Evaluating quality indicators for physical therapy in primary care

Objective. To evaluate measurement properties of a set of public quality indicators on physical therapy. Design. An observational study with web-based collected survey data (2009 and 2010). Setting. Dutch primary care physical therapy practices. Participants. In 3743 physical therapy practices, 11 274 physical therapists reporting on 30 patients each. Main Outcome Measure(s). Eight quality indicators were constructed: screening and diagnostics (n = 2), setting target aim and subsequent of intervention (n = 2), administrating results (n = 1), global outcome measures (n = 2) and patient's treatment agreement (n = 1). Measurement properties on content and construct validity, reproducibility, floor and ceiling effects and interpretability of the indicators were assessed using comparative statistics and multilevel modeling. Results. Content validity was acceptable. Construct validity (using known group techniques) of two outcome indicators was acceptable; hypotheses on age, gender and chronic vs. acute care were confirmed. For the whole set of indicators reproducibility was approximated by correlation of 2009 and 2010 data and rated moderately positive (Spearman's ρ between 0.3 and 0.42 at practice level) and interpretability as acceptable, as distinguishing between patient groups was possible. Ceiling effects were assessed negative as they were high to extremely high (30% for outcome indicator 6-95% for administrating results). Conclusion. Weaknesses in data collection should be dealt with to reduce bias and to reduce ceiling effects by randomly extracting data from electronic medical records. More specificity of the indicators seems to be needed, and can be reached by focusing on most prevalent conditions, thus increasing usability of the indicators to improve quality of care.