Pediatric Pulmonology

Geographic Variations in Cystic Fibrosis: An Analysis of the U.S. CF Foundation Registry Benjamin T. Kopp, MD,1,2* Lisa Nicholson, PhD,3 Grace Paul, MD,1 Joseph Tobias, Chandar Ramanathan, MD,1 and Don Hayes Jr., MD, MS1

MD,

4

Summary. Background: Emerging evidence suggests that the prevalence of pathogens common in cystic fibrosis (CF) may be unevenly distributed across the United States (U.S.). However, very little is known regarding the U.S. distribution of other CF modifiers such as medication usage and patient demographics. Therefore, we sought to determine if regional differences exist in the distribution of demographic variables and patient disease characteristics in the U.S. that may play a role in differential CF outcomes. Methods: Data were analyzed from the 30,896 subjects in the U.S. CF Foundation Patient Registry during the years 2007–2012, via geographical grouping of states based upon the Nationwide Inpatient Sample classification. Results: Significant differences in racial distribution were seen, including half of the total U.S. African–American CF population residing in the South. Both African–Americans and Hispanics had increased Medicaid usage (52.2%, 41.8%, respectively). Culture-reported pathogens were markedly different across the U.S., with the highest percentage of patients with Methicillin-resistant Staphylococcus aureus (41.9%), Pseudomonas aeruginosa (71.2%), and non-tuberculous mycobacterium (10.0%) in the South. The South region also had the lowest mean body mass index and forced expiratory volume in one second. Chronic medication usage such as inhaled tobramycin or macrolides followed P. aeruginosa distribution, while inhaled dornase alfa was most used in the West (84.7%). Comorbid conditions varied, with the highest percentage of depressed subjects in the Midwest (18.3%). Mean regional mortality rates were not statistically different among regions, although highest in each age grouping of the South. Conclusions: The U.S. has significant regional variations in CF demographics, insurance, pathogens, medication usage, and co-morbidities, without an overall impact on regional mortality. Regional variations in care practices should be studied further based on the findings. Pediatr Pulmonol. ß 2015 Wiley Periodicals, Inc. Key words: epidemiology; insurance; MRSA; depression. Funding source: Nationwide Children’s Hospital.

1 Section of Pulmonary Medicine, Nationwide Children’s Hospital, Columbus, Ohio.

Conflict of interest: The authors have no conflicts of interest to disclose.

2

Portions of this manuscript were presented at 2014 North American Cystic Fibrosis Conference.

3

Drs. Kopp and Hayes are funded through the CF Foundation for laboratorybased work. The remaining authors have no relevant disclosures.

Center for Microbial Pathogenesis, The Research Institute at Nationwide Children’s Hospital, Columbus, Ohio. Institute for Population Research, The Ohio State University, Columbus, Ohio.


4

Department of Anesthesiology & Pain Medicine, Nationwide Children’s Hospital, Columbus, Ohio. Contributions: BK conceptualized and designed the study, analyzed data, drafted the initial manuscript, and approved the final manuscript. LN carried out initial analyses, drafted methods, reviewed and revised the manuscript, and approved the final manuscript. GP and JT participated in analysis, reviewed and revised the manuscript, and approved the final manuscript. CR and DH contributed to study design, analysis, revised the manuscript, and approved the final manuscript.

ß 2015 Wiley Periodicals, Inc.

Correspondence to: Benjamin T. Kopp, MD, Nationwide Children’s Hospital, Section of Pulmonary Medicine, Columbus, OH 43205. E-mail: [email protected] Received 23 October 2014; Revised 29 December 2014; Accepted 23 January 2015. DOI 10.1002/ppul.23185 Published online in Wiley Online Library (wileyonlinelibrary.com).

2

Kopp et al.

INTRODUCTION

The composition of patients with cystic fibrosis (CF) in the United States (U.S.) and their ongoing care regimens is rapidly changing due to therapeutic advancements, newborn screening, and quality improvement efforts.1 Over the last decade the average life expectancy of patients with CF in the U.S. increased into the early 40s.2 Due to the longer life expectancy, the prevalence of comorbid conditions in CF including CF-related diabetes (CFRD) and depression has directly increased, complicating treatment regimens. Additionally, diagnostic advancements for CF-causing mutations have led to the expanded identification of milder forms of CF which may require less aggressive treatment regimens, further complicating traditional treatment regimens based on more classic disease.3 In conjunction with the changing landscape of CF patients, changes in the CF microbiome are also occurring. The increasing prevalence of existing pathogens such as methicillin-resistant Staphylococcus aureus (MRSA) and non-tuberculous mycobacterium (NTM) along with emerging pathogens such as Stenotrophomonas maltophilia leave uncertainty to the efficacy of current preventative antibiotic regimens aimed at the long-standing dominant pathogen in CF, Pseudomonas aeruginosa. Interestingly, in the U.S. it was recently noted that the time to initial P. aeruginosa acquisition has a higher association of spatial clustering in Southern states, indicating that cases are occurring in proximity to each other suggesting geographic dependence.4 NTM were also found to have significant spatial clustering in the U.S., with centering in some Southern and Western states. The significance of these geographic clusterings on clinical outcomes as well as the possible clustering of other pathogens common to CF remains unknown. Other known U.S. geographic variations in CF include markedly different inpatient hospital expenditures for adult CF patients across the U.S. with inpatient stays in the Western U.S averaging 9.86 days and $107,194 compared to 7.39 days and $38,205 in the Midwest.5 It is uncertain if these findings relate specifically to CF, geographical pathogen distribution, or to regional insurance differences as the aforementioned study did not have data available on CF outcomes or variables. Environmental factors may also be influencing regional outcomes including air pollution6 and proximity to major roadways affecting the frequency of exacerbations,7 manifesting as variability in the distribution of co-morbidities and pathogen acquisition across the U.S. Many unanswered questions persist regarding factors influencing regional differences in hospitalizations in CF, including the distribution of insurance types across the U.S. and variances in the regional pediatric CF Pediatric Pulmonology

population which has not been well described to date. As important, the impact of variances related to CF demographics across the U.S. is not known, especially in how it influences health outcomes and mortality in CF. Other countries, including Canada, have not shown socioeconomic disparities in hospitalizations regionally in CF, which infers regional variability or availability of care in the U.S.8 Therefore, with this study we sought to analyze the Cystic Fibrosis Foundation Patient Registry (CFFPR) to determine if differences exist in the distribution of CF patient variables across the U.S. and their impact on mortality, and to identify areas for further study that may explain regional variations in CF. We hypothesized that each geographic region would have a unique subset of pathogens and co-morbidities more common in their geographic location. METHODS Study Design

This study was deemed exempt from institutional review. Data were obtained from the CFFPR from 2007 up to 2012 for a total of 5 years. The CFFPR contains information on nearly 31,000 people who receive care at CF Foundation-accredited care centers in the U.S.2 Patients were classified according to their state of residence. All U.S. patients in the registry with available state information were eligible for analysis including new births and deaths during the study (n ¼ 30,896). All years of available subject data were used, and subjects had to have at least one year of data prior to death for study inclusion. All subjects had a confirmed diagnosis of CF as defined by two disease-causing mutations or a sweat chloride test 60 mmol/L. Study variables were identified at an individual level and then grouped by subject state into geographic regions. Regional Grouping

Regional state grouping was accomplished as follows based upon the Nationwide Inpatient Sample (NIS)9: Northeastern states (Connecticut, Maine, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont), Midwestern states (Illinois, Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, North Dakota, Ohio, South Dakota, Wisconsin), Southern states (Alabama, Arkansas, Delaware, District of Columbia, Florida, Georgia, Kentucky, Louisiana, Maryland, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, Virginia, West Virginia), and Western states (Alaska, Arizona, California, Colorado, Hawaii, Idaho, Montana, Nevada, New Mexico, Oregon, Utah, Washington, Wyoming).

Geographic Variations in CF

Data Variables

Demographic variables included age, gender, race, insurance status, and body mass index (BMI) and forced expiratory volume in one second (FEV1) as markers of disease severity. For BMI and FEV1, an average value was calculated for each individual per year based on at least two recordings. All FEV1 values available for young children less than 6 were included, but U.S. centers are variable in their practice of obtaining lung function in young children under 6. For insurance status, any combination of more than one type of the following entities was classified as multiple insurances: private, no insurance, Medicare, Medicaid, and military-based. Due to several combinations, further sub-dividing of multiple carriers was not statistically feasible. Age groups were stratified to compare adult and pediatric subjects. Adults were further classified up to 44 years of age, and over 44 years of age to correspond with available data from our previous study of the NIS.5 Pediatric subjects were grouped every 6 years to correspond with sensitivity analyses for poor FEV1 data for children less than 6 and to take into account school/developmental groupings. Subjects were classified into their respective groupings based on the age at which they entered the study. Co-variates including infection status, co-morbid conditions, and medications were identified by study year, and an average prevalence over the study period calculated. Infection status comprised bacterial, fungal, and NTM respiratory cultures either obtained from oropharyngeal, sputum, or bronchoalveolar lavage culture depending on the age and availability of samples. Some centers do not routinely culture fungi. To be considered positive for the study, patients had to have more than one positive culture during the study period. Analyses were consistent using 2 or more positive cultures. Medications were reported as chronic if used more than two courses in 1 year. Co-morbid variables of interest including depression, asthma, allergic bronchopulmonary aspergillosis (ABPA), CFRD, and cirrhosis are physician diagnosed at the center level. Four parameters had missing or incomplete data listed as percent rate of the total subjects included: Hispanics (3.8%), Inhaled antibiotics (3.2%), supplemental feeding (1.0%), and diabetes status (0.07%). Other parameters had less than 0.05% data missing. Statistical Analysis

In all tables categorical measures are presented by region as N’s and percentages and continuous measures as means and standard deviations. BMI and FEV1 are reported as means based on distribution of data and no significant change with median measurements, despite some patients having more measurements of BMI and FEV1 per year than others. Measured outcomes are reported as study period prevalences. x2 tests were

3

conducted to determine differences across all four regions, and between any two given regions, respectively. A P-value 44 years, 6.0%). In contrast, the South had a significantly lower proportion of adult subjects (18–44 years, 37.4%), (>44 years, 3.8%) and higher adolescent patients compared to the other regions (10–17 years, 24.6%). Racial subgrouping also differed significantly across the U.S. (P < 0.0001, Table 1). Caucasians constituted over 90% of subjects from all regions, but were heavily concentrated in the Midwest (94.6%). The South had a significantly higher proportion of African–Americans with CF (5.4%), and accounted for half of the total U.S. African–American CF population. Asian–Americans were more heavily concentrated in the West and Northeast (0.6% each). Hispanics were grouped under white and Pediatric Pulmonology

4

Kopp et al.

TABLE 1— U.S. CF Patient Demographics and Characteristics Via Regional Comparators From 2007–2012 Regional Comparison Statistics (Tukey)

Total Gender Male # (%) Age (years) 0–44 Race # (%) White Black A Indian Asian Other Mixed Insurance # (%) Private None Medicaid Medicare Military Other Multiple BMI1 (st dev) FEV12 % predicted

Total

West (1)

Midwest (2)

Northeast (3)

30,896 16035 (51.9)

6062 3074 (50.7)

8079 4203 (52.0)

6336 3363 (53.1)

10,419 5395 (51.8)

6648 3428 7121 12184 1515

(21.5) (11.1) (23.0) (39.4) (4.9)

1306 684 1324 2390 358

(21.5) (11.3) (21.8) (39.4) (5.9)

1800 873 1800 3223 383

(22.3) (10.8) (22.3) (39.9) (4.7)

1202 644 1437 2674 379

(19) (10.2) (22.7) (42.2) (6.0)

2340 1227 2560 3897 395

(22.5) (11.8) (24.6) (37.4) (3.8)

28680 1119 76 115 489 417

(92.8) (3.6) (0.2) (0.4) (1.6) (1.3)

5567 (91.8) 144(2.4) 29 (0.5) 37 (0.6) 178 (2.9) 107 (1.8)

7640 220 19 17 71 112

(94.6) (2.7) (0.2) (0.2) (0.9) (1.4)

5919 196 6 35 124 56

(93.4) (3.1) (0.1) (0.6) (2) (0.9)

9554 559 22 26 116 142

(91.7) (5.4) (0.2) (0.2) (1.1) (1.4)

14451 370 8031 463 398 232 6951 22.3 74.0

(46.8) (1.2) (26.0) (1.5) (1.3) (0.8) (22.5) (4.0) (26.0)

2890 77 1514 80 107 144 1250 22.2 73.9

3709 79 2029 96 41 26 2099 22.5 75.1

(45.9) (1.0) (25.1) (1.2) (0.5) (0.3) (26.0) (4.1) (26.0)

3306 64 1358 105 32 13 1458 22.5 74.8

(52.2) (1.0) (21.4) (1.7) (0.5) (0.2) (23.0) (3.9) (26.2)

4546 150 3130 182 218 49 2144 22.0 72.7

(43.6) (1.4) (30.0) (1.7) (2.1) (0.5) (20.6) (3.9) (25.6)

(47.7) (1.3) (25) (1.3) (1.4) (2.8) (20.6) (3.8) (26.3)

South(4)

1 vs 2

1 vs 3

1 vs 4

2 vs 3

2 vs 4

3 vs 4

* * *

* * * * *

* *

* *

* * *

* * *

* * * * * *

* * *

* * * * *

* * *

* * * *

* * *

*

*

*

*

*

*

* * *

* * * *

* *

* * * * *

*

*

* * *

* * *

*

“*”P < 0.05, statistically significant via Tukey comparison. 1 Body Mass Index (BMI), mean  standard deviation. 2 Forced Expiratory Volume in 1 second (FEV1), mean  standard deviation.

black categories in Table 1 based on self-reporting classification in the registry and missing reporting. However, of the 2118 total reported Hispanics (6.9%), the majority (n ¼ 782) were in the West (12.9% total west population) and South (n ¼ 770, 7.6%). Primary insurance usage varied greatly throughout the regions (P < 0.0001). Patients in the Northeast had a much higher carriage of private insurance only (52.2%) compared to other regions, contrasted with a significantly lower proportion of private insurance in the South (43.6%). Medicaid only usage correlated inversely with private insurance, with the highest Medicaid carriage in the South (30%), and lowest in the Northeast (21.4%). Over 20% of patients reported using more than one type of insurance, with the highest concentration in the Midwest (26.0%) and Northeast (23.0%). Overall, only 1.2% of total patients were reported to have no insurance, and no insurance was reported similarly between regions except for a higher percentage in the South (1.4%) compared to the Midwest (1.0%). Adults represented 76% of the uninsured population with approximately equal distributions in the 18–44 year group and older than 44 years. When ethnicity was compared against insurance status, a significant association was seen for Medicaid use and certain ethnic groups, with 52.2% of African–Americans in the South on Medicaid (P ¼ 0.001, no evidence of multi-collinearity), and increased Medicaid use for Pediatric Pulmonology

Hispanic–Americans in the West (41.8%) (P < 0.0001). There was no difference between these two groups in overall association. Nutritional status as reflected by mean BMI was significantly lower (22.0) for patients in the South compared to all other regions (22.2–22.5). BMI percentiles were similarly lower in the South (not shown). The mean percent predicted FEV1 as a measure of lung function was also significantly lower in the South (72.7%) compared to all other regions (73.9–75.1%). BMI and FEV1 were not affected by regional race, insurance variations, age, or other potential confounders. Microbial Prevalence

A summary of cumulative infectious pathogen growth by region is presented in Table 2. Pseudomonas aeruginosa growth was significantly higher in the South (71.4%) compared to all other regions, and also higher in the Midwest (68.9%) compared to the Northeast and West. Patients with at least one Methicillin-sensitive Staphylococcus aureus (MSSA) respiratory culture were significantly lower in the South (65.9%) compared to all other regions. However, subjects with positive MRSA respiratory cultures were significantly higher in the South (41.9%). Notably, the West only had 25.7% of patients with positive MRSA, significantly lower than other

Geographic Variations in CF

5

TABLE 2— Summary of Microbial Prevalence Via Regional Comparison For U.S. CF Patients Regional comparison statistics (Tukey) Total 1

MRSA # (%) MSSA2 # (%) NTM3 # (%) P. aeruginosa # (%) Burkholderia spp # (%) B. cenocepacia # (%) B. multivorans # (%) S. maltophilia # (%) A. baumanii # (%) A. fumigatus # (%) C. albicans # (%)

10829 21610 2512 21218 1491 150 244 8515 747 8463 7544

(35) (69.9) (8.1) (68.7) (4.8) (0.49) (0.8) (27.6) (2.4) (27.4) (24.4)

West (1)

Midwest (2)

1559 4337 583 4022 218 20 52 1565 138 1738 1,968

2927 5774 493 5566 331 36 57 2337 204 2351 1989

(25.7) (71.5) (9.6) (66.3) (3.6) (0.3) (0.9) (25.8) (2.3) (28.7) (32.5)

(36.2) (71.5) (6.1) (68.9) (4.1) (0.45) (0.7) (28.9) (2.5) (29.1) (24.6)

Northeast (3) 1982 4628 399 4191 375 52 71 1960 179 1618 1235

(31.3) (73) (6.3) (66.1) (5.9) (0.8) (1.1) (30.9) (2.8) (25.5) (19.5)

South (4) 4361 6871 1037 7439 567 42 64 2653 226 2756 2352

(41.9) (65.9) (10) (71.4) (5.4) (0.4) (0.6) (25.5) (2.2) (26.5) (22.6)

1 vs. 2

1 vs. 3

1 vs. 4

2 vs. 3

2 vs. 4

3 vs. 4

*

* * *

*

* *

*

* *

* * * * *

* * * * *

* *

* * *

*

*

* *

* *

* *

* * *

* * * * * * * *

“*”P < 0.05, statistically significant via Tukey comparison. 1 Methicillin resistant Staphylococcus aureus. 2 Methicillin sensitive Staphylococcus aureus. 3 Non-tuberculous mycobacterium.

regions. The presence of at least one NTM positive culture was highest in the West (9.6%) and South (10.0%). There was no significance difference in observed MSSA, MRSA, or P. aeruginosa acquisition for African– Americans or Hispanic–Americans from total regional population averages. Burkholderia species as a complex were significantly higher in the Northeast (5.9%) and South (5.4%), and lowest in the West (3.6%). Stenotrophomonas maltophilia was also most predominant in the Northeast (30.9%). In contrast, Fungal organisms Aspergillus fumigatus and Candida albicans were highest in the Midwest (29.1%, 32.5%) and West (28.7%, 24.6%).

Co-Morbid Conditions

The frequency of registry reported co-morbid conditions by geographic region are reported in Table 3. Asthma was significantly higher in the Northeast (30.8%) and South (30.6%) compared to the West (24.1%) and Midwest (24.3%). ABPA was not distributed similarly to asthma, with the only significant difference between the West (9.9%) and South (8.1%). This is also in contrast to the incidence of A. fumigatus positive cultures as previously noted. Patients with reported depression were significantly higher in the Midwest (18.3%). CF-related Diabetes (CFRD) was similar among regions.

TABLE 3— CF Patient Co-Morbidities and Medications Regional Distribution Regional comparison statistics (Tukey) Total Co-morbidities Asthma # (%) ABPA1 # (%) Cirrhosis # (%) CFRD2 # (%) Depression # (%) TOBI3 # (%) Therapeutics Aztreonam lysine # (%) Colistin # (%) Macrolides # (%) Dornase alfa # (%) 7% Saline # (%) Supplemental feeding # (%)

West (1)

Midwest (2)

Northeast (3)

South (4)

8562 1806 962 7418 5216 18,722

(27.7) (8.9) (3.1) (24.0) (16.9) (62.6)

1461 383 161 1435 984 3,568

(24.1) (9.9) (2.7) (23.7) (16.2) (61.1)

1962 464 262 2009 1477 4677

(24.3) (8.6) (3.2) (24.9) (18.3) (60.0)

1951 378 224 1457 1082 2768

(30.8) (9.8) (3.5) (23.0) (17.1) (61.4)

3188 581 315 2517 1673 6709

(30.6) (8.1) (3.0) (24.2) (16.1) (66.2)

5625 4100 16,351 24,275 16,850 21,323

(19.4) (13.7) (54.7) (81.2) (56.3) (69.7)

1,186 647 3,170 4,948 3,644 4,270

(20.9) (11.1) (54.3) (84.7) (62.4) (70.9)

1302 1254 4132 6155 3885 5502

(17.1) (16.1) (53.0) (79.0) (49.9) (69.0)

1200 882 3320 4771 3444 4177

(20.2) (14.4) (54.1) (77.7) (56.1) (66.6)

1937 1371 5729 8401 5877 7374

(19.8) (13.0) (56.5) (82.9) (57.8) (71.6)

1 vs. 2

1 vs. 3

1 vs. 4

2 vs. 3

2 vs. 4

*

* *

*

*

3 vs. 4

* * *

* *

* * * * *

* * * *

* * * *

* * * * *

* * * * *

*

* * *

“*”P < 0.05, statistically significant via Tukey comparison. 1 Allergic bronchopulmonary aspergillosis. 2 CF-related diabetes. 3 Tobramycin inhalation solution.

Pediatric Pulmonology

6

Kopp et al.

Chronic Medication Usage

Chronic medication usage by geographic region is reported in Table 3. In accordance with the prevalence of reported Pseudomonas cultures, chronic aerosolized Tobramycin usage (66.2%) and chronic macrolide therapy (56.5%) were significantly higher in the South compared to other regions. However, inhaled aztreonam lysine was similar among regions with the exception of a lower usage in the Midwest (17.1%). Notably, aztreonam lysine use markedly increased from 1.3% in 2008 to 19.9% in 2011 in concordance with its availability in the U.S. Inhaled colistin use was significantly higher in the Midwest (16.1%) and lowest in the West (11.1%). Dornase alfa use was significantly higher in the West (84.7%), and lowest in the Northeast (77.7%). Inhaled hypertonic saline was also reported significantly higher in the West (62.4%), and lowest in the Midwest (49.9%). Mortality

Age-adjusted regional mortality rates are presented in Table 4, and individual state age-adjusted mortality rates are presented in Supplementary Table 1. The South had the highest total age-adjusted mortality rates for all 5 age groupings, and the Northeast had the lowest rates for all groups except children less than 5 years old, but there were no statistically significant differences in age adjusted rates between regions based on the low amount

of overall deaths in the study period (Table 4, ANOVA Pvalue 0.77). However, it is noted that the summed age adjusted rate for the South is more than twice that of the Northeast (4563.18 vs. 1762.28, Table 4). We examined differences between each pair of regions using poisson regression modeling and find that the Northeast and South are significantly different (P ¼ 0.001). However, the combined effect of region and age group explains only slightly more than 1% of the variation in mortality, therefore region is not a significant predictor of mortality. Hawaii and Wyoming did not have any deaths recorded during the study period. For patients aged 18–44 years, age-adjusted mortality rates were greater than 250/ 100,000 persons for California, Florida, and Texas. For patients greater than 44 years, age-adjusted mortality rates were greater than 250/100,000 persons for California, Florida, Georgia, Ilinois, Indiana, Michigan, New York, Ohio, Pennsylvania, and Texas. There were no significant demographic or clinical variables associated with regional mortality rates. DISCUSSION

Geographical disparities distinctly exist with wide variations in insurance for chronic diseases.11,12 Although environment influences pathogenic colonization in CF, for example P. aeruginosa,13 very little is known about geographical influences on other important variables

TABLE 4— Regional Age-Adjusted Mortality Rates for U.S. CF Patients1

Region West West West West West Midwest Midwest Midwest Midwest Midwest Northeast Northeast Northeast Northeast Northeast South South South South South

Age at death child child child adult adult child child child adult adult child child child adult adult child child child adult adult

ages 0 to