American Journal of Infection Control 42 (2014) 288-93
Contents lists available at ScienceDirect
American Journal of Infection Control
American Journal of Infection Control
journal homepage: www.ajicjournal.org
Major article
The impact of influenza vaccination requirements for hospital personnel in California: Knowledge, attitudes, and vaccine uptake Katherine M. Harris PhD a, *, Lori Uscher-Pines PhD a, Bing Han PhD b, Megan C. Lindley MPH c, Suchita A. Lorick DO, MPH c a b c
RAND Corporation, Arlington, VA RAND Corporation, Santa Monica, CA Centers for Disease and Prevention, Atlanta, GA
Key Words: State policy Immunization Hospitals Survey research Difference-in-difference
Background: Seasonal influenza infections are a leading cause of illness, death, and lost productivity. Vaccinating health care personnel (HCP) can reduce transmission of influenza virus to patients and reduce influenza-related absenteeism, enabling the health care system to meet elevated demand for care during influenza outbreaks. Objectives: We evaluated the impact of California’s 2006 influenza vaccination requirement for hospital workers (requiring vaccination or signed declinations) on uptake and vaccination-related attitudes, beliefs, and knowledge among hospital HCP. Methods: We used a causal difference-in-differences approach to compare changes over the prior 10 years in the self-reported frequency of influenza vaccination for California hospital HCP and those from other states without similar laws using data from a stratified sample (N ¼ 3,529) of HCP drawn from online survey panels. We also examined cross-sectional differences in awareness of vaccination policies, promotion efforts, and attitudes toward influenza vaccination. All analyses used propensity score weighting to balance the observable characteristics of the 2 samples. Results: We found that compared with their counterparts in other states, California hospital HCP were (1) more likely to report working under a formal written policy for influenza vaccination, (2) no more likely to be vaccinated, and (3) less likely to report working for an employer who provided financial incentives for vaccination or rewarded or recognized employees for being vaccinated. Conclusion: Our results suggest that state-level vaccination requirements such as those enacted by California, may not be sufficient to increase uptake among hospital HCP. Copyright Ó 2014 by the Association for Professionals in Infection Control and Epidemiology, Inc. Published by Elsevier Inc. All rights reserved.
Seasonal influenza infections are a leading cause of illness, death, and lost productivity.1 Vaccinating health care personnel (HCP) can reduce transmission of influenza virus to patients2,3 and reduce influenza-related absenteeism, enabling the health care system to meet elevated demand for care during influenza outbreaks.4,5 The US Centers for Disease Control and Prevention’s
* Address correspondence to Katherine M. Harris, PhD, The RAND Corporation, 1200 South Hayes Street, Arlington, VA 22202. E-mail address: [email protected] (K.M. Harris). Supported by the US Centers for Disease Control and Prevention, Atlanta, GA. This work was performed as part of a cooperative agreement (1 U01 IP000416-02) with the US Centers for Disease Control and Prevention. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention or the California Department of Public Health. Conflicts of interest: None to report.
(CDC) Advisory Committee on Immunization Practices has recommended annual influenza vaccination of all HCP.6 For most of the past decade, however, less than half of all HCP received annual influenza vaccinations.7 The perception that voluntary programs are insufficient to generate substantial increases in influenza vaccination has led numerous experts, professional societies, advocacy organizations, and others to call for making influenza vaccination a condition of employment and/or professional privileges.8-12 However, data suggest that even vaccination requirements without strict penalties for noncompliance (defined here as job loss, loss of professional privileges, or visible identification of vaccination status) can increase vaccination among hospital HCP by 15 percentage points or more, depending on how they are implemented.13-16 The effectiveness of influenza vaccination requirements beyond those adopted by hospitals voluntarily remains unclear. The
0196-6553/$36.00 - Copyright Ó 2014 by the Association for Professionals in Infection Control and Epidemiology, Inc. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ajic.2013.09.030
K.M. Harris et al. / American Journal of Infection Control 42 (2014) 288-93
experience of California may provide lessons regarding the potential of relatively permissive state-level requirements. We define permissive requirements here as those that allow individuals to avoid vaccination by signing a declination statement without establishing an enforcement mechanism or penalties for noncompliance. As part of legislation to strengthen prevention of hospitalacquired infections, California became one of the first states to require influenza vaccination of all hospital-based HCP in 2006.17 As of March 2012, 7 other states have enacted similar laws.17,18 Specifically, California Health and Safety Code 1288.7 requires all general acute care hospitals to offer annually on-site influenza vaccinations to hospital employees at no cost and to require their employees to be vaccinated or declare in writing that they have declined vaccination. In accordance with CDC’s and the federal Healthcare Infection Control Practices Advisory Committee’s recommendations, in 2008 California Department of Public Health clarified requirements for compliance with the statute, including the provision of education to employees and other nonemployee HCP and the reporting to California Department of Public Health of annual vaccination and declination rates.19-21 The statute does not make vaccination a condition of employment, stipulate means through which hospitals should enforce requirements, or institute penalties for hospitals with low vaccination rates. As such, the California law can be considered a permissive state-level requirement in contrast to Rhode Island’s 2012 law, which requires unvaccinated HCP to wear masks during periods where influenza is widespread and includes specific penalties for noncompliance.22 Rhode Island’s law is currently being evaluated; however, at this time an evaluation of California’s law is critical to understanding whether success can be achieved without threatening HCP autonomy. To learn from California’s experiences, we evaluated the impact of California’s 2006 influenza vaccination requirement for hospital workers (requiring vaccination or signed declinations) on uptake and vaccination-related attitudes, beliefs, and knowledge among hospital HCP. Our results may assist policy makers to understand the potential effectiveness of state policies that strive to make the promotion of influenza vaccination a high priority without imposing strict penalties for noncompliance. METHODS Data sources In the absence of appropriate administratively documented vaccination coverage data, we drew a stratified sample of 3,529 hospital-based HCP in April 2010 from 3 online survey panels. These panels included (1) the Physicians’ Consulting Network (PCN) (N ¼ 846), a panel of approximately 70,000 HCP (mostly physicians) who agreed to participate in health care-related research in exchange for financial compensation; (2) eRewards (n ¼ 2,113) and GMI (n ¼ 99), which are 2 opt-in panels of Internet users and both of which include tens of thousands of individuals recruited through Web advertising who agreed to participate in surveys in exchange for small cash rewards and incentives; and (3) hospital-based HCP (n ¼ 471) recruited from KnowledgePanel, a nationally representative, online research panel of 50,000 individuals, who completed the same survey instrument as part of a concurrent CDC study to measure influenza vaccination among US HCP.23 All data sources are convenience samples with the exception of KnowledgePanel. To ensure representation of a diverse range of health care professionals and vaccination-related attitudes and behaviors, we prospectively recruited 4 strata or categories of HCP employed by or
289
working in general acute care hospitals: (1) physicians typically employed by hospitals (eg, anesthesiologists, surgeons), (2) nurses, (3) allied health professionals, and (4) nonclinical support staff and administrators. We recruited physicians, nurses, and selected types of allied health professionals through targeted e-mails using occupational information collected at the time of recruitment into their respective panels. We recruited others using panel-wide e-mails. When feasible, we balanced the number of sample members from each source residing inside and outside of California to minimize the potential for the confounding of sample source and exposure to CA’s vaccination requirement. Next, we excluded 297 respondents who resided in 7 states (Alabama, Illinois, Massachusetts, New Hampshire, Tennessee, Maine, and Oklahoma) that had laws similar to California’s during our study period. This exclusion ensured that the non-California “control” arm of the study only included HCP who were not subject to state-wide legislation requiring vaccination. This exclusion did not change our results. We also excluded 421 respondents who reported working in hospital-owned, ambulatory care practices. Seventy percent of eligible KnowledgePanel panelists and 22% of PCN panelists completed the survey. We cannot determine participation rates for opt-in panelists because the number of eligible people invited to participate is unknown. All 4 sample sources have been described in previous publications.23,24 After applying exclusion criteria, our analytic sample consisted of 1,201 California and 1,610 non-California hospital-based HCP. RAND’s Institutional Review Board approved the study design and survey protocols. Measures The survey measured vaccination-related behaviors, attitudes, and knowledge and exposure to workplace vaccination efforts. The survey asked 6 questions about influenza vaccination during the prior 10 years. We expected that HCP could accurately recall the frequency of influenza vaccination in the recent past, although not necessarily in specific terms. To aid recall, we started by asking about vaccination for seasonal and H1N1 influenza during the past season. We followed-up with 2 questions: (1) “Compared with ten years ago, are you more or less likely to get an influenza vaccination now than you were in the past?” and (2) “Please estimate how often you have been vaccinated for influenza over the past ten years.” We also gauged respondents’ exposure to workplace vaccination efforts. We started by asking respondents whether their employer had a “formal, written policy on influenza vaccination.” We measured respondents’ perceptions of working under a vaccination requirement by asking, “During this past influenza season, did your work site recommend or require that you be vaccinated for seasonal influenza?” To detect the presence of a requirement we asked whether unvaccinated HCP were required to sign a declination statement or faced consequences for not complying with their employer’s policy, including requiring a special badge, mask, or respirator, work reassignment, participation in an educational intervention, leave without pay, loss of benefits, and termination. The respondent was considered subject to a vaccination requirement if the respondent reported a declination requirement or any of the listed consequences of nonvaccination. The survey asked about other efforts to promote vaccination. These included on-site vaccination, reminders, publicizing the risks and benefits of vaccination, public identification of vaccinated staff, financial incentives, other rewards, and publicizing employee vaccination rates. We asked about vaccination-related attitudes and beliefs that could potentially be influenced by efforts to educate employees
290
K.M. Harris et al. / American Journal of Infection Control 42 (2014) 288-93
about vaccination benefits. Questions included agreement with statements about HCP susceptibility to influenza inflections and understanding of influenza transmission as described in CDCissued guidance on the topic.25 Finally, we asked about occupational and sociodemographic characteristics, characteristics of respondents’ hospitals, health status, and contact outside of work with individuals at increased risk of influenza.
the standard cutoff of P < .05, we claimed statistical significance only when the reported P value could achieve a more stringent false discovery rate (FDR) threshold of .05) to report working for an employer who provided financial incentives for vaccination, 2.6 percentage points less likely (P < .01; FDR < .05) to report working for an employer who rewarded groups of employees for vaccination, and 5.0 percentage points less likely (P < .01; FDR < .05) to report working for an employer who promoted competition between units to attain increased influenza vaccination coverage (not shown). To assess the robustness of our findings to the composition of our sample, we estimated propensity weighted differences separately by occupational grouping (ie, doctors, nurses, allied health professionals, and nonclinical administrative and support staff) and found similar results (not presented). Likewise, we excluded those not directly employed by the hospitals in which they were working and, again, found similar results (not shown).
Accurate measurement of policy impacts was complicated by the lack of preimplementation data and random assignment to vaccination policies. To address these challenges, we used a causal difference-in-differences (DID) approach, which has been shown in the literature to be superior to the classic regression DID approach with less bias and smaller mean-square error.26-28 The classic DID approach compares the differences in the mean outcome before and after the treatment between the treatment and control groups, which is usually implemented as a regression or analysis of covariance. The causal DID approach we used employs a propensity score weighting method to balance the observed characteristics between the 2 groups, which avoids the potential model misspecification in the classic DID. The propensity score is the probability that a respondent received treatment conditional on observed characteristics. In our DID approach, we compared changes across the 2 samples in the self-reported frequency of being vaccinated during the 4 years prior to the enactment of the California requirement (20012005) and 4 years following enactment (2007-2011). We defined the pre- and post- periods in this way to account for the phase-in of the requirement and to have time periods of equal length to increase comparability. We considered a statistically larger pre-post difference favoring California to be evidence that the California requirement increased vaccination among HCP compared with other states. We also compared reports of ever being vaccinated for influenza, being more likely to be vaccinated now compared with 10 years ago, being vaccinated during the 2010-2011 season, being subject to vaccination promotion efforts, and HCP attitudes and beliefs across the 2 samples to identify the mechanisms through which workplace promotion efforts contributed to any differential vaccination trends. For cross-sectional outcomes, we used a propensity score weighting method to eliminate further differences in observed characteristics between the non-California sample and the California samples that are confounded with sample membership. Similar to the principle of randomization, propensity score weighting was intended to maximize internal validity by balancing observable characteristics (eg, occupation) that both influence vaccination and differ across the 2 samples. The advantage of this approach over regression-based methods is that it does not assume a linear relationship between sample member characteristics and the outcome of interest. The propensity score is the estimated probability of being assigned to the treatment group conditional on respondent characteristics (as listed in Table 1). We weighted observations in the California sample and the non-California sample by reciprocal of the propensity score and the reciprocal of one minus the propensity score, respectively. We estimated the average treatment effect by applying a weight-adjusted 2-sample comparison procedure. In the absence of appropriate population benchmarks, we did not apply sampling weights to our data to generate population-based coverage estimates. The propensity score weighting estimation was implemented in STATA SE 11 (Stata Corp, College Station, TX) and by the weight-adjusted estimation option. Finally, we accounted for the multiple comparisons to avoid false findings of significance in the collection of all tests. Instead of using
RESULTS
DISCUSSION We used a DID approach to measure the impact of a 2006 law requiring hospital-based HCP in California to be vaccinated for influenza. To our knowledge, this study was the first to evaluate the impact of a state-level vaccination requirement. Using a variety of self-reported measures of vaccination receipt during the past decade, we found no evidence that California’s law was associated with higher self-reported vaccination among a large, diverse sample of the state’s hospital-based HCP following the law’s implementation, compared with hospital-based HCP in other states
K.M. Harris et al. / American Journal of Infection Control 42 (2014) 288-93
291
Table 1 Descriptive statistics by residential location Raw Respondent characteristic (%)
Propensity weighted CA, n ¼ 1,201
Non-CA, n ¼ 1,610
P valuey
.10
62.0
59.0
.27
78.9 7.1 6.0 8.0
Contents lists available at ScienceDirect
American Journal of Infection Control
American Journal of Infection Control
journal homepage: www.ajicjournal.org
Major article
The impact of influenza vaccination requirements for hospital personnel in California: Knowledge, attitudes, and vaccine uptake Katherine M. Harris PhD a, *, Lori Uscher-Pines PhD a, Bing Han PhD b, Megan C. Lindley MPH c, Suchita A. Lorick DO, MPH c a b c
RAND Corporation, Arlington, VA RAND Corporation, Santa Monica, CA Centers for Disease and Prevention, Atlanta, GA
Key Words: State policy Immunization Hospitals Survey research Difference-in-difference
Background: Seasonal influenza infections are a leading cause of illness, death, and lost productivity. Vaccinating health care personnel (HCP) can reduce transmission of influenza virus to patients and reduce influenza-related absenteeism, enabling the health care system to meet elevated demand for care during influenza outbreaks. Objectives: We evaluated the impact of California’s 2006 influenza vaccination requirement for hospital workers (requiring vaccination or signed declinations) on uptake and vaccination-related attitudes, beliefs, and knowledge among hospital HCP. Methods: We used a causal difference-in-differences approach to compare changes over the prior 10 years in the self-reported frequency of influenza vaccination for California hospital HCP and those from other states without similar laws using data from a stratified sample (N ¼ 3,529) of HCP drawn from online survey panels. We also examined cross-sectional differences in awareness of vaccination policies, promotion efforts, and attitudes toward influenza vaccination. All analyses used propensity score weighting to balance the observable characteristics of the 2 samples. Results: We found that compared with their counterparts in other states, California hospital HCP were (1) more likely to report working under a formal written policy for influenza vaccination, (2) no more likely to be vaccinated, and (3) less likely to report working for an employer who provided financial incentives for vaccination or rewarded or recognized employees for being vaccinated. Conclusion: Our results suggest that state-level vaccination requirements such as those enacted by California, may not be sufficient to increase uptake among hospital HCP. Copyright Ó 2014 by the Association for Professionals in Infection Control and Epidemiology, Inc. Published by Elsevier Inc. All rights reserved.
Seasonal influenza infections are a leading cause of illness, death, and lost productivity.1 Vaccinating health care personnel (HCP) can reduce transmission of influenza virus to patients2,3 and reduce influenza-related absenteeism, enabling the health care system to meet elevated demand for care during influenza outbreaks.4,5 The US Centers for Disease Control and Prevention’s
* Address correspondence to Katherine M. Harris, PhD, The RAND Corporation, 1200 South Hayes Street, Arlington, VA 22202. E-mail address: [email protected] (K.M. Harris). Supported by the US Centers for Disease Control and Prevention, Atlanta, GA. This work was performed as part of a cooperative agreement (1 U01 IP000416-02) with the US Centers for Disease Control and Prevention. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention or the California Department of Public Health. Conflicts of interest: None to report.
(CDC) Advisory Committee on Immunization Practices has recommended annual influenza vaccination of all HCP.6 For most of the past decade, however, less than half of all HCP received annual influenza vaccinations.7 The perception that voluntary programs are insufficient to generate substantial increases in influenza vaccination has led numerous experts, professional societies, advocacy organizations, and others to call for making influenza vaccination a condition of employment and/or professional privileges.8-12 However, data suggest that even vaccination requirements without strict penalties for noncompliance (defined here as job loss, loss of professional privileges, or visible identification of vaccination status) can increase vaccination among hospital HCP by 15 percentage points or more, depending on how they are implemented.13-16 The effectiveness of influenza vaccination requirements beyond those adopted by hospitals voluntarily remains unclear. The
0196-6553/$36.00 - Copyright Ó 2014 by the Association for Professionals in Infection Control and Epidemiology, Inc. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ajic.2013.09.030
K.M. Harris et al. / American Journal of Infection Control 42 (2014) 288-93
experience of California may provide lessons regarding the potential of relatively permissive state-level requirements. We define permissive requirements here as those that allow individuals to avoid vaccination by signing a declination statement without establishing an enforcement mechanism or penalties for noncompliance. As part of legislation to strengthen prevention of hospitalacquired infections, California became one of the first states to require influenza vaccination of all hospital-based HCP in 2006.17 As of March 2012, 7 other states have enacted similar laws.17,18 Specifically, California Health and Safety Code 1288.7 requires all general acute care hospitals to offer annually on-site influenza vaccinations to hospital employees at no cost and to require their employees to be vaccinated or declare in writing that they have declined vaccination. In accordance with CDC’s and the federal Healthcare Infection Control Practices Advisory Committee’s recommendations, in 2008 California Department of Public Health clarified requirements for compliance with the statute, including the provision of education to employees and other nonemployee HCP and the reporting to California Department of Public Health of annual vaccination and declination rates.19-21 The statute does not make vaccination a condition of employment, stipulate means through which hospitals should enforce requirements, or institute penalties for hospitals with low vaccination rates. As such, the California law can be considered a permissive state-level requirement in contrast to Rhode Island’s 2012 law, which requires unvaccinated HCP to wear masks during periods where influenza is widespread and includes specific penalties for noncompliance.22 Rhode Island’s law is currently being evaluated; however, at this time an evaluation of California’s law is critical to understanding whether success can be achieved without threatening HCP autonomy. To learn from California’s experiences, we evaluated the impact of California’s 2006 influenza vaccination requirement for hospital workers (requiring vaccination or signed declinations) on uptake and vaccination-related attitudes, beliefs, and knowledge among hospital HCP. Our results may assist policy makers to understand the potential effectiveness of state policies that strive to make the promotion of influenza vaccination a high priority without imposing strict penalties for noncompliance. METHODS Data sources In the absence of appropriate administratively documented vaccination coverage data, we drew a stratified sample of 3,529 hospital-based HCP in April 2010 from 3 online survey panels. These panels included (1) the Physicians’ Consulting Network (PCN) (N ¼ 846), a panel of approximately 70,000 HCP (mostly physicians) who agreed to participate in health care-related research in exchange for financial compensation; (2) eRewards (n ¼ 2,113) and GMI (n ¼ 99), which are 2 opt-in panels of Internet users and both of which include tens of thousands of individuals recruited through Web advertising who agreed to participate in surveys in exchange for small cash rewards and incentives; and (3) hospital-based HCP (n ¼ 471) recruited from KnowledgePanel, a nationally representative, online research panel of 50,000 individuals, who completed the same survey instrument as part of a concurrent CDC study to measure influenza vaccination among US HCP.23 All data sources are convenience samples with the exception of KnowledgePanel. To ensure representation of a diverse range of health care professionals and vaccination-related attitudes and behaviors, we prospectively recruited 4 strata or categories of HCP employed by or
289
working in general acute care hospitals: (1) physicians typically employed by hospitals (eg, anesthesiologists, surgeons), (2) nurses, (3) allied health professionals, and (4) nonclinical support staff and administrators. We recruited physicians, nurses, and selected types of allied health professionals through targeted e-mails using occupational information collected at the time of recruitment into their respective panels. We recruited others using panel-wide e-mails. When feasible, we balanced the number of sample members from each source residing inside and outside of California to minimize the potential for the confounding of sample source and exposure to CA’s vaccination requirement. Next, we excluded 297 respondents who resided in 7 states (Alabama, Illinois, Massachusetts, New Hampshire, Tennessee, Maine, and Oklahoma) that had laws similar to California’s during our study period. This exclusion ensured that the non-California “control” arm of the study only included HCP who were not subject to state-wide legislation requiring vaccination. This exclusion did not change our results. We also excluded 421 respondents who reported working in hospital-owned, ambulatory care practices. Seventy percent of eligible KnowledgePanel panelists and 22% of PCN panelists completed the survey. We cannot determine participation rates for opt-in panelists because the number of eligible people invited to participate is unknown. All 4 sample sources have been described in previous publications.23,24 After applying exclusion criteria, our analytic sample consisted of 1,201 California and 1,610 non-California hospital-based HCP. RAND’s Institutional Review Board approved the study design and survey protocols. Measures The survey measured vaccination-related behaviors, attitudes, and knowledge and exposure to workplace vaccination efforts. The survey asked 6 questions about influenza vaccination during the prior 10 years. We expected that HCP could accurately recall the frequency of influenza vaccination in the recent past, although not necessarily in specific terms. To aid recall, we started by asking about vaccination for seasonal and H1N1 influenza during the past season. We followed-up with 2 questions: (1) “Compared with ten years ago, are you more or less likely to get an influenza vaccination now than you were in the past?” and (2) “Please estimate how often you have been vaccinated for influenza over the past ten years.” We also gauged respondents’ exposure to workplace vaccination efforts. We started by asking respondents whether their employer had a “formal, written policy on influenza vaccination.” We measured respondents’ perceptions of working under a vaccination requirement by asking, “During this past influenza season, did your work site recommend or require that you be vaccinated for seasonal influenza?” To detect the presence of a requirement we asked whether unvaccinated HCP were required to sign a declination statement or faced consequences for not complying with their employer’s policy, including requiring a special badge, mask, or respirator, work reassignment, participation in an educational intervention, leave without pay, loss of benefits, and termination. The respondent was considered subject to a vaccination requirement if the respondent reported a declination requirement or any of the listed consequences of nonvaccination. The survey asked about other efforts to promote vaccination. These included on-site vaccination, reminders, publicizing the risks and benefits of vaccination, public identification of vaccinated staff, financial incentives, other rewards, and publicizing employee vaccination rates. We asked about vaccination-related attitudes and beliefs that could potentially be influenced by efforts to educate employees
290
K.M. Harris et al. / American Journal of Infection Control 42 (2014) 288-93
about vaccination benefits. Questions included agreement with statements about HCP susceptibility to influenza inflections and understanding of influenza transmission as described in CDCissued guidance on the topic.25 Finally, we asked about occupational and sociodemographic characteristics, characteristics of respondents’ hospitals, health status, and contact outside of work with individuals at increased risk of influenza.
the standard cutoff of P < .05, we claimed statistical significance only when the reported P value could achieve a more stringent false discovery rate (FDR) threshold of .05) to report working for an employer who provided financial incentives for vaccination, 2.6 percentage points less likely (P < .01; FDR < .05) to report working for an employer who rewarded groups of employees for vaccination, and 5.0 percentage points less likely (P < .01; FDR < .05) to report working for an employer who promoted competition between units to attain increased influenza vaccination coverage (not shown). To assess the robustness of our findings to the composition of our sample, we estimated propensity weighted differences separately by occupational grouping (ie, doctors, nurses, allied health professionals, and nonclinical administrative and support staff) and found similar results (not presented). Likewise, we excluded those not directly employed by the hospitals in which they were working and, again, found similar results (not shown).
Accurate measurement of policy impacts was complicated by the lack of preimplementation data and random assignment to vaccination policies. To address these challenges, we used a causal difference-in-differences (DID) approach, which has been shown in the literature to be superior to the classic regression DID approach with less bias and smaller mean-square error.26-28 The classic DID approach compares the differences in the mean outcome before and after the treatment between the treatment and control groups, which is usually implemented as a regression or analysis of covariance. The causal DID approach we used employs a propensity score weighting method to balance the observed characteristics between the 2 groups, which avoids the potential model misspecification in the classic DID. The propensity score is the probability that a respondent received treatment conditional on observed characteristics. In our DID approach, we compared changes across the 2 samples in the self-reported frequency of being vaccinated during the 4 years prior to the enactment of the California requirement (20012005) and 4 years following enactment (2007-2011). We defined the pre- and post- periods in this way to account for the phase-in of the requirement and to have time periods of equal length to increase comparability. We considered a statistically larger pre-post difference favoring California to be evidence that the California requirement increased vaccination among HCP compared with other states. We also compared reports of ever being vaccinated for influenza, being more likely to be vaccinated now compared with 10 years ago, being vaccinated during the 2010-2011 season, being subject to vaccination promotion efforts, and HCP attitudes and beliefs across the 2 samples to identify the mechanisms through which workplace promotion efforts contributed to any differential vaccination trends. For cross-sectional outcomes, we used a propensity score weighting method to eliminate further differences in observed characteristics between the non-California sample and the California samples that are confounded with sample membership. Similar to the principle of randomization, propensity score weighting was intended to maximize internal validity by balancing observable characteristics (eg, occupation) that both influence vaccination and differ across the 2 samples. The advantage of this approach over regression-based methods is that it does not assume a linear relationship between sample member characteristics and the outcome of interest. The propensity score is the estimated probability of being assigned to the treatment group conditional on respondent characteristics (as listed in Table 1). We weighted observations in the California sample and the non-California sample by reciprocal of the propensity score and the reciprocal of one minus the propensity score, respectively. We estimated the average treatment effect by applying a weight-adjusted 2-sample comparison procedure. In the absence of appropriate population benchmarks, we did not apply sampling weights to our data to generate population-based coverage estimates. The propensity score weighting estimation was implemented in STATA SE 11 (Stata Corp, College Station, TX) and by the weight-adjusted estimation option. Finally, we accounted for the multiple comparisons to avoid false findings of significance in the collection of all tests. Instead of using
RESULTS
DISCUSSION We used a DID approach to measure the impact of a 2006 law requiring hospital-based HCP in California to be vaccinated for influenza. To our knowledge, this study was the first to evaluate the impact of a state-level vaccination requirement. Using a variety of self-reported measures of vaccination receipt during the past decade, we found no evidence that California’s law was associated with higher self-reported vaccination among a large, diverse sample of the state’s hospital-based HCP following the law’s implementation, compared with hospital-based HCP in other states
K.M. Harris et al. / American Journal of Infection Control 42 (2014) 288-93
291
Table 1 Descriptive statistics by residential location Raw Respondent characteristic (%)
Propensity weighted CA, n ¼ 1,201
Non-CA, n ¼ 1,610
P valuey
.10
62.0
59.0
.27
78.9 7.1 6.0 8.0
