Trends of diagnostic adverse events in hospital deaths: longitudinal analyses of four retrospective record review studies

Introduction

Diagnostic safety is an important topic within the domain of patient safety. Although different methods result in different estimates, it is estimated that around 10–15 % of diagnoses result in error [1]. Furthermore, diagnostic errors are found to have a higher level of patient harm than other types of medical errors [2]. Despite this, research into diagnostic error and diagnostic safety has only gained increased attention over the last decade and diagnostic safety focused improvement measures have not yet been widely implemented.

Moreover, it remains unclear whether incidences, causes and consequences of diagnostic error have changed over time. A few studies to address patient safety trends in general have been conducted [3], 4] showing conflicting adverse event trends. These studies did also not specifically investigate diagnostic safety trends. Furthermore, most studies that report incidences use a variety of methods (e.g. record review, autopsy reports), in a variety of countries and populations, making these incidences incomparable over time.

In the Netherlands, adverse events (AEs) have been monitored every four years since 2004 using representative random samples of patient records [5], [6], [7], [8], [9]. An AE was defined as ‘an unintended outcome which resulted in temporary or permanent disability, death or prolonged care, caused by healthcare management rather than the disease’. The studies investigated the incidence, causes, and consequences of AEs in patients who either were discharged from the hospital or died during a hospital stay. AEs were categorised into various clinical processes, including the diagnostic process. These diagnostic adverse events (DAEs) were defined as ‘AEs that are predominantly related to the diagnostic process’. Importantly, each of these studies had the same methodology, except for the first study in 2004. This availability of data on both DAEs as other AE types across multiple years, creates the opportunity to investigate and compare longitudinal trends.

For the current study, we combined data of deceased patients from four of these Dutch AE studies to create a longitudinal dataset with data from 2008 to 2019. The research aim was to investigate how DAE incidence rates and causes changed over time and how they compare to other AE types.

Materials and methods

Data from four national Dutch patient record review studies (2008, 2011/2012, 2015/2016 and 2019; hereafter referred to as 2008, 2012, 2016, 2019) were combined to create a longitudinal dataset. Data from the first AE study in 2004 were available but not included due to substantial design differences. The datasets from 2016 to 2019 comprised only deceased patients because it was found that AEs were more common in this group, and therefore resulted in a higher power and more accurate incidence estimate. For these reasons, we decided to exclude the discharged patients from the 2008 and 2012 studies for analyses in the current study. This is in accordance with a previous longitudinal study using data on acutely admitted older adults from the same AE studies [10].

All initial AE studies were approved by the Medical Ethical Review Committee of the Amsterdam University Medical Center, the Netherlands. The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013).

Results

Trends of DAEs and compared to other AE types

Longitudinal analyses indicate that the incidence of DAEs decreased between 2008 (2.3 %) and 2012 (1.2 %, OR=0.52, 95 % CI: 0.32 to 0.83) and remained stable since (2016: 1.0 %, OR=0.87, 95 % CI: 0.51 to 1.53; 2019: 0.8 %, OR=0.81, 95 % CI: 0.46 to 1.43). The same pattern was found for DAEs with potentially preventable harm and potentially preventable death, see Supplementary Material, Appendix 2 and left side of Figure 1.

Figure 1:

Incidences of AE, AE with preventable harm, and AE with preventable death, separately for DAEs and other AE types, including indications of significant longitudinal differences for DAEs and indications of significant trend differences between DAE and other AE types.

Furthermore, the trend of DAE incidence between 2016 and 2019 was relatively stable which differed significantly with the increasing trend of other AE types during the same period (rate ratio=1.88, 95 % CI: 1.12 to 2.13, p=0.017; see Figure 1). This difference in trends was not found in other years and was not found for the DAE and AE subtypes with potentially preventable harm and potentially preventable death. In addition, DAEs were more often judged preventable than other types of AE (81.9 vs. 30.1 %; p<0.001) and, within preventable AEs, DAEs were more often the cause of preventable death (80.5 vs. 68.3 %, p=0.016).

Causes of DAEs and other AE types

Across all years, the mean number of cause types per AE was higher for DAEs than other AE types (1.64 vs. 1.28, p<0.001). Figure 2 shows DAEs were more often judged to be caused by human factors (83.1 vs. 28.7 %, p<0.001), organizational factors (29.4 vs. 8.5 %, p<0.001), and (protocol) violations (11.8 vs. 6.4 %, p=0.033), compared to other types of AE. Patient-related factors were less often implicated (32.4 vs. 53.3 %, p<0.001; see Figure 2).

Figure 2:

Causes of DAEs and other AE types as percentages within their own group.

Results of analyses of DAE causes over time show only a significant change in the percentage of DAEs caused by patient-related factors between 2016 and 2019 (29.5 and 58.6 % respectively, OR=3.40, 95 % CI: 1.20 to 9.66; see Figure 3).

Figure 3:

Causes of DAEs over time.

Discussion

This study analysed longitudinal data from four large Dutch adverse event (AE) studies with 19–20 hospitals each, focusing on diagnostic adverse event (DAE) incidence. Results show an initial decline between 2008 and 2012 after which the rate stabilized. Similar patterns were found for the subtypes potentially preventable DAEs and potentially preventable DAE-related deaths. Trends of DAEs and other AE types over time were similar except for the period between 2016 and 2019 were DAE incidence remained stable while incidence of other AE types increased. DAEs were more often preventable and had different causes compared to other AE types.

Several factors may have contributed to the observed trends in DAE incidence. The initial decrease coincided with activities of the National Dutch Patient Safety Programme (2008–2012), in which all Dutch hospitals participated [17]. The safety program was an initiative of various healthcare organisations (e.g. sector organisations for hospitals, medical specialists, nurses, and patients) and funded by the government [17], 18]. It comprised guidelines, education, and training on 10 themes which were chosen to reduce preventable harm in hospitals. Although the safety program did not include themes or implementations that were specifically focused on improving the diagnostic process, some had overlap with diagnostic safety. Most notably, three themes were focused on early recognition and monitoring of patients in certain conditions (i.e. sepsis, risk at vital organ failure, and frailty in older adults). Early identification of such patients could help physicians determine the correct diagnosis timely. Monitoring of these patients can help with recognizing deterioration of the patient’s condition, which is considered to be an aspect of diagnostic safety as well. Therefore, the safety program could have had impact on decreasing DAEs. Moreover, the improvements might have improved safety culture in general, extending its reach beyond the initial safety themes.

Furthermore, the current study only investigated AEs that were judged to have the diagnostic process as the primary contributing clinical process and these were compared to AEs with all other clinical processes (e.g. surgery, medication, or other clinical management). It is important to note that is possible that DAEs were related to other clinical processes as well, and vice versa, allowing DAEs to be improved by improving other aspects of health care.

In addition, we cannot rule out that there are other factors why the 2008 incidences of DAEs were higher than those of subsequent years, especially when considering the AE study in 2004. While this study had a slightly different methodology, it is reported that there was a significant increase between 2004 and 2008 for AE incidences [19].

DAEs were highly preventable and showed a different and more diverse cause profile, including more human and organisational cause types, and (protocol) violations compared with other AE types, and relatively few patient-related factors. The high preventability and the different cause profile might be related, as it is likely that more causes, especially human causes, can contribute to the preventability of the AE. Notably, the proportion of patient-related causes showed a sharp rise during the last period, perhaps related to growing patient complexity (e.g. more atypical presentations and comorbidities) [20]. Interestingly, this increase in patient-related factors has not affected the incidence of DAEs overall nor the subtypes of potentially preventable DAEs and potentially preventable DAE-related deaths. It is difficult to determine why this was observed, as it likely reflects a complex interplay of factors that cannot be fully explored with the data of this study.

The ‘(protocol) violations’ are defined as intended actions that deviate from the hospital’s protocols or guidelines. However, it is important to note that these actions generally not malicious, but arise through efficiency-thoroughness trade-offs (i.e. the choice between working thoroughly to help one patient vs. working more efficiently to help multiple patients) and could be linked to the Safety-II framework: While the Safety-I approach argues that healthcare professionals should adhere to protocols and guidelines, Safety-II emphasizes adaptivity and resilience [21]. This perspective allows healthcare professionals to deviate from existing protocols and guidelines if the situation calls for it.

Based on the current data, it is difficult to generate specific improvement strategies because the causes were studied using the main categories of the Eindhoven Classification Model which are quite broad, leaving the possibility for high heterogeneity of causes within each main category. Nevertheless, the current study has shown that the incidences of DAEs, potentially preventable DAEs and potentially preventable DAE-related deaths among patients who died in hospitals in the Netherlands have not further declined after 2012. The high preventability of DAEs, including preventable deaths, suggests there is room for improvement. Since the start of these AE studies in the Netherlands, no national programs have focused on improving diagnostic safety specifically. Results of the current study suggest that DAEs can react to improvements to general patient safety, as shown by the decrease between 2008 and 2012. However, DAEs could possibly benefit more from targeted safety interventions, especially since results suggest that DAEs have more diverse cause types compared to other AE types. Specific directions for such interventions as well as future research to improve diagnostic safety and reduce patient harm have been prioritized by both researchers and patients in two recent research agenda papers [22], 23].

Conclusions

The incidence of DAEs, potentially preventable DAEs, and potentially preventable DAE-related deaths in Dutch hospitals has decreased in 2012 compared to 2008 after the National Dutch Patient Safety Programme, but remained stable thereafter. While DAEs and other AE types were found to largely have the same incidence patterns, the incidence for other AE types increased relative to that of DAEs. DAEs remained highly preventable over the years and showed a different cause profile from other AE types with relatively more human and organisational causes, (protocol) violations and relatively few patient-related causes. These results suggest that DAEs are responsive to general patient safety improvements, but could benefit even more from targeted safety interventions.