Quality in healthcare: process

Quality in healthcare: process

Best Practice & Research Clinical Anaesthesiology Vol. 15, No. 4, pp. 555±571, 2001 doi:10.1053/bean.2001.0191, available online at http://www.ideali...

180KB Sizes 1 Downloads 83 Views

Best Practice & Research Clinical Anaesthesiology Vol. 15, No. 4, pp. 555±571, 2001

doi:10.1053/bean.2001.0191, available online at http://www.idealibrary.com on

5 Quality in healthcare: process Malcolm Pradhan*

MBBS, PhD

Director Health Informatics and Associate Dean for Information Technology, Faculty of Health Sciences, Adelaide University Adelaide, SA 5005 Australia

Michael Edmonds

B Med Sci (Hons)

Health Informatics Unit, Adelaide University Adelaide, SA 5005 Australia

William B. Runciman BSc (med), MBBCh, PhD, ANZCA, FFICANZCA, FRCA, FHKCA Professor and Head Department of Anaesthesia and Intensive Care President Australian Patient Safety Foundation, Adelaide, SA 5001 GPO Box 400, Australia

Recent studies have shown startling rates of adverse events and preventable mortality in hospitalised patients around the world. Research using root cause analysis and incident monitoring has improved our understanding of why these errors occur. These approaches are useful in identifying contributing factors and stakeholders involved in adverse and sentinel events. The factors that contribute to these events have been well described, and range from institutional and management decisions down to the individual health care professionals involved and the environment they work in. We discuss healthcare processes, and present a proactive approach of work¯ow process modelling using sequence diagrams to identify the factors involved. These diagrams can then be used in conjunction with simple calculations for risk analysis to prioritise the value of interventions at di€erent steps of the healthcare process. We discuss how these analyses can be used to plan interventions to improve patient safety and quality of healthcare. Key words: safety; quality of healthcare; risk assessment; delivery of healthcare; process assessment (healthcare).

IF IT AIN'T BROKE. . . In the business and manufacturing world a lot of time and money is spent in understanding business processes. In a globalised economy ecient business processes can mean the di€erence between ®nancial viability and economic failure. Getting a business process right is critical for companies to reduce overheads and wastage and to maximise the utilisation of their employees. Some of the names of these concepts may *All correspondence to: Malcolm Pradhan. Tel: ‡618 8303 423; Fax: ‡618 8223 3870; Email: malcolm. [email protected] 1521±6896/01/040555‡17 $35.00/00

c 2001 Harcourt Publishers Ltd. *

556 M. Pradhan, M. Edmonds and W. B. Runciman

be familiar, for example, just-in-time manufacturing, supply chain management, knowledge management. We can de®ne business processes as mechanisms used by an organization to create and deliver products and services to its customers. A work¯ow is the description and automation of the procedural rules that de®ne how a process is implemented in an organisation, including how participants transfer communications, documents and tasks for action.1 In stark contrast to the business world, healthcare has remained relatively impervious to the waves of management trends with respect to understanding and improving business processes; on the contrary, healthcare has taken pride in not being a `business' at all. Arguably, healthcare processes have not changed signi®cantly over the last century. Whereas some of the information in health care is now computerised, the way in which we interact with the information and each other during the delivery of healthcare remains remarkably constant. One could be tempted to think that the old adage `If it ain't broke don't ®x it' applies to healthcare processes; but in the last decade we have increasing and consistent evidence that the healthcare system is not as healthy as it should be. The ®gures are staggering, with an estimated 98 0002 people who die prematurely in US hospitals each year because of preventable harm. Proportionate estimates have been derived for the UK3 and Australia.4 In addition to mortality, preventable errors in healthcare result in signi®cant morbidity to patients, and generate huge additional costs to employers, healthcare payers and the community. In Western healthcare an estimated 5% of a hospital's activity is devoted to ®xing its errors.5 Preventable harm in healthcare has been an active area of research in the last 10 years that has improved our understanding of why medical errors occur, using methods such as incident monitoring6 and root-cause analysis. A convenient framework for understanding factors related to the pathogenesis of preventable harm was developed by Vincent and colleagues7 and comprises: 1. Institutional factors, such as the development and use of protocols and policies, the recruiting, training and rostering of sta€, the availability and adequacy of supervision, the supply and storage of materials and funding; 2. Organisational and management factors, such as organisational culture and management structure. Organisational or corporate culture can encourage adverse outcomes with a `press on regardless' approach such that poorly prepared patients are still exposed to high-risk procedures. Management decisions will in¯uence the availability and condition of equipment and infra-structure, as well as the number and expertise of sta€ employed; 3. Work environment factors, for example excessive workloads, inadequate training and skills for certain tasks, equipment availability and maintenance; 4. Team factors, such as team structure and communication; 5. Task factors, for example availability of information, task complexity. These factors may be emphasised by time limitations, poorly designed or impractical protocols or checklists or excessive false alarms that may induce a health care worker to take a short cut or respond inappropriately in a potentially hazardous situation; 6. Patient factors, such as comorbidities, language and communication. The more complex a patient's care, the more likely the interaction between di€erent factors will lead to an adverse outcome. Institutional and organisational factors, and human factors that lead up to an event are often termed `latent' factors.8 In addition, Reason produced a model of

Quality in healthcare: process 557

process-related factors that lead to preventable error, which we have adapted to highlight the factors listed above (Figure 1). Figure 1 reveals that a combination of factors permits hazards to become actual losses, harm or injury; these include both errors in judgement and knowledge and problems in healthcare processes. Commonly preventable errors in healthcare have been viewed as the fault of individuals rather than the result of a complex set of systemic problems. It is interesting to note that most quality initiatives, such as practice guidelines and the entire evidence-based medicine (EBM) movement, target implicitly so-called `human factors,' that is, the cognitive and knowledge de®ciencies of healthcare professionals. Although human reasoning is well documented as fallible9, the assumption of many practice improvement movements is that practice variation and sub-optimal patient outcomes are due to the failing of the individual practitioner to keep up with the literature and to apply `best evidence'. Problems due to systemic failures and poor quality processes are often overlooked by traditional approaches to practice improvement ± it is easier to blame an individual rather than try to understand a complex system of interactions. It is unlikely that we will be able to remove human error from our daily work. The aim of risk management is to e€ect change in organisational culture, processes, and structures to manage e€ectively potential adverse events and the opportunities to mitigate these events.10 The goal of this chapter is to present a framework and methods to model and communicate clinical processes so that we can understand the risk associated with each step of the delivery of patient care. The goal of understanding processes in healthcare is to pro-actively put into place initiatives to minimize the e€ect of inevitable human error in a complex and under-resourced health system. REACTIVE AND PRO-ACTIVE METHODS FOR UNDERSTANDING HEALTHCARE PROCESSES In the past most mapping of healthcare processes has been reactive analysis, that is, in response to a `sentinel event' or other trigger. A sentinel event is one that leads to permanent disability or death. Root-cause analysis (mishap analysis) is a well-established reactive technique for investigating the systemic and human factors associated with adverse events and near misses. Some accreditation organisations, such as the American Joint Commission on Accreditation of Healthcare Organisations (JCAHO) and the UK National Health Service (NHS), have implemented policies to ensure that all sentinel events in institutions are followed up by a root-cause analysis and subsequent action plan to rectify the cause.11,12 Root-cause analysis aims to identify all of the contributing factors that lead to an event, from the `proximal' factors, such as the sta€ and patient involved, through to the `distal', or latent, factors such as management decisions and institutional factors that play a role in the long term. In its fullest form root-cause analysis is a complex and time-consuming process, but has great bene®ts in identifying all factors involved that lead up to adverse events. Root-cause analysis also identi®es and addresses all of the stakeholders involved who represent each of the di€erent factors outlined above from the institution and management down to the primary doctors, service units and the patients themselves. Quality managers or consultants, rather than clinicians, usually perform root-cause analyses. Another reactive analysis method that has been used extensively in healthcare is incident monitoring. After an incident occurs ± whether it is an actual adverse event or a near miss ± a description of the incident is submitted on a paper or electronic form by

Organisational and management factors

Approach to risk management

Resources, policies

Staffing levels, supervision

Maintenance of equipment

Latent Factors

Institutional factors Patient factors

Comorbidities, drug interactions

Judgement and interpretation

Figure 1. Factors for the pathogenesis of preventable harm.

Time and staffing pressures

Task complexity

Error Producing Conditions (Contributing factors)

Task, work factors

Errors

Near miss

Protocols, alerts, safety mechanisms

Defence barriers, risk management

Harm, Injury

558 M. Pradhan, M. Edmonds and W. B. Runciman

Quality in healthcare: process 559

one of the people involved in the incident. In many cases a quality manager or equivalent receives the incident description and uses incident monitoring software to identify and code key pieces of information into a database. Incident monitoring software, such as the Australian Patient Safety Foundation's AIMS2 system, provides a structured approach to the description and coding of adverse events into structures such as: contributing factors; patient factors (if the patient was the subject of occurrence); environmental factors; stang factors; incident type (for example a fall, or medication incident); outcomes for the patient; consequences for the organisation; and, any follow-up action that is to be taken. Although not as extensive in its scope as root-cause analysis, incident monitoring reveals insights about healthcare processes and leads to system improvements.13 Incident monitoring is useful is detecting the occurrence of adverse events, focusing mainly on the proximal, short-term factors of an event, that is the human behavioural side more than the distal institutional or latent factors. Recent implementations of incident monitoring14±16 include both human and systems factors and are aimed at identifying any event that could, or did, harm anyone, or lead to a complaint.17 Root-cause analysis can then be used to obtain more detail from some of these incident reports. Incident reports are usually of obvious, witnessed or documented events, such as patient injuries or medication errors. The majority of circumstances that lead to harm are not identi®ed, as is demonstrated by the example that `falls' make up 40% of incidents reported in this way18, but only 3% of adverse events identi®ed by retrospective medical record review.6 To improve the quality of healthcare using pro-active approaches, we identify processes that are in need of improvement, or where errors could result in particularly serious outcomes and put into place interventions that will reduce the chance of error. Interventions that have been e€ective in reducing errors include computerised physician order entry with alerts19, bar coding, practice guidelines and decision support systems.20 Work¯ow modelling is a method that helps us understand the processes of healthcare delivery so we can manage risk and plan pro-active interventions. There has been relatively little research in modelling healthcare processes in the published medical literature. Existing work tends to focus on the internal quality control within individual laboratories21, in pharmacy22, billing and in radiology23, but not in the modelling of processes in the overall delivery of patient care. Work¯ow models can be quanti®ed with local data from hospital information systems, with data from the published literature, and from expert opinion to explore `what if' scenarios for interventions, and for input into a prioritisation process. The prioritisation step can include the quanti®cation of time and cost for each step in the work¯ow process, error rates and stang levels. These data can then identify where there is unnecessary redundancy and inecient process, or excessive cost. This analysis can then be used to predict where in the process it is most likely for errors to occur, and, from this, to plan interventions to improve that step. Identifying and intervening at potentially risky steps of the process will improve the overall safety and eciency of the entire process, and ideally prevent error before it occurs. We have used process models to simulate the e€ects of organisational changes in hospitals to assist administrators and clinical managers decide between complex options as they attempted to reduce practice variation and cost. In this task we used process models as both communication tools to managers, and as computational devices to simulate the e€ect of interventions such as the introduction of guidelines, ward closures and sta€ changes. Through the use of hospital discharge data and an understanding of work¯ows in a particular hospital, we identi®ed that a signi®cant number of patients remained in hospital an extra night because doctors were conducting ward rounds late

560 M. Pradhan, M. Edmonds and W. B. Runciman

in the afternoon. Using our model to simulate the e€ect of earlier ward round times and discharge planning on nurse stang and bed occupancy we were able to quantify the potential cost saving of changing this practice pattern, in this case over 1% of the annual hospital budget, and about 20% of their current budget over-run. Whereas the behaviour of late ward rounds had been an annoyance in the past, the magnitude of the cost had never been calculated, and therefore this problem was never prioritised. Considering that a person has a 200-fold greater chance of dying as a result of the care process in hospital compared to being in trac5, minimising the admission time of patients is also a good outcome. UNIQUE FEATURES OF HEALTHCARE PROCESSES The goals of modelling business processes are to improve the eciency of the organisation, lower costs, reduce inventory, improve product delivery time and to promote innovation. In general, business process and work¯ow modelling in business focuses on the following three areas: the automation of documents and forms through approval processes; supporting the collaboration between employees; and improving communication within an organisation. In addition to these areas, the manufacturing industry models customer demand, development, production, and ordering. We are more interested in the reduction of errors than improvements in eciency, although it is possible to achieve both through thoughtful process re-engineering. Before discussing modelling techniques for healthcare processes, we will make some broad observations about the nature of the health system that make it a challenging domain in which to model and change business processes. Governance One of the unique characteristics of the healthcare system is the hierarchical nature of the medical and nursing sta€, and the clear delineation of responsibilities.24 This de®cit of real teamwork in healthcare means that healthcare processes in hospitals that consider the totality of patient care are rarely planned, but have evolved without overall clinical governance. In general, a single doctor or unit is responsible for the outcome of a patient (the primary doctor); however, during the care process numerous requests are made to other areas of the hospital to provide speci®c services, such as laboratory tests, pathology tests, radiology and speciality consultations (service units). Usually, the service units do not play a role in the care of the patient beyond the speci®c request from the primary doctor. For example, if a patient is sent for a chest X-ray before day surgery and the radiologist detects a tumour, the responsibility of the radiologist is to report the X-ray result and send the report back to the doctor in charge of the patient; often in the same manner as all the normal results. Although an individual radiologist might ring the doctor in charge of the patient to explain an unusual result, this action is not usually required as part of the business process and therefore cannot be relied upon. Communicating information We will use the term channel to describe the method by which information travels in the healthcare system. In most hospitals the most common channels are paper forms and conversations.25 An increasing number of institutions are improving their level of

Quality in healthcare: process 561

computerisation; however, the adoption for clinical data has been slow. A very common problem in healthcare is that very few channels of communication are used for all information, irrespective of the urgency or impact of the information. In our chest X-ray example, the ®nding of a tumour on a routine pre-operative test without a history of lung masses on the request form should inform the radiology department that a diagnosis of cancer is now highly signi®cant in the care of the patient. In most cases, however, the report is not sent via an `urgent' channel in which the timely receipt of the information by the referring doctor is acknowledged. If such an urgent channel existed, any delays in reading the report after a pre-determined time, e.g. 1 day, should automatically trigger a reminder to the primary doctor via page or mobile phone. In the situation where urgent channels already exist, they tend to be based on the referring unit, such as the emergency department, rather than based on the clinical utility of the information. Because institutions rarely prioritise information delivery and acknowledgement of receipt, patients bear the unnecessary risk of clinically signi®cant results being lost in the background of normal results that have lower clinical importance. From a risk management perspective, the value of information of a test result depends on the ability of that information to change the current decision, and the magnitude of the consequences if the information is ignored; healthcare processes do not support directly the special treatment of high value information. Changing practice It is much easier to model healthcare processes than it is to change them. Compliance to new work¯ows and business practices is a major stumbling block in the implementation of clinical information systems and best-practice guidelines.26,27 It is worth noting that the current state of the patient not only determines the importance of information, like test results, but it also determines the likelihood of sta€ compliance to new processes. In the case of our pre-operative patient, if the screening anaesthetist excluded the patient from surgery while awaiting a test result from radiology it is more likely that the test result will be checked; the referring doctor is now waiting on the result before she can inform the surgeon, the patient and the operating theatre. If, however, the patient was allowed to remain on the operating list unless the new test result was signi®cantly abnormal, then the test is more likely to go unchecked because so many tests are ordered with little clinical utility.28,29 In this example the patient would be waiting a test to `rule-in' the surgery rather than `rule-out', and therefore there is more motivation by stakeholders to check the result. Stakeholder analysis is a critical step in business process re-engineering, and aligning stakeholder's values to the desired result is critical to the implementation of risk mitigation techniques. It is often easier to change the drivers and organisational structure to support a particular practice than it is to change each individual at a time.30 Computer information systems are widely used in healthcare today for returning laboratory results, and for administration purposes. Computerisation provides a convenient and potentially e€ective point of intervention to provide automation and support for healthcare processes. One of the main reasons for modelling healthcare processes is to target points in the work¯ow for intervention. Computerisation has demonstrated signi®cant bene®ts for physician order entry20,31,32, both for patients owing to reduced adverse drug reactions, and for institutions through reduced length of stay. Beyond physician order entry the opportunities for interventions, particularly in an inpatient setting, become harder for two reasons. The ®rst is that current clinical information systems do not capture key clinical detail using controlled terminology

562 M. Pradhan, M. Edmonds and W. B. Runciman

that can be used by decision support systems to automate work¯ow and reduce error. The second factor relates to the challenge of integrating information technology into the current work¯ow processes of clinicians in a way that ensures high compliance ± it is much harder to integrate computers into ward rounds than in outpatient clinics where doctors work from a desk. Clinical context Hospital information systems often only collect information for administration purposes, such as the patient's name, age, marital status and address. For work¯ow process purposes the clinical detail of the patient is required to determine the complexity of the case and the subsequent potential risks. The clinical context of a process is the variable complexity of each patient, such as comorbidities, concurrent medications and psychosocial factors that can make outcomes dicult to predict. The variability between patients means that the same process can have di€erent outcomes and there could be a need for various strategies to improve the health outcomes for each patient. This clinical information that is lacking from hospital information systems is vital for accurate work¯ow process modelling and analysis. Collection of this information would allow for automation of risk assessment. The lack of clinical details in the information systems also limits the interpretation that service units, such as laboratories, radiology or specialist consults, can apply to the results they generate. These service units depend on the clinical detail being provided on request forms, where it is often neglected. This lack of information makes it dicult for the service units to place a value of information on the test result, and there is a consequent lack of alerting systems for abnormal results that will change clinical decision making. Where clinical detail is available on the hospital information system, it is often dicult to capture it because of a lack of controlled terminology. METHODS FOR VISUALISING BUSINESS PROCESSES AND CALCULATING RISK A model is a simpli®cation of reality that enables communication and should improve our understanding of the problem area; it is perilous to assume that any model represents reality accurately. The purpose of a model is to gain shared insight into a problem, and to assist in decision-making; a good model is one that is useful in solving a problem, and a single model might not be useful in multiple problem areas without modi®cation. We simplify reality by making assumptions, and in any modelling procedure all assumptions should be explicit so those using the results of the analysis understand its limitations and its strengths. It is important to note that making the correct assumptions, and testing them, requires skill and practice. We promote graphical representations of our models so they are easier to communicate and therefore easier for people to understand what assumptions have gone into the creation of the model. Although most of the technical work in modelling is often carried out in a spreadsheet, our experience has been that a clear graphical representation of our models engages clinical sta€ and administrators alike in a way that a spreadsheet cannot. More signi®cantly, we are able to engage a wider pool of expertise to improve and re®ne our models through graphical methods before dwelling on the mathematics. To demonstrate practical techniques for modelling and improving healthcare processes we will use an example of a pre-operative screening clinic.

Quality in healthcare: process 563

Example: pre-operative screening clinic Many large hospitals run a pre-operative screening clinic to assess a patient's ®tness for day surgery, to reduce the risk of preventable error. The clinic work¯ow is quite complicated and demonstrates many features that exist in other areas of inpatient and outpatient care. The patient is interviewed by the clinic clerk, nurses, the surgical team and anaesthetists, and further information is requested and chased up from specialist consults, general practitioners, radiology departments and the biochemistry laboratories. The collation of this information is time-limited by the interval until the day of surgery, and is dependent on a complex information ¯ow and inter-professional dynamics. This process has many steps that are open to error, and is often held together by two or three dedicated team members coordinating the patient and information ¯ow so no information is lost. Using sequence diagrams for modelling healthcare processes We have been using sequence diagrams to model the process and information ¯ow between all of the stakeholders involved in the healthcare of a patient. From the work¯ow sequence model it is possible to identify high-risk processes, and therefore where interventions can be most bene®cial to reduce risk and improve patient safety. Importantly, the work¯ow sequence diagram is a good communication tool that highlights the interplay between di€erent stakeholders in an episode of patient care (Figure 2). The work¯ow sequence diagram is derived from the Uni®ed Modelling Language33, a set of diagrams and semantics used in software engineering. Software engineers discovered that while an organisational structure represents the position of software entities in a hierarchy, it does not provide insight into the way the entities interact to ful®l a task. Similarly in business, the organisational structure of a hospital or unit does not inform us how units and individuals interact to deliver patient care. Sequence diagrams are one of the graphical representations developed to demonstrate dynamic interaction of messages between entities in complex systems. Numerous other diagram formats exist but we have found the sequence diagram to be a good trade-o€ between accuracy and clarity. In sequence diagrams, entities that play a role in the process being modelled are represented along the top of the diagram. In healthcare processes we usually model units, individuals, or even organisations. For the episode of care being modelled, we represent an active participant in the process using a hollow rectangle. If a unit is involved in an episodic manner, then there can be several smaller rectangles, which can quickly identify problems with continuity of care. In the case of the pre-operative screening clinic, the episodic activity of the anaesthetist demonstrates clearly that there might be several di€erent anaesthetists involved in the care of the patient over time. Each horizontal arrow represents a message. In work¯ow sequence diagrams, we use solid arrows to represent requests for information or services and dashed arrows to represent new information being returned or delivered. The following patterns in work¯ow sequence diagrams represent work¯ow that may have a higher than expected risk: . Long vertical gaps between information request and information delivery signal time lapses that will make the information easier to overlook or forget in the absence of a reminder or alerting system;

full history, examination

lab/X-ray forms

In hospital Laboratory

lab results radiology results

patient goes to radiology

Day of operation

review of patient notes

Figure 2. Sequence diagram for a pre-operative screening clinic.

patient comes for operation

follow up of results

return call

return call

Specialist

community General Practitioner

review of patient notes

phone call if difficulty expected

Radiology

End of pre-operative screening session

blood sample, request form

surgical notes

previous reports from lab/radiology

anaesthetic forms/notes

lab/radiology results lab/radiology results

ECG

lab/radiology results

consent

procedure, risks, complications

waiting time blood sample, ECG

waiting time

procedure, risks, side-effects consent

Case Notes

phone calls to GPs/specialists

GP/specialist letters

Surgical Intern

previous reports from lab/radiology

Anaesthetist

In unit

questionnaire, obs

history, questionnaire, examination

waiting time

pre-op procedure, post-op care

questionnaire, obs, history

patient questionnaire

Patient Clinic Clerk Nurse registration computer printout

Anaesthetist at operation

operating theatre Surgeon

564 M. Pradhan, M. Edmonds and W. B. Runciman

Quality in healthcare: process 565

. The entity requesting the information is di€erent to the entity receiving the information. This makes overlooking results much more likely; . The information request is to an entity outside of the unit/organisational boundaries. For example, in a pre-operative clinic the anaesthetist might be required to contact a general practitioner or a specialist for patient information. If this information is not returned immediately, then the likelihood of information being lost or mistranscribed increases.

Figure 2 shows an example of complicated work¯ow that is open to risk and error at each of the points above. The pathway that blood test and other investigation requests follow from the requesting doctor, in this case the anaesthetist, is complex and inecient. There is a long vertical gap, or time lapse, the person receiving the information may not be the same that requested it, and it depends on an entity outside the unit. In this part of the work¯ow process the anaesthetist makes the request, and passes the paper form onto the nurse. The nurse takes a blood sample at the end of the patient's pre-operative screening session, and, every so often, all of the samples are sent to the laboratory. The laboratory sends the results of the test back to the preoperative clinic clerk, who ®les them with the patient's case notes. The nurse then checks these notes and passes on any signi®cant ®ndings to the anaesthetist, who will report their interpretation of these to the case notes. This process has risk in its many steps and over the long period it unfolds. Some pre-operative screening clinics have a mechanism in place to ensure results get back to the anaesthetist by using `show me' orders which require all results for that patient to be reviewed by the anaesthetist. These orders are normally only attached to high-risk patients or those where the results can change the decision to go ahead with surgery, but this does not account for all important test results. These orders are an example of an attempt at creating an `urgent' channel, but would not cover all potentially risky situations. From our analysis of a large pre-operative screening clinic we observed that the complicated work¯ow is largely unsupported by alerts and decision support. This lack of support places large demands on individual anaesthetists and nurses; the clinic has a reputation as a challenging place in which to work. We have found that the prevention of errors relies on the dedication of a few individuals to assiduously follow up results and to remember speci®c patient problems. The e€ectiveness of the clinic is therefore at considerable risk if sta€ changes or sta€ illness occur. Points for intervention may include increased automation, the ability for anaesthetists to highlight certain patients as important for follow up, the ability to mark certain patients as `rule-in', meaning that the operation is on hold unless certain criteria are met with respect to their health state. All of these interventions require investment but unfortunately the bene®ts accrued by the hospital through decreased length of stay resulting from e€ective pre-operative screening are not re-invested into improving the infra-structure of the clinic. It is not uncommon for healthcare organisations to have a `silo' approach to funding that hinders the improvement of quality in healthcare delivery. Simple calculations for risk analysis In order to prioritise interventions for system improvement, the risks and bene®ts of the process should be quanti®ed in some way. The degree of detail in the quanti®cation process depends very much on the purpose. Managers may be satis®ed with a broad indication of cost and bene®t for small changes that require little investment; for expensive or risky interventions they might require a much more detailed analysis with

566 M. Pradhan, M. Edmonds and W. B. Runciman Table 1. Likelihood/consequence grid for risk management. Consequences Likelihood

Insigni®cant

Minor

Moderate

Major

Catastrophic

A (almost certain) B (likely) C (moderate) D (unlikely) E (rare)

High Moderate Low Low Low

High High Moderate Low Low

Extreme High High Moderate Moderate

Extreme Extreme Extreme High High

Extreme Extreme Extreme Extreme High

Extreme ˆ extreme risk; immediate action required; High ˆ high risk; senior management attention needed; Moderate ˆ moderate risk; management responsibility must be speci®ed; Low ˆ low risk; manage by routine procedures.

detailed ®gures. Our approach is to start simply and to respond with the amount of detail required by decision makers. We have found, however, that the more local data that are used in the quanti®cation process the more compelling the analysis is to local stakeholders. Having an opinion leader supporting the proposed changes is also an important ingredient for success.27 Accessing local data can be complicated and can require information technology skills to clean and analyse data from hospital information systems. A simple and often e€ective way of prioritising quality improvement initiatives is to use a likelihood/consequences grid as described in the Australian Standards Risk Management framework.10 Table 1 is an example of the grid. In Table 1 the consequences range from insigni®cant to catastrophic that could involve large loss of life. In a similar way costs and risks of interventions can be categorised. If we move beyond this categorisation approach to more detailed risk analysis we ®nd that techniques can require complex mathematics to calculate accurately in the presence of inter-dependencies between processes, hierarchical models, and when forced to estimate variables from data or from indirect information. In recent years, computer intensive techniques have allowed expert modellers to construct and calculate complex risk models, even in cases where the distributions of information are non-linear and non-normal. However, these specialised techniques are well beyond the scope of this chapter, and we will concentrate on a simple method for calculating overall risk of failure in a healthcare process. In order to simplify the calculations we will assume that each step in the work¯ow is independent of other steps. So, if an individual patient has a blood test and an X-ray, the chance that the report is late in moving from the biochemistry laboratory to the primary doctor is not in any way in¯uenced by the report from radiology to the primary doctor. Although the calculations listed below might seem complicated, you will see from the worked examples that their use is relatively straightforward. Let us consider that an episode of patient care comprises a collection of m processes, where m is some number. The exact value of m can be obtained easily from sequence diagrams. Let us de®ne a process failure as an event when a process does not meet prede®ned quality standards, such as timeliness or accuracy. We will say that an error occurs if any one of the m processes fails to meet pre-de®ned standards. Therefore, the probability of an error occurring in an episode of patient care is the result of a failure in

Quality in healthcare: process 567

any of the 1 to m processes that comprise the episode. This can be calculated using equation (1) as follows: Pr‰ErrorŠ ˆ 1

m Y iˆ1

1

  Pr failurei :

…1†

In words, the probability of Error is one minus the probability that no failure will occur in each of the m processes that comprise the episode of care. For example, if every 5 out of every 100 laboratory results were delayed the probability of failure for the lab is Pr(failurelab) ˆ 0.05. Similarly, if delays in radiology reports were also 5 in 100 then Pr(failurexray) ˆ 0.05, and if doctors do not follow up 20 in every 100 test results in a timely manner then Pr(failureclinical) ˆ 0.20. Using equation (1) we can calculate the error rate resulting from a failure in any of these steps: h i      Pr‰ErrorŠ ˆ 1 1 Pr failurelab 1 Pr failurexray 1 Pr failureclinical ˆ1 ˆ1

…1 0:05†…1 0:05†…1 …0:95†…0:95†…0:80†

0:20†

ˆ 1 0:722 ˆ 0:278 The resulting error rate from this example is 27.8%. If there exist di€erences in the way patients are managed, then the expected error rates can be calculated for each situation in a similar way. To simplify calculations, we have used point or discrete probability estimates that do not take account of variation, which may be non-normal. To achieve error bounds, the above calculation can be repeated for worst-case values and best-case values, with the median values the most likely. But what is the e€ect of this error rate? Is it acceptable or dangerous? Process control charts, including those used in infection control, use an arbitrary value of 2 standard deviations to determine if something is `signi®cant'. The use of 2 standard deviations as a cut-o€ for importance is mathematically convenient but it does not re¯ect the actual signi®cance of the events in terms of the patient or the organisation, and is therefore inappropriate for risk management. In healthcare evaluation there is often confusion between a measurement, for example the number of infections that occur, and its value, the morbidity and increased length of stay resulting from the infection. Utility theory34,35 is a well-developed framework that assists us in moving from measurements, such as error rates, to values, or the impact of the errors. Multi-attribute utility theory is a method that incorporates complex value models in decision-making and is useful in healthcare contexts.36 For illustration purposes we will take the perspective of an organisation and use dollars to signify values, or in this case the loss to the institution due to errors in healthcare delivery. By combining the error rates derived using equation (1) and the loss to the institution due to errors, we can obtain a prioritisation list to determine how important it is for us to improve a speci®c work¯ow, and how much we should spend in doing so. It is important to understand that the goal of risk management is to reduce the probability of errors occurring and to maximise the likelihood of a good outcome for the patient; however, we cannot guarantee a good outcome. Good decision making and healthcare processes can still lead to bad outcomes for the patient. Similarly, bad processes can still result in good outcomes for patients.

568 M. Pradhan, M. Edmonds and W. B. Runciman

Let us de®ne a loss for an institution as the cost incurred for a preventable error as a result of a healthcare process. In the majority of cases, errors do not result in signi®cant morbidity to the patient;37 however, in particular patients the same error can result in a fatal injury often because of patient factors such as comorbidities. Let us assume that the patient complexity (Pt.Complexity) determines how sensitive a patient is to errors in the healthcare process and is measured by the loss incurred for that subgroup, and that there are n sub-groups of patients. The expected loss (ELoss) to the institution can be calculated by using the error rate previously calculated (Pr[Error]), the loss due to the error for patients in a sub-group of a given complexity (Loss[Pt.Complexity]), and the proportion of patients within the sub-group (Pr[Pt.Complexity]): n h i h i X Loss Pt:Complexityj  Pr Pt:Complexityj …2† ELoss ˆ Pr‰ErrorŠ  jˆ1

An example will make this clear. Let us say that in an adult tertiary hospital with an error rate for the timely checking of test results of 0.278 we have two sub-groups of patients who are going for day surgery. The majority (90%) are uncomplicated patients with little comorbidity for whom missing a test result could incur, on average, an increased length of stay of 0.2 days at $1000 per day, or $200. A second smaller group of patients (10%) have signi®cant comorbidities, and failure to check test results before surgery can incur an increased length of stay of 4 days at $1000 per day, or $4000. The ELoss per patient is therefore: ELoss ˆ …0:278†  ‰…0:9†…$200† ‡ …0:1†…$4000†Š ˆ …0:278†  ‰$580Š ˆ $161:24 The expected loss per patient is approximately $160. If the hospital sees approximately 10 000 patients per year, the loss will be $1 600 000. This ®gure is a direct cost, ignoring the cost of litigation, patient morbidity and costs to patients and employers. In a children's hospital where the majority of patients are uncomplicated (say 99%) and increased length of stay would be 0.1 days for this subgroup, the expected loss for 10 000 patients is approximately $400 000. The priority for improving this process would be lower in the children's hospital. The assumptions in our very simple example are very signi®cant as we have not modelled explicitly the cost of death. Death or severe morbidity must be carefully modelled as a special case, usually setting these events to a very high cost, even though the direct costs can be low; if a patient dies, their length of stay will be lower than if they were mildly injured. Consequently our use of length of stay as a surrogate measure of outcome will imply that death is a better outcome than mild injury. CONCLUSION Healthcare processes are complex because they involve many participants dealing with complex and important information. Although EBM and guidelines are important initiatives to reduce practice variation and human error, we acknowledge that human error will occur and that the healthcare process should be designed to minimise the harmful e€ects of this. In contrast to the business world in which there is a large

Quality in healthcare: process 569

investment in improving business processes, healthcare processes remain dicult to understand and to change. We have presented some techniques derived from management, software engineering, risk management and statistics to assist in understanding the risks and points for intervention in healthcare processes. The motivation for us all is to improve the quality of healthcare in a cost-e€ective manner. The reality is that healthcare has numerous barriers to change and that communication is a vital part of the process of system improvement. We stress that improving healthcare processes is a multidisciplinary task, and any initiative will bene®t greatly from a team approach: management sponsorship; quality managers; technical support; and strong clinical opinion leader support. Both reactive and pro-active techniques are important parts of the quality improvement toolkit. The tools we have described in this chapter are aimed at improving the communication of complex processes, and, if necessary, methods to quantify risk and cost. SUMMARY Recent studies have shown unacceptable rates of error in healthcare and subsequent adverse outcomes, including death. It has been well described that the majority of these errors are committed by people working in an imperfect system that is prone, or even conducive, to error. The situation is also unique in the healthcare setting because of factors such as clinical governance, poor communication channels, rapidly changing practice and the overall clinical context. Previous approaches to this problem have mainly been reactive techniques in response to adverse and sentinel events, such as root-cause analysis and incident monitoring. In this chapter we have outlined a pro-active method for modelling work¯ow process in healthcare that can be used to communicate process and identify potentially risky steps where intervention would be most bene®cial. Using the example of a preoperative screening clinic, we have shown that combined with simple risk calculations, these sequence diagrams are useful in risk analysis, not only from a measurement point of view, but also from a value point of view. Value is important in any decision making, and the perspective of all stakeholders must be considered, from institutional needs, down to the healthcare professionals and the patients themselves. Quantifying the work¯ow process models to incorporate all of these values can be a dicult and timeconsuming job, but the potential bene®ts from well-planned interventions based on these analyses are numerous. The methods described in this chapter should be used along side data collection and reactive methods for patient safety as part of a co-ordinated approach to identify and prevent error, to improve patient safety, to reduce cost, and to improve the overall quantity of healthcare. Practice points . encourage the implementation of in-house monitoring of adverse events and near-misses to identify common problems and patterns . encourage a no-blame culture for those who report adverse events and near misses - identify problems in the system not the person . use local data to understand potential areas for risk reduction . involve quality managers and involved sta€ to model and communicate clinical areas where risk is identi®ed and plan interventions for maximal bene®t

570 M. Pradhan, M. Edmonds and W. B. Runciman

Research agenda in trials related to quality measures, do not simply choose a `signi®cance' value of 0.05. Use the expected loss and bene®t to decide whether an initiative is clinically signi®cant. Current settings of alpha (0.05) and beta (0.8) are arbitrary and may not make sense for quality initiatives ± another form of medical ritual

REFERENCES 1. Fischer L. The Work¯ow Handbook 2001. Florida: Future Strategies Inc, 2000. 2. Kohn L, Corrigan J & Donaldson M. To Err is Human: Building a Safer Health System. Washington, DC: National Academy Press, 1999. 3. Department of Health. An Organisation with a Memory. London: Department of Health, 2000. * 4. Wilson RM, Runciman WB, Gibberd RW et al. The Quality in Australian Health Care Study. Medical Journal of Australia 1995; 163: 458±471. 5. Runciman W. Iatrogenic injury in Australia. Adelaide, South Australia: Australian Patient Safety Foundation, 2001, in press. 6. Webb RK, Currie M, Morgan CA et al. The Australian Incident Monitoring Study: an analysis of 2000 incident reports. Anaesthesia and Intensive Care 1993; 21: 520±528. * 7. Vincent C, Adams S & Stanhope N. A framework for the analysis of risk and safety in medicine. British Medical Journal 1998; 316: 1154±1157. * 8. Reason J. The contribution of latent human failures to the breakdown of complex systems. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences 1990; 327: 475±484. 9. Tversky A & Kahneman D. Judgment under uncertainty: heuristics and biases. Science 1974; 185: 1124±1131. *10. Standards Australia. Risk Management (AS/NZA 4360:1999). Strath®eld, NSW: Standards Association of Australia, 1999. 11. JCAHO. Sentinel Event Policy and Procedures. JCAHO, 1998. 12. NHS. Building a Safer NHS for Patients. Department of Health, 2001. *13. Runciman WB, Webb RK, Lee R & Holland R and The Australian Incident Monitoring Study. System failure: an analysis of 2000 incident reports. Anaesthesia and Intensive Care 1993; 21: 684±695. 14. Chopra V, Bovill JG, Spierdijk J & Koornneef F. Reported signi®cant observations during anaesthesia: a prospective analysis over an 18-month period. British Journal of Anaesthesia 1992; 68: 13±17. *15. Runciman WB. Report from the Australian Patient Safety Foundation: Australasian Incident Monitoring Study. Anaesthesia and Intensive Care 1989; 17: 107±108. 16. Short TG, O'Regan A, Lew J & Oh TE. Critical incident reporting in an anaesthetic department quality assurance programme. Anaesthesia 1993; 48: 3±7. 17. Runciman W. Incident monitoring, clinical anaesthesiology. Quality assurance and Risk Management in Anaesthesia 1996; 10: 333±356. 18. Goodwin MB & Westbrook JI. An analysis of patient accidents in hospital. Australian Clinical Review 1993; 13: 141±149. *19. Bates DW, Teich JM, Lee J et al. The impact of computerized physician order entry on medication error prevention. Journal of the American Medical Informatics Association 1999; 6: 313±321. *20. Hunt DL, Haynes RB, Hanna SE & Smith K. E€ects of computer-based clinical decision support systems on physician performance and patient outcomes: a systematic review. Journal of the American Medical Association 1998; 280: 1339±1346. 21. Truchaud A, Le Neel T, Brochard H et al. New tools for laboratory design and management. Clinical Chemistry 1997; 43: 1709±1715. 22. Peterson GM, Wu MS & Bergin JK. Pharmacist's attitudes towards dispensing errors: their causes and prevention. Journal of Clinical Pharmacy and Therapeutics 1999; 24: 57±71. 23. Schmidt J, Meetz K & Wendler T. Work¯ow management systems±a powerful means to integrate radiologic processes and application systems. Journal of Digital Imaging 1999; 12: 214±215. 24. Sexton JB, Thomas EJ & Helmreich RL. Error, stress, and teamwork in medicine and aviation: cross sectional surveys. British Medical Journal 2000; 320: 745±749.

Quality in healthcare: process 571 25. Parker J & Coiera E. Improving clinical communication: a view from psychology. Journal of the American Medical Informatics Association 2000; 7: 453±461. 26. Davis D, O'Brien MA, Freemantle N et al. Impact of formal continuing medical education: do conferences, workshops, rounds, and other traditional continuing education activities change physician behavior or health care outcomes? Journal of the American Medical Association 1999; 282: 867±874. 27. Davis DA & Taylor-Vaisey A. Translating guidelines into practice. A systematic review of theoretic concepts, practical experience and research evidence in the adoption of clinical practice guidelines. Canadian Medical Association Journal 1997; 157: 408±416. 28. Wattsman TA & Davies RS. The utility of preoperative laboratory testing in general surgery patients for outpatient procedures. American Surgeon 1997; 63: 81±90. 29. Dzankic S, Pastor D, Gonzalez C & Leung JM. The prevalence and predictive value of abnormal preoperative laboratory tests in elderly surgical patients. Anesthesia and Analgesia 2001; 93: 301±308. 30. Reason J. Human error: models and management. British Medical Journal 2000; 320: 768±770. 31. Evans RS, Pestotnik SL, Classen DC et al. A computer-assisted management program for antibiotics and other anti-infective agents. New England Journal of Medicine 1998; 338: 232±238. 32. Gawande A & Bates D. The use of information technology in improving medical performance Part II. Physiciansupport tools. Medscape, 2000. 33. Booch G, Jacobson I, Rumbaugh J & Rumbaugh J. The Uni®ed Modeling Language User Guide. New York: Addison-Wesley, 1998. 34. von Winterfeldt D & Edwards W. Decision Analysis and Behavioral Research. New York: Cambridge University Press, 1986. *35. Naglie G, Krahn MD, Naimark D et al. Primer on medical decision analysis: Part 3±Estimating probabilities and utilities. Medical Decision Making 1997; 17: 136±141. 36. Keeney RL & Rai€a H. Decisions with Multiple Objectives: Preferences and Value Tradeo€s. New York: Wiley and Sons, 1976. *37. Bates DW. Medication errors. How common are they and what can be done to prevent them? Drug Safety 1996; 15: 303±310.