FOR APLUS WRITER ONLY !

Increasing medication error reporting rates while reducing harm through simultaneous cultural and system-level interventions in an intensive care unit Katherine M Abstoss, 1Brenda E Shaw, 2Tonie A Owens, 2Julie L Juno, 2 Elaine L Commiskey, 3Matthew F Niedner 4 ABSTRACT Objective:This study analyses patterns in reporting rates of medication errors, rates of medication errors with harm, and responses to the Safety Attitudes Questionnaire (SAQ), all in the context of four cultural and three system-level interventions for medication safety in an intensive care unit.

Methods:Over a period of 2.5 years (May 2007 to November 2009), seven overlapping interventions to improve medication safety and reporting were implemented: a poster tracking ‘days since last medication error resulting in harm’, a continuous slideshow showing performance metrics in the staff lounge, multiple didactic curricula, unit-wide emails summarising medication errors, computerised physician order entry, introduction of unit-based pharmacy technicians for medication delivery, and patient safety report form streamlining. The reporting rate of medication errors and errors with harm were analysed over time using statistical process control.

SAQ responses were collected annually.

Results:Subsequent to the interventions, the reporting rate of medication errors increased 25%, from an average of 3.16 to 3.95 per 10 000 doses dispensed (p<0.09), while the rate of medication errors resulting in harm decreased 71%, from an average of 0.56 to 0.16 per 10 000 doses dispensed (p<0.01). The SAQ showed improvement in all 13 survey items related to medication safety, ve of which were signi cant (p<0.05).

Conclusion:Actively developing a transparent and positive safety culture at the unit level can improve medication safety. System-level mechanisms to promote medication safety are likely important factors that enable safety culture to translate into better outcomes, but may be independently ineffective in the face of poor safety culture.

INTRODUCTION In 2000, the landmark Institute of Medicine reportTo err is human: building a better healthsystemmade plain the alarmingly high rate of medication errors in the United States, citing that medication errors were estimated to account for approximately 7000 patient deaths annually. 1 Improving patient safety in all regards has become an increasing imperative in healthcare over the past decade, and many efforts have been aimed at better measuring and reducing medication-related harm. Medication errors are underreported for many reasons, including: fear of reprisal, the concern of lawsuits, time constraints, uncer- tainty of which incidents to report, concerns of implicating others and lack of feedback and evidence that information was valued and used for system change. 2e4 Traditionally, the focus was on the clinician’s error, and blame was followed by punishment. 5 However, punitive responses have led to errors being hidden, preventing the recognition, analysis and correction of the root causes and systemic problems. 156 Effective reporting of errors is critical to discovering process defects in complex systems, and transparent, non-puni- tive dissemination of ndings is crucial for organisational learning. 78 While implementing error-proo ng concepts for human factors is conceivable and well published, effective strategies for improving the safety culture are more complex and poorly understood. Safety culture can be de ned as ‘the product of individual and group values, attitudes, perceptions, competencies, and patterns of behaviour that determine the commitment to, and the style and pro ciency of, an orga- nisation’s health and safety management’. 9 Huanget alassociated poor frontline perceptions of management with hospital 1Department of Health Management and Policy, School of Public Health, University of Michigan, Ann Arbor, Michigan, USA 2CS Mott Children’s Hospital, Pediatric Intensive Care Unit Nursing, University of Michigan, Ann Arbor, Michigan, USA 3Health System Risk Management, University of Michigan, Ann Arbor, Michigan, USA 4Division of Pediatric Critical Care Medicine, University of Michigan, Ann Arbor, Michigan, USA Correspondence to Katherine M Abstoss, Department of Health Management and Policy, School of Public Health, University of Michigan, 1113 Olivia Avenue, Ann Arbor, MI 48104, USA; [email protected] Accepted 19 April 2011 Published Online First 20 June 2011 914 BMJ Qual Saf2011;20:914e922. doi:10.1136/bmjqs.2010.047233 Original research mortality rates, and low safety culture measures with increased length of stay. 10 And while culture in the microcosm of a medical unit is partly an emergent phenomenon, leaders also shape a unit’s culture.

Leaders communicate what they feel is important through symbolic actions, rewards and punishments. If leaders encourage a free and open environment of information ow, incident reporting will increase. 11 As individuals ‘buy into’ the mission of patient safety and quality improvement, a sense of ownership leads to empowerment. 11 Indeed, contrary to the thought that a high error reporting rate is a marker of less safe care, it may be an indicator of positive safety culture that is associated with less harm. 2 However, concomitant measures of errors, harm, and culture are necessary to discern such phenomena. A system designed for safety anticipates errors, designs processes with human factors and safety in mind, while creating a culture that encourages and rewards error reporting and includes team members in developing safeguards for error prevention. 12 13 While numerous structural and system changes have been proven effec- tive at reducing medication errors, the efforts reported herein focus more on the effects of safety culture trans- formation, and the combination of culture with system.

This analysis was undertaken to better understand the impact of both system-level and safety culture strategies deployed in our paediatric intensive care unit (PICU) on medication error reporting and harm. We hypothesised that both general and speci c cultural interventions would result in signi cant changes of safety culture measures that would be temporally associated with unit- level performance improvements in patient safetyd namely increased medication error reporting and decreased medication harm.

METHODS The University of Michigan’s C.S. Mott Children’s Hospital PICU tracks medication errors through a number of mechanisms, but primarily by voluntary reporting through a web-based portal that is secure and optionally anonymous. All submissions are systemati- cally categorised in terms of harm. Determinations of harm, including whether the harm was preventable, likelihood of recurrence, and strategies for improve- ment are all optionally and preliminarily adjudicated by the individual ling the report, then formally reviewed and adjudicated by risk management, the unit level nursing leadership, pharmacists (when indicated), and selectively by multidisciplinary review at a monthly morbidity and mortality conference. For the purposes of this analysis, medication errors include all reported errors (including near misses), categorised A through Iby the National Coordinating Council for Medication Error Reporting and Prevention; medication errors resulting in harm is the subset of all reported errors with a harm rating greater than ‘temporary and insig- ni cant’ or category E or greater. 14 Not all errors result in harm (eg, a near miss), and harm may be due to an error-free intervention (eg, an adverse drug reaction).

In this analysis, medication errors resulting in harm were de ned by the overlap between medication error and adverse drug reaction, and were therefore a discreet subset of all reported medication-related errors. 15 16 Harm generally re ected either patient intolerance (eg, pain or agitation) or a medical response to the harm (eg, uid bolus for low blood pressure). Medication dose dispensation in the PICU is captured though electronic pharmacy records, yielding denominators to calculate rates of error reporting and harm. The rates of all reported medication errors and medication errors resulting in harm are plotted on control charts and prospectively monitored by the PICU quality and safety committee. The primary outcome measure of this analysis is medication errors resulting in harm, with secondary process measures of medication error reporting rates and 13 relevant culture measures from the Safety Attitudes Questionnaire (SAQ). The University of Michigan Institutional Review Board screened this study and waived formal oversight, as it was deemed a retrospective analysis of prospective quality and management work devoid of subject-level data. On an annual basis, the SAQ is administered in the PICU as part of a routine unit assessment. The SAQ, a psychometrically validated instrument, includes 64 questions that capture the attitudes and perspectives of staff on speci c safety issues at a granular level. Thirteen of the elements are relatable to medication error or reporting safety concerns. 9 From May 2007 through November 2009, the PICU implemented seven overlapping interventions to promote, among other things, the reporting of medica- tion errors. Interventions were of two broad categories:

interventions to improve the safety culture and team- work around medications; and system or structural interventions to better enable safe medication adminis- tration and ease of error reporting. Many of the inter- ventions arguably had both systemic and cultural aspects, but were categorised based on the dominant purpose of the intervention. None of the interventions were level I error proo ng methods, but all were expected to reduce the likelihood of medication error occurrences, propa- gation, or repetition. Interventions emerged from stan- dard, institutional management processes (eg, committee meetings where patient safety is discussed).

Intervention selection and timing were based on BMJ Qual Saf2011;20:914e922. doi:10.1136/bmjqs.2010.047233 915 Original research a combination of management judgement, readiness of staff, and available resources. The interventions are outlined below, with their implementation dates listed in table 1.

Culture interventions Poster tracking ‘Days since last medication error with harm’ The PICU displays a poster in the unit’s staff lounge that historically tracked multiple quality-related adverse events, including nosocomial infection occurrences, unplanned readmissions, and unplanned extubations.

In November 2008, elds were added to this poster to relate medication errors resulting in harm to patients. In contrast to calculated error rates preferred by manage- ment, the poster expresses performance metrics in a way that is intuitive to the staff, including ‘days since’ the last adverse event, the record number of days with no adverse events, the average number of days between events and the number of events in the last 12 months.

The metrics were purposely chosen to re ect time periods and measurements that ‘make sense’ to the frontline staff. Staff members refer to the poster in order to promote a sort of competition or personal challenge, ‘How many days since? What record do we need to beat?’ The poster serves as a visual reminder that harmful events happen to speci c people in real timedthe data are not just audited performance metrics that are inattentively collected and reported.

Quality improvement channel The ‘QI channel’, which began in January 2009, plays a cycling slideshow on a large screen television in the PICU staff room. The content includes performance metrics, adverse event lessons learnt and action plans, policy updates, educational pearls regarding the science of quality improvement and patient safety, socialannouncements, humour and inspiration. It was created to maintain dynamic and contemporary communication with staff, in a format that recognises the challenges of running at 24-7-365 unit. The ‘QI channel’ continues to provide a way to continuously communicate a diverse array of information, using a venue that is palatable, interesting, continuous and dif cult to ignore. Quality improvement curriculum The PICU organised three didactic ‘mini-symposia’ targeting frontline healthcare workers with content focused on healthcare safety and quality improvement, including content on medication safety and reporting.

Sessions were held in January 2008, September 2008, and September 2009. In total, more than 40 h of continuing education were provided within the PICU over 20 months. The intent was to provide frontline workers with information on patient safety and quality improvement, while improving transparency in the unit.

Speci c themes contained in the sessions included medication safety (eg, electronic order entry), trans- parency (eg, the value of learning from errors), psychological safety (eg, non-punitive leadership response to error), and communication strategies (eg, closed loop communication).

Medication error emails The PICU sends periodic emails to its staff summarising recent medication error reports, including near misses and those resulting in harm. The rst email was distributed in January 2009. The emails explain the event, lessons learnt, and any resultant practice changes, without relating the names of the patients nor the clinician involved in the error. The emails also praise individual staff that identify and address potential errors before they occur, giving public recognition and thanks for their efforts as safety champions. The emails are written in a positive and non-punitive light, in order to encourage self-reporting of errors, shared awareness of errors, and to disseminate these opportunities for improvement widely.

System interventions Computerised physician order entry The University of Michigan Health System implemented a computerised physician order entry (CPOE) system, with the PICU going live in May 2007. CPOE allows clinicians to enter treatment instructions for patients electronically in a manner that theoretically increases ef ciency in communication, reduces errors related to handwriting or transcription, provides automated error- checking, and simpli es inventory. Much research has demonstrated how CPOE can improve medication safety. 17 Table 1 Cultural and systemic interventions to increase medication error reporting Dates of note 5/2007 Computerized physician order entry (CPOE) initiated in PICU 10/2007 PICU Medication Manager program commenced 1/2008 Quality improvement curriculum I* 9/2008 Quality improvement curriculum II* 9/2008 Patient safety report form revised 11/2008 ‘Days since’ poster addition re ects medication errors with harm* 1/2009 QI Channel commenced* 2/2009 Medication error emails commenced* 9/2009 Quality improvement curriculum III* 11/2009 Patient safety report form revised *Denotes cultural interventions.

916 BMJ Qual Saf2011;20:914e922. doi:10.1136/bmjqs.2010.047233 Original research PICU medication manager program C.S. Mott Children’s Hospital initiated a Medication Manager program in the PICU in October 2007, as described in detail elsewhere. 18 The Medication Managers are pharmacy technicians who provide the following services: delivering medications to the bedside, managing prioritisation of medication delivery, streamlining communication related to medication needs, restocking of automated medication dispensers, re lling drugs and replacing missing medications, monitoring drip rates and volumes, monitoring extra- corporeal medication/ uid supplies and rates, communicating with the pharmacy, performing hourly medication runs and in-between hour STAT runs, removing discontinued medications from the bedside/ unit, controlling insulin storage, and transferring medications to other patient care areas. By internal metrics, the use of Medication Managers in the PICU improved pharmacy and nursing communication, increased nurse’s time with the patient, and reduced medication waste, but no signal was appreciated regarding impact on medication errors.

Patient safety report form revisions In September 2008, several changes were made to the patient safety report form in order to make reporting easier. To assist reporters in locating the online form easily, a link to the portal was added to the intranet homepage. The form title was clari ed to ‘patient safety incident’ to focus reports from the patient perspective and incidents instead of persons. The link to the form was changed to route directly to the login page, elimi- nating the need to navigate the Risk Management website. In addition, the form itself was revised to reduce the number of mandatory elds and eliminate entire non-mandatory screens. The form was further revised in November 2009, when the multi-page form was consoli- dated into a single quick submission form with all required elds in a single area. These changes were in keeping with visual space management, lean and human factors engineering, and simply making it easy to do the right thing.

Statistical analysis Over the time period July 2006 to May 2010, the medi- cation error and medication error with harm reporting rates were plotted regularly on a prospective statistical control chart or u-chart ( gure 1). 19 Medication errors resulting in harm were established as the outcome measure, and the reporting rates and the culture measures (SAQ) as process measures. Statistical process control was the preferred method for real-time insight into the data, and the natural break points were identi- ed through special cause variation using the fourconventional rules, at which point mean lines were adjusted. 20 In addition, the number of exposures (ie, medications dispensed), between medication errors with harm was plotted on a chart for rare events or g-chart, to better understand the time at which the long error-free interval became signi cant. For the purposes of this analysis, the before and after time periods were split in February 2009, as this was the point when all new inter- ventions had been initiated. One-sided t-tests were used Figure 1 Unit level medication error reporting rate, medication errors with harm rate, and medication exposures between reported harmful events. The three control charts for the pediatric intensive care unit data share the same timeline.

The only dates shown correspond to the start and end of the analysis, as well as the months during which interventions occurred, corresponding totable 1. Asterisked dates correspond to culture interventions. All three control charts show special cause variation (points where mean line is adjusted), with increased medication error reporting, decreased medication-related harm (a subset of the former), and more exposures to dispensed medication between reported harmful events (revealing the earliest time when special cause variation occurred). BMJ Qual Saf2011;20:914e922. doi:10.1136/bmjqs.2010.047233 917 Original research for SAQ data analyses and retrospectively on medication error rates to generate more familiar p values; statistical signi cance was de ned as a p<0.05 without correction for multiple analyses.

RESULTS Over the study period, the overall rate at which medi- cation errors were reported increased, while the rate at which medication errors with harm occurred decreased.

Figure 1shows the number of medication error reports per 10 000 doses increasing from July 2006 to May 2010, with special cause variation occurring in December 2009.

At the same time, the medication errors with harm per 10 000 doses that were reported decreased, reaching special cause variation in November 2008, and again in July 2009. Lastly, the number of exposures between errors reported with harm increased over the period, crossing the control limit around November 2009, indicating special cause variation.

Traditional statistical analysis of the data was performed comparing the data before and after February 2009, when the last new intervention began.

The reported errors resulting in harm decreased from an average of 0.56 to 0.16 events per 10 000 doses, a 71% reduction (p<0.01). The overall error reporting rate during the same period increased from an average of 3.16 to 3.95 reports per 10 000 doses, an increase of 25% (p<0.09). Errors during the study period are categorisedby the National Coordinating Council for Medication Error Reporting and Prevention taxonomy in table 2. Contextual control data were gathered to better characterise the relative impact of some of the system interventions. Because the PICU is a unit within a chil- dren’s hospital within a health system, both the data for the health system as a whole, and the rest of the paedi- atric hospital (excluding the PICU) were used as control data to compare with the PICU results. Both CPOE implementation and patient safety report streamlining were system wide, so a reporting rate for all inpatient incidents was plotted on a control chart, annotated with these interventions ( gure 2). After CPOE imple- mentation, patient safety reports decreased in frequency, and streamlining the reporting process was not tempo- rally associated with any increased reporting. Speci cally looking at safety reports led for medication events within the children’s hospital (excluding PICU), a similar pattern was observed ( gure 3). Finally, safety reports led for medication-related harm within the children’s hospital (excluding the PICU) were plotted, and showed a fairly at rate across the entire period of analysis ( gure 4). The SAQ was administered annually, and selected line item results from April 2007 (n¼108, 88% response rate) and August 2009 (n¼85, 90% response rate) are summarised in table 3. During this interval, 10 measures of safety and ease of incident reporting increased, while three measures of dif culty or hesitation in reporting Table 2 Medication error reports categorised by the National Coordinating Council for Medication Error Reporting and Prevention Taxonomy 14 7/2006 to 2/2009 3/2009 to 5/2010 Summary Total medication error reports 128 75 Average reports per month 4.0 5.0 Reports per 10 000 doses dispensed 3.12 4.08 No harm AdCircumstances or events that have the capacity to cause error 6% 7% BdAn error occurred but the error did not reach the patient; ‘near miss’ 10% 7% CdAn error occurred that reached the patient, but did not cause patient harm42% 47% DdAn error occurred that reached the patient and required monitoring to con rm that it resulted in no harm to the patient and/or required intervention to preclude harm35% 36% Harm EdAn error occurred that may have contributed to or resulted in temporary harm to the patient and required intervention*8% 4% FdAn error occurred that may have contributed to or resulted in temporary harm to the patient and required initial or prolonged hospitalization*0% 0% GdAn error occurred that may have contributed to or resulted in permanent patient harm*0% 0% HdAn error occurred that required intervention necessary to sustain life* 0% 0% IdAn error occurred that may have contributed to or resulted in the patient’s death*0% 0% *Indicates errors that the UMHS PICU categorises as errors resulting in harm.

918 BMJ Qual Saf2011;20:914e922. doi:10.1136/bmjqs.2010.047233 Original research decreased, ve of which were statistically signi cant changes. In addition to the SAQ line items, we also derived the two major domains pertaining to teamwork and safety climate. Subsequent to 2007, both domains improved; teamwork climate increased from 52.8% to 71.8% agreement (p¼0.003), while safety climate increased from 54.6% to 63.4% agreement (p¼0.127).

DISCUSSION To our knowledge, this is the rst analysis associating unit-level interventions targeting safety culture with increased medication error reporting and decreased medication-related harm. Although multiple system-level changes occurred in 2007, it was only after a cluster of culture interventions in 2008e2009 that medication error reporting increased while errors with harm decreased. Given the multiple stepwise uncontrolled quality improvement efforts targeting medication safetyin the PICU, it is dif cult to determine which individual interventions had what attributable impactdonly that overall trends in medication safety improved with concomitant improvements in staff attitudes towards patient safety and error reporting. From a temporal perspective, the series of system-level interventions in 2007 did not appear to be suf cient to create the changes observed in measured outcomes of medication safety, especially when viewed in the context of institution-wide data. While the PICU’s reporting rate went up and harm went down, institutional reporting rates went down, and harm rates were at. While confounding factors exist, two main lines of argument could be used to explain the ndings: system-level interventions are not effective in a unit with poor safety culture; and/or improving safety culture can have a more substantial impact on safety outcomes than many system-level interventions. Yet these two are not mutually exclusive because an improved safety culture may only have achieved leverage in a system Figure 2 Institutional rate for all inpatient safety reports. The safety reporting rate throughout the healthcare system dropped sequentially, meeting special cause variation during or after each related system-wide intervention annotated. Figure 3 Children’s hospital rate for medication safety reports, excluding the pediatric intensive care unit. The reporting rate for medication-related safety throughout the children’s hospital- within-a-hospital dropped, meeting special cause variation during implementation of computerised physician order entry, but demonstrating only common cause variation after streamlining the reporting system. BMJ Qual Saf2011;20:914e922. doi:10.1136/bmjqs.2010.047233 919 Original research that allowed or facilitated it. Piecemeal approaches to cultural and system-level interventions may not have the same effect as an integrated approach.Factors that have been described to improve voluntary reporting include creating a supportive environment, establishing anonymity, using clear, web-based forms and Figure 4 Children’s hospital rate for medication safety reports associated with harm, excluding the pediatric intensive care unit.

The reporting rate for medication- related harm throughout the children’s hospital-within-a- hospital only demonstrated common cause variation during the interval of analysis. Table 3 Differences in responses to Safety Attitudes Questionnaire pre intervention and post intervention Questionnaire line item% Agree or strongly agree % p 2007 2009 Change If necessary, I know how to report errors that happen in this ICU89.6 98.8 10.3 0.003* I am encouraged by my colleagues to report any patient safety concerns I may have85.2 88.2 3.6 0.267 Medical errorsyare handled appropriately in this ICU 76.6 83.3 8.7 0.123 Information obtained through incident reports is used to make patient care safer in this ICU79.2 83.3 5.2 0.243 Personnel are not punished for errors reported through incident reports60.4 73.5 21.7 0.020* The culture in this ICU makes it easy to learn from the errors of others63.9 75.3 17.9 0.042* I have seen others make errors that had the potential to harm patients50.5 56.0 10.9 0.222 I have made errors that had the potential to harm patients 37.7 51.2 35.7 0.028* Error reporting is rewarded in this ICU 10.5 25.3 141.5 0.003* A con dential reporting system that documents medical incidents is helpful for improving patient safety84.1 89.4 6.3 0.144 In this ICU, it is dif cult to discuss errorsz17.8 12.9 27.1 0.185 In this ICU, it is dif cult to speak up if I perceive a problem with patient carez20.6 14.1 31.3 0.125 I may hesitate to use a reporting system for medical incidents because I am concerned about being identi edz17.1 10.7 37.5 0.080 In 2007, n¼108; in 2009, n¼85.

*p Values are statistically signi cant at the 95% level, using a one-sidedttest.

yMedical error de ned as any mistake in the delivery of care, by any healthcare professional, regardless of outcome.

zNegatively worded question for which a low score is desirable.

920 BMJ Qual Saf2011;20:914e922. doi:10.1136/bmjqs.2010.047233 Original research providing feedback to reports, all of which we have worked to implement. 21 Furthermore, the error reporting rate was an important outcome measure of meaningful engagement in safety practices (such as disclosure and transparency) in the PICU, and an important balancing measure to ensure that less measured harm was not simply a manifestation of decreased reporting. While the number of medication errors reported increased in the study period, it is not clear whether the actual number of errors, including near misses, changed. It is likely that the true frequency of medication errors remained fairly static, but that staff members were more likely to recognise and report them with an understanding that their reports are valued and that transparency is rewarded. This was re ected by signi cant improvements in questions from the SAQ such as ‘Error reporting is rewarded in this ICU’ and ‘Personnel are not punished for errors reported through incident reports’. The sense of teamwork related to error reporting improved as staff felt supported by colleagues in reporting errors and believed they were better enabled to learn from each others’ errors. The staff felt reports were valued as the ‘Information obtained through incident reports is used to make patient care safer in this ICU’ metric improved. The questionnaire also showed educational and system-level improvements, demonstrated by the increase in the staff’s comfort level in knowing how to report errors. While Hutchinsonet aldemonstrated a positive asso- ciation between incident reporting rates and measures of safety culture when comparing institutions; 2we demon- strate that changing the safety culture within a speci c unit may lead to improved reporting and reduced harm for a speci c incident type in a relatively short time period. The seemingly paradoxical discord between increased reporting and decreased harm is more likely re ective of a preferable state where unit leadership and frontline alike have greater awareness of risks emerging from a complex system and are therefore better able to prevent or mitigate untoward effects on patients. 210 These ndings further highlight that safety culture is largely a unit-level phenomenon responsive to unit-level interventions, and invite healthcare professionals to identify effective approaches to actively improve safety culture in their local care areas. This analysis illustrates an example of ‘interventional anthropology’, improving clinical and operational outcomes by improving teamwork and safety culture. 22e25 Much research has demonstrated how system-level interventions, such as CPOE and ease of reporting, can improve medication safety. 17 However, CPOE and other human factors engineering interventions have also been implemented without appreciable improvementsdand occasionally to apparent detriment. 26 27 The PICU datado not show an impact on medication safety either way that is temporally related to several of these system-level interventions. It is after interventions focused on changing the unit’s culture that a favourable impact on medication safety emerges. While improving the system may be important to achieve higher medication safety performance in some units, it may also be insuf cient in the context of a healthcare team struggling with safety and teamwork issues. Realistic evaluation, where leaders seek to understand how processes operate in a speci c context, may reveal that the area of greatest leverage on medication safety is the safety culture instead of speci c structures or processes within the system. 28 Indeed, the substantial improvements in medication safety and SAQ scores in the PICU seem to support the old business adage that ‘culture eats strategy for lunch.’ There are important limitations to note in this anal- ysis. First, voluntary medication error reporting is imperfect in capturing all events and types of medication errors. However, in this case, the proportion of medica- tion errors resulting in harm (category E or greater) is consistent with a previous study using trigger tools to identify medication errors through chart review. 29 Because it is the most common and economical method used, nding ways to optimise voluntary medication error reporting is of practical value; further, each system for detecting medication-related harm has strengths and weaknesses. 30 31 While no ideal control data exist in this unit-based longitudinal analysis, reporting and harm patterns within the institution and children’s hospital provide some context in which to interpret institution- wide systems interventions, as well as to isolate the effects of unit-level culture interventions. It should be noted that these comparative data included all safety reports and all reported harm, not only those relating to errors.

Additionally, the multiple overlapping interventions were undertaken in a prospective quality improvement manner, without waiting for signal or stabilisation in measures before further safety strategies were deployed.

This confounds any ability to assign discreet attribut- ability and poses limitations to the simple before versus after analyses presented. Therefore data are presented in the format used prospectively by the improvement teamdstatistical process control with annotated control charts.

CONCLUSION The data reveal an association between a unit’s safety culture and medication safety measures. Speci cally, improved cultural measures are paralleled by higher medication error reporting but lower medication-related harm. Further, medication error rates and harm appear to be responsive to interventions targeting safety culture BMJ Qual Saf2011;20:914e922. doi:10.1136/bmjqs.2010.047233 921 Original research over a fairly short timeframe. Actively developing a transparent and positive safety culture at the unit level can measurably improve medication safety. System-level mechanisms to promote medication safety are likely important factors that enable positive safety culture to translate into better outcomes, but may be ineffective by themselves in the face of poor safety culture. Methods described herein provide examples that could be adapted to other units or hospital settings.

AcknowledgementsThe authorship gratefully acknowledges the PICU frontline staff for their 24-7-365 hard work and dedication, and for achieving the results we proudly present in this manuscript. We also would like to thank Todd Benner (University of Michigan Health System, Medical Center Information Technology Applications Programmer) for providing us with the pharmaceutical data to run our analysis.

Competing interestsNone declared.

Provenance and peer reviewNot commissioned; externally peer reviewed.

REFERENCES1. Institute of Medicine.To err is human: building a safer health system.

Washington, DC: National Academy Press, 2000.

2. Hutchinson A, Young TA, Cooper KL,et al. Trends in healthcare incident reporting and relationship to safety and quality data in acute hospitals: results from the National Reporting and Learning System.

Qual Saf Health Care2009;18:5e10.

3. Lehmann DF, Page N, Kirschman K,et al. Every error is a treasure:

improving medication use with a nonpunative reporting system.

Jt Comm J Qual Patient Saf2007;33:401e7.

4. Taylor JA, Brownstein D, Christakis DA,et al. Use of incident reports by physicians and nurses to document medical errors in pediatric patients.Pediatrics2004;114:729e35.

5. Potylycki MJ, Kimmel SR, Ritter M,et al. Nonpunative medication error reporting.J Nurs Adm2006;36:370e6.

6. Wolf Z, Serembus JF. Medication errors: ending the blame-game.

Nurs Manage2004;35:41e8.

7. Conway J. Getting boards on board: engaging governing boards in quality and safety.Jt Comm J Qual Patient Saf2008;34:214e20.

8. Apold J, Daniels T, Sonneborn M. Promoting collaboration and transparency in patient safety.Jt Comm J Qual Patient Saf 2006;32:672e5.

9. Sexton JB, Helmreich RL, Neilands TB,et al. The safety attitudes questionnaire: psychometric properties, benchmarking data, and emerging research.BMC Health Serv Res2006;6:44.

10. Huang DT, Clermont G, Kong L,et al. Intensive care unit safety culture and outcomes: a US multicenter study.Int J Qual Health Care 2010;22:151e61.

11. Westrum R. A typology of organizational cultures.Qual Saf Health Care2004;13(Suppl II):ii22e7.12. Grif n FA, Haraden C. Patient safety and medical errors. In:

Ransom SB, Joshi M, Nash D, eds.The healthcare quality book.

2nd edn. Chicago, IL: Health Administration Press, 2008:243e68.

13. Dennison R. Creating an organizational culture for medication safety.

Nurs Clin North Am2005;40:1e23.

14. National Coordinating Council for Medication Error Reporting and Prevention.NCC MERP taxonomy of medication errors. 1998. http:// www.nccmerp.org/pdf/taxo2001-07-31.pdf (accessed 25 Oct 2010).

15. Morimoto T, Gandhi TK, Seger AC,et al. Adverse drug events and medication errors: detection and classi cation methods.Qual Saf Health Care2004;13:306e14. 16. Newman-Toker DE, Pronovost P. Diagnostic errorsdthe next frontier for patient safety.JAMA2009;301:1060e2.

17. Bates DW, Leape LL, Cullen DJ,et al. Effect of computerized physician order entry and a team intervention on prevention of serious medication errors.JAMA1998;280:1311e16.

18. Wagner D, Pasko D, Glenn D,et al. The medication manager: results of a medication at the bedside pilot in a pediatric teaching institution.

J Patient Saf2010;6:76e9.

19. Benneyan JC, Lloyd RC, Plsek PE. Statistical process control as a tool for research and healthcare improvement.Qual Saf Health Care2003;12:458e64.

20. Western Electric Company.Statistical quality control handbook. 1st edn. Indianapolis, Indiana: Western Electric Co, 1956:v, OCLC 33858387.

21. Stockwell DC, Slonim AD. Quality and safety in critical care units.

Intensive Care Med2006;21:199e210.

22. Deilkas E, Hofoss D. Patient safety culture lives in departments and wards: multilevel partitioning of variance in patient safety culture.

BMC Health Serv Res2010;10:85.

23. Smits M, Wagner C, Spreeuwenberg P,et al. Measuring patient safety culture: an assessment of the clustering of responses at unit level and hospital level.Qual Saf Health Care2009;18:292e6.

24. Miller MR, Silver P, Stuart J,et al. ‘CA-BSI Collaborative:

transforming care, culture, and leaders’.Annual Meeting for the National Association of Children’s Hospitals and Related Institutions 12 Oct 2009. Orlando, FL: 2009.

25. Niedner MF.‘Interventional anthropology: strategies to improve teamwork and safety culture to improve healthcare outcomes’.

University of Michigan Pediatric Grand Rounds, 2009.

26. Koppel R, Metlay JP, Cohen A,et al. Role of computerized physician order entry systems in facilitating medication errors.JAMA 2005;293:1197e203.

27. Han YY. Unexpected increased mortality after implementation of a commercially sold computerized physician order entry system.

Pediatrics2005;116:1506e12.

28. Pawson R, Tilley N.Realistic evaluation. London: Sage Publications Ltd, 1997.

29. Agarwal S, Classen D, Larsen G,et al. Prevalence of adverse events in pediatric intensive care units in the United States.Pediatr Crit Care Med2010;11:568e78.

30. Olsen S, Neale G, Schwab K,et al. Hospital staff should use more than one method to detect potential adverse events: incident reporting, pharmacist surveillance and local real-time record review may all have a place. Qual Saf Health Care2007;16:40e4.

31. Meyer-Massetti C, Cheng CM, Schwappach DL,et al. Systematic review of medication safety assessment methods.Am J Health Syst Pharm2011;68:227e40. PAGE fraction trail=8.75 922 BMJ Qual Saf2011;20:914e922. doi:10.1136/bmjqs.2010.047233 Original research