Patients With Diabetes Requiring ED Care for Hypoglycemia
Patients With Diabetes Requiring ED Care for Hypoglycemia
This study highlights the importance of triangulating different data sources to more accurately estimate the true incidence of hypoglycaemia in patients with diabetes requiring ED care. It also supports concerns raised in previous studies of the frequency of sulfonylurea-induced hospitalisation. Finally, it also suggests that patients with diabetes requiring ED care for hypoglycaemia have significant long-term mortality.
Coded data are a poor reflection of clinical practice, though this has improved in the last decade. In this study, only 34 out of 108 (31%) eligible episodes identified from AERs were recorded using the code for diabetes-related emergency calls. This could partly be because hypoglycaemic symptoms, signs and sequelae are non-specific. Therefore, a patient with hypoglycaemic seizure could easily be coded only as a seizure and not as diabetes-related emergency call. Had the AERs been searched by only including the code for diabetes-related emergency calls with no additional data obtained from HES and PAS, 131 of the 165 (79.4%) episodes would have been missed. Even when AERs were interrogated to identify all CBG<4 mmol/L, 34.5% of episodes would have been missed had HES and PAS data not been included. HES and PAS data coding also have significant limitations. Both rely on interpretation of case records by non-medical personnel who are seldom in discussion with clinicians about coding issues. Not surprisingly, this can result in missing and/or inaccurate data as seen in this study; PAS had more missing data than HES. Using PAS data on its own would have missed 72.7% of eligible episodes and HES on its own 44.2%. The unique feature of this study is the triangulation of data demonstrating that the number of eligible episodes is substantially greater than would have been otherwise realised. It is still possible for this study to have missed a small number of eligible episodes due to incorrect coding in all the three data sources. Nevertheless, this study suggests that previous UK publications that relied primarily on a single data source (usually ambulance records) are likely to have considerably underestimated the size of the problem, and future studies should consider including data from multiple sources to improve accuracy. Accurate data are important for several reasons, including their use to assess the quality of clinical services, to generate metrics and indicators for quality clinical outcomes, payment by results and even for appraisal and revalidation of consultants.
Recently, concerns have been raised about both inappropriate use of sulfonylureas in the elderly and hospitalisation due to sulfonylurea-induced hypoglycaemia. These concerns are supported by this study, which shows that one-third of hypoglycaemic episodes in patients with type 2 diabetes requiring ED care are exclusively related to sulfonylurea therapy. These patients were older, had worse renal function, more comorbidity and lower HbA1c than those on insulin. In addition, they suffered more prolonged hypoglycaemia and it is therefore essential that they be hospitalised for a longer period of observation. Furthermore, it could be argued that all patients who experience sulfonylurea-induced hypoglycaemia in the community even if they have fully recovered be transferred to ED for hospitalisation to prevent recurrence. However, this is not always the case; only a third of all hypoglycaemia-related ambulance callouts result in transportation to ED and what percentage of sulfonylurea-related events are transferred is unknown. In most patients treated onsite and not transported, paramedic personnel do not notify other healthcare professionals due to respect for patients' privacy and confidentiality. Also, the patients do not always inform their healthcare professionals for reasons such as ignorance, fear of local driving regulations, and so on. This is unfortunate as the opportunity for healthcare professionals to help patients prevent recurrent severe hypoglycaemia is lost.
This study also supplements the finding of significant long-term mortality in patients with diabetes requiring emergency care reported in a similar study from the USA. The 1-year mortality seen in this study was comparable to that reported. Additionally in that study, age and comorbidities were predictors of long-term mortality as in this study, but in contrast hospitalisation was not. The latter could be due to differences in the healthcare systems. In this study, type 2 diabetes was a risk factor for long-term all-cause mortality in those who required ED care for hypoglycaemia. However, when corrected for other risk factors such as comorbidities, renal function and age, this was no longer greater than those with type 1 diabetes.
There are several limitations to this study. Since HbA1c for the preceding 3 months of the hypoglycaemic episode was available in only 72 patients, the study included results for up to 12 months prior or 14 days after the episode, whichever was nearer. Even then it was not available for all the episodes. Due to the small sample size of patients in this study, there is possibility of a type II error that needs to be considered when interpreting the data. This study did not include mortality data for the local population with diabetes as this was unavailable. Therefore, whether hypoglycaemia requiring ED care is an independent risk factor for long-term mortality or only a marker for the degree of underlying illness is uncertain; the latter being more likely. A prospective well-designed larger study is required to answer this important question. Multiple hypoglycaemic events for an individual were not included as a risk factor in the Cox regression model for long-term mortality. However, a previous study that did include it in their analysis did not find it a significant risk factor for long-term mortality.
In summary, this study demonstrates that triangulation of multiple data sources provides a more accurate estimate of the incidence of hypoglycaemia in patients with diabetes requiring ED care. It also reports that one-third of hypoglycaemic episodes in patients with type 2 diabetes were associated with sulfonylurea therapy and that there is significant long-term mortality in patients with diabetes requiring ED care for hypoglycaemia. Further work is needed on whether hypoglycaemia requiring emergency medical care is an independent risk factor for long-term mortality or only an epiphenomenon.
Discussion
This study highlights the importance of triangulating different data sources to more accurately estimate the true incidence of hypoglycaemia in patients with diabetes requiring ED care. It also supports concerns raised in previous studies of the frequency of sulfonylurea-induced hospitalisation. Finally, it also suggests that patients with diabetes requiring ED care for hypoglycaemia have significant long-term mortality.
Coded data are a poor reflection of clinical practice, though this has improved in the last decade. In this study, only 34 out of 108 (31%) eligible episodes identified from AERs were recorded using the code for diabetes-related emergency calls. This could partly be because hypoglycaemic symptoms, signs and sequelae are non-specific. Therefore, a patient with hypoglycaemic seizure could easily be coded only as a seizure and not as diabetes-related emergency call. Had the AERs been searched by only including the code for diabetes-related emergency calls with no additional data obtained from HES and PAS, 131 of the 165 (79.4%) episodes would have been missed. Even when AERs were interrogated to identify all CBG<4 mmol/L, 34.5% of episodes would have been missed had HES and PAS data not been included. HES and PAS data coding also have significant limitations. Both rely on interpretation of case records by non-medical personnel who are seldom in discussion with clinicians about coding issues. Not surprisingly, this can result in missing and/or inaccurate data as seen in this study; PAS had more missing data than HES. Using PAS data on its own would have missed 72.7% of eligible episodes and HES on its own 44.2%. The unique feature of this study is the triangulation of data demonstrating that the number of eligible episodes is substantially greater than would have been otherwise realised. It is still possible for this study to have missed a small number of eligible episodes due to incorrect coding in all the three data sources. Nevertheless, this study suggests that previous UK publications that relied primarily on a single data source (usually ambulance records) are likely to have considerably underestimated the size of the problem, and future studies should consider including data from multiple sources to improve accuracy. Accurate data are important for several reasons, including their use to assess the quality of clinical services, to generate metrics and indicators for quality clinical outcomes, payment by results and even for appraisal and revalidation of consultants.
Recently, concerns have been raised about both inappropriate use of sulfonylureas in the elderly and hospitalisation due to sulfonylurea-induced hypoglycaemia. These concerns are supported by this study, which shows that one-third of hypoglycaemic episodes in patients with type 2 diabetes requiring ED care are exclusively related to sulfonylurea therapy. These patients were older, had worse renal function, more comorbidity and lower HbA1c than those on insulin. In addition, they suffered more prolonged hypoglycaemia and it is therefore essential that they be hospitalised for a longer period of observation. Furthermore, it could be argued that all patients who experience sulfonylurea-induced hypoglycaemia in the community even if they have fully recovered be transferred to ED for hospitalisation to prevent recurrence. However, this is not always the case; only a third of all hypoglycaemia-related ambulance callouts result in transportation to ED and what percentage of sulfonylurea-related events are transferred is unknown. In most patients treated onsite and not transported, paramedic personnel do not notify other healthcare professionals due to respect for patients' privacy and confidentiality. Also, the patients do not always inform their healthcare professionals for reasons such as ignorance, fear of local driving regulations, and so on. This is unfortunate as the opportunity for healthcare professionals to help patients prevent recurrent severe hypoglycaemia is lost.
This study also supplements the finding of significant long-term mortality in patients with diabetes requiring emergency care reported in a similar study from the USA. The 1-year mortality seen in this study was comparable to that reported. Additionally in that study, age and comorbidities were predictors of long-term mortality as in this study, but in contrast hospitalisation was not. The latter could be due to differences in the healthcare systems. In this study, type 2 diabetes was a risk factor for long-term all-cause mortality in those who required ED care for hypoglycaemia. However, when corrected for other risk factors such as comorbidities, renal function and age, this was no longer greater than those with type 1 diabetes.
There are several limitations to this study. Since HbA1c for the preceding 3 months of the hypoglycaemic episode was available in only 72 patients, the study included results for up to 12 months prior or 14 days after the episode, whichever was nearer. Even then it was not available for all the episodes. Due to the small sample size of patients in this study, there is possibility of a type II error that needs to be considered when interpreting the data. This study did not include mortality data for the local population with diabetes as this was unavailable. Therefore, whether hypoglycaemia requiring ED care is an independent risk factor for long-term mortality or only a marker for the degree of underlying illness is uncertain; the latter being more likely. A prospective well-designed larger study is required to answer this important question. Multiple hypoglycaemic events for an individual were not included as a risk factor in the Cox regression model for long-term mortality. However, a previous study that did include it in their analysis did not find it a significant risk factor for long-term mortality.
In summary, this study demonstrates that triangulation of multiple data sources provides a more accurate estimate of the incidence of hypoglycaemia in patients with diabetes requiring ED care. It also reports that one-third of hypoglycaemic episodes in patients with type 2 diabetes were associated with sulfonylurea therapy and that there is significant long-term mortality in patients with diabetes requiring ED care for hypoglycaemia. Further work is needed on whether hypoglycaemia requiring emergency medical care is an independent risk factor for long-term mortality or only an epiphenomenon.