Patrick Cawley, MD

With the continuing growth of the specialty of hospital medicine, the capabilities and performance of hospital medicine groups (HMGs) varies significantly. There are few guidelines that HMGs can reference as tools to guide self‐improvement. To address this deficiency, the Society of Hospital Medicine (SHM) Board of Directors authorized a process to identify the key principles and characteristics of an effective HMG.

METHODS

Final Approval

The final draft of the 10 principles and 47 characteristics was approved for publication at a meeting of the SHM Board of Directors in September 2013 (Figure 1).

jhm2119-fig-0001-m.png

RESULTS

A recurring issue that the workgroup addressed was the applicability of the characteristics from 1 practice setting to another. Confounding factors include the HMG's employment/organizational model (eg, hospital employed, academic, multispecialty group, private practice, and management company), its population served (eg, adult vs pediatric, more than 1 hospital), and the type of hospital served (eg, academic vs community, the hospital has more than 1 HMG). The workgroup has made an effort to assure that all 47 characteristics can be applied to every type of HMG.

In developing the 10 principles, the workgroup attempted to construct a list of the basic ingredients needed to build and sustain an effective HMG. These 10 principles stand on their own, independent of the 47 key characteristics, and include issues such as effective leadership, clinician engagement, adequate resources, management infrastructure, key hospitalist roles and responsibilities, alignment with the hospital, and the recruitment and retention of qualified hospitalists.

A more detailed version of the Key Principles and Characteristics of an Effective HMG is available in the online version of this article (see Supporting Information, Appendix, in the online version of this article). The online Appendix includes the rationales for each of the characteristics, guidance on how to provide feedback to the SHM on the framework, and the SHM's plan for further development of the key principles and characteristics.

DISCUSSION

To address the variability in capabilities and performance of HMGs, these principles and characteristics are designed to provide a framework for HMGs seeking to conduct self‐assessments and develop pathways for improvement.

Although there may be HMG arrangements that do not directly involve the hospital and its executive team, and therefore alternative approaches may make sense, for most HMGs hospitals are directly involved with the HMG as either an employer or a contractor. For that reason, the Key Principles and Characteristics of an Effective HMG is written for 2 audiences: the executive leadership of the hospital (most specifically the chief medical officer or a similar role) and the hospitalists in the HMG (most specifically the practice medical director). To address the key characteristics requires the active participation of both parties. For the hospital executives, the framework establishes expectations for the HMG. For the hospitalists, the framework provides guidance in the development of an improvement plan.

Hospital executives and hospitalists can use the key characteristics in a broad spectrum of ways. The easiest and least formalized approach would be to use the framework as the basis of an ongoing dialogue between the hospital leadership and the HMG. A more formal approach would be to use the framework to guide the planning and budgeting activities of the HMG. Finally, a hospital or health system can use the key principles and characteristics as a way to evaluate their affiliated HMG(s)for example, the HMG must address 80% of the 47 characteristics.

The Key Principles and Characteristics of an Effective HMG should be considered akin to the Core Competencies in Hospital Medicine previously published in the Journal of Hospital Medicine.[7] However, instead of focusing on the competencies of individual physicians, this framework focuses on the characteristics of hospitalist groups. Just as a physician or other healthcare provider is not expected to demonstrate competency for every element in the core competencies document, an HMG does not need to have all 47 characteristics to be effective. Effective hospitalists may have skills other than those listed in the Core Competencies in Hospital Medicine. Similarly, the 47 characteristics do not represent an exhaustive list of every desirable HMG attribute. In general, effective HMGs should possess most of the characteristics.

In applying the framework, the HMG should not simply attempt to evaluate each characteristic with a yes or no assessment. For HMGs responding yes, there may be a wide range of performancefrom meeting the bare minimum requirements to employing sophisticated, expansive measures to excel in the characteristic.

SHM encourages hospital leaders and HMG leaders to use these characteristics to perform an HMG self‐assessment and to develop a plan. The plan could address implementation of selected characteristics that are not currently being addressed by the HMG or the development of additional behaviors, tools, resources, and capabilities that more fully incorporate those characteristics for which the HMG meets only minimum requirements. In addition, the plan could address the impact that a larger organization (eg, health system, hospital, or employer) may have on a given characteristic.

As outlined above, the process used to develop the Key Principles and Characteristics of an Effective HMG was grounded in expert opinion and extensive review and feedback. HMGs that use the framework should recognize that others might have a different opinion. For example, characteristic 5.2 states, The HMG's compensation model aligns hospitalist incentives with the goals of the hospital and the goals of the hospitalist's employer (if different). There are likely to be experienced hospitalist leaders who believe that an effective HMG does not need to have an incentive compensation system. However, the consensus process employed to develop the key characteristics led to the conclusion that an effective HMG should have an incentive compensation system.

The publication of the Key Principles and Characteristics of an Effective HMG may lead to negative and/or unintended consequences. A self‐assessment by an HMG using this framework could require a significant level of effort on behalf of the HMG, whereas implementing remedial efforts to address the characteristics could require an investment of time and money that could take away from other important issues facing the HMG. Many HMGs may be held accountable for addressing these characteristics without the necessary financial support from their hospital or medical group. Finally, the publication of the document could create a backlash from members of the hospitalist community who do not think that the SHM should be in the business of defining what characterizes an effective HMG, rather that this definition should be left to the marketplace.

Despite these concerns, the leadership of the SHM expects that the publication of the Key Principles and Characteristics of an Effective HMG will lead to overall improvement in the capabilities and performance of HMGs.

CONCLUSIONS

The Key Principles and Characteristics of an Effective HMG have been designed to be aspirational, helping to raise the bar for the specialty of hospital medicine. These principles and characteristics could provide a framework for HMGs seeking to conduct self‐assessments, outlining a pathway for improvement, and better defining the central role of hospitalists in coordinating team‐based, patient‐centered care in the acute care setting.

With the continuing growth of the specialty of hospital medicine, the capabilities and performance of hospital medicine groups (HMGs) varies significantly. There are few guidelines that HMGs can reference as tools to guide self‐improvement. To address this deficiency, the Society of Hospital Medicine (SHM) Board of Directors authorized a process to identify the key principles and characteristics of an effective HMG.

METHODS

Final Approval

The final draft of the 10 principles and 47 characteristics was approved for publication at a meeting of the SHM Board of Directors in September 2013 (Figure 1).

jhm2119-fig-0001-m.png

RESULTS

A recurring issue that the workgroup addressed was the applicability of the characteristics from 1 practice setting to another. Confounding factors include the HMG's employment/organizational model (eg, hospital employed, academic, multispecialty group, private practice, and management company), its population served (eg, adult vs pediatric, more than 1 hospital), and the type of hospital served (eg, academic vs community, the hospital has more than 1 HMG). The workgroup has made an effort to assure that all 47 characteristics can be applied to every type of HMG.

In developing the 10 principles, the workgroup attempted to construct a list of the basic ingredients needed to build and sustain an effective HMG. These 10 principles stand on their own, independent of the 47 key characteristics, and include issues such as effective leadership, clinician engagement, adequate resources, management infrastructure, key hospitalist roles and responsibilities, alignment with the hospital, and the recruitment and retention of qualified hospitalists.

A more detailed version of the Key Principles and Characteristics of an Effective HMG is available in the online version of this article (see Supporting Information, Appendix, in the online version of this article). The online Appendix includes the rationales for each of the characteristics, guidance on how to provide feedback to the SHM on the framework, and the SHM's plan for further development of the key principles and characteristics.

DISCUSSION

To address the variability in capabilities and performance of HMGs, these principles and characteristics are designed to provide a framework for HMGs seeking to conduct self‐assessments and develop pathways for improvement.

Although there may be HMG arrangements that do not directly involve the hospital and its executive team, and therefore alternative approaches may make sense, for most HMGs hospitals are directly involved with the HMG as either an employer or a contractor. For that reason, the Key Principles and Characteristics of an Effective HMG is written for 2 audiences: the executive leadership of the hospital (most specifically the chief medical officer or a similar role) and the hospitalists in the HMG (most specifically the practice medical director). To address the key characteristics requires the active participation of both parties. For the hospital executives, the framework establishes expectations for the HMG. For the hospitalists, the framework provides guidance in the development of an improvement plan.

Hospital executives and hospitalists can use the key characteristics in a broad spectrum of ways. The easiest and least formalized approach would be to use the framework as the basis of an ongoing dialogue between the hospital leadership and the HMG. A more formal approach would be to use the framework to guide the planning and budgeting activities of the HMG. Finally, a hospital or health system can use the key principles and characteristics as a way to evaluate their affiliated HMG(s)for example, the HMG must address 80% of the 47 characteristics.

The Key Principles and Characteristics of an Effective HMG should be considered akin to the Core Competencies in Hospital Medicine previously published in the Journal of Hospital Medicine.[7] However, instead of focusing on the competencies of individual physicians, this framework focuses on the characteristics of hospitalist groups. Just as a physician or other healthcare provider is not expected to demonstrate competency for every element in the core competencies document, an HMG does not need to have all 47 characteristics to be effective. Effective hospitalists may have skills other than those listed in the Core Competencies in Hospital Medicine. Similarly, the 47 characteristics do not represent an exhaustive list of every desirable HMG attribute. In general, effective HMGs should possess most of the characteristics.

In applying the framework, the HMG should not simply attempt to evaluate each characteristic with a yes or no assessment. For HMGs responding yes, there may be a wide range of performancefrom meeting the bare minimum requirements to employing sophisticated, expansive measures to excel in the characteristic.

SHM encourages hospital leaders and HMG leaders to use these characteristics to perform an HMG self‐assessment and to develop a plan. The plan could address implementation of selected characteristics that are not currently being addressed by the HMG or the development of additional behaviors, tools, resources, and capabilities that more fully incorporate those characteristics for which the HMG meets only minimum requirements. In addition, the plan could address the impact that a larger organization (eg, health system, hospital, or employer) may have on a given characteristic.

As outlined above, the process used to develop the Key Principles and Characteristics of an Effective HMG was grounded in expert opinion and extensive review and feedback. HMGs that use the framework should recognize that others might have a different opinion. For example, characteristic 5.2 states, The HMG's compensation model aligns hospitalist incentives with the goals of the hospital and the goals of the hospitalist's employer (if different). There are likely to be experienced hospitalist leaders who believe that an effective HMG does not need to have an incentive compensation system. However, the consensus process employed to develop the key characteristics led to the conclusion that an effective HMG should have an incentive compensation system.

The publication of the Key Principles and Characteristics of an Effective HMG may lead to negative and/or unintended consequences. A self‐assessment by an HMG using this framework could require a significant level of effort on behalf of the HMG, whereas implementing remedial efforts to address the characteristics could require an investment of time and money that could take away from other important issues facing the HMG. Many HMGs may be held accountable for addressing these characteristics without the necessary financial support from their hospital or medical group. Finally, the publication of the document could create a backlash from members of the hospitalist community who do not think that the SHM should be in the business of defining what characterizes an effective HMG, rather that this definition should be left to the marketplace.

Despite these concerns, the leadership of the SHM expects that the publication of the Key Principles and Characteristics of an Effective HMG will lead to overall improvement in the capabilities and performance of HMGs.

CONCLUSIONS

The Key Principles and Characteristics of an Effective HMG have been designed to be aspirational, helping to raise the bar for the specialty of hospital medicine. These principles and characteristics could provide a framework for HMGs seeking to conduct self‐assessments, outlining a pathway for improvement, and better defining the central role of hospitalists in coordinating team‐based, patient‐centered care in the acute care setting.

With the continuing growth of the specialty of hospital medicine, the capabilities and performance of hospital medicine groups (HMGs) varies significantly. There are few guidelines that HMGs can reference as tools to guide self‐improvement. To address this deficiency, the Society of Hospital Medicine (SHM) Board of Directors authorized a process to identify the key principles and characteristics of an effective HMG.

METHODS

Final Approval

The final draft of the 10 principles and 47 characteristics was approved for publication at a meeting of the SHM Board of Directors in September 2013 (Figure 1).

jhm2119-fig-0001-m.png

RESULTS

A recurring issue that the workgroup addressed was the applicability of the characteristics from 1 practice setting to another. Confounding factors include the HMG's employment/organizational model (eg, hospital employed, academic, multispecialty group, private practice, and management company), its population served (eg, adult vs pediatric, more than 1 hospital), and the type of hospital served (eg, academic vs community, the hospital has more than 1 HMG). The workgroup has made an effort to assure that all 47 characteristics can be applied to every type of HMG.

In developing the 10 principles, the workgroup attempted to construct a list of the basic ingredients needed to build and sustain an effective HMG. These 10 principles stand on their own, independent of the 47 key characteristics, and include issues such as effective leadership, clinician engagement, adequate resources, management infrastructure, key hospitalist roles and responsibilities, alignment with the hospital, and the recruitment and retention of qualified hospitalists.

A more detailed version of the Key Principles and Characteristics of an Effective HMG is available in the online version of this article (see Supporting Information, Appendix, in the online version of this article). The online Appendix includes the rationales for each of the characteristics, guidance on how to provide feedback to the SHM on the framework, and the SHM's plan for further development of the key principles and characteristics.

DISCUSSION

To address the variability in capabilities and performance of HMGs, these principles and characteristics are designed to provide a framework for HMGs seeking to conduct self‐assessments and develop pathways for improvement.

Although there may be HMG arrangements that do not directly involve the hospital and its executive team, and therefore alternative approaches may make sense, for most HMGs hospitals are directly involved with the HMG as either an employer or a contractor. For that reason, the Key Principles and Characteristics of an Effective HMG is written for 2 audiences: the executive leadership of the hospital (most specifically the chief medical officer or a similar role) and the hospitalists in the HMG (most specifically the practice medical director). To address the key characteristics requires the active participation of both parties. For the hospital executives, the framework establishes expectations for the HMG. For the hospitalists, the framework provides guidance in the development of an improvement plan.

Hospital executives and hospitalists can use the key characteristics in a broad spectrum of ways. The easiest and least formalized approach would be to use the framework as the basis of an ongoing dialogue between the hospital leadership and the HMG. A more formal approach would be to use the framework to guide the planning and budgeting activities of the HMG. Finally, a hospital or health system can use the key principles and characteristics as a way to evaluate their affiliated HMG(s)for example, the HMG must address 80% of the 47 characteristics.

The Key Principles and Characteristics of an Effective HMG should be considered akin to the Core Competencies in Hospital Medicine previously published in the Journal of Hospital Medicine.[7] However, instead of focusing on the competencies of individual physicians, this framework focuses on the characteristics of hospitalist groups. Just as a physician or other healthcare provider is not expected to demonstrate competency for every element in the core competencies document, an HMG does not need to have all 47 characteristics to be effective. Effective hospitalists may have skills other than those listed in the Core Competencies in Hospital Medicine. Similarly, the 47 characteristics do not represent an exhaustive list of every desirable HMG attribute. In general, effective HMGs should possess most of the characteristics.

In applying the framework, the HMG should not simply attempt to evaluate each characteristic with a yes or no assessment. For HMGs responding yes, there may be a wide range of performancefrom meeting the bare minimum requirements to employing sophisticated, expansive measures to excel in the characteristic.

SHM encourages hospital leaders and HMG leaders to use these characteristics to perform an HMG self‐assessment and to develop a plan. The plan could address implementation of selected characteristics that are not currently being addressed by the HMG or the development of additional behaviors, tools, resources, and capabilities that more fully incorporate those characteristics for which the HMG meets only minimum requirements. In addition, the plan could address the impact that a larger organization (eg, health system, hospital, or employer) may have on a given characteristic.

As outlined above, the process used to develop the Key Principles and Characteristics of an Effective HMG was grounded in expert opinion and extensive review and feedback. HMGs that use the framework should recognize that others might have a different opinion. For example, characteristic 5.2 states, The HMG's compensation model aligns hospitalist incentives with the goals of the hospital and the goals of the hospitalist's employer (if different). There are likely to be experienced hospitalist leaders who believe that an effective HMG does not need to have an incentive compensation system. However, the consensus process employed to develop the key characteristics led to the conclusion that an effective HMG should have an incentive compensation system.

The publication of the Key Principles and Characteristics of an Effective HMG may lead to negative and/or unintended consequences. A self‐assessment by an HMG using this framework could require a significant level of effort on behalf of the HMG, whereas implementing remedial efforts to address the characteristics could require an investment of time and money that could take away from other important issues facing the HMG. Many HMGs may be held accountable for addressing these characteristics without the necessary financial support from their hospital or medical group. Finally, the publication of the document could create a backlash from members of the hospitalist community who do not think that the SHM should be in the business of defining what characterizes an effective HMG, rather that this definition should be left to the marketplace.

Despite these concerns, the leadership of the SHM expects that the publication of the Key Principles and Characteristics of an Effective HMG will lead to overall improvement in the capabilities and performance of HMGs.

CONCLUSIONS

The Key Principles and Characteristics of an Effective HMG have been designed to be aspirational, helping to raise the bar for the specialty of hospital medicine. These principles and characteristics could provide a framework for HMGs seeking to conduct self‐assessments, outlining a pathway for improvement, and better defining the central role of hospitalists in coordinating team‐based, patient‐centered care in the acute care setting.

The concept of improved inpatient diabetes control has been gaining attention in hospitals nationwide as a mechanism for improving patient outcomes, decreasing readmission rates, reducing cost of care, and shortening hospital length of stay.14 The growing recognition that glycemic control is a critical element of inpatient care has prompted several national agencies, including the National Quality Forum (NQF), University Health System Consortium (UHC), Centers for Medicare and Medicaid Services (CMS), and the Joint Commission (JC) to make inpatient diabetes control a focus of quality improvement efforts and outcomes tracking.1 There is a national trend toward the use of intravenous insulin infusion for tight glycemic control of stress‐induced hyperglycemia in postoperative intensive care unit (ICU) and medical ICU patients.5, 6 Consequently, there is a need for the development of a standardized approach for performance evaluation of subcutaneous and intravenous insulin protocols, while ensuring patient safety issues. The analysis of glucose outcomes is based on the systematic analysis of blood glucose (BG) performance metrics known as glucometrics.7, 8 This has provided a means to measure the success of hospital quality improvement programs over time.

The 2008 American Diabetes Association (ADA) Clinical Practice Recommendations endorse BG goals for the critically ill to be maintained as close as possible to 110 mg/dL (6.1 mmol/L) and generally <140 mg/dL (7.8 mmol/L).2 The American Association of Clinical Endocrinologists/The American College of Endocrinology guidelines recommend for ICU care BG in the range of 80110 mg/dL.1, 4 Regarding the non‐critically ill patients, the ADA recommends targets for fasting BG of <126 mg/dL (7.0 mmol/L) and all random BG 180200 mg/dL (1011.1 mmol/L).2 A limitation for these BG goals is hypoglycemia; the ADA endorses that hospitals try to achieve these lower BG values through quality improvement initiatives devised to systematically and safely reduce the BG targets.2

Materials and Methods

The Medical University of South Carolina (MUSC) is a 709‐bed tertiary‐care medical/surgical center located in Charleston, South Carolina. The medical center consists of 6 adult ICUs: medical intensive care unit, coronary care unit, cardiothoracic intensive care unit, neurosurgical intensive care unit, neurosurgical trauma intensive care unit, and surgical trauma intensive care unit. Overall, 14% of patients are in the ICUs, and 86% of patients are on the wards. MUSC has an extensive referral network including neighboring hospitals, rehabilitation centers, outpatient specialty treatment and imaging centers, and doctors' offices.

Intravenous Insulin Protocol

The HDTF initially reviewed 15 evidence‐based protocols and identified 5 desirable protocol characteristics. These characteristics included easy physician ordering (requiring only a signature), ability to quickly reach and maintain a BG target range, minimal risk for hypoglycemic events, adaptability for use anywhere in the hospital setting, and acceptance and implementation by nursing staff.16

The IV protocol, a web‐based calculator (Figure 1), was developed based on the concept of the multiplier by White et al.17 For this protocol, the IV infusion (IVI) rate is changed based on a formula that uses a multiplier (a surrogate for insulin sensitivity factor) and the difference between measured BG and target blood glucose (TBG). The calculator uses the following mathematical formula: rate of insulin infusion/hour = (current BG 60 mg/dL) 0.03.18, 19 Additionally, the protocol requires that enough insulin be infused to address severe hyperglycemia at initiation with a rapid reduction in the insulin infusion rate as BG normalizes. The protocol also permits an adjustment of the insulin rate by tenths of a unit per hour to maintain the BG in the center of the target range. The main variant of this protocol is the value of the starting multiplier. The web‐based calculator is currently being used in all 6 adult ICUs and on all of the adult medical‐surgical floors at MUSC.

mfig001.jpg

In early 2006, all adult ICUs were using our in‐house, web‐based intravenous insulin infusion calculator (IVIIC), which prompted more BG readings with intensification of insulin drip use.19 Specifically, initial monitoring for the IVIIC included BG readings every 24 hours. To avoid hypoglycemic events from occurring with the intensification of BG readings for the IVIIC, the BG monitoring frequency was increased to every hour in July 2006. Initial treatment for hypoglycemia was D50% (12.525 g), which tended to overcorrect BG. In July 2006, we revised the protocol using aliquots of D50% specific to the BG reading.19 This action has resulted in decreasing the glycemic excursions observed due to overcorrection of hypoglycemia.

BG target ranges to match the level of care are as follows: intensive care unit (80110 mg/dL); labor and delivery (70110 mg/dL); adult medical/surgical floors (80140 mg/dL); diabetic ketoacidosis (DKA)/hyperosmolar nonketotic coma (HHNK) (150200 mg/dL); neurosurgery ICU (90120 mg/dL); and perioperative patients (140180 mg/dL).20 These BG targets were created to satisfy the clinical requests of specific departments at MUSC. We have restricted starting the multiplier for DKA/HHNK at 0.01, to affect a slower rate of change and the multiplier for all others is set at 0.03.

Results

The baseline demographic characteristics of the four study groups are shown in Table 3. The four groups were found to be similar for gender distribution, mean age, and racial distribution. There were significant differences observed among hospital stay characteristics, insulin drip use, history of diabetes, ventilator support, kidney failure, dialysis, total parenteral nutrition (TPN), and red blood cell (RBC) transfusions. Overall, insulin drip use tended to increase over time. The percentage of patients with diabetes on admission or diagnosed during admission tended to decrease over time. This was likely due to an increase in the diagnosis and treatment of stress/steroid‐induced hyperglycemia during the hospital stay.

A total of 11,715 patient‐days, consisting of 56,401 individual BG readings obtained from 2,215 unique patients, were distributed across the 4 years. Table 4 presents the year‐specific patient‐day analysis. While the prevalence of mild (BG 5069 mg/dL) hypoglycemia was found to increase over the years studied (P < 0.01), the percentage of patient‐days with a mean BG in the range of 70180 mg/dL increased over the period of study (P < 0.01). The total hypoglycemia events <60 mg/dL are presented as comparative data to other studies.7 The percent of patient days with at least one BG < 70 mg/dL (reported in Table 4 as mild events) ranged from 3.72 in 2005 to as high as 10.71 in 2007; however, approximately one‐half of the hypoglycemic events are attributable to readings from BG 6069 since the proportion of patient days with a BG < 60 mg/dL was approximately one‐half that for BG < 70 mg/dL (Table 4). The prevalence of patient days with at least one moderate (BG 4049 mg/dL) or severe (BG < 40 mg/dL) hypoglycemia event was not found to increase in a linear manner. There was a statistical trend for potentially nonlinear relationship of year with moderate hypoglycemia and hyperglycemia.

Immediately following the implementation (year 2005), post hoc comparisons suggested that the rate of moderate hypoglycemia was lowest relative to the 3 other years, but no other statistical differences were observed. The year 2005 also had the lowest proportion of patient days with at least 1 hyperglycemic event.

The individual BG readings for the 2215 unique patients were also individually analyzed according to the methods of Goldberg et al.7 Even though no statistical tests were performed at the ward level, the descriptive data presented in Table 5 are consistent with the analysis of the patient‐day data. Several important features of the data are illustrated by Table 5. Most notably, the glycemic control at the hospital level is improved. The percentage of BG readings in the range of 70180 mg/dL increased annually whereas the mean BG values, the coefficient of variation, and the interquartile range (IQR) decreased annually.

Conclusions

Collectively, we have shown that implementing standardized insulin order sets including hypoglycemia, SC insulin, IV insulin, and IV to SC insulin transition treatment protocols at MUSC may generate the expected benefits for patient safety for this population of patients. The primary hypothesis that the rate of hypoglycemia and hyperglycemia would be lower after the implementation of these protocols was supported by the data, because the overall blood glucose control was markedly improved as a result of the protocols. However, the effect was strongest in 2005 (immediately following the protocol's implementation) and appeared to diminish some with time.

There were several other quality improvement measures initiated at MUSC that likely contributed to the decreasing rates of hypoglycemia and hyperglycemia. For example, comparing June 2004 with June 2007, the number of patients tested increased from 434 to 683. This increase could be attributed, in part, to a trend on medical/surgical services toward an increased focus on glucose monitoring.

When intensive glycemic control programs are implemented, hospitals should have a standardized, nurse‐driven hypoglycemia protocol.11 The success of such a hypoglycemia treatment protocol is demonstrated by the improvement observed at MUSC since the protocol was first implemented in October 2004.22

There are limitations that warrant consideration. A key limitation is that other procedural changes may have occurred during the years of study. Because the initial focus of the HDTF was to reduce hypoglycemic and hyperglycemic events, a multipronged approach was used, beginning with the treatment protocol but followed by other changes. These changes, while unmeasured in the current study, could have influenced the rate of hypoglycemia and hyperglycemia. Therefore, although the protocol that we developed has sound theoretical underpinnings, the improvement in glycemic control at other hospitals may vary. Second, because this was initially regarded as a quality improvement project for hospitalized patients with hypoglycemia and hyperglycemia, we did not evaluate morbidity, mortality, or other clinical outcome data other than BG targets and incidences of hypoglycemia and hyperglycemia. Third, there was no concurrent control group established for this study, rather the study used a retrospective, nonrandomized design with a historical control. As previously mentioned, we cannot rule out the idea that other changes occurred between the preprotocol and postprotocol interval to influence our results. Finally, there are statistical limitations to the research.

One limitation regarding the analysis of the BG data was the potential for an increased type I error (ie, false‐positive result) due to clustering of BG values within a patient and increased monitoring frequencies when a hypoglycemic or hyperglycemic event was observed. The generalized estimating equations directly addressed the first concern. In particular, the effective sample size for each participant was a function of the number of patient‐days and the correlation of patient‐day summaries. Therefore, patients with several highly‐correlated outcomes would contribute less to the analysis than other patients with the same number of patient‐days that were correlated to a lesser extent. As for the second concern, the patient‐day frequencies alleviate this problem and avoid the length‐of‐stay bias associated with a patient‐level (or patient‐stay) analysis. Power was less than planned due in part to the use of the patient‐day analysis instead of the originally designed ward‐level analysis. The change in the statistical design was a response to emerging evidence in the literature.7

In conclusion, the hypothesis that MUSC patients benefit from the use of standardized insulin order sets, hypoglycemia, and hyperglycemia treatment protocols, is supported by the data collected in this study. Because it has been recommended that a hypoglycemia and hyperglycemia prevention protocol as well as a hypoglycemia and hyperglycemia treatment protocol be in place, the HDTF will be focusing on the actual prevention of the hypoglycemic and hyperglycemic incidents occurring in the first place.2, 25 This may result in further reductions of hypoglycemic and hyperglycemic events. We have recently implemented hypoglycemia and hyperglycemia prevention policies at MUSC.

The concept of improved inpatient diabetes control has been gaining attention in hospitals nationwide as a mechanism for improving patient outcomes, decreasing readmission rates, reducing cost of care, and shortening hospital length of stay.14 The growing recognition that glycemic control is a critical element of inpatient care has prompted several national agencies, including the National Quality Forum (NQF), University Health System Consortium (UHC), Centers for Medicare and Medicaid Services (CMS), and the Joint Commission (JC) to make inpatient diabetes control a focus of quality improvement efforts and outcomes tracking.1 There is a national trend toward the use of intravenous insulin infusion for tight glycemic control of stress‐induced hyperglycemia in postoperative intensive care unit (ICU) and medical ICU patients.5, 6 Consequently, there is a need for the development of a standardized approach for performance evaluation of subcutaneous and intravenous insulin protocols, while ensuring patient safety issues. The analysis of glucose outcomes is based on the systematic analysis of blood glucose (BG) performance metrics known as glucometrics.7, 8 This has provided a means to measure the success of hospital quality improvement programs over time.

The 2008 American Diabetes Association (ADA) Clinical Practice Recommendations endorse BG goals for the critically ill to be maintained as close as possible to 110 mg/dL (6.1 mmol/L) and generally <140 mg/dL (7.8 mmol/L).2 The American Association of Clinical Endocrinologists/The American College of Endocrinology guidelines recommend for ICU care BG in the range of 80110 mg/dL.1, 4 Regarding the non‐critically ill patients, the ADA recommends targets for fasting BG of <126 mg/dL (7.0 mmol/L) and all random BG 180200 mg/dL (1011.1 mmol/L).2 A limitation for these BG goals is hypoglycemia; the ADA endorses that hospitals try to achieve these lower BG values through quality improvement initiatives devised to systematically and safely reduce the BG targets.2

Materials and Methods

The Medical University of South Carolina (MUSC) is a 709‐bed tertiary‐care medical/surgical center located in Charleston, South Carolina. The medical center consists of 6 adult ICUs: medical intensive care unit, coronary care unit, cardiothoracic intensive care unit, neurosurgical intensive care unit, neurosurgical trauma intensive care unit, and surgical trauma intensive care unit. Overall, 14% of patients are in the ICUs, and 86% of patients are on the wards. MUSC has an extensive referral network including neighboring hospitals, rehabilitation centers, outpatient specialty treatment and imaging centers, and doctors' offices.

Intravenous Insulin Protocol

The HDTF initially reviewed 15 evidence‐based protocols and identified 5 desirable protocol characteristics. These characteristics included easy physician ordering (requiring only a signature), ability to quickly reach and maintain a BG target range, minimal risk for hypoglycemic events, adaptability for use anywhere in the hospital setting, and acceptance and implementation by nursing staff.16

The IV protocol, a web‐based calculator (Figure 1), was developed based on the concept of the multiplier by White et al.17 For this protocol, the IV infusion (IVI) rate is changed based on a formula that uses a multiplier (a surrogate for insulin sensitivity factor) and the difference between measured BG and target blood glucose (TBG). The calculator uses the following mathematical formula: rate of insulin infusion/hour = (current BG 60 mg/dL) 0.03.18, 19 Additionally, the protocol requires that enough insulin be infused to address severe hyperglycemia at initiation with a rapid reduction in the insulin infusion rate as BG normalizes. The protocol also permits an adjustment of the insulin rate by tenths of a unit per hour to maintain the BG in the center of the target range. The main variant of this protocol is the value of the starting multiplier. The web‐based calculator is currently being used in all 6 adult ICUs and on all of the adult medical‐surgical floors at MUSC.

mfig001.jpg

In early 2006, all adult ICUs were using our in‐house, web‐based intravenous insulin infusion calculator (IVIIC), which prompted more BG readings with intensification of insulin drip use.19 Specifically, initial monitoring for the IVIIC included BG readings every 24 hours. To avoid hypoglycemic events from occurring with the intensification of BG readings for the IVIIC, the BG monitoring frequency was increased to every hour in July 2006. Initial treatment for hypoglycemia was D50% (12.525 g), which tended to overcorrect BG. In July 2006, we revised the protocol using aliquots of D50% specific to the BG reading.19 This action has resulted in decreasing the glycemic excursions observed due to overcorrection of hypoglycemia.

BG target ranges to match the level of care are as follows: intensive care unit (80110 mg/dL); labor and delivery (70110 mg/dL); adult medical/surgical floors (80140 mg/dL); diabetic ketoacidosis (DKA)/hyperosmolar nonketotic coma (HHNK) (150200 mg/dL); neurosurgery ICU (90120 mg/dL); and perioperative patients (140180 mg/dL).20 These BG targets were created to satisfy the clinical requests of specific departments at MUSC. We have restricted starting the multiplier for DKA/HHNK at 0.01, to affect a slower rate of change and the multiplier for all others is set at 0.03.

Results

The baseline demographic characteristics of the four study groups are shown in Table 3. The four groups were found to be similar for gender distribution, mean age, and racial distribution. There were significant differences observed among hospital stay characteristics, insulin drip use, history of diabetes, ventilator support, kidney failure, dialysis, total parenteral nutrition (TPN), and red blood cell (RBC) transfusions. Overall, insulin drip use tended to increase over time. The percentage of patients with diabetes on admission or diagnosed during admission tended to decrease over time. This was likely due to an increase in the diagnosis and treatment of stress/steroid‐induced hyperglycemia during the hospital stay.

A total of 11,715 patient‐days, consisting of 56,401 individual BG readings obtained from 2,215 unique patients, were distributed across the 4 years. Table 4 presents the year‐specific patient‐day analysis. While the prevalence of mild (BG 5069 mg/dL) hypoglycemia was found to increase over the years studied (P < 0.01), the percentage of patient‐days with a mean BG in the range of 70180 mg/dL increased over the period of study (P < 0.01). The total hypoglycemia events <60 mg/dL are presented as comparative data to other studies.7 The percent of patient days with at least one BG < 70 mg/dL (reported in Table 4 as mild events) ranged from 3.72 in 2005 to as high as 10.71 in 2007; however, approximately one‐half of the hypoglycemic events are attributable to readings from BG 6069 since the proportion of patient days with a BG < 60 mg/dL was approximately one‐half that for BG < 70 mg/dL (Table 4). The prevalence of patient days with at least one moderate (BG 4049 mg/dL) or severe (BG < 40 mg/dL) hypoglycemia event was not found to increase in a linear manner. There was a statistical trend for potentially nonlinear relationship of year with moderate hypoglycemia and hyperglycemia.

Immediately following the implementation (year 2005), post hoc comparisons suggested that the rate of moderate hypoglycemia was lowest relative to the 3 other years, but no other statistical differences were observed. The year 2005 also had the lowest proportion of patient days with at least 1 hyperglycemic event.

The individual BG readings for the 2215 unique patients were also individually analyzed according to the methods of Goldberg et al.7 Even though no statistical tests were performed at the ward level, the descriptive data presented in Table 5 are consistent with the analysis of the patient‐day data. Several important features of the data are illustrated by Table 5. Most notably, the glycemic control at the hospital level is improved. The percentage of BG readings in the range of 70180 mg/dL increased annually whereas the mean BG values, the coefficient of variation, and the interquartile range (IQR) decreased annually.

Conclusions

Collectively, we have shown that implementing standardized insulin order sets including hypoglycemia, SC insulin, IV insulin, and IV to SC insulin transition treatment protocols at MUSC may generate the expected benefits for patient safety for this population of patients. The primary hypothesis that the rate of hypoglycemia and hyperglycemia would be lower after the implementation of these protocols was supported by the data, because the overall blood glucose control was markedly improved as a result of the protocols. However, the effect was strongest in 2005 (immediately following the protocol's implementation) and appeared to diminish some with time.

There were several other quality improvement measures initiated at MUSC that likely contributed to the decreasing rates of hypoglycemia and hyperglycemia. For example, comparing June 2004 with June 2007, the number of patients tested increased from 434 to 683. This increase could be attributed, in part, to a trend on medical/surgical services toward an increased focus on glucose monitoring.

When intensive glycemic control programs are implemented, hospitals should have a standardized, nurse‐driven hypoglycemia protocol.11 The success of such a hypoglycemia treatment protocol is demonstrated by the improvement observed at MUSC since the protocol was first implemented in October 2004.22

There are limitations that warrant consideration. A key limitation is that other procedural changes may have occurred during the years of study. Because the initial focus of the HDTF was to reduce hypoglycemic and hyperglycemic events, a multipronged approach was used, beginning with the treatment protocol but followed by other changes. These changes, while unmeasured in the current study, could have influenced the rate of hypoglycemia and hyperglycemia. Therefore, although the protocol that we developed has sound theoretical underpinnings, the improvement in glycemic control at other hospitals may vary. Second, because this was initially regarded as a quality improvement project for hospitalized patients with hypoglycemia and hyperglycemia, we did not evaluate morbidity, mortality, or other clinical outcome data other than BG targets and incidences of hypoglycemia and hyperglycemia. Third, there was no concurrent control group established for this study, rather the study used a retrospective, nonrandomized design with a historical control. As previously mentioned, we cannot rule out the idea that other changes occurred between the preprotocol and postprotocol interval to influence our results. Finally, there are statistical limitations to the research.

One limitation regarding the analysis of the BG data was the potential for an increased type I error (ie, false‐positive result) due to clustering of BG values within a patient and increased monitoring frequencies when a hypoglycemic or hyperglycemic event was observed. The generalized estimating equations directly addressed the first concern. In particular, the effective sample size for each participant was a function of the number of patient‐days and the correlation of patient‐day summaries. Therefore, patients with several highly‐correlated outcomes would contribute less to the analysis than other patients with the same number of patient‐days that were correlated to a lesser extent. As for the second concern, the patient‐day frequencies alleviate this problem and avoid the length‐of‐stay bias associated with a patient‐level (or patient‐stay) analysis. Power was less than planned due in part to the use of the patient‐day analysis instead of the originally designed ward‐level analysis. The change in the statistical design was a response to emerging evidence in the literature.7

In conclusion, the hypothesis that MUSC patients benefit from the use of standardized insulin order sets, hypoglycemia, and hyperglycemia treatment protocols, is supported by the data collected in this study. Because it has been recommended that a hypoglycemia and hyperglycemia prevention protocol as well as a hypoglycemia and hyperglycemia treatment protocol be in place, the HDTF will be focusing on the actual prevention of the hypoglycemic and hyperglycemic incidents occurring in the first place.2, 25 This may result in further reductions of hypoglycemic and hyperglycemic events. We have recently implemented hypoglycemia and hyperglycemia prevention policies at MUSC.

The concept of improved inpatient diabetes control has been gaining attention in hospitals nationwide as a mechanism for improving patient outcomes, decreasing readmission rates, reducing cost of care, and shortening hospital length of stay.14 The growing recognition that glycemic control is a critical element of inpatient care has prompted several national agencies, including the National Quality Forum (NQF), University Health System Consortium (UHC), Centers for Medicare and Medicaid Services (CMS), and the Joint Commission (JC) to make inpatient diabetes control a focus of quality improvement efforts and outcomes tracking.1 There is a national trend toward the use of intravenous insulin infusion for tight glycemic control of stress‐induced hyperglycemia in postoperative intensive care unit (ICU) and medical ICU patients.5, 6 Consequently, there is a need for the development of a standardized approach for performance evaluation of subcutaneous and intravenous insulin protocols, while ensuring patient safety issues. The analysis of glucose outcomes is based on the systematic analysis of blood glucose (BG) performance metrics known as glucometrics.7, 8 This has provided a means to measure the success of hospital quality improvement programs over time.

The 2008 American Diabetes Association (ADA) Clinical Practice Recommendations endorse BG goals for the critically ill to be maintained as close as possible to 110 mg/dL (6.1 mmol/L) and generally <140 mg/dL (7.8 mmol/L).2 The American Association of Clinical Endocrinologists/The American College of Endocrinology guidelines recommend for ICU care BG in the range of 80110 mg/dL.1, 4 Regarding the non‐critically ill patients, the ADA recommends targets for fasting BG of <126 mg/dL (7.0 mmol/L) and all random BG 180200 mg/dL (1011.1 mmol/L).2 A limitation for these BG goals is hypoglycemia; the ADA endorses that hospitals try to achieve these lower BG values through quality improvement initiatives devised to systematically and safely reduce the BG targets.2

Materials and Methods

The Medical University of South Carolina (MUSC) is a 709‐bed tertiary‐care medical/surgical center located in Charleston, South Carolina. The medical center consists of 6 adult ICUs: medical intensive care unit, coronary care unit, cardiothoracic intensive care unit, neurosurgical intensive care unit, neurosurgical trauma intensive care unit, and surgical trauma intensive care unit. Overall, 14% of patients are in the ICUs, and 86% of patients are on the wards. MUSC has an extensive referral network including neighboring hospitals, rehabilitation centers, outpatient specialty treatment and imaging centers, and doctors' offices.

Intravenous Insulin Protocol

The HDTF initially reviewed 15 evidence‐based protocols and identified 5 desirable protocol characteristics. These characteristics included easy physician ordering (requiring only a signature), ability to quickly reach and maintain a BG target range, minimal risk for hypoglycemic events, adaptability for use anywhere in the hospital setting, and acceptance and implementation by nursing staff.16

The IV protocol, a web‐based calculator (Figure 1), was developed based on the concept of the multiplier by White et al.17 For this protocol, the IV infusion (IVI) rate is changed based on a formula that uses a multiplier (a surrogate for insulin sensitivity factor) and the difference between measured BG and target blood glucose (TBG). The calculator uses the following mathematical formula: rate of insulin infusion/hour = (current BG 60 mg/dL) 0.03.18, 19 Additionally, the protocol requires that enough insulin be infused to address severe hyperglycemia at initiation with a rapid reduction in the insulin infusion rate as BG normalizes. The protocol also permits an adjustment of the insulin rate by tenths of a unit per hour to maintain the BG in the center of the target range. The main variant of this protocol is the value of the starting multiplier. The web‐based calculator is currently being used in all 6 adult ICUs and on all of the adult medical‐surgical floors at MUSC.

mfig001.jpg

In early 2006, all adult ICUs were using our in‐house, web‐based intravenous insulin infusion calculator (IVIIC), which prompted more BG readings with intensification of insulin drip use.19 Specifically, initial monitoring for the IVIIC included BG readings every 24 hours. To avoid hypoglycemic events from occurring with the intensification of BG readings for the IVIIC, the BG monitoring frequency was increased to every hour in July 2006. Initial treatment for hypoglycemia was D50% (12.525 g), which tended to overcorrect BG. In July 2006, we revised the protocol using aliquots of D50% specific to the BG reading.19 This action has resulted in decreasing the glycemic excursions observed due to overcorrection of hypoglycemia.

BG target ranges to match the level of care are as follows: intensive care unit (80110 mg/dL); labor and delivery (70110 mg/dL); adult medical/surgical floors (80140 mg/dL); diabetic ketoacidosis (DKA)/hyperosmolar nonketotic coma (HHNK) (150200 mg/dL); neurosurgery ICU (90120 mg/dL); and perioperative patients (140180 mg/dL).20 These BG targets were created to satisfy the clinical requests of specific departments at MUSC. We have restricted starting the multiplier for DKA/HHNK at 0.01, to affect a slower rate of change and the multiplier for all others is set at 0.03.

Results

The baseline demographic characteristics of the four study groups are shown in Table 3. The four groups were found to be similar for gender distribution, mean age, and racial distribution. There were significant differences observed among hospital stay characteristics, insulin drip use, history of diabetes, ventilator support, kidney failure, dialysis, total parenteral nutrition (TPN), and red blood cell (RBC) transfusions. Overall, insulin drip use tended to increase over time. The percentage of patients with diabetes on admission or diagnosed during admission tended to decrease over time. This was likely due to an increase in the diagnosis and treatment of stress/steroid‐induced hyperglycemia during the hospital stay.

A total of 11,715 patient‐days, consisting of 56,401 individual BG readings obtained from 2,215 unique patients, were distributed across the 4 years. Table 4 presents the year‐specific patient‐day analysis. While the prevalence of mild (BG 5069 mg/dL) hypoglycemia was found to increase over the years studied (P < 0.01), the percentage of patient‐days with a mean BG in the range of 70180 mg/dL increased over the period of study (P < 0.01). The total hypoglycemia events <60 mg/dL are presented as comparative data to other studies.7 The percent of patient days with at least one BG < 70 mg/dL (reported in Table 4 as mild events) ranged from 3.72 in 2005 to as high as 10.71 in 2007; however, approximately one‐half of the hypoglycemic events are attributable to readings from BG 6069 since the proportion of patient days with a BG < 60 mg/dL was approximately one‐half that for BG < 70 mg/dL (Table 4). The prevalence of patient days with at least one moderate (BG 4049 mg/dL) or severe (BG < 40 mg/dL) hypoglycemia event was not found to increase in a linear manner. There was a statistical trend for potentially nonlinear relationship of year with moderate hypoglycemia and hyperglycemia.

Immediately following the implementation (year 2005), post hoc comparisons suggested that the rate of moderate hypoglycemia was lowest relative to the 3 other years, but no other statistical differences were observed. The year 2005 also had the lowest proportion of patient days with at least 1 hyperglycemic event.

The individual BG readings for the 2215 unique patients were also individually analyzed according to the methods of Goldberg et al.7 Even though no statistical tests were performed at the ward level, the descriptive data presented in Table 5 are consistent with the analysis of the patient‐day data. Several important features of the data are illustrated by Table 5. Most notably, the glycemic control at the hospital level is improved. The percentage of BG readings in the range of 70180 mg/dL increased annually whereas the mean BG values, the coefficient of variation, and the interquartile range (IQR) decreased annually.

Conclusions

Collectively, we have shown that implementing standardized insulin order sets including hypoglycemia, SC insulin, IV insulin, and IV to SC insulin transition treatment protocols at MUSC may generate the expected benefits for patient safety for this population of patients. The primary hypothesis that the rate of hypoglycemia and hyperglycemia would be lower after the implementation of these protocols was supported by the data, because the overall blood glucose control was markedly improved as a result of the protocols. However, the effect was strongest in 2005 (immediately following the protocol's implementation) and appeared to diminish some with time.

There were several other quality improvement measures initiated at MUSC that likely contributed to the decreasing rates of hypoglycemia and hyperglycemia. For example, comparing June 2004 with June 2007, the number of patients tested increased from 434 to 683. This increase could be attributed, in part, to a trend on medical/surgical services toward an increased focus on glucose monitoring.

When intensive glycemic control programs are implemented, hospitals should have a standardized, nurse‐driven hypoglycemia protocol.11 The success of such a hypoglycemia treatment protocol is demonstrated by the improvement observed at MUSC since the protocol was first implemented in October 2004.22

There are limitations that warrant consideration. A key limitation is that other procedural changes may have occurred during the years of study. Because the initial focus of the HDTF was to reduce hypoglycemic and hyperglycemic events, a multipronged approach was used, beginning with the treatment protocol but followed by other changes. These changes, while unmeasured in the current study, could have influenced the rate of hypoglycemia and hyperglycemia. Therefore, although the protocol that we developed has sound theoretical underpinnings, the improvement in glycemic control at other hospitals may vary. Second, because this was initially regarded as a quality improvement project for hospitalized patients with hypoglycemia and hyperglycemia, we did not evaluate morbidity, mortality, or other clinical outcome data other than BG targets and incidences of hypoglycemia and hyperglycemia. Third, there was no concurrent control group established for this study, rather the study used a retrospective, nonrandomized design with a historical control. As previously mentioned, we cannot rule out the idea that other changes occurred between the preprotocol and postprotocol interval to influence our results. Finally, there are statistical limitations to the research.

One limitation regarding the analysis of the BG data was the potential for an increased type I error (ie, false‐positive result) due to clustering of BG values within a patient and increased monitoring frequencies when a hypoglycemic or hyperglycemic event was observed. The generalized estimating equations directly addressed the first concern. In particular, the effective sample size for each participant was a function of the number of patient‐days and the correlation of patient‐day summaries. Therefore, patients with several highly‐correlated outcomes would contribute less to the analysis than other patients with the same number of patient‐days that were correlated to a lesser extent. As for the second concern, the patient‐day frequencies alleviate this problem and avoid the length‐of‐stay bias associated with a patient‐level (or patient‐stay) analysis. Power was less than planned due in part to the use of the patient‐day analysis instead of the originally designed ward‐level analysis. The change in the statistical design was a response to emerging evidence in the literature.7

In conclusion, the hypothesis that MUSC patients benefit from the use of standardized insulin order sets, hypoglycemia, and hyperglycemia treatment protocols, is supported by the data collected in this study. Because it has been recommended that a hypoglycemia and hyperglycemia prevention protocol as well as a hypoglycemia and hyperglycemia treatment protocol be in place, the HDTF will be focusing on the actual prevention of the hypoglycemic and hyperglycemic incidents occurring in the first place.2, 25 This may result in further reductions of hypoglycemic and hyperglycemic events. We have recently implemented hypoglycemia and hyperglycemia prevention policies at MUSC.