In this study, we collected relevant data from Taiwan’s National Health Insurance Research Database (NHIRD), which was launched by the Taiwanese government in 1995. This database contains the comprehensive medical information of approximately 98.29% of the population of Taiwan (approximately 23 million insured individuals) [56]; these data include the patient’s sex, date of birth, employment status, inpatient and outpatient diagnoses, medical procedures, drug use, treatment duration, and medical costs [57]. We obtained the patients’ baseline information from the Longitudinal Health Insurance Database 2000, a subset of the NHIRD, containing the data of 1 million randomly selected patients.

All data generated or analyzed during this study are included in this published paper and its Multimedia Appendix files. More detailed data sets are not publicly available due to restrictions set by the Taiwan Ministry of Health and Welfare regarding the NHIRD in Taiwan. Researchers interested in accessing the data from the NHIRD must obtain approval from the Health and Welfare. For further information on data availability and access, please refer to the “Data Availability” section.

Only patients with complete available data were included in this study. Patients with missing, inconsistent, or unknown records of baseline information, such as sex and birth year, were excluded from this study. The cohort included patients who underwent hip fracture surgery between January 1, 2000, and December 31, 2012. Additional exclusion criteria were as follows: unreasonable age (0 years old); younger than 60 years at baseline; with the outcome diagnosis prior to the start of the study (to prevent reverse causation bias); history of pathological fracture, open fracture, or major traffic accident before the initiation of the study; whose follow-up start date was the same as the follow-up end date. At last, participants were excluded if their survival date was before July 01, 2003, to ensure that the study population had a minimum of 10 years of exposure to air pollution. This criterion was established because the Environmental Protection Administration (EPA) data for this study was only available starting from July 1, 1993. Therefore, if a participant’s follow-up ended before July 1, 2003, their exposure to air pollution would be less than 10 years. The ICD-9 (International Classification of Diseases, Ninth Revision) codes established in the inclusion and exclusion criteria were listed in Multimedia Appendix 1.

We obtained information on the cumulative daily average levels of the aforementioned 12 air pollutants from 76 monitoring stations maintained by the Taiwan EPA, Executive Yuan. Data were collected for the period between July 1, 1993, and December 31, 2013. The cumulative daily average level of each pollutant was calculated over the 10 years before the end of follow-up. Then, the data were integrated with the patients’ residential postal codes obtained from their insurance registration data. Any change in residence during the assessment period was also considered. Specifically, changes in a patient’s place of residence during the study period were tracked by their information with insurance registration records, ensuring that any shifts in location were accounted for in the study’s data analysis to maintain accuracy and reliability.

All-cause mortality was the major outcome of interest. We followed up on each patient until we reached the primary end point, which was determined by 2 conditions: death (withdrawal from the National Health Insurance program) or end of the study period (December 31, 2013). The survival duration was calculated in months. Several confounders were identified and adjusted for, such as age, urbanization level, insurance amount, CCI score, hip fracture procedure, comedications, antiosteoporosis medication, ambient temperature, season, and lag 0-1. Patient data such as age, insurance amount, CCI score, hip fracture procedure, comedications, and antiosteoporosis medication were also obtained from the NHIRD database. Specifically, the insurance amount was evaluated as an average value for the assessment period of air pollution exposure. Comorbidities were defined as those occurring before the survival date. ICD-9 codes or Anatomical Therapeutic Chemical Classification codes for the definition of hip fracture procedure, comedications, and antiosteoporosis medication were detailed in Multimedia Appendix 1. Information on the meteorological factor of ambient temperature was collected from the EPA. Data on the urbanization level were recorded according to the patients’ residence at the beginning of the follow-up period in accordance with the classification proposed by Liu et al [58]. Furthermore, the season was defined on the basis of the date. We also included the 2-day moving average of current- and previous-day levels of air pollutants before the primary end point (lag 0-1) as a confounder, considering the largest effect estimate reported in the literature [59-61].

To identify the patient characteristics associated with air pollution, the cohort was divided into 3 tertiles on the basis of the level of exposure to each pollutant. The tertiles were compared using the chi-square test or 1-way ANOVA. Post hoc tests were performed to estimate between-tertile differences if significance was indicated. We plotted crude cumulative incidence curves of mortality within 1 year after hip fracture surgery for the 3 tertiles; between-tertile differences were assessed through log-rank tests. Hazard ratios (HRs) for exposure at 1 SD increment for 10 years were calculated using Cox regression models to evaluate the dose-response effects between air pollutants and mortality risk. The regression models were adjusted for the aforementioned confounders. All tests were 2-sided; statistical significance was set at P<.05. All analyses were performed using the MetaTrial Research Platform (Biomedica Corp).

The Research Ethics Committee of Taipei Medical University, Taiwan, approved this study (TMU-JIRB N202203088). Because of the anonymity of patient data in the NHIRD, the requirement of informed consent was waived.

After the inclusion of patients with complete available data (n=882,391), those who underwent hip fracture surgery during the study period were identified (n=9286). Patients who met the exclusion criteria (1349, 75, and 473 for the aforementioned 3 criteria, respectively) were further excluded from the analysis. Finally, 7426 patients were included in the analysis. Figure 1 illustrates the selection process.

Table 1 summarizes the characteristics of the included cohort (n=7426). The incidence of mortality was 929 (12.51%). The average age of the cohort was 78.54 years, and 2926 (39.40%) patients were men. The characteristics of patients exposed to each pollutant were assessed by dividing the cohort into 3 tertiles (Multimedia Appendices 2-13). Table 2 presents the mean and distribution values of each pollutant over the 10-year exposure period.

Table 3 presents the incidence of mortality within 1 year after hip fracture surgery across the tertiles, as well as the corresponding P value (1-way analysis of variance and post hoc tests). The cumulative incidence curves revealed substantial differences in the direction and strength of the associations between each air pollutant and the incidence of mortality (Figures 2-13).

To evaluate the dose-response effects between the levels of air pollutants and the risk of mortality, we used Cox regression models and calculated the HRs for exposure at 1 SD increment for 10 years. Table 4 presents the corresponding HRs. The findings suggest that a 1 SD increment in the average exposure levels of SO₂, CO, PM₁₀, PM_2.5, NO_X, NO, NO₂, THC, NMHC, and CH₄ was associated with 14%, 49%, 18%, 12%, 41%, 33%, 38%, 20%, 9%, and 26% significant increases in mortality risk, respectively. However, a significantly negative association was observed for O₃, with a 35% reduction in mortality risk. CO₂ tended to exhibit only a trend but not a statistical significance to reduced mortality risk.

In this study, long-term exposure to SO₂, CO, PM₁₀, PM_2.5, NO_X, NO, NO₂, THC, NMHC, and CH₄ was associated with an increased risk of all-cause mortality within 1 year after hip fracture surgery. In older individuals, a 1 SD increment in the average exposure levels of SO₂, CO, PM₁₀, PM_2.5, NO_X, NO, NO₂, THC, NMHC, and CH₄ was associated with 14%, 49%, 18%, 12%, 41%, 33%, 38%, 20%, 9%, and 26% increases in mortality risk, respectively. By contrast, we found a significant reduction (35%) in the HR for O₃ and a trend of potentially protective effects of CO₂.

In this study, long-term exposure to SO₂, CO, PM₁₀, PM_2.5, NO_X, NO, NO₂, THC, NMHC, and CH₄ was consistently found to be risk factors for mortality within 1 year after hip fracture surgery. Previous studies have found an association between osteoporosis and the levels of exposure to air pollutants, including PM_2.5, PM₁₀, CO, and NO₂ [62-65]. The same trends were observed in the risk of hip fracture associated with PM_2.5, SO₂, NO, NO₂, and O₃ [66,67]. Shi et al [35] were the first to investigate the effects of air pollution on mortality after hip fracture. In their nationwide cohort study in the United Kingdom, the levels of specific air pollutants, such as PM_2.5, PM₁₀, and NO₂ exhibited a moderately positive association with the increases in the incidence and 30-day mortality rates of hip fracture. Thus far, their study was the only one on the negative effects of air pollution on mortality after hip fracture. However, the focus on 30-day mortality limited the scope of the long-term prognosis assessment. Moreover, the representativeness and extrapolation of their findings were limited because of the lack of age-standardization of the incidence; the possibility of errors in air pollution assessment owing to the use of data acquired only from the hospital region; and the presence of key unadjusted confounders, such as sex, season, temperature, socioeconomic, and CCI score. The aforementioned limitations of their study and the complexity of air pollution resulted in an unclear dose-response effect. To overcome this problem, we analyzed many pollutants; the robust evidence obtained in this study may help policy makers assess the economic loss and burden associated with air pollution and devise effective prevention strategies.

Although the precise mechanisms underlying the association between air pollution and mortality remain unclear, we hypothesized the following mechanism: air pollution contributes to frailty in older adults, thus increasing their susceptibility to mortality after hip fracture surgery. Air pollution is an independent risk factor for mortality from respiratory and cardiac diseases [25,68]. Furthermore, air pollution modifies the aging process by interfering with biological pathways; these effects are supported by toxicological evidence indicating that air pollution enhances oxidative stress [43-52], activates systemic inflammation [36-42], and causes metabolic disorders [69-72], genetic and epigenetic alterations [73-81], and vector-mediated pathogen transmission [30]. Zanobetti and Schwartz [82] suggested that air pollution influences the frail population in the following manners: increasing the mortality rate, increasing recruitment into the frail group, and delaying recovery. This finding indicates that air pollution modifies aging through frailty. Moreover, air pollution may aggravate age-related decline and functional deterioration at the cellular, tissue, and organ levels [83], thus increasing individuals’ vulnerability to disease development and mortality incidence. Given the association of exposure to ambient air pollution with all-cause and specific morbidity and mortality, our hypothesis and the obtained significant HRs appear to be highly conceivable and concomitant.

We found a negative association between the risk of mortality after hip fracture and exposure to O₃ pollution. Various conclusions have been reported regarding the effects of O₃ exposure [84-88]. Studies reporting protective effects, which are consistent with our findings, have provided the following explanations. First, sunlight facilitates O₃ formation; O₃ is a seasonal pollutant because its level is higher in summer than in winter. Therefore, seasonal effects should be considered when analyzing the effects of O₃ on health [84,89]. In this study, season served as a confounder in our regression models for potentially adjusting for the seasonal effects. Second, O₃ is highly reactive and leads to the formation of other pollutants, such as NO₂ and particulate matter [84,87]. Thus, O₃ appears to be associated with other air pollutants; hence, the effects of O₃ on health should be considered together with those of other pollutants. Thus, in this study, we performed a Pearson correlation analysis of air pollutants (Multimedia Appendix 1). To address the confounding impact of other pollutants and to prevent potential issues related to multicollinearity, we used potential associations among air pollutants to evaluate the impacts of multiple pollutants, selectively regulating those based on their weak correlations with other air pollutants. We defined the absolute value of correlation coefficients of <0.3 to indicate a low strength of correlation. However, the negative association between mortality and O₃ pollution persisted (Multimedia Appendix 1) in the multiple-pollutant models of the targeted pollutants; thus, the second explanation was rejected. Thus, the precise mechanisms underlying the protective effects of O₃ remain unclear. Nevertheless, other theories may provide some insights into these mechanisms, as follows: the enhanced expression levels of endothelial and inducible nitric oxide synthase [90-92], improvement of hemorheological parameters and oxygen delivery [92,93], and neuroprotective effects [92,94].

CO₂ emission exhibits a nonsignificant negative correlation with mortality after hip fracture. For the correlation between exposure to CO₂ pollution and the risk of mortality within 1 year after hip fracture surgery, the HR was 0.88 (95% CI 0.76-1.02), with a P value of .09. This finding is similar to previous findings that CO₂ emission exerted positive but nonsignificant effects on longevity [95-97]. After adjustment for the urbanization level and insurance amount, our analysis revealed a positive but nonsignificant trend in the health effects of CO₂ exposure, which is consistent with the literature. However, to the best of our knowledge, this has so far not been explained nor recognized in the literature. The need for more extensive research in the future is evident, as there remain numerous unexplored facets and complexities within this subject that warrant thorough investigation to advance our knowledge and drive progress in the field.

Our study has several strengths. First, to the best of our knowledge, this study is the first to indicate that long-term exposure to air pollution contributes to mortality within 1 year after hip fracture surgery. Second, we analyzed many key air pollutants (n=12). Third, we assessed the mortality within 1 year of surgery, thus focusing on the long-term outcomes of hip fracture. Finally, we adjusted for most of the major confounders, such as sex, season, temperature, socioeconomic, and CCI score, which further strengthened our findings and provided robustness for our primary conclusion.

This study has some limitations. First, although smoking is a key factor influencing ambient air quality, we could not include it in the analysis because of data unavailability; this might have led to a potential bias. Second, the NHIRD does not include data on disease severity, clinical manifestations, laboratory findings, and mortality causes. Hence, we could not determine the specific causes of mortality. Third, the possibility of errors in the assessment of air pollution cannot be ignored. Fourth, because we regarded death as withdrawal from the National Health Insurance program, bias might have been introduced into our findings. Fifth, while we considered any changes in residence during the assessment period, it is worth noting that the residential postcodes may not always be updated accurately in the insurance registration data. This could result in inaccuracies when reflecting patient characteristics. Sixth, although we adjusted for several confounders, unmeasured or unknown confounders might have introduced bias into our results. Given the limitations of retrospective health insurance database cohorts, which may not encompass all potential confounding factors, the possibility of residual bias due to unmeasured variables remains. Seventh, although we aimed to analyze individual molecules separately, it is possible that we neglected the potential for nonlinear relationships or the possibility of mixed effects of other molecules. Finally, there is a potential that the current levels of air pollution may differ from the data used in our study. However, due to the availability of data, this is the most up-to-date information we can analyze. Furthermore, our paper highlights the health and societal hazards of air pollution, and we hope that our research findings can provide empirical support for air pollution control policies across various times and situations. Despite the aforementioned limitations, this study provides adequate, high-quality evidence to support our conclusion that long-term exposure to SO₂, CO, PM₁₀, PM_2.5, NO_X, NO, NO₂, THC, NMHC, and CH₄ is associated with an increased risk of all-cause mortality within 1 year after hip fracture surgery.

In this nationwide population-based retrospective study, we discovered that long-term exposure to SO₂, CO, PM₁₀, PM_2.5, NO_X, NO, NO₂, THC, NMHC, and CH₄ is correlated with an increased risk of all-cause mortality within 1 year after hip fracture surgery in Taiwanese adults older than 60 years. Furthermore, 1 SD increment in the average exposure levels of SO₂, CO, PM₁₀, PM_2.5, NO_X, NO, NO₂, THC, NMHC, and CH₄ was associated with 14%, 49%, 18%, 12%, 41%, 33%, 38%, 20%, 9%, and 26% increases in mortality risk, respectively. However, O₃ exposure was associated with a significant reduction (35%) in HR, whereas CO₂ exhibited a nonsignificant trend of potentially protective effects. Thus, our findings provide robust evidence that can be used by the government to devise and implement air pollution prevention strategies for reducing the burden on the health care system. Nonetheless, future studies with well-adjusted confounders are warranted to investigate the correlation between air pollution and hip fracture mortality and the underlying pathophysiologic mechanisms.