Implementation of a Prediabetes Risk Test for an Underserved Population in a Federally Qualified Health Center Free

The PRT, created by the CDC and ADA, has evolved over the years (7). The initial versions of these self- assessment questionnaires were developed for the National Health and Nutrition Examination Survey in 2004 or earlier to detect undiagnosed type 2 diabetes (8). Previously, the CDC and ADA had separate diabetes risk tests with different scoring systems. These tests had marginally different questions covering the same risk factors, including age, history of gestational diabetes, family history of diabetes, physical activity, and obesity (8). Poltavskiy et al. (8) compared these tests to evaluate their effectiveness in predicting prediabetes. Although both risk tests performed satisfactorily, the ADA’s version performed consistently better in identifying prediabetes (area under the curve 0.72–0.74 for the ADA test and 0.70–0.71 for the CDC test), likely because of the varying categories used for questions related to age and obesity. Both risk tests had high sensitivity (72–76%) in identifying prediabetes but lower specificity (54%) (8).

A study by Vanderwood et al. (9) evaluated noninvasive screening measures, including use of the ADA risk test, to identify individuals with prediabetes. This study was one of the first to evaluate the sensitivity and specificity of the ADA risk test in identifying individuals with prediabetes. The ADA risk test used at the time was the same seven-question risk test as the current version, but with a different scoring system. The results showed that the ADA risk test was highly sensitive (93.5%) in identifying individuals with prediabetes at the cutoff score (≥9 at the time) but had low specificity (17.1%) (9). These studies showed the PRT’s effectiveness in identifying prediabetes as well as diabetes (8,9). The current version of the PRT is a combination of the previous ADA and CDC tests and is used routinely to identify both diabetes and prediabetes. On the ADA website, this tool is called the Diabetes Risk Test; the CDC calls it the Prediabetes Risk Test (7,10).

Evidence supports the importance of early detection of prediabetes through screening in delaying and preventing progression to type 2 diabetes and associated microvascular and macrovascular diseases. Brannick and Dagogo-Jack (11) emphasized that prediabetes alone leads to pathophysiologic defects and increases the risk for macrovascular complications such as cardiovascular disease (CVD), stroke, and peripheral vascular disease. The defects of prediabetes include loss of β-cell volume and function, endothelial dysfunction, arterial stiffness, increased lipolysis, reduced incretin levels, impaired glucagon levels, and dysregulated cytokines (11).

Another study reported that signs of coronary artery disease and early diastolic heart failure were observed in people with prediabetes (12). This correlation stems from various factors, including 1) development of atherosclerosis from the coexistence of metabolic syndrome, 2) a correlation between increased glucose levels and cardiovascular risks, and 3) associations of prediabetes with increased high-sensitivity C-reactive protein and fibrinogen levels. In addition to macrovascular complications, microvascular complications such as retinopathy, nephropathy, and neuropathy are associated with prediabetes (12).

Multiple professional organization guidelines recommend prediabetes screening (4,13,14). The American Association of Clinical Endocrinologists (AACE), ADA, and the U.S. Preventive Services Task Force (USPSTF) all provide guidelines for prediabetes and type 2 diabetes screening in asymptomatic adults. As summarized in Table 1, these organizations outline screening criteria ranging from extensive lists of possible risk factors to simple consideration of age and BMI. The extensive screening criteria may pose a barrier for clinicians to screen patients during time-limited visits, and the inclusion of multiple factors can lead to inconsistent screening rates. The PRT created by the CDC and ADA ameliorates this problem by simplifying the process with a brief, self-administered questionnaire.

Along with differences in screening criteria, there are different diagnostic approaches among the professional organizations. AACE does not consider A1C to be a diagnostic test and recommends that an abnormal A1C should be confirmed by a fasting plasma glucose measurement or an oral glucose tolerance test (13). However, ADA considers A1C to be a valid diagnostic test but recommends that people with hemoglobinopathies receive a plasma glucose test for diagnosis (4).

The purpose of this QI project was to determine whether use of the PRT increased detection of prediabetes compared with usual care in adult patients at a federally qualified health center (FQHC). To achieve this goal, the number of prediabetes cases confirmed by A1C during project implementation was compared with the number detected and confirmed during the same time frame in the prior year. The specific project aims included 1) to identify people who were at increased risk of developing prediabetes using the PRT and 2) to evaluate the effectiveness of the PRT in appropriately identifying people with prediabetes in an underserved population.

The Iowa Model–Revised framework was used to guide this QI Project (15). This model uses a systematic approach to implement an evidence-based practice, with a focus on the application and sustainability of the practice. The Iowa Model–Revised step-by-step process includes 1) identify the issue; 2) state the question/purpose; 3) form a team; 4) assemble, appraise, and synthesize the body of evidence; 5) design and pilot the practice change; 6) integrate and sustain the practice change; and 7) disseminate the results. Three critical questions are included in the model’s flowchart, asking whether the topic is a priority, if there is sufficient evidence, and whether the change is appropriate for adoption. If the answer to any question is no, the flowchart redirects to the appropriate action to prepare for the next step in the project (15).

The QI project was conducted at a local FQHC in western Ohio. This FQHC has been serving the local community since 2008 and includes 30 medical professionals working in six clinics and one mobile health unit. According to its 2018 annual report, the center serves ∼17,700 patients annually. Before this project, there was no standard protocol and inconsistent prediabetes screening within the organization; such screening was left solely to providers’ discretion and practice style.

The project was implemented at one of the six clinics for 2 months, from September to November 2020. Participating providers included two physicians and two nurse practitioners whose scheduled patients during the project time frame were screened for prediabetes. Patient consent was not needed, as they were all established or new patients coming for scheduled visits with their primary care provider. Because this was a QI project implemented to improve standard care, it was exempt from Institutional Review Board review.

Eligible patients were determined based on inclusion and exclusion criteria. Because the project took place before the ADA lowered its age criterion for screening of asymptomatic adults, patients who were ≥40 years of age or had a BMI ≥25 kg/m² were included. Exclusion criteria included an ICD-10 code related to diabetes, prediabetes, or gestational diabetes or record of a point-of-care (POC) A1C or venous A1C test result within the last year. The project used a between-subjects, pre-/post- design with two independent groups and was implemented from 14 September to 13 November 2020.

The FQHC uses a printed report called a daily huddle report that shows details about the patients scheduled for visits each day. It includes each patient’s age, BMI, and last A1C result, as well as tasks that are due to be completed such as preventive screenings, depression screening, vaccines, and education. For patients who were eligible based on the inclusion and exclusion criteria, the PRT was listed as one of the tasks on the daily huddle report to identify at-risk patients and minimize unnecessary screening. Then the pre-implementation data from the prior year were compared with the post-implementation data to evaluate whether the PRT resulted in increased detection of prediabetes.

The intervention involved using the PRT, a self-administered questionnaire that assesses an individual’s risk for having prediabetes, on eligible patients (Supplementary Appendix S1). The questions on the PRT include age, sex, history of gestational diabetes, first-degree relative with diabetes, diagnosis of hypertension, physical activity, and weight. There are a total of seven categorical questions. Answer choices are presented as dichotomous variables in five questions and ordinal variables in two questions. Answer choices are assigned points that are summed at the end of the questionnaire to provide a total score. Individuals who score ≥5 are at increased risk for developing prediabetes (7).

Before implementation of the intervention, the project coordinator gave a presentation explaining the project and its workflow to all staff during a staff meeting. No specific training was required for the staff, as all of the skills performed for this project were within usual skill sets. In-person or video-remote interpreters were available for patients in need of such services to complete the questionnaire.

Figure 1 depicts the PRT implementation process. The intervention began with the patient service representative checking in patients for their scheduled visits and providing the PRT form per indication on the daily huddle report. The providers then reviewed patients’ completed PRT and ordered POC A1C testing for those who scored ≥5 on the questionnaire. Providers then discussed with patients the POC A1C results and the next steps if the POC A1C result was ≥5.7%.

The PRT scores, POC A1C results, and associated ICD-10 codes were documented in the electronic health record (EHR) per project protocol. After patients were screened, the diabetes registered nurse (RN) conducted audits in the EHR to identify patients newly diagnosed with prediabetes and mailed them an NDPP eligibility referral letter. The referral letter was the modified version of the Prevent Diabetes STAT initiative’s letter template for NDPP referral. Patients were advised to contact the diabetes RN if they were interested in participating in an NDPP lifestyle change program. For patients newly diagnosed with type 2 diabetes, the providers discussed results and management at their discretion. The diabetes RN was also available to provide diabetes-related education and resources.

Data were collected through EHR query. Pre-implementation data were collected retrospectively from the same providers’ patient panels and the same time frame in the prior year (14 September to 13 November 2019). Post-implementation data were from the implementation period (14 September to 13 November 2020) and were collected after the implementation period ended. All collected data were de-identified and included patients’ demographics, PRT scores, A1C results during the project period, and dates of prediabetes onset per ICD-10 code. Patient demographics in pre- and post-implementation groups were not matched. All data collected were reviewed only by the project coordinator and clinic staff on a need-to-know basis.

A between-subjects pre-/post- design with two independent groups was used to evaluate the intervention. The proportional frequencies of detected prediabetes or type 2 diabetes, as indicated by an A1C ≥5.7%, were compared between pre- and post-implementation groups using a Fisher exact test. Two-tailed significance testing was used with the level of significance set at 0.05. Odds ratio (OR) and 95% CI were calculated. Patients’ demographics were analyzed using descriptive statistics. Microsoft Excel spreadsheets were used to organize raw data, and data were de-identified before being entered into SPSS software, v. 26 (IBM), in which all statistical analyses were conducted.

The datasets generated and/or analyzed during this study are available from the corresponding author upon reasonable request.

Baseline demographics of all participants are presented in Table 2. A notable majority of the participants were White and middle-aged and had obesity. There were no significant differences in demographics for pre- and post-implementation groups (P >0.05).

In the post-implementation group, in the 2 months of the project, 285 participants were eligible for PRT screening. Of those, 51.2% (n = 146) were screened with the PRT. Of those screened, 52.1% (n = 76) scored ≥5 on the PRT. Furthermore, 75% (n = 57) of the 76 participants who screened positive had an A1C test. The remaining 19 participants who screened positive were not given an A1C test despite scoring ≥5 on the PRT. Of note, participants could have declined the A1C test. Of the 57 participants who received an A1C test, 35.1% (n = 20) had an A1C ≥5.7% and thus were newly diagnosed with prediabetes or type 2 diabetes (Figure 2).

In the pre-implementation group, 428 participants were eligible for prediabetes screening in the 2-month time frame in 2019. Of those, 4.2% (n = 18) were newly diagnosed with prediabetes or type 2 diabetes. Of those 18 participants, 8 had prediabetes and 10 had type 2 diabetes, confirmed by A1C results. In the post-implementation group, 285 participants were eligible for prediabetes screening in the corresponding time frame in 2020. Of those, 7% (n = 20) were newly diagnosed with prediabetes or type 2 diabetes. Of the 20 newly diagnosed patients, 18 had prediabetes and 2 had type 2 diabetes. A Fisher exact test showed an improvement in diagnosis of prediabetes from pre-implementation (n = 18, 4.2%) to post-implementation (n = 20, 7%), although this improvement was not statistically significant (P = 0.125). However, use of the PRT in an underserved population yielded clinically meaningful improvement (+2.8%) in prediabetes diagnosis between groups (OR 1.72, 95% CI 0.89–3.31).

A notable difference between the two groups is that the captured data from the usual care (pre-implementation) group did not specify whether A1C tests were completed for the purposes of screening. Interestingly, analysis of the data showed that A1C levels in the pre-implementation group ranged from 4.8 to 12.7%, whereas A1C in the post-implementation group ranged from 4.1 to 7.1%. Only 45% (n = 8) of the diagnosed participants in the pre-implementation group had prediabetes, and the remainder (n = 10) had type 2 diabetes. In contrast, in the post-implementation group, 90% (n = 18) of the diagnosed participants had prediabetes, and the remainder (n = 2) had type 2 diabetes. The wide range of A1C results in the pre-implementation group supports the assumption that not all A1C tests performed during this period were done for the purposes of screening for prediabetes or type 2 diabetes and that some patients may already have had diagnosed diabetes but were not assigned a correct diagnosis in the EHR.

For fidelity, the number of participants who were eligible for PRT, who completed PRT, and who received an A1C test were recorded. The NDPP eligibility letter and local diabetes education handouts were mailed out to all participants with newly diagnosed prediabetes or type 2 diabetes.

This QI project examined how the implementation of the PRT improved detection of prediabetes in an underserved population. This standardized screening approach yielded clinically meaningful effects. OR calculation showed that individuals who were screened with the PRT were 1.7 times more likely to have a diagnosis of prediabetes or diabetes compared with those who were not screened with the PRT. It should be noted that only two-thirds of participants identified as being at high risk based on their PRT score then received an A1C test. This inconsistency could have been the result of medical staff missing opportunities or of participants declining the test. Although not measured, uninsured participants were more likely to decline filling out the PRT and/or to decline the POC A1C test because they would have had to pay out of pocket for any additional testing performed in the clinic.

It is important to acknowledge that prediabetes screening was not performed in the ideal setting, as FQHCs face many challenges that are not issues in private primary care clinics. These challenges include funding limitations, expanding patient needs, and staff shortages. The patient populations at FQHCs have more complex health concerns, including those related to socioeconomic issues, behavioral health, and multiple chronic conditions (16). However, it is worth highlighting that the project screened approximately half of the eligible participants under these circumstances. More than half of the screened participants scored ≥5 on the PRT, which suggests that the PRT was used in a pertinent patient population. The sensitivity and specificity of the PRT in this project coincided with those found in other studies that evaluated the use of the ADA’s diabetes risk test to identify prediabetes (8,9).

Several factors could have influenced the outcomes of this project. The QI project was performed in one of six locations at the FQHC, which limits the generalizability of the results. Implementation of the project during the coronavirus disease 2019 pandemic limited its ability to perform at its full potential because of limited in-person office visits and additional pandemic-related protocols in place for office visits. Because the PRT was only available in paper form, it was not available to individuals having telehealth appointments. The project compared two independent groups with uncorrelated samples, and different groups of providers were involved, although the patient panels were inherited from one provider to another. This feature further reduces generalizability and may prevent extrapolating the results to a larger population.

A major barrier encountered in this project was the inability to capture the full breadth of data in the EHR. This problem may be common in both FQHCs and community clinics. The EHR software currently used in the project location has many pitfalls that limit big-data collection and analysis. There was also insufficient EHR training and information technology (IT) support available. This was evident when the project coordinator had to perform a manual data-cleaning process. Additionally, only about half of the PRT scores were documented in the EHR because there was no standard template available. As a result, the project coordinator had to sort through the raw data and add missing PRT scores into the Excel data spreadsheets. Furthermore, only ∼67% of the patients with an A1C ≥5.7% were assigned an appropriate ICD-10 code. This shortcoming prevented the project coordinator from relying on ICD-10 coding alone to identify newly diagnosed patients. Instead, the project coordinator manually examined the raw data to ensure that all patients with abnormal A1C results were included as diagnosed for the purpose of this project. The inconsistent ICD-10 coding could have resulted from delayed results from venous A1C testing ordered at provider discretion, which could have influenced the ICD-10 codes assigned during visit transcription.

Some patients were incorrectly identified as eligible for prediabetes screening on the daily huddle report because of inaccurate or missing ICD-10 codes from past encounters. Those patients were excluded from screening either by the patients themselves or by providers during their visit.

Active participation in this prediabetes screening project gave patients and providers an opportunity to discuss the risk factors for developing prediabetes. Supported by many studies and organizations such as the CDC and ADA, the PRT is a noninvasive, cost-effective method of identifying both prediabetes and type 2 diabetes. Expanded use of this screening tool will increase public awareness and enable more people to participate in diabetes prevention programs.

One crucial component to expanding application of the PRT will be to improve its availability and feasibility in a variety of settings. As noted above, major obstacles encountered in this project were the challenges of capturing data within the existing EHR system and inadequate IT support. These problems can be overcome by providing additional EHR training and IT support to facilitate real-time problem-solving. Use of a paper-based PRT and manual documentation of PRT scores led to reduced participation and incomplete data collection. A more effective application of this tool would include making it available in a digital form with automatic documentation of the data collected in the EHR system.

Because underserved populations are at increased risk for developing prediabetes and type 2 diabetes, implementing mechanisms to actively screen these high-risk populations would be beneficial by identifying prediabetes early and implementing intervention to prevent progression and thereby curb the type 2 diabetes epidemic. The targeted screening strategy described here, along with modifications to overcome the noted limitations, can enhance prediabetes screening in underserved populations. Developing clinic-specific individualized plans, adopting standardized protocols, and incorporating the PRT in EHR systems will facilitate expanded implementation of the PRT in the future.

The authors thank Julie A. Thompson, PhD, of Duke University for statistical consultation. The authors also thank Chava Sonnier, Diane Baughman, and Shelley Fouts at the Community Health Centers of Greater Dayton and Elena Turner at Duke University for their contributions.

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

M.M.K. implemented the project, gathered and analyzed the data, and wrote the manuscript. K.E.K. was the chair of the project and reviewed and edited the manuscript. B.I.P. and K.L. were committee members for the project and reviewed and edited the manuscript. M.M.K. 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.