Identifying and preventing complications for patients in hospital

Older people and those who are frail or have chronic diseases are at particular risk of multiple complications

The 2021 World Health Organization (WHO) global safety goals1 include eliminating preventable hospital‐acquired complications occasioned by modifiable causes of harm.2 Despite efforts to achieve this aim, hospital patients still experience complications with alarming frequency.1,3 In Australia, one in four patients hospitalised for at least one night experience complications, about half of which are potentially preventable.3

This issue of the MJA includes the retrospective, observational cohort analysis by Duke and colleagues of administrative data (2015–2018) on 1.56 million acute multiple‐day hospital episodes for adults in 38 public hospitals with intensive care units in South Australia and Victoria.4 The authors estimated the prevalence of hospital‐acquired complications for thirteen classes of the Australian list of high priority complications,5 as well as the relative contributions of patient and hospital factors to their occurrence.

Similar to the findings of a previous report,3 Duke and his colleagues found that complications had been recorded for almost one in ten admissions (9.72 per 100 separations; 95% CI, 9.67–9.77 per 100 separations), and that frailty and multimorbidity were more frequent among people who experienced complications than those who did not. Multiple complications were recorded for about 3% of admissions, with a mean of 1.4 complications for these separations. Complication event rates were highest for hospital‐acquired infections (3.64 per 100 separations; 95% CI, 3.61–3.67 per 100 separations), renal failure (2.13 per 100 separations; 95% CI, 2.11–2.15 per 100 separations), and delirium (1.80 per 100 separations; 95% CI, 1.78–1.82 per 100 separations).

Duke and colleagues used two multilevel model structures to assess the relative contributions of patient and hospital characteristics to the occurrence of hospital‐acquired complications in a care episode for each complication class. First, propensity scores for the likelihood of a patient experiencing a complication during their admission were estimated in two‐level probit models including selected on‐admission patient characteristics, with care episodes nested within hospitals. Second, after controlling for year, unexplained variance at the care episode, patient characteristic (clusters defined by propensity scores), hospital, and hospital type levels was estimated in four‐level probit regression models and used to calculate intra‐class correlation coefficients (ICCs). ICCs at the patient characteristic level were substantially larger than at the other two levels (hospital level; hospital peer group level), indicating that variation in complications was chiefly attributable to patient on‐admission characteristics. The authors concluded that hospital‐acquired complication rates are influenced more by patient‐related factors than by hospital characteristics.

These findings should be interpreted with caution. First, Duke and colleagues used propensity scores in a novel manner, clustering care episodes according to the likelihood of complications. We are unaware of any precedent for this approach. Propensity scores are typically used to balance treatment groups in terms of pre‐treatment characteristics.6,7 Hindsight bias was possible because complications and conditions such as frailty were retrospectively identified on the basis of information that may not have been available during the care episode; however, using on‐admission data may have reduced this bias. Finally, analysis of administrative datasets entails risks of underreporting and is subject to data quality limitations.8

Despite these limitations, we agree with the recommendations by Duke and his co‐authors. We need to validate the contribution of on‐admission patient factors to hospital‐acquired complications in order to identify patients most likely to benefit from targeted interventions. Consistent with other reports,9,10 the authors found that frailty and multimorbidity were more frequent among patients who experienced hospital‐acquired complications, leading them to suggest that these factors may be the antecedents of unavoidable complications. However, identifying these patients on admission can be challenging; for example, in Australia there is no generally accepted definition of frailty, nor a standard instrument for identifying it.11

Further, distinguishing between the importance of patient and hospital factors for hospital‐acquired complication rates is challenging in the complex, adaptive systems that characterise health care. Many hospital patients do not receive optimal preventive care,12,13 and harm prevention strategies for those with multiple risks are inadequate. Harm prevention imposes considerable cognitive, workload, and documentation burdens on clinicians.14 Uncertainty about best practice multi‐risk harm prevention stems from the fact that strategies for managing individual risks can be contradictory if patients have several comorbid conditions.9,15 Clinicians require integrated guidance about caring for patients at risk of several complication types.9,11,15

The final recommendation by Duke and his colleagues concerns the need for validated clinical indicators that are meaningful to clinicians and based on quality data. Indicators must be useful for detecting care below the level required for accurately identifying preventable complications. Reliable multi‐risk prediction models, feasible interventions of proven effectiveness for patients with multiple risks, and consensus regarding appropriate outcome measures each require further investigation.

The study by Duke and colleagues adds to the body of knowledge about hospital‐acquired complications, and provides an Australian perspective of a ubiquitous problem. Their findings indicate that we need to re‐think harm prevention for the patients at greatest risk. Models of comprehensive preventive care and composite tools for easing the burden on clinicians by integrating and prioritising harm prevention activities are urgently needed to replace the plethora of risk‐specific tools currently used in hospitals.