Evaluation of Noncompletion Bias and Long-Term Adherence in a 10-Year Patient-Reported Outcome Monitoring Program in Clinical Routine.

BACKGROUND

Currently there is little knowledge on real-life sustainability of routine patient-reported outcome (PRO) measurement and the representativeness of collected data.

OBJECTIVES

The investigation of routine PRO with regard to noncompletion bias and long-term adher- ence, considering the potential impact of mode of assessment (MOA) (paper-pencil vs. electronic PRO [ePRO]) and patient characteristics.

METHODS

At our department, routine PRO measurement in oncological patients is being done since 2005 using different MOA (paper-pencil assessment until 2011 and ePRO assessment from 2011 onward). We analyzed two different patient groups: patients eligible in both periods (both-MOA group) and patients eligible in only one period (one-MOA group). The primary outcome was PRO noncompletion (100% missing questionnaires). The secondary outcome was poor PRO adherence (>20% missing questionnaires). Multivariate logistic regression models were developed, testing the impact of MOA and patient characteristics on the outcomes in the different patient groups.

RESULTS

Data from 1484 eligible patients were included in the analyses. Most of the patients could be included in PRO assessment at least once. PRO noncompletion rates were clearly higher during paper-pencil assessment (odds ratios between 2.72 and 4.31), as were poor PRO adherence rates (odd ratio 2.23). Analyses of potential bias by patient characteristics showed that male patients had a higher risk of poor adherence. Other factors with significant impact were age, country, and cancer diagnosis, but results were indecisive.

CONCLUSIONS

ePRO increased the feasibility of our clinical routine PRO data for retrospective analyses by increasing completion rates. In general, potential completion bias regarding certain patient characteristics requires attention before generalizing results to the respective populations.