Author Response

We thank Freburger, Mielenz, and Fritz for their insightful commentaries on our article.¹ Both commentaries^2,3 help to situate this research within the larger context of efforts stimulated by the Institute of Medicine (IOM) to make health care providers more accountable for clinical quality.^4,5 The IOM has recommended development of incentives that encourage quality improvement by rewarding providers’ performance, which would realign financial incentives in health care delivery.⁵ Although some administrators in today's reimbursement system realize that they can deliver care at lower cost by using an extender of care, such as a physical therapist assistant instead of a physical therapist, our findings suggest that greater use of physical therapist assistants negatively affects the quality of health care delivery. We believe that research such as ours as well as initiatives such as pay-for-performance^5,6 or value-based purchasing⁷ strategies will encourage administrators to support clinicians in developing models of care designed to produce good outcomes efficiently.

As the commentaries illustrate, our report leaves many unanswered questions and unresolved issues. Perhaps most importantly, we must acknowledge that our study did not include process measures to indicate whether or not the care delivered was evidence-based (ie, was consistent with professional knowledge). Data suggest that patients with low back pain syndromes have better outcomes with skillful physical therapy clinical decision making and application of specific clinical skills, such as use of patient classification^8,9 and clinical prediction rules.^10,11 We agree with Fritz in that we would expect care that is consistent with current professional knowledge (ie, as assessed via process measures) would be an essential aspect of quality assessment. Although our study had no direct measure of “care consistent with current professional knowledge,” we suggest that, in our study, this type of care can be considered a latent variable that, although not directly observed, can be inferred from the physical therapist assistant utilization variable.

We believe that it will be beneficial for future researchers to develop process measures for care that are consistent with current professional knowledge. These process measures could be used to assist in the evaluation of quality of care in physical therapy. Such measures could be used to assess the use of standardized evidence-based patient evaluation and management for patients with certain conditions. One example that has already been used in the literature pertains to the compliance with guidelines for postacute care of patients with stroke.^12,13

We also concur with Fritz's comments that our categorization of physical therapy clinic settings as hospital-based or “other” did not capture important administrative elements that may be related to quality. This comment points to 2 limitations: the inherent limitations of performing secondary analyses of pre-existing data and the difficulty categorizing types of clinics. Even if the FOTO data contained more extensive levels of the type of clinic in which the patient was treated, the validity of that variable could be questioned, given the difficulty of categorizing types of clinics. Although Fritz suggests use of electronic medical records as a way of assessing the benefits of structural aspects of patient management, we are unaware of outcomes data collection systems that integrate electronic medical record and electronic outcomes collection for the purpose of assessing clinical practice improvement where outcomes are integrated with patient demographic, process, and structural variables outside of the Maccabi Healthcare Services in Israel.¹⁴

As integrated data collection systems emerge, future analyses could examine the relationship between process and structural measures, as Fritz suggests. In this light, we recognize the need to extend our research to understand process and structural factors related to the quality of physical therapy delivery in other settings such as inpatient hospitals, nursing homes, and home care. Thus, we believe that future research would be facilitated by a large, prospectively collected, uniform data set for physical therapy across settings that contain data shown to meet the appropriate mathematical and clinical assumptions necessary for equating outcomes measures.¹⁵

We believe that future research modeled on prior mixed-methods research on expert therapists^16,17 could be done to compare average, high-quality, and low-quality clinics. Such research might begin by using quantitative methods to profile clinics by quality and then purposefully sample clinics and study them using qualitative methods. This work would enable an understanding of similarities and differences in clinic culture, use of electronic medical record use, quality monitoring and improvement initiatives, and the process of delegation of care and supervision of physical therapist assistants.

Freburger and Mielenz have suggested that the FOTO outcome measure may have been less sensitive than condition-specific measures such as the Roland-Morris Disability Questionnaire and the Oswestry Disability Index and that this may have biased our findings toward the null. Although head-to-head comparison of sensitivity of these measures has not, to our knowledge, been conducted, we do not believe that this is a central weakness of our approach because the responsiveness of the FOTO outcome variable in patients with low back pain has previously been documented as strong.^6,16 We do suspect, however, that there are additional aspects of successful rehabilitation and hallmarks of quality practice that would not be fully captured by any discharge health-related quality-of-life measure. For example, most recognize that one of the primary goals of physical therapy is to enhance patients’ self-efficacy (ie, confidence in managing their own health). This is clearly an important goal of therapy in low back pain where recurrence is likely and the patient needs problem-solving skills to handle minor setbacks. Thus, we believe that further research to develop and test different quality measures in physical therapy is needed.

We agree with Freburger and Mielenz that future research should evaluate other measures such as episode costs and units of Current Procedural Terminology codes rather than visits. Although we think that using number of visits as a measure of utilization has high face validity, this method does not take into consideration how therapy is billed or the associated costs of other services a patient with low back pain might use. FOTO does not collect data on Current Procedural Terminology unit utilization, so this type of analysis was not possible. However, we believe that future research evaluating the relationship between costs and outcomes of physical therapy is particularly important in light of the finding that hospital-based outpatient physical therapy is costlier than office-based care.¹⁸ Thus, we recommend that future studies on outcomes also examine costs and methods of billing for physical therapy services.

We thank Freburger, Mielenz, and Fritz for their thoughtful comments and contribution to the discussion regarding quality measurement in physical therapy. We hope to see increased attention to and interest in this area of research reflected in this and other physical therapy journals, in the American Physical Therapy Association's clinical research agenda, and, ultimately, in the funding priorities of institutes and foundations.

• American Physical Therapy Association