Abstract
Background: The Osteopathic Survey of Health Care in America (OSTEOSURV) is a decennial national telephone survey. Its goals are to monitor longitudinal trends in awareness, beliefs, utilization, and patient satisfaction relating to osteopathic physicians and to assess beliefs and attitudes regarding contemporary health care issues in the United States. The questionnaire was validated during the administrations of OSTEOSURV-I and OSTEOSURV-II in 1998 and 2000, respectively. In OSTEOSURV 2010, the contemporary health care issues of interest are patient-centered care and the Patient Protection and Affordable Care Act of 2010.
Methods: The target population was household residents of the United States aged 18 years or older. A total of 10,308 random landline telephone numbers were dialed using a computer-assisted telephone interviewing system to acquire 1000 completed interviews between July 23, 2010, and October 1, 2010. The response, cooperation, and contact rates as defined by the American Association for Public Opinion Research were comparable to those of other national telephone surveys. The survey provides an estimated margin of error no greater than 3% to 4% for both general items and for those relating to the subset of respondents claiming to be aware of osteopathic physicians. Because respondents were older and more likely to be female than referents in the general population, the observed responses will be weighted by age and sex to reflect the US Census estimates for persons aged 18 years or older in 2010.
Discussion: OSTEOSURV 2010 was successfully fielded as the latest national telephone survey relevant to osteopathic medicine and contemporary US health care issues. Data analysis should yield important new findings relating to osteopathic physicians, patient-centered care, and the Patient Protection and Affordable Care Act that may not be readily observed through other national health care data sets. While underrepresented in this survey, which excluded cell phone–only participants, young adult respondents were reflective of their national age referents with regard to health insurance coverage and general health status. Thus, it appears likely that statistical weighting by age and sex of the OSTEOSURV 2010 data will minimize potential bias in estimates of health-related items. Rapidly evolving technology and sociocultural transitions will necessitate changes in the design of OSTEOSURV 2020.
The Osteopathic Survey of Health Care in America (OSTEOSURV) is a decennial national telephone survey. Its main goals are to monitor longitudinal trends in awareness, beliefs, utilization, and satisfaction relating to osteopathic physicians and to assess beliefs and attitudes regarding contemporary health care issues in the general population of the United States. The previous administrations of OSTEOSURV-I in 1998
1 and OSTEOSURV-II in 2000
2 validated the OSTEOSURV questionnaire, including the core anchor items relating to osteopathic physicians and other survey domains.
3
The Osteopathic Survey of Health Care in America complements the large national surveys that provide data sets for secondary analysis in the realm of health services and policy research. Two important advantages of these national surveys are that they are generally representative of the US population and that they are statistically powerful. For example, the National Ambulatory Medical Care Survey has been used to study the practice patterns of osteopathic physicians in delivering family and general medical care as well as preventive medicine services,
4 managing low back pain,
5 and providing medical care at academic health centers.
6 However, there are often 2 major limitations of such national data sets. Most obviously, the research hypotheses are limited to those soluble with the data provided within the survey questionnaires or interviews. Secondly, because the data are frequently collected through physician offices, hospitals, or third-party administrators, the patient perspectives of health care delivery are frequently lacking. The Osteopathic Survey of Health Care in America fills these voids by addressing osteopathic medicine and related contemporary health care issues and by asking patients about their beliefs, utilization, and satisfaction relating to health care.
The contemporary health care issues of interest in OSTEOSURV 2010 are beliefs about patient-centered care and the Patient Protection and Affordable Care Act of 2010 (PPACA). Patient-centered care was highlighted by the Institute of Medicine in 2001,
7 wherein it recommended that health care should be patient centered, in addition to being safe, effective, timely, efficient, and equitable. Six commonly accepted dimensions of patient-centered care are the following: respect for patients' values, preferences, and expressed needs; coordination and integration of care; information, communication, and education; physical comfort; emotional support (eg, relieving fear and anxiety); and involvement of family and friends.
7,8
The PPACA has been described as the most consequential social legislation of our generation, arguably transforming all aspects of health care in the United States by mandating minimum health insurance coverage, providing federal subsidies to qualified individuals to help offset the cost of such mandated coverage, imposing extensive new requirements on the health insurance industry, and placing additional regulations on the practice of medicine.
9 Since its enactment on March 23, 2010, there has been much debate about the positive and negative implications of health care reform, including PPACA implementation and its potentially transformative effect on the US health care system.
The institutional review boards at the University of North Texas and the University of North Texas Health Science Center both approved OSTEOSURV 2010 prior to its administration.
The OSTEOSURV 2010 questionnaire (
1) was designed to be easily understood and completed by respondents having a basic knowledge of the English language. It includes 65 items within 6 domains (
Table 1). Forty items had Likert-scale responses, 24 had categorical responses, and 1 item was open-ended. Five items were used to compute the Medical Outcomes Study Short Form-36 general health score.
10 Screening items with branch points were used to acquire data on the subsets of respondents claiming either to be aware of or to have ever received health care from an osteopathic physician.
The target population for OSTEOSURV 2010 was household residents of the United States who were aged 18 years or older. Random digit dialing was used to generate the survey sample after identifying all landline telephone exchanges in the United States and determining the relevant proportional household estimates. Pre-survey letters were sent by the University of North Texas Survey Research Center to a subset of 5048 households identified using a reverse directory of listed telephone numbers. The intent of these letters was to increase participation by explaining the purpose of the survey and notifying recipients that a telephone interviewer would be calling in about 1 week. In addition, all eligible contacts were promised a $10 gift card redeemable through a national retailer as an incentive if they were to complete the telephone interview. Household residents without a landline (ie, those with only cell phones or no telephone service) were not included in the sampling frame. A total of 10,308 random numbers were dialed by interviewers using a computer-assisted telephone interviewing system.
The intent of OSTEOSURV 2010 was to survey adults using household landlines. Consequently, persons in business or government offices, group quarters, or other non-household settings were excluded from participation. Also excluded from the survey were persons who could not understand and respond in English and persons with physical or mental impairments that precluded participation in the telephone interview.
We determined the disposition of all randomly dialed numbers between July 23, 2010, and October 1, 2010 (
Figure). A total of 5780 numbers were unreachable despite 8 call attempts (on the sixth call attempt, whenever possible, a voice mail message was left to explain the purpose of the call and to provide a toll-free number to complete the telephone interview). Among the 3451 contacts with confirmed eligibility, 1036 (30%) respondents participated in telephone interviews. However, only the 1000 respondents who completed the interview (ie, remained on the telephone until the final question, regardless of whether all preceding questions were answered) will be included in future OSTEOSURV 2010 analyses. The median time required by these respondents to complete the questionnaire was 16 minutes (interquartile range, 4 minutes).
The American Association for Public Opinion Research standard definitions and corresponding equations (
2) were used to compute the response, cooperation, and contact rates
11 for OSTEOSURV 2010, using their Outcome Rate Calculator (version 3.1; American Association for Public Opinion Research, Deerfield, Illinois). Response rate 3, cooperation rate 3, and contact rate 2 were purposely selected to facilitate a comparison of OSTEOSURV with a variety of other national landline surveys conducted and reported by the Pew Research Center
12 (
Table 2).
The 1000 respondents who completed interviews in OSTEOSURV 2010 provided a margin of error no greater than 3.1% at the 95% confidence level for items with no missing responses, and no greater than an estimated 3.3% at the 95% confidence level for items with missing responses. For items based only on the subset of respondents claiming to be aware of osteopathic physicians, the margin of error was estimated to be no greater than 4% at the 95% confidence level. For items based only on the subset of respondents claiming to have ever received health care from an osteopathic physician, the margin of error was estimated to be no greater than 7% at the 95% confidence level.
We assessed the basic sociodemographic characteristics of the 1000 respondents who completed the OSTEOSURV 2010 telephone interview (
Table 3). These respondents were older and more likely to be female than referents in the US general population.
13 The modal age category of respondents was 55 to 64 years, and 58% were female, compared with the corresponding population parameters of 45 to 54 years of age and 51% female. Hispanics were underrepresented in our survey (5%), as compared with the general population (14%).
Statistical weighting of the OSTEOSURV 2010 data set was performed (
Table 3). This involved weighting the observed responses by age and sex to reflect the US Census estimates for persons aged 18 years or older in 2010.
13 Aside from the intended effects relating to age and sex, this statistical weighting modestly increased Hispanic representation in the survey (7%), with small or no changes in the remaining sociodemographic characteristics.
Thus far, the OSTEOSURV 2010 data have been managed, analyzed, and statistically weighted with SPSS version 17.0.3 (IBM SPSS Statistics; Chicago, Illinois). Subsequent data management and analyses will be conducted at The Osteopathic Research Center using the relevant software and statistical techniques for the specific hypotheses to be studied.
The Osteopathic Survey of Health Care in America 2010 is supported by a grant from the Osteopathic Heritage Foundation (Columbus, Ohio). The sponsor provides financial support only and has no role in the design and conduct of the survey; the collection, management, analysis, and interpretation of the data; or the preparation, review, or approval of manuscripts for publication.
The Osteopathic Survey of Health Care in America 2010 was successfully fielded as the latest in a series of decennial national telephone surveys relevant to osteopathic medicine and contemporary health care issues in the United States. Analyses of the data provided by the 1000 respondents who completed the interview should yield important new findings relating to osteopathic physicians, patient-centered care, and the PPACA that may not be readily observed through other national health care data sets.
Since our previous administrations of OSTEOSURV-I and OSTEOSURV-II, many more Americans have acquired household technologies, such as answering machines, voice mail, caller ID, and call blocking, that can be used to screen telephone calls and avoid participating in telephone surveys. However, research suggests that the effects of nonresponse bias are minimal despite decreasing response rates in telephone surveys over time.
14 Although fielded 5 to 14 years later, the OSTEOSURV 2010 response rate was only slightly lower than the response rates typically reported in 80 random-digit dialing surveys conducted by another university-based survey research center between 2000 and 2004
15 and in 114 random-digit dialing surveys conducted by 14 major survey organizations servicing news media and government agencies between 1996 and 2005.
16 Further, the response, cooperation, and contact rates for OSTEOSURV 2010 were comparable to those achieved in a variety of national landline surveys conducted by the Pew Research Center in 2007 and 2008.
12
The exclusion of cell phone–only participants from OSTEOSURV 2010 may raise questions regarding the potential for biased results. About 18% of US households use cell phone–only telecommunications,
13 and it is widely acknowledged that cell phone–only usage is most prevalent in young adults. Thus, the underrepresentation of young adults may be among the most problematic aspects of landline surveys. Only 3% and 8%, respectively, of OSTEOSURV 2010 respondents were in the 18- to 24-year-old and 25- to 34-year-old subgroups. As a historical frame of reference, 13% and 20%, respectively, of OSTEOSURV-I respondents were in the corresponding age subgroups. While exclusion of cell phone–only participants generally has not biased the overall findings of landline surveys, there is evidence of biased estimates for certain variables in young adults.
17
Among persons aged 18 to 25 years, the differences between landline and cell phone–only respondents appear to be much more profound for technology usage, such as e-mailing (17% differential) and text messaging (31% differential), than for health-related issues, such as smoking (1% differential) and health insurance coverage (5% differential).
17 It is quite possible that beliefs and attitudes relating to health care are not yet formulated and entrenched in young adults. Therefore, excluding cell phone–only participants (predominantly younger adults) from landline surveys relating to health care may not materially affect the results. Although they were underrepresented in OSTEOSURV 2010, respondents in the 18- to 24-year-old subgroup were representative of their population age group with regard to health insurance coverage and general health status. We found that 69% in both the unweighted and weighted analyses for this age subgroup reported having health insurance. These findings are similar to those reported by comparably aged landline (67%) and cell phone–only (72%) respondents.
17 The mean Medical Outcomes Study Short Form-36 general health scores of 18- to 24-year-old respondents in OSTEOSURV 2010 were 79 (95% confidence interval, 72-85) and 76 (95% confidence interval, 73-79) in the unweighted and weighted analyses, respectively, compared with the age-specific population norm of 77.
10 The 25- to 34-year-old respondents in OSTEOSURV 2010 were similarly comparable in health status to their age-specific referents. Thus, it appears likely that statistical weighting by age and sex of the OSTEOSURV 2010 data will minimize potential bias in estimates of health-related items.
Hispanics have been substantially underrepresented in telephone surveys, in part because the interviews are usually conducted in English. The OSTEOSURV 2010 sample is not unlike that of an experimental study conducted in 2003, wherein Hispanics represented only 7% of the respondents, although they comprised 12% of the US population.
14 To date, budgetary constraints have precluded administering OSTEOSURV in Spanish or using a dual-frame design to include cell phone–only respondents. Instead, statistical weighting based on the US Census estimates for age and sex in 2010 will be used to analyze the OSTEOSURV 2010 results. However, it is readily apparent that rapidly evolving technology and sociocultural transitions in the United States will necessitate changes in the methodology of OSTEOSURV 2020.