Trends in Incidence of Hospitalization for Hypoglycemia and Diabetic Ketoacidosis in Individuals With Type 1 or Type 2 Diabetes With and Without Severe Mental Illness in Denmark From 1996 to 2020: A Nationwide Study Free

Individuals with severe mental illness (SMI), including schizophrenia, bipolar disorder, or major depression, have a two to three times higher risk of developing type 2 diabetes (1). Cohort studies have also reported that individuals with SMI and type 1 or type 2 diabetes are at increased risk of developing micro- and macrovascular diabetes complications (2,3) and, in some settings, receive lower-quality diabetes care (4).

Acute diabetes complications, including hypoglycemia and diabetic ketoacidosis (DKA), are serious and potentially life-threatening, but preventable, conditions (5,6). High amounts of alcohol consumption, drug abuse, comorbidities, and low levels of social support increase the risk of hypoglycemia and DKA (7–9). Additionally, impaired hypoglycemia awareness (10), factors associated with low socioeconomic status (11), and polypharmacy (7) increase the risk of recurrent hypoglycemia and DKA events. These factors are more prevalent in individuals with SMI (12,13), making them a potentially vulnerable group in terms of developing recurrent hypoglycemia and DKA.

Yet, only three previous studies have explored the risk of hospitalization for hypoglycemia and DKA in individuals with SMI and diabetes. These studies reported a 1.4–3.2 times higher risk of hospitalization for hypoglycemia and DKA among those with versus those without SMI (14–16). Only one of these studies examined recurrent acute complication events (16). Additionally, previous studies had several limitations, including use of a composite measure of hypoglycemia and DKA (14), lack of segregation by type of diabetes (14,16), and inclusion of a relatively short follow-up of 3–5 years (15,16). Two studies exclusively focused on schizophrenia (14,15), with none examining bipolar disorder. Furthermore, recent research has demonstrated that the disparity of risk of diabetes and its complications between individuals with and without SMI is most pronounced in the younger age groups (17) and has widened over time (18), but no study, to our knowledge, has examined how the risk of acute diabetes complications vary with age and calendar year. We sought to address these limitations and gaps using a large, nationwide, Danish cohort study to compare, among individuals with type 1 or type 2 diabetes, the incidence of hospitalization for hypoglycemia and DKA between 1996 and 2020 in individuals with versus those without SMI, by first and subsequent events, age, and calendar year.

Using a dynamic cohort study design, we identified all individuals with type 1 or type 2 diabetes diagnosed when they were older than 17 years, between 1 January 1996 and 30 June 2020. Individual-level data were linked by the unique personal identification number given to Danish residents (19). We identified individuals with diabetes using a nationwide Danish diabetes register (20) that uses information from six different registers (for a detailed description of the register, including the accuracy and completeness of diabetes ascertainment and a flow diagram of the study population, see Supplementary Material [pp. 1–3] and Supplementary Fig. 1). Individuals were followed from date of diabetes diagnosis until date of emigration, date of death, or end of follow-up (30 June 2020).

SMI diagnoses were obtained from the Danish Psychiatric Research Register, which contains all admissions to psychiatric inpatient facilities since 1969 and visits to all outpatient and emergency psychiatric departments since 1995 (21). In line with previous studies (22), we defined SMI as schizophrenia, bipolar disorder, or major depression, using the ICD-10 and corresponding ICD-8 diagnosis codes (Supplementary Table 1). We included a composite category comprising one or more SMIs. Separate SMI groups were not mutually exclusive, so some individuals appear in more than one group. Date of onset of SMI was defined as the first psychiatric assessment date, from 1969 onward. We considered SMI to be a time-varying variable, with individuals contributing to the non-SMI group until a diagnosis of SMI, after which they contributed to the appropriate SMI group.

The outcome of the study was hospitalizations for hypoglycemia and DKA (for ICD-10 codes, see Supplementary Table 1) in the period 1994–2020. This information was obtained from the Danish National Patient Register (23), which contains information on all inpatient admissions to all hospitals in Denmark since 1977 and outpatient and emergency contacts since 1995.

Confounders and mediators were identified based on prior evidence and a directed acyclic graph was constructed to illustrate the causal network of associations between SMI and acute diabetes complications (Supplementary Fig. 2). Confounders included age, sex, diabetes duration, calendar year, educational level, and country of origin. The Danish Civil Registration System (19) was used to obtain information on age, sex, and country of origin. Individuals were considered immigrants if they were born outside Denmark or had parents born outside Denmark and without Danish citizenship (19). Immigrants were categorized by country of origin as either Western or non-Western countries. Information on educational level was collected from the Education Register (24) and defined as the highest achieved education at date of diabetes diagnosis and categorized as primary, lower secondary, upper secondary, short cycle tertiary, bachelor’s degree or equivalent, master’s degree or equivalent, and doctoral degree or equivalent. The Danish diabetes register was used to extract information on diabetes duration.

A multistate framework was applied to the data set (25). The follow-up of each individual was subdivided at the date of each hypoglycemia or DKA event. Because a diabetes diagnosis (as defined in the diabetes register) does not necessarily come before an event, the number of hypoglycemia or DKA events prior to start of follow-up for each individual was counted (0.4% of the study cohort had a hypoglycemia event and 0.3% had a DKA event before diabetes diagnosis). This enabled us to classify each interval of follow-up by the number of previous events (specifically, none to six or more) without losing information on events occurring before diabetes diagnosis (see the multistate model in Supplementary Fig. 3). Follow-up time was split along calendar years in 1-year intervals.

Incidence rates (IRs) of hypoglycemia and DKA were modeled using Poisson regression, with the next event as outcome and the log person-years as offset. This approach allowed us to assess the IRs and incidence rate ratios (IRRs) for each first and subsequent complication event between individuals with and without SMI. We conducted separate analyses for type 1 and type 2 diabetes and composite and individual SMIs. There were sufficient data to include up to four and six events in the models for individual and composite SMI groups, respectively. For the combined SMI group, we modeled the IRs of each further complication event as a function of SMI, age (as a continuous variable), calendar year (as a continuous variable), and covariates, including interaction terms between SMI and age or calendar year. This yielded age- and calendar year–specific IRs between individuals with and without SMI. Covariates were included in the models in two steps. The basic adjusted model included age, sex, diabetes duration, and calendar year, and the fully adjusted model additionally included education and country of origin. The effects of current age, diabetes duration, and calendar year were modeled as natural splines with five knots. The knots were located so that the hypoglycemia and DKA events were evenly distributed between the knots. We examined whether number of events, age, and calendar year modified the association between SMI and hypoglycemia or DKA. Likelihood ratio tests were used to compare models with and without the interaction term. Complete data were available for all variables except highest educational level and country of origin, for which 9% and 0.4%, respectively, of the study population had missing information. An additional missing category was included within each of these variables. Statistical analyses were performed using R, version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria; www.R-project.org). The Epi-package, version 2.48, was used to apply the multistate models (26).

The study was approved by the Danish Data Protection Agency. Register-based studies do not require consent, according to Danish law. All data were anonymized before release to the researchers.

Data supporting this study are obtainable from Statistics Denmark. Access to these data is subject to restrictions and access requires an application and permission.

We included 20,160 individuals with type 1 diabetes and 413,449 with type 2 diabetes in this study. Among those with type 1 diabetes, 596 (3%) had schizophrenia, 203 (1%) had bipolar disorder, and 1,186 (6%) had major depression. The proportions were comparable among individuals with type 2 diabetes (Table 1). Of all individuals with an SMI diagnosis, 27,397 (75%) received the SMI diagnosis before the diabetes diagnosis. Median follow-up was 8.9 years (interquartile range [IQR] = 3.9–15.3) for individuals with type 1 diabetes and 7.2 years (IQR = 3.2–11.7) for individuals with type 2 diabetes. Irrespective of type of diabetes, those with SMI were more likely to be women, have been diagnosed with diabetes at a younger age, have a lower level of education, and have a non-Western origin (Table 1).

Among individuals with type 1 diabetes, one or more hypoglycemia and DKA events were more common among individuals with than without SMI (25% vs. 16% for hypoglycemia and 21% vs. 12% for DKA, respectively). There were fewer events among those with type 2 diabetes, but patterns by SMI status were similar (Table 1).

The number of events, age, and calendar year modified the association between SMI and hypoglycemia or DKA (Supplementary Table 2); therefore, associations between SMI and hypoglycemia or DKA are presented from analyses stratified by these variables.

Among individuals with type 1 diabetes, the overall crude IR of first hypoglycemia and DKA event during follow-up were 16.2 and 8.1 per 1,000 person-years, respectively, for individuals without SMI and 27.4 and 14.0 per 1,000 person-years for individuals with SMI (Table 2).

In fully adjusted models, SMI was associated with a higher IRR of hypoglycemia across all event numbers except event six and above, with the IRR decreasing with increasing number of events. SMI was also associated with a higher IRR for the first and second DKA events. IRRs were largest for first events (IRR 1.77 [95% CI 1.56–2.00] for hypoglycemia; and 1.78 [95% CI 1.50–2.11] for DKA; Table 2).

The pattern of results was broadly similar when investigating by individual SMI disorder, but with fewer events resulting in reduced statistical power and lower precision of some estimates. The IRR of recurrent hypoglycemia was greater for individuals with schizophrenia compared with those with bipolar disorder or major depression (Table 3).

Among individuals with type 2 diabetes, the overall crude IR of first hypoglycemia and DKA event during follow-up were 4.2 and 0.7 per 1,000 person-years, respectively, for individuals without SMI and 6.1 and 1.4 per 1,000 person-years, respectively, for individuals with SMI (Table 2).

In fully adjusted models, SMI was associated with a statistically significantly higher IRR of hypoglycemia across almost all numbers of events. The highest IRR of hypoglycemia was observed for the first event (IRR 1.64; 95% CI 1.55–1.74) and the sixth and later events (IRR 1.70; 95% CI 1.14–2.26). Individuals with SMI had a higher IRR of the first DKA event (IRR 1.85; 95% CI, 1.64–2.09), but a similar IRR of subsequent events (Table 2).

Overall, the results were similar when investigating by the individual SMI disorders. However, individuals with schizophrenia had higher IRRs of recurrent hypoglycemia events when compared with those with bipolar disorder or major depression. The IRRs of DKA were mostly elevated for those with bipolar disorder across the number of events, whereas it possibly decreased for those with schizophrenia and major depression (Table 3).

In the type 1 diabetes group, the IRs of first hypoglycemia event decreased with increasing age until age 40 years, after which the IRs increased. This pattern was more pronounced in individuals with SMI as compared with those without SMI (Fig. 1A). In the type 2 diabetes group, the IRs for first hypoglycemia event decreased with increasing age until age 60 years, after which it increased in individuals with and in those without SMI (Fig. 1C). Irrespective of diabetes type, hypoglycemia IRs were higher for individuals with SMI compared with those without SMI across the entire age span, with differences being largest at younger ages (<50 years old) (Fig. 1A and C). For example, in the type 1 diabetes group, the incidence rate difference (IRD) for hypoglycemia between individuals with and without SMI was 24.4 cases per 1,000 person-years at age 25 years and 10.6 cases per 1,000 person-years at age 75 years (calculations are based on Fig. 1A).

In the type 1 diabetes group, the IRs of first DKA event decreased with increasing age. Furthermore, the IRs of first DKA event was higher for those with than those without SMI, up to 60 years of age, after which rates by SMI status were similar. The difference was greatest in the age span 25–40 years. (Fig. 1B). For example, the IRD in DKA between individuals with and without SMI was 27.5 cases per 1,000 person-years at age 25 years and 3.5 cases per 1,000 person-years at age 60 years (calculations are based on Fig. 1B). In the type 2 diabetes group, the IR of first DKA event decreased with age in individuals without SMI until 60 years of age, after which it stabilized. In the SMI group, the IR increased until 50 years of age, after which it decreased. The DKA IR was higher for those with than those without SMI over the entire age span, except for the 30–49 age group, for which it was similar (Fig. 1D).

In the type 1 and type 2 diabetes groups, the IRs of first hypoglycemia event increased from 1996 to 2005, although the increase was small in the type 2 diabetes group, after which the IRs decreased in individuals with and those without SMI. The excess IR of hypoglycemia in individuals with SMI remained unchanged between 1996 and 2020 (Fig. 1E and G). In the type 1 diabetes group, the IRs of first DKA event increased with increasing calendar year from 1996 to 2015, irrespective of SMI status. After 2015, the IR stabilized for individuals without SMI, whereas it decreased in individuals with SMI (Fig. 1F). In the type 2 diabetes group, the DKA IRs increased from 1996 to 2008, after which it decreased in both comparison groups, with the excess IR for those with compared with those without SMI persisting over time (Fig. 1H).

IRs of second and third hypoglycemia and DKA events are presented in Supplementary Figures 4 and 5. Overall, the figures show that the differences in IRs of second and third hypoglycemia and DKA events between individuals with and without SMI were similar across age and calendar year.

In this cohort study, based on nationwide registers, we found an increased risk of first and recurrent hypoglycemia and DKA hospitalizations in individuals with versus those without SMI, regardless of type of diabetes. SMI was associated with a 1.7- and 1.8-fold higher risk of first hypoglycemia event and first DKA event, respectively. As expected, hypoglycemia and DKA IRs were higher in type 1 diabetes compared with type 2 diabetes. Furthermore, individuals with schizophrenia faced a particularly heightened risk of recurrent hypoglycemia events compared with the other SMI disorders. Additionally, the IRD of hypoglycemia in both diabetes groups and DKA in the type 1 diabetes group between individuals with and without SMI was more prominent in younger individuals (<50 years old) and remained stable over calendar years.

Previous studies examining the risk of hospitalization for hypoglycemia and DKA for individuals with SMI have reported similar findings (14–16). Goueslard et al. (15) found that among individuals with type 1 diabetes who were aged 15–35 years, those with schizophrenia had a 3.2- and 2-times higher risk of first hypoglycemia and DKA event, respectively, compared with individuals without schizophrenia. The study population was younger than in the present study, which may account for the slightly higher excess risk of outcomes observed in this previous study. Becker and Hux (14) used a composite measure of acute complications comprising hypoglycemia and DKA and found a 74% higher risk of a first event for individuals with schizophrenia, which aligns with our findings. Katon et al. (16) found that major depression was associated with a 42% higher risk of hypoglycemia. We found a somewhat higher excess risk, likely due to differences in study populations. The former study identified the study population from a survey with a response rate of 54% (16), whereas we included an unselected and nationally representative cohort of individuals with diabetes.

Our study extends the findings of previous studies by showing that individuals with SMI are at increased risk of not only the first but also subsequent events of hypoglycemia and DKA and that individuals with schizophrenia are at particularly high risk of recurrent hypoglycemia events. Several factors are likely to contribute to the higher risk of recurrent acute complications of diabetes among individuals with SMI. Many episodes of severe hypoglycemia are treated in the community by family, friends, and ambulance services (27). Individuals with SMI, who are more likely to live alone and to lack social support (28), may experience more hospitalizations due to limited access to immediate assistance in managing severe hypoglycemia. Suicide attempts involving insulin overdose are more prevalent among individuals with SMI (29), offering another explanation of the excess risk. Psychiatric symptoms and medications with sedative effects can hinder the ability to perceive and recognize severe hypoglycemia events, which also increases the risk. Additionally, comorbidities, alcohol consumption, and impaired hypoglycemia awareness could also contribute to the increased risk of hypoglycemia in individuals with SMI (8,30). We found that individuals with schizophrenia, the most severe of the three SMIs, are at particular risk of recurrent hypoglycemia events, which may reflect that those with schizophrenia experience more of the aforementioned explanatory factors.

Missed insulin treatments or inadequate insulin therapy are among the important risk factors for DKA (9). However, studies suggest that the adherence to diabetes medication is similar or better in individuals with SMI (31). Although sodium-glucose cotransporter-2 inhibitors are a potential choice for individuals with SMI, due to their protective renal and cardiovascular effects and weight reduction, they also pose an increased risk of DKA (32). However, a recent Danish study reported that the use of sodium-glucose cotransporter-2 inhibitors among individuals with SMI was not higher than among those without SMI (33). Furthermore, although recent technological advances such as sensors and pumps affect risk of DKA, their use has not yet been examined by SMI status. Given observed socioeconomic disparities in access to these technologies (34), their use is likely lower among those with SMI. However, there is a lack of studies examining access to newer diabetes technologies among individuals with SMI. Furthermore, some studies suggest certain antipsychotic drugs may increase the risk of DKA (35), yet limited evidence necessitates further research.

As expected, we observed markedly higher IRs of hypoglycemia and DKA in the type 1 diabetes group compared with the type 2 diabetes group. This discrepancy is probably related to the absolute insulin deficiency and the reliance on exogenous insulin therapy among individuals with type 1 diabetes, making them more prone to acute complications. We found that the risk of hypoglycemia and DKA between individuals with and without SMI was particularly elevated in younger individuals (<50 years of age), whereas no age-related trend was observed for DKA in the type 2 diabetes group. However, low numbers of DKA events in this group could contribute to chance findings. Previous studies have reported a higher excess risk of type 2 diabetes and micro- and macrovascular diabetes complications in individuals with SMI in younger age groups than in older age groups (3,17). We extend these findings by demonstrating that the risk of acute complications is similarly higher for younger individuals (<50 years old). SMI is commonly diagnosed in young adulthood, when individuals may encounter challenges due to psychotic symptoms and insufficient disease control (36). They may be more inclined to delay or skip meals, miss insulin treatments, and consume more alcohol, thereby increasing the risk of hypoglycemia and DKA.

For individuals with and those without SMI, hypoglycemia IRs slightly increased from 1996 to 2005, followed by a decrease. These time trends are in line with previous research (37), and the decrease probably reflects less use of sulfonylureas and insulins and increased access to diabetes technologies. The IRs of hypoglycemia hospitalizations were in line with previous studies examining such hospitalizations (37), but lower when compared with studies examining hypoglycemia events that did not require hospitalizations (38). The decrease in IRs from 2006 to 2020 was most pronounced for the type 2 diabetes group, possibly reflecting increased use of newer glucose-lowering medications that have a lower risk of hypoglycemia compared with sulfonylureas and insulins. For both types of diabetes, DKA IRs increased until 2007, with a continued increase for individuals with type 1 diabetes, albeit less pronounced, and a subsequent decrease in the type 2 diabetes group. These time trends align with previous research (39) and might reflect better treatment options. The continued increase after 2007 in DKA IRs in the type 1 diabetes group, contrary to the decrease in the type 2 diabetes group, may be attributed to improved access to diabetes technologies among individuals with type 1 diabetes, because insulin pump therapy has been associated with an increased risk of DKA (40). The excess risk of hypoglycemia and DKA in individuals with versus without SMI was stable over calendar years. Although the SMI disparity in type 2 diabetes incidence potentially has been widening (18), we did not observe this pattern for acute complications, which is reassuring.

Our study provides a significant contribution to an area that has received limited research attention. The main strength of the study lies in the use of population-based Danish registers, which allowed us to analyze a nationally representative and unselected cohort of individuals with diabetes. The large study population and long follow-up of up to 24 years allowed us to investigate the trends in incidence of recurrent hypoglycemia and DKA events by SMI status, including individual SMI disorders. To our knowledge, no previous study has examined this in individuals with bipolar disorder. Unlike previous studies that solely focused on the first occurrence (14,15), we examined recurrent events of hypoglycemia and DKA, providing a more complete understanding of the association between SMI and acute diabetes complications. Additionally, we distinguished between type 1 and type 2 diabetes, something that former studies have not done (14,16).

The study also has several limitations. Hypoglycemia can result from overtreatment with insulin or insulin secretagogues (8). In our study, we did not account for prescribed antidiabetes medication; therefore, the higher risk of hypoglycemia observed in individuals with SMI may reflect differences in medications and their use between individuals with and without SMI. A recent Danish study found that a higher proportion of individuals with type 2 diabetes and SMI were treated with insulin during the first 10 years after diabetes diagnosis compared with those without SMI (33). Additionally, lower-cost diabetes medications, like insulins and sulfonylureas, have higher hypoglycemia risk (41) and may be used by greater proportions of individuals with SMI, as a consequence of their lower cost (42). However, it is important to note that our study aimed to assess the overall risk of acute diabetes complications rather than investigate the specific mechanisms through which SMI affects acute diabetes complications. Consequently, we did not adjust for the identified mediators. Nonetheless, some of these mediators, like substance use disorder and smoking, may contribute to the development of schizophrenia and could be considered confounders (43,44). Therefore, we adjusted our analysis for substance use disorder, which led to a slight attenuation of the effect estimates (Supplementary Table 3). However, given that 34% of the individuals with SMI and substance use disorder received the latter diagnosis after the SMI diagnosis, potential overadjustment is suspected. Unfortunately, the lack of data on smoking prevented adjustment in our analysis. Moreover, educational level does not necessarily reflect all aspects of socioeconomic status, potentially leading to residual confounding from this and other factors. Furthermore, there is a possibility that some individuals with type 2 diabetes who solely managed their condition through lifestyle changes may have been overlooked in the beginning of the study period (Supplementary Material, see pp. 2–3). However, we expect the misclassification to be minimal. Additionally, we only had information on hospitalizations for hypoglycemia and, therefore, could not include severe and milder hypoglycemia events managed outside the hospital setting.

The findings from this study highlight the importance of improving prevention of hypoglycemia and DKA among individuals with SMI and diabetes. Although most studies have focused on the first event of these complications, this nationwide study demonstrates that individuals with SMI also have a higher risk of subsequent events and there is a particularly high risk associated with SMI for younger individuals (<50 years old). Thus, preventive initiatives must include targeting of the young population. The risk for hypoglycemia was high across all numbers of events; therefore, closer clinical care should be given to individuals with prior hypoglycemia events, particularly for individuals with schizophrenia, because we found the highest risk in this group. Potential approaches to prevent acute complications among individuals with type 1 diabetes and SMI may involve revising glycemic targets. For those with type 2 diabetes, a shift to a treatment regimen excluding insulin or sulfonylurea could be considered. Additionally, enhancing glycemic control through the use of a continuous glucose monitoring device may be effective for some individuals. Another affective approach to prevent acute diabetes complications is diabetes self-management education (9). Thus, considering this intervention for individuals with diabetes and SMI could be beneficial. However, modifications would be necessary to accommodate their needs and circumstances. More research is required to improve the understanding of psychotropic and glucose-lowering medications, including newer technologies, in the risk of acute diabetes complications. Additionally, it is also important to also investigate the association between other mental disorders and acute diabetes complications.

In conclusion, individuals with SMI have a markedly higher risk of first and subsequent hypoglycemia and DKA hospitalizations, regardless of diabetes type. Individuals with schizophrenia are particularly at risk for recurrent hypoglycemia hospitalizations. SMI disparities in incidence of hypoglycemia and DKA (only in type 1 diabetes) events were greatest in younger individuals (<50 years old) and stable across calendar years. Therefore, improved prevention of acute diabetes complications in individuals with SMI is needed, especially for younger individuals and those with schizophrenia. Studies should focus on identifying the underlying mechanisms that contribute to this elevated risk and developing effective interventions.

Acknowledgments. This work was performed in partial fulfillment of the requirements for a PhD degree for S.H.S.

Funding. This work received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Duality of Interest. G.S.A., M.E.J., and B.C. hold shares in Novo Nordisk A/S. S.H.W. attends meetings of the Scottish Study Group for Care of Diabetes in the Young that have been supported by Novo Nordisk. M.E.J. has received research grants from Amgen, Astra Zeneca, Boehringer Ingelheim, Novo Nordisk, and Sanofi Aventis. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. M.E.J., G.S.A., M.E.B., and S.H.S. were involved in the conceptualization and the design of the study. The data acquisition was carried out by S.H.S., V.K., and G.S.A. S.H.S. and B.C. had full access to all the data and assessed the underlying data. B.C. detailed the statistical method. S.H.S. performed the data management and analysis; drafted the manuscript, which all authors commented on; and is the guarantor of this work and takes responsibility for the integrity of the data and the accuracy of the data analysis. All authors approved the final manuscript.

Prior Presentation. This work was presented in part in an oral presentation at the European Diabetes Epidemiology Group Conference 2024, Pesaro, Italy, 20–23 April 2024.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Elizabeth Selvin and Alka M. Kanaya.