Hyperglycaemia in the hospital setting is associated with increased patient morbidity and mortality.1^–6 Control of blood glucose in the hospital setting improves patient outcomes and reduces hospital costs.1–3 7 Insulin is the most effective therapy for achieving prompt glycaemic control in the hospital. However, insulin has been listed as one of the top five high-risk medications (ie, likely to cause injury when misused) by the Joint Commission on Accreditation of Healthcare Organizations.8

Appropriate and safe use of insulin in the hospital is complicated by lack of familiarity with the different insulin preparations on the part of prescribers, the need for contingencies in insulin dosing for variability in glycaemia or nutritional status, the need for combination insulin regimens to achieve glycaemic control and the risk for hypoglycaemia.9^–11 It is not uncommon for home diabetes medications to be discontinued on admission and for sliding scale insulin (SSI) to be the sole form of hyperglycaemic management.10^–12 Despite the frequency with which SSI is used in the hospital setting, there is evidence demonstrating that SSI is ineffective at achieving and maintaining desired levels of glycaemic control in the majority of hospitalised patients with diabetes and hyperglycaemia.11 An overdependence on SSI insulin together with a lack of provider ownership, poor understanding of how to manage diabetes and inadequate monitoring contribute to uncontrolled blood glucose in the hospital setting.1314

As diabetes therapies intensify and new medications enter clinical practice, hospitals can expect to see increasingly complex diabetes regimens for inpatients. As diabetes medication orders in the hospital become more complex, systems that improve the clarity of these schedules can help minimise risk of patient injury.15 The creation of guidelines and order sets that guide the ordering of insulin as basal, bolus and correctional therapies been recommended by leading diabetes groups.716^–20

Assessment of problems

Medication errors related to the prescribing, transcription and administration of insulin are commonly reported at our institution. In fact, insulin is usually either the first or second most common drug class involved in reported medication errors each month. In 2006, 142 errors attributed to insulin therapy and eight attributed to oral hypoglycaemic agents were submitted to the MEDMARX National Data Base for Medication Errors (https://www.medmarx.com). Of the 113 errors that were reviewed for cause, 62% were attributed to administration and 22% were due to incorrect transcribing of an order.

Observed prescribing errors included reliance on SSI as standalone therapy, failure to continue basal insulin following admission to hospital in patients requiring insulin, failure to designate type of insulin being ordered, and lack of insulin adjustment during changes in nutritional status. Administration errors typically relate to variability in the timing of an insulin injection relative to meals. All types of insulin error have contributed to events of hypoglycaemia and/or hyperglycaemia. Transcription errors include misinterpretation of the abbreviation “u” for units, which can result in administration of an incorrect dose if the error goes undetected by a nurse or pharmacist. It is highly likely that the frequency of errors due to insulin and oral agents is higher than what is reported as these reports are highly dependent on individual behaviour and willingness to submit a report.2122

To address these issues, the multidisciplinary Diabetes Patient Safety Committee (DPSC) adopted a phased approach to improving the use of insulin in the hospital by introducing a standardised diabetes oral medication order form and a subcutaneous insulin order form (see supplementary figs 1 and 2 online, respectively).

METHODS

This intervention took place in a 716-bed tertiary care academic medical centre comprising 26 non-critical care units and 13 critical care units. A multidisciplinary DPSC was convened at this institution in 2001 at the request of the hospital patient safety committee to address a series of adverse events related to uncontrolled glucose levels.23 The initial task of this committee was the development and implementation of a hypoglycaemia treatment protocol (HTP) for any patient with blood glucose <3.9 mmol/l (<70 mg/dl).23 By focusing on prevention of severe hypoglycaemia, the committee identified and addressed areas related to inpatient management which were contributing to uncontrolled glucose levels in the hospital.

An early step in the evolution toward standardised order sets for inpatients with diabetes focused on standardising orders for SSI and developing guidelines for judicious use of SSI. Following implementation of the SSI protocol, the incidence of prescribing errors and hyperglycaemia declined.24 Although this was an important first step, the problems of SSI monotherapy and discontinuation of home diabetes medications continued.

By using chart review of patients with blood glucose values >16.7 mmol/l (>300 mg/dl) or <2.2 mmol/l (<40 mg/dl), the DPSC identified the following system barriers to effective ordering of scheduled insulin therapy:

• absence of a section for use of rapid-acting insulin analogues on the SSI form;

• absence of a very low dose scale for insulin sensitive patients;

• absence of readily available insulin prescribing guidelines;

• lack of prompts for ordering scheduled insulin or for dose adjustment when nutritional intake changed.

In addition, a diabetes knowledge survey administered to incoming physician house staff across different specialties demonstrated knowledge deficits in the areas of glycaemic goal setting, methods for adjusting diabetes medications for changes in nutrition or in response to a hypoglycaemic event, and switching patients from intravenous to subcutaneous insulin.25

Interventions

Following assessment of staff educational needs as well as pharmacy and system issues, the DPSC determined that a diabetes order set would address the identified barriers. Therefore, the committee developed a diabetes order set consisting of a prescriber order form for oral medications (oral medication order form: see supplementary fig 1 online) and insulin (subcutaneous insulin order form: see supplementary fig 2 online), and a one-page summary of inpatient diabetes management guidelines (guideline for inpatient diabetes management: see supplementary fig 3 online). Clinical staff across hospital departments provided feedback.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 A comparison of types of insulin order prescribed on the order set unit and the control unit. SSI, sliding scale insulin; basal, basal insulin alone; basal bolus, combination with nutritional or short-acting insulin, both alone or in combination with SSI.

The oral medication order form lists oral diabetes agents grouped by class and includes prechecked orders to hold certain medications when the patient is fasted and, in the case of the biguanide metformin, after contrast administration. The subcutaneous insulin order form prompts prescribers to select basal, nutritional and correctional insulin. The prescriber enters the numbers of units of insulin and checks boxes for insulin type which provides automatic contingencies for changes in nutritional intake. In addition, there is the option to order rapid-acting insulin analogues and very-low-dose correctional insulin. The title “sliding scale” was changed to “correction insulin” to reflect that this insulin is intended to cover excursions in blood glucose above desired levels. To minimise confusion among staff when a change in one component of the scheduled insulin is made, a prechecked box at the top of the form states that all previously ordered insulin is discontinued, mandating that the prescriber rewrite all insulin doses with any changes.

Following extensive discussions and feedback from the department of nursing, hospitalists in the division of general internal medicine and the pharmacy and therapeutics committee, the diabetes order set was piloted on two organ transplant units. Feedback from doctors, nurses and pharmacists on these units resulted in minor modifications to the order sets.

Measures

Approval for a project to compare frequency of use, prescribing trends and glycaemic outcomes on a unit using the diabetes order set (order set unit) with a unit not yet using the order set (control unit) was obtained from the Quality Improvement Review Committee of the University of Pittsburgh Medical Center Health System.26^–28 The order set unit was one of the two units that participated in the pilot of the order set. The control unit was not involved in the pilot of the order set. The quality improvement project was conducted during the planned implementation of the diabetes order set.

Patients with an order for a diabetes medication and at least 3 days of blood glucose data, hospitalised between November 2005 and February 2006 on the order set unit and the control unit were included in the project. A period of 3 days was selected for the measurement to allow determination of changes in response to glucose values outside the desired range while on the same unit. The types of order written for diabetes medications, the use of the insulin order form, incidence of episodes of hypoglycaemia (<2.2 mmol/l (<40 mg/dl) and <3.85 mmol/l (<70 mg/dl)) and hyperglycaemia (>13.75 mmol/l (>250 mg/dl)) were collected for each patient and comparisons made of the groups. Differences in the proportions were analysed using two-sided Fisher exact tests.

Data collection also included history of diabetes and comorbidity to assist an expert panel in evaluating appropriateness of inpatient diabetes management. The expert panel, blinded to the use of the diabetes order set, and composed of an endocrinologist, advance practice nurse certified diabetes educator (CDE) and a pharmacist CDE, reviewed the de-identified blood glucose values, diabetes medication orders and pertinent comorbidities for each patient to determine the appropriateness of the diabetes orders. The panel judged each order as appropriate or inappropriate diabetes management based on current guidelines.7 The proportion of appropriate orders in each group was analysed using a two-sided Fisher exact test.

A satisfaction survey was administered to prescribers, nurses and secretaries on the order set unit on completion of data collection. The survey included four questions addressing form availability, ease of use, and perceived effect on patient care (appendix).

RESULTS

We reviewed 3 days’ hospital data on a total of 70 patients. There were 55 orders for diabetes medications in 35 patients on the order set unit and 50 orders in 35 patients on the control unit. The order set was used for 71% of all orders on the order set unit and for 10% on the control unit. The use of the order set on the control unit was attributed to prescriber crossover from the order set unit.

Orders for diabetes medication were divided into five groups:

• Group 1: SSI alone;

• Group 2: oral diabetes medications (ODM) ± SSI;

• Group 3: basal insulin ± SSI;

• Group 4: basal insulin + ODM ± SSI;

• Group 5: basal bolus insulin ± SSI.

Types of diabetes therapy prescribed were evaluated by both unit assignment and order set use (fig 1). There were significantly more orders for basal/bolus insulin therapy with the diabetes order set than without it (p<0.001). There was a trend towards fewer orders for SSI monotherapy when orders were written in the diabetes order set (p = 0.156). Basal/bolus insulin was more frequently prescribed on the order set unit than the control unit (p = 0.008).

There was no significant difference in the number of patients with an episode of mild hypoglycaemia (2.2–3.8 mmol/l (40–69 mg/dl)) between the two units. There were no episodes of severe hypoglycaemia (<2.2 mmol/l (<40 mg/dl)) on either unit. More patients experienced hyperglycaemia (>13.8 mmol/l (>250 mg/dl)) on the order set unit than the control unit (60% vs 35%, respectively; p = 0.015); however, the frequency of corticosteroid therapy was higher on the order set unit than on the control unit (46% vs. 17%, respectively; p = 0.019). On both units, patients who received corticosteroids were more likely to have moderate hyperglycaemia (9.9–13.75 mmol/l (180–250 mg/dl)) than those who did not receive steroids (p = 0.014). Among patients not receiving steroid therapy, there were no differences in the frequency of hyperglycaemia between the units.

The expert panel identified 91% of the orders written with the diabetes order set and 80% of those written without it as appropriate (p = 0.137). The most common reasons for inappropriate orders were inadequate dose changes for hyperglycaemia and inappropriate use of SSI (table 1).

A high degree of satisfaction with the diabetes order set was described by all members of the healthcare team. Of the 52 respondents, 80% reported they liked using the order set, 87% agreed that the order set was easy to use, 94% indicated that it was easily accessible, and 88% felt that it improves patient care.

Following completion of the quality improvement project, the order set was implemented on all non-critical care units, a transitional care unit and a rehabilitation unit over a period of 6 months. Members of the DPSC arranged educational sessions with nursing personnel to review the order sets and the rationale for their implementation. Presentations at nursing and pharmacy leadership meetings and small group educational sessions for prescribers, nurses, unit secretaries and pharmacists accompanied the dissemination of the order forms. The guideline for inpatient diabetes management was made available on laminated pocket cards for prescribers. A contact person from the DPSC was available for each unit, to address questions, provide support and collect feedback. Critical care units were also included in the implementation programme, as the order sets allow ordering of transition to subcutaneous insulin therapy as patients are transferred.

DISCUSSION

We found that basal/bolus insulin regimens were ordered more frequently and SSI monotherapy less frequently when the newly introduced diabetes order set was used. Fewer orders were deemed inappropriate on the order set unit compared with the orders on the control unit, although this did not reach significance. This trend towards more physiological and appropriate insulin prescribing was a primary objective of creation and implementation of the diabetes order set. The guideline for inpatient diabetes management, which is part of the diabetes order set, educates and directs the prescriber to use all recommended components for rational insulin therapy in the hospital, including basal, bolus and correction insulin. It is possible that an order form which includes a list of oral agents and another that includes sections and names for basal and nutritional insulin components may have prompted physicians to order more scheduled insulin.29

The three members of the expert panel for this project were not involved in the development and implementation of the diabetes order set and were blinded to its use. The panel deemed 91% of diabetes medications ordered with the order set as appropriate. This percentage is higher than that reported in a prior expert review of inpatient diabetes management.3031 Possible reasons for this difference include the subjective nature of determining what is appropriate, prior clinical experience of panel members, methods used to determine “appropriateness”, institutional culture, and the patient population. The high percentage of appropriate orders with the diabetes order set as well as the 80% “appropriate” diabetes orders without it may have also been affected positively by the hospital-wide educational efforts accompanying implementation of the order set.

There are several limitations of this project. One is that the investigational period was limited to two hospital units, which also limited the number of doctors ordering insulin. It is possible that the results would have differed on other units represented by other medical specialties. Another limitation is the relatively small numbers of subjects included (35 subjects on each unit), which limits the generalisability of the results. There was a desire on the part of the institution to disseminate an order set that could serve as a guide to appropriate use of diabetes medications, which limited our ability to study more patients.

The diabetes order set is currently available throughout the hospital on the computerised “print on demand” system. Computer access, time to log on to the system and printer malfunction at times have presented barriers to the use of the form. At our institution, as at many others, there are plans to move toward computerised provider order entry (CPOE). CPOE has been suggested by some groups as an additional way to reduce diabetes medication errors.71718 The diabetes order set was developed with input from the hospital order set project manager to ensure that it would be easily converted into the CPOE system.

The data collected regarding reasons for “inappropriate” orders, as determined by the expert panel review, will be used to guide prescriber education. Nurse and doctor internet-based learning modules on diabetes medications have been developed based on identified educational needs.

In the hospital, the use of sliding scale monotherapy is often ingrained in the culture and passed down from mentor to trainee.12 Although hospitals are beginning to recognise the importance of inpatient glycaemic control, changing that culture and the longstanding practice of ordering SSI monotherapy will take time, education and reinforcement. This diabetes order set, which includes prescribing guidelines, provides a tool in the process of educating and encouraging hospital staff to prescribe insulin in a more physiological manner.

The Diabetes order set was generally well accepted as evidenced by the high prescriber use (71%) on the order set unit even though the use of the order forms to order diabetes medications was not mandatory at the time of data collection. One potential reason for the high acceptance and satisfaction is that the order forms require less writing than traditional handwritten methods for writing insulin orders. Clearly, the use and success of a diabetes order set depends on medical staff being aware of it, understanding it and liking it enough to use it correctly and consistently.