Article Text

The impact of interruptions on the duration of nursing interventions: a direct observation study in an academic emergency department
  1. Gai Cole1,
  2. Dicky Stefanus2,
  3. Heather Gardner3,
  4. Matthew J Levy3,
  5. Eili Y Klein3
  1. 1Emergency Medicine, Johns Hopkins Medicine, Baltimore, Maryland, USA
  2. 2Carey Business School, Johns Hopkins University, Baltimore, Maryland, USA
  3. 3Emergency Medicine, Johns Hopkins University, Baltimore, Maryland, USA
  1. Correspondence to Dr Gai Cole, Emergency Medicine, Johns Hopkins Medicine, 5801 Smith Ave, Davis Building, Suite 220, Baltimore, MD 21209, USA; gcole4{at}


Background Interruptions to nursing workload may contribute to procedural failures and clinical errors impacting quality/safety of care, but the impact of interruptions on the duration of these activities has not been closely scrutinised. This study analyses the effect of interruptions to care provided by nurses and clinical technicians on the length of clinical procedures and interventions (excluding the length of the interruption).

Methods An observational time study of the effect of interruptions on common nursing interventions in the emergency department (ED) of a large academic medical centre was conducted. This study used direct observations of nurses and clinical technicians while delivering care to patients.

Results The average time spent on an uninterrupted intervention was 296.47 s (median:185.15, SD:319.05), while interrupted interventions took 682.02 s (median:589.63, SD:504.59). Controlling for intervention type and other potential confounding factors using multiple linear regression found that interrupted interventions were 121.36 s (95% CI 79.57 to 163.15) longer, a 19 percentage point increase (95% CI 11.31 to 26.89), than an intervention without (excluding the length of the interruption). Family/patient interruptions effected duration the most while staff interruptions affected the intervention time the least.

Discussion Our findings are consistent with outcomes of studies in non-healthcare domains, but are contrary to a study of ED physicians, suggesting differential responses to interruptions by physicians and nurses. Future studies on interruptions in healthcare should thus be discipline specific. Though the effect of interruptions on intervention length is only about 2 min, in an ED setting, this can increase patient risks and costs. To better focus efforts to reduce interruptions future research should focus on further separation of interruption type (eg, urgent vs routine or unnecessary).

  • Emergency department
  • Healthcare quality improvement
  • Nurses
  • Human factors
  • Interruptions

Statistics from

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.


Interruptions frequently occur in healthcare settings1 and are common in busy, high-risk, clinical settings such as the emergency department (ED).2 Prior studies have shown that emergency nurses experience high rates of interruptions: about 11–12 interruptions hourly.3 ,4 The consequences of such frequent interruptions include an increase in the number of procedural failures and clinical errors.5–7 Studies across clinical settings (including the ED) have described the rate3 ,4 ,8 and sources9 ,10 of interruptions in nursing care, but the results have focused on the impact these had on the precision of clinical activities (eg, correct dosage provided to the patient). Conversely, the impact of interruptions on the duration of these activities (eg, length of time to administer the dosage) has not been closely examined. Whether or not these interruptions (and their source) have an impact on the length of a clinician's interventions remains an open discussion.

The length of an intervention is important because of the direct relationship between the duration of care and the patient's overall length of stay (LOS) in the ED.11 Lengthier interventions raise LOS. Negative outcomes related to increased LOS in the ED are well documented12 and include an increase in the proportion of patients who leave without being seen by a physician,13 ,14 increased patient morbidity,15 ,16 and low patient satisfaction.17–19 Further, when low acuity interventions take longer, the effect is to keep the clinician from providing higher acuity interventions to other patients. This is true across the spectrum of ED patients, and within geographical subcategories of patients such as those seen in the acute care or urgent care sections of the department. Lengthier nursing interventions may also increase risk by delaying nurses from pulling patients from the waiting room and into the next step in their care. Finally, the positive correlation between ED LOS and hospital costs is well documented,20–22 but certain costs extend to patients as well. Facility costs (which include nursing costs) comprise an average of 34% to 42% of ED costs for all patients and average 50.9% of patient costs associated with ‘non-urgent’ ED visits.23 Since ED patients are charged for the duration of nursing care provided (not a ‘room and board’ charge like on inpatient units) lengthier nursing interventions equate directly to higher ED bills.24

Previous studies in non-healthcare settings have found that interruptions negatively impact work strategy,25 decision making26 and resumption lag,27 as well as lengthen task completion time.28 ,29 Such adverse effects can pose negative consequences to patient outcomes in high-risk healthcare situations such as emergency care and resuscitation. Measuring the impact of interruptions on task duration using observational studies in actual healthcare settings poses an intrinsic challenge: The longer the task the greater the probability it will be interrupted. This phenomenon is known as length-sampling bias.2 Controlling for this bias, a study by Westbrook et al30 found that ED physicians were able to complete interrupted tasks in less time than uninterrupted ones. It also found a high percentage of the interrupted tasks went uncompleted and some interrupted tasks were truncated to ‘catch-up;’ a practice which may pose its own negative consequences to patient safety. Interruptions, however, don't only affect physicians, but also nurses who are the healthcare provider with the greatest amount of direct patient contact.31 We sought to develop a better understanding of the effect(s) of interruption occurrence on the duration of complex activities for nurses (while excluding the time of the actual interruption). To do so, we investigated the impact of interruptions on the length of discreet, clearly defined (in written policy) clinical interventions performed regularly by nurses.


Study setting

This was an observational time study of common nursing interventions in the adult and paediatric ED of a large academic hospital: The Adult ED is a 67-bed facility comprised of 22 acute care, 2 critical care, 16 urgent care, 17 observation and 8 psychiatric beds. The paediatric ED is a 30-bed facility consisting of 22 acute care, 2 critical care, 3 triage and 3 psychiatric beds. The two EDs also share six robust trauma/resuscitation suites. Together the EDs serve over 104 000 patients annually. While these are two geographically adjoining facilities with different patient populations, their operational functions are similar and are considered one ED in this analysis.

A priori, we stratified all 91 documented ED nursing interventions using a factor which combined their duration and their frequency. For cost reasons, of the 91 interventions, the study focused only on the 35 that comprised ∼90% of the nursing minutes in the ED. These interventions were performed by either nurses or clinical technicians (who are staff that perform routine patient care and procedures of a specialised and technical nature under the direct supervision of a nurse). At the time of study, the ED was comprised of 161 nurse FTEs (full time equivalent) and 28 clinical technician FTEs. Each FTE represents 40-h-per-week of staffing. The study limited participation only to nurses and clinical technicians with a minimum of 12 months of work experience in their current position. The participants, including the patients, clinical technicians and nurses, were aware of and had the option to opt out of the study.

Personnel excluded from analysis were students, temporary employees or agency workers. Those procedures eligible for observation included those performed by eligible personnel. Procedures excluded from this analysis consisted of those performed during a disaster or involving a Joint Commission sentinel event; procedures that involved patients who required communications assistance; procedures that involved patients 18 year or older in the paediatric ED; and those patients under airborne, special or droplet isolation precautions. Observations excluded from analysis were those in which the observer made a recording error, the observer was interrupted or the observer had to engage in patient care to assist during the observation. The study period was from 10 December 2012 to 18 June 2013. This study was exempted from review by the Institutional Review Board.

Data collection

This study used direct observations of nurses and clinical technicians while delivering care to patients in the course of their ED clinical encounter over the course of the study period. Data collection was performed by one of five trained observers. All observers were ED-employed registered nurses who had an average of 19 years of working experience in the department (median 13 years, range 8–36 years). Prior to initiating the study, observers participated in a 90-min training session given by a senior Clinical Nurse Specialist who was a member of the research team. Observations were done in 4–8 h dedicated observation shifts that were randomised by location within the ED, by weekday versus weekend, and by time of day, in which the observer sought out eligible activities as they occurred in the course of routine patient care. No additional randomisation by ED census was conducted. To maximise utilisation of limited, and expensive, time study resources (ie, nurses conducting the observations), a stopping rule to the observation collection process was applied using a fixed coefficient of variation (Cv). Use of coefficient of variation as a stopping rule has been described elsewhere.32–34 The study, a priori used the stopping rule of Cv≤50% described by Köbrich.32 Thus, interventions with high variability were subjected to additional sampling.

Observers were provided detailed written definitions of each nursing intervention to ensure consistency in the start and stop times of their observations. Similar to Westbrook et al,30 interruptions were defined as situations where a nurse ceased a current intervention in order to attend to an external stimulus. This study categorised interruptions into five types: Family/Patient Interruption, Staff Interruption, Phone/Pager Interruption, Another Procedure Interruption and Other Interruption (table 1).

Table 1

Interruption types and definitions

Intervention timing was accomplished using TimeStudy (V.1.3 nuVizz). This software enabled observers to select the nursing intervention type, capture the intervention start, pause the timer whenever clinical care was interrupted while simultaneously specifying the interruption type, and resume the procedure time when appropriate. For each observation, a reference number, the date and time of the observation, the observer's unique identifier, the intervention type, total time to complete the intervention (excluding the interruption time), and the cumulative length and type of each interruption was recorded. Multiple interruptions of the same type were not counted as separate interruptions.

In addition to intervention time, other variables were also recorded because of assumed importance to the effect of an interruption. These were shift, location, employee type, a patient modifier and department. It was hypothesised that nursing interruptions would be more frequent during busier shifts. The ED functions across three primary 8-h shifts (day, evening and night). Day shifts generally see the highest arrival rates, but evening shifts tend to be busiest due to the build-up of arrivals over the course of the day. It was thought that location may be associated with intervention time, so location was added as a variable. From the location anticipated to have the fastest intervention time to the slowest, these were (1) trauma room, (2) triage, (3) rapid assessment room, (4) patient room, and (5) psychiatric room. It was hypothesised that the length of interventions might vary by employee type. Employees providing the intervention were categorised as nurse or clinical technician. To better understand the impact certain patient populations have on the length of interventions, a patient modifier was added. Data was collected for a narrow scope of patients whose clinical interventions were hypothesised to take longer than patients not in this scope. Patients whose ability or willingness to follow instructions was compromised were classified as: combative, neurologically devastated or intoxicated (table 2). For neurologically devastated patients, only paediatric cases were considered. The final variable was the department in which the intervention occurred. It was hypothesised that interventions in a paediatric setting would take longer than in an adult setting. The rationale for this was that paediatric patients (especially younger ones) follow directions less often, are less likely to remain still during an intervention, are more likely to interrupt the caregiver, and that their parents are more likely to interrupt the nurse or technician with questions than the family member of an adult patient.

Table 2

Patient modifiers and definitions

Data analysis

Unpooled two sample t tests, multiple linear regression analysis, and quantile regression analysis were conducted to investigate whether occurrences of interruptions had an impact on the duration of nursing interventions. Welch's t test ascertained whether the average length of uninterrupted interventions differed from interventions that were interrupted at least once. Multiple linear regression analysis was used to quantify the impact of an interruption on an intervention controlling for intervention type, observer, shift, location, employee type, patient modifier and department. Relative differences in the effect of an intervention were calculated as the percentage difference from the mean of the intervention length by interruption type.

While linear regression provides the effect of interruptions on the mean intervention time, quantile regression allows for an examination of the effects of the independent variables across the entire distribution of intervention times. By examining the effect of interruptions across the conditional distribution of intervention times, the effect of length-sampling bias can be quantified, which would be expected to be more of an issue the longer the intervention. The dependent and independent variables of this analysis were identical to those of the multiple linear regression analysis described above.

The occurrence of each interruption type and its effect on intervention time was also examined. The type of interruption replaced the occurrence of an interruption in the regression defined above, otherwise all other control variables remained the same. All statistical analyses were performed with R V.3.1.0 for Windows.35 The quantreg package36 was used to perform quantile regression analysis.


Across 7 months of observation, the study captured 904 individual observations for 35 interventions. Of these 35 interventions assessed, 28 interventions had interrupted and uninterrupted observations. These 28 interventions (see table 3 and online supplementary table S1) consisted of 723 unique observations, of which 273 (37.76%) were interrupted at least once. These 723 observations reflect 88.8 h of clinical care time. Categorising the interruptions yielded the following: 63 observations with staff interruption, 70 with family/patient interruption, 22 with phone/pager interruption, 23 with another procedure interruption and 170 categorised as other interruption.

Table 3

Interrupted intervention types observed

The average time spent on an uninterrupted intervention was 232.64 s (median, 114.18, SD, 296.26), for all 35 procedures, and 296.47 s (185.15, SD: 319.05) for the uninterrupted interventions of the 28 interventions that had interrupted and uninterrupted interventions. In comparison, the time spent on an intervention with an interruption was significantly different (p<0.01) taking an average of 682.02 s (589.63, SD: 504.59) (figure 1). This is an unadjusted increase of 130%. Standardising by intervention type, we found that the average relative length of an intervention without an interruption was significantly shorter than average, −7.6% (95% CI −12.2% to −3.0%), while for interventions with an interruption it was significantly longer, 12.5% (95% CI 6.2% to 18.8%).

Figure 1

Intervention time plots. The bands inside the box plots represent the median times for interventions. (A) Is the difference between the time spent on an uninterrupted intervention and the time spent on an intervention with an interruption, while (B) describes the relative difference from the mean of an intervention length. The ends of the plots (‘whiskers’) represent the shortest intervention time within 1.5 IQR of the lower quartile, and the longest intervention time within 1.5 IQR of the upper quartile.

Multiple linear regression analysis (table 4) also found that the occurrence of an interruption had a significant effect (p<0.001) on the duration of an intervention, and that interventions with an interruption were 121.36 s longer on average (95% CI 79.57 to 163.15) or a 19 percentage point increase (95% CI 11.31 to 26.89) over the mean length of an intervention.

Table 4

Multiple linear regression analysis

While the results suggested that clinical technicians tended to complete interventions faster than the nurses, no effect on time was noted based on shift or department. Interventions performed in trauma rooms (the reference group) tended to be the fastest, while the nursing interventions performed in patient rooms and psychiatric rooms were significantly slower. Of the patient types, while intervention times for neurologically devastated patients were significantly longer, the percentage increase was not statistically significant. For intoxicated patients the inverse was true.

Quantile regression analysis (figure 2) demonstrated that there is a consistent positive effect of interruptions across the distribution of intervention times (see online supplementary figure S1 for distribution). The effect of interruptions on interventions increased with the length of the intervention. Shorter interventions, or those less than 7 min in length, were increased by about 1 min, while those above 11 min long experienced longer delays.

Figure 2

Quantile regression result for interruption coefficient. Quantile regression analysis demonstrated that there is a consistent positive effect of interruptions across the distribution of intervention times. However, the positive slope in the figure shows that longer interventions tend to experience longer delays. On average, interventions that were interrupted were 121.36 s longer, represented by the solid horizontal line bisecting the figure. The dashed lines denote the upper and lower bounds (79.57 s and 163.15 s, respectively) of a 95% CI around this mean.

The type of interruption also had a significant effect on the duration of a nursing intervention (see online supplementary table S2), with the occurrence of family/patient interruptions having the greatest effect on the duration of an intervention increasing intervention lengths by 184.75 s (95% CI 122.42 to 247.07). The occurrence of staff interruptions affected the intervention time the least increasing the intervention length only 89.10 s (95% CI 25.83 to 152.37). The significances of other variables on the intervention times were consistent with the first regression analysis.


Accounting for length-sampling bias, the outcomes of this study suggest that interruptions increased the duration of an intervention (excluding the duration of the interruption from the intervention time). While this study design did not allow for an explanation of the cause of this phenomenon, previous controlled laboratory design studies in non-healthcare domains have suggested a similar relationship.28 ,29 ,37 A study on interruptions by Altmann and Trafton27 might explain this phenomenon. Their research suggested that an interruption causes a resumption lag (the time needed to ‘collect one's thoughts’ and restart a task after an interruption). Thus, when care delivery is interrupted, the intervention time would be longer as a result of this lag. This likely explains only some of the observed increase in time as such lags are short and interruptions do not always yield resumption lag.38 Other studies that have suggested causality or association between interruptions and procedural error5 ,39 ,40 might also inform an explanation of this phenomenon. It is conceivable that clinical staff who recognised the error during the intervention period decided to redo some part of the intervention, thus increasing its length. Alternatively, in an effort to avoid errors (rather than correcting them), nurses may perform task steps again41 increasing intervention time. One such example we observed, after being interrupted while setting up an infusion pump, a nurse would repeat the last few pump inputs so as not to miss one accidentally. Further, nurses may revalidate task completion post interruption; for example confirming the accuracy of critical pump inputs such as medication concentration and infusion rate.

The quantile regression analysis revealed that longer interventions tend to experience longer delays. This is likely due to length-sampling bias; in other words, tasks with longer durations have a higher probability of being interrupted and of being interrupted more frequently. When aligned with our previous analysis suggesting that interruptions increased nursing intervention time, the delay time would continue increasing as the interventions experienced more interruptions. This hypothesis could not be tested because the experimental design did not record the number of same-type interruptions that occurred during each observation. However, by normalising the data to be a percentage increase relative to the mean of that procedure, we find similar results to the absolute change, suggesting indeed that interruptions greatly increase the intervention time (excluding interruption time) regardless of the length of the intervention.

Some interruptions have intrinsic value or purpose while others do not; yet both types are potentially harmful.1 Efforts to reduce interruptions in clinical settings have demonstrated mixed outcomes.1 Making clinical care more resilient to the deleterious effects of interruptions may add more utility than efforts to diminish their frequency. A few strategies that have demonstrated success are: (1) Using checklists coupled with ‘explicit hold’ techniques (marking the interruption point so that checklist performance resumes at that step, avoiding repetition or revalidation) for critical, lengthy or complex nursing interventions;41–43 (2) training clinical staff to complete a task or subtask within a procedure before attending to an interruption which minimises the disruptive effects of an interruption;44 ,45 and (3) using ‘behavioural strategies’ to improve task resumption, such as physical or environmental cues (eg, holding ECG leads while addressing an interruption) employed as reminders to return nurses to their primary intervention quickly and reliably.38 ,41

Given the rigorous training performed a priori, all observers should have recorded the intervention time to the same standard. The regression model, however, indicated that Observer 2's observations were significantly longer than the reference group (which was Observer 1). Further investigation revealed that Observer 2 recorded only 16 different types of interventions and 40 individual observations, a relatively low number compared with those by other observers. A Welch two sample t test was conducted to compare the times recorded by Observer 2 and those recorded by other observers, considering only the 16 intervention types. The test failed to reject (p=0.4109) that the average intervention time recorded by Observer 2 was higher, indicating this observer recorded intervention types that tended to take longer and all observers actually performed the time study to the same standard.

The results of this study are not consistent with the study by Westbrook et al, which explored the effect of interruption occurrence on physicians’ task completion time in the ED. It revealed that interrupted tasks were actually completed in a shorter time than uninterrupted tasks.30 There are some possible explanations for this apparent discrepancy. First, Westbrook et al's30 adjustment for length-sampling bias incorporates a simplified assumption that interruptions occur according to a ‘homogeneous Poisson process’. In contrast, our study used quantile regression and did not introduce any external interference to the data. Second, while both studies were conducted in the ED setting, one focused on physician interventions during weekdays while the other on nursing interventions at all times of operation. Westbrook et al hypothesised that interrupted physicians compensate by working faster and by cutting corners, a hypothesis that was also suggested by Mark, Gudith and Klocke.25 This might not be the route that nurses and clinical technicians took because, in contrast to physicians who exercise more discretion in their practice, nursing practice is more protocol driven.46–48 The performance of nursing interventions are substantially driven by explicit, stepwise protocols which may preclude one's initiative to ‘cut corners’. Detailed, step-by-step criteria for each intervention measured in this study exist as part of this ED's standards of practice. Timmermans and Berg found that nurses tend “…to hold more systematized, less individualistic, conceptions of clinical work than doctors and appeared to be more fastidious in adhering to documented procedures.”47 In addition, the presence of observers, themselves senior nurses in the department, may have caused observed nurses to comply with protocols even more strictly out of concern for this scrutiny.46 ,48–50 Perhaps for fear of being labelled a ‘bad nurse’50 for not following procedures, nurses did not appear to compromise intervention steps despite the presence of interruptions.

Contrary to physicians, nurses tend to emphasise adherence to standardised protocols and view guidelines as essential in providing safe, high quality care.46–48 ,51 When examining how clinicians cope with interruptions, researchers should consider nurses’ and physicians’ divergent views on protocol adherence. Research suggests these professions view clinical practice in disparate ways and that they ascribe heterogeneous values to the utility of procedural guidelines in informing their own clinical practice.52 ,53 Such unequal approaches may impact how these groups contend with interruptions to patient care.


The outcome of interest in this study was the intervention time. It failed to consider another important piece of nursing practice—the quality of the care being delivered. Future research should evaluate patient safety and outcomes as interruptions increase.

A further limitation was this study's focus on only 28 emergency nursing interventions within a single hospital, confining the generalisability of the findings. However, results are fairly generalisable in terms of facility layout, nurse-to-patient ratio, supply type and availability, and other generic ED features, because these factors were controlled for via their inclusion in the interruption time rather than in the intervention time. In other words, there should be no significant difference between one ED in which a nurse has to walk a lengthy distance to gather supplies from a locked cabinet and one in which supplies reside in an unlocked cabinet right outside the patient room. Our study design classified the non-value-added time spent walking to and unlocking the supplies as an interruption and not a component of patient care.


Consistent with the outcomes of studies in non-healthcare domains, this research demonstrates that the presence of interruptions had a negative impact on the duration of nursing care in the ED. On average, clinical interventions that were interrupted tended to be about 2 min longer than those that were not. This experiment also revealed that longer interventions tend to experience longer delays. Furthermore, this study suggested that the occurrence of family/patient interruption had the greatest effect on the time to complete an intervention. While seemingly trivial, in an emergency care setting, additional minutes translate into increased patient risk. When examined across the dozens of interventions a nurse performs during an 8 h shift, this increased time and risk can become significant. From a financial perspective, additional nursing time translates to additional costs for the patient. In the ED, nursing time is billed directly. Several minutes of extra nursing time, can increase the patient's hospital bill significantly. Likewise, if a procedure is interrupted and must be subsequently repeated, the patient's bill may increase. Organisationally, such interruptions lead to inefficient practice and a drag on productivity. Combined, these factors should provide some incentive to clinical leaders and administrators to explore methods to reduce the interruptions that are most detrimental to care delivery, and, perhaps more realistically, implement methods to make nursing practice more resilient to the effects of interruptions.


View Abstract

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.


  • Contributors GC and HG designed and oversaw the study; DS, EYK and GC performed the statistical analysis; all authors contributed to the writing of the manuscript.

  • Competing interests None declared.

  • Ethics approval Johns Hopkins Bloomberg School of Public Health Institutional Review Board.

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

  • Data sharing statement All data is available upon request to the corresponding author.