Identifying Racial Disparities Within Healthcare Costs in New York State
Healthcare in the United States has been subject to much scrutiny, with many important topics of discussion occurring with regards to access of care, cost, and quality. Unlike many western countries, the United States does not have universal healthcare. The United States relies on a direct-fee system for its healthcare cost structure, where patients under the age of 65 are expected to pay for medical costs themselves, with additional help provided through employers’ insurance policy.
At a time when racial tensions are high in the United States, with ongoing protests that highlight the issue of systemic equality and within the United States. With that in mind, I wanted to highlight one aspect of systemic inequality and investigated cost discrepancies between races with regard to healthcare costs.
I looked at a dataset focusing on hospital discharges in the state of New York between 2014 and 2016, in order to gain insights on the aforementioned issue. The data was collected through The Statewide Planning and Research Cooperative System (SPARCS). The data field captured patient characteristics, diagnoses, treatments, services, and charges incurred at the time of discharge. The data does not contain any protected health information that would be in violation of the Health Insurance Portability and Accountability Act (HIPAA). The 3 datasets i.e. 2014, 2015, and 2016, contained over 5 million records of patient data, which I merged together and analyzed to look at some of the trends within those 3 years, to explore potential discriminatory discrepancies among different races in the New York State hospital system.
I began by looking at general trends within the data. Looking strictly at how the average healthcare cost differed between different race groups. The data had split all the patients into 4 racial groups: White, Black/African American, Asian/Native American/Pacific Islander, and Multi-Racial, which encompassed any mix of the previous 3 races mentioned. Beginning with this general perspective, I delved into the data.
As shown above, White patients incurred the lowest average costs at the time of their discharge at $13,922. The next closest group was the Asian/Pacific Islander/Native American category, which incurred a total cost of $15,086 on average, with Black/African American and Multi-Racial patients receiving an average even higher than that ($15,846 and $20,124 respectively).
I wanted more detail about the situation, and to see if these situations existed within specific conditions, such as with other dimensions of data baked into the analysis. By looking deeper into the data, I would see where some of the biggest problem areas are that may go into explaining this racial disparity in cost in greater detail. The first area I looked into was whether there were any trends in the data if the analysis contained an additional dimension of age groups being taken into account.
Looking at the trends, it seems as though the average cost that patients incur increases for every racial group as the age of the patient increases. The increases to costs associated with age seem to be consistent across each racial group, with only African American patients not experiencing a drop in average cost for those at 70 years or older compared to patients between the ages of 50 and 69. However, looking again at racial disparity, White patients tend to have a lower overall cost at every age group listed within the data.
As Americans have to pay for each night that they take up a hospital bed, I wondered whether the minority groups that received higher costs happened to experience longer stays during their hospital visits. I decided to look at the what the average length of stay was for each group against their total costs to see if they could partially explain the differences in cost between racial groups for the healthcare service they get.
This graph does help explain some of the disparities a little better, as Black patients on average stay at the hospital almost a day longer than the lowest group of Asian Americans/Pacific Islanders/Native Americans. However, even within this data, the Asian/Pacific Islander/Native group still had a higher cost at discharge than their White counterparts despite having a slightly shorter average length of stay, which still confounded the reason as to why White patients had lower costs, highlighting some disparities again between race and costs.
I next wondered if this disparity may be present within only certain types of hospital visits, such as emergency room visits or urgent care. This data also had a column that indicated the Type of Admission that each patient fell under. These admissions ranged from Elective surgeries to Newborns, as well as 3 other categories listed below. When looking at this data split by race and measured against average cost, similar trends appeared as the previous graphs.
As seen in the graph above, White patients on average also paid the lowest costs in 4 out of the 5 categories within the data that they were labelled under, with only White patients who came in for Trauma-related reasons experienced an average total cost that was not the lowest out of the 4 racial groups in the data. This was also the only case where the Asian/Native/Pacific Islander races and Black races had a lower average cost at discharge. What surprised me the most was the cost difference in Newborn hospital discharges, where I didn’t anticipate seeing such a large gap between the least and most expensive average cost for each group. It appears as though families of non-White races experience much higher costs — sometimes double — compared to their White counterparts. Once again, despite adding another level of dimensionality to better understand the data, a similar conclusion came up with few exceptions. When I delved deeper into these categories, looking up specific disease diagnoses (CCS.Diagnosis.Description Column in the diagram below), while controlling for severity of illnesses and comparing them across each race, I found more disturbing figures within the data. For patients that had a similar description of their diagnosis, and similar severity indices attached to their respective diagnoses, there were still racial differences with the average costs incurred at discharge. The severity of the illness increases with each number (1 = Mild, 2 = Moderate, 3 = Major, 4 = Extreme).
As the diagnosis description for patients is a lengthy list, especially when split into respective severities, I have posted a snapshot of the data for 4 diagnoses. Some of the trends within this snapshot that I noticed were that not only were White patients often incurring lower costs for their healthcare visit for each diagnosis, the increases in average cost as the severity of the diagnosis worsened was not as high as minority groups in the same situation. For example, those suffering heart attacks (acute myocardial infarction) may be paying around the same price for healthcare charges with a diagnosis with the least severity on the index, but a minority could be charged almost $20,000 more on average than their White counterpart if the diagnosis was coded at the most severe index, highlighting a disturbing trend between the respective racial groups.
At this point, I want to point out a few caveats within the data that may play a factor into the discoveries that I have displayed thus far. For one, White patients outnumber every other minority group at a ratio of approximately 3:1. This had the potential to skew the results when looking at average healthcare costs between each group. This was also the reason I did not make any definitive conclusions about the multi-racial group throughout the article, as there are only 55,000 records within the data for this group.
To mitigate for the difference in sample size between racial groups, I also looked at the median of these various comparisons and found similar results to my previous findings, most of which I have posted below. With the exception of the multi-racial race group, it was already unlikely that outliers would largely affect the conclusions made given such a large sample of each race within the data. However, to be more accurate, I analyzed the same data with the median cost for each group rather than the average. Below are the results of the visualizations presented throughout the article with the median used instead of the mean. You’ll find that the results are closer, however the conclusions remain largely the same.
When using the median, the White and Asian groups come closer together in their healthcare costs, however a Black patient could still spend almost $1000 more than a White patient by the time they are discharged from the hospital.
Although the results aren’t quite as egregious as when I used the average cost per patient, the conclusions also largely remain the same when looking at the differences between each group at different age groups. White patients often pay up to $2000 less than some of their minority counterparts at each specific age bracket.
Looking at the median costs for various admissions, I found the merit of using the median rather than the average, as there was a little more equality between racial groups. At the time of discharge for elective procedures and trauma cases, White patients actually experienced higher median costs for their hospital visit and experienced lower median costs than every other minority group in emergency, newborn, and urgent care admissions. Out of the 3 groups with reputable sample sizes, Black patients never incurred the lowest median costs for any admission type, however were in the middle for elective and trauma admissions costs, with Asian/Pacific Islander/Native Americans being charged the least in these 2 categories.
Finally, looking at specific diagnoses with their associated severity and the median costs associated with them between racial groups, I found that even when looking at the median costs between each race, White patients were still paying less for specific diagnoses at every level of severity. Once again, the differences between each race are not as egregious as when the average was used, however White patients again incurred less costs at each level of severity for its respective diagnosis, and experienced smaller increases in cost with increasing severity in the diagnoses compared to their minority counterparts.
I looked at some of the literature on this topic that might explain some of the findings within this data. After reading an article by Dr. Wayne J. Riley, that found that Black patients were less likely than White patients to receive proactive diagnostic and vascularization procedures despite having similar risk characteristics[1], which could explain why the most severe cases of acute myocardial infarction are associated with a median cost that is $7000 greater in Black patients than White patients within this dataset.
Having looked at the paper by Dr. Riley as to why these cost differences may exist, it would be wrong to conclude that costs associated are due to the hospital voluntarily charging minorities more than their White counterparts when it came to healthcare costs. Rather, the analysis made demonstrates one of the by-products of a systemic problem where minorities are at a disadvantage and incur a higher cost in more ways than one. For example, Black and Latino communities are far more likely to not have adequate health insurance, which leads to fewer preventative services being available to them, as well as being subject to poorer health outcomes and only turning to hospitals when they are in desperate need of care[1]. According to Becker’s Hospital review, 9% of the New York state population is uninsured.[2]
Socio-economic status is another issue, as those within poorer communities may experience greater debilitation to their health due to certain environmental pressures. Household income, employment status and education also influence health inequity between races. In the future, better health education for at-risk communities, more inclusive insurance plans and public policy could be some ways to address and mitigate some of the underlying problems that lead to rising healthcare costs for disadvantaged communities.
To conclude, systemic inequalities have been perpetuated between various racial groups in this institution given the data that I analyzed, and not just as a sweeping generalization. In almost every dimension of data, there is some discrepancy between the costs associated with getting medical attention at a hospital between people of different races. This also lends support to the claim that many minority communities do not have the same resources in their neighbourhoods as their White counterparts, leading to many services being underfunded, which contributes to poorer health outcomes as people in these situations are more likely to visit a hospital as a last resort rather than as a preventative measure. It is my hope that in the future, hospitals and other important societal institutions are looked at holistically as it is clear that this is a multi-faceted issue that requires revamping many social, educational and health systems simultaneously in order to achieve optimal outcomes for every single person.
References:
1) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3540621/
2) https://www.beckershospitalreview.com/lists/healthcare-in-new-york-state-10-things-to-know.html
2014 Dataset:
2015 Dataset:
2016 Dataset:
