Thomas and Petersen (2003) classified methods of studying errors and adverse events into eight broad groups and reviewed the respective advantages and disadvantages of each method. In their paper, they use the term error to include terms such as mistakes, close calls, near misses and factors that contribute to error. In a later chapter we will discuss the difficulties of defining and classifying errors, but in this section the term error is used as a catch-all for any incident that does not involve patient harm. The terms ‘near miss’ and ‘close call’ are seldom clearly defined but broadly speaking refer to incidents in which harm was only narrowly avoided; this includes both incidents which never developed to the point of actually harming a patient and those in which prompt action averted disaster.

Hindsight bias is mentioned several times in the table. What is hindsight bias? The term derives from the psychological literature and in particular from experimental studies showing that people exaggerate in retrospect what they knew before an incident occurred–the ‘knewitall along’ effect. After a disaster, with the benefit of hindsight, it all looks so simple and the ‘expert’ reviewing the case wonders why the clinician involved couldn’t see the obvious connections. Looking back, the situation actually faced by the clinician is inevitably grossly simplified. We cannot capture the multiple pathways open to the clinician at the time or the unfolding story of a clinical encounter. Still less can we capture the pressures and distractions that may have been affecting clinical judgement, such as fatigue, hunger and having to deal with several other patients with complex conditions.

Retrospective reviews of medical records aim to assess the nature, incidence and economic impact of adverse events and to provide some information on their causes. Adverse events are defined as an unintended injury caused by medical management rather than the disease process that results in some definite injury or, at the very least, spent on additional days in hospital (Box 4.1). Definitions are critical in patient safety and one has to be constantly aware of differences in terminology. For instance, a study by Andrews et al. (1997) in the United States showed a 17.7% rate of serious adverse events in a surgical unit, much higher than most other studies. However, their definition of adverse event was different from that usually employed and they used observation rather than record review, as most other studies do. These are not flaws; the study is a good one. The point is that one has to be careful about definitions when interpreting findings and comparing studies.

The basic record review process is as follows. In phase I, nurses or experienced record clerks are trained to identify case records that satisfy one or more of 18 well-defined screening criteria – such as death, transfer to a special care unit or re-admission to hospital within 12 months. These have been shown to be associated with an increased likelihood of an adverse event (Neale and Woloshynowych, 2003). In phase II, trained doctors analyse positively screened records in detail to determine whether or not they contain evidence of an adverse event using a standard set of questions. The basic method has been followed in all the major national studies, though modifications of the review form and data capture have been developed (Woloshynowych, Neale and Vincent, 2003). In France, Phillipe Michel used prospective review, in the sense that record review is carried out close to the time of discharge on a previously defined set of patients and, in some cases, combined with interviews with staff (Michel et al., 2004).

As the examples show, many patients suffer increased pain and disability from serious adverse events. They often also suffer psychological trauma and may experience failures in their treatment as a terrible betrayal of trust. Staff may experience shame, guilt and depression after making a mistake, with litigation and complaints imposing an additional burden (Vincent, 1997). These profoundly important aspects of patient safety, generally given far too little attention, are considered in Chapters 8 and 9.

The financial cost of adverse events, in terms of additional treatment and extra days in hospital, are considerable and vastly greater than the costs of litigation. One of the most consistent findings from the record reviews is that, on average, a patient suffering an adverse event stays an extra six to eight days in hospital. An extra few days in hospital is, clinically speaking, an unremarkable event and it is not necessarily particularly traumatic or unpleasant for the patient. However, when the sums are done and the findings extrapolated nationally the costs are staggering. In Britain, the cost of preventable adverse events is £1 billion per annum in lost bed days alone (Vincent, Neale and Woloshynowych, 2001). The wider costs of lost working time, disability benefits and the wider economic consequences would be greater still. The Institute of Medicine report (1999) was able to estimate that in the United States total annual national costs (lost income, lost household production, disability, healthcare costs) were between $17 billion and $29 billion for preventable adverse events and about double that for all adverse events; healthcare costs accounted for over one half of the total costs incurred. Even when using the lower estimates, the total national costs associated with adverse events and preventable adverse events represent approximately 4% and 2%, respectively of national health expenditure (Kohn, Corrigan and Donaldson, 1999).

Costs of direct hospital care, essentially additional time in hospital, have recently been estimated from the Dutch adverse events study finding that about 3% of all bed days and 1% of the total health budget could be attributed to preventable adverse events. The real overall costs are probably a good deal higher, as this estimate does not include additional treatments and investigations or any of the associated societal costs discussed above. Remember also that these estimates are confined to the hospital sector; we have no idea of the additional costs of adverse events in primary care or mental health.

A significant percentage of adverse events are associated with a surgical procedure. For instance, in the Utah Colorado Medical Practice Study, the annual incidence rate of adverse events amongst hospitalized patients who received an operation was 3%, of which half were preventable. Some operations, such as extremity bypass graft, abdominal aortic aneurysm repair and colon resection, were at particularly high risk of preventable adverse events (Thomas et al., 2000b; Thomas and Brennan, 2001).

In the United Kingdom, complication rates for some of the major operations are 20–25% with an acceptable mortality of 5–10% (Vincent et al., 2004). However, at least 30–50% of major complications occurring in patients undergoing general surgical procedures are thought to be avoidable. In Canada, Wanzel et al. (2002) prospectively monitored the presence and documentation of complications for all 192 patients admitted over a two month period to a general surgical ward. 75 patients (39%) of the patients suffered a total of 144 complications, 2 of which were fatal, 10 life threatening and 90 of moderate severity. Almost all the complications were documented in the patient’s notes but only 20% were reviewed at the weekly morbidity and mortality rounds; about one-fifth of complications were due, in part, to error. Many adverse events classified as operative are, on closer examination, found to be due to problems in ward management rather than intra-operative care. For instance Neale, Woloshynowych and Vincent (2001) identified preventable pressure sores, chest infections, falls and poor care of urethral catheters in their study of adverse events, together with a variety of problems with the administration of drugs and intravenous fluids.

Retrospective review of medical records, like any other research method, has its limitations and the findings of the studies have to be interpreted with due regard to the methodological limitations. Adverse events that are not recorded in the notes, or at least cannot be discerned from the notes, will not be detected and therefore record review probably provides a lower estimate of the actual scale of harm. The process of record review also necessarily relies on implicit clinical judgement, and agreement between reviewers, particularly on judgements of preventability, have often only been moderate (Neale and Woloshynowych, 2003). Great efforts have been made to strengthen the accuracy and reproducibility of these judgments by training, by the use of structured data collection, by duplicate review with re-review and by resolution of disagreements; however, even with training the reliability of such judgments is only moderate. Nevertheless, following a series of careful studies, Kieran Walshe concluded that the recognition of adverse events by record review had moderate to good face, content and construct validity with respect to quality of care in a hospital setting (Walshe, 2000).

These and other methodological issues have come to the fore in debates about the number of deaths due to adverse events, particularly after the headline capturing claims that up to 98000 Americans were dying each year from adverse events in hospital. The methodological arguments are too complex to summarize in their entirety here, but it is important to note that the figures have been challenged, and to give a flavour of the arguments. For instance, one research team argued, following estimates of the death rates in hospital at the time of the study, that the patients who reportedly died from adverse events in the Harvard study were already severely ill and likely to die anyway (McDonald, Weiner and Hui, 2000). In a further challenge to the figures, Hayward and Hofer (2001) compared the findings with their own review of the standard of care of patients who died in hospital while having active, as opposed to palliative, care. They found that only 0.5% of patients would have lived longer than three months, even if they had all had optimal care. So, yes, there were some deaths which were perhaps preventable, but the great majority of these people were already very ill and would have died anyway.

In a reply to McDonald and colleagues, Lucian Leape (2000) noted that some people seemed to have the impression that many of the deaths that had been attributed to adverse events were minor incidents in the care of people who were severely ill and likely to die anyway. He pointed out that terminally ill patients had been excluded from the study, but agreed that there were a small group of patients (14% of deaths attributed to adverse events) who had been severely ill; for these patients the adverse event had tipped the balance. However, for the remaining 86%, the deficiencies in the care they received were a major factor leading to the death:

Examples include a cerebrovascular accident in a patient with atrial flutter who was not treated with anticoagulants, overwhelming sepsis… in a patient with signs of intestinal obstruction that was untreated for 24 hours, and brain damage from hypotension due to blood loss from unrecognized rupture of the spleen.
(LEAPE, 2000)

The issue of the incidence of adverse events in patients who died and their preventability has been addressed more recently in a major study in The Netherlands (Zegers et al., 2009). The records of 7926 patients were reviewed across 21 hospitals: 3943 admissions of discharged patients and 3983 admissions of hospital patients who died in 2004. A large sub-sample of deceased hospital patients was included to determine the incidence of potentially preventable deaths more precisely compared to previous international studies. Of these patients, 663 experienced a total of 744 adverse events, with 10% of patients suffering two or more adverse events. For deceased patients, the incidence of adverse events was 10.7%, and a rate of 5.2% for preventable adverse events. The incidence of adverse events was significantly higher therefore than for living patients. About half of the patients with preventable adverse events had a life expectancy of more than one year; the exact contribution of the adverse event to the death is not clear but the implication is that life was shortened by some months for these people. The authors estimate that in 2004 around 1735 deaths (95%CI 1482–2032) in Dutch hospitals were potentially preventable. We should note that the terminology is slightly confusing here, in that the adverse events described here are not the death itself, but serious problems in patients’ care that led to harm which in turn hastened death. We should also remember that an adverse event near the end of life should not only be assessed by the extent to which death was hastened; contracting Clostrium difficile or sustaining a major adverse drug reaction in one’s final days may turn a potentially relatively peaceful passing into a nightmare of pain and suffering.

The power of the major adverse event studies is that they reveal the overall scale of harm to patients and also, to some extent, the nature and causes of harm. In the following sections we will examine two of the major types of harm, healthcare nosocomial infection and adverse drug events. We will then address a further important question of who is most vulnerable to harm.

Nosocomial infection, or healthcare associated infection (HCAI), is the commonest complication affecting hospitalized patients. In the Harvard Medical Practice Study, a single type of hospital acquired infection, surgical wound infection, was the second largest category of adverse events (Burke, 2003). Currently 5–10% of patients admitted to hospital in Britain and the United States acquire one of more infections; millions of people each year are affected. In a massive survey of over 75 000 patients in 2006, Smyth et al. found a prevalence rate of 7.59% in the United Kingdom (Smyth et al., 2008). In the United States, 90000 deaths a year are attributed to these infections, which add an estimated $5 billion to the costs of care. Intensive care units sustain even higher rates, approximately 30% of patients being affected, with an impact on both morbidity and mortality (Vincent, 2003).

Four types account for about 80% of nosocomial infections: urinary tract infections, often associated with catheter use; bloodstream infections, often due to intravascular devices, surgical site infection and pneumonia. Each of these four types may arise in more than one way and may be due to one or more different bacterial species. Intravenous lines are a particular potent source of infection, and the chance of infection is increased the longer the line remains in place. This is particularly disturbing as lines inserted into patients are often not being used. In one study, a third of patients in a general hospital setting had intravenous lines or catheters inserted; one-third of the lines were not in active use; 20% of the cannulas inserted were never used at all and overall 5% of the lines in use led to an unpleasant complication (Baker, Tweedale and Ellis, 2002). Not all infections are necessarily preventable by any means, with overcrowding and understaffing being important contributory factors (Clements et al., 2008). However, there is a consensus that many could be avoided by interventions such as the proper use of prophylactic antibiotics before surgery and hand hygiene campaigns amongst staff. Despite many studies, and massive campaigns, there is still widespread failure to adhere to basic standards of hand hygiene and it is hugely difficult to bring about change.

Infection control has for decades been seen as a public health problem and tackled by specialist doctors and infection control nurses, rather than linked with general quality improvement work. The emergence of the patient safety movement has energized and supported infection control, leading to those involved widening their remit to monitor antibiotic use as well as infection and to associate themselves with the broader drive to make healthcare safer (Burke, 2003). Patient safety in turn may be able to learn much from the techniques of infection control, particularly in the methods of surveillance, rapid response to problems and epidemiological analyses. Infection control requires, amongst other things, careful specification of the types of infection coupled with both a rapid response to outbreaks and systematic, routine surveillance and monitoring.

atient safety, in the form described in this book, has primarily developed in advanced, relatively well resourced healthcare systems. However, the safety of healthcare is of huge concern in poorer countries where infections are the leading cause of mortality. Deaths and morbidity from disease dominate, but the risks of infection from healthcare itself are terrifying. To get a sense of the scale of problems facing developing healthcare systems, we will look briefly at the question of injection safety, drawing on a comprehensive review by Yvan Hutin and colleagues (Hutin, Hauri and Armstrong, 2003). This review is one of a number of safety related programmes established by the World Health Organization, concerning such mattersas the safety of blood products, chemical safety, vaccine and immunization safety, drug safety and medical device safety.

During the 20th century, injection use has increased tremendously, and injections are now probably the commonest healthcare procedure. Many injections given to provide treatment in developing countries are in fact unnecessary, as oral drug treatment would be equally or more effective. The belief in the power of injections, as opposed to pills, is one reason for the continuation of this practice. The dangers come from the reuse of syringes without sterilization, with syringes often just being rinsed in water between injections. This should not be seen as simply due to poor training or low standards; in a poor country everything is reused, simply because there is no alternative. Although lack of knowledge and poor standards play a part, the danger is hugely compounded by the basic lack of resources and the need to reuse any item of equipment if at all possible.