Quality and Predictors of Diabetes Care Among Patients With Schizophrenia: A Danish Nationwide Study

Schizophrenia represents a major global health challenge and is associated with excess mortality (1,2). About one-third of the mortality excess is attributed to unnatural causes, including suicide and accidental death, and two-thirds is attributed to natural causes, including chronic medical comorbidities (2–4). Diabetes appears to be a particularly important contributor to the 10 to 20 years of shortened life expectancy among patients with schizophrenia (3–8). Therefore, special health care attention is warranted, focusing on early diagnosis and effective treatment of diabetes. Increasing evidence, however, indicates deficient diabetes care among patients with schizophrenia compared with the general population (9–12). Although studies have suggested disparities, the magnitude varies, and methodological limitations, including non–population-based study designs and lack of detailed data on the provided care, have been noted. Variation between studies not only emerges from methodological limitations but may also emerge from differences in the organization and financing of health care systems.

In Denmark, health equity is a stated priority in the public health care system, which is funded mainly though taxes. The universal health coverage provides all Danish residents regardless of their socioeconomic status equal access to care at public hospitals (13,14). The quality and equity of care provided by the Danish health care system have been routinely monitored for nearly two decades by a nationwide multidisciplinary initiative (15,16). This setting offers a unique opportunity to examine whether equal access also implies equivalent and sufficient diabetes care for individuals with serious comorbidities, such as schizophrenia, and if not, whether specific factors predict potential quality gaps in such vulnerable populations. Thus we conducted a nationwide population-based cohort study to examine the quality of diabetes care as reflected by its meeting specific process performance measures of care among Danish individuals with and without schizophrenia and to identify potential patient-, provider-, and system-specific predictors of quality of diabetes care among individuals with schizophrenia.

Methods

In Denmark, patients with schizophrenia are exclusively treated at public psychiatric hospitals, whereas individuals with diabetes are managed by general practitioners (GPs) or in specialized outpatient hospital clinics. Each patient contact with the health care system is recorded in public registries with a unique civil registration number assigned to all Danish residents. This number provides unambiguous linkage of information at the individual level among various data sources (13,14). We used data from the following population-based registries: the Danish Clinical Registries and the Danish Psychiatric Central Research Register (PCRR). Reporting to the registries is mandatory for all Danish hospitals.

The Danish Clinical Registries

In 2000, a nationwide multidisciplinary initiative was launched with the objective of systematically monitoring and continuously auditing the quality of care provided by the Danish health care system. In the following years, population-based clinical registries were established to document care for important diseases, including schizophrenia (the Danish Schizophrenia Registry [DSR]) and diabetes (the Danish Adult Diabetes Registry [DADR]). These registries hold data on disease-specific performance measures of care and several prognostic factors (15–18). An expert board representing national scientific societies and professional organizations followed a structured process to develop these performance measures on the basis of recommendations from national clinical guidelines with the highest strength of evidence. The DSR documents the following areas: diagnostic evaluation; antipsychotic treatment, including side effects; screening for cardiovascular risk factors, including, for example, glycated hemoglobin (HbA1c); family intervention; psychoeducation; suicide risk assessment; and postdischarge care. Referral of patients with schizophrenia who have increased levels of HbA1c to GPs or to specialized clinics varies. The DADR covers domains of cardiometabolic control and screening for complications. All data are prospectively collected on the basis of documentation in medical records by use of standardized registration forms with detailed instructions. The registration form is completed for all inpatients with schizophrenia at discharge from psychiatric hospitals and once a year for patients with schizophrenia followed at hospital outpatient clinics. For individuals with diabetes, quality-of-care data are recorded annually at all hospital outpatient clinics; similar diabetes care data are collected from GPs; however, completeness of data collection remains low.

The registries include records for 93% of all hospital inpatients, 92% of all hospital outpatients with schizophrenia, and approximately 90% of all hospital outpatients with diabetes in the Danish health care system (17,18). Almost all patients with schizophrenia who have been in contact with the secondary health care system can thus be identified from the DSR since its initiation in 2004, and most individuals with diabetes treated at hospital outpatient clinics can be identified from the DADR since 2005.

The PCRR

The PCRR has maintained a systematic collection of all psychiatric diagnoses and contacts, including dates of each admission to and discharge from Danish psychiatric hospitals, since 1969 and the start and end dates of all hospital outpatient and emergency room contacts since 1995 (19–21).

Study Population

Individuals with diabetes.

The study population included all individuals (≥18 years) treated for diabetes at hospital outpatient clinics and recorded in the DADR between January 1, 2005, and December 31, 2013. The DADR includes individuals with the following ICD-10 codes for diabetes: E10.0–E11.9 and E13.0–E14.9 (22).

Individuals with diabetes and schizophrenia.

Using the DSR, outpatients with diabetes and comorbid schizophrenia were identified (ICD-10 codes F20.0–F20.99). We included inpatients and outpatients with schizophrenia recorded in the DSR between January 1, 2004, and December 31, 2011. Patients were excluded if the diagnosis of schizophrenia was not recorded before their first diabetes contact (N=29) or if more than five years elapsed between the recording of the schizophrenia diagnosis and the subsequent diabetes contact (N=23), which left 83,813 individuals with diabetes—for a total of 302,638 diabetes contacts to be included in the analysis. A total of 106 hospital outpatient clinics were represented. Within this sample, 669 (.8%) individuals with 1,681 diabetes contacts had schizophrenia recorded before their diabetes contact.

Predictors of Diabetes Care

Patient-, provider-, and system-specific factors were examined as potential predictors of the quality of diabetes care among individuals with diabetes and schizophrenia. Drawn from the DADR, DSR, and PCRR, the patient-specific predictors included age, sex, smoking, alcohol or drug abuse (illegal drugs, benzodiazepines, or central stimulants), duration of schizophrenia at the time of the hospital outpatient contact for diabetes, and score on the Global Assessment of Functioning (GAF), ranging from 1 to 100, to evaluate the overall psychosocial functioning of individuals with schizophrenia (23,24).

The provider-specific predictor included the quality of schizophrenia care preceding the diabetes contact, because inadequate quality of schizophrenia care may be associated with a subsequent deficient quality of diabetes care. Table 1 lists the 12 assessed process performance measures of care, collected from the DSR, for hospital inpatients and outpatients with schizophrenia, including incident patients, defined as individuals who were diagnosed as having schizophrenia within the past year (17). The quality of schizophrenia care was assessed overall by dividing the number of performance measures received by the patient with the number of relevant measures for each individual.

Collected from the DADR, system-specific predictors included contact volume, which was defined as the average number of contacts in each hospital outpatient diabetes clinic per year from 2005 to 2013. The contact volume was divided into four quartiles; low (quartile 1, ≤519 diabetes contacts per year), medium (quartile 2, >519–895), high (quartile 3, >895–1,620), and very high (quartile 4, >1,620).

Quality of Diabetes Care

With data from the DADR, the quality of diabetes care was assessed with 11 process performance measures of care relevant to hospital outpatients with diabetes (18) (Table 2). Furthermore, we summarized the quality of diabetes care overall by dividing the number of performance measures of care received by the patient with the number of relevant measures for each individual.

Statistical Analyses

Binary regression was used to examine the association between schizophrenia and the quality of diabetes care by estimating the relative risk (RR) of receiving the individual process performance measures of diabetes care and the overall quality of care among individuals with and without schizophrenia. A high quality of diabetes care was defined by using a pragmatic cut point as having received ≥80% of the relevant measures. To assess the robustness of our findings, we also repeated the analysis by using alternative thresholds ranging from 60% to 90%. The analyses were stratified according to sex, age, and HbA1c level.

Binary regression was used to elucidate the effect of patient-, provider-, and system-specific predictors on quality of diabetes care among individuals with diabetes and schizophrenia. The potential predictors included sex, age, smoking, substance abuse, GAF score, duration of schizophrenia, quality of schizophrenia care, and clinic contact volume. A high quality of schizophrenia care was defined as having received ≥80% of the relevant measures. This analysis was repeated with different cut points of 70% and 90% to evaluate the strength of the results. The adjusted analysis adjusted for predictors that were significantly associated with the outcome of interest in the unadjusted analysis. Clustering of individuals within hospital outpatient diabetes clinics was taken into account in all analyses by correcting the 95% confidence intervals (CIs) for robust estimates of the variance. A two-sided p value ≤.05 was considered to be significant. Stata, version 11.2 special edition, was used for all analyses.

Results

Compared with individuals with diabetes only, those with diabetes and schizophrenia were younger and a larger proportion had type 2 diabetes, and they were more likely to be smokers and to have a higher body mass index and a higher HbA1c level (Table 3). In both groups of outpatients, most received 80% to 100% of the relevant recommended process performance measures.

Individuals with diabetes and schizophrenia were less likely than those with diabetes only to receive high-quality diabetes care (receipt of ≥80% of all relevant process performance measures) (RR=.91) (Table 4). This pattern was also evident when the analysis used alternative thresholds between 60% and 90% for defining high-quality care, with RRs ranging between .94 (CI=.91–.97) and .82 (CI=.76–.88) (data not shown). Moreover, those with diabetes and schizophrenia were less likely than those with diabetes only to receive several individual process performance measures, including blood pressure monitoring (RR=.98), treatment with antihypertensive drugs (RR=.83) and ACE (angiotensin-converting enzyme) /ATII (angiotensin II receptor) inhibitors (RR=.72), screening for albuminuria (RR=.96), eye examination at least once every second year (RR=.97), and foot examination (RR=.96). No clear associations were found for the remaining measures (Table 4). We found no evidence of systematic interaction when stratifying the analyses according to sex, age, and HbA1c (data not shown).

In particular, age, drug abuse, duration of schizophrenia, and contact volume were associated with the quality of diabetes care [see online supplement]. The clearest associations were found for drug abuse and contact volume. In this case, drug abuse was significantly associated with a reduced likelihood of receiving a foot examination (adjusted RR=.88). Moreover, drug abuse was not significantly associated with a reduced chance of receiving high-quality diabetes care (unadjusted RR=.89, CI=.73–1.09), including a lower chance of receiving antihypertensive medication (unadjusted RR=.15, CI=.02–1.03) and several other measures.

In clinics with very high contact volume, the likelihood of receiving high-quality diabetes care was significantly greater than in clinics with low contact volume (adjusted RR=1.44), including treatment with ACE/ATII inhibitors (adjusted RR=3.19). When the analysis used cut points of 70% and 90% for defining high-quality schizophrenia care, the pattern remained, showing no clear association between a high level of quality of schizophrenia care and deficient diabetes care (data not shown).

Discussion

Our results indicate that individuals with diabetes and schizophrenia were less likely than those with diabetes only to receive overall high-quality diabetes care. Furthermore, those with diabetes and schizophrenia were less likely to receive several individual process performance measures of care, including blood pressure monitoring and screening for albuminuria; treatment with antihypertensive drugs, including ACE/ATII inhibitors; eye examination at least once every second year; and foot examination. Among those with diabetes and schizophrenia, patient- and system-specific factors were associated with the deficient quality of diabetes care, including young age, drug abuse, short duration of schizophrenia, and low clinic contact volume.

Strengths and Limitations

The strengths of the study include the nationwide population-based design, a large sample, and detailed information about the quality of diabetes care. Nevertheless, the data in the DSR and DADR are collected in routine clinical settings by various clinicians; thus variations by practice and errors may occur. However, thorough efforts are made to ensure data uniformity and accuracy by providing detailed instructions for the data collection in standardized registration forms, with explicit data definitions. In addition, regular multidisciplinary structured audits are conducted locally, regionally, and nationally to critically assess data validity and provide continuous feedback to hospitals. To ensure high validity, data collection for the DSR and DADR focuses on a limited set of variables. Organizations are not required to maintain extensive data sets for reporting; such data sets may include incomplete and poor-quality data. Because of the limited data, however, unmeasured factors may have influenced the quality of diabetes care, including, for example, measures of disease severity other than the GAF score and racial-ethnic background, although the Danish population is in general considered to be homogeneous.

Confounding was of minor concern because the included process performance measures were considered relevant for all included individuals independent of psychiatric diagnosis, patient characteristics, and the treating hospital clinic.

This study included only hospital outpatients diagnosed as having diabetes and being treated for diabetes. The impact of schizophrenia on diabetes care may thus be considerably larger than reflected by our findings because individuals with diabetes and schizophrenia are commonly undiagnosed, not in treatment, or treated only by GPs.

Comparison With Previous Studies

The diabetes performance measures used in our study are similar to those used elsewhere, including the HEDIS measures in the United States (9–12). However, none of the prior studies were based on a nationwide population, and variation in the examined psychiatric disorder makes direct comparison difficult. Also, our results are from a system that provides universal health care, and they may not be generalizable to other health services, such as the U.S. system, which indicates the need for further studies of the quality of diabetes care among patients with schizophrenia in various settings. The limited number of studies of the association between mental illness and possible predictors of diabetes care have methodological inconsistencies (10,11). Previous U.S. studies reported that patient-specific (gender, age, race-ethnicity, and number of general medical comorbidities), provider-specific (primary care workforce), and state-specific (reimbursement rates) factors predicted the quality of care (10,11). Financial incentives may, nonetheless, further complicate comparison of results with findings from a system with universal coverage. To our knowledge, only studies in cardiovascular care have assessed the association between mental illness and the organizational structure of a health care system as a potential predictor of quality of care (25–27). U.S. studies reported that patients with mental illness were more likely than those without mental illness to be admitted to low-quality hospitals and treated by low-quality cardiac surgeons for their cardiovascular comorbidities (25–27).

This study demonstrated that the quality disparities experienced by individuals with diabetes and schizophrenia are a nationwide problem, even in a country with free access to care. Despite the younger age of individuals with diabetes and schizophrenia, their general medical health status was worse than that of individuals with diabetes alone, and those with schizophrenia were more likely to demonstrate unfavorable lifestyle behaviors, including smoking. These findings indicate possible suboptimal prophylactic medical services. Even though the sample with diabetes and schizophrenia had exacerbated general medical problems, our results indicate that they were less likely than those with diabetes alone to receive high-quality diabetes care and several individual process performance measures, including treatment with antihypertensive drugs, blood pressure monitoring, screening for albuminuria, and eye and foot examination. In this context, the relatively long intervals between diabetes follow-up checks may constitute a particular challenge in maintaining contact and providing sufficient treatment. Moreover, the mental disorder itself may affect medication adherence and awareness of diabetes complications, including reduced sight and incipient foot ulcers. Cognitive and communicative limitations may prevent this vulnerable population from seeking and receiving care.

However, it must be acknowledged that the observed differences in diabetes care were modest, which may indicate that outpatient clinics are capable of accommodating individuals with diabetes and schizophrenia. Outpatient clinics may primarily treat patients who have mild cases of schizophrenia and who are more likely to attend and receive diabetes checks. The universal health care system and the establishment of the DADR that ensures guideline-recommended care for individuals with diabetes, supported by regular audits, should also be taken into consideration when interpreting the results. The remaining individual process performance measures, including glycemic monitoring, antidiabetic medication, cholesterol monitoring, lipid-lowering medication and eye examination at least once every fourth year, were equally distributed among individuals with and without schizophrenia.

We observed that drug abuse and contact volume had the clearest association with the quality of diabetes care. Individuals with schizophrenia and documented drug abuse were less likely than those without documented drug abuse to receive a foot examination, and clinics with very high contact volume were more likely than those with low contact volume to provide high-quality diabetes care, including treatment with ACE/ATII inhibitors. Individuals with diabetes and schizophrenia and concomitant drug abuse may have difficulty attending clinics for follow-up checks. Clinics with very high contact volume may have better resources, more specialized treatments, and personnel who have greater clinical experience with psychiatric patients.

We can only speculate about the mechanisms underlying the observed deficiencies in diabetes care and predictors of such deficiencies. For individuals with diabetes and schizophrenia, navigating between the psychiatric and general medical care systems may constitute a particular challenge. Physician barriers may include minimal experience with psychiatric patients and discomfort with or stigmatizing attitudes toward these patients. In addition, system factors, such as variations in resources by clinic and region, lack of coordination between systems, and time constraints because of the need to attend to patients’ comorbid conditions, may underlie disparities. A further intermediary factor may be disengagement from treatment. Individuals with diabetes and schizophrenia may find the treatment unsuitable to their needs or may experience difficulties in their relationship with the treating physician and may thus disengage from care (28–30).

Conclusions

This nationwide population-based cohort study demonstrated that schizophrenia was associated with a lower likelihood of receiving high-quality diabetes care and several individual process performance measures of diabetes care. However, the absolute differences in diabetes care between those with and without schizophrenia were modest. Predictors of the quality of diabetes care among individuals with schizophrenia included patient- and system-specific factors. We encourage further population-based studies of the association between schizophrenia and quality of diabetes care in health care systems with free and equal access to services. To effectively target quality improvement interventions, it is important to identify individuals with diabetes and schizophrenia, who are at particular risk of receiving deficient care.