Nurses spent approximately one third of their time with patients and this did not change over time. A subset of this is the provision of direct care which significantly increased from 20% in year 1 to 25% in year 3. There are surprisingly few studies which have sought to quantify the amount of time nurses spend in direct care activities with patients and we have identified no study which has examined changes over time. Hendrich et al  in a study of multiple units at Kaiser Permanente in the United States reported an average of 19.3% of nurses' time (approximately 81 minutes per shift) was spent with patients. Using a diary method among 30 nurses in a Swedish hospital, Furaker  found around 38% of nurses' time was spent with patients. In year 3 the nurses in our study had moved to allocation of patients to nursing teams, but this appears to have had no effect on proportion of time spent with patients.
A central question is the extent to which this amount of time ensures safe care. Surveillance of patients by nurses has been identified as important to detect patients who are deteriorating. Research by Aiken and colleagues  has highlighted the relationship between nurse surveillance and patient safety. Surveillance relies on frequent interactions to be able to constantly monitor patients' conditions and provide opportunities to respond. On average we found each direct care task consumed approximately 80 seconds, and in an average hour nurses performed approximately 10 direct care tasks. However we were unable to assess how these tasks were distributed and this is likely to make a substantial difference to patient care. For example, 10 direct care tasks completed in quick succession leaves patients with no nurse contact for the remainder of the hour. However 10 tasks distributed evenly across the hour would provide much greater opportunity for surveillance. Further work is underway to develop methods to assess the sequencing of task distribution.
Few researchers have reported the amount of time which nurses spend on individual tasks. Nurses' work was characterised by a pattern of rapidly changing short tasks. Our findings are consistent with available evidence and suggest a general trend in the nature of nursing work on hospital wards. On average nurses in our study changed tasks every 55 seconds. Cornell et al examined time spent in 29 task categories in a direct observational study on two wards and reported a similar high rate of task-switching with an average of 88 tasks per hour. We grouped work into 10 broad tasks and found a rate of 72 tasks per hour. Cornell et al also found task length was short with only 5% of tasks lasting longer than two minutes. The implications of this rapid task changing activity in real-world settings have been underexplored. Experimental studies demonstrate that task-switching leads to increased errors and slower task performance [19, 27–29]. One of the posited reasons for the slower performance when task-switching occurs is the cognitive effort required in reconfiguring the taskset which can involve both shifting attention to the new task while also inhibiting attention to a previous task . Importantly these 'switch costs' have been shown to occur regardless of the participant's familiarity or training in the tasks performed . The availability of preparation time prior to a task-switch has been shown in some cases to reduce switch costs [28, 31]. The rapidity of task-switching found in the present study suggests nurses receive limited time to prepare for new tasks.
Our results demonstrate the reliance that nurses have on formal information sources to complete their work. Around 30% of their time involved tasks where formal paper records were used and this proportion increased over time. This was not due to greater demands on nurses to document information, as time spent in documentation did not increase over the two year period. The increased reliance on formal information sources may be a response to a decrease in access to information from other sources given that there was a significant fall in face to face professional communication and an increase in time nurses spent completing tasks alone. Use of computers constituted a very small amount of nurses' overall work, but increased over time. It is likely that with the introduction of greater computerisation, for example with computerised medication management and clinical documentation systems, time spent completing tasks with a computer will increase substantially.
A significant decline in time spent in transit was found. This may be related to greater access to computerised information sources and a decrease in seeking information face to face (as evidenced by the reduction in professional communication) both of which reduce the need to travel to obtain information. However without a focused study it is not possible to confirm the role of these factors in this result.
We found that nurses spent approximately 6% of their time multi-tasking and experienced approximately two interruptions per hour. Multi-tasking is an important component of health professionals' work and Australian doctors have been found to spend 20% of their time multi-tasking . The majority of multi-tasking involved communication with patients or other health professionals and is a required feature of health care work which has rarely been quantified. Along with the results about interruptions it adds further evidence of the non-linear nature of clinical work.
Most concerning was that the highest proportion (27%) of all interruptions occurred during medication tasks. Further, 25% of medication tasks were undertaken in parallel with another task, most frequently professional communication. Kosit et al , in a study of interruptions in an emergency department, reported nurses were interrupted on average of 3.3 times per hour and that the highest proportion of interruptions (27%) occurred during medication tasks, as we found. Interruptions during medication tasks have been shown to be directly associated with the rate and severity of medication administration errors by nurses . While nurses experienced rates of interruptions lower than their medical colleagues[15, 19], their concentration during medication tasks suggests this task is at specific risk and interventions to reduce interruptions during this process are required [34, 35].
No previous studies have followed nurses' work patterns over time. We found four broad categories of tasks consumed close to 80% of nurses' time (direct care, indirect care, professional communication, and medication tasks) in both periods. While the proportion of time in direct and indirect care increased, the amount in professional communication decreased significantly from 24% to 19% and the average time per communication task almost halved from 59 seconds to 33. While time spent in documentation decreased, but not significantly, (from 10% to 7% in year 3), the average time spent in each documentation task became significantly shorter (an average of 92 seconds to 33). Both these changes in the shortening of communication activities may be related to an increased reliance upon electronic communication and the greater use of clinical information systems reducing both the level of verbal communication required and the amount of documentation. While medication task time overall did not change, and continued to consume around 19% of nurses' time, the average medication task increased a small, though significant amount from an average of 42 seconds to 47 seconds per task. This may reflect the increased complexity of medication management among hospitalised patients requiring additional time, particularly in administration of medications. However whether this is an adequate amount of time is unclear. Research conducted by our team at this site on several wards, including the study wards, demonstrated high medication administration error rates and poor compliance with some medication administration procedures[33, 36]. For example, direct observation revealed that in less than 50% of administrations did nurses correctly check patients' identification prior to drug administration. The extent to which this reflects intentional deviation of practice or a response to time pressures is unknown. However there is good evidence that current practice is resulting in a high rate of medication administration errors [33, 36].
The results provide little support for an increase in the amount of inter-disciplinary care or communication over time. Nurses experienced a dramatic increase in time spent completing tasks alone, from 28% (average 2.5 hours per shift) to 39% (average 3.5 hours per shift). This was largely due to a significant decline in the time spent with other nurses which fell from 54% to 41% of nurses' time. Interestingly, the results suggest that the requirement that certain tasks be completed with a colleague (the number of tasks completed with others did not change) may have led to nurses in year 3 completing joint tasks in significantly shorter times than in year 1. The average time for collaborative tasks with another nurse fell from 80.6 seconds to 53.9 seconds. This may reflect changes in nursing practice and/or compensation for a decrease in the availability of other nurses. For example, the move to team based allocation of patients may have led to nurses having a smaller pool of colleagues (ie those in their team) from whom to seek assistance in year 3. The amount of time nurses spent in professional communication significantly declined. There was little change in collaboration or communication with other health professionals which remained at very low levels. Cornell  also reported low levels of interactions between nurses and non-nursing colleagues making up around 2.8% of their time. On average nurses in our study spent approximately 3.6% of their time (18.4 minutes per shift) completing tasks with a doctor, while Cornell reported nurses in her US hospital spent only 0.5% (approximately 2.6 minutes per shift) of their time with doctors. Thus while the literature on the value of improved inter-disciplinary communication expands, our results suggest no evidence of increased interaction. Nurses on our study wards did not increase their level of engagement with other professionals. Further, the amount of time they worked collaboratively with other nurses substantially declined. This occurred in the context of both wards moving to a team-based nursing model. The impact of decreased collaborative task completion on care provision in terms of quality or efficiency is unknown and is worthy of consideration in future studies. While our study did not measure the content or quality of communication, the finding that the average length of a professional communication task almost halved between years 1 and 3 (from 59 seconds to 33) suggests little time is available for detailed information exchange about patient care.
The results reflect work patterns on two wards at one hospital and thus may not generalise to other hospitals with very different nursing practices. Our study examined weekday work. The results may not be representative of evenings or weekends. We used a direct observational approach, and while nurses may have changed their behaviours because they were being observed, the likelihood of dramatic change is low due to the extended length of the study, reducing the chance of sustained behavioural change on busy hospital wards. Observational studies of clinicians in-situ have suggested that the extent of behaviour change is minimal [17, 38, 39]. Strengths of our study include the longitudinal study design, consistency of methods and the data collection technique which accounted for multi-tasking, all of which have extended previous work in this area.