Conclusions
This is the first study to describe systematically the epidemiology of adverse events causing or contributing to the death of RRT patients. In this national, population-based cohort, direct complications of RRT were responsible for 2.1% of deaths, and in 3.5% of deaths, areas of concern were identified that may have or did lead to the death of a patient. The areas of concern largely fell into five main categories: recognition and management of hyperkalaemia, safe prescribing, out of hours care, the prevention and management of infection and haemodialysis vascular access. These factors were both organizational and human in nature, and no deaths were considered to be directly attributable to the failure or misuse of RRT-related medical devices. These findings suggest that errors are not a common cause of avoidable mortality in the RRT population and that efforts to improve the safety of RRT should focus not just on the technical aspects of RRT, but also consider the many other potential sources of harm in this complex group of patients.
Recent years have seen an increasing focus on patient safety across all areas of health care, partly in response to the publication of landmark reports on this issue in both the USA and the UK. Among general hospital inpatients, the prevalence of preventable death has been estimated to be in the range of 4.1–6.0%. Although the definition differs somewhat, the most analogous figure estimated in our study (inpatient deaths where concerns were identified that may have or did contribute to death) was 4.0%, which is in line with these previous estimates. It is difficult to know if the slightly lower estimate in our study is a real difference, or just reflects differences in methodology and definitions. As far as we are aware, there are no previous published estimates of preventable death in the RRT population, and studies concerning patient safety in this population have largely focused on medical device-related harm, novel adverse drug reactions [5, 16], dialysis water purification and blood-borne infections. Strategies to improve patient safety in dialysis units have emphasized the importance of effective communication, falls prevention, reducing medication error, correct dialysis equipment preparation and infection control. In our cohort, no cases of harm due to failure or misuse of RRT-related equipment were documented. This study demonstrates that the five main factors identified in the review of the most serious adverse events largely concern aspects of the everyday care of RRT patients and not the more technical aspects of dialysis or kidney transplantation. This is consistent with studies of adverse safety events in the pre-dialysis chronic kidney disease population, where the commonest causes of harm are from infection, falls, electrolyte/metabolic disturbance and medication error. Patients with CKD are at increased risk of a whole range of adverse events compared with the general hospital population, not just those directly related to CKD itself. Strategies to prevent harm to RRT patients therefore involve all health professionals who care for RRT patients, not just nephrology specialists and those working in dialysis units.
The proportion of deaths attributable directly to RRT complications estimated in this study is small, although the size of the estimate is obviously dependent on what is classified as a RRT complication. For example, dialysis access-related septicaemia was not included in this category, and was instead included in the infection group. We were unable to specifically classify instances of septicaemia as either access related or non-access related, but analyses from the same cohort demonstrate significantly (adjusted OR 6.9) higher risk of death from septicaemia in patients using tunnelled central venous catheters. Hyperkalaemia and other electrolyte disturbances have been implicated as a contributing cause of the high rate of sudden cardiac death in the RRT population (9.2% in our cohort), but were only classified as such here where there was evidence to support death from this cause. Within the CKD population, episodes of hyperkalaemia are associated with markedly increased risk of death (OR 8.0 for CKD Stage 5 patients with severe hyperkalaemia), and patients with Stage 5 CKD are at highest risk of hyperkalaemia. The true contribution of hyperkalaemia, and other electrolyte disturbances, to RRT-related mortality, therefore, may be higher than the estimate in this study. Our study also demonstrates that hyperkalaemia-related mortality may be caused not just by its occurrence, but also by errors in how it is managed and treated. These errors were both organizational and human factors related, and imply that efforts to prevent avoidable hyperkalaemia-related mortality are warranted. In particular, clear protocols concerning how, where and by whom patients at high risk of hyperkalaemia should be managed might avoid otherwise preventable deaths. In addition this study suggests that failure to recognise the signs and symptoms of hyperkalaemia is contributing to avoidable mortality. Professional guidelines on the management of hyperkalaemia have recently been published in the UK for the first time, and this may provide a useful opportunity to improve the awareness, understanding and management of this condition.
The only previous published report of safety lapses in a dialysis cohort was a retrospective cohort study of 41 dialysis patients admitted to a surgical service. Multiple lapses of care were identified in 78% of the cohort, relating to provision of appropriate meals, dose adjustments during prescribing and the administration of IV fluids. As with this study, inappropriate dosing of strong opioids was identified as a common error. Medication errors have previously been identified as one of the most common safety events in dialysis units, particularly during transition between care settings or clinical teams. Medication errors have also been found to be common in patients with other stages of CKD, largely as a consequence of failing to adjust correctly for reduced drug clearance and the prescription of nephrotoxic medications. More broadly, studies of patient safety lapses in other patient populations have identified medication errors as a frequent source of harm, with an incidence of adverse events due to medication errors of 2.3% in hospitalized patients. A number of approaches to reducing medication-related harms have been developed and our study suggests that a particular focus on the safe and effective use of opioids in patients with ERF is warranted.
Health-care-associated infection is one of the most significant sources of iatrogenic harm, and, in our study, was identified as contributing to the deaths of 9.6% of patients. However, only a minority of these deaths had been identified as being of high concern, which suggests that episodes of health-care-associated infection were in many instances not felt to have been avoidable or the result of medical error. Infection in the RRT population is a complex problem: it is both common (the prevalence of septicaemia in dialysis patients is over 100 times that of the general population ) and multifactorial, associated with high rates of hospitalization, dialysis-specific infection risks and immunocompromise as a consequence of renal impairment, comorbidity and immunosuppressive therapy. In addition, experience suggests that infection in RRT patients, although the direct cause of death in only a minority, frequently contributes to a multifactorial decline in health status in the period before death. Identifying what proportion of infections in this population is avoidable is, therefore, difficult. However, interventions in other complex, high-risk populations have led to large reductions in health-care-associated infection and suggest that there may be a significant opportunity to reduce infection-related deaths in the RRT population through strategies such as care bundles and minimizing the use of central venous catheters.
This was a mortality-based study; we, therefore, have no data concerning patient safety incidents in patients who did not die, nor those which occurred in earlier time points in the treatment history of patients who then went on to die during the study period. Whilst there are likely to be common themes, the nature and prevalence of non-fatal medical errors may be different to those that result in death. Data were drawn from a single country and it is possible that they do not translate to other health settings or populations. Some of the data included in this study are based on the opinions of individuals. There was evidence of inter-observer variation, as evidenced by differences between assessors in the proportion of patients where major concerns were identified, and thus the accuracy of the estimate of the proportion of deaths on RRT that are of major concern is difficult to quantify. Kappa statistics were not estimated because of the difficulties of achieving equal information between observers: in many instances, the assessments concerning factors contributing to death were made using information from sources other than the paper case record, such as electronic care records, team meetings and personal communication with members of the care team. Previous studies using retrospective case record review have also found inconsistency between assessors even when using validated tools , and the development of review methodologies that allow objective assessments to be made of the types of complex information available outside the written case record would improve the accuracy of future studies. Future studies might also usefully focus on quantifying and classifying the non-fatal adverse events in the RRT population, since research on this population is lacking.