Our research showed that the SMR had significantly decreased over the 12 years in the enrolled ICUs. There was a strong link between a higher admission patient volume and a lower SMR. Patients in both low and high-admission groups showed decreasing SMR trends.
In the present study, we calculated the SMR based on the expected mortality rate from the APACHE II score. APACHE II is a widely applied scoring system in critical care medicine. It has demonstrated a strong accuracy in predicting patient prognosis (20, 21). Our study revealed a median SMR of 0.324 and a mean of 0.528. In previous research, SMRs calculated using the APACHE IV model reached values from 0.4 to 0.89 (16, 22), consistently below 1. However, this does not negate the predictive value of the score for prognosis. Many studies used statistical methods to standardize them to avoid overestimating or underestimating expected mortality rates (23–27). Other studies employ composite scores or new predictive models to calculate expected mortality rates (11–13, 28), increasing the SMR to 1. A study in 21 ICUs from Finland, Estonia, and Switzerland also reported a declining trend in SMR from 2008 to 2017 (P < 0.001) (12). This is consistent with the results of our study, indicating a continuous improvement in the overall quality of critical care medicine.
In our study, 3h SSC bundles compliance, 6h SSC bundles compliance, and the proportion of DVT prophylaxis were negatively associated with SMR, suggesting a better value in three indicators corresponding to a better prognosis. Studies on quality control indicators from China concluded that process factors need more attention than outcome factors (29). Unplanned transfer to ICU and ICU re-admission rate within 48h (24h) also negatively correlated with SMR. Some studies suggested that ICU re-admission rate was associated with poorer clinical outcomes (30). Maharaj et al. conducted a retrospective study of 682,975 patients, arguing that ICU re-admission rate as a performance indicator might not reflect ICU healthcare quality, which supported our result here. This might be because only patients who survived after the initial admission had the possibility for re-admission to the ICU, leading to a selection bias that could mislead ICU performance measurement (31). We also found that the CRBSI and CAUTI incidence rates were positively correlated with SMR. Although the VAP incidence rate declined, it had no significant correlation with SMR (P = 0.528). Therefore, the management of VAP will remain a critical quality control focus. Our VAP incidence rate had a median of 0 and an average of 6.29 per 1,000 ventilator days. The incidence rate of VAP in this study was lower than those reported in many studies (8 to 18 per 1000 ventilator days) (32, 33). However, compared with another study on the VAP incidence rate in China in 2019 (5.58 per 1000 ventilator days) (34), there was still room for improvement. The proportion of ICU in total inpatient bed occupancy positively correlated with SMR. Some research results suggested that an increased bed capacity without a proportional increase in medical staff might impact healthcare quality, potentially lowering patient outcomes (33).
ICUs were stratified into two groups based on patient admission volume, with the high-admission group demonstrating a more favorable prognosis (P < 0.05). ICUs experiencing a greater influx of patients are typically either tertiary hospitals, which boast more beds and a broader patient base, or specialty ICUs, characterized by relatively stable patient sources and shorter average lengths of stay. Such ICUs are likely to have staff possessing superior diagnostic and treatment capabilities, or they may cater to patients presenting with less severe illnesses, both factors contributing to improved clinical outcomes as evidenced by lower SMRs. Research has indicated that in higher-level medical facilities or scenarios where the ICU's share of total inpatient bed occupancy is sufficiently large, an increase in ICU patient volume is correlated with reduced mortality rates (3, 4, 7).
In the ICU in the tertiary hospitals, there was no significant difference in the decreasing trend of SMR between the low and high-admission groups (P = 0.276). This suggested that the improvement in the level of medical care in tertiary hospitals was less influenced by patient admission volume. This was in line with the positioning of tertiary hospitals as comprehensive referral centers providing specialized healthcare services. In contrast, for the ICU in secondary hospitals, both the low and high-admission groups exhibited a decreasing trend with an inconsistent tendency (P = 0.048), mainly due to a mild SMR decrease in the low-admission group in secondary hospitals from 2016 to 2022. To explain this, we analyzed various indicators in secondary hospitals during 2016–2022 (supplementary file 5). We found a significant increase in the VAP incidence rate in the low admission group while a significant decrease in the VAP incidence rate in the high admission group. This suggested that inadequate infection control for VAP in the low admission group was one of the significant factors influencing SMR. Our study also revealed a significant upward trend in three process indicators for the high admission group, namely 3h SSC bundles compliance, 6h SSC bundles compliance, and microbiology detection before antibiotics. In contrast, there were no improvements in these three indicators in the low-admission group. This highlighted the requirement for SSC bundle adherence and microbiological testing before antibiotics as the critical points for future quality control efforts in the low-admission group.
The decreasing trends of the two groups in the general ICUs were not parallel (P = 0.018). The SMR for the high admission group significantly decreased from 2017 to 2020, likely due to adoption of the sepsis 3.0 guidelines, updated ARDS guidelines, and the National Quality Control Center to enhance quality control indicators and the continuous development and standardization of critical care medicine. From 2020 to 2022, both groups showed a mild upward trend in SMR, although without statistically significant differences. This might be associated with the comprehensive impact of the COVID-19 pandemic in 2020, leading to a surge in critically ill pneumonia patients with a high mortality rate or SMR (35). The decreasing trend in SMR between the two groups in specialty ICUs showed no significant difference (P = 0.511), indicating that the specialty ICUs had relatively stable patient sources and were less affected by the COVID-19 pandemic.
Our study had certain limitations. Firstly, data on patient demographics and treatments were not collected, which potentially limited its comparisons with other studies and the generalizability of the study results. Secondly, data collection and reporting bias were commonly seen in a retrospective study. Future prospective studies are required to validate our study results.