Abstract
There is considerable variability in the clinical course of disease in cystic fibrosis (CF). Although currently unidentified modifier genes might explain some of this heterogeneity, other factors are probably contributory. Socioeconomic status (SES) is an important predictor of health status in many chronic polygenic diseases, but its role in CF has not been systematically evaluated. We performed a historical cohort analysis of pediatric CF patients in the United States using National Cystic Fibrosis Foundation Patient Registry (NCFPR) data for 1986 to 1994, and used Medicaid status as a proxy for low SES. The adjusted risk of death was 3.65 times higher (95% confidence interval [CI]: 3.03 to 4.40) for Medicaid patients than for those not receiving Medicaid. The percent predicted FEV1 of surviving Medicaid patients was less by 9.1% (95% CI: 6.9 to 11.2). Medicaid patients were 2.19 times more likely to be below the 5th percentile for weight (95% CI: 1.91 to 2.51) and 2.22 times more likely to be below the 5th percentile for height (95% CI: 1.95 to 2.52) than were non-Medicaid patients. Medicaid patients were 1.60 times more likely to require treatment for a pulmonary exacerbation (95% CI: 1.29 to 1.98). There was no difference in the number of outpatient clinic visits for Medicaid and non-Medicaid patients. We conclude that low SES is associated with significantly poorer outcomes in children with CF. Barriers in access to specialty health care do not seem to explain this difference. Further study is indicated to determine what adverse environmental factors might cluster in CF patients of low SES to cause worse outcomes.
Recent research has established that the pathophysiologic basis of cystic fibrosis (CF) lies in different gene mutations that encode abnormalities in the structure of the cystic fibrosis transmembrane conductance regulator (CFTR) protein, leading to physiologic defects in ion transport (1). The course of CF is quite heterogeneous, and much of the variability in mortality and morbidity is unexplained by current conceptualization of the disease. Although certain genotypes are clearly associated with normal pancreatic function, genotype is a poor predictor of pulmonary disease and eventual patient outcome (2) in the general population of CF patients. For unclear reasons, female sex is associated with a younger median age at death (3). Socioeconomic status (SES) is an important determinant of outcome in many diseases (4, 5) and may also play a role in CF. Decreased survival has been found in the United Kingdom among CF patients of lower social class (6), and in the United States, reports from individual CF clinics have suggested that low SES and lack of health insurance are associated with poorer outcomes (7, 8). We report here the results of a study designed to evaluate longitudinally the impact of low SES on CF mortality and morbidity by evaluating a large cohort of patients in the United States.
The National Cystic Fibrosis Foundation Patient Registry (NCFPR) includes all patients seen at Cystic Fibrosis Foundation-accredited care centers in the United States. The registry is supervised and maintained by Cystic Fibrosis Foundation personnel and has been updated annually since 1966. In 1994, it was estimated to include 83% of all patients with diagnosed CF in the United States and over 90% of CF deaths reported in the Vital Statistics of the United States (9). The Foundation distributes annually a patient data collection questionnaire that is completed by clinic personnel at each care center. The registry collects basic demographic data, diagnostic information, clinical data (including spirometry, microbiology, complications, and treatments), and health care utilization data. The data collection form has been repeatedly modified and refined through the years, and in 1994, it consisted of 46 questions.
Our historical cohort study used registry data for the years 1986 to 1994. Study subjects consisted of all patients under 20 yr of age who had had at least one measurement in the NCFPR during those years. Patients who had received an organ transplant were excluded. The factor of interest was low SES, but since the CF registry questionnaire contains no items specifically addressing SES, we used information on health insurance coverage to impute SES. Medicaid insurance status was used as a proxy for low SES (10). Minimum Medicaid eligibility criteria are determined by the U.S. government, and although individual states may lower the requirements, they are always tied to family income and its relationship to the poverty level established by the U.S. government (11). Patients were included in the Medicaid group if they were covered by Medicaid for every year in which they were listed in the registry during the study period, even if they reported an additional source of health insurance. This group was compared with patients who never received Medicaid during the same period. In order to verify the validity of our approach, we performed a parallel analysis in which we evaluated the effect of Medicaid status in the year in which CF was diagnosed; the results were similar to those we report here.
Since the uninsured might form another high-risk group, we compared, in a separate analysis, patients without health insurance for varying periods with those who were never without coverage (commercial, Civilian Health and Medical Program of the Uniformed Services [CHAMPUS], or Medicaid).
Because poverty might have a different effect on the health of urban and rural dwellers, we obtained information about the urbanization of patients' residence from 1990 U.S. census data according to zip code (12). We evaluated this both as an interval-scale variable (% of population living in a rural area) and as a categorical variable (more than 50% rural population).
The study had two components. The mortality analysis compared time to death from any cause. The starting point was 1986 or the year of first entry into the registry, and the endpoint was death, survival to age 20 years, or the end-study year of 1994, whichever came first. The morbidity analysis compared pulmonary function, height and weight, and the need for intravenous antibiotic therapy for pulmonary exacerbations of CF among survivors in 1994. We also evaluated factors that might contribute to morbidity and mortality, including health care utilization, acquisition of pathogenic bacterial flora, and age at diagnosis.
Data reported in the 1994 NCFPR for all nontransplant–recipient patients under 20 yr of age were used to evaluate pulmonary function, height and weight, and health care utilization. The average of four quarterly measurements of FEV1 and FVC (available from the NCFPR for the first time in 1994) was used as an index of pulmonary disease (13). Average FEV1 and FVC were used because episodic pulmonary exacerbations experienced by CF patients are associated with reversible declines in pulmonary function (14), and any single measurement may not be representative of a patient's true status. Percent predicted values for FEV1 and FVC are calculated by the National Cystic Fibrosis Foundation through use of a modification of the Knudson (15) equations that removes the lower boundary for height for all six age- and sex-specific groups (9). For African–Americans, the Hsu (16) equations are utilized. Our analysis of pulmonary function was limited to patients 5 yr of age or older because of lower age limitations in the normative equations.
We used the average of the four reported height and weight percentiles recorded in the NCFPR in 1994 as our indicator of growth. The National Cystic Fibrosis Foundation calculates percentiles from normalized growth reference curves that are produced by the National Center for Health Statistics of the Centers for Disease Control (9).
The periodic exacerbations of pulmonary infection experienced by CF patients often require hospitalization and intravenous antibiotics (14), and significantly affect these patients' quality of life. We evaluated the need for intravenous antibiotic treatment as a secondary indicator of disease morbidity.
Age, race, and genotype are factors that could affect morbidity and mortality, and so were initially considered as potential confounders in all of our statistical models. We classified genotyped patients into three groups, base on the presence of 0, 1, or 2 alleles having the ΔF508 mutation. Genotype was eventually dropped from all models because its inclusion did not alter the estimate of Medicaid effect, but caused instability of the model because of the number of patients for whom this information was unavailable in 1994. Pancreatic sufficiency is associated with mutations causing milder CF lung disease and reduced mortality (17) and was therefore used as a proxy for genotype. Patients were categorized as either pancreatic-insufficient or -sufficient depending upon whether or not they were reported to use pancreatic enzyme replacement. Use of this variable as a proxy for genotype resulted in no change in point estimate but gave better precision because it allowed inclusion of ungenotyped patients.
Initial bivariate analysis of normally distributed continuous variables was done with Student's two-sample t test (two sided), and Wilcoxon's rank-sum test was used for variables that were not normally distributed. Fisher exact test was used to evaluate categorical variables.
For the mortality analysis, we used Kaplan–Meier estimation of the proportion of subjects surviving at any point during follow-up, and the log-rank statistic to assess differences between the survival curves. A Cox proportional hazards regression was initially fitted, using Medicaid status and the potential confounders mentioned previously, but the proportional hazards assumption for the covariates was found to be invalid. In light of this, a stratified (also referred to as an extended) Cox model (18) was used to assess the effect of Medicaid status, with adjustment for age, race, sex, pancreatic status, and year of entry into the registry. The SAS Proc Phreg program was used to perform this part of the analysis (SAS Institute Inc., Cary, NC).
For the morbidity analysis, we used linear regression analysis of 1994 patient data to evaluate differences between interval-scaled outcome measures, and we used logistic regression to calculate odds ratios for categorical outcomes (hospitalization, treatment of pulmonary exacerbations, acquisition of bacterial organisms). Because utilization of medical care is affected by severity of illness, models for evaluating outpatient visits and treatment for pulmonary exacerbations included FEV1 to control for degree of pulmonary dysfunction. Because the distribution of age at diagnosis was skewed severely to the right, the median was modeled through quantile regression.
Regression models were constructed using forward stepwise regression, permitting reexamination at every step of the variables incorporated in the model in previous steps. Variables were chosen for testing on the basis of prior judgment of their probable relevance. Interaction terms containing the Medicaid variable were also evaluated, and terms whose inclusion produced a difference significant at the p = 0.10 level were preserved in the model. Logistic models were evaluated with Pearson's chi-square goodness-of-fit test. All data points beyond 3 SD from the mean for included variables were examined individually and were dropped if they were clearly erroneous.
A total of 25,626 patients had at least one entry in the NCFPR between 1986 and 1994, of whom 20,390 were under the age of 20 yr, had all insurance data recorded, and had never received an organ transplant. Median follow-up of the group was 8 yr. Of the 20,390 patients in the eligible group, 1,894 (9.3%) received Medicaid during every year for which they were listed in the registry, and 13,476 (66.1%) were never covered by Medicaid. Patients in the Medicaid group were younger and more likely to be African-American (Table 1).
| Mortality Analysis: 1986–1994 | Morbidity Analysis: 1994 | |||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| All | Never on Medicaid | Always on Medicaid | All | Never on Medicaid | Always on Medicaid | |||||||||||||||||||
| Total | 20,390 | 13,476 | 1,894 | 13,804 | 8,454 | 1,535 | ||||||||||||||||||
| Race* | ||||||||||||||||||||||||
| African–American | 719 | 4% | 243 | 2% | 193 | 10% | 521 | 4% | 148 | 2% | 158 | 10% | ||||||||||||
| White, other | 19,671 | 96% | 13,233 | 98% | 1,701 | 90% | 13,283 | 96% | 8,306 | 98% | 1,377 | 90% | ||||||||||||
| Genotype† | ||||||||||||||||||||||||
| ΔF508 homozygote | 3,437 | 51% | 2,345 | 51% | 352 | 50% | 2,789 | 52% | 1,817 | 52% | 322 | 51% | ||||||||||||
| ΔF508 heterozygote | 2,412 | 36% | 1,671 | 37% | 242 | 35% | 1,910 | 36% | 1,274 | 37% | 222 | 35% | ||||||||||||
| No ΔF508 | 884 | 13% | 554 | 12% | 106 | 15% | 673 | 13% | 392 | 11% | 93 | 15% | ||||||||||||
| Total genotyped | 6,733 | 100% | 4,570 | 100% | 700 | 100% | 5,372 | 100% | 3,483 | 100% | 637 | 100% | ||||||||||||
| Not genotyped‡ | 13,657 | 67% | 8,906 | 66% | 1,194 | 63% | 8,432 | 61% | 4,971 | 59% | 898 | 59% | ||||||||||||
| Pancreatic enzyme use* | 19,432 | 94% | 12,711 | 93% | 1,864 | 98% | 12,673 | 92% | 7,660 | 91% | 1,490 | 97% | ||||||||||||
| Female† | 9,593 | 47% | 6,374 | 47% | 894 | 47% | 6,507 | 47% | 4,002 | 47% | 720 | 47% | ||||||||||||
| Age, yr (mean ± SD)§ | 11.6 ± 6.0 | 12.1 ± 6.0 | 7.3 ± 6.1 | 9.7 ± 5.3 | 10.1 ± 5.3 | 5.7 ± 4.7 | ||||||||||||||||||
| Age if spirometry performed, yr§ | 12.4 ± 3.9 | 12.5 ± 3.9 | 11.3 ± 3.9 | |||||||||||||||||||||
Between 1986 and 1994, 151 (8.0%) Medicaid and 808 (6.0%) non-Medicaid patients died (Table 2). Figure 1 shows the Kaplan–Meier survival curves for the two groups, which diverged significantly (log-rank p < 0.0001). When stratified by age, this relationship was not significant for children aged 0 to 5 yr (p = 0.3113), but was similarly divergent for those in the 5 to 10-, 10 to 15-, and 15 to 20-yr age groups (p < 0.0001). The relative risk (RR) of death for Medicaid patients during the study period was 2.02 (95% confidence interval [CI]: 1.69 to 2.41) by Cox regression. When adjusted for sex, race, age, and pancreatic enzyme use, the RR was 3.65 (95% CI: 3.03 to 4.40). The increased mortality appeared to be explained by differences in baseline pulmonary function, because when FEV1 at year of entry was included in the model, the increased RR of death associated with Medicaid status ceased to be statistically significant. The mortality of patients who were uninsured for 25% of the study period did not differ from that of fully insured patients (adjusted RR = 0.97; 95% CI: 0.67 to 1.43). Varying the definition of “no insurance” to any level between 10% and 90% of the years for which the patient was listed in the registry still revealed no increased risk of death for the uninsured group.
| Unadjusted Difference | Adjusted Difference | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Primary Outcomes | Medicaid | Non-Medicaid | Point Estimate | 95% Confidence Interval | p Value | |||||||
| Mortality | 8.0% | 6.0% | RR = 2.02 | RR = 3.70* | (3.06, 4.46) | < 0.001 | ||||||
| Pulmonary function | ||||||||||||
| FEV1 (% pred) | 78.1% | 84.8% | 6.7 | 9.2* | (7.1, 11.4) | < 0.001 | ||||||
| FVC (% pred) | 96.0% | 94.3% | 4.7 | 6.6* | (4.6, 8.5) | < 0.001 | ||||||
| Growth | ||||||||||||
| Weight percentile | 28.6%ile | 34.3%ile | 5.7 | 7.8* | (6.3, 9.4) | < 0.001 | ||||||
| Weight ⩽ 5%ile | 29.1% | 16.7% | OR = 2.04 | OR = 2.19* | (1.91, 2.51) | < 0.001 | ||||||
| Height percentile | 25.0%ile | 33.9%ile | 8.9 | 8.9* | (7.4, 10.5) | < 0.001 | ||||||
| Height ⩽ 5%ile | 33.4% | 17.3% | OR = 2.39 | OR = 2.11* | (1.86, 2.41) | < 0.001 | ||||||
| Treatment of pulmonary exacerbations | ||||||||||||
| Any hospitalization | 43.4% | 25.9% | OR = 2.20 | OR = 1.64† | (1.33, 2.04) | < 0.001 | ||||||
| Any home IV antibiotics | 5.3% | 10.7% | OR = 0.47 | OR = 0.46† | (0.32, 0.65) | < 0.001 | ||||||
| Any exacerbation | 44.5% | 28.6% | OR = 2.00 | OR = 1.58† | (1.27, 1.96) | < 0.001 | ||||||
| Secondary outcomes | ||||||||||||
| Median Age at Diagnosis (days) | 131 | 157 | 26 | −11* | (−5, 28) | 0.173 | ||||||
| Number of CF Clinic Visits | 4.5 | 4.2 | −0.3 | 0.13† | (0.40, −0.14) | 0.398 | ||||||
Fig. 1. Kaplan–Meier estimate of the proportion of Medicaid and non-Medicaid patients remaining alive, adjusted for age, race, and sex. Between 1986 and 1994, there were 151 deaths among 1,894 Medicaid patients and 808 deaths among 13,476 patients who never received Medicaid. Stratified Cox proportional hazards regression gave a RR for dying of 3.70 (95% CI: 3.06 to 4.46). The Kaplan–Meier curves are significantly divergent (log-rank p < 0.0001). When stratified by age, the Kaplan– Meier curves were similarly divergent, with p < 0.0001 for the age groups of 5 to 10 yr, 10 to 15 yr, and 15 to 20 yr, but were not significant for the age group of 0 to 5 yr.
[More] [Minimize]A total of 19,517 patients were reported in the 1994 registry, of whom 13,804 were under the age of 20 yr and had not received an organ transplant. Of this group, 1,535 (11.1%) had received Medicaid for every year in which they were listed in the registry during the study period, and 8,454 (61.3%) had never received Medicaid. Patients in the Medicaid group were younger, primarily owing to a larger number of patients under the age of 5 yr (Table 1). Patients in the Medicaid group were also more likely to be African–American and pancreatic-insufficient.
The mean FEV1 of Medicaid patients was 78.1% predicted, as compared with 84.8% predicted for non-Medicaid patients. When adjusted for age, pancreatic status, sex, and race, the average FEV1 of Medicaid patients was less by 9.2% predicted (95% CI: 7.1 to 11.4; p < 0.001) than that of non-Medicaid patients (Table 2). The difference between the groups was slightly age-dependent, increasing by 0.54% per year of age (95% CI: −0.02 to 1.06; p = 0.06) (Figure 2). Table 3 shows differences in FEV1 associated with other variables in the regression model. Other significant predictors of pulmonary function were age, sex, and pancreatic enzyme use. Rural residence, lack of health insurance, and African–American race were not independent predictors of FEV1, but the effect of race approached statistical significance. As shown in Table 2, results for FVC were similar to those for FEV1. The average adjusted difference in FVC between Medicaid and non-Medicaid patients was 6.5% predicted, and this difference also varied with age, increasing at a rate of 0.52% predicted per year (95% CI: 0.03 to 1.0; p < 0.05).
Fig. 2. Relationship of % predicted FEV1 to age for Medicaid and non-Medicaid patients. The plot shows the adjusted mean FEV1 by age for Medicaid and non-Medicaid patients listed in the NCFPR in 1994. Shaded areas represent 95% confidence intervals (CI). The difference between the groups, adjusted for age, race, sex, and ΔF508 genotype, was 9.2 (95% CI: 7.1 to 11.4). The difference in FEV % predicted increased by 0.52 (95% CI: −0.02 to 1.06) per year of age. A similar relationship existed for FVC.
[More] [Minimize]| Adjusted Difference in % Predicted FEV1 | 95% CI | p Value | ||||
|---|---|---|---|---|---|---|
| Medicaid recipient | −9.1% | (−11.2, −6.9) | < 0.001 | |||
| African–American | −3.5% | (−7.5, 0.4) | 0.08 | |||
| Genotype (compared with ΔF508 homozygote) | ||||||
| ΔF508 heterozygote | 1.1% | (−0.9, 3.1) | 0.30 | |||
| No ΔF508 | 2.1% | (−1.1, 5.2) | 0.20 | |||
| Age (per year) | −2.4% | (−2.5, −2.3) | < 0.001 | |||
| Nonuse of pancreatic enzymes | 6.1% | (3.4, 8.9) | < 0.001 | |||
| Female | −1.2% | (0.0, 2.3) | 0.04 | |||
| Rural residence | 0.6% | (−1.0, 2.2) | 0.49 | |||
| No health insurance | −0.3% | (−4.7, 4.1) | 0.88 |
Medicaid patients were more likely than non-Medicaid patients to have positive cultures for Pseudomonas aeruginosa and Burkholderia cepacia, but when adjusted for FEV1 there was no statistical difference between the two groups (data not shown).
Lack of health insurance was not associated with a significant difference in pulmonary function (data not shown).
CF patients who were Medicaid recipients were smaller than those who were not. Medicaid patients had a mean weight at the 28.6th percentile (median: 18th percentile), as compared with 34.3rd percentile (median: 28th percentile) for the non-Medicaid group. The mean difference in percentile was 7.8 (95% CI: 6.3 to 9.5; p < 0.001) when adjusted for age, race, sex, and pancreatic enzyme use. As illustrated in Figure 3, the most striking difference between the groups was that 29.1% of the Medicaid patients were at or below the 5th percentile for weight, as compared with 16.7% of the non-Medicaid patients, giving an adjusted OR for this measure of 2.31 (95% CI: 2.02 to 2.51; p < 0.001). There was no linear relationship between age and the odds of being at or below the 5th percentile; the prevalence of underweight was highest at both extremes of age.
Fig. 3. Height and weight percentiles for Medicaid and non-Medicaid patients. No linear relationship existed between height or weight percentiles and age, in contrast to the case with pulmonary function. The overall distribution was shifted to the left in Medicaid patients, and the most striking difference was the likelihood of being at or below the 5th percentile for weight (OR = 2.19; 95% CI: 1.91 to 2.51) and height (OR = 2.11; 95% CI: 1.86 to 2.41).
[More] [Minimize]Similar relationships were seen with height. The mean height percentile was 25.0 for Medicaid patients (median: 15th percentile), and was 33.9 (median: 27th percentile) for non-Medicaid patients. When adjusted for age, race, sex, and pancreatic enzyme use, the difference in percentiles was 9.1 (95% CI: 7.6 to 10.7; p < 0.001). Medicaid patients were 2.22 times more likely than non-Medicaid patients to be at or below the 5th percentile for height (95% CI: 1.95 to 2.52; p < 0.001), and the odds did not vary with age.
Uninsured patients did not show any deficit in height or weight as compared with the rest of the study population.
As shown in Table 2, 44.5% of Medicaid patients received intravenous antibiotic treatment for a pulmonary exacerbation in 1994, as compared with 28.6% of the non-Medicaid group. When adjusted for age, race, sex, and pancreatic status, the odds of an exacerbation were 2.38 (95% CI: 2.11 to 2.69; p < 0.001) higher in the Medicaid group. Even when their worse pulmonary function was considered in the model, the odds of Medicaid patients receiving intravenous antibiotic treatment were 1.58 times higher (95% CI: 1.27 to 1.96; p < 0.001) than those of non-Medicaid patients. Treatment primarily took the form of hospitalization; Medicaid patients were less likely than non-Medicaid patients to receive intravenous antibiotics on an outpatient basis.
Age at diagnosis was examined as an indicator of problems in access to care during infancy. The mean age at diagnosis for the population was 550 d (1.5 yr), but the median was 146 d (0.4 yr). Patients in the Medicaid group had CF diagnosed at a slightly earlier median age (131 d versus 157 d; p < 0.001). With adjustment for significant confounders (age, race, sex, pancreatic status), the difference was not statistically significant (Table 2). This remained true when patients in whom CF was diagnosed at birth because of meconium ileus, prenatal diagnosis, or newborn screening were excluded.
Underutilization of ambulatory specialty care by the poor was not apparent. Medicaid patients averaged 4.5 clinic visits per year, as compared with 4.2 visits in the non-Medicaid group (p < 0.001). This difference was not significant after adjustment for confounders including pulmonary function (Table 2).
Medically indigent CF patients suffer more serious consequences of their disease than does the general CF population. They have more than a threefold greater risk of death, and survivors have significantly worse pulmonary function and growth. This increased severity of disease seems to begin in infancy, and the magnitude of the difference changes minimally with age. Medicaid patients are also more likely to receive hospital treatment for a pulmonary exacerbation of CF, even when their relatively worse pulmonary disease is taken into account. There was no difference in the age at diagnosis or in the number of ambulatory CF clinic visits of Medicaid and non-Medicaid patients, so a disparity in access to health care does not appear to explain their relatively worse course.
Our findings are generally congruent with those of other studies that have found worse health outcomes in indigent populations (4), but patients with CF are different for several reasons. In contrast to other chronic conditions whose etiology is multifactorial, the genetic basis of CF is well-defined (1). Although it can be speculated that a more complete characterization of CF genotype, including the possible role of modifier genes, might lead to closer genotype–phenotype correlation, more of the variability in disease outcome for the CF population as a whole can at present be attributed to SES rather than to genotype.
A unique characteristic of CF is that it is predominantly a disease of white persons. This makes it easier to control for the confounding relationship of race with SES, a difficult problem when examining other chronic diseases (4). A recent report described differences in growth but not in pulmonary function in a group of African–American CF patients matched for age, sex, and genotype with a white population, but SES was not taken into consideration (19). Race did not appear to exert a significant independent effect on disease outcome in our analysis, although in view of the relatively small number of African–American CF patients in our study, a type II error might have been made.
Decreased access to care is an important cause of SES-related adverse health outcomes in many but not all populations (4, 5, 11, 20, 21). It does not appear to play an influential role in CF. Medically indigent CF patients received the same number of outpatient specialty visits as did their privately insured cohorts. We found no differences in outcomes for patients who lacked health insurance, and no rural/urban differences. The fact that our Medicaid group was diagnosed with CF at the same median age as the comparison population implies that access to primary care and opportunities for appropriate diagnostic testing were equal in the two groups in infancy. Given that our Medicaid patients were of smaller height and weight throughout infancy and had worse pulmonary function at age 5 yr, it is conceivable that these children were more symptomatic at an early age and should have had CF diagnosed earlier, but this is speculative. Access to primary care could not otherwise be evaluated with the data available in the NCFPR.
It is striking that the deficiencies in height, weight, and pulmonary function exhibited by Medicaid patients were similar at all ages, with a minimally larger difference seen in older patients. This suggests that early adverse exposures may be of paramount importance in the course of CF. A Canadian study of normal children also reported worse pulmonary function in children of low SES, and the magnitude of the difference was independent of age (22). An alternative explanation for the relatively constant difference in pulmonary function in our study could be that a higher mortality rate results in the loss of more severely compromised patients.
Overall, the adverse effect of poverty on health in CF is probably mediated by many of the same factors that have been shown to be relevant in other chronic diseases (4). Nutritional inadequacies associated with poverty might worsen pulmonary function as well as growth, by compromising immunologic responses to lung infection (20). We found that Medicaid patients were more likely to show low height and weight, but it is unclear whether this was a cause or an effect of respiratory deficiency. Exposure to indoor or outdoor air pollution is more prevalent in the poor and may hasten the progression of pulmonary disease. Cigarette smoking is more prevalent in low SES groups (23), and exposure to environmental tobacco smoke has been linked to decreased pulmonary function in children with CF (24). Respiratory virus infections, particularly with respiratiory syncytial virus (25), may occur earlier and more frequently in economically disadvantaged children, initiating airway inflammation at a younger age (26). Increased stress is another hypothesized cause of greater disease severity in the poor (20). Psychological stress affects immune function (27) and increases susceptibility to infection (28). Stress may also have a negative impact on family function, leading to decreased adherence to prescribed medical regimens (29). Adherence might also be worse in disadvantaged families because they are less well educated about the disease (30, 31). Furthermore, although it appears that Medicaid patients receive adequate specialty care, they might still experience difficulties obtaining access to primary health care, especially before the diagnosis of CF is made.
In the present study, data on insurance status as well as other outcomes were derived from medical records, and may not have been completely reliable. It seems likely, however, that any misclassification error would occur equally in all patients regardless of disease severity, and this nondifferential misclassification would be expected to decrease rather than to increase the association between outcome and exposure (32). A greater potential problem is the appropriateness of Medicaid status as an indicator of low SES. Medicaid eligibility is closely tied to family income, but medical expenses are subtracted from income to determine eligibility (33), so sicker patients are more likely to receive Medicaid. We attempted to minimize this potential bias by limiting our comparison to patients who received Medicaid in every year throughout the period of analysis. We also excluded adult patients, in whom health problems are more likely to influence SES and insurance status. Moreover, a parallel analysis that compared all patients receiving Medicaid in the year of their CF diagnosis with those who had private insurance at the time of diagnosis gave similar results. Thus, it appears that Medicaid eligibility precedes the development of more severe disease.
It has recently been asserted that longitudinal analysis of pulmonary function data, using statistical methodologies such as mixed-model analysis, is the best approach to describing and comparing pulmonary function trends in CF (34). However, 1994 was the first year in which multiple pulmonary function measurements were recorded in the NCFPR, and we believe that the average of these measurements provides a more valid indicator of patients' respiratory status than do the single annual measurements reported before that year. Our analytical approach evaluated age-adjusted averages, and not the rate of decline in individual patients; but the regression coefficient for FEV1 (−2.35% predicted per year of age) closely approximates annual changes found in several recent longitudinal studies of children with CF (35).
Medically indigent patients form a subgroup whose mortality and morbidity are significantly worse than those of the population of CF patients as a whole. Our study shows that factors other than genotype have a significant effect on CF phenotype. Presumably, medically indigent patients suffer more severe consequences of CF, because low SES is associated with a clustering of detrimental environmental influences. The findings of our study have several important implications regarding therapy for and further research into the mechanism of CF. First, we can readily identify in the poor a group of patients at risk for adverse outcomes. As a public health measure, it might be appropriate to offer this group more intense standard treatment (e.g., more antibiotics, more antiinflammatory therapy) than to a lower risk group, but the efficacy of this approach would need to be proven. On the other hand, if the specific causes of the link between poverty and more severe lung disease in CF are identified, then more specific effective therapies could be offered to these patients. This knowledge would also lead to improvements in outcome for others who are not of low SES but who may, nevertheless, be exposed to any of these risk factors. Second, because of its association with adverse health outcomes, low SES should be considered as a potential confounder in all clinical research trials, with appropriate adjustments made for it. Additionally, a better understanding of environmental risk factors and how they mediate adverse outcomes would provide insight into the basic biology and pathogenesis of CF lung disease.
| 1. | Davis PB, Drumm M, Konstan MWCystic fibrosis—state of the art. Am J Respir Crit Care Med154199612291256 |
| 2. | Dean M, Santis GHeterogeneity in the severity of cystic fibrosis and the role of CFTR gene mutations. Hum Genet931994364368 |
| 3. | Rosenfeld M, Davis R, FitzSimmons R, Pepe M, Ramsey BGender gap in cystic fibrosis mortality. Am J Epidemiol1451997794803 |
| 4. | Adler NE, Boyce WT, Chesney MA, Folkman S, Syme SISocioeconomic inequalities in health. JAMA269199331403145 |
| 5. | Mackenbach JP, Kunst AE, Cavelaars AE, Groenhof F, Geurts JJSocioeconomic inequalities in morbidity and mortality in western Europe: the EU Working Group on Socioeconomic Inequalities in Health. Lancet349199716551659 |
| 6. | Britton JEffects of social class, sex, and region of residence on age at death from cystic fibrosis. BMJ2981989483487 |
| 7. | Schechter MS, Margolis PARelationship between socioeconomic status and disease severity in cystic fibrosis. J Pediatr1321998260264 |
| 8. | Curtis JR, Burke W, Kassner A, Aitkin MLAbsence of health insurance is associated with decreased life expectancy in patients with cystic fibrosis. Am J Respir Crit Care Med155199719211924 |
| 9. | FitzSimmons SCThe changing epidemiology of cystic fibrosis. J Pediatr122199319 |
| 10. | Liberatos P, Link BG, Kelsey JKThe measurement of social class in epidemiology. Epidemiol Rev10198887120 |
| 11. | Cartland JD, McManus MA, Flint SSA decade of Medicaid in perspective: what have been the effects on children? Pediatrics911993287295 |
| 12. | U.S. Census Bureau. STF 3B, File 01, Tables P1-P13. Washington, DC: U.S. Government Printing Office; 1990. |
| 13. | Kerem E, Reisman J, Corey M, Canny GJ, Levison HPrediction of mortality in patients with cystic fibrosis. N Engl J Med326199211871191 |
| 14. | Packe GE, Hodson MEChanges in spirometry during consecutive admissions for infective pulmonary exacerbations in adolescent and adult cystic fibrosis. Respir Med8619924548 |
| 15. | Knudson RJ, Lebowitz ME, Holberg CJ, Burrows BChanges in the normal maximal expiratory flow-volume curve with growth and aging. Am Rev Respir Dis1271983725734 |
| 16. | Hsu KH, Jenkins DE, Hsi BP, Bourhofer E, Thompson V, Tanakawa N, Hsieh GSVentilatory functions of normal children and young adults— Mexican–American, white, and black; I. Spirometry. J Pediatr9519791423 |
| 17. | Hamosh A, Corey MCorrelation between genotype and phenotype in patients with cystic fibrosis. N Engl J Med329199313081313 |
| 18. | Cox DR, Oakes D. Analysis of survival data. London: Chapman and Hall; 1984. |
| 19. | Hamosh A, FitzSimmons SC, Macek M, Knowles MR, Rosenstein BJ, Cutting GRComparison of the clinical manifestation of cystic fibrosis in black and white patients. J Pediatr1321998255259 |
| 20. | Bor DH, Epstein PRPathogenesis of respiratory infection in the disadvantaged. Semin Respir Infect61991194203 |
| 21. | Pincus T, Esther R, DeWalt DA, Callahan LFSocial conditions and self-management are more powerful determinants of health than access to care. Ann Intern Med1291998406411 |
| 22. | Dimissie K, Ernst P, Hanley JA, Locher U, Menzies D, Becklake MSocioeconomic status and lung function among primary school children in Canada. Am J Respir Crit Care Med1531996719723 |
| 23. | Haire-Joshu D, Morgan G, Fisher JEBDeterminants of cigarette smoking. Clin Chest Med41991711724 |
| 24. | Kovesi T, Corey M, Levison HPassive smoking and lung function in cystic fibrosis. Am Rev Respir Dis148199312661271 |
| 25. | Abman SH, Ogle JW, Butler-Simon N, Rumack CM, Accurso FJRole of respiratory syncytial virus in early hospitalization for respiratory distress of young infants with cystic fibrosis. J Pediatr1131988826830 |
| 26. | Khan TZ, Wagener JS, Bost T, Martinez J, Accurso FJ, Riches DWEarly pulmonary inflammation in infants with cystic fibrosis. Am J Respir Crit Care Med151199510751082 |
| 27. | Kiecolt-Glaser JK, Malarkey WB, Chee M, Newton T, Cacioppo JT, Mao H, Glaser RNegative behavior during marital conflict is associated with immunological down-regulation. Psychosom Med551993395409 |
| 28. | Cohen S, Tyrrell DA, Smith APPsychological stress in humans and susceptibility to the common cold. N Engl J Med3251991606612 |
| 29. | Patterson JM, Budd J, Goetz D, Warwick WJFamily correlates of a 10-year pulmonary health trend in cystic fibrosis. Pediatrics911993383389 |
| 30. | Henley LD, Hill IDErrors, gaps, and misconceptions in the disease- related knowledge of cystic fibrosis patients and their families. Pediatrics85199010081014 |
| 31. | DiMatteo MREnhancing patient adherence to medical recommendations. JAMA27119947982 |
| 32. | Walter SD, Irwig LMEstimation of test error rates, disease prevalence and relative risk from misclassified data: a review. J Clin Epidemiol411988923937 |
| 33. | Holahan J, Rowland D, Feder J, Heslam DExplaining the recent growth in Medicaid spending. Health Aff121993177193 |
| 34. | Corey M, Edwards L, Levison H, Knowles MLongitudinal analysis of pulmonary function decline in patients with cystic fibrosis. J Pediatr1311997809814 |
| 35. | Konstan MW, Byard PJ, Hoppel CL, Davis PBEffect of high-dose ibuprofen in patients with cystic fibrosis. N Engl J Med3321995848854 |
