34) 376 (29) 0.2444   
Sociopsychological factors       
 Education    0.0117   
  ≤Some college 517 (35) 89 (45) 428 (34)  Reference  
  College graduate 340 (23) 37 (19) 303 (24)  0.739 (0.390–1.399)  
  >4-year college graduate 600 (41) 73 (37) 527 (42)  1.109 (0.630–1.950)  
 Family income    0.0006   
  <$75,000/year 518 (37) 93 (45) 425 (35)  Reference  
  ≥$75,000/year 884 (63) 101 (52) 783 (65)  0.672 (0.400–1.131)  
 U-M workforce status    0.0281   
  Employee 857 (61) 131 (69) 726 (59)  Reference  
  Retiree 156 (11) 14 (7) 142 (12)  0.862 (0.347–2.142)  
  Dependent 398 (28) 44 (23) 354 (29)  0.755 (0.423–1.346)  
Structural factors       
 Prediabetes awareness 734 (50) 135 (68) 599 (47) <0.0001 0.814 (0.463–1.429)  
 History of gestational diabetes mellitus 137 (9) 26 (13) 111 (9) 0.0509 0.978 (0.451–2.119)  
 Family history of diabetes 757 (51) 125 (63) 632 (49) 0.0006 0.867 (0.536–1.403)  
 Knowledge of type 2 diabetes risk factors 6.5 ± 2.1 6.4 ± 2.1 6.5 ± 2.1 0.6726   
Perceived threat = serious + susceptibility + worry 8.6 ± 1.4 9.3 ± 1.3 8.5 ± 1.4 <0.0001 1.176 (0.979–1.413)  
Perceived benefit: personal control over diabetes risk 3.3 ± 0.6 3.2 ± 0.6 3.3 ± 0.6 0.0084 1.032 (0.678–1.570)  
Self-efficacy for diabetes prevention 1.7 ± 0.5 1.8 ± 0.6 1.7 ± 0.5 0.0074 1.127 (0.705–1.803)  
Cues to action       
 Clinic outreach or doctor encouragement 493 (33) 82 (41) 411 (32) 0.0140 0.719 (0.428–1.210)  
 Doctor offered metformin 225 (15) 157 (79) 68 (5) <0.0001 73.82 (43.27–125.93) 65.82 (41.49–104.42) 

Data are n (%) or mean ± SD unless otherwise indicated. Perceived threat: the seriousness of diabetes, susceptibility to diabetes, and worry about diabetes.

*

Adjustment for age, sex, BMI, education, family income, U-M workforce status, prediabetes awareness, history of gestational diabetes mellitus, family history of diabetes, perceived threat (the seriousness of the threat of diabetes, worry about diabetes, susceptibility do diabetes), perceived benefits, self-efficacy, clinic outreach or doctor encouragement, and doctor offering metformin.

Adjustment for age, sex, BMI, and doctor offering metformin.

When we examined the domains of the HBM, for metformin users there were higher levels of perceived threat, lower levels of perceived benefits, and higher levels of self-efficacy and reported more cues to action than to nonusers (Table 1). With the first logistic model, we examined all of the domains of the HBM to determine which were independently associated with metformin use. The only HBM domains that entered the model were higher levels of perceived threat (odds ratio [OR] 1.202; 95% CI 1.021–1.414, P = 0.0267) and cues to action (if a doctor had offered them metformin [OR 64.280; 95% CI 40.608–101.752, P < 0.0001]). Perceived benefits (OR 0.989; 95% CI 0.0672–1.454, P = 0.9540), self-efficacy (OR 1.401; 95% CI 0.923–2.127, P = 0.1138), and cues to action including clinic outreach or doctor encouragement to enroll in the National DPP were not independent predictors of metformin use (OR 0.668; 95% CI 0.421–1.059, P = 0.0858). In the next model we examined whether perceived threat was associated with other potentially important demographic, clinical, sociopsychological, and structural factors. The variables that could potentially impact the association between perceived threat and metformin use were female sex (P < 0.0001), higher BMI (P < 0.0001), worse self-rated health (P < 0.0001), hypertension (P = 0.0006), prediabetes awareness (P < 0.0001), history of gestational diabetes mellitus (P < 0.0001), family history of diabetes (P < 0.0001), and greater knowledge of risk factors for type 2 diabetes (P < 0.0001).

With the next logistic model we examined whether perceived threat and cues to action were independent predictors of metformin use after adjustment for other demographic, clinical, sociopsychological, and structural factors. The most descriptive fully adjusted multivariate model predicting metformin use included age, sex, BMI, education, income, U-M workforce status, prediabetes awareness, history of gestational diabetes mellitus, family history of diabetes, perceived threat, perceived benefits, self-efficacy, and cues to action (Table 1). In this most descriptive fully adjusted multivariate model, younger individuals, individuals with higher BMIs, and individuals who reported that their doctor recommended metformin were more likely to use metformin. No other variables in the most descriptive fully adjusted model were statistically significant. While perceived threat was significantly associated with metformin use before the other significant factors were entered into the model, its significance diminished when the other factors were included. When a stepwise logistic regression was used to identify the most parsimonious model, only female sex, younger age, higher BMI, and the cue to action of a doctor recommending metformin therapy entered the final model predicting metformin use (Table 1).

People with prediabetes who use metformin for diabetes prevention are younger and more likely to be women and have higher BMIs and worse self-rated health than those who do not use metformin. Metformin users are also more likely to be aware of their prediabetes and have a personal history of gestational diabetes mellitus and a family history of diabetes than nonusers. When also considering the domains of the HBM including perceived threat, perceived benefits, self-efficacy, and cues to action, we found that the only independent predictors of metformin use among people with prediabetes were younger age, female sex, higher BMI, and a doctor offering metformin for diabetes prevention.

Diabetes prevention requires a patient-centered approach and shared decision-making between patients and clinicians. In a previous study of 24 patients newly diagnosed with prediabetes investigators found that no primary care physician recommended participation in a National DPP or the use of metformin. Instead, primary care physicians provided general lifestyle recommendations to lose weight and be more active (21). Another qualitative study found that although physicians acknowledged that lifestyle interventions were unrealistic for many patients with prediabetes and had concerns about access to and social support for participation in these programs, physicians preferred the National DPP to metformin as a “more effective” treatment option (22). We found that the simple cue to action of a doctor offering metformin for diabetes prevention was strongly associated with metformin use. While shared decision-making between the patient and the clinician is important, discussion of the option of using metformin generally needs to be initiated by the clinician. Even though clinicians may believe that patients do not want to take metformin for diabetes prevention (6,22), a study found that >90% of primary care patients with prediabetes reported a willingness to take metformin if they were not successful with lifestyle modification (23). And, more importantly, all individuals in that study reported a desire for a conversation regarding both lifestyle intervention and metformin as treatment options for prediabetes (23).

In another qualitative study, when asked specifically about metformin use as an option for diabetes prevention, physicians expressed skepticism about prescribing it before the onset of diabetes and indicated that they were unclear about the appropriate time to prescribe metformin on the continuum from prediabetes to diabetes (22). In another study, findings indicated that >25% of physicians did not believe that metformin reduces the risk of diabetes in patients with prediabetes (6). Taken together, the results of these studies suggest that there is a need for improved clinician awareness on the benefits of metformin for diabetes prevention.

Metformin has been shown to be more effective for diabetes prevention in subgroups of people with prediabetes including those who are younger and have higher BMIs, higher fasting glucose levels, and higher diabetes risk (3,2426). We found that these characteristics were associated with metformin use, suggesting that some clinicians are prescribing metformin to the subgroups that are most likely to benefit. However, one additional characteristic that we found to be overwhelmingly important was the cue to action of a doctor offering metformin for diabetes prevention. Clinicians should initiate this conversation with patients and use shared decision-making to decide when to initiate treatment for each individual patient. This takes time, and clinicians can only address a limited number of issues during each visit. While lifestyle interventions may be the most effective and should be routinely recommended to adults with prediabetes, use of metformin is appropriate for many people with prediabetes and should be offered as a treatment option.

We recently conducted a study in the same population with prediabetes to assess factors associated with enrollment in the National DPP (16). In that study, we found that older age, female sex, and higher BMI were associated with National DPP enrollment. In contrast, in the current study we found that use of metformin was associated with younger age. We hypothesize that younger adults with prediabetes may have more demands placed on their time due to work and family responsibilities and might have less time to devote to a lifestyle intervention. In the other study we also found that for National DPP enrollees there were higher levels of perceived benefits defined as perceived personal control over diabetes risk. Perceived benefits were not a significant predictor of metformin use, suggesting that those who elect metformin for diabetes prevention may also be less confident in their abilities to affect behavior change to prevent type 2 diabetes.

Our study has several limitations. While our survey did include questions about metformin use, its focus was on National DPP enrollment. It was not designed to evaluate preferences for metformin use or reasons for choosing metformin either alone or in combination with the National DPP. In addition, because of the study design, a relatively small number of metformin users were surveyed compared with nonusers. We relied on pharmacy claims data for assessment of prescriptions filled for metformin and we dichotomized the variable as ever using metformin without assessing adherence. Some patients may have obtained supplies of metformin at no cost from pharmacies that did not report use to the health plan. Also, some individuals who used metformin may have progressed to type 2 diabetes and used it for diabetes treatment rather than diabetes prevention. Finally, this study was conducted at only one site, and its results might not be generalizable to other settings.

Our results highlight the need for patient-centered care and shared decision-making when it comes to diabetes prevention. While we found that no intrinsic factors from the HBM model were independent determinants of metformin use, external cues to action and specific sociodemographic and clinical factors were important. In future studies investigators should better define the barriers to metformin use from both the patient and clinician perspective so that targeted interventions can be designed and implemented to overcome these barriers and facilitate diabetes prevention with metformin among appropriate high-risk individuals.

This article contains supplementary material online at https://doi.org/10.2337/figshare.20218520.

Acknowledgments. The authors thank Marsha Manning, Manager, Medical Benefits and Strategy at the University of Michigan; Ashley Weigl, Associate Director, MHealthy; and Dr. Marc D. Keshishian and Dawn Beaird, Blue Cross Blue Shield of Michigan, for their contributions to this project.

Funding. This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, grant R01 DK109995.

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

Author Contributions. L.M.N. researched data, wrote the manuscript, and reviewed and edited the manuscript. T.E.H. and K.L.J. contributed to the discussion and reviewed and edited the manuscript. W.H.H. researched data, contributed to the discussion, and reviewed and edited the manuscript. W.H.H. 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.

1.
Aziz
Z
,
Absetz
P
,
Oldroyd
J
,
Pronk
NP
,
Oldenburg
B
.
A systematic review of real-world diabetes prevention programs: learnings from the last 15 years
.
Implement Sci
2015
;
10
:
172
2.
Ely
EK
,
Gruss
SM
,
Luman
ET
, et al
.
A national effort to prevent type 2 diabetes: participant-level evaluation of CDC’s national diabetes prevention program
.
Diabetes Care
2017
;
40
:
1331
1341
3.
Knowler
WC
,
Barrett-Connor
E
,
Fowler
SE
, et al.;
Diabetes Prevention Program Research Group
.
Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin
.
N Engl J Med
2002
;
346
:
393
403
4.
Hostalek
U
,
Gwilt
M
,
Hildemann
S
.
Therapeutic use of metformin in prediabetes and diabetes prevention
.
Drugs
2015
;
75
:
1071
1094
5.
Rhee
MK
,
Herrick
K
,
Ziemer
DC
, et al
.
Many Americans have pre-diabetes and should be considered for metformin therapy
.
Diabetes Care
2010
;
33
:
49
54
6.
Tseng
E
,
Yeh
HC
,
Maruthur
NM
.
Metformin use in prediabetes among U.S. adults, 2005-2012
.
Diabetes Care
2017
;
40
:
887
893
7.
Moin
T
,
Li
J
,
Duru
OK
, et al
.
Metformin prescription for insured adults with prediabetes from 2010 to 2012: a retrospective cohort study
.
Ann Intern Med
2015
;
162
:
542
548
8.
Hurst
TE
,
McEwen
LN
,
Joiner
KL
,
Herman
WH
.
Use of metformin following a population-level intervention to encourage people with pre-diabetes to enroll in the National Diabetes Prevention Program
.
BMJ Open Diabetes Res Care
2021
;
9
:
e002468
9.
Mainous
AG
3rd
,
Tanner
RJ
,
Scuderi
CB
,
Porter
M
,
Carek
PJ
.
Prediabetes screening and treatment in diabetes prevention: the impact of physician attitudes
.
J Am Board Fam Med
2016
;
29
:
663
671
10.
Bandi
K
,
Vargas
MC
,
Lopez
A
, et al
.
Development and evaluation of a prediabetes decision aid in primary care: examining patient-reported outcomes by language preference and educational attainment
.
Sci Diabetes Self Manag Care
2021
;
47
:
216
227
11.
Trinkley
KE
,
Malone
DC
,
Nelson
JA
,
Saseen
JJ
.
Prescribing attitudes, behaviors and opinions regarding metformin for patients with diabetes: a focus group study
.
Ther Adv Chronic Dis
2016
;
7
:
220
228
12.
Becker
MH
.
The Health Belief Model and personal health behavior
.
Health Educ Monogr
1974
;
2
:
324
508
13.
Rosenstock
IM
,
Strecher
VJ
,
Becker
MH
.
The Health Belief Model and HIV risk behavior change
. In
Preventing AIDS: Theories and Methods of Behavioral Interventions
.
DiClemente
RJ
,
Peterson
JL
, Eds.
Springer US
,
1994
, p.
5
24
14.
West
R
,
Godinho
CA
,
Bohlen
LC
, et al
.
Development of a formal system for representing behaviour-change theories
.
Nat Hum Behav
2019
;
3
:
526
536
15.
Herman
WH
,
Joiner
K
,
Hurst
T
,
McEwen
LN
.
The effectiveness of a proactive, three-level strategy to identify people with prediabetes in a large workforce with employer-sponsored health insurance
.
Diabetes Care
2021
;
44
:
1532
1539
16.
Joiner
KL
,
McEwen
LN
,
Hurst
TH
,
Adams
MP
,
Herman
WH
.
Domains from the health belief model predict enrollment in and National Diabetes Prevention Program among insured adults with prediabetes
.
J Diabetes Complications
2022
;
36
:
108220
17.
Rosenstock
IM
.
Historical origins of the Health Belief Model
.
Health Educ Monogr
1974
;
2
:
328
335
18.
Walker
EA
,
Mertz
CK
,
Kalten
MR
,
Flynn
J
.
Risk perception for developing diabetes: comparative risk judgments of physicians
.
Diabetes Care
2003
;
26
:
2543
2548
19.
Rochefort
C
,
Baldwin
AS
,
Tiro
J
,
Bowen
ME
.
Evaluating the validity of the risk perception survey for developing diabetes scale in a safety-net clinic population of English and Spanish speakers
.
Diabetes Educ
2020
;
46
:
73
82
20.
Joiner
K
,
Speight
J
,
Piatt
G
.
A Spanish-language version of the Type 2 Diabetes Stigma Assessment Scale: psychometric properties in U.S. Latinos with T2D
.
Ann Behav Med
2020
;
54
(
Suppl. 1
):
S39
21.
Hafez
D
,
Nelson
DB
,
Martin
EG
,
Cohen
AJ
,
Northway
R
,
Kullgren
JT
.
Understanding type 2 diabetes mellitus screening practices among primary care physicians: a qualitative chart-stimulated recall study
.
BMC Fam Pract
2017
;
18
:
50
22.
Kandula
NR
,
Moran
MR
,
Tang
JW
,
O’Brien
MJ
.
Preventing diabetes in primary care: providers’ perspectives about diagnosing and treating prediabetes
.
Clin Diabetes
2018
;
36
:
59
66
23.
O’Brien
MJ
,
Moran
MR
,
Tang
JW
, et al
.
Patient perceptions about prediabetes and preferences for diabetes prevention
.
Diabetes Educ
2016
;
42
:
667
677
24.
Moin
T
,
Schmittdiel
JA
,
Flory
JH
, et al
.
Review of metformin use for type 2 diabetes prevention
.
Am J Prev Med
2018
;
55
:
565
574
25.
Sussman
JB
,
Kent
DM
,
Nelson
JP
,
Hayward
RA
.
Improving diabetes prevention with benefit based tailored treatment: risk based reanalysis of Diabetes Prevention Program
.
BMJ
2015
;
350
:
h454
26.
Herman
WH
,
Pan
Q
,
Edelstein
SL
, et al.;
Diabetes Prevention Program Research Group
.
Impact of lifestyle and metformin interventions on the risk of progression to diabetes and regression to normal glucose regulation in overweight or obese people with impaired glucose regulation
.
Diabetes Care
2017
;
40
:
1668
1677
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/journals/pages/license.