BACKGROUND AND SIGNIFICANCE

Chronic obstructive pulmonary disease (COPD) is now the third leading cause of death nationally.1 Tobacco cessation has been shown to be the most important intervention to slow the progression of disease.2 Unfortunately, up to one-third of patients admitted to the hospital for a COPD exacerbation continue to smoke.3,4 Hospitalization is a potentially effective time to provide evidence-based treatment for tobacco cessation.5 Patients are already in a tobacco-free environment with access to nurses, physicians and pharmacists able to facilitate treatment. However, the number of patients treated for tobacco cessation following hospitalization is very low.6

The reasons underlying the low rate of treatment for tobacco abuse among hospitalized patients are largely unknown, and likely reflect a complex interaction of patient and clinician behaviors, as well as the results of tobacco control efforts on the part of hospitals. Studies of the treatment of tobacco abuse in the outpatient setting have demonstrated significant disparities in the provision of medications for tobacco cessation by age, race,7 sex, presence of psychiatric disorders,8 tobacco-related diagnoses,9,10 and a number of other patient characteristics. COPD frequently coexists with other health disparities,11 and may itself be a risk factor for health-related disparities. Little data exists as to whether the disparities in treatment seen in the outpatient setting persist in the inpatient setting, and how they might interact with inpatient tobacco control efforts. While there are a number of studies examining the use of inpatient tobacco control measures, there is less data on the association of these measures with treatment after discharge. It is therefore important to examine the interplay of inpatient tobacco control measures in context with other patient characteristics that may influence treatment for tobacco addiction after discharge.

The goals of this study were to assess the existence of disparities in treatment with tobacco cessation medications following discharge for exacerbation of COPD, examining a range of patient characteristics, and to assess the association of inpatient tobacco control processes with post-discharge treatment for tobacco use in the context of these patient characteristics.

METHODS

Design, Setting and Participants

We conducted a cohort study of current smokers discharged from hospital for a COPD exacerbation within the Veterans Affairs (VA) Veterans Integrated Service Network (VISN)-20. This study was approved by the VA Puget Sound Health Care System Institutional Review Board (#00461).

Data Source

We utilized information from the VISN-20 data warehouse that collects data using the VA electronic medical record, commonly utilized for research purposes.12–14 The warehouse obtained information on demographics, pharmacy records of medications dispensed, hospital and outpatient diagnoses, and dates of death. In addition, we utilized “health factors,” electronic entries coded by staff at the time of an encounter to describe patient health behaviors, and the text of chart notes that were available for electronic query and review.

Study Cohort

We identified all patients aged ≥ 40 hospitalized within the VISN-20 between 2005 and 2012 with a primary discharge diagnosis of COPD based on ICD-9 codes (491, 492, 493.2, and 496), or a diagnosis from the admission note indicating an exacerbation of COPD. Specific query terms for the admission diagnoses were selected based on consensus among several clinicians as appropriately indicating COPD exacerbation. We limited inclusion to current smokers aged ≥ 40 years to improve the specificity of COPD diagnosis, excluding patients who died during the initial admission or had no smoking status identified.

To establish tobacco status, we built on previously developed and validated methodology15 and performed truncated natural language processing, identifying phrases in the medical record that reflected patients’ tobacco status and querying all notes from the day of admission up to 6 months prior. We manually examined these records to confirm the tobacco status. If no status was indicated in the notes, we used the status encoded by the most recent health factor. This combined method allowed us to identify a baseline status for all but 77 of the 3580 patients admitted for COPD.

Outcome and Exposures

In order to assess prescriptions associated with the hospitalization, the primary outcome was the dispensation by the pharmacy of any approved medication to aid in smoking cessation within 48 hours of discharge, including: nicotine patch, short-acting nicotine replacement therapy (NRT), varenicline and buproprion.

Potential predictors were assessed administratively in the year prior to the index date or during the index hospitalization. We examined a range of individual characteristics and markers of tobacco control efforts, including: demographics (age, sex, race), psychosocial characteristics (history of drug or alcohol abuse, homelessness, depression, psychosis, or post-traumatic stress disorder [PTSD]),16 smoking-related comorbidities (Charlson comorbidity index,17 history of coronary artery disease, congestive heart failure, hypertension, stroke, or lung cancer), markers of COPD severity (prescription for steroids in the past year, mechanical or noninvasive ventilation during hospitalization, count of controller inhalers prior to admission), and inpatient tobacco control processes (treatment with smoking cessation medications as an inpatient, nurse-based counseling prior to discharge). Counseling at discharge was determined by nurses recording completion of a discharge process focused on smoking cessation. The presence of comorbid and psychiatric conditions were assessed using a validated method utilizing ICD-9 codes.16

Statistical Analysis

We used Stata 13 (College Station, TX) software. Chi-squared tests and t-tests were used to assess bivariate associations. We built a logistic regression model using methods as described in Applied Logistic Regression Third Edition. 18 In the initial model, we included all variables that attained p ≤ 0.2 in bivariate models containing each predictor and the outcome of interest, clustered by site, with robust standard errors. We then created another preliminary model dropping all variables with p > 0.1 in the initial multivariable model, retaining key demographic variables (age, sex, race), regardless of significance. We examined the β coefficients to determine whether a 20 % change was found when comparing this to the initial full model. We then re-introduced individual variables to the model to try to account for this variation in a parsimonious manner. Nested models were compared with robust Wald tests. Finally, each excluded variable was once again added back to the model to ensure that no significant variables had been excluded, yielding the final model. Model specification was assessed using the linktest function. The area under the curve of the final model was 0.77, with 84.1 % of subjects correctly categorized. The final regression equation is available in the Online Appendix 1.

Sensitivity Analyses

Buproprion has dual indications for smoking cessation and as an antidepressant. Given that there may have been misclassification in the exposure and outcome related to the use of buproprion in this capacity, we performed sensitivity analyses excluding dispensation of buproprion at discharge or during the hospital stay. In addition, due to concerns that the inclusion of subjects with an admission diagnosis of COPD may have created a less specific cohort of subjects with COPD, we performed sensitivity analysis repeating the logistic regression model building exercise among only those subjects with a primary discharge diagnosis code of COPD.

RESULTS

We identified 1511 patients who met our inclusion criteria. Of these, 255 (16.9 %) patients were dispensed a tobacco cessation medication within 48 hours of discharge. Overall, patients were primarily older white males. In unadjusted analyses, patients who were dispensed a medication were on average younger and more likely to be white, with a lower mean Charlson index. Patients who were dispensed a medication were less likely to carry a diagnosis of stroke, congestive heart failure, hypertension, or coronary artery disease. Patients who were dispensed a medication at discharge were more likely to have been treated with tobacco cessation medications as an inpatient and to receive documented brief counseling from the nurse as part of the discharge process. (Table 1)

Table 1. Characteristics of Veterans Dispensed or Not Dispensed Medications for Tobacco Addiction Within 48 Hours of Discharge Following Exacerbation of COPD*
Full size table

Use of Medications

Overall, 17.5 % (n = 264) of patients were treated with a tobacco cessation medication while inpatient. During the inpatient stay, the majority of those treated received the nicotine patch or buproprion, with minimal use of other tobacco cessation medications. Less than half of subjects treated with medications as an inpatient were dispensed medication after discharge (44.7 %, n = 114). Following discharge, of those patients who received a medication (n = 255), the majority (74.1 %, n = 189) were dispensed monotherapy with the nicotine patch, with the remainder receiving other therapies including combination therapy (Figs. 1 and 2).

Figure 1.
figure 1

Results of cohort selection among patients admitted to the Veterans Affairs Veterans’ Integrated Service Network 20 for exacerbation of COPD

Full size image
Figure 2.
figure 2

Among Veterans hospitalized for exacerbation of COPD who were treated with medications, choice of tobacco cessation medications provided 1) within 48 hours of discharge and 2) during hospitalization. NRT = nicotine replacement therapy

Full size image

Disparities in Treatment with Tobacco Cessation Medications

In adjusted analyses, several demographic characteristics were associated with decreased odds of receiving medications. For each year older, the odds of receiving medications at discharge declined by approximately 4 % (OR 0.96, 95 % CI 0.95–0.98). In comparison to white patients, black patients had significantly lower odds of receiving medications at discharge (OR 0.34, 95 % CI 0.12–0.97). Patients with a history of psychosis were less likely to be dispensed medications (OR 0.40, 95 % CI 0.31–0.52). Among patients with higher comorbidity index, each additional point was associated with lower odds of treatment (OR 0.89, 95 % CI 0.82–0.96). Finally, several comorbidities and markers of COPD severity were associated with decreased odds of receiving medications, including hypertension, and treatment with steroids in the past year (Table 2).

Table 2. Predictors of Treatment with Tobacco Cessation Medications Within 48 Hours of Discharge for Exacerbation of COPD* n tot = 1511
Full size table

Association of Tobacco Control Processes with Receipt of Medications

Patients who underwent documented, nurse-provided counseling prior to discharge were significantly more likely to be dispensed tobacco cessation medications after discharge in comparison to those who were not (OR 3.08, 95 % CI 2.02–4.68). The strongest predictor of receipt of medications after discharge was treatment with tobacco cessation medications as an inpatient (OR 5.95, 95 % CI 3.19–11.11). (Table 2)

Sensitivity Analyses

Given the possibility of misclassification of patients treated with buproprion for depression as patients treated for tobacco addiction, the analysis was repeated excluding treatment with buproprion from the outcome. We found a similar relationship with inpatient treatment and post-discharge treatment (OR 6.17, 95 % CI 3.18–12.01), with no significant differences in the other predictors. In addition, when the model building was repeated among only those subjects with a primary discharge diagnosis of COPD (n = 1087), there were no significant differences in inference or magnitude of the observed effects.

DISCUSSION

In this large cohort of Veteran smokers discharged following an exacerbation of COPD, we found a low rate of treatment for smoking cessation with medications, both during the hospitalization and following discharge. Black race, older age, psychiatric disease, and medical comorbidity were associated with fewer tobacco cessation medications dispensed at discharge, while the receipt of inpatient tobacco control processes was strongly associated with more tobacco cessation medications dispensed.

The proportion of patients treated with medications for nicotine withdrawal during the inpatient hospital stay was very low, though similar to previous studies examining the use of these medications among smokers admitted to the hospital.19–21 Most smokers with COPD have significant tobacco dependence.22,23 Untreated patients are therefore likely suffering from significant withdrawal symptoms during their stay,24 or continuing to smoke while hospitalized,25 neither of which are desirable. However, we found that patients who were treated with medications for nicotine dependence during their stay had much higher odds of being dispensed cessation medications at discharge. A similar study of hospitalized patients demonstrated an increase in the use of NRT after discharge among patients treated with NRT as inpatients.26 Collectively, these data suggest that initiating treatment while in hospital may be an important intervention.

Using these medications while in the hospital may reassure patients that these cessation aids are safe and effective for them individually, and give them personal experience to draw upon when leaving the hospital. We also found a higher odds of treatment among patients who received nurse-delivered pre-discharge counseling. Although there is concern about the effectiveness of this low-intensity counseling intervention alone to produce sustained quit rates,27,28 an associated increase in the use of pharmacotherapy for tobacco cessation may prove to be effective, and is supported by our results. Pairing this nurse-based intervention with other processes that facilitate treatment, such as changes to electronic discharge order sets, may further improve the use of medications.

It is unclear what barriers exist to providing treatment to these patients, whether at the patient, provider, or system level. In terms of the use of the “5A’s” of tobacco cessation,29 all of our subjects were identified as smokers in the medical record (Ask), and the majority underwent documented brief pre-discharge counseling (Advise), but few were ultimately provided with tobacco cessation medications (Assist). In general, the majority of hospitalized smokers indicate an interest in quitting,30 which is at odds with the low level of treatment.

Several disparities in treatment that have been identified in the outpatient setting appear to persist at discharge from an inpatient setting. Black outpatients less frequently receive advice to quit, and subsequently receive medications less often than white patients.31,32 A prior study of patients admitted for myocardial infarction indicated that black patients are less likely to be advised to quit smoking,7 suggesting that this disparity in treatment exists in other inpatient settings. Although the number of black subjects in our study was low, the difference in the odds of treatment is striking. Other studies indicate that, despite being at high risk for tobacco-related conditions, older patients and those with worse health status are counseled and treated less frequently,33–35 though this varies somewhat by setting. It is possible that our findings are explained by that patients with COPD may lack the motivation to make a quit attempt or prefer to quit without medications, and that these characteristics vary by age and race. However, it is possible that these findings represent bias on the part of physicians.36

The reasons that underlie the variations in the use of medications for tobacco cessation for patients who are older with more comorbidities are unclear, but may reflect safety concerns and fatalism on the part of patients and physicians. Though multiple studies have indicated that these therapies are safe to use in patients with a variety of medical conditions,37,38 many patients have erroneous beliefs that nicotine replacement is dangerous or ineffective,39,40 and these beliefs may be more common among patients who are older and sicker. Physicians and patients who are faced with continued tobacco use by a patient of advanced age or with significant medical comorbidity may feel fatalistic about the success and benefits of quitting.41 This may in turn contribute to lower rates of treatment. Finally, patients who are older and sicker have more medical conditions competing for physician time and attention, which can limit the time available for discussion of tobacco use. The use of dedicated counselors, rather than dependence on physicians, may help alleviate this difficulty.

The lower observed odds of treatment among patients with a history of psychosis was unsurprising. Psychiatric patients have extremely high rates of tobacco use compared to the general population,42 and suffer from well-documented difficulties in trying to quit.43 Data indicate that psychiatric patients state an interest in quitting that is similar to smokers overall, but are unlikely to receive medications to treat nicotine addiction.44 This may stem from providers’ perceptions that treating these patients is futile or may worsen their psychiatric symptoms.45 Therefore, this group may need tailored interventions to ensure that they are adequately treated for tobacco abuse.

This study has several limitations related to the observational nature of the data. We know whether patients received treatment during or after admission, but we do not know the reasons behind lack of treatment. We do not have information on patient motivation, level of addiction, or whether medications were offered by hospital staff and refused. Patients were identified administratively, and data was drawn from administrative records and chart notes that were not documented for research purposes. Measurement of tobacco status was self-reported as part of routine care, though prior data indicates this form of assessing smoking status in from healthcare interactions is generally valid.46,47 Moreover, we examined self-reported smokers, and social desirability more often leads smokers to report being non-smokers. If patients purchased nicotine replacement or obtained prescription medications outside of the VA system, this would not have been captured. We feel this is likely to be rare among patients admitted to a VA facility due to financial incentives,48 and we have excellent capture of medications provided within the VA system. Finally, although representative of the Veterans who seek care in the VISN-20, there were very few women and Hispanic patients included, which limits the ability to generalize to these groups. Women are commonly found to suffer from disparities in tobacco treatment,34,49 which would have been poorly captured in our study.

This study has several strengths. We were able to assess the provision of tobacco cessation treatment within an entire VISN, comprised of varied hospital settings including academic and rural centers. We included all smokers with a COPD exacerbation, regardless of motivation to quit. All data collection occurred after the initiation of the first JCAHO quality measures surrounding tobacco treatment in 2004, when tobacco treatment among inpatients became a priority. Finally, we have excellent completeness of pharmacy records during the hospital stay and at the time of discharge.

In summary, few patients were dispensed medications to treat tobacco addiction among this cohort of Veteran smokers discharged following exacerbation of COPD. Although our study was performed in a VA setting, similar disparities in treatment have been found in a wide variety of settings, including nationally representative samples,9 ambulatory settings,10 and other groups of hospitalized patients,6,7 suggesting that these results may be generalizable to other hospital practice settings. In addition, while our subjects had high rates of mental illness, this is also typical of the population of smokers50,51 and the population of patients with COPD.52,53 We found a strong association between inpatient tobacco control processes and the provision of pharmacotherapy for smoking cessation after discharge. This suggests that mechanisms to promote guideline adherent tobacco control policies during hospitalization can have beneficial effects on use of these medications after discharge. At a system level, our data suggest that hospitals may implement novel treatment approaches, such “opt out” strategies for treatment of tobacco cessation treatment while hospitalized. Other interventions may focus on the use of integrated electronic order sets for tobacco control, dedicated inpatient tobacco counselors, and education initiatives for patients and staff. Further research is needed to help improve the delivery of tobacco cessation medications and counseling to inpatients, and to alleviate the observed disparities in treatment.