Neighborhood Health and Outcomes in Idiopathic Pulmonary Fibrosis

Abstract

Rationale

Living in a disadvantaged neighborhood has been associated with worse survival in people with idiopathic pulmonary fibrosis (IPF), however, prior studies have only examined the impact of neighborhood health on outcomes in IPF as a composite measure.

Objectives

To investigate the association between neighborhood health and disease severity, measured by pulmonary function at presentation, and death in follow-up, with an additional focus on the contributions of the neighborhood’s underlying physical and social factors to these outcomes.

Methods

In a retrospective study of participants from the University of California, San Francisco, IPF Cohort (2001–2020), geocoded home addresses were matched to the California Healthy Places Index (HPI), a census-tract measure of neighborhood health. The HPI comprises 25 indicators of neighborhood health that are organized into eight physical and social domains, each of which is weighted and summed to provide a composite HPI score. Regression models were used to examine associations between the HPI as a continuous variable, in quartiles, and across each physical and social domain of the HPI (higher values indicate greater advantage) and forced vital capacity (FVC) percent predicted (% predicted), diffusing capacity of the lung for carbon monoxide (Dl_CO) % predicted, and death, adjusting for demographic and clinical covariates. We also studied the interaction between disease severity at presentation and neighborhood health in our time-to-event models.

Results

In 783 participants with IPF, each 10% increase in HPI was associated with a 1% increase in FVC % predicted and Dl_CO % predicted (95% confidence intervals [CIs] = 0.55, 1.72; and 0.49, 1.49, respectively). This association appeared primarily driven by the economic, education, access, and social HPI domains. We also observed differences in the associations of HPI with mortality depending on disease severity at presentation. In participants with normal to mildly impaired FVC % predicted (⩾70%) and Dl_CO % predicted (⩾60%), decreased HPI was associated with higher mortality (hazard ratio = 2.91 Quartile 1 vs. Quartile 4; 95% CI = 1.20, 7.05). No association was observed between the HPI and death for participants with moderate to severely impaired FVC % predicted and Dl_CO % predicted.

Conclusions

Living in disadvantaged neighborhoods was associated with worse pulmonary function in participants with IPF and was independently associated with increased mortality in participants with normal to mild physiological impairment at presentation.

Idiopathic pulmonary fibrosis (IPF) is a progressive, age-related form of interstitial lung disease (ILD) that leads to fibrosis of the underlying lung tissue (1). As the disease worsens, patients experience lung function decline and debilitating respiratory symptoms (1). Recent work examining the impact of contextual factors on outcomes in IPF has demonstrated that living in economically disadvantaged areas is associated with worse survival (2, 3). Because living in disadvantaged areas also contributes to an increased burden of symptoms and disability in other chronic diseases (4–6), neighborhood health may be a modifiable risk factor of disease morbidity in individuals with IPF.

Neighborhood health is defined by the availability of physical assets (e.g., quality of buildings/institutions/recreational space, access to services, and housing conditions) and the social surroundings (e.g., community cohesion, violence, and customs) that make up the local environment. These local conditions are influenced by policies that determine resource distribution across communities and influence exposure to environmental hazards, resulting in a significant impact on health at the individual level (7). Neighborhood conditions are often characterized by the level of disadvantage or deprivation in the area, and they are commonly captured by composite measures of aggregate-level socioeconomic indicators and features of the built environment (8). These composite measures are thought to more holistically capture neighborhood health compared with studies using single indicators and have been previously used to examine outcomes in IPF (3, 9). Missing from these studies, however, is the simultaneous examination of the composite neighborhood health measure and its underlying physical and social components. By studying both concurrently, our understanding of how the broad neighborhood context and its specific factors may be contributing to worse outcomes in IPF would be enhanced, leading to the design of more efficient interventions and policies that target low-quality neighborhood conditions.

The California Healthy Places Index (HPI) was developed by the Public Health Alliance of Southern California as a tool to assess neighborhood health across the state of California. The HPI is unique because its indicators are organized into social and physical health domains that can be summed to provide an aggregate measure of neighborhood health, or each domain can be examined individually to study their impact on health outcomes (10). Using the HPI, we sought to further examine the association between neighborhood health and disease severity in participants with IPF, measured by pulmonary function at presentation, and death, with an additional focus on the unique contributions of each social and physical domain of the HPI.

These results were previously presented at the American Thoracic Society International Conference in San Francisco, California, in May 2022, and at the 20th International Colloquium on Lung and Airway Fibrosis in Reykjavik, Iceland, in October 2022. This work was also published in abstract form (11).

Methods

Study Population

Participants with IPF were retrospectively identified from the University of California, San Francisco (UCSF), ILD Cohort, a longitudinal observational cohort consisting of a subset of participants seen in the UCSF ILD Clinic between August 2001 and February 2020. The diagnosis of IPF was made by a multidisciplinary team consisting of pulmonologists, chest radiologists, and pulmonary pathologists per consensus recommendations (1). In the present study, cohort participants with IPF were eligible for inclusion if they were 18 years or older, had pulmonary function tests (PFTs) documented within 12 months of their initial clinic visit, lived in California at the time of the initial visit, and had an address that could be geocoded. This study was approved by the UCSF Institutional Review Board (#10-01592), and written informed consent was obtained from all participants at the time of parent cohort enrollment.

Geocoding

Geocoding for each participant was performed using the Texas A&M University batch geocoding services (12). Home addresses recorded at the initial visit were used to obtain latitude and longitude, as well as census Federal Information Processing System codes for each census tract, county, and state using 2010 Decennial Census boundaries.

Primary Predictor

The California HPI 2.0 is a publicly available and evidence-based tool derived from 25 physical and socioeconomic indicators measured at the census-tract level. Indicators are further organized into eight “action areas,” or domains, which are well-described social determinants of health and linked to actionable interventions at the policy level: economic, education, housing, access, neighborhood, pollution, transportation, and social domains. Each of these are scored and weighted on the basis of their association with life expectancy at birth and summed to construct the composite HPI score (see the data supplement for source materials and score construction) (10, 13). We examined the HPI as a composite score (0–100), with a higher HPI score indicating more advantaged neighborhood conditions. Following the author’s recommendations and the current application by the California Department of Public Health (10, 14), we also compared associations across HPI quartiles, comparing the least advantaged neighborhoods to the most advantaged. This is also concordant with prior studies that applied quartile analysis of neighborhood disadvantage in IPF (3, 9), and for further comparability, we repeated analyses using the Area Deprivation Index (ADI), a publicly available composite index of neighborhood disadvantage, in place of the HPI (3, 8). The ADI ranges from 0 to 100, with lower scores indicating greater neighborhood advantage. The most advantaged quartiles for each index were set as the reference group (Quartile 4 [Q4] for the HPI and Quartile 1 [Q1] for the ADI) following the standard approach in health disparities research (15). Last, we separately examined each of the eight composite domain scores of the HPI to determine the direction of association and the magnitude of effect on selected outcomes.

Outcomes

Key outcomes of interest included forced vital capacity (FVC) percent predicted (% predicted) and diffusing capacity of the lung for carbon monoxide (Dl_CO) % predicted obtained closest to the date of the initial clinic visit. Because PFTs were obtained in laboratories that were both internal and external to UCSF, % predicted values were standardized using the Global Lung Initiative reference equations. We use the “Global Lung Initiative other” reference equation for FVC % predicted, because the appropriateness of including socially constructed variables such as race and ethnicity in pulmonary function equations is undergoing evaluation, and we wanted to avoid the possibility of masking effects that may be disproportionately experienced by certain racial or ethnic groups (16, 17). Time to death was included as an outcome with vital status and date of death determined using the National Death Index (18). Lung transplant was treated as a competing risk for death, and participants were censored at the end of the study period if nondeceased or if they had not undergone lung transplantation. Time to death or censoring was defined from the date of the first clinic visit.

Statistical Analysis

Complete case analysis was used in all models, and data were considered missing at random. Demographic and clinical covariates were selected a priori on the basis of an extensive literature review. These included age, sex, and race and ethnicity, which were included as confounders; and smoking history, asthma diagnosis, chronic obstructive pulmonary disease diagnosis, body mass index (BMI), and antifibrotic use at presentation, which were treated as mediators and adjusted for in our models to examine the direct effect of neighborhood health on each outcome of interest (19–27). Asthma was excluded from models examining death, and BMI was excluded from models examining associations with Dl_CO % predicted (see data supplement for a description of covariates and justification for exclusion). We used multivariate linear regression analyses to examine the associations between the HPI with FVC % predicted and Dl_CO % predicted. All model assumptions were met except for linearity for BMI, and the polynomial term was included in the models.

The association between HPI and time to death was assessed using the Fine-Gray subdistribution hazard model, where lung transplantation was treated as a competing risk, and models were performed in a staged fashion for consistency with prior studies (3). In the first analysis, we adjusted for clinical and demographic confounders and mediators. In the second analysis, to isolate the residual direct effects of HPI on mortality, we further adjusted for FVC % predicted and Dl_CO % predicted. Last, because pulmonary function decline is a strong predictor of mortality in IPF (19), the effects of neighborhood disadvantage on clinical outcomes, which occur upstream to individual-level behaviors and disease characteristics (7), may be difficult to detect from these downstream factors in more advanced disease. Therefore, to isolate the independent effects of neighborhood health on mortality, we included an interaction term between disease severity at presentation and the HPI in our time-to-event models. We first created a binary variable for FVC % predicted and Dl_CO % predicted. FVC % predicted values ⩾70% and Dl_CO % predicted values ⩾60% were defined as normal to mildly impaired lung function per interpretative guidelines for PFTs (27), and all values below these thresholds were considered moderate to severely impaired.

In all models, we scaled the continuous HPI and domain scores by 10% to aid in interpretability. Analyses using the ADI were performed using the same models and statistical analyses for comparison. P values ⩽0.05 were considered statistically significant, and for interaction studies, P values up to ⩽0.1 were considered clinically relevant and set a priori as the threshold to report stratum-specific estimates for normal to mildly impaired versus moderate to severely impaired pulmonary function groups. When testing each domain as an outcome, a false discovery rate was applied to account for multiple comparison testing (28). Models were run using the Stata software statistical package, Version 16.1/SE (29). Figures were created using Excel; Stata, Version 16.1/SE; and RStudio (30).

Results

Study Population

A total of 795 participants met the inclusion criteria. Six participants were excluded for missing PFTs or because they were obtained more than 12 months from the initial clinic visit, and another 6 were excluded because they lived in census tracts with a population less than 1,500 or with ⩾50% living in group quarters and, as a result, were missing an HPI score. The remaining 783 participants were included in the analysis (Table 1). The median duration of follow-up was 785 days (interquartile range = 347, 1,435). Participants lived predominantly in the San Francisco Bay Area, Northern California, and the Central Valley (Figure 1). The participant age, in years, was (mean ± SD) 71 ± 8.5, and participants were predominantly men. The median HPI percentile was 72.9 (interquartile range = 50.0, 92.0), and raw HPI scores were normally distributed. Further, the mean raw HPI score in our cohort was greater than the mean score observed within the state of California (P < 0.001) (see Figure E1 in the data supplement), with the majority of participants seen at UCSF living in neighborhoods with a percentile of 50 or greater (see Figures E2 and E3). The median ADI score was 12, with most participants living in neighborhoods with a score less than 40 (see Figure E4). The most advantaged HPI quartile (Q4) showed the highest frequency of men, non-Hispanic White race and ethnicity, and urban location and the lowest use of supplemental oxygen compared with the other groups.

Table 1.

Baseline characteristics of participants with idiopathic pulmonary fibrosis by neighborhood health quartile

	Quartile 1: Least Advantaged	Quartile 2	Quartile 3	Quartile 4: Most Advantaged
Characteristic	196 (25.0)	196 (25.0)	196 (25.0)	195 (24.9)
Age, yr, mean (SD)	70.3 (8.1)	69.3 (8.5)	71.6 (8.5)	73.9 (8.3)
Urban/rural residence, n (%)
Rural	26 (13.3)	26 (13.3)	9 (4.6)	3 (1.5)
Urban	170 (86.7)	170 (86.7)	187 (95.4)	192 (98.5)
Sex, male, n (%)	131 (66.8)	144 (73.5)	152 (77.6)	153 (78.5)
Race and ethnicity, n (%)
Asian/Pacific Islander	12 (6.2)	15 (7.7)	26 (13.4)	20 (10.3)
White	136 (70.1)	151 (77.8)	141 (72.7)	154 (79.0)
Hispanic/Latinx	34 (17.5)	17 (8.8)	25 (12.9)	15 (7.7)
Other*	12 (6.2)	11 (5.7)	2 (1.0)	6 (3.1)
BMI, mean (SD)	28.6 (5.5)	28.8 (5.7)	26.8 (4.1)	27.2 (4.2)
Current/former smoker, n (%)	130 (66.3)	140 (71.4)	130 (66.7)	126 (64.6)
Supplemental oxygen use, n (%)	53 (27.7)	39 (20.5)	30 (15.7)	25 (13.0)
Cough, n (%)	172 (88.7)	170 (86.7)	164 (84.5)	165 (85.1)
Shortness of breath, n (%)	175 (92.1)	177 (95.2)	166 (88.3)	156 (87.6)
GERD, n (%)	48 (24.6)	61 (31.3)	65 (33.7)	60 (31.1)
Asthma, n (%)	37 (19.1)	34 (17.3)	24 (12.3)	20 (10.3)
COPD, n (%)	49 (25.1)	35 (17.9)	50 (25.8)	34 (17.6)
FVC, mean (SD)	2.5 (0.8)	2.6 (0.8)	2.6 (0.8)	2.8 (0.9)
FVC % predicted, mean (SD)†	73.9 (19.5)	75.6 (19.7)	77.4 (20.1)	83.5 (20.9)
FEV1, mean (SD)	2.0 (0.7)	2.2 (0.6)	2.2 (0.6)	2.3 (0.7)
FEV1% predicted, mean (SD)	74.8 (19.6)	77.8 (20.0)	77.7 (18.9)	85.0 (19.8)
DlCO, mean (SD)	10.8 (4.1)	11.9 (4.4)	12.0 (4.7)	12.7 (4.3)
DlCO, % predicted, mean (SD)†	47.2 (16.9)	49.9 (16.3)	51.6 (19.5)	54.8 (16.8)
Antifibrotic use, n (%)‡	16/65 (24.6)	15/65 (23.1)	17/65 (26.2)	17/65 (26.2)
Follow-up time, d, median (IQR)§	777 (296, 1,503)	733 (316, 1,403)	798 (413, 1,306)	859 (365, 1,603)

Definition of abbreviations: BMI = body mass index; COPD = chronic obstructive pulmonary disease; Dl_CO = diffusing capacity of the lung for carbon monoxide; FEV₁ = forced expiratory volume in 1 s; FVC = forced vital capacity; GERD = gastroesophageal reflux disease; IQR = interquartile range; % predicted = percent predicted.

N = 783. Data are presented as mean (SD), median (IQR; 25%, 75%), or n (%).

^*Of the 31 participants in the race and ethnicity category “Other,” racial groups self-identified as Black or African American (39%), American Indian (29%), Middle Eastern (16%), multiple races (6%), or other/not specified (10%).

^† FVC % predicted and Dl_CO % predicted were calculated using the Global Lung Initiative reference equations.

^‡ Antifibrotic use is defined as use of pirfenidone or nintedanib use at initial presentation after Food and Drug Administration approval on October 31, 2014.

^§ Categorical statistical analyses were performed using Kruskal–Wallis testing because of skewed distribution.

Figure 1.

Distribution of the Healthy Places Index (HPI) by neighborhood quartile at the census tract level. (A) HPI quartiles displayed for all census tracts with an available HPI in the state of California. (B) HPI quartiles displayed for census tracts of all participants from the UCSF, Interstitial Lung Disease Cohort. UCSF = University of California, San Francisco.

Neighborhood Health and Disease Severity

Each 10% increase in the HPI percentile was associated with a 1.14% (adjusted 95% confidence interval [CI] = 0.55, 1.72; P < 0.001) increase in FVC % predicted at presentation. All HPI domains except for housing, neighborhood, and pollution were associated with FVC % predicted, with the largest effects observed in the economic, education, access, and social domains (Figure 2A). We observed congruent findings when examining across HPI quartiles (Figure 3A).

Figure 2.

Mean adjusted pulmonary function measures by Healthy Places Index domain. (A) Forced vital capacity percent predicted. (B) Diffusing capacity of the lung for carbon monoxide percent predicted. Solid lines represent 95% confidence intervals for each coefficient estimate. Asterisks indicate values that remained statistically significant after the false discovery rate was applied.

Figure 3.

Mean adjusted pulmonary function measures by Healthy Places Index quartile. (A) FVC % predicted. (B) Dl_CO % predicted. Dl_CO = diffusing capacity of the lung for carbon monoxide; FVC = forced vital capacity.

Each 10% increase in HPI percentile was associated with a 0.99% increase (adjusted 95% CI = 0.49, 1.49; P < 0.001) in Dl_CO % predicted at presentation. All HPI domains except for pollution were associated with Dl_CO % predicted, with the economic, education, access, and social domains showing the greatest magnitude of association (Figure 2B). We observed congruent findings when examining across HPI quartiles (Figure 3B).

We found similar associations between the ADI and FVC % predicted and Dl_CO % predicted, with higher ADI scores (greater disadvantage) showing an association with lower pulmonary function measures at presentation (see Table E1).

Neighborhood Health and Mortality

A total of 474 participants (60%) died during the study period. Each 10% increase in neighborhood health was associated with a 5% lower risk of mortality over time (adjusted hazard ratio = 0.95; 95% CI = 0.91, 0.98; Table 2). We observed similar findings when examining across HPI quartiles (Table 2). After adjusting our time-to-event models for disease severity at presentation (FVC % predicted and Dl_CO % predicted), the association of HPI with mortality risk disappeared (Table 2). Although the economic and education domains were associated with death after adjustment for clinical and demographic variables, this finding was no longer present in these domains, or any other domain, after adjustment for pulmonary function measures (see Tables E2 and E3). We found similar associations between ADI and mortality risk (see Tables E4 and E5).

Table 2.

Analyses of Healthy Places Index and death

Variable	Multivariate Analysis of Demographic + Clinical Variables*			Multivariate Analysis of Demographic + Clinical Variables + Pulmonary Function at Presentation†
	HR	CI	P Value	HR	CI	P Value
HPI (continuous)	0.95	0.91, 0.98	0.01	0.99	0.95, 1.04	0.8
HPI quartile
4	(reference)			(reference)
3	1.12	0.84, 1.48	0.4	0.92	0.69, 1.24	0.6
2	1.26	0.94, 1.68	0.1	0.94	0.69, 1.28	0.7
1	1.51	1.14, 1.99	0.004	1.06	0.79, 1.41	0.7

Definition of abbreviations: BMI = body mass index; CI = confidence interval; COPD = chronic obstructive pulmonary disease; Dl_CO = diffusing capacity of the lung for carbon monoxide; FVC = forced vital capacity; HPI = Healthy Places Index; HR = hazard ratio. Values in bold are statistically significant. Lung transplantation was treated as a competing risk. Continuous HPI score was scaled by 10% for interpretability.

^*Variables included age, sex, race and ethnicity, smoking history, COPD diagnosis, BMI, and antifibrotic use.

^† Variables included age, sex, race and ethnicity, smoking history, COPD diagnosis, BMI, antifibrotic use, FVC % predicted, and Dl_CO % predicted.

When examining the interaction between disease severity at presentation and the HPI in our time-to-event analyses (see Table E6), we found an association between the HPI and mortality for participants with normal to mildly impaired lung function values at presentation. However, this relationship was absent in those participants with moderate to severely impaired lung function values. For example, in participants with FVC % predicted values ⩾70% at presentation, reduced HPI scores (Q1 vs. Q4) was associated with an 82% increase in mortality risk (hazard ratio = 1.82; 95% CI = 1.21, 2.74; Figure 4A and Table E7). We found similar results both when examining normal to mildly impaired Dl_CO % predicted values (Figure 4B and Table E8 in the online supplement), by including both FVC % predicted and Dl_CO % predicted in models together, and when examining the HPI as a continuous variable (see Tables E7–E9). We did not perform further analyses of the HPI domains and death stratified by pulmonary function at presentation.

Figure 4.

Cumulative incidence of death by Healthy Places Index quartile is displayed for participants with (A) forced vital capacity percent predicted ⩾70% and (B) diffusing capacity of the lung for carbon monoxide percent predicted ⩾60% from the initial clinic visit up to 2,000 days or 5.48 years of follow-up time.

Discussion

In this single-center study conducted in California, we found that individuals with IPF living in the least advantaged neighborhoods presented with worse disease severity and that living in these areas increases the risk of death for those with earlier disease. Further, the relationship between neighborhood health and more advanced IPF appears to be predominantly driven by social rather than physical neighborhood components, a finding that has not been previously examined. We believe that this supports the critical role of addressing social determinants of health in the delivery of ILD care, both to improve access and health outcomes in populations living in the least advantaged neighborhoods.

Our finding of the effect of disease severity on the relationship between HPI and mortality risk is significant and novel. We hypothesize that this interaction may reflect a critical role for neighborhood health in the natural history of IPF. We know that IPF is a complex disease characterized by senescence of the alveolar epithelium, a fibroproliferative alveolar microenvironment, and increased susceptibility of the pulmonary interstitium to exacerbation and disease worsening from a variety of intrinsic and extrinsic factors (31). Because neighborhood health is an upstream determinant of individual-level health outcomes (7), the increased risk of death observed early in the disease course of IPF for participants residing in least advantaged neighborhoods may be masked by downstream pathophysiologic mechanisms once the disease severity reaches a critical threshold, where these processes may become the predominant driver of mortality later in the course of disease. This finding deserves further follow-up, as it suggests that efforts to address these social determinants early in the disease course are critical to improving the health of patients with IPF living in disadvantaged areas.

We found that most of the domain scores were individually associated with pulmonary function at presentation. For both FVC % predicted and Dl_CO % predicted, the economic, education, access, and social domains demonstrated the greatest magnitude of association with these physiologic measures. Although complex interrelationships exist within and between these domains of neighborhood health, our findings appear to demonstrate that social, rather than physical indicators, are the predominant drivers of the relationship between neighborhood health and IPF severity. Although more studies are needed to elucidate the reason for these findings, we hypothesize that living in disadvantaged areas exposes individuals to fewer resources and fractured access to care compared with living in more advantaged communities (32, 33). For example, residents of high-poverty neighborhoods often have fewer employment opportunities and are more likely to receive a lower quality education, both of which are known to affect access to care (34, 35). Although few studies have examined the effect of socioeconomic and contextual factors on access to care in IPF, there is an increased recognition that interventions and policies that allocate resources to less advantaged communities are especially impactful for populations with rare diseases because of the high burden of disability, costly and limited treatment options, and the increased dependence on access to specialty care for early disease recognition (36–38). In addition to these longer term interventions targeting neighborhood quality, more proximal effects of neighborhood disadvantage on health outcomes, such as poor access to high-quality ILD care, can be targeted with interventions such as telehealth, which are effective in improving access to specialty care (39).

It is interesting that we did not observe an association between the pollution domain and our pulmonary function measures at presentation, despite prior work demonstrating this relationship in IPF and other fibrotic ILDs (40, 41). Our findings may be related to the composite nature of the pollution domain, where differences in the direction of association for each individual indicator may be obscured when examined as an aggregate measure. Further, the source material to construct the domain is temporally nonspecific; an average value of each air pollutant is obtained over a time period of 1 to 3 years, and cumulative exposure is not measured. Thus, our findings should be interpreted with caution, and we suggest that conclusions about the impact of pollutants on pulmonary function measures in IPF should not be drawn from this study.

A relationship between neighborhood disadvantage and disease severity at presentation has not been universally described. Prior work from Goobie and colleagues examined neighborhood deprivation in participants with fibrotic ILD, using the ADI as a measure of neighborhood disadvantage for a single-center U.S. population (University of Pittsburgh ILD Cohort) and the Canadian Index of Multiple Deprivation for the multicenter Canadian population (Canadian Registry for Pulmonary Fibrosis) (42). In the subgroup of participants with IPF, Goobie and colleagues found no association between neighborhood deprivation and FVC % predicted or Dl_CO % predicted at the time of presentation in either population (3). We believe that the robustness of our findings across multiple analytic approaches (continuous and categorical) and indices (HPI and ADI) strengthen confidence in their validity. In addition, similar results were found in an earlier report from our group using the CalEnviroScreen 3.0, an index of neighborhood disadvantage comprising indicators that measure exposure to pollution and socioeconomic conditions (9). It is quite possible that regional/national differences in access to health care and other important social determinants of health may have lessened the association’s statistical strength in the work from Goobie and colleagues.

There is more consistency in the literature regarding the impact of neighborhood disadvantage on mortality risk. Although findings from the Canadian cohort in the study by Goobie and colleagues showed no independent association between neighborhood deprivation and mortality risk, an increased mortality risk was observed for participants with IPF living in the most disadvantaged neighborhood quartile from the U.S. center (3). They hypothesized that differences in the U.S. and Canadian healthcare systems might be responsible for lessening this disparity. A separate study examining participants with IPF in France showed that living in low-income cities was associated with mortality risk, independent of similar demographic and pulmonary physiology covariates (2). Heterogeneity in both disease severity at presentation and mortality risk across studies is likely due, in part, to unique national-, state-, and local-level differences in neighborhood conditions, population characteristics, regional referral patterns, and healthcare payor systems and policies. Multicenter studies conducted across regions are needed to further examine the complex relationship between neighborhood health, including the area’s underlying social and physical conditions, and outcomes in IPF such as pulmonary function decline and death.

There are several important limitations of this study. First, the retrospective nature of this study limits our ability to further explore the associations described and does not account for individual-level socioeconomic variables that may be contributing to disease severity or mortality. Prospective, multicenter studies are sorely needed. Second, the relatively high socioeconomic status of our participants may impact the external validity of our results; further, the HPI is a tool that is currently only applicable to populations living in California. Next, the initial clinic visit represents the time that participants were first seen in the ILD care center, which may not correspond to the date of IPF diagnosis. Last, our findings of worse pulmonary function in participants living in the most disadvantaged neighborhoods could be due to selection bias, where only patients with the most complex or advanced disease from these areas are referred to ILD care centers. However, our findings of worse survival for participants from the least advantaged areas with normal to mildly impaired physiology at presentation highlights that patients with earlier disease are reaching ILD centers.

In conclusion, we found that participants in California with IPF living in the least advantaged neighborhoods presented to an ILD care center with worse disease severity, an association primarily driven by the neighborhood’s social factors. This translated to increased mortality risk for those with earlier disease, suggesting that early interventions to address the health effects of social determinants and less advantaged neighborhoods are critically important. Researchers, community organizations, funders, and policymakers should come together to design and conduct multicenter, multiregional, prospective studies that more rigorously define and assess social determinants and other factors important to neighborhood health in patients with IPF; more robustly identify the key factors associated with poor health outcomes; and design interventions to address these factors through clinical practice and health policy.