Background

Tallaght Hospital, Dublin, Ireland, is a 661-bed publicly funded tertiary academic teaching hospital affiliated with Trinity College, Dublin. In 2008, the hospital's Antimicrobial Stewardship programme targeted once daily gentamicin dosing and TDM practice as a focus for continuous improvement. Our hospital had four different gentamicin dosing and monitoring policies, each tailored to specific patient groups: critical care patients, patients receiving renal replacement therapy, paediatric patients and, finally, the remainder of the general adult patient population. Drug prescribing practice consisted of paper based physician order entry. Performance in dosing and monitoring was perceived to be generally good in our critical care, paediatric and renal patient groups, and drug error reports were rare. This was attributed to good stakeholder familiarity with gentamicin use, and process flexibility due to nurse venesection, dedicated staff training and a specialised staffing complement with relatively low staff turnover. However, error reports had identified regular non-adherence to policy in the general adult patient population where many of these factors were not present.

In this population, each patient was cared for by hospitalist medical staff, with nursing and clinical pharmacist support. Each hospitalist team consisted of at least three medical staff headed by a specialist consultant, with a medical team to patient ratio of roughly 1:10 to 1:30. Consultant specialists were permanent staff members, but more junior medical staff were rotational and changed jobs every 3–6 months. Nurses, clinical pharmacists and phlebotomists were ward based and tended not to change roles as regularly. Venesection was performed by phlebotomy staff with some medical staff involvement. While all patients were located on one geographic hospital site, most patients were not cohorted by medical team or specialty.

Gentamicin usage process error occurred despite readily accessible evidence based guidelines in a medicines usage handbook, regular staff training and clinical pharmacist intervention where deviation from policy occurred. Guidelines and access to patient laboratory results were also readily available electronically on the ward, with each ward having at least one networked computer per 10 patients. Improvement efforts focused on this patient population of approximately 300 patients annually.

Using Six Sigma to improve practice

A multidisciplinary stakeholder improvement team was formed. This team included pharmacists, medical microbiologists, a non-consultant hospital doctor, a nurse, a phlebotomist and laboratory staff representatives. The project lead undertook ‘Green Belt’ training in the Six Sigma DMAIC methodology, and mentoring was provided by a local Six Sigma Master Black Belt. A member of the senior hospital management team who was also Green Belt trained was the project sponsor. The overall aim of the project was to improve the quality of gentamicin usage when treating Gram-negative infection. The probability of gentamicin treatment efficacy in Gram-negative infection correlates with serum peak concentration of the drug post dose: it exhibits concentration dependent killing.13 It follows therefore that underdosing due to poor dose selection and missed doses was potentially causing suboptimal treatment outcomes in our patients. This project attempted to improve the consistency of the attainment of more appropriate daily dosing of gentamicin, while also ensuring that the TDM process facilitated adequate safety monitoring without itself causing inadvertent dose omission or delay. It was hypothesised that treatment efficacy would improve through better translation of evidenced based dosing into local practice. This improvement effort also needed to be at least cost and staff-neutral for the organisation.

A Project Charter was agreed which identified all stakeholders and confirmed the basic process. The project was carefully scoped to ensure clarity on desired end-points, and an outline timetable was agreed. A series of working group meetings were held throughout the improvement project to work through the DMAIC improvement cycle. Ethics committee approval was obtained for data collection and intervention at the outset of the project, and all interventions were overseen by the hospital Drugs and Therapeutics Committee.

Define

A high level view of the gentamicin dosing and monitoring process was obtained using a Suppliers, Inputs, Process, Outputs and Customer analysis. The dosing and monitoring process was also mapped by the improvement team in detail using process mapping techniques. This exercise enabled greater understanding of the realities of the process from the perspective of the operatives involved. The improvement team then identified five process measures that were deemed critical to quality (CTQ) in achieving optimal dosing and drug monitoring with aminoglycosides.

1. The patient is prescribed the correct dose as defined by hospital policy for his or her weight and renal function.

2. The patient's first trough level is taken on the correct day as per protocol.

3. The patient's trough level blood sample is taken in the correct time window.

4. The patient received doses immediately after the trough level had been taken, unless the patient fell into a predefined group of patients where a level result was required before redosing.

5. All patients with high trough levels were managed in line with hospital guidelines.

Particular emphasis was placed on compliance with practice around the first trough level, as analysis suggested that failure to follow protocol at the outset of the process was much more likely to be associated with process breakdown throughout therapy. The number of patients who experienced an episode of disruption to gentamicin therapy due to TDM process breakdown was also recorded. This was defined as one or more episodes of unintended inappropriate delay in excess of 4 h or one or more unintended complete dose omissions per treatment episode. This had to occur due to TDM process breakdown rather than intentional prolongation of dosage interval or dose omission as per protocol. It was identified by the investigating patient medical record to differentiate deliberate dose omission due to a dosage change from unintentional dose omission due to process breakdown. Where a documented reason for the omission was not present, the medical team was interviewed by the investigators to confirm if omission was intentional, or identify if it was a result of miscommunication or misinterpretation of guidelines.

Measure

A prospective baseline audit of practice was conducted in March 2008 in consecutive patients over 4 weeks. In all, 21 patients were identified from laboratory drug assay requests for gentamicin, and each patient on gentamicin was audited at least 1 week after his or her first level was processed to capture practice over that week. This represented a convenience sample, and included one-twelfth of all patients treated with gentamicin in this patient group in 2008. Consecutive sampling was chosen to avoid selection bias. This sample size also achieved widespread measurement of practice across the hospital, while providing a practical timeframe to facilitate timely improvement. Where treatment continued for more than a week, patients were followed until course completion. The clinical teams were not informed that audit was being undertaken as it was thought that this information might have prompted a change in behaviour. However, if an error was identified in patients where treatment continued at the time of measurement, investigator intervention occurred as necessary.

None of the patients complied with all five CTQ parameters, and our process defect rate was therefore 100%. (figure 1) Moreover, 58% of patients had therapy unintentionally disrupted due to process breakdown. Of these, 47% had a delay of more than 4 h in drug administration, and 11% missed at least one dose unintentionally due to process breakdown. Both of these error types were regarded as clinically significant. When examining the group as a whole, one in seven planned doses was missed due to process breakdown. Calculation of the cumulative underdosing per day of therapy was conducted by adding up the total dose that would have been achieved had hospital policy been followed without inadvertent dose omission, compared with the cumulative dose administered. This revealed that the total dose administered was 29% less than the planned dose as per policy. The average daily dose (mg/kg)/day of gentamicin therapy was 3.39 mg/kg.

Analyse

A root cause analysis exercise, informed by baseline audit data and process mapping, was conducted with all stakeholders (box 1). A number of key factors which were implicated in poor performance emerged, and these influenced our improvement plan.

Improving the accuracy of sample timing

Inaccurate timing of level sampling was commonly identified. Root cause analysis revealed a deficit in phlebotomist information on blood sample timing requirements and that phlebotomy services were not aligned to around the clock sample timing. Outside of phlebotomy hours, medical staff were assigned to take serum samples. Audit revealed that these samples were sometimes omitted or taken in the wrong time window. Further analysis suggested that drug assay sampling was low down on the list of priorities for busy on-call medical staff, with resultant poor performance. The short timeframe available for sampling also contributed to inaccurate timing of blood sample taking.

Reassignment of phlebotomist resources and improving staff awareness were considered as a potential solution. Review of the published literature revealed other potential options. A dedicated team assigned to TDM management has been suggested as an optimal approach to address this issue.14 Alternatively, expanding the time window for taking blood samples would provide more flexibility for blood sample collection. We decided to opt for the latter as additional staffing resources were not available, and TDM specialisation could result in deskilling of general staff.

Three potential options to improve sample timing flexibility were considered. The Hartford nomogram15 and the Urban–Craig16 nomogram were examined, as both improved sample time flexibility and are well validated. They were both radically different to baseline practice at our hospital and other local hospitals with whom we shared a large common pool of junior medical staff. Using either would require junior medical staff in particular to undertake significant training, as they tended to change jobs and hospitals most often, and most responsibility would be placed on them to ensure good adherence with both nomogram processes. Moreover, these solutions did not fully address the identified root causes of phlebotomist availability and awareness of when sample timing was required. Indeed, error as a result of poor performance in sample timing had been reported with the use of nomograms elsewhere.17

Modification of the current in-house process which specifically addressed the root causes of poor performance was also considered. Pharmacokinetic evaluation of gentamicin clearance18 explored the potential for extending the functional definition of a trough level from 23 to 24 h postdose to a more accommodating 18–24 h postdose. Standardisation of dosing timing and phlebotomy level taking timing were also considered as outlined in figure 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Dosage administration timing standardisation methodology.

A dosing interval of 19–28 h over the first three doses would result. All patients with an estimated creatine clearance of over 10 ml/min would receive three doses over the first 72 h of therapy. Using this modified policy had the potential for better process performance than baseline practice or the Hartford/Urban–Craig nomograms in our hospital, and more conservative dosing could potentially be compensated for by less inappropriate dosage delay or omission. The option of individualising therapy in patients who required more aggressive therapy would also remain available to clinicians. Crucially, the timing of administration and sample timing would be managed by nurses and phlebotomists who were less rotational than junior medical staff and more likely to be familiar with in-house processes. The junior medical staff role was still critical, but was considered to be more straightforward in this putative process.

Failure modes and effects analysis

A failure modes and effects analysis7 was conducted to assess the relative risks and potential mitigation strategies required if we were to adopt either the Hartford and Urban–Craig nomograms or our own modified inhouse process. A potentially higher risk of process breakdown with nomogram usage was identified. However, all dosing and monitoring strategies would require a full programme of staff education, clear and accessible guidelines for prescribers, pharmacists and nurses, and regular monitoring of performance with intervention and feedback. Guided by this analysis, our new inhouse process was adopted (figure 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

The new gentamicin dosing and therapeutic drug monitoring process.

Improve

A pilot of our new policy began on two wards over a 4-week period. This was accompanied by full stakeholder education. A revised gentamicin intravenous administration monograph for nurses which provided a standard operating procedure outlining the nurse role in the process was also introduced.

Ten patients were recruited. Improved process performance was observed, with less delay in dosing due to elevated levels than previous practice. The new protocol was subsequently introduced across all other adult specialties. Full stakeholder education was provided by a multidisciplinary education team of trained clinical pharmacists and nursing clinical placement coordinators over a 4-week period. A repeat of the baseline audit was conducted in a further 25 consecutive patients in August 2009. Statistical analysis was conducted using a Two Proportions and a Fisher's exact test.