7/11/06

Two research posters using FOTO data were presented at the recent 2006 AcademyHealth Annual Research Meeting in Seattle, WA.

Dr. Hart presented findings from the recent Centers for Medicare and Medicaid funded pay-for-performance feasibility study. Co-authored by Jerome B. Connolly, Kathy Scott and Mark W. Werneke, the poster entitled, “Development of an Outcomes-Based Pay-for-Performance Process for Outpatient Physical and Occupational Therapy”, highlighted the major findings of the retrospective analyses of data from 189,088 adult patients with orthopedic or neurological impairments. The analyses were used to develop the statistical models that were the mathematical foundation of FOTO’s new risk-adjusted Value Purchasing Payment Algorithm© that could be used to reimburse providers based on the functional improvement of their patients and the number of treatment visits used. The abstract is currently posted for review at  http://www.academyhealth.org/2006/abstracts/PartII/QualityMeasuringReportingRewardingPerformance.pdf.

The authors concluded that the results supported a risk-adjusted pay-for-performance model based on a measure of effectiveness could be developed for patients receiving outpatient therapy. The predictive validity of the model was supported, and a simulation using the model and payment algorithm suggested overall reimbursement could be reduced if payments were dependent on patient outcomes and number of treatment visits used. Of importance, results of the simulation suggested a shift of delivery of therapy 1) away from patients who are not benefiting and are being treated longer than predicted and 2) towards patients who have improving functional status and are treated using the least number of visits appropriate to attain the predicted functional improvement.

These findings could have implications for policy and practice because a value-based purchasing process that is based on clinical outcomes and independent of treatment or provider may encourage providers to use evidence-based rehabilitation efficiently and may allow payers to redistribute resources to patients who are benefiting from treatment. Furthermore, the value-based purchasing process could be an alternative to the annual per beneficiary therapy caps for outpatient rehabilitation under Medicare Part B.

Dr. Resnik’s poster, co-authored by Dawei Liu, Vince Mor, Sharon-Lise Normand and Dennis L. Hart, emphasized the need for sophisticated risk-adjustment before profiling of providers of physical therapy can be accomplished in a reasonable manner. The poster, entitled, “Profiling Providers of Outpatient Physical Therapy”, highlighted the results of sophisticated 3-level hierarchical linear modeling that was used to interpret the relation between measures of structure, process and outcomes and shed light on how providers of physical therapy should be profiled. The sample consisted of 16,281 patients with a variety of lumbar syndromes who were treated in 114 outpatient clinics. The models nested patients within therapists and therapists within clinics. In addition, several risk-adjustment variables were added to the model to control for the effect of important independent variables. Inverse probability weighting was used to control for bias due to missing follow-up data. Further, statistical methods were used to control for patient selection bias into clinics by clinic reputation or referral source. Clinics were profiled by risk-adjusted patient functional status at discharge. The association between clinic quality profile and the effect of staffing and caseload characteristics were tested using logistic regression.

The authors demonstrated that unadjusted (i.e., not risk-adjusted) profiles of clinic quality misrepresented average and lower than average performing clinics as determined by patient self-report of functional status. Therefore, statistical risk-adjustment techniques of functional status were needed to more accurately profile clinics on quality. There were no staffing or caseload characteristics associated with the clinic quality profiles, and number of treatment visits was not associated with clinic quality profiles.

The authors concluded that they could profile clinics on quality using a measure of functional status, but no differences in staffing patterns, clinic size or volume of patients with low back pain syndromes affected clinic quality profiles. Number of treatment visits was not different between clinics profiled on quality. The poster is posted on http://www.academyhealth.org/2006/abstracts/PartII/QualityMeasuringImprovingQuality.pdf.

When taken together, these two studies emphasize the 1) value of the large FOTO data set for advancing the assessment of risk-adjusted measures of functional outcomes used in the development of provider profiling and pay-for-performance methods, 2) need for sophisticated statistical methods to assist in the interpretation of these effectiveness data sets, and 3) need for both large data sets and sophisticated statistical methods if research is to appropriately advance public policy related to payment of outpatient therapy or provider profiling.