gms | German Medical Science

Introduction: Patient-reported outcome measures (PROMs) have the potential to improve the quality of patient-centered care by providing services and support tailored to the individual’s needs. A concept to assess and quantify patients’ active engagement and their ability of self-managing their health status is therefore crucial in health care practice and research. The Patient Activation Measure-13 (PAM-13) is a 13-item self-administered questionnaire relating to knowledge, skills, confidence and behaviors critical for individuals to take responsibility for their own health [1]. From the 13 items an overall PAM-score is calculated, according to which patients can be categorized into one of four progressive stages of activation [1]. As all PROMs, the PAM-13 is highly susceptible to missing data and analyses can be compromised when missingness is handled inappropriately, leading to biased results or reduced power [2].

State of the art: In Germany, two validation studies of the German PAM-13-D were published by (a) Brenk-Franz et al. [3] and (b) Zill et al. [4] in 2013. They differ with respect to their recommendations on how to handle missing data: (a) only questionnaires with ≥ 7 items were included in the analysis. In case of missingness in the remaining datasets, the total score was divided by the number of completed items and multiplied by 13. The resulting sum raw scores were transformed into natural logarithm and from the logit metric to a standardized metric ranging from 0-100; (b) PAM-questionnaires with < 9 items present were excluded. For patients with less missingness, missing values were imputed using expectation-maximization procedure. The score was then calculated by adding up the raw scores and mapping the value onto a scale of 0–100.

Concept: As no conclusive guidelines on the handling of missing data in the PAM-13-D exist, the objective of our study was to compare four different approaches: (a) and (b) as described above, (c) multiple imputation (MI) by item regardless of the number of missing items, (d) MI at raw sum score level.

Implementation: Real-world data from a controlled longitudinal implementation study (CCC-integrativ) on a 3-month interprofessional complementary and integrative healthcare (CIH) counseling program for oncology patients was used [5]. In analogy to the primary endpoint of the original study, our analysis focused on pre-/post-intervention (T1, T2) PAM-13 data. In a first step, the data was reduced to complete cases, then we mimicked the missingness patterns observed in the original CCC-integrativ with different overall proportions of missingness ranging from 5% to as high as 60%. The simulation process was repeated 100 times for stable results. Different performance measures will be used to measure the accuracy and error of the simulated results with respect to the true values.

Lessons learned: Simulations are currently work-in-progress, first results are expected by the time of the conference. Differences of the approaches will be discussed and conclusions on the best method drawn to provide clinicians with clearer recommendations on how to appropriately account for missing answers in the PAM-13-D.

The authors declare that they have no competing interests.

The authors declare that a positive ethics committee vote has been obtained.