Electronic Health Record Classification of Tobacco Smoke Exposure and Cotinine Levels in Hospitalized Pediatric Patients

A convenience sample of 0- to 17-year-old patients who were hospitalized on a general medical unit between December 2016 and June 2017 in a US children’s hospital and had EHR documentation of positive or negative TSE status were recruited during day hours by study staff. TSE status was assessed by a physician or nurse with 1 of 2 prompts: “tobacco smoke exposure” or “smokers in the home.” Children were classified as having a positive TSE status if either answer was “yes” (n = 100) or a negative TSE status if both answers were “no” (n = 99). Children were excluded if they had a tracheostomy (n = 2) or were smokers (n = 10). Caregiver consent on all children and child assent on children age 11 and older were obtained. Saliva samples were obtained from all children at the time of study enrollment; the mean collection time was 18 hours and 43 minutes (SD 0.68 hours) after admission. Saliva was tested for cotinine by Salimetrics LLC²¹  using enzyme-linked immunosorbent assay (ELISA) techniques; the level of detection was 0.15 ng/mL. Previous research has classified children with salivary cotinine values >1.0 ng/mL as having positive biochemical verification of TSE.^22–24  We assessed the robustness of using the >1.0 ng/mL cut point by building the same models and using different cotinine cut points (ie, >0.70, >0.85, and >1.25 ng/mL), and overall, our results showed consistent findings (see footnote in Table 2). Thus, we have classified children with cotinine levels of >1.0 ng/mL as having positive biochemical verification of TSE, hereafter classified as positive cotinine levels.

Of the 199 participants, 171 had sufficient saliva volume to have cotinine measured and were included in the study. Caregivers reported sociodemographic characteristics, and children’s EHRs were abstracted for the following: past medical history (PMH) and International Classification of Diseases, 10th Revision discharge diagnosis (specifically asthma, bronchiolitis, and pneumonia), number of emergency department (ED) and urgent care (UC) visits, hospitalizations for a 6-month period before and after the index hospitalization, and number of revisits (ie, ED, UC, or hospitalizations) in a 30-day period. This study was approved by our hospital’s institutional review board.

We conducted χ² tests to assess the agreement between the proportions of children with positive or negative EHR classification of TSE status and biochemical verification of exposure or no exposure among all patients and by patients within age categories. Then, we assessed the relationship between sociodemographic and clinical variables and cotinine levels by building a series of crude and adjusted logistic regression models while controlling for age, race, and ethnicity. We also evaluated which demographic characteristics were associated with insurance status and assessed whether insurance status accounted for the variance independent of age. R version 3.3.0²⁵  was used for all statistical analyses, and associations were considered significant at α < 0.05.

The mean age of children in this study was 5.1 years (SD 3.7; range 0.04–17), 53.2% were boys, and 62% had public insurance or were self-pay. Of the sample, 74.9% were white, 18.7% were African American, and 96.5% were non-Hispanic. There were no differences based on child age, sex, race, ethnicity, insurance type, or TSE status between participants who did (n = 171) and did not (n = 28) have saliva analyzed for cotinine. A total of 102 (59.6%) children had positive cotinine levels. Table 1 presents the cross-tabulations of EHR classification of TSE status and positive or negative cotinine levels. Overall, 71% (121 of 171) of EHR classifications were correct given the 1.0 ng/mL cutoff of salivary cotinine indicating exposure. Specificity was 0.77 (53 of 69) and sensitivity was 0.67 (68 of 102). Of the 102 patients with positive cotinine levels, 68 (67%) were correctly identified as TSE case patients in the EHR. Of the 69 patients with negative cotinine levels, 53 (77%) were correctly identified as unexposed in the EHR. The negative predictive value was 0.61 (53 of 87); 61% were confirmed unexposed with negative cotinine levels, and 39% (34 of 87) were misclassified as unexposed. The positive predictive value was 0.81 (68 of 84); 19% (16 of 84) were misclassified as exposed.

Positive cotinine levels indicative of TSE decreased as child age increased, with 76.5% of 0- to 1-year-olds having positive cotinine levels, followed by 2- to 4-year-olds (66.7%), 5- to 9-year-olds (48.6%), and 10- to 17-year-olds (36.6%; Table 2). Because younger children had higher levels of cotinine overall, we examined if the false-negative results were primarily in younger children. Although 52.9% (n = 36) of children 0 to 1 year old were classified as positive TSE status in the EHR, 76.5% (n = 52) of children in this age group had positive cotinine levels. We found similar discrepancies in children 2 to 4 years old; 37.0% (n = 10) had positive TSE status in the EHR, but 66.7% (n = 18) had positive cotinine levels. In children 5 to 9 years old, 40.0% (n = 14) had positive TSE status in the EHR, but 48.6% (n = 17) had positive cotinine levels. However, in older children age 10 to 17 years, 58.5% (n = 24) had positive TSE status in the EHR, but 36.6% (n = 15) had positive cotinine levels.

We examined if there were associations between sociodemographic characteristics of participants or clinical factors and positive cotinine levels (Table 2). Univariate logistic regression models indicated that 0- to 1-year-olds (odds ratio [OR] 5.63; 95% confidence interval [CI] 2.41–13.14; P < .001) and 2- to 4-year-olds (OR 3.47; 95% CI 1.25–9.63; P = .02) were more likely to have positive cotinine levels than 10- to 17-year-olds. Patients with public insurance or who were self-pay were 8.7 times (95% CI 4.31–17.71; P < .001) more likely to have positive cotinine levels than patients with private insurance. A total of 57.3% of children with a PMH of asthma or bronchiolitis had positive cotinine levels. Those with a PMH of bronchiolitis (OR 2.34; 95% CI 1.06–5.20; P = .04), a bronchiolitis diagnosis (OR 3.69; 95% CI 1.52–8.99; P = .004), or asthma and/or bronchiolitis diagnosis (OR 2.62; 95% CI 1.35–5.08; P = .004) were at increased risk of having positive cotinine levels compared with those without these PMH or discharge diagnoses. Patients who had ≥1 ED visit (excluding an ED visit associated with the index hospitalization) within the past 6 months were 2.4 times (95% CI 1.27–4.42; P = .007) more likely to have positive cotinine levels compared with those with no ED visits.

Age, insurance type, and asthma and/or bronchiolitis diagnosis remained statistically significant in the adjusted models (Table 2). While assessing whether insurance status accounted for the variance independent of age, we found that patients with public insurance or self-pay were more likely to be 0 to 1 year old (P = .02), be African American (P = .02), be of other race (P = .05), or have positive TSE status in the EHR (P < .001).

This study tested the accuracy of EHR documentation of hospitalized pediatric patients’ TSE status using cotinine verification. In our study, we found a lack of concordance with EHR classification of TSE status and biochemically verified TSE. Specifically, we found that 39% of the children who were thought not to be exposed to tobacco smoke on the basis of their EHR were actually exposed on the basis of positive cotinine levels. Conversely, we found that 19% of children who had positive TSE documented in the EHR were not exposed on the basis of negative cotinine levels. Our findings were similar to those of other studies in different settings. A study of children with asthma recruited from health departments and asthma clinics²⁶  found poor agreement across all clinics between EHR documentation and parent report, although agreement was higher within asthma specialty clinics (κ 0.410 compared with 0.205). In 1 study of infants and young children in a primary care clinic, reported TSE based on 1 question about exposure to passive TSE was 13% compared with 55% with cotinine-verified exposure.¹⁶  In another study of children hospitalized with asthma, parents were asked 3 questions about potential TSE, and 65% reported no exposure, yet 70% of children had positive salivary cotinine levels.¹³ 

Factors that may lead to discrepancies we found in the EHR compared with cotinine results include a lack of standardized methods to assess TSE via parent reports, the use of an outdated TSE status, underreporting of child TSE, and parents’ lack of awareness of all TSE sources.^10,13,27,28  The latter possibility is important to consider given the many sources of potential TSE in children’s environments of which parents may be unaware.^29,30  We were unable to assess different TSE sources (eg, electronic cigarettes and multiunit housing) and locations of exposure (eg, home, day care, cars, and public areas) because we did not collect these data. Future research should examine these factors among hospitalized pediatric patients to provide further insight into reporting discrepancies. Furthermore, our finding that 19% of children with positive TSE status were actually unexposed on the basis of negative cotinine levels may have been due to the timing of saliva collection. Some saliva samples were obtained later during the hospitalization, which may have resulted in decreased cotinine levels because cotinine has a half-life of 16 to 18 hours.¹²  However, the mean collection time was ∼19 hours with little variation; thus, the cotinine levels of the children in this study population may have been higher if obtained earlier in the hospitalization.

Our results indicate that children age 4 years and younger had higher rates of biochemically validated TSE compared with children age 5 to 9 years and 10 to 17 years. This may be because younger children have decreased opportunities to leave environments where cigarettes are smoked compared with older children. In contrast to other research that reported higher TSE rates in non-Hispanic African American children,^13,16,31  we did not find higher TSE rates in this population. However, consistent with other studies, we found that those with TSE were >8 times more likely to have public or self-pay insurance (a proxy for low income).^13,31,32  Similar to previous studies,^19,20,33  we found that children with a diagnosis of asthma or bronchiolitis were twice as likely to have TSE. However, if PMH of asthma or bronchiolitis alone were used to determine who should be screened for TSE, then 57.3% of exposed children would have been missed. Thus, these findings suggest that screening should be considered in all hospitalized children regardless of their sociodemographic background, PMH, or current diagnoses. Additionally, the high prevalence of TSE in this population is consistent with our previous work in patients in the pediatric ED,^34,35  and these high prevalence rates provide further impetus to provide parental cessation interventions in the pediatric hospital setting. Previous research of parental smokers in the hospital setting indicates that parents are interested in receiving counseling and that after brief counseling, self-reported quitting and TSE behavior changes are encouraging.^36–39  More research is needed to further develop and test the efficacy of these interventions.

Limitations of this study include the use of a convenience sample of participants, which limits generalizability. We assessed cotinine in saliva using the ELISA method, which is not as sensitive or specific as other methods, such as liquid chromatography mass spectrometry⁴⁰ ; however, the use of ELISA to measure cotinine does have good sensitivity and specificity.⁴¹  The screening questions that were used in the EHRs were nonspecific and did not account for many factors that affect children’s TSE levels, such as the type of tobacco or nicotine product and the locations, amount, and frequency of exposure. Future research and quality-improvement projects should expand the TSE screening questions to account for other factors affecting exposure and address tobacco or nicotine product use of any form by parents.

Our findings indicate that it is not adequate to rely on EHR classification or parental reports of exposure. Almost 40% of participants in the current study were misclassified as nonexposed to tobacco smoke. We found that a high proportion of hospitalized children with both respiratory and nonrespiratory PMHs were exposed to tobacco smoke. Given the significant health benefits of a smoke-free environment for hospitalized children, the reliance on EHR classification to determine TSE status may result in missed opportunities for reducing TSE in this vulnerable population. Our findings provide support for the recommendation to adopt a universal approach to TSE-related screening. The regular collection of cotinine assays would greatly reduce the underreporting biases that are prevalent during pediatric health care visits and improve the accuracy of EHR TSE documentation.^13,18,33  If this is not feasible because of costs or other concerns, incorporating standardized and expanded TSE screening and counseling into pediatric hospital settings, regardless of the child’s diagnosis, could significantly decrease TSE and related morbidity in children.⁴