To describe and understand the burden of out-of-pocket expenses in patients with rheumatoid arthritis (RA).
To describe and understand the burden of out-of-pocket expenses in patients with rheumatoid arthritis (RA).
We studied out-of-pocket expenses and their burden in 8,545 US patients with RA. We determined direct medical costs, out-of-pocket expenses, the burden of out-of-pocket expenses, household income, and measures of RA severity and outcome. In addition, patients were classified into 3 groups based on the level of burden caused by out-of-pocket expenses: no or limited problem (I am able to pay the bills without much problem); moderate problem (paying the bills takes away some money I need for other activities); and a great problem (I can't purchase all of the medications or medical care that I need).
A total of 43.6% of patients reported problems paying medical bills after insurance payments and 9.0% reported severe or great problems. Problems with expenses were associated with measures of RA severity, but also and particularly with lower household income and absence of health insurance. The proportion of household income that was consumed by out-of-pocket spending for the 3 groups was 2.4%, 7.2%, and 19.2%, respectively, and the percentage of patients meeting the 185% poverty level for these groups was 12.3%, 24.4%, and 51.3%, respectively.
The out-of-pocket burden is substantial, particularly in those <65 years of age. Out-of-pocket expenses exert their severity predominantly on those with the most severe RA who have the least ability to pay. Household income is the primary determinant of out-of-pocket burden, followed by RA severity, and type of health insurance.
Out-of-pocket expenses are an increasing burden to persons seeking health care (1–5). Compared with Australia, Canada, New Zealand, and the UK in 1998, more patients in the US were having problems with medical bills (18%), were not filling a prescription because they could not afford it (17%), and reported having bills over $250 that were not covered by insurance (29%) (6). Among Medicare recipients, 7% reported cost-related poor adherence to treatment, and adherence was related to prescription coverage level (2). Among the poor, cost-related poor adherence increased to 20% (2). In 1996, the mean annual out-of-pocket spending per person for persons with ≥3 comorbid conditions was $1,134 (1).
Out-of-pocket expenses have attracted interest in rheumatoid arthritis (RA) as part of studies of direct medical costs, primarily from the societal perspective (7). However, out-of-pocket expenses are clearly of separate interest to patients, too (8, 9). Only one RA study addressed out-of-pocket expenses from the perspective of the patient. A detailed report on 136 patients disaggregated out-of-pocket expenses into domains of cost (e.g., hospital, drugs, etc.) and concluded that out-of-pocket expenses were a substantial burden to patients (7).
What does it mean to patients that out-of-pocket expenses are “high” or are a “substantial burden”? And who suffers the burden? Who is injured? We hypothesized that the primary factor in the real world burden of out-of-pocket expenses at the level of the individual patients was household income, and that this burden would also be related to RA severity. This study investigates these hypotheses, as well as the effect on health insurance.
Participants in this study were part of the National Data Bank for Rheumatic Diseases (NDB) ongoing longitudinal observational study of RA outcomes (10, 11). Beginning in 1998, NDB participants were recruited from the practices of US rheumatologists and were followed prospectively with semi-annual, detailed, 28-page questionnaires. Participants are volunteers who have a variety of rheumatic diseases, and are not compensated for their participation. In 2005, a question was added to the surveys regarding out-of-pocket expenses. The sample for the current study was composed of a randomly selected semi-annual observation from each of the 8,545 RA patients who responded to the out-of-pocket questions from 2005 through 2007 (response rate 97.4%). To maximize complete data, in case patients had complete data at one observation but not at another, the random selection process first selected from among patient observations with complete out-of-pocket data, then selected from the remaining patient observations. The mean study year was 2006.
Demographic variables included age, sex, ethnicity, education level, marital status, medical insurance, and total household income. Household income was assessed with a multiple choice question, “Which income group comes closest to your total household income in the last year from all sources before taxes?” Eleven choices were available, ranging from <$10,000 to ≥$100,000, in $10,000 increments. Disabled status variables were based on patient self-report. Specifically, we asked patients which of the following categories best characterized their current work status: paid work, housework, student, unemployed, disabled, or retired. We characterized a patient as having self-reported work disability if he or she indicated disabled to the above question, and as receiving US Social Security disability benefits if he or she indicted so on a direct question. Validation studies have demonstrated the reliability of the disability assessments (12).
To determine poverty levels, we used the Health and Human Services poverty guidelines for the 48 contiguous states for the years 1998–2008 (13). Poverty guidelines are simplified versions of the federal poverty thresholds, and are based on income level in a given year and the number of persons in a household. A level of 185% of the Health and Human Services poverty guideline, selected for this study, is a commonly used measure of poverty, and is used to determine eligibility for the School Breakfast and Lunch programs. To calculate poverty levels we used the midpoint of the income categories (e.g., $25,000 for the $20,000–$29,999 category). For the ≥$100,000 category we used $100,000 as the midpoint.
To determine out-of-pocket expenses we asked the following question: “Between (specific 6-month period), approximately how much did you spend out of pocket on your medical expenses (this includes expenses for medication, doctor visits, x-rays, laboratory tests, hospitalizations and more)? Do not include what you paid for health insurance or any costs reimbursed by insurance.”
To determine the burden of out-of-pocket expense, we asked the following question referring to the 6-month survey period, “Drug, doctor, and hospital costs may or may not be partially or fully paid by your insurance. How much of a financial problem are your drug and medical bills after receiving all insurance reimbursement?“ Responses could be “No problem or limited problem (I am able to pay the bills without much problem); a moderate problem (paying the bills takes away some money I need for other activities); or, a great problem (I can't purchase all of the medications or medical care that I need).” Direct medical costs, adjusted to 2007, were calculated from hospitalization, treatment, and utilization data as previously described (14).
Comorbidity was measured by a patient-reported composite comorbidity score (range 0–9) comprised of 11 present or past comorbid conditions including: pulmonary disorders, myocardial infarction, other cardiovascular disorders, stroke, hypertension, diabetes mellitus, spine/hip/leg fracture, depression, gastrointestinal ulcer, other gastrointestinal disorders, and cancer (15, 16).
RA assessment measures included the Health Assessment Questionnaire (HAQ) functional disability index (DI) (17), the Medical Outcomes Study Short Form 36 (SF-36) physical component (PCS) and mental component (MCS) summary scales (18), and visual analog scales (VAS) for pain and patient global severity. For the purposes of the multivariable analyses, we considered HAQ DI, pain, global severity, and use of prednisone, analgesics, and opioids as RA severity variables. We also collected data on quality of life using the EuroQol (EQ-5D) utility score. The EQ-5D is a 5-item questionnaire that assesses function (3 questions), mood (1 question), and pain (1 question) (19). Scoring was accomplished using US tariffs (weights) (20, 21).
Linear regression was used to analyze the trend across the 3 ordered groups of Table 1, log- transforming cost data. Multivariable ordered logistic regression was used to analyze predictors of the categories of out-of-pocket burden. In these analyses continuous predictors were dichotomized at the median, and private (nongovernment) insurance was combined into a single variable. We used this information to predict the probability of mild, moderate, or severe out-of-pocket financial burden, adjusting for all other variables in the model.
|Variable||No difficulty||Moderate difficulty||Great difficulty|
|Financial burden, %||56.5||34.6||9.0|
|Age <65 years, %||52.8||35.4||11.8|
|Age ≥65 years, %||61.3||33.4||5.3|
|Age, mean ± SD years||63.4 ± 12.5||61.9 ± 12.6||57.4 ± 12.4|
|Sex, % female||77.3||81.5||86.6|
|Non-Hispanic white, %||94.9||94.3||89.7|
|College graduate, %||36.1||24.5||20.3|
|Not married, %||24.4||28.9||37.9|
|Median income, dollars||55,000||35,000||25,000|
|Poverty level (185%)||12.3||24.4||51.3|
|RA duration, median years†||15.3||14.4||14.2|
|Self-reported disabled, %||8.0||17.0||36.0|
|Social Security disability, %||10.1||19.5||33.9|
|Lifetime TJR, %‡||24.4||26.0||24.5|
|Comorbidity index, mean ± SD (range 0–9)§||1.6 ± 1.5||2.0 ± 1.6||2.5 ± 1.9|
|SF-36 physical score, mean ± SD||35.8 ± 10.8||31.1 ± 10.0||26.6 ± 8.4|
|SF-36 mental score, mean ± SD||53.8 ± 9.6||49.2 ± 11.2||42.9 ± 12.6|
|EQ-5D utility, mean ± SD (range −0.11, 1)||0.79 ± 0.16||0.71 ± 0.18||0.59 ± 0.23|
|HAQ DI score, mean ± SD (range 0–3)||0.9 ± 0.7||1.2 ± 0.7||1.5 ± 0.7|
|VAS pain, mean ± SD (range 0–10)||3.1 ± 2.5||4.2 ± 2.7||5.8 ± 2.7|
|Global severity, mean ± SD (range 0–10)||3.0 ± 2.3||4.0 ± 2.4||5.2 ± 2.5|
|Fatigue, mean ± SD (range 0–10)||3.6 ± 2.8||4.8 ± 2.9||6.1 ± 2.8|
|Biologic agents, %‡||47.7||52.0||44.1|
|Any analgesic, %||37.8||44.9||55.7|
|Payment and cost|
|Out-of-pocket expense, mean ± SD dollars (n = 4,836)||1,226 ± 1,966||2,408 ± 3,149||3,338 ± 7,711|
|Out-of-pocket expense, median dollars (n = 4,836)||600||1,400||1626|
|Outpatient costs, mean ± SD dollars||846 ± 913||1,080 ± 1,155||1,335 ± 1,530|
|Drug costs, mean ± SD dollars‡||7,721 ± 11,970||8,790 ± 11,883||7,522 ± 11,522|
|Hospital costs, mean ± SD dollars||900 ± 3,406||1,321 ± 4,343||1,599 ± 4,577|
|Total medical costs, mean ± SD dollars||9,470 ± 12,531||11,194 ± 12,844||10,481 ± 12,774|
Except for out-of-pocket expenses, the range of missing data for all the variables used in the study was 0–4.79%. Of the 8,545 observations in the study, 15.7% had missing values and 1.6% of all values were missing overall (7,203 were all present). For the variables used in the multivariable regression analyses, 147 of 8,545 values (0.02%) were missing. Because missing data were few, we used single imputation to replace missing data. The imputations were generated via randomly drawn models from a Bayesian posterior distribution based on cases in which the imputed variable was present (22). Because of substantial missing data for out-of-pocket expenses (43.4%), this variable was excluded from regression analyses and missing data were not imputed for it. Data were analyzed using Stata, version 10.1 (23). The level of statistical significance was set at 0.05 and all tests were 2-tailed.
Beginning with the NDB questionnaire of January 2005, we asked patients to state their approximate out-of-pocket expenses for all medical expenses, and 3,709 (43.4%) of patients did not provide an answer to that question. Comparing patients who provided data with those who did not, 54.4% versus 58.1% had no out-of-pocket burden, 35.6% versus 33.7% had a moderate out-of-pocket burden, and 10.0% versus 8.2% had a severe out-of-pocket burden. The data of the present versus absent groups differed for the out-of-pocket responses (χ2 = 11.5, P = 0.003).
Followup interviews with those patients with missing data indicated that they were unable to determine the out-of-pocket amount because of charge and reimbursement complexity, and temporal issues. Results of multivariable logistic regression indicated that patients who could not answer the question were more likely to be ≥65 years of age, odds ratio (OR) 2.2 (95% confidence interval [95% CI] 2.0–2.4), below the median income of study participants, OR 1.4 (95% CI 1.3–1.5), not college graduates, OR 1.3 (95% CI 1.2–1.5), and had higher HAQ DI scores, OR 1.1 (95% CI 1.0–1.2). However, those with missing out-of-pocket cost data were more likely to have no problems paying their medical and drug bills after receiving insurance payments (58.1% versus 54.4%).
Overall, 43.6% of RA patients had problems paying medical and drug bills after insurance payments and 9.0% reported a severe or great burden—being unable to purchase all the medications or care they needed because of out-of-pocket medical expenses (Table 1). This burden was substantially greater for patients <65 years of age (11.8%) compared with those age ≥65 years (5.3%). Among the patients providing out-of-pocket expense data, the mean ± SD annualized out-of-pocket expenses were $1,798 ± $3,314 and the median $1,000. Mean and median expense for the 4 quartiles were Q1 $106 and $90, Q2 $679 and $620, Q3 $1,557 and $1,500, and Q4 $5,221 and $4,000, respectively (Figure 1). For patients in the 90th percentile of out-of-pocket expenses the mean and median were $8,157 and $6,000.
To better understand the relationship between difficulty in paying for medications and demographic and disease severity, we studied the 8,545 patients who provided payment problem data (Table 1). There was a step-wise increase in RA severity or adverse status as groups went from no problem to great problem with out-of-pocket expenses. The demographic variables of sex, ethnicity, college education, nonmarried status, and median income that are usually associated with lower income states were associated with out-of-pocket severity states. However, this was not true for age in which lower age was associated with out-of-pocket severity, perhaps reflecting change in insurance status and reduced income need in those ≥65 years of age. The mean ± SD number of comorbid conditions was 1.8 ± 1.6, not including RA.
RA related variables in Table 1 were also more abnormal with increasing out-of-pocket burden. In addition, the results in the severe burden group were quite abnormal. For example, the HAQ DI score was 1.5, the patient global severity was 5.2, and the work disability rates were more than 3 times greater than in the mild group, among other severe abnormalities. Quality of life as measured by the EQ-5D was very low (0.59) in the severe burden group.
By contrast, total medical costs were not greater in the severe out-of-pocket group compared with the moderate group (Table 1). This was the result of reduced drug costs in the severe group. However, out-of-pocket expenses increased across the groups, although the difference between moderate and severe groups was less than might be expected given the severity rating. Biologic therapy was not associated with the trend of out-of-pocket burden. Patients receiving biologic agents had mean out-of-pocket expenses of $1,906 compared with $1,678 for those who did not receive biologic agents, a difference of $228. The total semi-annual medical cost for patients receiving biologic agents was $17,508 compared with $3,129 for those not receiving biologic agents. By logistic regression, a $10,000 increase in household income was associated with a 6% increase in the odds of biologic agent use (OR 1.06, 95% CI 1.04–1.05).
The relationship of out-of-pocket severity to health insurance and annual household income is shown in Table 2. Severe out-of-pocket expense burden occurs most often in those with no insurance (60.5%), those receiving Medicare disability (21.5%), and those receiving Medicaid (14.8%), all of which are groups that are associated with RA severity and reduced income. In addition, the severe out-of-pocket burden is mostly associated with household income classes of $35,000 or less (Table 2). The median household income of those without health insurance or with Medicare disability or Medicaid was $25,000 compared with $45,000 for those with other insurance types. The proportion of household income that was consumed by out-of-pocket spending was 2.4%, 7.2%, and 19.2% for the no burden, moderate burden, and great burden groups, respectively, and the 185% poverty levels for these groups were 12.3%, 24.4%, and 51.3%, respectively (and 19.8% overall).
|Variable||N||None or mild, %||Moderate, %||Severe, %|
|Annual household income, dollars|
To understand the relative role of insurance, demographics, and RA severity on out-of-pocket severity we performed multivariable ordered logistic regression for key study variables, adjusted for all other variables in the model (Table 3). In addition, we predicted the probability of moderate and severe out-of-pocket burden, also after adjusting for all other variables in the model. The OR, Z score, and predicted probability of severe burden provides information about the most important predictor variables. Income below the median, with an OR of 3.03, a Z score of −20.6, and a severe probability of 0.110, had the most impact on out-of-pocket burden, followed by the absence of health insurance, and VAS pain. In these analyses prednisone and biologic agent use were not significantly associated with the severity of out-of-pocket burden. Compared with private insurance, Medicare and Medicaid insurance were associated with a lower risk of out-of-pocket burden. Total direct medical costs were not associated with out-of-pocket burden. To provide more information about household income than is contained in the above/below median analysis of Table 3, we performed a similar analysis to that of Table 3 using all of the categories of household income. As shown in Figure 2, burden increases substantially with decreasing household income. In additional analyses of data from Table 3 we added a dummy variable to identify those who did or did not provide out-of-pocket expense data. This variable was not significant in the model.
|Variable||Percentage of sample||OR (SE)||Z score||P||Probability moderate burden (95% CI)||Probability severe burden (95% CI)|
|All patients||0.369 (0.556–0.579)||0.064 (0.059–0.069)|
|College graduate, %||30.7||0.78 (0.04)||−4.8||0.000||0.335 (0.318–0.353)||0.054 (0.048–0.059)|
|Age 65 years, %||43.2||0.77 (0.10)||−1.9||0.052||0.345 (0.315–0.374)||0.056 (0.047–0.065)|
|Sex, % female||79.6||1.11 (0.07)||1.7||0.092||0.372 (0.360–0.384)||0.065 (0.059–0.070)|
|Not married, %||27.2||1.09 (0.06)||1.6||0.114||0.376 (0.358–0.394)||0.066 (0.059–0.073)|
|RA disease treatment|
|Pain > median||48.2||1.69 (0.10)||9.3||0.000||0.419 (0.404–0.435)||0.082 (0.075–0.090)|
|HAQ DI score > median||49.0||1.43 (0.08)||6.6||0.000||0.404 (0.389–0.419)||0.076 (0.069–0.083)|
|Patient global > median||49.3||1.34 (0.08)||5.2||0.000||0.394 (0.379–0.410)||0.072 (0.066–0.079)|
|Opioids, %||21.9||1.31 (0.09)||3.9||0.000||0.410 (0.388–0.433)||0.079 (0.069–0.088)|
|Prednisone, %||31.6||1.10 (0.05)||1.9||0.063||0.380 (0.364–0.397)||0.067 (0.061–0.074)|
|Any analgesic, %||41.8||0.91 (0.05)||−1.5||0.128||0.358 (0.342–0.375)||0.060 (0.054–0.067)|
|Biologic agents, %||48.8||0.99 (0.08)||−0.2||0.881||0.370 (0.351–0.388)||0.064 (0.057–0.071)|
|Private insurance||41.3||0.419 (0.389–0.450)||0.082 (0.068–0.096)|
|No insurance, %†||2.0||8.95 (1.53)||12.9||0.000||0.504 (0.465–0.543)||0.359 (0.283–0.435)|
|Medicaid, %†||6.1||0.59 (0.07)||−4.3||0.000||0.264 (0.224–0.305)||0.037 (0.028–0.046)|
|Medicare disability†||8.5||1.38 (0.12)||3.8||0.000||0.422 (0.392–0.451)||0.083 (0.070–0.096)|
|Medicare, %†||36.3||0.64 (0.09)||−3.1||0.002||0.313 (0.278–0.348)||0.048 (0.039–0.057)|
|Medicare HMO, %†||5.8||0.86 (0.12)||−1.1||0.265||0.341 (0.292–0.390)||0.055 (0.041–0.069)|
|Total costs > median||50.0||1.09 (0.09)||1.1||0.272||0.375 (0.357–0.394)||0.066 (0.059–0.073)|
|Income < median||46.2||3.03 (0.16)||−20.6||0.000||0.471 (0.456–0.486)||0.110 (0.101–0.119)|
There are 2 separate but related aspects of out-of-pocket expenses, the expenses themselves and the burden the expenses place on patients with RA. With respect to out-of-pocket expenses in RA, we found such expenses to be high, $1,798. At the 4th quartile the mean out-of-pocket expenses were $5,221, and at the 90th percentile they were $8,157. In the general US population in 1996, out-of-pocket spending for persons with ≥3 comorbid conditions was $1,134 (1). This translates to approximately $1,728 in 2007 dollars. Our out-of-pocket cost data were obtained 10 years later, in an era of increasing copayments (3, 4), but were approximately the same as the US data, although the mean number of comorbid condition in this study was 1.8. However, differences between the age, sex, ethnic, and educational characteristics of the US population and the current study population, as well as differences in era, make true comparisons difficult.
Cost sharing, or increasing out-of-pocket expenses, is an important public concern in patients with RA (3–5), but the immediate consequences of out-of-pocket expenses occur at the level of the individual patient, reinforcing the difference between a global societal prospective and a patient perspective. An increase in cost sharing can be of little consequence to patients with ample financial resources, but can be devastating to those who lack them.
We characterized patients into 3 out-of-pocket expense burden groups according to their self-report. Patients having any difficulty with out-of-pocket expenses had more severe RA and socioeconomic disadvantage than the 56.5% without problems (Table 1). This was particularly true for the 9% with severe problems. For example, the HAQ DI and pain scores (1.5 and 5.8, respectively) in this group are at the 65th percentile of such scores in the NDB, and were much greater than those in the no problem groups. Of particular interest was the rate of Social Security and self-reported disability, which was 3 to 4 times greater than in the no problem groups. But most striking, and of central importance, was the median household income difference, $25,000 versus $55,000. Differences were less extreme between the moderate and no burden group, but were still clearly discernible and important.
In addition, the proportion of household income that was consumed by out-of-pocket spending was 2.4%. 7.2%, and 19.2% for the no burden, moderate burden, and severe burden groups, respectively. At the same time, the 185% poverty levels for these groups were 12.3%, 24.4%, and 51.3%, respectively (the overall sample 185% poverty rate was 19.8%). These data suggest that in addition to the loss or postponement of medical services and treatments, there is a direct, personally important loss of available income.
We have shown previously that patients with RA have reduced income compared with workers doing similar jobs, and that household income is related to the degree of functional disability (24). In the current study, median household income was lowest in those who had no insurance, Medicare disability, or Medicaid insurance ($25,000) compared with $45,000 for other types of insurance. This suggests that out-of-pocket burden reflects reduced income more than type of health insurance. A similar conclusion can be drawn from the data shown in Table 3.
One interesting finding of this study was that the use of biologic agents was not significantly associated with financial burden. This was true both in the unadjusted (Table 1) and adjusted analyses (Table 3). Biologic agents are very expensive treatments and often are associated with high copayments, and we found that higher household income was associated with increased use of biologic agents. It seems likely that other factors that were not assessed by our model explained the nonassociation of biologic agents and out-of-pocket financial burden. One possible explanation is that biologics were offered preferentially to those with the ability to absorb out-of-pocket expenses or with insurance that had little or no copayment requirement.
The problems of out-of-pocket burden occur in societies that do not provide universal free health care or that require substantial copayments for medical care. In the current study, persons without health insurance had the greatest adjusted probability of severe out-of-pocket burden, 35.9%, compared with the overall severe burden of 6.4% (Table 3). Approximately 25% of the US population between the ages of 18 and 65 years are without health insurance (25). In the current study 3.4% of those <65 years of age had no health insurance. However, current study participants are not typical of the US population (30.7% are college graduates and 94.2% are non-Hispanic whites), and the actual percentage of RA patients without health insurance in the general population is certainly much higher. The lack of guaranteed health insurance is a characteristic of the US and stands in “stark contrast to the situation in virtually every other developed country” (26). Given the severe RA and the socioeconomic disadvantage of those without health insurance, one recommendation that can be drawn from this study is the need for universal health care.
A severe out-of-pocket burden was reported by 9% of the study patients. When those <65 are considered, 11.8% reported a severe out-of-pocket burden. We found that this group had severe illness, including very abnormal PCS and MCS scores, increased disablement, high levels of HAQ functional disability and pain, and an EQ-5D utility score of 0.59—indicating severely decreased quality of life. The finding that those with the most severe illness have the greatest out-of-pocket burden further suggests the recommendation for guaranteed, affordable health care as a matter of public policy. While affordable health, here measured by out-of-pocket burden, is not the main contributor to public health status (27), it is a signal consideration at the level of the individual patient.
Our study has a number of limitations. One limitation relates to the estimate of out-of-pocket expenses. More then 40% of patients were unable to provide out-of-pocket data because of their complexity. We had no means to validate patients' reports, and the degree of reliability of such reports is not known. This limitation is similar to that faced by most studies, and has often been addressed by estimating a fixed proportion for out-of-pocket expenses, a method that is satisfactory for group data, but less so for individual patient data. In addition, the questions we asked regarding problems paying bills implied that “no problem” meant no problem paying bills or having funds for other activities, “moderate problem” meant that the problem in paying bills meant postponing other activities, and a “severe problem” meant postponing other activities and not being able to purchase all needed medical care. These categories were not completely symmetrical and the implied limitations were not stated for all categories; this may have led to some misunderstanding by patients.
In summary, although out-of-pocket expenses and recent increases in copayments have important societal effects, at a patient level they cause a substantial burden in those with less income. In patients with RA, income level is very often a function of RA severity itself. As a result, the burden caused by out-of-pocket expenses fall most heavily on those RA patients with the most severe illness, consuming considerable household income, as well as limiting available services. The most important contributors to severe out-of-pocket burden are household income, absence of heath insurance, and RA severity.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Wolfe had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Wolfe, Michaud.
Acquisition of data. Wolfe, Michaud.
Analysis and interpretation of data. Wolfe, Michaud.
The authors thank Mary Charlton, PhD, University of Nebraska, for her helpful suggestions.