Anne M. Gadomski, MD, MPH

Intervention

The Smoking Cessation service for inpatients began on April 3, 2006. Upon admission, all patients were screened regarding their current smoking status. The nurse asked the patient if s/he currently smoked and then entered the responses into the hospital electronic medical record (EMR). A current smoker was defined as smoking every day or some days within the past 30 days. A roster of newly admitted current smokers was electronically transmitted to the Respiratory Care office daily. Only current smokers received counseling. The Smoking Cessation Specialist (SCS) subsequently saw inpatients within a 24‐hour time frame of admission, except for weekends and holidays. Each patient received 1 to 2 intensive follow‐up counseling sessions during hospitalization. An average of 10 patients per day were seen.

The goal of the inpatient smoking cessation service was to counsel patients on the health effects of smoking, address nicotine withdrawal symptoms, explain the different pharmacotherapies available, advise on how to quit, give self‐help materials, counsel family members, and refer to the New York State (NYS) Smokers' Fax‐to‐Quit program.9 Following the consult, the SCS documented the encounter in the patient's chart, including recommendations for nicotine replacement therapy (NRT) or bupropion (varenicline was not addressed as it was not on the formulary). The chart documentation informed the physician and nursing staff of the intervention and included the date/time, stage of change, and support action taken.

The SCS was part‐time, had nursing training, smoking cessation training,10 and was also trained by the Seton Health Cessation Center in the Butt Stops Here Program.11 She also implemented a performance improvement plan to increase the provision of smoking cessation counseling, increase NRT or bupropion prescriptions to smokers admitted to the hospital, and increase referrals to the NYS telephone quitline through the Fax‐to‐Quit program for outpatient resources and help following hospital discharge. The Fax‐to‐Quit program allows health care providers to refer patients to the NYS quitline via fax, with the patient's signature (patient permission) on the fax to quit form. After hospital discharge, the quitline then contacts the patient at a time that the patient requested.

The SCS visited patients with all admitting diagnoses on the medical, surgical, and special care units who were current smokers. Inpatients admitted to psychiatry, obstetrics, and the intensive care unit (ICU) were not seen by the SCS, except for ICU patients referred by a physician. Inpatients who had short stays or who were admitted and discharged in 1 day or during the weekend were not seen.

The intervention included either a brief 3‐minute to 5‐minute intervention or a more intensive intervention, that required 10 to 20 minutes (18 minutes average). The length of the intervention was determined by how receptive the patient was to the intervention. All interventions began with patient identification, an introduction to the SCS, and an explanation of the purpose of the visit. The SCS then inquired about the patient's comfort level vis‐a‐vis nicotine withdrawal and if s/he was receiving any NRT while in the hospital (NRT on the inpatient formulary included the nicotine patch or gum). If the patient was receptive to counseling, the SCS then began to work through the 5 A's, as described in the 2000 DHHS Clinical Practice Guidelines.12 The 2000 DHHS Clinical Practice Guidelines were used because the 2008 update had not been released at the time this study was initiated in 2006. These include: asking about smoking status, advising on how to quit, assessing readiness to quit, and assisting in arranging treatment options that include pharmacotherapy, counseling, as well as referral to the NYS Smokers' Quitline. A workbook was provided to reinforce counseling but was not necessarily used during counseling session. A compact disc (CD) with relaxation exercises5 was provided to those inpatients who were interested in stress reduction. If family members were present, and were also smokers, they were included in the counseling session, if willing. Each patient was offered a referral via the Fax‐to‐Quit program to continue treatment on an outpatient basis.

If the patient was not motivated to quit or declined the consult, the visits were short and focused on the patient's experience with nicotine withdrawal. These patients were also given self‐help materials and, if possible, the relevance of and roadblocks to quitting were reviewed. Patients were prompted to think about why quitting was relevant and often the reason for hospitalization was used to motivate the quitting process.

Upon hospital discharge, the patient's primary care provider was notified of the cessation intervention by a letter from the SCS. The letter described the intervention and stated whether or not the patient agreed to be referred to the Fax‐to‐Quit Program.

Study Participants

Patients were recruited from July 1, 2006 (after the smoking ban went into effect) through June 1, 2008. Inpatients who currently smoked were informed of this study and were asked to sign informed consent to participate after they were seen by the SCS. Current smokers of all admitting diagnoses were recruited into the study. Patients provided informed consent for a telephone interview 6 months posthospital discharge. A written Health Insurance Portability and Accountability Act (HIPAA) release was obtained to allow access to an individual's specific EMR.

A comparison group of inpatients who were also current smokers, but who did not receive the intervention were also contacted six months after hospitalization. Reasons for not receiving the intervention included the fact the SCS was part‐time and also took a leave of absence during the study and therefore could not see all inpatients who currently smoked. Other reasons for not receiving the intervention include too short a stay for the SCS to see the patient or the patient was out of the room for tests or procedures when the SCS was available. These patients provided informed consent to be interviewed 6 months after hospital discharge and HIPAA consent for access to their medical record. Not all inpatients in the comparison group provided written HIPAA release for use of their medical record; therefore, these patients were excluded because their baseline demographic and diagnostic data were missing.

Sample‐size considerations were driven around having adequate numbers of subjects to measure the prevalence of smoking cessation at 6 months post‐hospital discharge with an acceptable degree of precision. Prevalence estimates from previous studies for 6‐month cessation typically range from 20% to 30%, with cessation rates as high as 67% (this estimate applies to postmyocardial infarction patients.)13 For conservative estimation, we used 50% as the 6‐month prevalence of cessation in the current study, which placed binomial variance at its theoretical maximum. In this case, a sample of 300 subjects provides a margin of error of 0.058 for a 95% confidence interval around this point estimate.

Data Sources

The hospital EMR database was used to monitor several components of the program: nursing screening, smoking cessation counseling, and pharmacy dispensing of NRT and bupropion. The screening data were also used to monitor the proportion of current smokers admitted during the study period. Elements of the EMR were used to define the following covariates: patient age, gender, ethnicity, and the primary discharge diagnosis (via International Statistical Classification of Diseases and Related Health Problems, 9th edition [ICD9] codes) and readmission during the six month follow‐up period. Mean length of stay (LOS) was computed. The Elixhauser Comorbidity Index that utilizes ICD9 codes was used for comorbidity risk adjustment.14

Study participants were contacted by phone 6 months posthospital discharge. Data collection began July 1, 2006 and was completed January 1, 2009. The interview focused on self‐reported point prevalence of smoking and 6‐month quit status. The point prevalence for self‐reported abstinence was derived from the question Do you now smoke cigarettes every day, some days, or not at all?15 Self‐reported quit status was derived from question Have you quit smoking since you were discharged from the hospital? In addition, respondents were queried about their number of years smoked, post‐hospital discharge number of quit attempts, and cessation efforts (NRT, self‐help groups, quitline use, etc.). Last, they were surveyed about barriers to cessation (exposure to secondhand smoke, rules about smoking in the home or car), educational level, employment, and health ratings.

To determine the status of those lost to follow‐up, administrative and EMR databases for appointments and follow‐up visits were accessed to determine if the patient was alive during the 6 months between discharge and the follow‐up call. To confirm mortality, we searched the Internet, Ancestry.com, and/or local newspaper obituaries for dates of death for all patients to validate that they had not died during the 6‐month follow‐up. World wide web searches can identify 97% of deaths listed in the Centers for Disease Control and Prevention (CDC)/National Death Index, which is considered the gold standard in epidemiologic studies.16

Analysis

Univariate analysis of all covariates was completed to examine the normal distribution curves for these variables. Bivariate correlation analysis of all the independent variables by study group was performed to assess comparability of the study groups at baseline. The self‐reported cessation outcomes were calculated by dividing the number of patients who said they were not using tobacco or had quit, at 6 months posthospital discharge by the number of individuals in the study group at baseline minus those who had died. Both the intent to treat method, which assumes patients lost to follow‐up were still smoking, and the responder method, which does not include nonresponders in the analysis, were used to adjust the denominators for these outcomes.

Multivariate regression analysis was then used to model receipt of the intervention as predictor of self‐reported quit status adjusted for significant covariates. Statistical significance was defined by a P value of less than 0.05.

Survival analysis was employed to model differences in mortality between the study groups, controlling for any baseline imbalances (eg, comorbidity). Because baseline data were used in this model, the model includes only patients with signed a HIPAA release.

Internal review boards of our hospital and the NYS Department of Health reviewed and approved this study.

Limitations

This study targets a later phase in research progression from hypothesis development, pilot studies, efficacy (empirically supported) trials, effectiveness trials (real‐world settings), and dissemination studies.31 Because this study addresses the effectiveness rather than the efficacy of inpatient smoking cessation counseling, the use of a quasiexperimental rather than randomized controlled clinical trial design led to measured differences in the study groups at baseline. An important imbalance arose in the intervention group that had twice the percentage of patients with cardiovascular‐related discharge diagnoses as the comparison group. While we were able to adjust for these differences in our analysis, there may be unmeasured differences due to the fact that the inpatients were not randomized to the study groups.

The outcomes of this study cannot be attributed to any one component of the intervention (eg, NRT) vs. the combined effect of the inpatient smoking cessation program. The program components were implemented simultaneously in order to maximize synergistic effects; therefore, the effects of program components are difficult to disaggregate.

The results are limited by the validity of self‐report of smoking status. It is well known that research studies which validate smoking status biochemically have lower efficacy (OR, 1.44; 95% CI, 0.992.11) than those that do not validate smoking status (OR, 1.92; 95% CI, 1.262.93).32 Although it was impractical in this effectiveness study to biochemically validate smoking status 6 months posthospital discharge, we have documented a significant difference between the study groups that confirms the direction of the effect, if not the effect size.

Self‐reports tend to underestimate smoking status in population studies33; however, the discrepancy between self‐reported smoking and biochemical measurements among clinical trial participants is small.34 However, a small but significant bias toward a socially desirable response in intervention groups compared to control groups of 3% with carbon monoxide and 5% for cotinine has been documented.35 If social desirability bias is operative in this study, and if we apply the above correction factor of 4% to correct this classification error, then the difference between the intervention and comparison group would be 10 percentage points (40% in the intervention group vs. 30% in the comparison group using the intention to treat estimates). That difference is still clinically relevant.

The observed difference between the intervention and comparison group is underestimated because the comparison group was exposed to smoking cessation as well both at the time of admission and following discharge (Table 4). The comparison group in this study could thus be viewed as a usual care group rather than a control group. That exposure does cloud the measurement of quit rates as the comparison group is contaminated to some degree by exposure to various cessation methods. The impact of this exposure is to reduce the effect size observed in this study or underestimate the effect of the inpatient smoking cessation counseling because the comparison group was exposed other cessation methods, although to a lesser extent.

The social desirability bias inherent in self‐reported smoking status may increase the effect size while the use of comparison group that received usual care may decrease the effect size. Because neither of these biases could be measured in this study, it is impossible to say whether they negated each other.

As with any administrative database, use of EMR as a data source in this study led to missing data that precluded use of certain variables in the analysis. In addition, lack of written and signed HIPAA releases also precluded inclusion of several inpatients, mostly in the comparison group, in the analytic database. However, it is reassuring that the results of the 6‐month survey did not differ significantly when these individuals were included or excluded from a separate analysis of the survey data. Last, our study population is almost 100% Caucasian thus limiting how generalizable the results are to more heterogenous patient populations.

Intervention

The Smoking Cessation service for inpatients began on April 3, 2006. Upon admission, all patients were screened regarding their current smoking status. The nurse asked the patient if s/he currently smoked and then entered the responses into the hospital electronic medical record (EMR). A current smoker was defined as smoking every day or some days within the past 30 days. A roster of newly admitted current smokers was electronically transmitted to the Respiratory Care office daily. Only current smokers received counseling. The Smoking Cessation Specialist (SCS) subsequently saw inpatients within a 24‐hour time frame of admission, except for weekends and holidays. Each patient received 1 to 2 intensive follow‐up counseling sessions during hospitalization. An average of 10 patients per day were seen.

The goal of the inpatient smoking cessation service was to counsel patients on the health effects of smoking, address nicotine withdrawal symptoms, explain the different pharmacotherapies available, advise on how to quit, give self‐help materials, counsel family members, and refer to the New York State (NYS) Smokers' Fax‐to‐Quit program.9 Following the consult, the SCS documented the encounter in the patient's chart, including recommendations for nicotine replacement therapy (NRT) or bupropion (varenicline was not addressed as it was not on the formulary). The chart documentation informed the physician and nursing staff of the intervention and included the date/time, stage of change, and support action taken.

The SCS was part‐time, had nursing training, smoking cessation training,10 and was also trained by the Seton Health Cessation Center in the Butt Stops Here Program.11 She also implemented a performance improvement plan to increase the provision of smoking cessation counseling, increase NRT or bupropion prescriptions to smokers admitted to the hospital, and increase referrals to the NYS telephone quitline through the Fax‐to‐Quit program for outpatient resources and help following hospital discharge. The Fax‐to‐Quit program allows health care providers to refer patients to the NYS quitline via fax, with the patient's signature (patient permission) on the fax to quit form. After hospital discharge, the quitline then contacts the patient at a time that the patient requested.

The SCS visited patients with all admitting diagnoses on the medical, surgical, and special care units who were current smokers. Inpatients admitted to psychiatry, obstetrics, and the intensive care unit (ICU) were not seen by the SCS, except for ICU patients referred by a physician. Inpatients who had short stays or who were admitted and discharged in 1 day or during the weekend were not seen.

The intervention included either a brief 3‐minute to 5‐minute intervention or a more intensive intervention, that required 10 to 20 minutes (18 minutes average). The length of the intervention was determined by how receptive the patient was to the intervention. All interventions began with patient identification, an introduction to the SCS, and an explanation of the purpose of the visit. The SCS then inquired about the patient's comfort level vis‐a‐vis nicotine withdrawal and if s/he was receiving any NRT while in the hospital (NRT on the inpatient formulary included the nicotine patch or gum). If the patient was receptive to counseling, the SCS then began to work through the 5 A's, as described in the 2000 DHHS Clinical Practice Guidelines.12 The 2000 DHHS Clinical Practice Guidelines were used because the 2008 update had not been released at the time this study was initiated in 2006. These include: asking about smoking status, advising on how to quit, assessing readiness to quit, and assisting in arranging treatment options that include pharmacotherapy, counseling, as well as referral to the NYS Smokers' Quitline. A workbook was provided to reinforce counseling but was not necessarily used during counseling session. A compact disc (CD) with relaxation exercises5 was provided to those inpatients who were interested in stress reduction. If family members were present, and were also smokers, they were included in the counseling session, if willing. Each patient was offered a referral via the Fax‐to‐Quit program to continue treatment on an outpatient basis.

If the patient was not motivated to quit or declined the consult, the visits were short and focused on the patient's experience with nicotine withdrawal. These patients were also given self‐help materials and, if possible, the relevance of and roadblocks to quitting were reviewed. Patients were prompted to think about why quitting was relevant and often the reason for hospitalization was used to motivate the quitting process.

Upon hospital discharge, the patient's primary care provider was notified of the cessation intervention by a letter from the SCS. The letter described the intervention and stated whether or not the patient agreed to be referred to the Fax‐to‐Quit Program.

Study Participants

Patients were recruited from July 1, 2006 (after the smoking ban went into effect) through June 1, 2008. Inpatients who currently smoked were informed of this study and were asked to sign informed consent to participate after they were seen by the SCS. Current smokers of all admitting diagnoses were recruited into the study. Patients provided informed consent for a telephone interview 6 months posthospital discharge. A written Health Insurance Portability and Accountability Act (HIPAA) release was obtained to allow access to an individual's specific EMR.

A comparison group of inpatients who were also current smokers, but who did not receive the intervention were also contacted six months after hospitalization. Reasons for not receiving the intervention included the fact the SCS was part‐time and also took a leave of absence during the study and therefore could not see all inpatients who currently smoked. Other reasons for not receiving the intervention include too short a stay for the SCS to see the patient or the patient was out of the room for tests or procedures when the SCS was available. These patients provided informed consent to be interviewed 6 months after hospital discharge and HIPAA consent for access to their medical record. Not all inpatients in the comparison group provided written HIPAA release for use of their medical record; therefore, these patients were excluded because their baseline demographic and diagnostic data were missing.

Sample‐size considerations were driven around having adequate numbers of subjects to measure the prevalence of smoking cessation at 6 months post‐hospital discharge with an acceptable degree of precision. Prevalence estimates from previous studies for 6‐month cessation typically range from 20% to 30%, with cessation rates as high as 67% (this estimate applies to postmyocardial infarction patients.)13 For conservative estimation, we used 50% as the 6‐month prevalence of cessation in the current study, which placed binomial variance at its theoretical maximum. In this case, a sample of 300 subjects provides a margin of error of 0.058 for a 95% confidence interval around this point estimate.

Data Sources

The hospital EMR database was used to monitor several components of the program: nursing screening, smoking cessation counseling, and pharmacy dispensing of NRT and bupropion. The screening data were also used to monitor the proportion of current smokers admitted during the study period. Elements of the EMR were used to define the following covariates: patient age, gender, ethnicity, and the primary discharge diagnosis (via International Statistical Classification of Diseases and Related Health Problems, 9th edition [ICD9] codes) and readmission during the six month follow‐up period. Mean length of stay (LOS) was computed. The Elixhauser Comorbidity Index that utilizes ICD9 codes was used for comorbidity risk adjustment.14

Study participants were contacted by phone 6 months posthospital discharge. Data collection began July 1, 2006 and was completed January 1, 2009. The interview focused on self‐reported point prevalence of smoking and 6‐month quit status. The point prevalence for self‐reported abstinence was derived from the question Do you now smoke cigarettes every day, some days, or not at all?15 Self‐reported quit status was derived from question Have you quit smoking since you were discharged from the hospital? In addition, respondents were queried about their number of years smoked, post‐hospital discharge number of quit attempts, and cessation efforts (NRT, self‐help groups, quitline use, etc.). Last, they were surveyed about barriers to cessation (exposure to secondhand smoke, rules about smoking in the home or car), educational level, employment, and health ratings.

To determine the status of those lost to follow‐up, administrative and EMR databases for appointments and follow‐up visits were accessed to determine if the patient was alive during the 6 months between discharge and the follow‐up call. To confirm mortality, we searched the Internet, Ancestry.com, and/or local newspaper obituaries for dates of death for all patients to validate that they had not died during the 6‐month follow‐up. World wide web searches can identify 97% of deaths listed in the Centers for Disease Control and Prevention (CDC)/National Death Index, which is considered the gold standard in epidemiologic studies.16

Analysis

Univariate analysis of all covariates was completed to examine the normal distribution curves for these variables. Bivariate correlation analysis of all the independent variables by study group was performed to assess comparability of the study groups at baseline. The self‐reported cessation outcomes were calculated by dividing the number of patients who said they were not using tobacco or had quit, at 6 months posthospital discharge by the number of individuals in the study group at baseline minus those who had died. Both the intent to treat method, which assumes patients lost to follow‐up were still smoking, and the responder method, which does not include nonresponders in the analysis, were used to adjust the denominators for these outcomes.

Multivariate regression analysis was then used to model receipt of the intervention as predictor of self‐reported quit status adjusted for significant covariates. Statistical significance was defined by a P value of less than 0.05.

Survival analysis was employed to model differences in mortality between the study groups, controlling for any baseline imbalances (eg, comorbidity). Because baseline data were used in this model, the model includes only patients with signed a HIPAA release.

Internal review boards of our hospital and the NYS Department of Health reviewed and approved this study.

Limitations

This study targets a later phase in research progression from hypothesis development, pilot studies, efficacy (empirically supported) trials, effectiveness trials (real‐world settings), and dissemination studies.31 Because this study addresses the effectiveness rather than the efficacy of inpatient smoking cessation counseling, the use of a quasiexperimental rather than randomized controlled clinical trial design led to measured differences in the study groups at baseline. An important imbalance arose in the intervention group that had twice the percentage of patients with cardiovascular‐related discharge diagnoses as the comparison group. While we were able to adjust for these differences in our analysis, there may be unmeasured differences due to the fact that the inpatients were not randomized to the study groups.

The outcomes of this study cannot be attributed to any one component of the intervention (eg, NRT) vs. the combined effect of the inpatient smoking cessation program. The program components were implemented simultaneously in order to maximize synergistic effects; therefore, the effects of program components are difficult to disaggregate.

The results are limited by the validity of self‐report of smoking status. It is well known that research studies which validate smoking status biochemically have lower efficacy (OR, 1.44; 95% CI, 0.992.11) than those that do not validate smoking status (OR, 1.92; 95% CI, 1.262.93).32 Although it was impractical in this effectiveness study to biochemically validate smoking status 6 months posthospital discharge, we have documented a significant difference between the study groups that confirms the direction of the effect, if not the effect size.

Self‐reports tend to underestimate smoking status in population studies33; however, the discrepancy between self‐reported smoking and biochemical measurements among clinical trial participants is small.34 However, a small but significant bias toward a socially desirable response in intervention groups compared to control groups of 3% with carbon monoxide and 5% for cotinine has been documented.35 If social desirability bias is operative in this study, and if we apply the above correction factor of 4% to correct this classification error, then the difference between the intervention and comparison group would be 10 percentage points (40% in the intervention group vs. 30% in the comparison group using the intention to treat estimates). That difference is still clinically relevant.

The observed difference between the intervention and comparison group is underestimated because the comparison group was exposed to smoking cessation as well both at the time of admission and following discharge (Table 4). The comparison group in this study could thus be viewed as a usual care group rather than a control group. That exposure does cloud the measurement of quit rates as the comparison group is contaminated to some degree by exposure to various cessation methods. The impact of this exposure is to reduce the effect size observed in this study or underestimate the effect of the inpatient smoking cessation counseling because the comparison group was exposed other cessation methods, although to a lesser extent.

The social desirability bias inherent in self‐reported smoking status may increase the effect size while the use of comparison group that received usual care may decrease the effect size. Because neither of these biases could be measured in this study, it is impossible to say whether they negated each other.

As with any administrative database, use of EMR as a data source in this study led to missing data that precluded use of certain variables in the analysis. In addition, lack of written and signed HIPAA releases also precluded inclusion of several inpatients, mostly in the comparison group, in the analytic database. However, it is reassuring that the results of the 6‐month survey did not differ significantly when these individuals were included or excluded from a separate analysis of the survey data. Last, our study population is almost 100% Caucasian thus limiting how generalizable the results are to more heterogenous patient populations.

Intervention

The Smoking Cessation service for inpatients began on April 3, 2006. Upon admission, all patients were screened regarding their current smoking status. The nurse asked the patient if s/he currently smoked and then entered the responses into the hospital electronic medical record (EMR). A current smoker was defined as smoking every day or some days within the past 30 days. A roster of newly admitted current smokers was electronically transmitted to the Respiratory Care office daily. Only current smokers received counseling. The Smoking Cessation Specialist (SCS) subsequently saw inpatients within a 24‐hour time frame of admission, except for weekends and holidays. Each patient received 1 to 2 intensive follow‐up counseling sessions during hospitalization. An average of 10 patients per day were seen.

The goal of the inpatient smoking cessation service was to counsel patients on the health effects of smoking, address nicotine withdrawal symptoms, explain the different pharmacotherapies available, advise on how to quit, give self‐help materials, counsel family members, and refer to the New York State (NYS) Smokers' Fax‐to‐Quit program.9 Following the consult, the SCS documented the encounter in the patient's chart, including recommendations for nicotine replacement therapy (NRT) or bupropion (varenicline was not addressed as it was not on the formulary). The chart documentation informed the physician and nursing staff of the intervention and included the date/time, stage of change, and support action taken.

The SCS was part‐time, had nursing training, smoking cessation training,10 and was also trained by the Seton Health Cessation Center in the Butt Stops Here Program.11 She also implemented a performance improvement plan to increase the provision of smoking cessation counseling, increase NRT or bupropion prescriptions to smokers admitted to the hospital, and increase referrals to the NYS telephone quitline through the Fax‐to‐Quit program for outpatient resources and help following hospital discharge. The Fax‐to‐Quit program allows health care providers to refer patients to the NYS quitline via fax, with the patient's signature (patient permission) on the fax to quit form. After hospital discharge, the quitline then contacts the patient at a time that the patient requested.

The SCS visited patients with all admitting diagnoses on the medical, surgical, and special care units who were current smokers. Inpatients admitted to psychiatry, obstetrics, and the intensive care unit (ICU) were not seen by the SCS, except for ICU patients referred by a physician. Inpatients who had short stays or who were admitted and discharged in 1 day or during the weekend were not seen.

The intervention included either a brief 3‐minute to 5‐minute intervention or a more intensive intervention, that required 10 to 20 minutes (18 minutes average). The length of the intervention was determined by how receptive the patient was to the intervention. All interventions began with patient identification, an introduction to the SCS, and an explanation of the purpose of the visit. The SCS then inquired about the patient's comfort level vis‐a‐vis nicotine withdrawal and if s/he was receiving any NRT while in the hospital (NRT on the inpatient formulary included the nicotine patch or gum). If the patient was receptive to counseling, the SCS then began to work through the 5 A's, as described in the 2000 DHHS Clinical Practice Guidelines.12 The 2000 DHHS Clinical Practice Guidelines were used because the 2008 update had not been released at the time this study was initiated in 2006. These include: asking about smoking status, advising on how to quit, assessing readiness to quit, and assisting in arranging treatment options that include pharmacotherapy, counseling, as well as referral to the NYS Smokers' Quitline. A workbook was provided to reinforce counseling but was not necessarily used during counseling session. A compact disc (CD) with relaxation exercises5 was provided to those inpatients who were interested in stress reduction. If family members were present, and were also smokers, they were included in the counseling session, if willing. Each patient was offered a referral via the Fax‐to‐Quit program to continue treatment on an outpatient basis.

If the patient was not motivated to quit or declined the consult, the visits were short and focused on the patient's experience with nicotine withdrawal. These patients were also given self‐help materials and, if possible, the relevance of and roadblocks to quitting were reviewed. Patients were prompted to think about why quitting was relevant and often the reason for hospitalization was used to motivate the quitting process.

Upon hospital discharge, the patient's primary care provider was notified of the cessation intervention by a letter from the SCS. The letter described the intervention and stated whether or not the patient agreed to be referred to the Fax‐to‐Quit Program.

Study Participants

Patients were recruited from July 1, 2006 (after the smoking ban went into effect) through June 1, 2008. Inpatients who currently smoked were informed of this study and were asked to sign informed consent to participate after they were seen by the SCS. Current smokers of all admitting diagnoses were recruited into the study. Patients provided informed consent for a telephone interview 6 months posthospital discharge. A written Health Insurance Portability and Accountability Act (HIPAA) release was obtained to allow access to an individual's specific EMR.

A comparison group of inpatients who were also current smokers, but who did not receive the intervention were also contacted six months after hospitalization. Reasons for not receiving the intervention included the fact the SCS was part‐time and also took a leave of absence during the study and therefore could not see all inpatients who currently smoked. Other reasons for not receiving the intervention include too short a stay for the SCS to see the patient or the patient was out of the room for tests or procedures when the SCS was available. These patients provided informed consent to be interviewed 6 months after hospital discharge and HIPAA consent for access to their medical record. Not all inpatients in the comparison group provided written HIPAA release for use of their medical record; therefore, these patients were excluded because their baseline demographic and diagnostic data were missing.

Sample‐size considerations were driven around having adequate numbers of subjects to measure the prevalence of smoking cessation at 6 months post‐hospital discharge with an acceptable degree of precision. Prevalence estimates from previous studies for 6‐month cessation typically range from 20% to 30%, with cessation rates as high as 67% (this estimate applies to postmyocardial infarction patients.)13 For conservative estimation, we used 50% as the 6‐month prevalence of cessation in the current study, which placed binomial variance at its theoretical maximum. In this case, a sample of 300 subjects provides a margin of error of 0.058 for a 95% confidence interval around this point estimate.

Data Sources

The hospital EMR database was used to monitor several components of the program: nursing screening, smoking cessation counseling, and pharmacy dispensing of NRT and bupropion. The screening data were also used to monitor the proportion of current smokers admitted during the study period. Elements of the EMR were used to define the following covariates: patient age, gender, ethnicity, and the primary discharge diagnosis (via International Statistical Classification of Diseases and Related Health Problems, 9th edition [ICD9] codes) and readmission during the six month follow‐up period. Mean length of stay (LOS) was computed. The Elixhauser Comorbidity Index that utilizes ICD9 codes was used for comorbidity risk adjustment.14

Study participants were contacted by phone 6 months posthospital discharge. Data collection began July 1, 2006 and was completed January 1, 2009. The interview focused on self‐reported point prevalence of smoking and 6‐month quit status. The point prevalence for self‐reported abstinence was derived from the question Do you now smoke cigarettes every day, some days, or not at all?15 Self‐reported quit status was derived from question Have you quit smoking since you were discharged from the hospital? In addition, respondents were queried about their number of years smoked, post‐hospital discharge number of quit attempts, and cessation efforts (NRT, self‐help groups, quitline use, etc.). Last, they were surveyed about barriers to cessation (exposure to secondhand smoke, rules about smoking in the home or car), educational level, employment, and health ratings.

To determine the status of those lost to follow‐up, administrative and EMR databases for appointments and follow‐up visits were accessed to determine if the patient was alive during the 6 months between discharge and the follow‐up call. To confirm mortality, we searched the Internet, Ancestry.com, and/or local newspaper obituaries for dates of death for all patients to validate that they had not died during the 6‐month follow‐up. World wide web searches can identify 97% of deaths listed in the Centers for Disease Control and Prevention (CDC)/National Death Index, which is considered the gold standard in epidemiologic studies.16

Analysis

Univariate analysis of all covariates was completed to examine the normal distribution curves for these variables. Bivariate correlation analysis of all the independent variables by study group was performed to assess comparability of the study groups at baseline. The self‐reported cessation outcomes were calculated by dividing the number of patients who said they were not using tobacco or had quit, at 6 months posthospital discharge by the number of individuals in the study group at baseline minus those who had died. Both the intent to treat method, which assumes patients lost to follow‐up were still smoking, and the responder method, which does not include nonresponders in the analysis, were used to adjust the denominators for these outcomes.

Multivariate regression analysis was then used to model receipt of the intervention as predictor of self‐reported quit status adjusted for significant covariates. Statistical significance was defined by a P value of less than 0.05.

Survival analysis was employed to model differences in mortality between the study groups, controlling for any baseline imbalances (eg, comorbidity). Because baseline data were used in this model, the model includes only patients with signed a HIPAA release.

Internal review boards of our hospital and the NYS Department of Health reviewed and approved this study.

Limitations

This study targets a later phase in research progression from hypothesis development, pilot studies, efficacy (empirically supported) trials, effectiveness trials (real‐world settings), and dissemination studies.31 Because this study addresses the effectiveness rather than the efficacy of inpatient smoking cessation counseling, the use of a quasiexperimental rather than randomized controlled clinical trial design led to measured differences in the study groups at baseline. An important imbalance arose in the intervention group that had twice the percentage of patients with cardiovascular‐related discharge diagnoses as the comparison group. While we were able to adjust for these differences in our analysis, there may be unmeasured differences due to the fact that the inpatients were not randomized to the study groups.

The outcomes of this study cannot be attributed to any one component of the intervention (eg, NRT) vs. the combined effect of the inpatient smoking cessation program. The program components were implemented simultaneously in order to maximize synergistic effects; therefore, the effects of program components are difficult to disaggregate.

The results are limited by the validity of self‐report of smoking status. It is well known that research studies which validate smoking status biochemically have lower efficacy (OR, 1.44; 95% CI, 0.992.11) than those that do not validate smoking status (OR, 1.92; 95% CI, 1.262.93).32 Although it was impractical in this effectiveness study to biochemically validate smoking status 6 months posthospital discharge, we have documented a significant difference between the study groups that confirms the direction of the effect, if not the effect size.

Self‐reports tend to underestimate smoking status in population studies33; however, the discrepancy between self‐reported smoking and biochemical measurements among clinical trial participants is small.34 However, a small but significant bias toward a socially desirable response in intervention groups compared to control groups of 3% with carbon monoxide and 5% for cotinine has been documented.35 If social desirability bias is operative in this study, and if we apply the above correction factor of 4% to correct this classification error, then the difference between the intervention and comparison group would be 10 percentage points (40% in the intervention group vs. 30% in the comparison group using the intention to treat estimates). That difference is still clinically relevant.

The observed difference between the intervention and comparison group is underestimated because the comparison group was exposed to smoking cessation as well both at the time of admission and following discharge (Table 4). The comparison group in this study could thus be viewed as a usual care group rather than a control group. That exposure does cloud the measurement of quit rates as the comparison group is contaminated to some degree by exposure to various cessation methods. The impact of this exposure is to reduce the effect size observed in this study or underestimate the effect of the inpatient smoking cessation counseling because the comparison group was exposed other cessation methods, although to a lesser extent.

The social desirability bias inherent in self‐reported smoking status may increase the effect size while the use of comparison group that received usual care may decrease the effect size. Because neither of these biases could be measured in this study, it is impossible to say whether they negated each other.

As with any administrative database, use of EMR as a data source in this study led to missing data that precluded use of certain variables in the analysis. In addition, lack of written and signed HIPAA releases also precluded inclusion of several inpatients, mostly in the comparison group, in the analytic database. However, it is reassuring that the results of the 6‐month survey did not differ significantly when these individuals were included or excluded from a separate analysis of the survey data. Last, our study population is almost 100% Caucasian thus limiting how generalizable the results are to more heterogenous patient populations.

Policy Implementation

As prior studies have shown that institution of a smoke‐free medical campus policy involves much more than just posting signage,9, 10 a detailed multidisciplinary work plan was implemented starting 1.5 years prior to the date our policy went into effect on July 1, 2006. The Implementing a Smoke‐Free Environment plan, produced by the University of Michigan,11 which includes a 15‐step checklist, was used to guide this policy change.12 As part of that plan, employees were offered on‐site smoking cessation services, including nicotine replacement therapy (NRT), and 150 employees participated in this program prior to July 1, 2006. Staff, community, and patient education was also completed. A new campus map delineating the smoke‐free border was disseminated. Signage was posted in areas used in the past for smoking. In addition to implementing this plan, an inpatient smoking cessation service was started 3 months prior to July 1, 2006. In addition to supporting the enforcement of the smoke‐free medical campus, our inpatient smoking cessation program was designed to help inpatients with nicotine withdrawal as well as smoking cessation, if they were ready to quit.

Data Collection and Analysis

The inpatient electronic medical record (EMR) was used to monitor the smoking status of patients admitted to hospital on a monthly basis. On admission to the hospital, the admitting nurse screened patients for current smoking status. This information was entered into the EMR starting in April 2006; therefore, pre‐ban screening data were limited to 2 months prior to the ban. Inpatients too sick to complete this screening process, women admitted for labor and delivery, and inpatients boarded in the emergency department were not screened. No identifiers were used in compiling these monthly data.

Nursing reports of inpatients signing out against medical advice (AMA) were compiled in order to compare incidence of AMA pre‐ban to post‐ban. AMA documentation in our hospital takes the form of a structured incident report that is reliably documented by nursing staff and signed by the attending physician of service.

Computerized inpatient doctors' orders to pharmacy for NRT, dispensed as gum or patch, were monitored 2 years preinitiation and postinitiation of the inpatient smoking cessation service on April 1, 2006. As varenicline was nonformulary and bupropion was used for other indications than smoking cessation, these medications were not included in this review. The Chow test was used to measure and test for significant breaks in a time series analysis of the NRT orders.

New York State law requires an annual occupational health review to be completed by every hospital employee. At our hospital, this review included a question on tobacco use Do you smoke or chew tobacco? Although there has been a smoker/nonsmoker differential in the rates offered for supplemental life insurance since 1992, there were no wellness credits or other incentives for medical insurance offered in employee benefits that may predispose employees to underreport tobacco use. Using this question, employees were categorized as self‐reported current smokers or chew users. Employee smoking rates were estimated using different denominators to validate the direction of the trend. First, self‐reported smoking rates were compared pre‐ban and post‐ban among a stable cohort of hospital employees (n = 489), defined as hospital‐based employees with anniversary dates from March to June who reported in both 2005 and 2007. The McNemar test was used to test the statistical significance of the 2 smoking rates of paired replicates in this stable cohort of employees reporting pre‐ban and post‐ban. Second, all employees in the database reporting smoking status pre‐ban, March to June 2005, and then post‐ban, March to June 2006 and 2007, were compared in order to monitor trends in employee smoking overall. A t‐test was used to compare the statistical significance of the difference in the overall rates of smoking among all employees pre‐ban and post‐ban.

Internal review boards of our hospital and the New York State Department of Health reviewed and approved this study.

Inpatient Outcomes

An average of 959 patients were admitted per month in the 18‐month period pre‐ban (January 2005 to June 2006) vs. 988 per month in the 23‐month period post‐ban (July 2006 to September 2008). A monthly average of 89% of inpatients were screened for tobacco use when admitted. The monthly average for the percentage of inpatients who currently smoke has been approximately 21.6% following the implementation of the smoke‐free hospital plan. There has been little variation (Figure 1) in the percentage of inpatients who smoke pre‐ban and post‐ban except for the startup period in 2006 and the onset of the 2007 respiratory illness season.

Figure 1

Among all inpatients who currently smoke, 69.8% received a brief nursing intervention at the time of admission and 25% received an inpatient visit from our part‐time smoking cessation specialist.

The percentage of inpatients who signed out against medical advice (AMA) with the reason of having to smoke was 13.8% (4/29) 6 months pre‐ban, and 13.6% (3/22) 6 months post‐ban. In 2007, there were no inpatients who signed out AMA stating that they needed to smoke. Because the reason for signing out AMA may be underreported, we also examined the rate of smoking among all inpatients who sign out AMA. Six months pre‐ban, this percentage was 48.3% (14/29), but increased 6 months post‐ban to 59% (13/22). In 2007, the percentage of smokers among inpatients who sign out AMA leveled off at 50.8% (29/57).

Review of computerized inpatient prescription orders shows that orders for NRT nearly tripled after the inpatient smoking cessation service started April 1, 2006 (3 months prior to the ban) (Figure 2). Inpatient orders for these medications increased from 832 in a 2‐year period before the ban (April 1, 2004 to March 31, 2006) to 2475 in the 2 years following the initiation of the inpatient smoking service (April 1, 2006 to March 31, 2008). The Chow test is highly significant for a break point in June 2006 (P = 0.008), 1 month prior to the ban.

Figure 2

Employee Smoking Rates

Among a cohort of 489 hospital‐based employees reporting in both 2005 and 2007, 12% reported smoking in 2005 and 7.5% in 2007 (McNemar was significant at P < 0.001). Two employees reported using chewing tobacco in 2005 and only 1 in 2007.

Including all hospital employees reporting any 1 year during their anniversary dates, the self‐reported smoking rates were 14.3% (n = 624) in March to June 2005, 14.8% (n = 661) in March to June 2006, and 9.4% (n = 1,112) in March to June 2007 (P < 0.0002). Because promotions change the anniversary date, and the database was expanded in 2007 to include new hires and managerial staff, these estimates represent the point prevalence among employees whose anniversary dates fall between March and June.

Policy Implementation

As prior studies have shown that institution of a smoke‐free medical campus policy involves much more than just posting signage,9, 10 a detailed multidisciplinary work plan was implemented starting 1.5 years prior to the date our policy went into effect on July 1, 2006. The Implementing a Smoke‐Free Environment plan, produced by the University of Michigan,11 which includes a 15‐step checklist, was used to guide this policy change.12 As part of that plan, employees were offered on‐site smoking cessation services, including nicotine replacement therapy (NRT), and 150 employees participated in this program prior to July 1, 2006. Staff, community, and patient education was also completed. A new campus map delineating the smoke‐free border was disseminated. Signage was posted in areas used in the past for smoking. In addition to implementing this plan, an inpatient smoking cessation service was started 3 months prior to July 1, 2006. In addition to supporting the enforcement of the smoke‐free medical campus, our inpatient smoking cessation program was designed to help inpatients with nicotine withdrawal as well as smoking cessation, if they were ready to quit.

Data Collection and Analysis

The inpatient electronic medical record (EMR) was used to monitor the smoking status of patients admitted to hospital on a monthly basis. On admission to the hospital, the admitting nurse screened patients for current smoking status. This information was entered into the EMR starting in April 2006; therefore, pre‐ban screening data were limited to 2 months prior to the ban. Inpatients too sick to complete this screening process, women admitted for labor and delivery, and inpatients boarded in the emergency department were not screened. No identifiers were used in compiling these monthly data.

Nursing reports of inpatients signing out against medical advice (AMA) were compiled in order to compare incidence of AMA pre‐ban to post‐ban. AMA documentation in our hospital takes the form of a structured incident report that is reliably documented by nursing staff and signed by the attending physician of service.

Computerized inpatient doctors' orders to pharmacy for NRT, dispensed as gum or patch, were monitored 2 years preinitiation and postinitiation of the inpatient smoking cessation service on April 1, 2006. As varenicline was nonformulary and bupropion was used for other indications than smoking cessation, these medications were not included in this review. The Chow test was used to measure and test for significant breaks in a time series analysis of the NRT orders.

New York State law requires an annual occupational health review to be completed by every hospital employee. At our hospital, this review included a question on tobacco use Do you smoke or chew tobacco? Although there has been a smoker/nonsmoker differential in the rates offered for supplemental life insurance since 1992, there were no wellness credits or other incentives for medical insurance offered in employee benefits that may predispose employees to underreport tobacco use. Using this question, employees were categorized as self‐reported current smokers or chew users. Employee smoking rates were estimated using different denominators to validate the direction of the trend. First, self‐reported smoking rates were compared pre‐ban and post‐ban among a stable cohort of hospital employees (n = 489), defined as hospital‐based employees with anniversary dates from March to June who reported in both 2005 and 2007. The McNemar test was used to test the statistical significance of the 2 smoking rates of paired replicates in this stable cohort of employees reporting pre‐ban and post‐ban. Second, all employees in the database reporting smoking status pre‐ban, March to June 2005, and then post‐ban, March to June 2006 and 2007, were compared in order to monitor trends in employee smoking overall. A t‐test was used to compare the statistical significance of the difference in the overall rates of smoking among all employees pre‐ban and post‐ban.

Internal review boards of our hospital and the New York State Department of Health reviewed and approved this study.

Inpatient Outcomes

An average of 959 patients were admitted per month in the 18‐month period pre‐ban (January 2005 to June 2006) vs. 988 per month in the 23‐month period post‐ban (July 2006 to September 2008). A monthly average of 89% of inpatients were screened for tobacco use when admitted. The monthly average for the percentage of inpatients who currently smoke has been approximately 21.6% following the implementation of the smoke‐free hospital plan. There has been little variation (Figure 1) in the percentage of inpatients who smoke pre‐ban and post‐ban except for the startup period in 2006 and the onset of the 2007 respiratory illness season.

Figure 1

Among all inpatients who currently smoke, 69.8% received a brief nursing intervention at the time of admission and 25% received an inpatient visit from our part‐time smoking cessation specialist.

The percentage of inpatients who signed out against medical advice (AMA) with the reason of having to smoke was 13.8% (4/29) 6 months pre‐ban, and 13.6% (3/22) 6 months post‐ban. In 2007, there were no inpatients who signed out AMA stating that they needed to smoke. Because the reason for signing out AMA may be underreported, we also examined the rate of smoking among all inpatients who sign out AMA. Six months pre‐ban, this percentage was 48.3% (14/29), but increased 6 months post‐ban to 59% (13/22). In 2007, the percentage of smokers among inpatients who sign out AMA leveled off at 50.8% (29/57).

Review of computerized inpatient prescription orders shows that orders for NRT nearly tripled after the inpatient smoking cessation service started April 1, 2006 (3 months prior to the ban) (Figure 2). Inpatient orders for these medications increased from 832 in a 2‐year period before the ban (April 1, 2004 to March 31, 2006) to 2475 in the 2 years following the initiation of the inpatient smoking service (April 1, 2006 to March 31, 2008). The Chow test is highly significant for a break point in June 2006 (P = 0.008), 1 month prior to the ban.

Figure 2

Employee Smoking Rates

Among a cohort of 489 hospital‐based employees reporting in both 2005 and 2007, 12% reported smoking in 2005 and 7.5% in 2007 (McNemar was significant at P < 0.001). Two employees reported using chewing tobacco in 2005 and only 1 in 2007.

Including all hospital employees reporting any 1 year during their anniversary dates, the self‐reported smoking rates were 14.3% (n = 624) in March to June 2005, 14.8% (n = 661) in March to June 2006, and 9.4% (n = 1,112) in March to June 2007 (P < 0.0002). Because promotions change the anniversary date, and the database was expanded in 2007 to include new hires and managerial staff, these estimates represent the point prevalence among employees whose anniversary dates fall between March and June.

Policy Implementation

As prior studies have shown that institution of a smoke‐free medical campus policy involves much more than just posting signage,9, 10 a detailed multidisciplinary work plan was implemented starting 1.5 years prior to the date our policy went into effect on July 1, 2006. The Implementing a Smoke‐Free Environment plan, produced by the University of Michigan,11 which includes a 15‐step checklist, was used to guide this policy change.12 As part of that plan, employees were offered on‐site smoking cessation services, including nicotine replacement therapy (NRT), and 150 employees participated in this program prior to July 1, 2006. Staff, community, and patient education was also completed. A new campus map delineating the smoke‐free border was disseminated. Signage was posted in areas used in the past for smoking. In addition to implementing this plan, an inpatient smoking cessation service was started 3 months prior to July 1, 2006. In addition to supporting the enforcement of the smoke‐free medical campus, our inpatient smoking cessation program was designed to help inpatients with nicotine withdrawal as well as smoking cessation, if they were ready to quit.

Data Collection and Analysis

The inpatient electronic medical record (EMR) was used to monitor the smoking status of patients admitted to hospital on a monthly basis. On admission to the hospital, the admitting nurse screened patients for current smoking status. This information was entered into the EMR starting in April 2006; therefore, pre‐ban screening data were limited to 2 months prior to the ban. Inpatients too sick to complete this screening process, women admitted for labor and delivery, and inpatients boarded in the emergency department were not screened. No identifiers were used in compiling these monthly data.

Nursing reports of inpatients signing out against medical advice (AMA) were compiled in order to compare incidence of AMA pre‐ban to post‐ban. AMA documentation in our hospital takes the form of a structured incident report that is reliably documented by nursing staff and signed by the attending physician of service.

Computerized inpatient doctors' orders to pharmacy for NRT, dispensed as gum or patch, were monitored 2 years preinitiation and postinitiation of the inpatient smoking cessation service on April 1, 2006. As varenicline was nonformulary and bupropion was used for other indications than smoking cessation, these medications were not included in this review. The Chow test was used to measure and test for significant breaks in a time series analysis of the NRT orders.

New York State law requires an annual occupational health review to be completed by every hospital employee. At our hospital, this review included a question on tobacco use Do you smoke or chew tobacco? Although there has been a smoker/nonsmoker differential in the rates offered for supplemental life insurance since 1992, there were no wellness credits or other incentives for medical insurance offered in employee benefits that may predispose employees to underreport tobacco use. Using this question, employees were categorized as self‐reported current smokers or chew users. Employee smoking rates were estimated using different denominators to validate the direction of the trend. First, self‐reported smoking rates were compared pre‐ban and post‐ban among a stable cohort of hospital employees (n = 489), defined as hospital‐based employees with anniversary dates from March to June who reported in both 2005 and 2007. The McNemar test was used to test the statistical significance of the 2 smoking rates of paired replicates in this stable cohort of employees reporting pre‐ban and post‐ban. Second, all employees in the database reporting smoking status pre‐ban, March to June 2005, and then post‐ban, March to June 2006 and 2007, were compared in order to monitor trends in employee smoking overall. A t‐test was used to compare the statistical significance of the difference in the overall rates of smoking among all employees pre‐ban and post‐ban.

Internal review boards of our hospital and the New York State Department of Health reviewed and approved this study.

Inpatient Outcomes

An average of 959 patients were admitted per month in the 18‐month period pre‐ban (January 2005 to June 2006) vs. 988 per month in the 23‐month period post‐ban (July 2006 to September 2008). A monthly average of 89% of inpatients were screened for tobacco use when admitted. The monthly average for the percentage of inpatients who currently smoke has been approximately 21.6% following the implementation of the smoke‐free hospital plan. There has been little variation (Figure 1) in the percentage of inpatients who smoke pre‐ban and post‐ban except for the startup period in 2006 and the onset of the 2007 respiratory illness season.

Figure 1

Among all inpatients who currently smoke, 69.8% received a brief nursing intervention at the time of admission and 25% received an inpatient visit from our part‐time smoking cessation specialist.

The percentage of inpatients who signed out against medical advice (AMA) with the reason of having to smoke was 13.8% (4/29) 6 months pre‐ban, and 13.6% (3/22) 6 months post‐ban. In 2007, there were no inpatients who signed out AMA stating that they needed to smoke. Because the reason for signing out AMA may be underreported, we also examined the rate of smoking among all inpatients who sign out AMA. Six months pre‐ban, this percentage was 48.3% (14/29), but increased 6 months post‐ban to 59% (13/22). In 2007, the percentage of smokers among inpatients who sign out AMA leveled off at 50.8% (29/57).

Review of computerized inpatient prescription orders shows that orders for NRT nearly tripled after the inpatient smoking cessation service started April 1, 2006 (3 months prior to the ban) (Figure 2). Inpatient orders for these medications increased from 832 in a 2‐year period before the ban (April 1, 2004 to March 31, 2006) to 2475 in the 2 years following the initiation of the inpatient smoking service (April 1, 2006 to March 31, 2008). The Chow test is highly significant for a break point in June 2006 (P = 0.008), 1 month prior to the ban.

Figure 2

Employee Smoking Rates

Among a cohort of 489 hospital‐based employees reporting in both 2005 and 2007, 12% reported smoking in 2005 and 7.5% in 2007 (McNemar was significant at P < 0.001). Two employees reported using chewing tobacco in 2005 and only 1 in 2007.

Including all hospital employees reporting any 1 year during their anniversary dates, the self‐reported smoking rates were 14.3% (n = 624) in March to June 2005, 14.8% (n = 661) in March to June 2006, and 9.4% (n = 1,112) in March to June 2007 (P < 0.0002). Because promotions change the anniversary date, and the database was expanded in 2007 to include new hires and managerial staff, these estimates represent the point prevalence among employees whose anniversary dates fall between March and June.