Clinical decision-making by midwives: managing case Complexity

Journal of Advanced Nursing, 1997, 25, 265–272 Clinical decision-making by midwives: managing case complexity Jane Cioffi RN BAppSc (Adv Nsg) GradDipEd MAppSc(Nsg) PhD FRCNA Lecturer, School of Health, Faculty of Health, Humanities and Social Ecology, University of Western Sydney, Hawkesbury and Roslyn Markham MAPS MA PhD Senior Lecturer, Department of Psychology, The University of Sydney, Australia Accepted for publication 19 February 1996 CIOFFI J. & MARKHA M R. (1997) Journal of Advanced Nursing 25, 265–272 Clinical decision-making by midwives: managing case complexity In making clinical judgements, it is argued that midwives use ‘shortcuts’ or heuristics based on estimated probabilities to simplify the decision-making task. Midwives (n=30) were given simulated patient assessment situations of high and low complexity and were required to think aloud. Analysis of verbal protocols showed that subjective probability judgements (heuristics) were used more frequently in the high than low complexity case and predominated in the last quarter of the assessment period for the high complexity case. ‘Representativeness’ was identified more frequently in the high than in the low case, but was the dominant heuristic in both. Reports completed after each simulation suggest that heuristics based on memory for particular conditions affect decisions. It is concluded that midwives use heuristics, derived mainly from their clinical experiences, in an attempt to save cognitive effort and to facilitate reasonably accurate decisions in the decision-making process. I NTRODUCTI ON Competency in clinical decision-making is the very least a patient should expect from a nurse, legally and ethically. However, there is limited understanding of the processes used by nurses in making their clinical judgements (Grier 1976, Tierney 1987). If these were understood, then nurses could be offered appropriate education in decisionmaking. As a result of this, patient outcomes might be improved. The two main approaches to the study of decisionmaking in nurses are the ‘prescriptive’ and ‘descriptive’. The prescriptive approach targets how decisions ought to be made (e.g. Corcoran 1986, Hughes & Young 1990), and the descriptive approach focuses on how decisions are Correspondence: Dr J. Cioffi, School of Health, Faculty of Health, Humanities and Social Ecology, University of Western Sydney, Hawkesbury, Bourke St, Richmond, NSW 2753, Australia. © 1997 Blackwell Science Ltd actually made by nurses (e.g. Benner 1984, Jones 1988). The present study is descriptive, as it is concerned with the processes nurses use to make judgements rather than on the outcomes and results of their decision-making processes. An examination of the clinical thinking of nurses in patient assessment situations can provide information about the possible reasoning processes that underlie the assessments nurses make and the subsequent diagnoses (e.g. Broderick & Ammenthorp 1979, Gordon 1980, Tanner et al. 1987). The nature of the clinical situations themselves undoubtedly affects the processes used. Some assessment situations are much more complex than others, involving more ‘unknowns and uncertainties’ (Hammond 1966, Carnevali et al. 1984, Tierney 1987). In fact, nurses often have to make judgements when there are few ‘knowns’ and ‘certainties’. Studies of nurses’ decisionmaking have shown that the processes used are taskdependent (Corcoran 1986, Yocom 1986). 265 J. Cioffi and R. Markham The very complexity of these decision-making tasks necessitates short-cuts in reasoning, i.e. the task may be so vast that the nurse would be quite unable to consider all possibilities within the time restrictions. In this type of situation the nurse will have to make a judgement about what is the most probable scenario. These estimations of probability may depend on factors such as memories of similar events they have experienced or read about, or the details they happen to focus upon at the time. As a result, individual differences in the probability estimates of nurses are to be expected. Nurses in different clinical settings will be exposed to different types of cases with varying frequencies. Furthermore, there will be diversity within the particular types of cases to which they are exposed. The nurse may have preconceived notions of what is expected in a case, but this is likely to be challenged by what is found in actual practice. The transformation of textbook knowledge into skilled clinical knowledge by practice has been discussed by both psychologists (e.g. Anderson 1982, 1983, 1987) and nurse educators (e.g. Benner 1982) as the process of skill acquisition. Heuristic decision-making It is argued that through experience, individuals develop ‘rules of thumb’ that they rely on when making decisions. These rules of thumb processes are often referred as heuristics. In the social sciences, the term ‘heuristics’ is used to denote principles that reduce complex tasks of assessing probabilities and predicting values to simpler operations. (In cybernetics and artificial intelligence ‘heuristic’ is usually used to describe a procedure for searching out an unknown goal by using a method which cuts down on the amount of searching required.) Studies have shown that certain heuristic principles are used for estimating the probability of solutions (Kahneman & Tversky 1973, Tversky & Kahneman 1973, 1974, 1982). It is argued that they are easy and fast to use and usually result in reasonably valid inferences (Abelson & Levi 1985). Heuristic knowledge is tacit rather than explicit. It is acquired because it works most of the time. However, the use of heuristic techniques does not guarantee solutions; they are used to reduce the number of possibilities that are considered, and the correct solution may be one that is ignored. As a consequence of heuristic activity, people will be expected to vary in the degree of confidence they have regarding the likelihood of a particular outcome in a specific situation (Bernouilli 1954). The ‘classic’ heuristic principles that have been identified are: ‘representativeness’, ‘availability’ and ‘anchoring and adjustment’ (Tversky & Kahneman 1974). Investigation of the use of these heuristic methods by nurses, in complex decision-making tasks, may shed light on the choices that they actually make in the course of 266 their nursing activities about the likelihood of particular events. Howell & Burnett (1978) propose that these types of heuristics are relied on more heavily as the level of complexity increases in a judgement task. Representativeness Nurses would be expected to use the ‘representativeness’ heuristic process when judging the probability that certain signs and symptoms in patients indicate a particular clinical condition that the nurse has previously encountered. For example, the triage nurse, when assessing a patient with chest pain, might consider the extent to which the patient’s pattern of presentation is more representative of cardiac, musculoskeletal or gallbladder involvement. This type of assessment exemplifies ‘representativeness’ (Kahneman & Tversky 1972, Kruglanski & Ajzen 1983), which is the most commonly used heuristic principle in decision-making (Nisbett et al. 1983). The nurse needs to determine the possible alternative decisions that could be made and the probability of each outcome based on their prior experiences, both practical and theoretical (Kuipers et al. 1988). Such likelihood judgements, however, are affected by known risks associated with certain clinical conditions; for example, breech presentation, with the possible accompanying fetal complications. Both psychological (Azjen 1977, Bar Hillel 1985) and medical (Balla 1985) studies have shown that when conditions have potentially serious effects, the incidence of the conditions in a given population is overestimated. Availability The ‘availability’ heuristic principle is characterized by the ease with which instances of similar conditions come to mind (Tversky & Kahneman 1973, Friedlander & Stockman 1983). It is likely to be the heuristic process employed when events are thought of more readily in terms of particular cases. Instances that are easily recalled are assumed by the reasoner to be more probable than those less easily recalled. If a nurse holds a very vivid memory, for example, of a particular breech birth experienced some time in the past, then memory of this particular case is available to be used when assessing any other patient who may be similar in some way to that case. Anchoring and adjustment When using the ‘anchoring and adjustment’ heuristic principle the decision-maker starts from an ‘anchor point’ or baseline and adjusts from this anchor point to take account of individual patient characteristics and arrive at a final estimate (Tversky & Kahneman 1974). These adjustments involve consideration of related signs and symptoms, and will result in probability estimates for particular decisions (Kuipers et al. 1988). For example, if a patient complained of calf pain the nurse would consider the patient’s risk of © 1997 Blackwell Science Ltd, Journal of Advanced Nursing, 25, 265–272 Clinical decision-making by midwives deep venous thrombosis (DVT). If this patient had been on bed-rest, had a previous history of DVTs and had had a splenectomy in the last 72 hours, then the risk of DVT would be higher than for the patient with calf pain who had been active, had no history of DVT and was being admitted for a day surgical procedure for removal of a basal cell carcinoma from the forehead. Heuristic clinical decision-making and task complexity The importance of the use of heuristic methods by nurses has been noted in two clinical judgement studies (Broderick & Ammenthorp 1979, Bennett 1980), in which contrived experimental situations of a medical type were used (Bennett 1980). Additionally, some studies which have reported descriptions of intuitive judgements used in various clinical situations can be interpreted as providing indirect evidence for the use of heuristics by nurses, although they were not interpreted in these terms by the investigators (e.g. Pyles & Stern 1983, Benner & Tanner 1987, Schraeder & Fischer 1987, Rew, 1988). To date, no study has been reported which explores the effects of varying levels of complexity on the use of heuristics by nurses in assessment situations. Complex situations are inherently uncertain, because of the many alternatives that are available for consideration by the decision-maker (Carroll & Johnson 1990). Patient assessment situations are almost always uncertain, so it seems highly probable that heuristic techniques are used by nurses in decisionmaking in their clinical practice. Furthermore, nurses would be expected to employ heuristics to a greater extent as the level of complexity of the decision-making task increases, because of the greater number of alternatives that need to be considered. In the present study, the relationship between the use of heuristics and task complexity was examined in clinical decision-making tasks. Heuristic approaches were identified from the verbal protocols of midwives as they attempted to determine patient conditions. It was expected that a greater number of heuristic processes would be used in the more complex case than in the simpler, and it was anticipated that the ‘representativeness’ heuristic principle would be used more frequently than ‘availability’ and ‘anchoring and adjustment’. THE STUDY Method Design The focus of the study was the examination of the processes of clinical decision-making as performed by certified and student midwives in the patient assessment phase. Midwifery was selected as the nursing domain, as it involves a high level of autonomous practice. Simulated assessment situations were designed to approximate as closely as possible to those occurring in the ‘real world’, by using actual clinical case studies. The midwives were instructed to think aloud while attempting to reach a decision in each simulated case. This ‘think aloud’ procedure allowed the collection of extensive data about the reasoning processes undertaken by each nurse. This could then be analysed in order to deduce the use of heuristics (Elstein et al. 1982). A postexperimental report, completed by the midwives at the end of each case, allowed further deductions about heuristic approaches. Sample The sample consisted of 30 volunteer midwives of various levels of experience. They came from midwifery units in teaching and district hospitals. The sample size was restricted because of the vast amounts of data generated for analysis by each subject and the complex and timeconsuming nature of the protocol analysis process (Ericsson & Simon 1984, Elstein et al. 1990). Instrument Two patient assessment cases were developed from the actual case records of childbirth patients, to provide simulated assessment situations. The assessment situations selected, i.e. ‘uncomplicated established labour’ and ‘antepartum haemorrhage’, were of low and high complexity (uncertainty) respectively. The low complexity case was identified as such because relevant information was available and there were predictable relationships between the signs and symptoms; the high complexity case involved relationships between the signs and symptoms that were not easily predictable, and there was a reduced level of relevant information (Cosier & Dalton 1980). A minimal profile describing the patient’s presentation to the birth unit and a series of question–answer items were devised for each case. The questions were those that a panel of expert midwives (n=10) judged as likely to be asked by a midwife during the assessment, and addressed all aspects of each case (Baussell 1986). The experts determined the appropriate answers to the questions, and both questions and answers were placed on a master sheet for use by the investigator. The experts were experienced midwives from both faculty and hospital settings (mean years of experience, 11.2). For both uncomplicated established labour and antepartum haemorrhage, the items were judged to be at least 92% necessary and 90% sufficient. Both the panel of experts and the midwife subjects assessed the level of complexity (uncertainty) of each case simulation. As predicted, the low uncertainty case simulation, uncomplicated established labour, was judged by both groups to have a much greater degree of relevant © 1997 Blackwell Science Ltd, Journal of Advanced Nursing, 25, 265–272 267 J. Cioffi and R. Markham information and to contain a much higher level of predictable relationships than the high uncertainty case simulation, antepartum haemorrhage. A short report form was constructed to enable the nurses to provide information about the vividness, recency and recall of similar cases, and the estimates of base rates they held for the incidence of both clinical conditions. Recency was evaluated by estimating the number of days that had elapsed since they had seen the clinical condition, with the vividness of the condition in memory being rated on a scale of ‘very’, ‘quite’, ‘somewhat’ and ‘not vivid’. This information allowed aspects of the ‘availability’ and the ‘representativeness’ heuristic processes to be identified. Procedure Each subject was interviewed individually and the verbal protocols were tape-recorded for the two counterbalanced case simulations. Before the first case was presented the subject was instructed in the ‘think aloud’ technique. This was followed by a practice session in the technique (as recommended by Ericsson & Smith (1984)). The subject was then informed that, following the presentation of a patient introductory statement, they would be required to seek further patient information by asking the investigator specific questions, similar to those that would be asked during an assessment in actual practice. The investigator answered the questions asked by the subject, using the standardized answers from the master sheet. The subjects were requested to say aloud everything they were thinking from the time they first received the initial patient statement until they believed that they had diagnosed the patient’s condition. They were regularly prompted to ‘think aloud’. Each protocol lasted approximately 10–15 minutes. At the end of each protocol, the subject completed a post-interview report form before starting the next protocol. Protocol analysis Analysis of the 60 protocols was carried out by two raters, both experienced midwives (mean years of midwifery experience, 14) who were trained in coding and categorizing protocol segments. When the protocols were first examined, it was agreed that all the statements could be coded into a small number of categories ( Jones 1988). It was decided initially to allocate the protocol segments to these categories before attempting to identify the heuristic processes. This procedure allowed a clear identification of those parts of the protocol which involved inference and previous knowledge, in contrast to those that simply involved gathering data from the investigator, repetition of information previously collected, irrelevant exclamations or anticipation of actions that might be taken. 268 From the protocol segments involving inferences and previous knowledge, the raters categorized those judgements that implied some estimate of probability, using the approach of Kuipers et al. (1988). Two of the major types of heuristics, ‘representativeness’ and ‘anchoring and adjustment’ were identified from the protocols, together with a combination of these two, ‘representativeness combined with anchoring and adjustment’. Examples of each of these are: ‘ovoid-shaped uterus so technically speaking the baby should be a longitudinal lie’ (representativeness); ‘gravida 7, parity 6, — probably going to be very very quick to deliver’ (anchoring and adjustment); ‘no increase in bleeding, half a cup of bright red blood, moderately soaked pad — may have been a slightly heavier than usual show’ (anchoring and adjustment combined with representativeness). Inter-rater reliability for coding the segments of the transcripts was 91% and for categorizing the heuristics it was 94%. The third major heuristic, ‘availability’, could only be identified from reports made by the midwife at the end of each case, as it is based on factors such as vividness and recency of experience with particular cases. This heuristic was considered separately, after the completion of the main analyses. The data from the subjects’ reports collected after each case were only analysed descriptively, because of the vast range of scores and the preferred index of central tendency was the median value. RESULTS It should be noted that accurate diagnoses were reached by 100% of the midwives in both case assessments. Midwives were found to use heuristics in their decisionmaking processes. As the protocol length varied between subjects (from 12 to 101 segments), the number of heuristic processes used by each subject was calculated relative to the number of segments identified in each protocol, i.e. for each subject, the total number of heuristic processes identified in each protocol was summed and divided by the total number of segments for that subject in that protocol, to give proportion scores. The means and standard deviations (SD) of the proportion scores based on the number of heuristic processes used in the lower and higher complexity cases were 0·074 (SD=0·108) and 0·149 (SD= 0·189), respectively. A related t-test found that heuristic approaches were verbalized more by nurses in higher than in lower complexity assessment situations, as predicted (t(degrees of freedom 29)=−2·93, P<0·01). Each protocol was then further divided into quartiles over time, based on the total number of segments identified in that protocol. The frequency of heuristic processes that occurred in each quartile relative to the total frequencies was calculated. The means and standard deviations for each quartile of the lower and higher complexity cases are © 1997 Blackwell Science Ltd, Journal of Advanced Nursing, 25, 265–272 Clinical decision-making by midwives given in Table 1. The proportion of heuristic processes increased between the first and last quartile for both cases. However, this increase was much greater in the higher complexity case. These differences were tested using an ANOVA with lower and higher cases as the betweensubject factor and quartiles as the within-subjects factor. The main effect for quartiles was significant (F(degrees of freedom 3)=6·131, P<0·001) as was the interaction of quartile and complexity (F(3)=2·946, P<0·01), but no main effect for complexity was found. Post hoc Newman–Keuls comparisons of the proportions of heuristic processes between the lower and higher complexity cases at quartiles one, two, three and four revealed that nurses verbalized a greater proportion of heuristics in the fourth quartile in the higher than the lower complexity case ( P<0·001) but showed no differences for the other quartiles. The patterns of heuristic activity within the lower and higher complexity cases were tested separately. In the higher complexity case, the use of heuristics in the first quartile was significantly lower than in the second, third and fourth quartiles ( P<0·05); and the first, second and third quartiles were lower than the fourth ( P<0·05). In the lower complexity case, no significant differences between the quartiles were found. The proportion of each heuristic type in the protocol was calculated for each subject by dividing the frequency of heuristic processes for each type by the frequency of all heuristic processes used by the subject in that case assessment. The means and standard deviations for each of the heuristic types were calculated for lower and higher complexity cases and are presented in Table 2. Significant differences were found between the mean proportions of the different heuristic types (F(2)=73·832, P<0·001) and for the interaction between heuristic type and complexity ( F(2)=3·877, P<0·05), but no significant main effect of complexity was found. Newman–Keuls tests showed that in both the lower and higher complexity cases, the proportion of the ‘representativeness’ heuristic type was greater than both the ‘anchoring and adjustment’ and ‘anchoring and adjustment combined with representativeness’ heuristic Table 1 Means and standard deviations of proportions of heuristic strategies for the lower and higher complexity cases Case n Quartile Mean SD Lower complexity 30 30 30 30 30 30 30 30 First Second Third Fourth First Second Third Fourth 0·121 0·249 0·232 0·197 0·049 0·213 0·217 0·421 0·254 0·350 0·331 0·293 0·099 0·225 0·221 0·323 Higher complexity Table 2 Means and standard deviations of proportions for the three heuristic types ‘representativeness’, ‘anchoring and adjustment’ and ‘anchoring and adjustment combined with representativeness’, in lower and higher complexity cases Lower complexity case Higher complexity case Representatives Mean 0·512 SD 0·435 0·763 0·340 Anchoring and adjustment Mean 0·093 SD 0·187 0·020 0·068 Anchoring and adjustment and representatives Mean 0·195 0·117 SD 0·315 0·217 types ( P<0·01) (see Table 2). That is, in their decisionmaking processes, midwives relied on the ‘representativeness’ heuristic activity more than on other types of heuristics, regardless of the level of complexity of the case. However, the proportion of ‘representativeness’ heuristic processes to be verbalized was greater in the higher than the lower complexity case ( P<0·05). From the report forms completed at the end of each protocol collection, frequencies were calculated for the nurses’ recall of similar cases and the recency of and vividness of memory for particular cases. The findings suggested that the midwives used heuristic techniques involving memories of particular cases during the decision-making process. Recall of cases similar to the type currently being assessed was greater in lower (77%) than higher (57%) complexity cases. Furthermore, this lower complexity case (‘uncomplicated established labour’) was reported to have been experienced more recently (median 2 days) than the higher complexity case (‘antepartum haemorrhage’, median 35 days). Memories of uncomplicated established labour were reported to be ‘very vivid’ by 75% of nurses, whereas those for antepartum haemorrhage were stated to be ‘very vivid’ by only 25%. The report form also asked the midwives to estimate the base rate for each clinical condition from incidences in their own clinical experience. The medians of these rate estimations were calculated; then they were compared with the specific hospital incidences reported in the NSW Midwives Data Collection (1992). As different hospitals were involved, the median values and ranges for ‘antepartum haemorrhage’ and ‘uncomplicated established labour’ in each hospital childbirth population are presented separately in Table 3. The base rates for ‘antepartum haemorrhage’ were overestimated by the nurses from three of the four hospitals. The estimated base rate for ‘uncomplicated established labour’ by midwives from each hospital could © 1997 Blackwell Science Ltd, Journal of Advanced Nursing, 25, 265–272 269 J. Cioffi and R. Markham Table 3 Actual hospital incidences and medians and ranges of midwives’ base rate estimates of antepartum haemorrhage and uncomplicated established labour. All data are percentages Clinical condition Hospital code Antepartum haemorrhage A B C D Uncomplicated established labour A B C D Median 5 3·5 3·5 1·5 80 80 80 90 Range Hospital incidence 2–35 1–10 3–4 — 1·8 6·0 1·5 0·8 40–90 50–99 60–90 80-95 Not available not be compared to specific hospital rates, as this clinical category was not available in the NSW Midwives Data Collection (1992). DI SCUSSION The results of this study are important in showing that midwives relied on heuristics in simulated clinical decision-making situations and employed heuristic processes to a greater extent the more complex the clinical case. Thus, in patient assessment situations, where complexity is created by a lack of relevant clinical information and limited predictability of relationships, midwives may be expected to use a high number of heuristic strategies. This finding supports Howell & Burnett’s (1978) suggestion that as task complexity increases, there will probably be a heavier dependence on heuristics, and confirms a similar conclusion by Tversky & Kahneman (1973, 1974) regarding non-clinical tasks. Heuristic techniques were used in an effective manner, as shown by the high accuracy of diagnoses in this study. In the higher complexity case, the midwives used increasing proportions of heuristic techniques over time in making their diagnostic judgements. The results suggest that when uncertainty is not resolved by the patient information that is collected over the assessment period, the midwife relies increasingly on the use of heuristics in an effort to determine the patient diagnosis. This is consistent with Tversky & Kahneman’s (1974) findings that judgements tend to be in terms of probability when the decisionmaking situation remains uncertain. Hence, midwives in conditions of uncertainty take short-cuts in reasoning as a way of simplifying the complexities of their judgement tasks. The ‘representativeness’ heuristic process was found to be relied on more than other types in both lower and higher complexity cases, supporting the claim of Nisbett et al. (1983) that the ‘representativeness’ heuristic technique is 270 the one most often used in decision-making. This suggests that nurses are characterizing events in terms of categories of clinical conditions (Kahneman & Tversky 1972, Kruglanski & Ajzen 1983), judging the probability of the patient’s presenting signs and symptoms as belonging to previously experienced clinical entities. The ‘representativeness’ heuristic was used significantly more in higher than in lower complexity cases, suggesting that there is greater influence from prior clinical experiences in the higher complexity case. Thus, midwives with more previous experiences would be expected to have an advantage in complex decision-making situations. ‘Representativeness’ is influenced by knowledge about base rates. Midwives’ overestimation of base rates of antepartum haemorrhage may be associated with the perception that this condition can precipitate both maternal and fetal complications. This is consistent with the findings of Azjen (1977), Bar Hillel (1985), and Balla (1985) regarding the potency of causal significance. Such overestimation may create bias in the judgements midwives make in clinical practice, and may result in adverse patient outcomes. This suggests that midwives should be made very aware of the actual base rates of various clinical conditions in their hospitals and be informed of possible biasing effects if they estimate incorrectly. Midwives’ reports of recent experiences with the clinical conditions indicated that ‘uncomplicated established labour’ had been experienced more recently than ‘antepartum haemorrhage’. This is expected, as ‘uncomplicated established labour’ is the most common presentation in clinical practice. In addition, more nurses reported more ‘very vivid’ memories of ‘uncomplicated established labour’ than ‘very vivid’ memories of antepartum haemorrhage. Recency and vividness are evidence of the use of the ‘availability’ heuristic process by midwives. This heuristic strategy should thus favour increased probability estimates of uncomplicated established labour and thus may counteract overestimation of the probability of risky but less common clinical conditions. The tendency of midwives in this study to recall similar cases to that under consideration has also been noted in previous studies with nurses from different clinical settings (Pyles & Stern 1983, Benner & Tanner 1987, Schraeder & Fischer 1987). This form of thinking has been identified as ‘intuitive’ by these authors and, more specifically, as the ‘similarity recognition’ aspect of intuition (Dreyfus & Dreyfus 1986, Benner & Tanner 1987). Intuitive judgement can therefore be considered to have an heuristic component. In summary, this study addresses some important issues about decision-making by midwives. In daily practice, midwives must often cope with clinical decision-making situations in which there can be little certainty, because patient information is usually incomplete. Therefore, reasoning in uncertainty is quintessential to clinical © 1997 Blackwell Science Ltd, Journal of Advanced Nursing, 25, 265–272 Clinical decision-making by midwives judgement. In the complex case in the present study, midwives were shown to adapt their decision-making strategies by increasing the use of those heuristic techniques based on estimated probabilities. These probabilities were derived from their knowledge based on personal experiences and from other processes. The accuracy of such subjective probability judgements will, of course, be dependent on the appropriateness of prior experience and the ease of retrieval from memory of these experiences. Implications Although midwives were found to use repertoires of heuristic strategies and to require mastery of heuristics to assure accurate judgements for patients, it is questionable how conversant they are with the role that heuristic activity plays in their decision-making. As retrieving information from memory and inferring probabilities is central to the judgement process, clinicians should be made aware of the fact that their estimations of probability are dependent on their personal experiences and they should be encouraged to examine their decision-making processes for possible biases created by their past experiences. Thus, practising midwives will require continual updating of factors that may influence their clinical decision-making, e.g. base rate information. The influence in decision-making of previous experience has implications for service providers, beginning practitioners and nurse educators. The more experienced nurses are concerning the relevant clinical phenomena, the more likely they are to manage complex decisionmaking situations adequately and to make appropriate judgements about patients. There are also implications for the development of educational programmes for nurses. The development of skilled clinical knowledge in a nurse depends on a process of making adjustments to preconceived notions and expectations, by repeated encounters with somewhat similar practical situations. Beginning practitioners need to be cognisant of this critical adjustment process and seek placements in service settings that will help it to happen. In addition, educators need to be aware of the importance of varied clinical experiences when developing their programmes. CONCLUSION In conclusion, the findings of this study expand our knowledge of clinical decision-making. However, there is a need for clinicians to actively pursue further investigations into the processes involved in making clinical judgements, so that nurses can understand and thus control the processes that lead to diagnoses and actions. The outcome of this may be increased control over the quality of care given and enhanced professional value in the healthcare service and the community. Acknowledgement This study was partly funded by the NSW Nurses’ Registration Board. References Abelson R.P. & Levi A. (1985) Decision making and decision theory. In Handbook of Social Psychology, vol. 1 3rd edn (Lindzey G. & Aronson E. eds). Random House, New York, pp. 231–309. Ajzen I. (1977) Intuitive theories of events and effects of base rate information on prediction. Journal of Personality and Social Psychology 35, 303–314. Balla J.I. (1985) The Diagnostic Process. A Model for Clinical Teachers. Cambridge University Press, Cambridge. Bar Hillel M. (1980) The base rate fallacy in probability judgement. Acta Psychologica 44, 211–233. Bausell R.B. (1986) A Practical Guide to Conducting Empirical Research. Harper and Row, New York. Benner P. (1984) From Novice to Expert: Excellence and Power in Clinical Nursing, Addison Wesley, Menlo Park, California. Benner P. & Wrubel J. (1982) Skilled clinical knowledge: the value of perceptual awareness. Nurse Educator 7, 11–17. Benner P. & Tanner C. (1987) Clinical judgement: how expert nurses use intuition. American Journal of Nursing 87, 23–31. Bennett M. (1980) Heuristics and the weighting of base rate information in diagnostic tasks by nurses. PhD thesis, Monash University, Melbourne. Bernouilli D. (1954) Exposition of a new theory on the measurement of risk. Econometrica 22, 23–36. Broderick M.E. & Ammenthorp W. (1979) Information structures: an analysis of nursing performance. Nursing Research 28, 106–110. Carroll J.S. & Johnson E.J. (1990) Decision Research. A Field Guide. Sage, Newbury Park. Carnevali D.L., Mitchell P.H., Wood N.F. & Tanner C.A. (1984) Diagnostic Reasoning in Nursing. J.B. Lippincott, Philadelphia. Corcoran S. (1986) Decision analysis: a step-by-step guide for making clinical decisions. Nursing and Health Care 7, 149–154. Cosier R.A. & Dalton D.R. (1988) Presenting information under conditions of uncertainty and availability: some recommendations. Behavioral Sciences 33, 272–281. Dreyfus H.L. & Dreyfus S.E. (1986) Mind over Machine. The Free Press, New York. Elstein A.S., Rovner D.R., Holzman G.B., Ravitch M.M., Rothert M.L. & Holmes M.M. (1982) Psychological approaches to medical decision-making. American Behavioral Scientist 25, 557–584. Elstein A.S., Shulman L.S. & Sprafka S.A. (1990) Medical problem solving, a ten year retrospective. Evaluation and The Health Professions 13, 5–36. Ericsson K.A. & Simon H.A. (1984) Protocol Analysis. MIT Press, Cambridge, Massachusetts. Friedlander M.L. & Stockman S.J. (1983) Anchoring and publicity effects in clinical judgement. Journal of Clinical Psychology 39, 637–643. Gordon M. (1980) Predictive strategies in diagnostic tasks. Nursing Research 29, 39–45. © 1997 Blackwell Science Ltd, Journal of Advanced Nursing, 25, 265–272 271 J. Cioffi and R. Markham Grier M.R. (1976) Decision-making about patient care. Nursing Research 25, 105–110. Hammond K.R. (1966) Clinical inferences in nursing. A psychologist’s viewpoint. Nursing Research 15, 27–38. Howell W.C. & Burnett A.A. (1978) Uncertainty measurement: a cognitive taxonomy. Organisational Behaviour and Human Performance 32, 45–68. Hughes K. & Young W. (1990) The relationship between task complexity and decision making consistency. Research in Nursing and Health 13, 189–197. Jones J. (1988) Clinical reasoning in nursing. Journal of Advanced Nursing 13, 185–192. Kahneman D. & Tversky A. (1972) Subjective probability: a judgement of representativeness. Cognitive Psychology 3, 430–454. Kahneman D. & Tversky A. (1973) On the psychology of prediction. Psychological Review 80, 237–251. Kahneman D. & Tversky A. (1982) Variants of uncertainty. In Judgement under Uncertainty: Heuristics and Biases (Kahneman D., Slovic P. & Tversky, A. eds), Cambridge University Press, New York, pp. 509–520. Kruglanski A.W. & Ajzen I. (1983) Bias and error in human judgement. European Journal of Social Psychology 13, 1–44. Kuipers B., Moskowitz A.J. & Kassirer J.P. (1988) Critical decisions under uncertainty: representation and structure. Cognitive Science 12, 177–210. New South Wales Midwives Data Collection (1992) Epidemiology and Health Services Evaluation, NSW Health Department, North Sydney. 272 Nisbett R.E., Krantz D.H., Jepson D.H. & Kunda Z. (1983) The use of statistical heuristics in everyday inductive reading. Psychological Review 90, 339–363. Pyles S.H. & Stern P.H. (1983) Discovery of nursing gestalt in critical care nursing: the importance of the gray gorilla syndrome. Image. The Journal of Nursing Scholarship 15, 51–59. Rew L. (1988) Intuition in decision making. Image: Journal of Nursing Scholarship 20, 150–154. Schraeder B.D. & Fischer D.K. (1987) Language and designs for probability judgements. Cognitive Science 9, 309–339. Tanner C.A., Padrick K.P., Westfall V.E. & Putzier D.J. (1987) Diagnostic reasoning strategies of nurses and nursing students. Nursing Research 36, 358–363. Tierney A.J. (1987) A view on clinical judgement and decision making from the perspective of the nursing process. In Clinical Judgement and Decision Making: The Future with Nursing Diagnosis, Proceedings of the International Conference (Hannah K.J., Reimer M., Mills W.C. & Letourneau S. eds), John Wiley, New York, pp. 260–267. Tversky A. & Kahneman D. (1973) Availability: a heuristic for judging frequency and probability. Cognitive Psychology 5, 207–232. Tversky A. & Kahneman D. (1974) Judgement in uncertainty: heuristics and biases. Science 185, 1124–1131. Yocom C. (1986) Influence of initial nursing educational preparation on patient assessment. (Doctoral dissertation, University of Illinois at Chicago, 1985). Dissertation Abstracts International, 46, 2629 B. © 1997 Blackwell Science Ltd, Journal of Advanced Nursing, 25, 265–272