Spinal Manipulation Services vs. Prescription Drug Therapy for Long-term Care

Research Design

This description applies to project specific aim 2.

Overview of Design and Methods: FFS Medicare beneficiaries will be surveyed and compared by cohort for differences in response. Subjects will be randomly sampled from each of the four cohorts (A, B, A2B \& B2A) identified in SA1.

Survey Instrument and Question Formulation: A validated survey instrument will be used to collect subjective patient data. The survey will include questionnaires intended to evaluate 1) self-reported QOL, 2) satisfaction with care, and 3) beliefs regarding back pain and its treatment. For assessment of self-reported QOL a version of the SF12 (Ware '96) will be used and modified to account for time lapse between treatment for LBP and administration of the survey, and to elicit treatment-specific responses The SF -12 has been previously validated for measuring QOL among elderly patients, (Jakobsson '07) and patients using prescription drugs. (Naveiro-Rilo '14) For assessment of satisfaction with care a 0-10 numeric scale will be used. For assessment of beliefs a modified version of the validated LBP Treatment Beliefs Questionnaire will be used. (Dima '15). A pre- test of all the survey instruments will be conducted by distributing a prototype to a sample of 100 Medicare beneficiaries seen for LBP at the SCU Health Center in Whittier, CA. Responses to the survey pre-test will inform any need to edit the questionnaire for comprehensibility and ease of use, thus helping to ensure instrument face validity. For ease of comprehension by older subjects, the survey questions will be printed in large font and will be carefully worded to be brief, unambiguous, and free of bias.

Survey Methods: ResDAC will initiate contact with random samples of beneficiaries who meet criteria for inclusion in the cohorts identified in SA1. The Beneficiary Contact Service has reviewed the survey plan with the PIs and provided an official cost estimate for this service, which is routinely provided by the BCS with strict attention to patient protection. Initial contact will be in the form of a Beneficiary Notification Letter, signed by the CMS Privacy Officer. The letter will alert beneficiaries to the opportunity to voluntarily participate in a healthcare survey. Recipients will be informed that they may decline participation via enclosed reply forms, and will be given phone numbers to call CMS personnel for additional information. After three weeks, the Beneficiary Contact Service will supply the investigators with contact information for eligible beneficiaries (those who did not decline to participate). The survey will be commenced by mailing the printed survey, cover letter with informed consent form based on NCCIH guidelines, and a postage-paid return envelope with detailed information about the survey. Participants can contact study personnel if they have any questions. Follow up by phone after every two weeks or as needed will be used to increase the rate of response.

Outcomes Measurement and Statistical Analysis: Testing will be done for between-cohort differences in self-reported QOL, satisfaction with care, and beliefs about treatments for LBP. Survey responses between the four groups will be compared using Pearson chi-square tests and ANOVA. Demographic characteristics [e.g. sex as a biological variable, and age - because age-related cognitive decline can affect survey responses (Wolinsky '15)] will be controlled using linear regression for continuous survey items, proportional odds logistic regression for ordinal items and multinomial regression for categorical items. In the multivariable regression models equivalence between the four groups will be evaluated using likelihood ratio or Wald tests. Given the multiplicity of testing (e.g. multiple groups for multiple items), type I error inflation will be considered through used of Bonferonni corrections or approaches for false discovery rates. Parametric tests of numeric data can be used to yield unbiased answers when analyzing Likert scale responses. (Norman '10)