In-hospital mortality as an outcome indicator for air pollution health risk assessment

Numerous studies have delved into the adverse impact of air pollution on health , such as increased hospital visit risks of cardiovascular diseases and respiratory diseases, and identified the association between short-and long-term exposure to ambient air pollution and increased risks of developing and dying from diseases, such asincluding cardiovascular diseases, respiratory diseases and so on (1)(2)(3). The global exposure to this risk is rapidly intensifying. The Global Burden of Disease Study 2019 identified ambient particulate matter pollution as one of the risk factors with the largest increases in exposure from 2010 to 2019, and ambient air pollution led to approximately 6.67 million deaths globally in 2019 (4). We have noticed that recently an increasing number of studies have focused on a health outcome-in-hospital mortality. Therefore, to ground our discussion, we reviewed articles published in English over the last 10 years that focus on this specific outcome (Table 1) (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18). This viewpoint aims to highlight the significance of in-hospital mortality as an outcome indicator and the potential of utilizing hospitalization records to advance our understanding of the air pollution-health linkage.In-hospital mortality directly reflects the short-term and severe health deterioration caused by air pollution exposure among patients, serving as a critical health outcome indicator. Although in-hospital mortality account for only a subset of total mortality, evidence indicates consistent trends in pollutant effects between in-and out-of-hospital mortality. For instance, a Chinese study found that despite the number of out-of-hospital ischemic heart disease (IHD) deaths exceeded in-hospital deaths several times, fine particulate matter (PM2.5) was significantly associated with an increased risk of both in-and out-of-hospital IHD mortality (19). A study analyzing daily mortality data from 1989 to 2000 in U.S. found that elevated particulate matter concentrations were significantly associated with increased risks of both in-and out-of-hospital all-cause mortality (20). Similarly, an Italian multi-city study revealed that PM10 increases significantly linked to higher all-cause mortality risks, with no significant difference in effects on in-and out-of-hospital mortality (between-group comparison P=0.817) (21). These findings collectively support the validity of in-hospital mortality as an outcome indicator for assessing the mortality risks due to ambient air pollution.Nevertheless, from the perspective of the complexity of causal association analysis, in-hospital mortality has unique research value compared to out-of-hospital mortality. First, deaths occurring in hospitals benefit from more comprehensive clinical evaluations that minimize misclassification of death causes (22). In contrast, out-of-hospital deaths are prone to classification errors in cause of death and frequently require autopsies to verify causes (23,24). Second, determining causality in out-ofhospital deaths is complex due to influencing variables including the quality of pre-hospital emergency care (25) and latent disease progression (26). These critical confounders are frequently undocumented in mortality records, leading to underreporting of true contributing factors (27). However, the information on important risk factors for death (such as the baseline health status and treatment) of hospitalized patients are more comprehensive, which helps reduce confounding factors during the analysis and form more rigorous causal inferences. Finally, compared with out-of-hospital deaths, (even down to the minute), which supports the construction of hourly exposure lag models and is beneficial for identifying the critical exposure period and strengthening causal inference.Most studies in Table 1 utilize existing hospitalization records to evaluate the relationship between air pollution exposure and in-hospital mortality risk. This approach represents a valuable form of medical data mining, transforming routine clinical data into valuable research findings. Using existing hospitalization records to analyze the impact of air pollution on in-hospital mortality offers the following advantages: First, standardized hospitalization records ensure data consistency. Taking China's hospitalization record system as an example, the front page of medical records adopts structured data formats, offering information on patient demographics, comorbidities and complication diagnoses, in-hospital mortality outcomes, surgeries, invasive treatments, and so on. For nationwide hospitalization registration information, standardized medical records and quality control procedures related to medical insurance provide highly consistent and well quality-controlled data for conducting large-scale, multi-center investigations.Second, hospitalization records provide more critical clinical information than mortality registration systems. Geographic information in hospitalization records (e.g., patient residential addresses) can be used to correlate high-resolution air pollution monitoring data for more precise exposure assessment. These extensive clinical data enable researchers to: 1) explore susceptible subgroups through stratified analysis (e.g., patients of different age groups or with varying comorbidities); 2) examine the influence of socioeconomic factors (marital status, occupation, etc.).Most importantly, these studies can be conducted without extra recruitment or sample collection costs. By analyzing existing clinical data, studies can provide substantial evidence for public health decision, such as identifying the benefits of air pollution control and vulnerable populations requiring more protection.Despite the aforementioned advantages, there are some issues that need further attention when utilizing hospitalization data for air pollution related in-hospital mortality research.First is the potential exposure bias. Current studies usually use pre-admission exposure (e.g., one week before admission) as individual exposure, which did not assess in-hospital exposure effects. This may introduce exposure bias, particularly for patients with long hospital stays. We recommend routinely restricting in-hospital death windows (e.g., within 30 days) as sensitivity analysis to check result robustness. Future studies should take indoor air pollution during hospitalization by indirectly estimating indoor air pollution based on indoor-outdoor correlation coefficients from prior studies (28,29), or directly measuring indoor pollution levels. Relying on monitoring station data to assess exposure may inaccurately reflect individual exposure due to spatial heterogeneity. We suggest integrating We recommend assessing exposure by using high-spatial-resolution pollution measurement methods (e.g., combined satellite retrievals and ground-based measurements (30)) based on patients' address information to improve exposure assessment precision.Second, selection bias must be considered. Rapid-onset diseases may cause death before hospital admission. In some regions, cultural preferences may lead to a higher likelihood of home deaths near the end of life (31). These factors can result in an underestimation of in-hospital mortality burden associated with air pollution, as demonstrated by discrepancies between in-hospital mortality data and population-level mortality data (32)(33)(34). We suggest combining hospitalization records with death registry data to analyze both in-hospital and peri-hospitalization deaths. This would help assess the impact of unrecorded peri-hospitalization deaths and provide valuable insights for future research. Additionally, cross-regional studies can help minimize the influence of cultural differences on the results.Finally, confounding control and generalizability need improvement. As shown in Table 1, most existing studies use logistic regression models at the individual level to estimate air pollution related in-hospital mortality. However, lifestyle factors (e.g., smoking status) and genetic backgrounds are often unavailable in hospitalization records, potentially introducing unmeasured confounding in logistic regression. We recommend adopting case-crossover designs to control unmeasured timeinvariant confounders through self-matching (35,36). And the results from case-crossover studies approximate relative risks (35) and thus offer greater generalizability than odds ratios from logistic regression. Furthermore, multicenter studies can further enhance the generalizability of the findings by reducing regional heterogeneity.Investigating in-hospital mortality as an outcome indicator for air pollution health effects complements current air pollution health studies. Utilizing hospitalization records for such research is a feasible and economical method. We recommend maximizing the utility of available medical records to conduct multicenter, high-quality studies, offering scientific evidence for public health policy against pollution-related health risks. Same as above PM2.5 and its componentsIn-hospital mortality in stroke patientsLong-term exposure to PM2.5 and specific PM2.5 components were associated with higher risk of in-hospital mortality in stroke patients.Lin, 2024Same as above PM1, PM2,5, PM10In-hospital mortality among AMI patientsSame as above Short-and long-term exposure to PM1, PM2.5 and PM10 were significantly associated with higher risk of inhospital mortality in AMI patients.Lin, 2024Same as above PM2.5 and its componentsIn-hospital mortality in AMI patientsShort-and long-term exposure to PM2.5 and its components were significantly associated with higher risk of inhospital mortality in AMI patients.Lai, 2024 (18) Shanxi, China PM2. 5

Air pollutants · Air pollution · Burden of disease · Cardiovascular health · Disease · Disease burden · Particulates · Risk assessment · Air Quality and Health Impacts · Air Quality Monitoring and Forecasting · Occupational and environmental lung diseases · Pollution

• The Case-Crossover Design

  Malcolm Maclure•American Journal of Epidemiology•1991

  Incluir item

• Global burden of 87 risk factors in 204 countries and territories, 1990–2019

  Open Access•Christopher J L Murray, Aleksandr Y Aravkin et al.•The Lancet•2020

  Incluir item

• Ambient fine particulate matter chemical composition associated with in-hospital case fatality, hospital expenses, and length of hospital stay among patients with heart failure in China

  Open Access•Zhihan Lai, Jingyi Zhang et al.•Journal of Global Health•2024

  Incluir item

• Burden of cardiovascular disease attributed to air pollution

  Open Access•Amir Hossein Khoshakhlagh, Mahdiyeh Mohammadzadeh et al.•Globalization and Health•2024

  Incluir item

• Subnotificação de óbitos por Aids no Brasil

  Open Access•Rondinelle Alves do Carmo, Gabriela Moreira Policena et al.•Ciência & Saúde Coletiva•2021

  Incluir item

• Out-of-Hospital Covid-19 Deaths

  Elizabeth B Pathak, Rebecca B Garcia et al.•American Journal of Public Health•2021

  Incluir item