Article Text

Self-reported violations during medication administration in two paediatric hospitals
  1. Samuel J Alper1,
  2. Richard J Holden2,
  3. Matthew C Scanlon3,
  4. Neal Patel4,
  5. Rainu Kaushal5,
  6. Kathleen Skibinski6,
  7. Roger L Brown7,
  8. Ben-Tzion Karsh8
  1. 1Human Factors Practice, Exponent Failure Analysis Associates, Chicago, IL USA (work completed while a PhD student in the Department of Industrial and Systems Engineering of the University of Wisconsin)
  2. 2Departments of Medicine and Biomedical Informatics, Vanderbilt University, Nashville, TN, USA (work completed while a Postdoctoral Fellow at the School of Medicine and Public Health at the University of Wisconsin)
  3. 3Department of Pediatrics, Medical College of Wisconsin, Milwaukee, WI USA
  4. 4Department of Pediatrics, Vanderbilt University, Nashville, TN, USA
  5. 5Departments of Pediatrics, Public Health, and Medicine, Weill Cornell Medical College, New York, New York, USA
  6. 6Department of Pharmacy, St Mary's Hospital, Madison, Wisconsin, USA
  7. 7School of Nursing and Department of Family Medicine, University of Wisconsin-Madison, Madison, Wisconsin, USA
  8. 8Department of Industrial and Systems Engineering and Systems Engineering Initiative for Patient Safety, University of Wisconsin, Madison, WI USA
  1. Correspondence to Dr Ben-Tzion Karsh, Industrial and Systems Engineering, University of Wisconsin-Madison, 1513 University Avenue, Madison, WI 53706-1572, USA; bkarsh{at}engr.wisc.edu

Abstract

Content Violations of safety protocols are paths to adverse outcomes that have been poorly addressed by existing safety efforts. This study reports on nurses' self-reported violations in the medication administration process.

Objective To assess the extent of violations in the medication administration process among nurses.

Design, setting and participants Participants were 199 nurses from two US urban, academic, tertiary care, free-standing paediatric hospitals who worked in a paediatric intensive care unit (PICU), a haematology-oncology-transplant (HOT) unit or a medical-surgical (Med/Surg) unit. In a cross-sectional survey, nurses were asked about violations in routine or emergency situations in three steps of the medication administration process.

Main outcome measure Self-reported violations of three medication administration protocols were made using a seven-point 0–6 scale from ‘not at all’ to ‘a great deal’.

Results Analysis of variance identified that violation reports were highest for emergency situations, rather than for routine operations, highest by HOT unit nurses, followed by PICU nurses and then Med/Surg unit nurses, and highest during patient identification checking, followed by matching a medication to a medication administration record, and then documenting an administration. There was also a significant three-way interaction among violation situation, step in the process, and unit.

Conclusions Protocol violations occur throughout the medication administration process and their prevalence varies as a function of hospital unit, step in the process, and violation situation. Further research is required to determine whether these violations improve or worsen safety, and for those that worsen safety, how to redesign the system of administration to reduce the need to violate protocol to accomplish job tasks.

  • Patient safety
  • compliance
  • medication administration
  • violation
  • human error
  • human factors
  • implementation science
  • information technology

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

Hospitalised patients may experience an average of one medication error per day.1 Errors at the medication administration stage of the medication use process are especially likely to result in patient harm because, with the exception of high-risk medications, no independent healthcare provider checks a medication before it is administered.2 3 To reduce errors in the medication administration process, protocols and technologies have been implemented.1 4 However, the success of protocols and technologies depends on compliance. Non-compliance with protocols, here referred to as violations, represents a path to medical injury that requires further study.5 Violations are actions that break rules, policies, protocols or procedures.6 In the case of medication administration, this would apply to any rule, policy or procedure regarding how steps in the administration process should be carried out.

Violations in the medication administration process are known to occur,7–9 but to date no quantitative data exist on the extent or prevalence of violations. This study of paediatric nurses in two hospitals is the first to examine the extent to which nurses knowingly violate medication administration protocols. Because the violations are self-reported, intentional violations and unintentional violations recognized as violations after the fact are captured. Intentional violations are those committed knowing that rules, policies or norms are being broken.10 11 This study is also the first to examine how violations vary between two hospitals, between three types of hospital units, between three specific steps in the medication administration process, and between violations committed routinely versus in emergency situations.

Methods

This study was conducted as part of a broader study to evaluate the impact of a barcoded medication administration system on medication administration errors and on end users (nurses, pharmacists and pharmacy technicians). This study analyses cross-sectional data from a paper survey administered at two hospitals before either one implemented barcoded medication administration. The study was approved by the University of Wisconsin-Madison Institutional Review Board (IRB) and by the IRBs of both study hospitals.

Setting and sample

Two urban, academic, tertiary care, free-standing paediatric hospitals participated in the study. Hospital A had 222 beds and was located in the Midwest USA. Hospital B had 162 beds and was located in the South USA. Three units were studied in both hospitals: a paediatric intensive care unit (PICU), a haematology-oncology-transplant (HOT) unit and a general medical/surgical (Med/Surg) unit.

Full-time nurses (24 h per week or more) who provided patient care in the study units were eligible to participate. The sampling frame was 203 nurses from hospital A and 144 from hospital B. Surveys were distributed to all nurses in the sampling frame. Data were collected during November–December 2005 and March–May 2006 at hospitals A and B, respectively. Response rates of 59.6% and 54.2% were attained in hospitals A and B, respectively. Descriptive statistics are provided in table 1.

Table 1

Descriptive statistics

Measures

The questions about violations were developed based on previous research10 12 and are provided in table 2. Respondents were instructed to think of the past 30 days.

Table 2

Questions and response scale used in study

Design and procedures

Pretesting

Questionnaire items were evaluated using expert review and cognitive interviewing.13 Expert reviews were completed by safety trained experts, nurse researchers, a pharmacist co-investigator and two questionnaire design experts to assess readability, word choice, question clarity and question content. Sixteen cognitive interviews were conducted with nurses at large teaching hospitals to appraise content validity.14 Through these interviews, it was determined that nurses had a shared understanding of the protocols relevant to the questions and of the meaning of breaking protocol. In addition, nurses reported interpreting the response scale as one of frequency. That is, when asked, for example, ‘why did you choose to circle the 2 (“some”) instead of the 3 (“moderate”)’ the nurses would respond that they only violated some of the time, not a moderate amount of the time.

Publicity

The study team worked with nurse managers on the study units and with the research coordinators at each hospital to determine where to hang flyers advertising the study. One flyer provided information about the study; another listed ‘frequently asked questions’ and answers to those questions. Some nurse managers sent emails to nursing staff encouraging them to participate in the study and/or made the PowerPoint presentation of the study available electronically.

Survey administration

When possible, members of the research team attended staff meetings to speak with nurses and hand out surveys. All survey packets were delivered in personalised envelopes which contained the survey, an information sheet (consent form not requiring a signature), a stamped, addressed envelope in which to put the completed survey and mail it back to the survey processing centre, and a $5 cash incentive. In addition to, or in lieu of, staff meetings, the research team presented the study to nurses during the report at the beginning of the nurse's shift or during informal inservices. When few surveys remained to be delivered, the research team attempted to find each nurse to provide an individual presentation of the study and to deliver that nurse's survey. Before giving surveys to subjects, the research team endeavoured to speak with each of the subjects to tell them about the study, how to participate in the study, and about measures taken to ensure their confidentiality. After the presentation, time was left for subjects' questions. Then, subjects were given their survey packets. After survey distribution started, a series of three reminders was used to encourage nurses to complete the survey. To protect confidentiality, all nurses received identical-looking reminders.

Analysis

A four-factor mixed measures factorial model15 16 was carried out in SPSS V.16.0 using a two (hospital: hospital A, hospital B) by three (unit: PICU, HOT, Med/Surg) by three (step: matching the medication to the medication administration record (Match-MAR), checking patient identification (ChxID), and documenting administration of the medication (DOC)) by two (violation situation: routine, emergency) design. Step and violation situation were within-subject variables. Hospital and unit were between-subject variables. Contrasts were run for significant interaction terms. The analysis of variance (ANOVA) was performed on the untransformed data and on the rank transformed data and very similar results were obtained. Therefore, based on Conover,17 the results of the ANOVA with untransformed data are presented here.

Interaction contrasts were carried out using Graphpad Software (t-test calculator). A false discovery rate (FDR) approach was used to correct for family-wise error rate for numerous interaction contrasts. FDR is the expected proportion of false-positive findings among all the rejected hypotheses. Compared with the typical approach (eg, Bonferroni, Sidak, etc), FDR is not as conservative and provides a good balance between discovery of statistically significant effects and limitation of false-positive occurrences (see details elsewhere18).

Results

Table 3 provides means and SDs for each step of the medication administration process in each hospital, each unit within each hospital, overall for both hospitals, and for each type of unit across hospitals. Additionally, the percentage of nurses reporting ever violating (a response >0) is reported. Online appendix 1 shows the raw distributions of responses in a mosaic plot; the plot provides the complete distribution of responses on the seven-point response scale for: both violation situations, both hospitals, the three types of units, and the three steps of the medication administration process.

Table 3

Means, standard deviations, and per cent self-reporting ever violating for questions about violations

Significant effects (main effects)

The main effects (testing for the effects of individual variables, without regard for interactions or moderating effects) for the variables of unit (F (2,1158)=3.47, p<0.05), step in the process (F (2,1158)=17.52, p<0.05) and violation situation (F (1,1158)=77.63, p<0.05) were significant. Effect sizes for significant effects, based on Cohen's D, were 0.26 for unit, 0.26 for step and 0.49 for violation situation. For unit, violation scores were highest for HOT units, followed by PICUs and then Med/Surg units. For step, violation scores were highest for ChxID, then Match-MAR and finally DOC. The main effect for violation situation showed that violation scores were higher in emergency situations. The main effect for hospital was not significant (F (1,1158)=2.80, p=0.096).

Significant effects (interaction effects)

Interaction effects, which test the hypothesis that the effect of variable A on the outcome differs depending on the level of some variable B, are more illuminating than the main effects as they test whether and how main effects are moderated by another variable. The four-way interaction among the four main effects was not significant. There was a significant unit by step by violation situation interaction (F (4,1158)=3.24, p<0.05). To ease discussion of the three-way interaction, the two-way interaction between unit and step is discussed for each level of violation situation. Figures 1A,B show the three-way interaction.

Routine situations

Contrasts for the three-way interaction show that there were no differences among the three units for DOC or Match-MAR for routine violations (figure 1A). However, for ChxID, the HOT units had significantly higher routine violation scores than the PICUs and Med/Surg units. In the PICUs and HOT units, routine violations in ChxID were reported more than violations in Match-MAR. Across units, significantly more violations were self-reported in ChxID than in DOC.

Figure 1

The three-way interaction of unit × process × situation. (A) Variation of unit and process within routine situations. (B) Variation of unit and process within emergency situations. ChxID, checking patient identification; DOC, documenting administration of the medication; HOT, haematology-oncology-transplant unit; Match-MAR, matching the medication to the medication administration record; Med/Surg, medical-surgical unit; PICU, paediatric intensive care unit.

Emergency situations

In emergency situations (figure 1B), there were significant unit differences for all three steps of the medication administration process. For Match-MAR, the PICUs and HOT units had significantly higher emergency violation scores than the Med/Surg units. For ChxID, all three units significantly differed from each other; the HOT units had higher violation scores than the PICUs and Med/Surg units; the PICUs had significantly higher violation scores than the Med/Surg units. Finally, for DOC, the HOT units had significantly higher violation scores than both the PICUs and Med/Surg units.

Across units, there were no significant differences between reported violations in Match-MAR and ChxID. In the PICUs and HOT units, there were fewer reported violations in DOC than in Match-MAR or ChxID; in the Med/Surg units, there were no differences between steps.

Discussion

This study provided the first attempt at quantifying violations in the medication administration process. The results show that, based on nurses' self-reports, depending on the unit of care, the violation situation and the step of the medication administration process, violations were reported by as few as 33.3% to as many as 90.8% of nurses (defined as any non-zero response); reports of violations were not limited to one or two aberrant scenarios. This study also provided the first assessment of the extent to which violations vary by structure and process variables; violations do vary depending on the violation situation, the care unit and the particular step of the medication administration process.

The study confirms findings from qualitative studies that show violations of protocols and technology use policies occur in the medication administration process.7–9 19 20 This study extends prior findings by showing the extent of the violations and the structure and process variables that contribute to varying violation rates.

Significant effects (main effects)

Main effects were found for unit, step and violation situation, but not for hospital. The null finding for hospital is as important as the significant main effects. Not finding hospital differences demonstrates that our findings were not isolated to what might otherwise have been labelled a problematic hospital. Instead, we found similar self-reported violation levels at two highly regarded paediatric hospitals. We cannot say for sure if other paediatric hospitals have similar levels of violations, but our findings do not rule out that violations in medication administration are a systemic phenomenon.

The data also show that violations during emergency situations are more likely to be reported than routine violations, that violations are more likely to be reported in HOT units, followed by PICUs and then by Med/Surg units, and that violations are more likely to be reported in the ChxID process, followed by Match-MAR and then by DOC. The significant three-way interaction between unit, step and violation situation provides a more accurate picture of those variables than the main effects, and is discussed next.

Significant effects (interaction effects)

In routine situations, violations of the protocol for ChxID are reported more often on the HOT units than on the other two units. Greater reporting of ChxID violations in HOT units compared with other units may be partially explained by longer patient stays in HOT units due to the nature of patients' diseases. Because patients are on the unit for a long time, nurses are better able to get to know the patients and may perceive that following the protocol for checking identification of these more familiar patients is less necessary. There may be a misunderstanding of the need for the identification check as per protocol: the check goes beyond verifying that the nurse knows the patient; it also ensures the medication in hand matches the patient. A nurse could accidentally take the wrong medication for the right patient or have the ‘right’ medication, but get distracted and go into the wrong patient's room—in both cases, simply recognising the patient in the room may not prevent the misadministration.

In emergency situations, the picture is different. Typically, the HOT units had the highest reported levels of violations in emergency situations, followed by PICUs and then Med/Surg units. Why the HOT units had the highest reported levels requires further investigation. The reason for their reporting of higher levels of violations of Match-MAR may be related to patients having long lengths of stay, but it is not clear why violations of DOC would be higher.

Interestingly, in emergency situations, violations of Match-MAR were reported more than violations of ChxID and DOC in all units (though not significantly more in all cases). This differs from routine situations, in which violations of ChxID were most reported. This may be because the MAR was located in a patient's chart and not necessarily available at the bedside, unlike the patient identification. Another possibility is that the MAR may be less relevant during an emergency; emergency medications may be ordered and administered without first being entered into the MAR.

Violations in healthcare settings

Rules, policies and procedures assume a work environment that is favourable to compliance. In contrast, healthcare delivery, especially in inpatient PICUs and HOT units, is characterised by time pressure, high acuity patients, and a need for problem solving. In those environments, medication administration is highly complex and filled with interruptions to manage and challenges to resolve.21–24 When the reality of a clinical environment does not match the environment assumed by designers of rules or policies, violations are to be expected—because there is not enough time to comply with protocols, because the cost of compliance is perceived as higher than the cost of violating or because different protocols may conflict.5 6 8 25 The presence of conflicting goals is common in complex systems like healthcare26 and makes it more likely that violations will occur regularly.

Importantly violations do not imply less safety; violations may reduce, have no impact on or even improve safety.10 27 28 While violations may increase the likelihood and severity of adverse events,11 29 violations can also be necessary or even desired (by a patient or family), such as when compliance would slow down a process that requires speed to save a life.30 In such cases, it is the clinician who, while technically violating a rule, might be in fact making ‘the difference between total disaster and a small accident, or no accident at all’.31

Limitations

This study has four main limitations. First, the questions captured self-reports of violations rather than the actual number of violations. Therefore, we cannot report actual rates of violations. Our data only give a sense of the extent of intentional violations that nurses were able and willing to self-report. Violations committed unknowingly would not have been captured in this analysis unless recognised after the fact. While observing violations is possible and may have provided a more accurate count of violations, differentiating between intentional and unintentional violations is impossible because the observer cannot know the subject's intentions. Further, nurses may choose not to violate while being observed because violations are socially undesirable.

Second, only two hospitals participated. This limits the generalisation of the findings. However, most published patient safety studies have occurred in a single institution. The lack of a main effect for hospital suggests that the extent of the violations is not limited to one hospital. A related point is that we did not adjust for the demographical differences between the hospitals so we do not know if any of the existing results could be explained by differences in, for example, job experience. This should be addressed in future research. Third, because of the question wording, we do not know the specific violation behaviours that respondents had in mind when they reported committing violations, generally. For example, we do not know if a reported violation of ChxID meant only checking one patient identifier instead of two, checking no identifier at all or violating the protocol in some other way. Future research will need to obtain such specifics.

Fourth, there could be response bias. The response rates were acceptable, but we do not know how non-respondents differed from respondents. The response rate is high enough, however, that even if all non-respondents reported not violating at all, the data would show high levels of reported violations. Finally, though not a limitation per se, this study does not explain how to address violations. There exists general knowledge about violation causes in work domains6 and there is emerging evidence from healthcare,7 8 20 32 but more research is needed.

Conclusion

This study demonstrates that violations occur in the medication administration process across hospitals, units and steps in the process. Importantly, factors such as the unit in which a nurse works, whether or not a nurse is in an emergency violation situation and the step of the medication administration process all interact to influence intentional violations. This finding is important as it provides empirical evidence that structural and process variables influence intentional violations. Just as safety scientists urge a shift from blaming people for medical errors to studying the causes of errors, we strongly advocate not blaming clinicians for violations, but rather searching for a more systems-oriented causal explanation. It is, after all, the causes of violations that need remediation.

Acknowledgments

The authors would like to thank Scot Barnett and Nancy DiCanio for their assistance in editing this manuscript. The authors also thank the Barcode Study Team and the many nurses who participated in the Barcode Study and made this work possible.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:

Footnotes

  • Funding This study was funded in part by a grant from the Agency for Healthcare Research and Quality (R01 HS013610, Karsh PI). RJH was supported by a pre-doctoral training grant from the National Institutes of Health (1 TL1 RR025013-01) and a post-doctoral training grant from the Agency for Healthcare Research and Quality (5 T32 HS000083-11).

  • Competing interests None.

  • Ethics approval Ethics approval was provided by University of Wisconsin-Madison, Children's Hospital of Wisconsin, Vanderbilt Children's Hospital.

  • Provenance and peer review Not commissioned; externally peer reviewed.