Adult leukemia in Florida 2010-2019: diverse, aging population as an indicator for the United States

Leukemias are blood cancers that cause significant morbidity and mortality.¹ Globally, in 2022, there were ∼410 000 new cases (2.2% of all cancer cases) of adult leukemia (defined as leukemia diagnosed in those aged ≥20 years because patients aged 15-19 years are considered adolescents²) and 270 000 total deaths (2.9% of all cancer deaths).³ Although recognized for decades, little research has been performed to uncover the reasons for geographic variation in the distribution of leukemia. For example, countries with the highest leukemia incidence rates are in regions of Australia and New Zealand, North America, and Europe, whereas the incidence rates for countries in Asia and Africa are generally low.^3,4 The geographic variation in leukemia burden reflects different distributions of leukemia risk by age, sex, race/ethnicity, behavioral risks (eg, obesity and smoking rates), and environmental exposure (eg, benzene and radiation).^5,6 In the United States, the 2016-2020 age-adjusted incidence rate of leukemia is 13.9 cases with 6.0 deaths per 100 000 population per year.⁷ Most diagnoses were in older adults, as age-adjusted incidence and death rates in those aged ≥65 years were 60.8 and 36.6 per 100 000 persons per year, respectively, compared with 7.2 and 1.6 in people younger than 65 years.⁷ Among the 3 dominant racial/ethnic groups, the non-Hispanic White (NHW) population has the highest age-adjusted incidence rate, followed by the Hispanic population, and then non-Hispanic Black (NHB) population.⁷ Similar trends are found with respect to mortality rates (ie, number of deaths divided by number of population at risk), although NHB population generally has higher odds of dying (ie, number of deaths divided by number of leukemia diagnosis).⁸ 

In the past decade, leukemia incidence and mortality rates have trended down in the United States.^9-11 However, the benefits of improved leukemia control were not shared equally among all racial/ethnic groups. The average ratio of leukemia mortality rate to incidence rate for the NHB population from 2011 to 2020 was 0.49, compared with 0.43 and 0.39 in the NHW and Hispanic population, respectively.⁷ Population-based studies have also revealed survival disparities by leukemia subtypes with respect to socioeconomic status.^12-15 As the percent of the population older than age 65 years is expected to increase in the coming decade,¹⁶ it is important to study the epidemiology and disparities associated with all subtypes of leukemia in the United States to identify and address survival disparities. Of all states in the United States, Florida has the highest leukemia incidence rate.⁶ This may be partially explained by the Florida population comprising of 17.6%, 17.3%, and 22.6% individuals aged ≥65 years (ie, older adults are at high risk for leukemia) for years 2000, 2010, and 2020, respectively.¹⁷ The continuous growth of older populations in Florida is fueled by interstate migration of retirees from northern states (eg, New York, Pennsylvania, and Ohio) to Florida.¹⁸ Florida also has a strong ethno-geographic identity, with the Hispanic and NHB populations from Latin America and the Caribbean islands concentrated in central and southeastern Florida.^13,19,20 Florida, however, is not part of the Surveillance, Epidemiology, and End Results (SEER) registry system,²¹ and previously cited studies of leukemia epidemiology^9,10,12,13 using SEER data did not include data from Florida, potentially underestimating the US leukemia burden. This paper presents the findings from a comprehensive investigation of leukemia incidence rate, mortality rate, and sociodemographic disparities by race/ethnicity in Florida. As health care issues related to racial/ethnic diversity and population aging are becoming relevant to all US states, findings from our study of leukemia epidemiology and disparities in Florida represent a bellwether for the entire United States.

The Florida Cancer Data System (FCDS), the state’s official cancer registry, provides cancer incidence rates and mortality data. All adults aged ≥20 years diagnosed with leukemia in Florida from 2000 to 2019 were included in this analysis. Data from 2010 to 2019 were used to calculate average incidences and mortalities in the last decades (Tables 1 and 2), whereas those from the entire period from 2000 to 2019 were used to analyze yearly trends of leukemia incidences and mortalities in the last 2 decades (Figures 1 and 2). These cases were grouped by the 3 major racial/ethnic groups (ie, NHW, Hispanic, and NHB), which account for 95% of Florida population. Leukemia cases from other racial groups were not included in the analysis due to low prevalence (low statistical power) in the population. Subtypes of leukemia were identified with histology codes of the International Classification of Diseases for Oncology third edition.²² We separated acute myeloid leukemia (AML) cases into acute promyelocytic leukemia (APL; International Classification of Diseases histology code 9866) and non-APL (ie, specific AML histologies other than APL, including 9840, 9861, 9865, 9867, 9869, 9871-9874, 9895-9897, 9898, 9910-9911, and 9920). Acute lymphoblastic leukemia (ALL) includes B-cell ALL (9811-9818 and 9826) and T-cell ALL (9837). Note that B-cell ALL cases diagnosed before 2010 were identified by codes 9727, 9728, 9835, and 9836. Chronic lymphocytic leukemia cases were identified by code 9823 and chronic myeloid leukemia (CML) by 9863, 9875 to 9876, and 9945 to 9946.

Statewide cumulative incidence and mortality rates from 2011 to 2019 were calculated by dividing the total numbers of leukemia cases by subtypes for NHW, Hispanic, and NHB with 2010 population from the US Census. Age adjustment of the cumulative incidence and mortality rates were made by applying 2000 standard population weights per the National Cancer Institute (NCI) SEER.²³ Socioeconomic data were sourced from the 5-year estimates of 2015 American Community Survey (ACS) of the US Census, including median household income (ACS subject table S1901: income in the past 12 months), percent population with no insurance (B27010: types of health insurance coverage by age), percent population with bachelor’s degree (S1501: educational attainment), and percent workers with no vehicles (B08141: means of transportation by vehicles available). These socioeconomic variables are important factors that can affect the mortality risk of patients with cancer, because they determine available resources and propensity for accessing required leukemia care services. The Population Level Analysis and Community Estimates (PLACES) data released by the Centers for Disease Control and Prevention provided census tract-level estimates of behavioral risk that can affect the chance for survival of a patient with leukemia, including prevalence of smoking, obesity, lack of physical activity, and routine physician checkup. Note that the PLACES data document behavioral risk factors in the population and do not contain socioeconomic variables. ACS data produced by the US Census are frequently used authoritative data source for socioeconomic variables. To identify locations where leukemia treatment is available in Florida, listings of oncologists and hematologists treating leukemia were downloaded from the state of Florida Department of Health Practitioner Profile Search website,²⁴ supplemented with information from Google Maps, practice website searches, and calling individual offices to inquire about treatment of leukemia. Cancer centers were also included and not limited to NCI-designated cancer centers.

It is important to note that the analyses we performed for socioeconomic and geographical disparities (Tables 3 and 4; Figure 1) are based on census tract data in FCDS. SEER data based on county-level data cannot reflect leukemia disparities within a county.

For comparison of means among the 3 groups of poverty levels in Table 3, 1-way analysis of variance (ANOVA) appropriate for comparing means of multiple groups was used. P values from ANOVA tests for all variables are included in Table 3. For comparison of means between urban vs rural areas, t test was used. Linear regression was used to analyze factors associated with leukemia mortality odds by races/ethnicities, including access to leukemia treatment services, socioeconomic variables, and behavioral risk factors.

To characterize geographical disparities associated with leukemia mortality odds in Florida, we performed a geospatial hot spots analysis to identify census tract clusters where leukemia mortality odds (2010-2019) are significantly higher than statewide averages for each of the 3 racial/ethnic groups.²⁵ A geospatial hot spot refers to a location where the average value (eg, number of leukemia deaths) for a group of neighboring locations (ie, identified within a predefined measure of proximity surrounding the location) is significantly higher than the average value of a study area. Analysis of geospatial hot spots essentially identifies patterns of clustering with high values in a predefined area. We used ArcGIS Pro 3.0, which implements the Getis-Ord Gi∗ statistic, to identify clusters of census tracts where leukemia mortality odds are significantly higher than the average in Florida.²⁶ A key input to a Getis-Ord hot spot analysis is the number of neighboring locations (ie, k nearest neighbors) used to calculate the Gi∗ statistics.²⁷ We used Moran’s I²⁸ spatial autocorrelation analysis²⁹ to determine the number of census tracts neighbors (ie, the value of k) where spatial clustering is most intense for calculation of the Getis-Ord Gi∗ statistic. With this procedure, we identified hot spots for leukemia mortality odds for Hispanic, NHW, and NHB populations in Florida. After the hot spots were identified, we overlaid locations of leukemia treatment providers on the hot spot maps to identify regions in Florida where access to leukemia care may be attributable to high mortality odds. In conjunction with the hot spot analysis, we also created a variable for every census tract in Florida that measures the number of leukemia doctors that can be reached within 30 miles of a census tract. Depending on traffic conditions and hours of day in different parts of Florida, time required to travel 30 miles can vary greatly from 30 minutes to an hour. We chose the distance of 30 miles to measure accessibility of leukemia treatment services based on a paper that suggests the propensity for one to visit a doctor can significantly decline when travel time required to reach services lasts for 30 to 60 minutes.³⁰ 

Consistent with findings from previous studies, the most common leukemia subtype is chronic lymphocytic leukemia, followed by non-APL, AML, and CML, with the NHW population having the highest incidence rates and oldest average ages for these 3 subtypes (Table 1). Overall, the NHW population dominated the number of leukemia cases and incidence rates and the NHB population was often the youngest at diagnosis. Variations of population-based crude and age-adjusted mortality rates by leukemia subtypes and racial/ethnic groups generally followed the same patterns as incidence rates; however, the NHB population had the highest odds of dying from leukemia for most subtypes.

To assess how population aging could affect leukemia control in the coming decades, Table 2 shows leukemia incidence and mortality rates by races/ethnicities for 2 age groups: cases diagnosed in adults younger than 60 years and those aged ≥60 years. Except for APL and T-cell ALL, the numbers of cases diagnosed in adults aged ≥60 years exceed those in the younger group by a large margin for all subtypes. The numbers of mortality and crude mortality rates in Table 2 follow the same distribution patterns as the number of cases diagnosed and incidence rates. Survival disparities for NHB patients with leukemia are uncovered by the highest mortality odds in the NHB population for all subtypes and age groups, except for APL and CML cases aged ≥60 years, for which the NHW population show the highest mortality odds.

Table 2 also reveals the numbers of leukemia cases per year by patients’ birthplaces with respect to US states or foreign countries. The birthplace data are often missing for cancer survivors in the FCDS database because FCDS links cancer registry data with death certificate data from Florida Department of Vital Statistics to fill in the birthplace data.³¹ For all Florida adult leukemia cases with valid birth state information, only ∼25% of the cases were born in Florida. The same ratio is lower at 20% when only cases aged ≥60 years are considered. The top 4 states where Florida patients with leukemia migrated from are New York (19.2% of the cases), Pennsylvania (5.9%), Ohio (4.9%), and New Jersey (4.7%), all of which are within the top 10 states with the highest number of annual leukemia case count in the United States.⁶ New York, Pennsylvania, and Ohio are also the top 5 states from which retirees most often move, whereas Florida is the number 1 US destination for retirement.¹⁸ All the statistics cited above indicate that most NHW patients with leukemia in Florida were retirees who developed leukemia in Florida.

To assess whether the findings in Tables 1 and 2 based on data from the last decade (2010-2019) may carry over to the coming decades, we evaluated the trends of leukemia cases and mortality rates for the past 2 decades from 2000 to 2019. Figure 1A reveals a time-series plot of yearly leukemia cases and deaths for all racial/ethnic groups for those aged younger than 60 years and aged 60 years and older in this period. The number of leukemia cases grew from 2000 to 2019 with an estimated annual growth rate of 0.55% in adults aged ≥60 years and 0.1% in those younger than 60 years. The increasing trends of leukemia are associated with population growth in Florida over the years. In contrary, mortality rates by year demonstrated decreasing trends in both age groups, which can be attributed to improving leukemia treatment in the last 2 decades. Figure 1B reveals the same plot for yearly leukemia incidence rates and mortality rates, derived by dividing yearly numbers of leukemia cases and deaths with annual populations estimated by SEER based on US Census data for the same period in Florida. Because the growth rate in yearly leukemia cases in Florida is slower than the population growth rate, crude incidence rates by year demonstrate decreasing trends in both age groups.

Previous studies on US leukemia epidemiology found that patients from rural areas and/or areas with lower socioeconomic conditions tend to have worse survival rates than their counterparts.^12,32,33 To assess socioeconomic disparities associated with leukemia in Florida, we geocoded the leukemia cases to census tracts by patients’ places of residence and separated the cases by rural and urban counties designated by the Florida Department of Health.³⁴ To assess how access to leukemia treatment services varies by poverty levels and rurality, we also added a variable that measures the number of leukemia specialists that can be reached within 30 miles. Table 3 shows the summaries of leukemia cases, mortality odds, treatment accessibility, socioeconomic variables, and behavioral risk factors for the 3 racial/ethnicity groups by 3 levels of poverty in urban vs rural counties. Comparing urban vs rural areas, living in rural counties is associated with statistically significant higher mortality odds, regardless of race/ethnicity. In addition, rural regions also have significantly worse socioeconomic conditions and higher prevalence of behavioral risks, potentially contributing to higher mortality odds. Within the rural areas, all socioeconomic and behavioral risk factors deteriorate as poverty level goes up. Although mortality odds also reveal a worsening pattern as poverty levels go up in rural areas, ANOVA tests for mortality odds among poverty-level groups are not significant, due to higher mortality odds variance resulting from a small sample size (only 220 census tracts in all Florida rural counties). For urban areas, all ANOVA tests for mean comparison among the poverty groups are significant, suggesting that addressing disparities associated with socioeconomic and behavioral risk factors is important for eliminating leukemia mortality disparities in urban populations.

Table 4 shows results of linear regression analyses for which mortality odds by races/ethnicities served as the dependent variables, whereas all socioeconomic and behavioral risk factors included in Table 3 were entered into regression analyses as independent variables in combinations to test for their significance in the associations with mortality odds (NHW, Hispanic, and NHB). Only variables with significant t statistics were retained in the final equations. The NHW population is the largest racial group (56%) in Florida with substantial number of populations in both urban and rural counties. Median household income and number of leukemia doctors within 30 miles are negatively associated with NHW mortality odds. Observing mortality odds variation on a map (Figure 2) confirms that accessibility of leukemia treatment services is a factor for NHW mortality odds as clusters of census tracts in central Florida with sparse leukemia treatment service coverage were identified as statistically significant hot spot of NHW mortality odds. The NHB population also spreads between urban and rural counties in Florida. However, regression analysis reveals that leukemia treatment services within 30 miles is not a factor for NHB mortality odds, but poverty level is, suggesting that disparities associated with financial difficulty preclude the NHB population from acquiring necessary care services regardless of where they live. Behavioral risk factors in smoking and lack of physical activity also contribute to NHB mortality odds. The Hispanic population in Florida mostly resides in urban areas without issues of leukemia treatment service accessibility by locations. Regression analysis reveals that Hispanic mortality odds are positively associated with the percentages of population lacking physical activity and no health insurance and negatively with percentage of population with regular physician checkup, confirming the importance of addressing health disparities associated with socioeconomic conditions.

Figure 2 illustrates that hot spots of high leukemia mortality odds are generally located in areas of lower income levels throughout the state. However, leukemia specialists tend to be located in population centers where income levels are at or above average. In northern Florida, clusters of census tracts identified with high mortality odds for the NHB and NHW populations are in rural areas where lack of easy access to leukemia treatment providers may be attributable to impaired survivorship for the patients. In urban areas in central and southeastern Florida, hot spot clusters for the Hispanic and NHB populations are mostly in areas associated with lower income, which generally manifest in worse overall social determinants of health that can affect the availability and quality of supportive care needed for survivorship.

According to population projections from the US Census, by 2030 all baby boomers will be older than 65 years and 1 in 5 people in the United States will be of retirement age.³⁵ The number of annual cancer diagnoses in the United States is also projected to increase by 67% from 2010 to 2030 as the population becomes disproportionately older.³⁶ We demonstrated a strong association between the number of leukemia diagnoses in the population and aging in Florida, where retirees have migrated from other states for decades.¹⁸ As Florida is currently the second state in the United States (ie, behind Maine) for percentage of population aged ≥65 years³⁷ and the number 1 destination for retirement,¹⁸ we project that the number of leukemia cases in Florida will continue to increase in the coming decades, although leukemia incidence rate may decrease because the growth rate of older populations currently outpaces the rate of increasing leukemia diagnoses. We caution that the expected increase in adult leukemia cases will not be unique to Florida. Other US states expecting disproportionately large NHW populations aged ≥65 years in the coming decades may also experience an increasing count of leukemia cases. For example, together with Florida, California, Texas, New York, and Pennsylvania are the top 5 states in the United States for NHW population and total populations aged 40 to 64 years.⁶ The same 5 states are also currently the top 5 for annual count of leukemia diagnoses (2016-2020).⁶ We thus expect that demographic trends in the 5 states will lead to increasing leukemia cases in the coming decades. However, extrapolating these data to the entire United States would require more detailed analysis of aging and ethnic trends which may be of interest to future studies.

Although decreasing incidence rates generally reflect lower leukemia risk in the population, which may be attributable to reduced prevalence of smoking and other risk factors, increasing diagnoses place higher demand on health care services. The American Society of Clinical Oncology cautioned policy makers and the public in 2007 about the potential shortage in oncology services due to rising demand driven by an aging population and reducing supply caused by retiring providers who are baby boomers themselves.³⁸ A critical consequence of a shortage of oncology services is worsening disparities in cancer outcomes between socioeconomic classes and geographic regions as service shortages are more likely to leave the disadvantaged population without needed care for survival. In 2020, Shih et al³⁹ reported findings on the state of cancer care in the United States, based on their analyses of county-level oncology demand (ie, crude incidence rate of cancers) and supply (ie, densities of oncology workforce). They found that counties in the US Midwest and Central South had higher cancer incidence rates but lower densities of the oncology workforce. Their findings are consistent with what we found in Florida in that providers are typically located in urban areas and that rural residents are more likely to encounter unmet demand for oncology services. However, with geospatial analysis, we were able to pinpoint areas where oncology services are desperately needed to address survival disparities in adult leukemia. We recommend that other states or countries performing epidemiologic analyses use these geospatial methods to uncover potential disparities within their own geographic areas.

On the basis of our findings, we note that potential remedies on the supply side of oncology services recommended by the American Society of Clinical Oncology to address service shortages, such as increasing oncology fellowship slots, delaying retirement of current providers, and engaging nurse practitioners and physician assistants for oncology services,³⁸ may not fully address cancer survival disparities in rural and underserved areas due to the practical difficulties of recruiting and placing providers to those areas. Isaac et al⁴⁰ demonstrated that oncology services supplemented with telehealth approaches by an academic cancer center for patients with AML in rural Virginia can achieve results comparable to services accessible to urban patients. We recommend that similar approaches using telehealth be followed by the NCI-designated cancer centers and other treatment providers located closest to the rural and underserved areas in Florida to address socioeconomic and regional disparities associated with leukemia outcomes. In addition, we also observed that socioeconomic conditions and behavioral risks frequently associated with lower socioeconomic classes, such as lack of regular physician checkup, smoking, lack of physical activity, and obesity, play a bigger role in leukemia mortality for minority populations than geographical access to treatment services. Risk reduction and financial support for care services through community outreach and engagement approaches are effective strategies to address these disparities. We recommend academic cancer centers or local governments with community outreach and engagement programs follow the approaches of geospatial data analytics demonstrated in this paper to identify regions that are at high risk for increased leukemia mortality odds for effective intervention.

J.T. is supported by the National Institute of General Medical Sciences/National Institutes of Health (NIH; R35GM151109), the Doris Duke Charitable Foundation, the Edward P. Evans Foundation, and the National Cancer Institute Cancer Center Support Grant to Sylvester Comprehensive Cancer Center (P30CA240139).

The content of this work is solely the responsibility of the authors and does not necessarily represent the official views of the NIH or agencies that published data used for this study.

Contribution: M.S.L. and J.T. had full access to all data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis; M.S.L., R.E.K., N.S.E., E.N.K., and J.T. conceptualized and designed the study; M.S.L., R.E.K., C.D.A., D.B.D., and J.T. prepared the original draft of the manuscript; N.S.E., R.R.B., J.M.W., and M.A.S. reviewed and edited the manuscript; M.S.L., R.E.K., C.D.A., D.B.D., and J.T. were involved in data curation and investigation; and E.N.K. and J.T. provided supervision.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Justin Taylor, Medicine, University of Miami Sylvester Comprehensive Cancer Center, 1501 NW 10th Ave, Miami, FL 33136; email: [email protected].