Fragmentation in Care and the Potential for Human Error

Doctors rely on information provided by nurses to make treatment decisions. This information is usually documented but most of the time, according to our observations, it is provided verbally by the nurse either voluntarily or upon the doctor's request. At the same time, doctors may possess some types of information that need to be shared with the nurse. This includes, diagnostic or prognostic information and information about treatment goals and reasons underlying treatments, which could help the nurse to deliver care more efficiently and be more alert to changes in the baby's condition. Cases were reported where Doctors had stipulated care and treatment actions without notifying nurses of their motivations. For example, to enable the baby to develop self-sufficiency, it is sometimes necessary to withdraw some aspects of treatment to prompt the baby to cope. However, this will subject the baby to temporary discomfort. Where the nurse is unaware of the Doctors motivation in this respect, he/she may act in a way to alleviate discomfort, by adjusting sedation, for example. In this way, the Doctor’s strategy for treatment is compromised.

Relationships between staff appear to owe more to the traditional roles of doctors and nurses than to any rational assessment of how such relationships and lines of communication should best be organised and supported in neonatal intensive care. Thus, certain demarcations in responsibility are assumed, with nurses ‘nursing and caring’, and doctors ‘diagnosing and planning treatment’. In some cases, the distinction is explicitly stated, with the implication that nurses have little discretion.

In reality, these distinctions are blurred, with both types of staff engaging in both types of activity. For example, nurses may facilitate diagnosis and treatment planning by anticipating and collecting relevant information in order to inform doctors. In one incident reported to us, an experienced nurse was prompted to question the validity of a ‘normal’ blood oxygen reading. Taken at face value, the baby’s health was entirely satisfactory. However, this nurse had taken note of prior treatment interventions and had noted that four hours earlier, on the previous shift, the baby had been given an increase in oxygen and no change to this had been made since then. The Nurse extrapolated from the current condition to the next occasion for routine monitoring and was concerned that the baby would be at risk from excess oxygen. She had no discretion to act directly, but was at liberty to call the Consultant who then made the appropriate adjustment.

Equally, doctors must appreciate the importance of 'minimal handling' and ensure that treatment interventions do not compromise recovery. Minimal handling is a care principle of utmost importance for nurses because it is this that allows the baby to rest — this is one of the most effective means by which a baby can recover. Specialist Consultants and Registrars appreciate this, but the less experienced SHOs may not. It is clear that there can be conflict between the inexperienced SHO and the experienced Nurse. Effective senior nurses are strong characters indeed!

If this anticipation and co-operation did not occur, then it would be difficult for NICU teams to work so effectively. Therefore, examination of the organisation of NICU teams and the information infrastructure that supports them is important in resolving performance problems.

Cooperation depends upon communication. There are a number opportunities or channels for communication.

Shift handover: Shift handovers occur between nurses going off shift handing over responsibilities to other nurses and doctors going off shift handing over responsibilities to other doctors. The role of shift handover is to facilitate a smooth transition from one shift to the next without interruption in care. It is intended to help oncoming staff quickly to update their knowledge of the babies and to identify changes that occurred since they were last on shift. It can helps them to put the babies in order of severity and to prioritise attention. It can provide information on things to look out for, on unusual and important events that occurred during the previous shift, events that may have been recorded but may be missed amidst the detail of other recorded information. It can provide a summary of what needs to be done and of any changes in the care plan that were decided during the previous shift.

However, there are risks inherent in handover. For example, some important information may be omitted, there may be ambiguity in the verbal communication leading to misperceptions, erroneous inferences and suppositions. Biases may be introduced in the information search of the oncoming shift (Grusenmeyer, 1991). The outgoing shift may fail to direct the oncoming shift's attention to important events and aspects of care. The oncoming shift may fail to memorise or take into account some transmitted information or may fail to check the validity and completeness of the transmitted information before starting care activities and continue with a wrong practice ‘inherited’ from the previous shift (Grusenmeyer, 1995).

In some of the NICUs that we observed, handover was done from memory and neither the outgoing nor the incoming shift kept notes. Moreover, oncoming nurses often started their care activities without consulting the records. Shift change is indeed a critical phase of work in terms of safety; a higher number of accidents have been found during shift changeover in other contexts than during the shift or at the end of the shift (Godimus et al., 1977, cited in Grusenmeyer, 1991). Gopher et al. (1989) found peaks of errors immediately following each shift change of nurses in a respiratory intensive care unit. According to Grusenmeyer (1991), there is a period when the oncoming operator is trying to gain information and is not "fully operational" for his tasks which increases the likelihood of incidents.

Our observations of handovers revealed a vast range of personal styles that were difficult to assess from the viewpoint of their adequacy. A major problem that arose was that staff were conscious of being watched and tended to be more comprehensive that they might otherwise have been — this was as much to help inform us than simply that they were on their best behaviour. We witnessed a wide range of styles. In some cases, a nurse would systematically work through a standard template of observations, while in other cases the nurse would be briefer. Sometimes the on-coming nurse would ask questions. Sometimes the nurse coming on shift was already familiar with the baby and little had changed. As babies become more stable and staff are more confidant of their prognosis, the handover communication centred more around the families and their increasing involvement in treatment. Sometimes, very little information would be conveyed and the on-coming nurse would move into a nursing routine that revealed nothing of the previous exchange. This suggested, that in many cases, the handover served little purpose in terms of obtaining information. Indeed handover could serve simply as a ritual to enable people to engage and disengage with the task or with coming to work. As a means of providing information, it could exist to confirm that all is as would be expected for a baby of these characteristics. It could provide specific information about crises or any observations concerning the baby’s health or response to specific treatments.

Ward rounds: Ward rounds are a different context for information exchange in which all staff concerned with a baby are present. They occur daily and involve doctors going to each cot, reviewing and discussing the case, examining the baby and deciding on whether revisions to the treatment plan are warranted. The nurse in charge of the baby is usually present to provide information and to comment on the baby's progress. The doctors write the new treatment plan in the medical notes and the nurse updates the care plan in the light of the doctors' latest instructions.

Substantial verbal exchange of information takes place during ward rounds but only part of it is recorded. Doctors rarely record the reasons behind their decisions, although they may communicate them verbally to the nurse in charge of the baby or other doctors present. The nurse may not record the doctor's reasons in the nursing notes (it is not a type of information that she is required to record) and may not inform her shift successor about them (because she forgot or did not consider it important). Verbal information can easily get lost and staff often commented on this. Of course, people cannot record everything and read everything. It would be too time consuming and inefficient. Therefore, judging which information is important and would be useful to the next person coming on shift seems to be an essential skill in intensive care.

NICU babies require continuous monitoring, as they can quickly become very unstable and their appearance may not sufficiently indicate an impeding deterioration. Moreover, only a limited number of invasive tests can be made on neonates, which makes continuous, non-invasive monitoring even more important (Cunningham et al., 1992). The interdependency of staff in the delivery of care suggests that if the forms of communication exchange were in some way deficient, then care could suffer. In the description of NICUs given above it is clear that care is fragmented as staff change shift and as they engage at different stages of the baby’s treatment and care.

Fragmentation of care between doctors and nurses can also reflect their respective traditions and philosophies. For example, some doctors are not concerned with how the baby is establishing feeds; this is considered one of the nurses' main activity. Nurses are interested in maintaining the baby's comfort, whilst doctors are more interested in delivering treatment. Delivering treatment often compromises the baby's comfort, providing a constant source of tension between doctors and nurses. This gap between doctors and nurses is reflected in separate handovers and separate records which result in overlapping information but also in valuable information not getting across, simply because staff think that colleagues are not interested. Some members of both the nursing and medical staff have pointed out that it is necessary to have joint doctor-nurse treatment decisions. This suggests that they acknowledge the fragmentation in care and its potential consequences.

In cases where demarcation within teams has evolved through local practices, it is difficult to predict where performance failures due to inadequate staff interactions may arise. What constitutes an individual task is unclear, as is the relationship between individual tasks. This is made more complicated due to local variations in team organisation and relationships. To take account of these variations, we have adopted the strategy of first examining the general function of delivering neonatal intensive care and then seeing how this may be resolved into different tasks. This has been done using Hierarchical Task Analysis (Annett et al. 1971; Shepherd 1995, 1997). Hierarchical Task Analysis is a method of describing tasks in terms of the operations that aim to meet functional goals and plans which describe the conditions under which goals must be met to satisfy the conditions of their superordinate goal.

Figure 1 shows the top level of this analysis illustrating how staff must take information from the previous shift (1) to inform their principle activities of maintaining equipment (2) and maintaining care and treatment (3). At an appropriate stage of the day, the wardround is conducted to review each babies development and to enable information exchange to take place between different members of the team. Finally, responsibility must be handed over to the next shift (5).

The importance of this part of the analysis is to set the context in which care and treatment are delivered. Indeed, the elements of handover (1 & 5) cannot be understood without reference to what goes on within care and treatment (3). The general nature of the model at this level reflects how different members of staff may be engaged. Thus all NICU staff are concerned with handover, care and treatment and being engaged in ward rounds. We have included the operation concerned with maintaining equipment and supplies. If this is not done to a satisfactory standard, then the standards of treatment and care are at risk. We observed a critical event where a piece of equipment was called for during an emergency. First the technician bringing it went to the maternity ward instead of the NICU, thereby causing delay. Then, when the equipment arrived it was found to be unservicable and an alternative strategy for dealing with the problem had to be pursued. The potential consequences of this sort of event do not need to be spelled out. Clearly all staff must issue unambiguous instructions to each other and certain staff must be given responsibility to make routine investigations to guarantee the availability of such crucial equipment. This applies whether or not these staff are members of the NICU or other staff within the hospital.

The central activities that we wish to address are described in Figure 2. This shows how the tasks involved in ‘Care for and treat babies in NICUs’ (3) are set out. These functions are reasonably standard and would apply to most NICUs irrespective of their specific organisation.

The Principal Goals of Managing Care and Traetment: We have focused on the aspects of treatment and care provided in NICUs. This is seen as the external control of immature organ systems to facilitate development to a state where the baby may be safely discharged, transferred to another unit or be terminally cared for. For example, the babies may not be able to sustain their cardiorespiratory function and regulate their temperature and fluid homeostasis and, therefore, require assistance. Such control should be administered in a way that does not damage the baby's organ systems (e.g., lung or eye damage due to excessive mechanical ventilation is a common side effect of NICU treatment). External control should be gradually reduced and eventually removed, so that the baby can regulate its own bodily functions. Both provision of and weaning from support require careful assessment, continuous monitoring and adjustment. Moreover, they require dealing with routine deviations as well as tackling unexpected problems.

Figure 2 shows how care and treatment depends upon a cycle of activities including monitoring key indicators of the baby's condition and monitoring the adequacy of current care and treatment plans; delivering routine care and treatment; treatment planning; and communication. Some task elements, such as the monitoring tasks must be conducted throughout. Others are done in response to certain events, for example, the baby entering the ward or various unacceptable conditions arising.

For the purpose of discussion, we have categorised the types of tasks identified by the task analysis as procedural tasks, monitoring tasks, care and treatment planning tasks, care management support and communication tasks.

Procedural : These tasks are ‘settle the baby and make initial measurements’ (1) and ‘carry out treatment/care and specified monitoring activities according to current schedule’ (7). Care and treatment planning tasks are proceduralised to a large extent, due to the difficulty of diagnosing and treating highly unstable, highly interacting organ systems. Most of the time, treatment consists of selecting an established method and appropriately adapting it to accommodate individual differences and ongoing responses (Shalin & Bertram, 1996).

The word ‘procedural’ denotes tasks that are generally executed in a fixed way and are often straightforward, requiring little discretion. Procedural tasks in the context of neonatal intensive care include obtaining routine measurements such as temperatures, obtaining blood samples, administering drugs, delivering oxygen, giving fluids, inserting a catheter, intubating/extubating, suctioning etc. These are the ‘normal’ routines of nursing and treatment delivery. On further analysis it becomes clear that each of these is subject to variation, some of which requires judgements to be made by the nurse or SHO carrying out the task and some of which require information to be supplied by the doctor responsible for the treatment plan. Invasive treatments or the administration of drugs to alleviate pain, for example, will yield different responses from different babies. If a nurse recognises that one method of treatment is unsuitable for a particular baby, then an alternative method for achieving the same nursing goal must be found. The nurse will then know what to do on the next occasion and should inform colleagues to ensure that the baby is not discomforted unnecessarily. Thus the team can learn about the characteristics of individual babies. It soon becomes obvious that these ‘procedural’ tasks all rely on information being provided by colleagues to ensure they are dealt with most effectively. It is also clear that each task provides opportunities for staff to learn new things about caring for and treating the baby to communicate to other members of the team.

It must also be acknowledged, though, that some treatments are intended to be followed irrespective of some temporary discomfort to the baby. In these cases, the routes to follow must be made explicit to staff and reasons given. This must be done in a way in which instructions and motivations can be conveyed beyond the members of staff who were present when the original decision was taken. If this is not done, then nursing staff may promote comfort at the expense of the treatmentplan.

Monitoring: Three types of monitoring are identified. First, ‘maintain baby’s comfort’ (8), is concerned with nursing decisions. Nurses will ensure that babies are not too cold or have been discomforted by an item of equipment slipping out of position, for example.

Second, nurses are engaged in reviewing and evaluating whether the baby is progressing satisfactorily in accordance with the current treatment plan’ (9). Here, their job is to make general observations of the baby and to take various measurements from which judgements are made about the baby’s progress. They will record the information or draw problems to the attention of a doctor. At one level this task appears straightforward. The nurse needs to know the target parameter, the target value and the tolerances to be accepted. It can be assumed that there is little judgement and expertise involved, a view held by some doctors. In practice, monitoring can be a complex task. Where a nurse is heavily loaded, he or she must prioritise and, therefore, needs to understand factors associated with criticality and recoverability in order to direct attention in an optimal way. Where a parameter is seen to go outside acceptable tolerance, the nurse must investigate the problem. Is a poor pulse signal the result of the pulse sensor being dislodged by the baby moving or is there general concern? Is the current acceptable reading of oxygen in the blood the sign of a healthy baby or is it a transitory state following an earlier intervention? If the nurse failed to anticipate these problems or failed to account of the factors that would direct an effective schedule of monitoring behaviour, then the nurse would be less effective, despite what doctors might say about the level of discretion a nurse has.

The third type of monitoring is ‘monitor adequacy of current treatment plan’ (3). If the baby’s response to treatment is unsatisfactory, then the doctor might decide to be more cautious and revise the treatment plan. Alternatively, the baby might be progressing well and the doctor decides that treatment must move forward. For example, the nurse might notice that the baby is making spontaneous attempts to breathe and so the doctor might decide the baby should be weaned off the ventilator. Where such action is taken, the treatment plan is revised. The nurse needs to know this so that greater attention can be given to the baby during this critical phase of change and because other vital signs will be disturbed caused by anxiety brought on by the change.

Care and treatment planning: These tasks include ‘specify treatment/care plan’ (2) and ‘revise treatment/care plan’ (4). The importance of the process of prescribing a treatment plan has been made clear. This is the responsibility of doctors. Their task is to make medical judgements, either to identify a problem or to prescribe what to do next. Much neonatal care treatment planning is done in conditions of uncertainty. Because of the fragility of the organism and the idiosyncratic nature of some problems, doctors are often uncertain. Their task is not necessarily one of diagnosing, then confidently prescribing. They are often making their best judgement about what treatment will help stabilise the baby and allow it to settle before progressing the treatment. As a consequence, they must rely upon the effectiveness of the nursing staff in obtaining information and of nursing staff and SHOs in delivering care and treatment. Paramount, of course is the effective monitoring in between to ensure that treatment is progressing well. This means that unless nurses and SHO fully understand the motivations of doctors in prescribing a course of treatment and unless they understand the parameters to which they must now attend, the effectiveness of treatment might be compromised. The interelationship between task elements and, hence, staff, is again emphasised.

Treatment planning directs the strategy for care and treatment. It may involve diagnosis or simply prescription of an established programme of treatment on the basis of assessment of the baby's condition. Sometimes treatments are prescribed with little confidence, but something has to be done. Treatment planning is invariably followed by treatment actions, and then careful monitoring and assessment to determine treatment effectiveness. Thus, a treatment plan should inform monitoring and testing activities in terms of parameters to be monitored and diagnostic tests, routine or non-routine, to be carried out.

Care management support: This is seen as the management of the care and treatment interventions. It was emphasised to us that one of the prime treatments that babies can be given was to allow them to rest. Unnecessary disturbances will arrest progress. Exercising some degree of forward planning by care staff will ensure that everything that needs to be done will be done, but that it is coordinated to minimise upset. Thus, preparing and revising the schedule of treatment/care activity for the baby (5 & 6), needs to be taken seriously and respected as a task. We have seen strong-willed senior nurses drive this process very effectively, because they know that unless they do so, other staff may cause interruptions to care which could have been avoided.

Communication: Communications are vital. Both ‘record/relate significant observations, events and inferences’ (10) and ‘provide information to next shift’ (11) depend upon staff understanding what is of significance to promote treatment and care, above and beyond the information that will be routinely monitored. The information judged to be significant must be formally recorded or committed to memory. Where information is obtained during a complex procedure or decision making task, then the carer may simply forget to make a record of it.

Our observations and task analysis have revealed a number of opportunities for error inherent in the NICU system. Errors are frequent and often relate to unavailability of information. Fortunately, the system can often prevent such errors from developing into incidents. Defence mechanisms include duplication, staff memory and shiftwork. Thus, information that has been omitted from one record may exist in a different record. Information that has not been recorded may be remembered because it was verbally expressed at some point. Information that may be missed by one shift may be picked up by the next shift. However, it is evident that such defences are informal and sometimes may not work or may work to the detriment of the system. For example, duplication in records is time consuming and may lead to transcription errors resulting in inconsistencies. Memory of verbal exchanges is notoriously unreliable. NICU staff repeatedly comment on how often information gets lost or altered through verbal communication. The involvement of various members of staff in the care of an individual baby increases the likelihood of information being picked up but at the same time increases the likelihood of information getting lost. There is a big difference between a member of staff obtaining a piece of information and using it to perform a task and a member of staff obtaining a piece of information and communicating to a colleague who will then use it to perform that task.

The potential sources of error revealed by our fieldwork are plausible yet remain hypothetical. The logic of the care task, observations made of events and incidents and comments made by various NICU staff give credence to these suggestions, but they do not consitute categorical evidence.

Our current phase of research is to investigate these potential weaknesses further and to investigate ways in which problems might be overcome. Thus, we have developed baby simulations whose health can be monitored, where problems to be introduced in their development and treatments can be administered. Our current simulation has been programmed on Stella™, a modelling package available to run on desktop computers. Individual babies are represented by running the baby simulation on different iMac™ computers. Different ‘babies’ are given different initial states, different recovery characteristics and different problems to contend with. We are using the simulations to invistigate how engagement with different tasks influences the manner in which people handle information and how they communicate with colleagues. Thus, we are investigating, for example, the benefits of different ways of training people to think about their task, different forms of representing records and different protocols for briefing colleagues. We also plan to investigate the development of teams over time, observing how their skills and relationships develop and how, as a consequence, they learn to share information.

We hope to establish how various elements, including computer displays, recording formats, training and protocols for verbal communication, can be designed to improve the handling of information in contexts such as these. The outcome of such work can never be in the form of a categorical finding that we can be confident of applying to real situations and we are fully aware of the constraints of this approach.. However, through such methods we would seek support for more viable design hypotheses than can be obtained through observational studies alone.

Our studies of neonatal intensive care have revealed a complex working environment in which different people with different expertise and responsibilities and working across different shifts, must collaborate in order to maintain the care and treatment of babies at extreme risk. Through observations and task analysis, we have identified a wide range of variations of practice, even though the basic processes of care are common between units.

A particular observation has been the complex interdependency between different sorts of care task, especially from the perspective of sharing information. The need for staff to collaborate is paramount. Yet there are clear indications that this collaboration can be compromised by problems of communication brought about by the fragmentation of tasks across staff.

Several aspects of the NICU context lend themselves to redesign, including the format of records, the effectiveness of training and protocols for shift handovers. Making recommendations in this respect depends upon further research into the design of information artifacts. This is one reason why simulation studies are recommended to supplement observational studies, even though the simulation of these tasks severely compromises the realities of the real NICU context.

One area that has been noted, which cannot be studied in a task simulation, concerns the traditions of the medical and nursing professions. The staffing of NICUs owes much to the traditional relationships that have emerged between doctors and nurses. Yet, if NICU staffing were to be specified from the perspective of good team and job design, the recommended allocation of responsibilities would probably look very different. There is scope for more clearly identifying how nurses undertaking monitoring and care tasks can more effectively anticipate the requirements of medical staff undertaking treatment planning. And there is scope for prescribing better ways for medical staff to convey their plans to the staff who will be responsible for carrying them through.

Annett, J., Duncan, K.D., Stammers, R.B. and Gray, M.J. Task Analysis. HMSO, 1971.

Cunningham, S., Deere, S., Elton, R. A., & McIntosh, N. Neonatal physiological trend monitoring by computer. International Journal of Clinical Monitoring and Computing, 9, pp 221-227, 1992.

Gopher, D., Olin, M., Badihi, Y., Cohen, G., Donchin, Y., Bieski, M., & Cotev, S. The Nature and Causes of Human Errors in a Medical Intensive Care Unit. Paper presented at the Perspectives. Proceedings of the Human Factors Society 33rd Annual Meeting, Denver, Colorado, 1989.

Grusenmeyer, C. La relève de poste - Une phase critique du travail en équipes successives. Cahiers de notes documentaires, 144, pp 407-418, 1991.

Grusenmeyer, C. Shared Functional Representation in Cooperative Tasks-The Example of Shift Changeover. The International Journal of Human Factors in Manufacturing, 5(2), pp 163-176, 1995.

Merenstein, G. B., & Gardner, S. L. Handbook of Neonatal Intensive Care. (3rd ed.). St. Louis: Mosby-Year Book, Inc., 1993.

Room, R. Alcohol and ethnography: a case of problem deflation? Human Organization,. 25, no. 2, pp.169-78,1984.

Shepherd, A. Task Analysis as a Framework for Examining HCI Tasks. Perspectives on HCI - Diverse Approaches. (ed. A. Monk and G. Gilbert), London, Academic Press: pp 145 - 174, 1995.

Shepherd, A. HTA as a framework for task analysis. Ergonomics. 41(11), pp1537 – 1552, 1998.

Shalin, V. L., & Bertram, D. A. (1996). Functions of expertise in a medical intensive care unit. Journal of Experimental and Theoretical Artificial Intelligence, 8(3-4), 209-227.

We would like to acknowledge all of the staff of Neonatal Intensive Care Units who have given us their time and demonstrated their expertise.

Dr Andrew Shepherd, Cognitive Ergonomics Research Group, Department of Human Sciences, Loughborough University, Loughborough Leics, LE11 3TU, UK. Telephone: +44 (0)1509 223028.
Fax: +44 (0)1509 223940. Email: A.Shepherd@lboro.ac.uk.

Dr Shepherd is an applied psychologists with particular interests in task analysis, training, interface design, team .

Dr Olga Kostopoulou, Cognitive Ergonomics Research Group, (see above).
Telephone: +44 (0)1509 263171 ext. 4271.
Fax: +44 (0)1509 223940. Email: O.Kostopoulou@lboro.ac.uk.

Dr Kosopoulou is an applied psychologists with particular interests in diagnosis and faultfinding with a particular interest in medical settings.