Introduction

An estimated 30.3 million people in the United States (US) are living with diabetes [1], a complex chronic disease that is a leading cause of death [2], negatively affects quality of life [3], and is associated with high levels of healthcare utilization [4]. In 2017 in the US, an estimated $237 billion in direct medical expenditures was attributable to diabetes [4]. Persons diagnosed with diabetes have medical costs that are approximately 2.3 times that of persons without diabetes [4], and complications of diabetes represent a significant driver of healthcare costs and utilization in persons with diabetes [5].

Acute and long-term complications of diabetes can often be prevented through ongoing high-quality outpatient medical care [6]. Unfortunately, the quality of outpatient diabetes care may be suboptimal for patients who have comorbid mental health conditions or substance use disorders (collectively termed “behavioral health conditions”). For example, compared to those without substance use disorders, people with comorbid diabetes and alcohol or drug use disorders are less likely to receive recommended hemoglobin A1c (HbA1c) screening [7–9], low-density lipoprotein cholesterol testing (LDL-C) [7–9], microalbuminuria testing [9], foot sensory exams [10], retinal eye exams [7, 8, 10], or nephropathy screening [7], and they are more likely to have poor HbA1c and LDL-C control [7, 8].

The relationship between mental health conditions and the quality of outpatient diabetes treatment is less clear. Depression and anxiety are associated with deficits in diabetes care quality based on many of the measures described previously [8, 9, 11]. However, a literature review of 43 studies found that 65.1% identified no significant association between diabetes care quality and bipolar disorder, schizophrenia, or other psychoses, while 25.6% found lower care quality and 9.3% found higher care quality in persons with these conditions [12]. Differential grouping of clinically diverse behavioral health conditions may explain some of this variation across studies, as studies examining behavioral health conditions individually found that disparities in diabetes care quality varied substantially by condition [8, 11]. Further, people may experience multiple behavioral health conditions, and as the number of different behavioral health conditions increases the likelihood of receiving high-quality diabetes care decreases [8].

While the quality of diabetes care received by persons with both behavioral health conditions and diabetes has been well-studied, the association between behavioral health comorbidities and consequential diabetes-related potentially preventable hospitalizations (PPHs) in individuals with diabetes is relatively under-explored. PPHs, also called prevention quality indicators, are hospitalizations for ambulatory-care sensitive conditions (ACSCs)–conditions that should typically not require hospitalization if patients receive timely, high-quality, community-based primary care or early interventions [13]. Diabetes is one of many ACSCs [13].

Most studies examining PPHs do not focus specifically on diabetes-related PPHs in persons with diabetes. Instead, they use data from general populations [14–19] or veterans [20]. While three studies examined the relationship between comorbid behavioral health conditions and diabetes PPHs in individuals with diabetes, these studies had important limitations. They used non-standard definitions of diabetes PPHs [21], excluded high-prevalence behavioral health conditions (e.g., anxiety, adjustment disorder) [21], did not examine the impact of multiple co-occurring behavioral health conditions [9, 21, 22], or did not investigate whether behavioral health conditions were associated with increased risk of multiple diabetes PPHs over time [9, 21, 22]. Further, these studies focused on people with public insurance [21, 22] or persons residing outside the United States (US) [9], so the relationship between diabetes PPHs and behavioral health conditions has not been explored within the large, heterogeneous population of commercially insured persons in the US with diabetes [23].

The current study examines the relationship between diabetes PPHs and behavioral health conditions in a sizable sample of commercially insured working-age persons with diabetes in the US. We improved on past research a number of ways. In order to more fully characterize the relationship between behavioral health conditions and diabetes-related PPHs, we included a broad range of conditions (i.e., schizophrenia, bipolar, depression and other non-bipolar mood disorders, anxiety, adjustment disorder, alcohol use disorder, and drug use disorder). We also explored the relationship between multiple behavioral health conditions and diabetes PPHs. Additionally, we analyzed individuals’ data over a three-year period to examine recurring diabetes PPHs over time.

Specifically, we hypothesized that each behavioral health condition would be independently associated with an increased likelihood of experiencing at least one diabetes-related PPH as well as multiple diabetes PPHs during a three-year period. Further, we hypothesized that, relative to those with no or only one behavioral health condition, persons with multiple behavioral health conditions would have a greater likelihood of experiencing at least one diabetes-related PPH as well as multiple diabetes PPHs.

Materials and methods

The North Texas Regional Institutional Review Board at the University of North Texas Health Science Center reviewed and approved this project as exempt category research.

Results

Of the 4 million people in the available Optum Clinformatics data, 229,179 were adults ages 20 to 64 with diabetes. We excluded 140 of those individuals due to missing socio-demographic information. Thus, 229,039 individuals met inclusion criteria for the study. A total of 4,521 diabetes PPHs were experienced by 3,246 of these persons (3,246/229,039; 1.42%; 95% Confidence Interval [CI] 1.37–1.47%). In addition, 623 individuals (623/229,039; 0.27% CI 0.25%-0.29%) experienced more than one diabetes PPH. Further, 47,701 (20.83%; CI 20.66%-20.99%) of persons in the study had at least one behavioral health condition; 34.30% (16,361/47,701; CI 33.88%-34.73%) of those persons had more than one behavioral health condition.

Relationships between diabetes PPHs and behavioral health conditions

The 20.83% of persons with one or more behavioral health conditions experienced 43.62% (1,972/4,521; CI 42.18%-45.07%) of all diabetes PPHs, and the 7.14% of persons with more than one behavioral health condition experienced 24.77% (1,120/4,521; CI 23.54%-26.05%) of all diabetes PPHs.

Mental health conditions.

In unadjusted analyses, all mental health conditions (i.e., schizophrenia, bipolar, depression/other mood disorder, anxiety, and adjustment disorders) were associated with an increased likelihood of having at least one diabetes PPH and increased counts of diabetes PPHs (p<0.001 for all; see Table 1). However, in the adjusted analysis, only two mental health conditions, schizophrenia and depression/other mood disorder, were independently associated with an increased likelihood of having at least one diabetes PPH (p<0.001 for both; see Table 2). Similarly, depression/other mood disorder was the only mental health condition independently associated with increasing counts of PPHs in those with at least one PPH (p<0.001) although schizophrenia approached significance (p = 0.095; see Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of persons with diabetes by diabetes-related potentially preventable hospitalization (PPH) count, based on medical and pharmacy claims data with 2011–2013 service dates (n = 229,039).

PPH counts are truncated at three in Table 1 for presentation purposes, but counts were not truncated in statistical analyses.

https://doi.org/10.1371/journal.pone.0212955.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Adjusted odds ratios (aOR) and adjusted risk ratios (aRR) of potentially preventable hospitalizations for diabetes (PPH) in individuals with and without behavioral health conditions (n = 229,039).

Logit results examine findings regarding the presence or absence of any PPH, whereas the negative binomial results examine the findings regarding the increasing counts of PPHs in individuals with one or more PPH. Analyses were adjusted for sociodemographic and physical health covariates. For full Model 1 results see S4 File.

https://doi.org/10.1371/journal.pone.0212955.t002

Substance use disorders.

Substance use disorders were also associated with diabetes PPHs. In unadjusted and adjusted analyses, both alcohol use disorders and drug use disorders were associated with an increased likelihood of having at least one diabetes PPH (p<0.001 for all), and in unadjusted analyses both were associated with increasing counts of diabetes PPHs (p<0.001 for both; see Tables 1 and 2). However, in adjusted analyses, only drug use disorders were associated with an increased likelihood of increasing numbers of diabetes PPHs in persons with at least one PPH (p<0.001); this relationship was nonsignificant for alcohol use disorders (p = 0.964). Results of bivariate and multivariable analyses of individual behavioral health variables are in Tables 1 and 2, respectively, and detailed Model 1 results are in S4 File.

Behavioral health complexity.

Increasing behavioral health complexity as measured by a count of diagnosed behavioral health conditions was associated with an increased likelihood of having one or more diabetes PPH in unadjusted analysis (p<0.001; see Table 1) and adjusted analysis (p<0.001 for all categories; see Table 3). Further, increasing behavioral health complexity was associated with increasing counts of diabetes PPHs in unadjusted (p<0.001; see Table 1) and adjusted analyses (p<0.001 to p = 0.015). See Table 3 for findings related to behavioral health complexity from Model 2, the multivariable model examining this variable, and see S4 File for detailed results of this model.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Adjusted odds ratios (aOR) and adjusted risk ratios (aRR) of potentially preventable hospitalizations for diabetes (PPH) by behavioral health (BH) complexity level (n = 229,039).

Logit results examine findings regarding the presence or absence of any PPH, whereas the negative binomial results examine the findings regarding the increasing counts of PPHs in individuals with one or more PPH. Behavioral health complexity represents a simple count of each person’s behavioral health conditions, including both mental health conditions and substance use disorders. Analyses were adjusted for sociodemographic and physical health covariates. For full Model 2 results see S4 File.

https://doi.org/10.1371/journal.pone.0212955.t003

Our sensitivity analysis indicated that the findings described above were robust to variations in the operational definition of our chronic pain covariate (see S5 File).

Discussion & conclusions

PPHs for diabetes were disproportionately concentrated in individuals with co-occurring behavioral health conditions. While 20.8% of persons with diabetes had a comorbid behavioral health condition, this group experienced 43.6% of all diabetes PPHs. These findings suggest that the outpatient care of individuals who have both diabetes and behavioral health conditions may be suboptimal, and this suboptimal care results in measurably poorer health outcomes.

However, contrary to our hypotheses, not all persons with comorbid diabetes and behavioral health conditions are at increased risk of outpatient care quality deficits as measured by diabetes PPHs; the relationship varied by condition. Depression/other mood disorder was associated with an increased likelihood of having at least one diabetes PPH and recurring PPHs in those with at least one PPH. These results align with our hypotheses and reiterate previous findings that depression increases risk of diabetes related hospitalizations and other ACSCs [9, 16]. Conversely, our adjusted analyses did not identify significant independent associations between diabetes PPHs and anxiety or adjustment disorders. These results are in contrast to those of Mai et al. [9] who found that adjustment disorders and anxiety disorders (termed “neurotic disorders” in their paper) were associated with an increased likelihood of experiencing a hospitalization for diabetes complications. The disparate findings may be due to differing methodologies in categorizing patients with multiple co-occurring behavioral health conditions. We included all behavioral health conditions in the analysis separately, whereas Mai et al. [9] assigned patients to a single behavioral health condition based on a hierarchy [30]. Patients with both anxiety and depression were deemed to have anxiety, and patients with both adjustment disorder and depression were deemed to have adjustment disorder. Thus, the aforementioned association between depression and diabetes PPHs may explain their positive results.

Additionally, we found that schizophrenia was associated with an increased likelihood of having at least one diabetes PPH, although schizophrenia was not associated with recurring PPHs over time. However, bipolar was only associated with an increased likelihood of any diabetes PPH and increased counts of diabetes PPHs in unadjusted analyses. These findings are consistent with that of Druss and colleagues [22]. They examined US Medicaid data and identified significant positive associations between schizophrenia and hospitalizations for ACSCs of all types but did not find significant relationships between ACSC hospitalizations and bipolar disorder. In contrast, our findings are inconsistent with that of Davydow and colleagues [17] who identified independent positive associations between diabetes PPHs and both schizophrenia and bipolar in a Danish sample of patients, and they are inconsistent with Mai et al. who identified no association between schizophrenia and experiencing at least one hospitalization for diabetes in an Australian sample of patients [9]. Our findings are also different from those of Leung and colleagues [21] who identified a decreased risk of diabetes PPHs in individuals with schizophrenia who were enrolled in Medicare or Medicaid in one US state. On the other hand, they too found no association between diabetes PPHs and bipolar [21]. The inconsistency in research findings regarding the relationship between diabetes PPHs and these serious mental illnesses mirror the inconsistency in the results of past research on the relationship between serious mental illness and the quality of outpatient diabetes care received [12], as would be expected given that diabetes PPHs are indicators of outpatient care quality deficits [13]. A number of factors influence the quality of care that a patient receives, including patient, delivery system, and insurance characteristics [12]. It is likely that these factors underlie the variations in findings described above; our study focused on US patients enrolled in commercial insurance plans, while the other studies examined patients from other countries [9, 17] or US patients with public insurance [21, 22].

However, findings regarding substance use disorders are uniform across studies. Consistent with past research [9, 21, 22], we found that persons with diabetes who have comorbid alcohol or drug use disorders had an increased likelihood of having ≥1 diabetes PPH. Similar findings have been observed in other populations [18]. We also expanded on past research by identifying that, while recurring diabetes PPHs were not observed in individuals with alcohol use disorders and at least one diabetes PPH, people with both diabetes and drug use disorders who experience at least one diabetes PPH have an increased risk of experiencing recurring diabetes PPHs over time. The observed relationships between substance use disorders and diabetes PPHs are likely consequences of the diabetes care quality deficits experienced by persons with co-occurring substance use disorders [7, 8, 10]. Given that persons with comorbid drug use disorders are significantly more at risk for multiple PPHs than subjects with other comorbid behavioral health conditions, future research is needed to understand how best to prevent repeat PPHs in this population. Preventing PPHs in this population presents unique challenges because repeated relapses are common, and–by definition–individuals with such disorders may neglect other areas of their life while spending a significant amount of time obtaining, using, or recovering from their substance(s) of choice [31].

Our results illustrate the importance of looking beyond individual behavioral health conditions, as over one-third of persons with comorbid diabetes and behavioral health conditions were diagnosed with multiple behavioral health conditions. Persons in this group were at even greater risk of experiencing diabetes PPHs. In total, approximately one quarter of diabetes PPHs were experienced by the 7.1% of persons with diabetes who were diagnosed with ≥1 behavioral health condition, and the risk of PPHs increased as the number of behavioral health conditions increased. These are important findings, as we identified no previous studies examining the association between increasing PPH risk and increasing numbers of behavioral health conditions. However, our results are consistent with past research that found that the likelihood of receiving recommended screenings (e.g., HbA1c, LDL-C) decreased and the likelihood of having poor lipemic or glycemic control increased as the number of behavioral health conditions increased [8]. Our results build on those findings, suggesting that these care quality deficits are likely resulting in measurably poorer outcomes for people with multiple or complex behavioral health conditions.

Enhanced outpatient care strategies that improve care quality may reduce the risk of PPHs in persons with diabetes and multiple behavioral health conditions as well as those with diabetes and either co-occurring schizophrenia, depression, or substance use disorder. An evidence-based model of care that may mitigate diabetes PPH risk in these populations is to better integrate behavioral and physical healthcare services. Mental health patients receiving integrated care that included both mental health services and physical health care generally receive a higher quality of care, including higher-quality services for prevention of complications related to diabetes [32]. There is also a need to integrate substance use screening into diabetes treatment programs for better diabetes management and substance use prevention and treatment [33]. Accordingly, current practice guidelines encourage shared responsibility between physical and behavioral health providers and recommend ongoing integrated physical and behavioral health care [34]. However, behavioral health specialists may not feel responsible for providing standard diabetes care services to their patients, and some do not have the training to manage diabetes in their patients [35]. Additionally, providers may experience reimbursement-related challenges when implementing integrated healthcare programs [36]. Hence, a multilevel cultural shift backed by payers, providers, and policymakers will be required to create a functioning and sustainable integrative care system.

The outpatient care quality deficits that likely underlie the PPHs observed in this study may also be related in part to provider attitudes and beliefs. Stigma related to mental health and fear of being labeled impedes people from seeking services for their mental health needs [37]. For persons who do seek treatment, healthcare providers may have negative attitudes towards and consequently provide a poor quality of care to patients with mental health and/or substance use disorders [38, 39]. Stigma in healthcare settings often results from lack of awareness or skills and workplace culture [38]. By addressing these issues, outpatient care quality can be improved for the vulnerable population of persons with comorbid diabetes and behavioral health conditions.

Additionally, the majority of care that patients with chronic conditions receive is self-care [40], so the importance of self-care in reducing PPH risk in the population of persons with co-occurring diabetes and behavioral health conditions cannot be understated. Behaviors such as healthy eating, blood sugar monitoring, medication adherence, and physical activity are all associated with positive diabetes outcomes [41]. Outpatient providers play a key role in encouraging self-care [41]. Given that substance abuse disorders and depression may be associated with suboptimal self-care [42, 43], outpatient providers should consider providing self-management support to persons in these groups in order to facilitate self-care [40, 44].

Furthermore, community-based diabetes self-management programs have shown promise in improving diabetes-related knowledge and health outcomes [45, 46]. For example, in a pilot test and subsequent randomized control trial of a community-based program delivered by a team of nurses and peer educators, researchers observed significant improvement in diabetes knowledge among people with diabetes and mental health disorders after 60 weeks [47, 48]. Another pilot test of a telehealth intervention integrated with nurse case management services displayed improvement in diabetes related outcomes among people with mental health conditions at 6 months [49]. However, few randomized controlled trials have examined the effectiveness of community-based diabetes self-care interventions for persons with serious mental illness [50] so more research is needed.

While our study provides important insights into the relationships between diabetes PPHs and behavioral health conditions, there are limitations. The cross sectional observational design, while consistent with other studies examining PPHs [14], disallows us from making causal statements about the relationship between diabetes PPHs and behavioral health conditions. Further, we were unable to include individuals who were 18 or 19 years of age. While the logic that we used to identify diabetes PPHs is applicable to persons 18 and older [13], our data source grouped individuals into 5-year age groups (e.g., 15 to 19 years). Additionally, while claims data provide important insights into patients’ healthcare conditions and typically reflect accurate diagnoses, only diagnoses associated with services and included on claims can be analyzed. Patients with undiagnosed conditions or with conditions not reflected in claims would not be identified as having these conditions.

Similarly, claims data only contains information relevant to payers, so we were unable to control for many socioenvironmental variables (e.g., household composition, social support, stress) that would be expected to vary greatly within the heterogeneous commercially insured population. We were also unable to control for modifiable risk factors and individual lifestyle choices that have previously been found to be associated with poor physical health in persons with behavioral health conditions (e.g., lack of exercise, poor nutrition, motivation [51]). In addition, individuals in our sample were continuously enrolled in commercial insurance plans for a three-year period, so persons without stable coverage and persons who migrate into Medicaid or other coverage are not represented by this analysis.

We provide important insights regarding the strong association between drug use disorders and experiencing multiple diabetes PPHs, but different drugs (e.g., cannabis, opioids, sedatives) may have differing associations with PPH. We did not examine non-alcohol substances separately, so additional research is warranted. Finally, behavioral health complexity measured by unique diagnosed condition counts does not reflect the range of morbidities that may be observed within persons with a single condition such as schizophrenia or substance use disorder.

While our study provides new information about the relationship between behavioral health conditions and diabetes PPHs, some questions remain unanswered. We did not investigate whether improvements in the quality of outpatient care provided to persons with behavioral health conditions mitigates their risk of diabetes PPHs. When future research is conducted to explore this topic, a number of standard quality measures could be leveraged, including Health Effectiveness Data and Information Set (HEDIS) measures (e.g., Diabetes Monitoring for People with Diabetes and Schizophrenia which examines HbA1c testing, Diabetes Care for People with Serious Mental Illness which examines HbA1c control)[52]. Further, the relationship between atypical antipsychotics, commonly used for treatment of psychotic disorders, bipolar disorder, and some forms of depression, is well known. Therefore, future studies needed to explore whether psychiatric medications or adherence to those medications increases or decreases the risk of diabetes PPHs in the populations of persons with behavioral health conditions.

We provided important insights regarding the strong association between drug use disorders and experiencing multiple diabetes PPHs, but different drugs (e.g., cannabis, opioids, sedatives) may have differing associations with PPH. We did not examine non-alcohol substances separately, so additional research is warranted. And, while our study provides new information regarding diabetes PPHs in persons with diabetes, future studies are needed to examine whether individual behavioral health conditions are each independently associated with an increased likelihood of experiencing at least one PPH or increasing numbers of PPHs in persons with other chronic ambulatory care sensitive conditions (e.g., chronic obstructive pulmonary disease, congestive heart failure, asthma).

Despite the limitations, this work provides important, actionable insights into the relationship between diabetes PPHs and behavioral health conditions. In total, 43.6% of all diabetes PPHs were experienced by the 20.8% of persons with diabetes and a comorbid behavioral health condition. Specifically, patients with depression, schizophrenia, substance use disorders, or multiple behavioral health conditions are at increased risk of experiencing at least one diabetes PPH. Patients with depression, drug use disorder, or multiple behavioral health conditions are especially at-risk, as they are susceptible to recurring diabetes PPHs over time. Given that diabetes PPHs are often preventable with timely, high-quality outpatient care, providers, policymakers and payers should implement targeted interventions and enhanced care models to improve the quality of clinical and self-care received by these vulnerable patients.

Supporting information

Acknowledgments

The authors gratefully acknowledge Thomas Lane, NCPS, CRPS, who provided assistance in identifying literature relevant to the study content, and Thaddeus Miller, DrPH, MPH, who provided important intellectual contributions during the review and interpretation of preliminary data analysis results and prior presentations of those results.

The findings and conclusions described herein are those of the authors and do not necessarily represent the positions of Magellan Health, Inc., or other institutions with which the authors are affiliated. The authors have no other financial disclosures to report.