Assessment of accuracy of data obtained from patient-reported questionnaire (PRQ) compared to electronic patient records (EPR) in patients with lung cancer.

Abstract

40 Background: Cigarette smoking, alcohol consumption and co-morbidities are important determinants of health in lung cancer patients. The gold standard for obtaining accurate data is PRQ. The purpose of this study is to ascertain the accuracy of abstracting health-related behaviour data from retrospective chart review compared to data directly obtained from PRQ in a lung cancer patient population. Methods: 731 lung cancer patients completed a PRQ related to lifetime tobacco use, alcohol consumption and co-morbidity. Relevant smoking, alcohol and co-morbidity data was collected independently from EPR. Results: Ever/never status for smoking showed almost perfect agreement (k=0.95) between PRQ and EPR and surpassed all other health behavioural measures and co-morbidity agreement values. Both the sensitivity and specificity were high (0.94 and 0.99 respectively). The calculation of pack-years from EPR and PRQ showed substantial agreement (k=0.77); However, categorizing the smoking status into current/ former / never, resulted in moderate agreement (k=0.46). Alcohol ever/ never status agreement was moderate (0.43) with high sensitivity (0.90) but low specificity (0.50). Agreement for co-morbidities varied by condition showing moderate to substantial agreement for hypertension (K=0.57), heart attack (K=0.80) and diabetes (K=0.76) while fair to slight agreement (K<0.4) was seen in the others. Specificity was 0.86 or higher for co-morbidity conditions and was consistently higher than the sensitivity. Conclusions: EPR may be used as a reliable surrogate to PRQ in determining ever/never smoking status and lifetime smoking exposure. Evaluation of current/former/never smoking status and alcohol consumption is best determined by PRQ. Diabetes, hypertension and heart attack are more accurately reported in the PRQ than other co-morbidities. Patients tend to report absence of a medical condition more accurately than the presence of it. Missing EPR data related to smoking pack years, alcohol consumption and lung co-morbidities is concerning and suggests more synoptic reporting by physicians would improve opportunities for research.