Strengths and limitations
We observed a significant association between PM2.5 and the first hospital admission related to diseases of the kidney and the urinary system at low concentrations of air pollution, suggesting that the health benefit per unit decrease in the concentration of PM2.5 might be consistent across concentrations below the current NAAQS. This potential saturation effect (lower air pollution risk at high ranges of annual air pollution) can be explained by smaller risk changes in annual mean air pollution concentration in regions with higher baseline air pollution levels.41 Furthermore, from the non-linear concentration-response curves, we found no evidence of a threshold value (a concentration below which air pollution exposure does not affect kidney outcomes) for harmful pollution.
The large sample from the Medicare part A FFS cohort enabled us to estimate air pollution risks relating to kidney diseases among subpopulations.24 However, the results should be carefully interpreted because of small sample sizes for some subpopulations, such as racial minority groups and people eligible for Medicaid, compared with others. Further research is needed that investigates the association between air pollution and kidney disease in potentially vulnerable subpopulations, including potential differences in diagnosis and access to healthcare, especially given the high number of people with undiagnosed disease. The estimated associations between air pollution and kidney outcomes were generally higher in white participants and those not eligible for Medicaid. This finding conflicts with results from previous US studies reporting higher air pollution mortality risk in non-white populations and those eligible for Medicaid,24 43 although it is consistent with some other study results. A few plausible explanations exist; one possibility is competing risks. A previous study in Taiwan reported that people with comorbidities showed a lower risk related to PM2.5 and CKD development.20 Another study in Korea reported that patients with CKD and healthy lifestyles (normal weight, non-smokers, and non-drinkers) showed a higher air pollution mortality risk than patients with CKD and non-healthy lifestyles.44 These findings suggest the possibility of competing risk between air pollution and biological or behavioral factors that affects the incidence of kidney disease; the results by race and Medicaid eligibility could be affected by the potential competing risks of underlying health conditions and other risk factors. Another possibility is that the longer life expectancy of high income and white populations could affect the results of this study. The life expectancy of those with high income is longer than that of low income populations,45 and white populations also have a longer life expectancy than black populations.46 Because the older population is more vulnerable to the incidence of kidney disease,11 47 the relatively longer life expectancy of Medicaid non-eligible and white populations could be related to the higher air pollution related risks on kidney outcomes in this study. Finally, because low income and racial or ethnic minority populations generally have less access to medical facilities,48 underdiagnosis could potentially affect the results. Subgroup results should be examined in future studies with more detailed clinical and socioeconomic indicators.
This study has several limitations. Firstly, the outcome of first hospital admission with diagnosis codes for kidney diseases has limitations when interpreted as the onset of kidney disease. According to the Centers for Disease Control and Prevention, most adults with CKD do not recognize they have CKD,11 and hospital admission can occur at more advanced stages of the disease or for treating complications attributable to kidney disease. Moreover, kidney disease is generally diagnosed through laboratory tests (eg, the estimated glomerular filtration rate, creatinine, and the accumulation of end products of nitrogen metabolism),35 49–51 and previous studies have reported that the incidence of kidney disease was substantially underestimated when ICD codes were used to define kidney disease.52 53 Therefore, the hospital admission records do not fully represent disease incidence, and our cohort probably undercounted the total kidney disease incidence. Additionally, because the use of ICD codes limits the classification of individual disease and the small sample size of specific disease, we did not investigate specific kidney diseases except for CKD and instead used the total kidney and urinary system diseases based on earlier published studies.28–34 Therefore, we were not able to investigate kidney and urinary disease separately, although an association exists between some kidney and urinary diseases.30–33 Therefore, our study did not investigate the potential heterogeneity of biological mechanisms among specific kidney and urinary diseases related to air pollution.
Secondly, the potential competing events (eg, hypertension, diabetes, and cardiovascular diseases; not death) that might affect the first hospital admission related to kidney disease35 49 54 were not considered in this study because of limited data availability. Thirdly, patients with CKD defined in this study represent a subset of those with total kidney disease; these patients were not independent of the total number of patients with kidney and urinary system disease. Therefore, the apparent air pollution related risks for total kidney and urinary system disease could be due to confounding with CKD,55 and bias could exist in the results of total kidney and urinary system disease resulting from estimating the effects on CKD. Furthermore, additional research is needed on the association between air pollution and CKD in relation to diagnosis, especially given the large number of participants with undetected disease. In this study CKD was identified as the primary diagnosis in less than 1%; most participants with CKD were identified through hospital admission for comorbidities.54 Other areas of interest are the impacts of air pollution on recurrent events. This study investigated first hospital admissions only, but subsequent hospital admissions might also be of interest. Importantly, the association between air pollution and CKD was present when primary and secondary ICD codes were considered, but not when the primary code only was considered, which warrants further investigation.
Fourthly, our results only represent the Medicare part A FFS population that does not include all Medicare beneficiaries (the Medicare FFS population covers up to around 75% of the Medicare population).23 We had no information on Medicare Health Maintenance Organization claims, therefore, we could not cover the entire Medicare population. Fifthly, although we adjusted for several neighborhood level indicators, Medicare claims do not include extensive individual level data on behavioral and socioeconomic risk factors, which could be crucial confounders. Additionally, owing to limitations of the study data, we were not able to consider the potential effect modifications by comorbidities and CKD stages. These unmeasured individual level confounders might help to elucidate the biological mechanisms underlying the observed association between air pollution and kidney disease, and they should be looked at in future studies. Finally, future research could also investigate complex air pollution mixtures. We focused on traffic related air pollutants of NO2 and PM2.5, although each of these pollutants also has non-traffic sources, and actual air pollution mixtures are complexes with other pollutants and different chemical structures of PM2.5. We also used state-of-the-science air pollution prediction data, which are only available until 2016; therefore, future studies should investigate the association between air pollution and kidney disease with more recent data.