Social and Psychosocial Determinants of Health Associated With Uncontrolled Diabetes in a Federally Qualified Health Center Population

This retrospective cohort study included patients with a diagnosis of diabetes who were 18–89 years of age and had at least one medical evaluation by a primary care provider at Salud Family Health Centers during the 12-month time frame from 1 September 2016 to 31 August 2017. Patients with a diagnosis of steroid-induced or gestational diabetes were excluded. Diagnoses used in these criteria were identified using International Classification of Diseases, 10th Revision (ICD-10), codes. This study was approved by the Colorado Multiple Institutional Review Board.

All A1Cs of included patients that were measured during the study period were collected. Each patient’s highest A1C was defined as the index. If a patient had an index A1C >9% or no A1C tested during the study period, the A1C was categorized as “uncontrolled,” whereas those with A1C values ≤9% were categorized as “controlled,” using language consistent with HRSA clinical quality measures (3). Data were pulled from the time of the index event. If data were unavailable during the time of the index event, the latest datapoint for that variable during the study period was collected.

All data were extracted using Bridge-IT data-management software automated reports from the electronic medical record (EMR) (eClinical Works, Westborough, Mass.). A random sampling of records was manually reviewed to ensure accuracy of the data. Measures collected are listed in Table 1. Patients’ age, sex, ethnicity, preferred language, insurance status, and education level were collected from the patient information section of the EMR, and BMIs were calculated from data in the EMR vitals section. Medication use was identified using documented active medications reconciled as “taking” or “continued” or “refilled” during the study time frame. Appointment adherence rates were calculated by dividing the number of completed patient visits by the total number of visits scheduled (including no-show, cancelled, or incomplete visits) during the study period. Poor adherence to appointments was defined as missing ≥20% of visits, consistent with other literature. Mental health diagnoses were extracted using ICD-10 codes. Annual Patient Health Questionnaire-2 (PHQ-2) screening results were collected, with a score ≥3 indicating the presence of depressive symptoms (15). Data for all other variables were part of a standard questionnaire embedded in the EMR and completed by Salud patients on an annual basis. Additional details are shown in Table 2.

In patients with diabetes, we aimed to determine patient-specific risk factors independently associated with having an A1C >9% or no A1C tested during the study period compared patients with an A1C ≤9%.

Univariate analyses were performed between the two groups using χ² for categorical variables and Student t test for continuous variables. Adjusted analyses were performed using multivariable regression with group set as the dependent variable and the patient characteristics listed in Table 1 entered as potential independent variables. Stepwise selection was used to identify the best-fitting model, weighting variables according to their contribution to the model according to the Akaike information criterion, and the proportion of missing data. Two-sided α was defined as 0.05, and results are presented as odds ratios (ORs) and 95% CIs. All analyses were conducted using SAS software version 9.4 (SAS Institute, Cary, N.C.).

A total of 6,185 patients met the inclusion criteria. Of these, 3,220 (52%) were in the “controlled” A1C group, and 2,965 (48%) were in the “uncontrolled” A1C group (Table 1). The uncontrolled A1C group included 926 patients with no A1C test, which was ∼15% of the total population. More than 50% of patients in both groups were female. The highest proportion of patients who identified as Hispanic or Latino were in the uncontrolled A1C group, and these patients made up 71% of this group. The controlled A1C group had a larger proportion of obese patients compared to the uncontrolled group (58 vs. 48%). Regarding diabetes medications, metformin use was similar between groups, with more frequent use of other antidiabetic agents and insulin in the uncontrolled A1C group (Figure 1). With psychiatric medications, a higher proportion of patients in the controlled A1C group were prescribed an antidepressant compared to the uncontrolled A1C group (26 vs. 22%, P <0.001) (Figure 2).

Multivariable analyses were performed on the 4,774 patients with complete data (Table 3). Patients with poor adherence to appointments during the study time frame were three times more likely to be in the uncontrolled A1C group (OR 3.24 [95% CI 2.30–4.57]). Another statistically significant predictor of being in the uncontrolled A1C group was a positive PHQ-2 screening result for depression (OR 1.35 [95% CI 1.12–1.62]). For the continuous variables of BMI and age, with each unit increase in BMI or year of age, the odds of being in the uncontrolled group decreased (OR 0.98 [95% CI 0.97–0.99]). Nonmodifiable risk factors with statistically significant results were Hispanic or Latino ethnicity (OR 1.29 [95% CI 1.13–1.47]) and female sex (OR 0.76 [95% CI 0.68–0.86]). For additional comparison between patients with an A1C ≤9%, those with an A1C >9%, and those with no A1C test done, see Supplementary Table S1.

It is understood that social and psychosocial health factors may commonly affect the ability of patients with diabetes to achieve glycemic goals. However, the specific impact of these factors on a large, predominantly Hispanic, FQHC-treated population had not yet been explored (2). Defining individual characteristics associated with higher A1C values is essential in developing interventions to improve care for these patients.

This study found that patients who missed at least 20% of their scheduled clinic appointments were three times more likely to be in the uncontrolled A1C group. Reasons for missed visits were not known. These results are consistent with other studies showing that poor adherence to appointments negatively affects achievement of glycemic targets. In a study of 4,253 patients with diabetes in rural Virginia, for each 10% increase in missed appointments, the odds of having an A1C >9% increased by 24% (15). Additional focus on patient-specific barriers to appointment adherence would allow for potential improvements in patient engagement and rate of clinic follow-up.

Psychological and disease-related distress have also been shown to affect patients’ ability to achieve glycemic targets. The clinics in this study administer the PHQ-2 assessment tool annually to screen for depression (16). This is the first study to demonstrate an association between PHQ-2 screening results and A1C values. In this population, patients with a positive PHQ-2 were more likely to have an A1C >9% or untested A1C. This result indicates that even “possible depression” may be an independent risk factor for uncontrolled diabetes. Another study of 463 patients with type 2 diabetes found that diabetes distress (accounting for perceived burden of the diabetes regimen, interpersonal and provider-related distress, and emotional burden of the disease), but not depression, increased the risk of uncontrolled diabetes (17). Because this study did not find a relationship between number of mental health diagnoses and A1C results, patients with comorbid depression and other mental health symptoms of concern may be better identified by conducting more frequent screens and ensuring that additional diagnostic screens are completed when warranted. This procedure would allow for a targeted, team-based approach that integrates support from behavioral health providers into patients’ plans for meeting A1C goals.

Incorporating behavioral health providers on the care team and improving identification and treatment of depression may have further benefits to this patient population. In this study, a higher proportion of patients in the controlled A1C group were prescribed an antidepressant compared to the uncontrolled A1C group (26 vs. 22%, P <0.001). This was without controlling for a diagnosis of depression. This trend is consistent with results of a previous study, which found that patients with depression taking an antidepressant were twice as likely to achieve an A1C <7% than those experiencing depression without antidepressant medications (18). In patients with a diagnosis of depression and uncontrolled symptoms, adding an antidepressant medication could improve mood while also indirectly affecting patients’ ability to achieve glycemic goals.

This study found that a larger proportion of patients in the controlled A1C group were obese compared to the uncontrolled A1C group (58 vs. 48%, P <0.001). Obesity is often thought to complicate the achievement of glycemic targets. Although these results seem to contradict this, the average BMI of patients in this study was 32 kg/m², and the majority of patients in the study were obese. Additionally, many patients with an A1C above goal may experience weight loss with extremely elevated glucose readings, which may obscure this result.

Age was another patient characteristic that had an effect on achievement of glycemic goals. With each year of age gained, the likelihood of being in the uncontrolled A1C group decreased. This result is consistent with what has been observed in another study of more than 318,000 Hispanic patients with type 2 diabetes that found that patients >65 years of age were more likely to achieve an A1C ≤7% than younger adult subgroups (19). These results indicate that achieving glycemic targets in younger and middle-aged adult populations may be more challenging, and using practice methods that cater to the needs of younger generations may help improve diabetes care in these patients.

This study has certain limitations. Data were extracted through automated reports from the EMR, and a random sampling of charts were reviewed to assess the accuracy of these data. However, results are limited by missing or incomplete data that may result from inconsistent documentation in the EMR. In patients without an A1C completed during the study period, demographic and social characteristics were collected from the most recent encounter. Although the goal was to ensure that all data were directly correlated to the time of the index A1C event, some patients without data that directly correlated to this time point were included to allow for the greatest amount of data to be evaluated. Patients without an A1C tested during the study period were considered to have uncontrolled diabetes. Although this may have affected these results, this categorization method is consistent with the way quality metrics in FQHCs are reported.

It is important to acknowledge that many medications for mental health may have multiple indications and that a prescription for a medication may not always indicate patient adherence to that medication. Some medications, such as benzodiazepines, were not included in our study due to the potential for “as needed” use and to avoid including prescriptions for one-time doses. Thus, the use of these medications and their effects on A1C may be underestimated.

Additionally, using positive PHQ-2 screenings can underestimate the actual prevalence of depression in our patient population and may not capture the degree to which emotional distress could contribute to poorly managed diabetes (20). Although depression may contribute, other psychosocial health factors such as anxiety, substance use disorder, or post-traumatic stress disorder could also impede management of chronic diseases (21,22). Therefore, the strength of psychosocial health as a predictor in this study is likely underestimated. Screening practices in these clinics are currently being revised to improve the frequency of screening and reduce reporting bias. It may also be prudent to consider the use of screening tools for other psychosocial health conditions in future studies to better understand the overlap between not meeting glycemic targets and mental health.

Poor adherence to appointments and a positive PHQ-2 result were associated with having uncontrolled A1C values (A1C >9% or untested A1C) in our FQHC population. Targeting patients who screen positive for possible depression on the PHQ-2 may improve access to necessary services and indirectly improve their ability to achieve glycemic goals. In addition, identifying and helping patients overcome barriers to appointment adherence may aid in reducing the proportion of patients with elevated or untested A1C values.

The authors thank Arti Dhar, who assisted with electronic collection of data, and Frank Jadwin, LCSW, and Jonathan Muther, PhD, who assisted with data interpretation and drafting of the manuscript.

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

J.R. and E.R.K. researched data, wrote the manuscript, contributed to discussion, and edited the manuscript. S.J.B., J.L.P., and J.J.S. researched data, contributed to discussion, and edited the manuscript. E.R.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.