Abstract
Background/Aims Cigarette smoking is an important risk factor for multiple serious health outcomes and has significant consequences for health care management and research. However, the completeness and accuracy of the smoking information in the electronic medical record (EMR) system remain unknown. We assessed the accuracy of diagnosis coding for cigarette smoking in the EMR using self-reported smoking status collected from Kaiser Permanente Southern California (KPSC) male members who participated in the California Men’s Health study (CMHS).
Methods Ever smokers in the CMHS were identified based on their self-reported smoking status in the initial (2002–2003) and/or follow-up (2006) study surveys. Smoking diagnosis coding (SDC) was ascertained based on tobacco-related ICD-9 diagnosis codes, CPT procedure codes, and smoking history from KPSC’s integrated EMR system after the initial CMHS survey date. Sensitivity and positive predictive value (PPV) of the SDC were calculated. Logistic regression was used to study the association between SDC and demographics and clinical characteristics among self-reported ever- smokers.
Results A total of 13,539 (65.9%) of 20,554 self-reported ever smokers from the CMHS had SDC in the EMR. The overall sensitivities were 19%, 58%, and 65% during the time periods of 2003–2005, 2006–2008, and 2009–2010, respectively. The sensitivities were higher among those with higher co-morbidity scores and smoking-related chronic diseases, especially chronic obstructive pulmonary disease [78% (95% CI: 74–82%) in 2009–2010]. The overall sensitivity increased to 92% among self-reported current smokers from the CMHS in 2009–2010. The PPV ranged from 76% to 94% during the study period. In multivariable Logistic regression, increased health service utilization and smoking duration were associated with a higher likelihood of having SDC among self-reported ever-smokers. On the other hand, Asian race and Spanish language preference was linked to a lower likelihood of having SDC.
Discussion We found that completeness of SDC in the EMR increased significantly over time. Of note, Asian men and men with Spanish language preference may be at risk of being under coded for smoking history. Although considerable progress has been made, results highlight the need for greater efforts by the health plan to screen for and code patients’ tobacco use.
