Appendix E:  Arthritis Foundation/American College of Rheumatology Survey of Pediatric and Internist Rheumatologists

Data Sources

In conjunction with the American College of Rheumatology, the Arthritis Foundation created and fielded a survey of pediatric rheumatologists and internist rheumatologists in the United States.  The survey was not formally pilot tested and was fielded using a Web-based format.  All physician members of the American College of Rheumatology (n=3627) received an email message describing the survey and providing a link to the online survey.  The email message was sent on March 9, 2004 and a reminder on March 23, 2004. 

Data Cleaning

The Arthritis Foundation provided access to a comma-delimited version of the survey data.  Raw data were converted to a usable format by performing multiple data manipulations.   Observations that were obvious duplicates, (i.e., observations that were identical to another observation for all variables, (n=22) were deleted).  Observations from non-physicians, non-rheumatologists, retired physicians (n=38), fellows (n=21), physicians practicing overseas (4), physicians with missing data that prevented categorizing them as pediatric or internist rheumatologists (3), and records with no data (4) were excluded.  All impossible values for zip code were recoded to missing; all non-US zip codes to missing were recoded to missing.  For questions in a series (e.g., “How much time do you spend on each of the following activities?”), recoding of values to missing was done only if all questions in the series do not have data; otherwise, blank values were recoded to zero.  Reponses of “don’t know” for percentage of Medicaid and uninsured patients were considered missing.

Several questions asked respondents to report time amounts.  If respondents reported a range of values (e.g., 10-15 hours), the lower end of the reported range was used.  If respondents report a value of greater than or less than (e.g., >40 hours), the value reported was used, without the greater than or less than qualifier.  If the total number of hours per week summed to 100, it was assumed that the provider had entered in percentages; these were converted to hours using an assumed workweek of 49 hours.

For questions that require a “Yes/No” response, binary variables were created to indicate an affirmative response.  For open-ended questions that requested specification, categorical variables were created that collapsed the responses into a smaller number of categories.

Methods

Respondents were classified as either adult or pediatric rheumatologists or internist rheumatologist using data on their Board certification status and/or their self-description.  Internist rheumatologists were further classified as involved in the care of children if they reported that children represented at least some of their patients.  Internist rheumatologists who specified that they did not treat children with rheumatic diseases were considered to not be involved in the care of children.

Bivariate analyses were performed using Pearson c² for categorical variables and two-sided, two-sample t-tests for comparisons of means.  In addition, logit analyses were performed to assess the independent association between distance to the nearest pediatric rheumatologist and the likelihood that an internist rheumatologist treated children; distance data were derived from the HRSA Bureau of Health Professions’ Area Resource File.

Results

Of the 3,637 emails sent, 300 failed to reach the intended recipient; 3,337 individuals received the email message; of this number, 706 unique survey responses were received.  Among the respondents 63 were ineligible due to:  residence outside the U.S. (4), retirement from patient care or misclassification as a rheumatologist (38), or current fellowship status (21) yielding an effective sample of 3,274.  Of the 706 responses received, 633 were eligible and could be used in the analyses; the effective response rate was 19.3 percent.  Given that approximately 224 physician members of the American College of Rheumatology are pediatric rheumatologists, a figure that includes both fellows and practicing physicians, it was estimated that the response rate among pediatric rheumatologists was considerably higher, closer to 55 percent.

Table E.1.  Provider Type,¹ Arthritis Foundation/American College of Rheumatology Survey

¹Current fellows (n=21) excluded from analyses

Of all respondents, 17.4 percent were pediatric rheumatologists and 82.6 percent were internist rheumatologists (Table B.1).  Of the internist rheumatologists in patient care, 55.2 percent spent at least some patient care time treating children. 

Bivariate Analyses

Pediatric rheumatologists were compared to internist rheumatologists who did and did not treat children for all responses (Table B.2).  In addition, characteristics of internist rheumatologists involved in the care of children were compared with those who did not treat children.  Pediatric rheumatologists spent significantly more time than internist rheumatologists in teaching and research and less time in-patient care.  Compared to internist rheumatologists who did not treat children, internist rheumatologists who treated children spent significantly more time in patient care and less in research.  Pediatric rheumatologists were, as expected, significantly more likely to spend most of the practice time caring for children.  Of those internist rheumatologists involved in the care of children, the overwhelming majority reported that children comprise less than 10 percent of their patient care time.

Pediatric rheumatologists were younger than internist rheumatologists and more likely to be women.  Internist rheumatologists who treated children were slightly more likely to be men, however, than those who did not treat children.  Internist rheumatologists who treated children were more likely to be white and less likely to be Asian than those who did not treat children.

Table E.2.  Provider and Practice Characteristics by Provider Type, Arthritis Foundation/American College of Rheumatology Survey

Characteristics of Pediatric Rheumatologists

Nearly 90 percent of responding pediatric rheumatologists worked in an academic setting or teaching hospital (Table B.3).  These providers received the majority of their referrals from pediatricians (61.7 percent); other physician specialties contributed far less to referrals.  Almost two-thirds of responding pediatric rheumatologists reported that the current wait time for an appointment was greater than 2 weeks.  

Table E.3. Provider and Practice Characteristics:  Pediatric Rheumatologists Arthritis Foundation/American College of Rheumatology Survey

*Mean number of non-physician clinicians per practice is 1.4.

About half of pediatric rheumatologists used a non-physician clinician in their practice; in most cases the non-physician clinician was a nurse practitioner.  Pediatric rheumatologists with a non-physician clinician in their practice spent about half of their practice time working with this person. 

About one-third of pediatric rheumatologists planed to decrease their time in clinical care in the next 5 years, with 33.1 percent the average planned decrease.  The primary reason for planning a decrease in time was obtaining salary support from a research source (45.5 percent); however, many also report retirement (21.2 percent), salary support from a business source (18.2 percent), and other reasons (27.3 percent).  Interestingly, one-third of those who planned to decrease their patient care time within the next 5 years also had reported already decreasing their patient care time in the previous 5 years.

Just over one-fourth of pediatric rheumatologists have decreased time in clinical care in the preceding 5 years, with an average reduction of 32.2 percent.  The primary reason reported was obtaining salary support from a research source (39.3 percent).  Many providers also reported decreasing their patient care time because another pediatric rheumatologist joined their practice (32.1 percent); they changed employers or career (17.9 percent), or other reasons (25.0 percent).