Year: 2023
Type: editorial
Abstract: In the study by Shoag et al. (1) in this issue of the Journal, investigators capitalized on publicly available home value information provided by the real estate marketplace company Zillow to understand the relationship between individual wealth and cancer survival. Estimates of home values from Zillow, known as “Zestimates,” were linked with cancer diagnoses and survival outcomes within the Ohio Cancer Incidence Surveillance System. Analyses also included more conventional area deprivation indices as a measure of census block–level economic and material disadvantage. The authors found that increasing (logged) home prices, adjusted for age and stage at diagnosis and area deprivation index, were associated with a modest decrease in the hazard of all-cause and cancer-specific mortality (hazard ratio = 0.92 and hazard ratio = 0.95 per SD, respectively), which equated to roughly an 8 percentage point decrease in all-cause mortality for each 40% increase in estimated home price. The association between home price and mortality was substantially more prominent for patients living in less deprived census block groups. For example, in low deprivation (ie, resource-rich) neighborhoods, the top vs bottom quartile of home prices were associated with a staggering 13.1-year vs 5.0-year overall survival. The overall conclusion is that home prices are correlated with survival in patients with cancer and that this survival advantage is differentially distributed within neighborhoods with similar area deprivation indices. We applaud the authors on an innovative investigation of socioeconomic and geographic disparities in cancer using a freely available, modern data source to augment other types of social determinants of health information—often measured at the census block or county level—with individual household-level data. These data add to a longstanding and robust body of evidence that dates back at least 6 decades that has previously shown that lower socioeconomic status—at both the individual and area levels—is consistently associated with worse overall health outcomes, including cancer outcomes (2,3). Importantly, prior evidence has shown that although absolute income and wealth at the individual household level are important, the relative differences in income and wealth—inequalities across groups—are also substantial, if not even stronger, determinants of health outcomes in diverse populations. These associations have been observed over time, across countries and cultures, and span the life course and multiple different types of health outcomes, from maternal, child, and infant mortality and disease-specific mortality to overall life expectancy (4). A 2016 study demonstrated longer life expectancy of 15 years and 10 years, respectively, among US male and female patients in the top 1% vs bottom 1% of income, based on tax records (5). For Americans with cancer, the intersection of race, individual socioeconomic status, and area-level disadvantage contributes to longstanding, well-documented disparities in cancer mortality (6,7). Sociologists, demographers, economists, and epidemiologists alike have contributed to this evidence, and it has motivated development of some of the most cited public health theories, models, and frameworks of the modern era that share a common question: “How can we explain and intervene in the process through which material wealth influences health?” Shoag and colleagues (1) explored their data in numerous subgroup and sensitivity analyses by cancer type, stage, and era, all showing the same general patterns. The largest socioeconomic disparities were observed in melanoma; lymphoma; and breast, prostate, and uterine cancers, which seems plausible because early detection and timely access to effective treatments are critical for these cancers, which have widely documented barriers to care access by socioeconomically disadvantaged groups. The introduction of novel, effective therapies has varied widely by cancer site. For example, 2011 marked the introduction of immune checkpoint therapy for melanoma, but this therapy was not introduced until 2014 and later for lung, kidney, and other diseases. The treatment for malignant brain gliomas, in contrast, remains noncurative to this day. In the current study, brain cancer–specific mortality was not linked with home price and area deprivation index. Importantly, in fully adjusted models, including area deprivation indices, home price was not consistently independently associated with cancer-specific mortality but was associated with overall survival, suggesting that other noncancer factors, such as other comorbidities that are socioeconomically driven, may be more influential causes of death. The study tends to intermix results that reflect overall survival with cancer-specific survival. The differences in overall survival are far more pronounced, but this is not emphasized. It is worth considering alternative interpretations of the data because there are several potential sources of bias and confounding regarding the use of Zestimates to proxy individual household-level income and wealth. Perhaps most importantly, the study used Zestimates from 2022, whereas the study cohort was diagnosed between 1996 and 2016. Therefore, the house price estimates used were obtained anywhere from 6 to 26 years after a patient’s cancer was diagnosed. The study authors suggest that the delay is not problematic because the rank order pricing of housing should be stable. There is reasonable concern, however, that this delay could have introduced substantial bias. Someone dealing with an aggressive cancer who has a poor performance status (or survival) would not be expected, for example, to be undertaking aggressive renovations that could increase or maintain the future home valuation. Another potential source of bias is from missing data. Roughly half the study sample had to be excluded due to missing demographic, Zillow, or area deprivation index data, and it was noted that the missingness was strongly nonrandom. Patients with missing home prices had higher rates of area deprivation than those with available home prices. If missingness of home prices is correlated with unfavorable area deprivation index metrics, it is also possible that missing prices may be particularly common for rural communities, those with stagnant housing markets, or “older” neighborhoods in which few houses turn over ownership because Zillow cites inadequate house and nearby transactional data as 1 of the most common causes for a Zestimate being unavailable. In contrast, one could imagine that younger, healthy, actively working, educated professionals would be the most likely to change housing frequently, thereby correlating higher Zestimates with more favorable patient characteristics. Zillow estimates themselves are created by an algorithm, speculative, and imperfect. The study largely assumes that Zillow price estimates are the gold standard, but no comparator is offered, such as actual home sale prices, which could conceivably be gathered from state records. If Zillow prices were artificially inflated in high-income areas or exhibit heteroskedasticity, one could conceive of additional potential sources of related bias. The question of whether patients owned vs rented the house in which they live would also be expected to be highly correlated with personal wealth and income (and misclassified for those patients with cancer who rented their house but were “assigned” a Zestimate for their landlord). These nuances were not disambiguated in this study. How could such potential sources of bias be detected or addressed? There is a precedent for the use of alternative, even more direct individual-level income data, such as tax returns (5). Actual house sale prices or property tax evaluations may be other options, each of which could address the temporal problem of using 2022 price estimates. It also would have been possible and helpful to dive into the differences in demographics and stage by Zestimate and missingness. Were Zestimates correlated with stage of cancer diagnosis? Age? Marital status? A deeper understanding of the causal mechanisms underlying such relationships would be fertile soil for future research because these mechanisms are the key to developing and implementing solutions to address socioeconomic disparities in mortality. We know from prior work that individual household income and wealth are distributed differentially across demographic and geographic subgroups. We also know that economic advantage influences receipt of cancer screening, timely cancer diagnosis, timely and guideline-concordant treatment, and monitoring of symptoms and progression during treatment through multiple pathways, including through access to employment, health insurance, health information and awareness, social and economic capital, and ability to travel for care. How to intervene successfully in these multilevel, transtheoretical social determinants of health is the real billion-dollar investment question. For example, interventions to directly improve material well-being, such as unconditional cash transfers or guaranteed incomes, may lead to better health outcomes. Other interventions are focused on expanding access to health insurance or addressing employment protections for people with cancer. These are promising areas for future policy work because we know that it is not speculative home price estimates that matter but rather the cumulative effects of complex, interwoven socioeconomic dynamics. In conclusion, the study by Shoag et al. (1) is an interesting study that uses a novel approach to proxy household income and wealth through home price valuations. This descriptive study helps move measurement toward individual-level metrics of socioeconomic status in addition to area-level measures of socioeconomic deprivation. The largest signal, however, is for overall survival, not cancer-specific survival. The study yields almost as many interesting questions as it does results and points toward future work to better understand both individual household-level and area-level socioeconomic determinants of health disparities. No new data were generated or analyzed for this editorial. Michaela A. Dinan, PhD (Conceptualization; Supervision; Writing—original draft; Writing—review & editing), Stephanie Wheeler, PhD (Conceptualization; Writing—original draft; Writing—review & editing). No funding was used for this editorial. M.A.D. has no relevant conflicts to disclose. S.B.W. receives grant funding paid to her institution from Pfizer and AstraZeneca for research unrelated to the topic of this editorial.
Source: JNCI Cancer Spectrum
Authors: Michaela A Dinan, Stephanie B Wheeler
Language: English
Subfield: Health
Field: Social Sciences
Domain: Social Sciences
Open Access status: gold
Estimated APC paid: $1,654
OpenAlex ID: W4388684673