Patterns of engagement with the health care system and risk of subsequent hospitalization amongst patients with diabetes

Diabetes affects approximately one in ten adults in Canada [1, 2] with treatment costs estimated to exceed $12 billion dollars per year [3]. A large component of costs are attributed to the direct costs of inpatient care, despite the fact that diabetes is a chronic condition generally amenable to outpatient treatment [4‐6]. Patients with diabetes have an increased risk of hospitalization compared to those without diabetes [7‐10]. Further, repeated hospitalizations are also common and, although observed in a smaller proportion of patients with diabetes, may represent a disproportionate burden on the health care system [11‐15].

Since re-hospitalization is common and an important driver of morbidity and costs in diabetes [11‐15], identifying which patients are at highest risk of subsequent hospitalization is relevant. Given the multiple factors which might contribute to re-hospitalization, the Andersen Behavioral Model for health care utilization can be used as a framework to identify the important patient and provider/system level factors [16]. Prior studies have identified patient level factors including demographic (age, sex, race, and socioeconomic status), clinical (comorbidity, diabetes duration), and behavioral factors (glycemic control, and adherence to medication) [11‐15, 17], though many of these studies were restricted to pediatric or elderly patients with diabetes limiting the generalizability of their findings. Furthermore, few studies have assessed aspects of patient care related to patterns of engagement with the health care system, including use of primary care, emergency rooms, and the circumstances around hospital discharge (including destination and whether people leave against medical advice).

Given the limitations in previous research and the burden that repeated hospitalization places on the health care system, we used population-based data to determine the association between patterns of health care engagement (health resource utilization and discharge disposition), and the risk of subsequent hospitalization among patients with diabetes.

In this large, population-based cohort of adults with diabetes and at least one hospitalization, we found that certain patterns of engagement with the health care system prior to the initial hospitalization, specifically higher use of the emergency department, and limited or increased use of primary care, were associated with an increased risk of subsequent hospitalization. Moreover, patients discharged against medical advice were more likely to be re-hospitalized. Given the financial burden that in-patient care places on the health care system, the ability to identify patients at highest risk of subsequent hospitalization is not only hypothesis generating, but may help healthcare providers target resources to high-risk patients.

Our results add to those from previous studies. Smith et al. found that the number of emergency department visits in the 6-month period prior to hospitalization was a significant predictor of 90-day repeat hospitalization among patients with chronic disease, some of whom had diabetes [31]. Our results extend this finding to a large cohort of patients with diabetes, and demonstrate similar risk associated with both all-cause and cause-specific subsequent hospitalization. The higher risk of re-hospitalization among patients with diabetes who have a higher use of the emergency department visits may reflect a sicker patient population with multi-morbidity [5, 32, 33], though we noted the same association after controlling for measured comorbidity. It is also possible it reflects patients with more severe diabetes, or those who are prone to hospitalizations related to hypo or hyperglycemia; hospitalizations that might be prevented by appropriate access and use of primary care services [34, 35]. Regardless of the cause, higher use of the emergency department does appear to identify a group of patients at higher risk of re-hospitalization.

We also observed a relationship between the number of primary care physician visits and risk of subsequent all-cause hospitalization. In various chronic disease populations, increased primary care accessibility and use has been associated with decreased risk of hospitalization, especially for ambulatory care sensitive conditions [36, 37]. In diabetes populations specifically, multiple physician visits have been shown to be associated with risk of first hospitalization [38]. Our results suggest that a higher number of primary care physician visits are also associated with a greater risk of subsequent hospitalization, possibly because those with multiple visits are sicker patients who require more complex care. Though our results suggest that the majority of patients have adequate access to primary care services, as observed by the rate of use both before and after hospital discharge, it may be that our current model of providing care is not adequate for management of complex patients in a primary care setting. A multidisciplinary approach to chronic disease care has been proposed and shown to reduce the risk of hospitalization specifically in patients with diabetes [39]. In addition, we found a potential increased risk of subsequent all-cause hospitalization in patients with no general physician visits prior to their index event. While limited to a small proportion of the study population and only significant in our sensitivity analysis, these results support previous literature showing that limited access is associated with increased hospitalization risk in chronic disease populations [40‐42]. Future work is required to identify characteristic of these patients at high risk and determine whether the absence of health care use in the period prior to hospitalization represents limited access or health behaviors in which a patient chooses not to seek care.

A unique finding of our study was the association between discharge disposition, whether a patient left against medical advice, and risk of repeated hospitalization. Specifically, we found that patients discharged to palliative or long-term care were less likely to have a subsequent hospitalization (possibly given the competing risk of death) whereas those who left against medical advice were significantly more likely. In a cohort of elderly patients with diabetes identified within the California State Inpatient Dataset, Kim et al. found that a discharge disposition other than home was associated with an increased risk for an unscheduled subsequent hospitalization (OR: 1.28; 95% CI: 1.24–1.32) [14]. However, their dichotomous analysis could not determine how different transitions of care place patients with diabetes at different risk for subsequent hospitalization. Patients with diabetes who leave against medical advice represent a high-risk group worthy of future study to better understand the circumstances surrounding the discharge against medical advice.

Our study should be interpreted in light of its limitations. First, there are a number of factors that place patients at increased risk of subsequent hospitalization, including severity of disease, and thus, the possibility of residual confounding exists given our administrative data sources. However, we were able to adjust for a number of patient and clinical characteristics, including laboratory tests, which represent proxy measures of disease severity. Second, we were unable to determine whether a patient had a regular primary care physician, or the level of coordination available during the transition from the hospital to community care. Continuity of care and physician accessibility has been associated with improved outcomes, particularly in diabetes [40‐42]. Our inability to adjust for these factors may also confound the observed associations. Finally, we assessed all-cause and cause-specific subsequent hospitalizations irrespective of the index hospitalization diagnosis, which makes the interpretation of these associations less clear. However, patients with diabetes often suffer from various micro and macro-vascular complications, and studies have shown that patients with diabetes are often hospitalized due to one or more of these complications [43]. Further, any hospitalization (regardless of type) represents a burden on the health care system. For these reasons, we elected to consider all hospitalizations that occurred among the study population.

Despite these limitations, our study has a number of strengths. We utilized population-based data within a single province of Canada, which provides a unique opportunity to comprehensively assess the issue of subsequent hospitalization in patients with diabetes. We also grounded this work in a recognized framework for the study of health care utilization (the Andersen Behavioral Model), and our results highlight the need for researchers and clinicians to consider health resource use and discharge disposition in context of the known patient-level and clinical factors that place diabetes patients at risk for repeated hospitalization. Unlike many patient and clinical characteristics, these associations observed could be considered modifiable and represent areas that require further exploration.

In summary, we found certain patterns of engagement with the health care system are associated with an increased risk of subsequent hospitalization among patients with diabetes, including increased frequency of emergency department visits, limited and increased use of primary care visits, as well as leaving the hospital against medical advice. We acknowledge that repeated hospitalization is a complex topic that requires an understanding of the multiple patient, provider and system level factors that influence it. Though subsequent hospitalization may represent progression in the natural history of the patient’s underlying disease, or the consequences of poor coordination of care following discharge, our findings should be considered hypothesis-generating and represent an important step in the development of strategies to identify and intervene on patients at high risk of re-hospitalization. Our results also highlight the need to consider the patterns of health care engagement when studying re-hospitalization among patients with diabetes. Inclusion of these factors may ultimately improve predictive accuracy of this outcome in future studies.