Data

We used data from the 2006 Harvard School of Public Health (HSPH) Pandemic Influenza Survey, a random digit–dial survey sponsored by the HSPH Project on the Public and Biological Security. The survey was conducted to provide information with regard to the public’s reaction to the possible use of social distancing and other nonpharmaceutical interventions during a severe outbreak of pandemic influenza. Survey questions assessed willingness and ability to comply with public health recommendations in 3 domains: home, school, and work.

The target population was adults >18 years of age, who lived in the United States. During September 28 through October 25, 2006, International Communications Research (Media, PA, USA) surveyed a representative sample of adults >18 years of age, including an oversample of adults with children <18 years of age in the household. Sampling procedures are described elsewhere (8). Response rate was 36% and cooperation rate was 75%, which produced a total of 1,697 completed interviews (8). Because adults with children were oversampled to gauge the possible effect of community mitigation on families, data were weighted to reflect the actual proportion of the total adult population. In addition, to compensate for nonresponse bias and unequal probability of selection and to ensure that demographic groups were represented in their actual proportion in the adult population, we weighted sample data to the most recent US census data available from the Current Population Survey (www.census.gov/cps) for gender, age, race, region, and education.

Employment characteristics for the full sample were 50% employed full time, 13% employed part time, 36% unemployed, and 1% unknown. Our study focused on employment-related constraints that may limit ability to comply with isolation recommendations; therefore, we analyzed responses from only the 1,101 respondents who were employed either full or part time (Table 1). A small percentage of employed respondents refused to answer some items in the survey (e.g., 15% refused income, 3% education, 2% race).

Outcome Variables: Indicators of Ability to Comply

Employed respondents were asked a series of questions to assess real or perceived constraints with regard to their ability to comply with pandemic influenza mitigation strategies that require isolation from the workplace. The 4 dependent variables were positive responses to the following statements: “It is likely that I or a member of my household would lose a job or business as a result of having to stay home for 7–10 days” (dichotomized; “very” and “somewhat” likely represented a positive response and “not too” and “not at all” likely represented a negative response); and “I would have serious financial problems if I stayed away from work for 7–10 days, 1 month, or 3 months.” The latter questions used a split sample, whereby only respondents who answered “no” to the 7–10 day duration were asked about financial problems that they would have at 1 month, and only those who answered “no” to having serious financial problems at 1 month were asked about the 3-month period. Many unadjusted prevalence estimates for each outcome variable (Table 2) have been described in a descriptive study that used the 2006 HSPH Pandemic Influenza Survey (8).

Predictor Variables

To assess the effect of potentially modifiable employment-related constraints on compliance with recommendations that require missing work, we chose key predictor variables that represented selected employment characteristics; i.e., inability to work from home, lack of pay when absent from work, and self-employment status (Table 2). With regard to employment-related constraints, <1% of respondents refused to answer 2 questions.

Covariates

To assess the potentially disproportionate difficulties that low-income and urban populations may face if asked to stay home from work in the event of a serious outbreak, we included income and urban residence in all models. Other sociodemographic and personal characteristics assessed were education, race/ethnicity, age, sex, self-reported health status, and self-reported knowledge of pandemic influenza.

Statistical Analysis

Structured analytical approaches have improved predictions of behavior in emergency situations, modeling the joint effects of several factors on planned behavior (3). Therefore, to identify factors that may lead to disproportionate vulnerability in the event of a serious outbreak, we used multivariable logistic regression to model the predicted probability that some groups of working adults (delineated by employment characteristics such as inability to work from home, lack of pay when absent from work, and self-employment) may be less able than identified referent groups to comply with pandemic influenza mitigation strategies that require voluntary isolation from work.

Each outcome used 1 full model; all models controlled simultaneously for all key predictors as well as covariates such as income, urban residence, age, race/ethnicity, education, gender, self-reported health status, and self-reported knowledge of pandemic influenza. We conducted a complete case analysis and analyzed only those 1,101 respondents who reported being employed either full or part time. Tests of significance were estimated at p<0.05, and 95% confidence intervals (CIs) were reported for all odds ratios (ORs). To adjust for unequal probabilities of selection and for potential nonresponse bias, we applied individual weighting factors to all estimates. The analysis was conducted with SAS software version 9.2 (Cary, NC, USA) by using the PROC SURVEYLOGISTIC procedure and probability sampling weights.

Perceived Likelihood of Losing Job or Business

Of the employed respondents, 28% reported that they likely would lose their job or business as a result of having to stay home from work for 7–10 days in the event of a pandemic influenza outbreak (Table 2). Our multivariable models elucidated differential vulnerability by lack of paid sick leave, income level of respondents, and urban residence (Table 3). Those respondents who would not be paid if kept from work were almost 5× more likely than those who would receive pay (OR 4.72) to say that they would likely lose their job or business as a result of having to stay home from work for 7–10 days.

Respondent income also was associated with reported likelihood of losing a job or business. Those who earned <$30,000 per year were 4× more likely than those who earned >$75,000 per year (OR 4.31) and those who earned $30,000–$49,000 per year and $50,000–$74,000 per year were ≈2× more likely than those who earned >$75,000 per year (ORs 1.70 and 2.08, respectively) to say that they would likely lose their job or business as a result of having to miss work in the event of a serious outbreak.

Urban residence was associated with limited ability to comply with recommendations that require missing work. Respondents living in urban areas were ≈70% more likely than those living in rural areas to say that they would likely lose their job or business as a result of having to stay home for 7–10 days in the event of an outbreak (OR 1.66).

Perceived Likelihood of Experiencing Serious Financial Problems

Certain employment characteristics and respondent income levels were associated with the likelihood that working adults would experience serious financial problems and thus be less able to comply with isolation recommendations, if required to miss work for long periods of time (Table 3). Respondents who were self-employed were twice as likely as those who worked for an employer to say that they would experience serious financial difficulties if isolated from work for 7–10 days (OR 2.09). Those who were not able to work from home were significantly more likely than those who were able to work from home to say that they would experience serious financial problems if isolated from work for durations of 7–10 days, 1 month, and 3 months (ORs 1.57, 1.91, 1.65, respectively).

Respondent income also was associated with likelihood of experiencing serious financial problems if the respondent were kept from work because of an outbreak of pandemic influenza, although the models showed interesting patterns, depending on the duration of isolation. If isolated from work for 7–10 days, those who earned <$30,000 per year were 3× more likely than those who earned >$75,000 per year to say that they would experience substantial financial problems (OR 3.26). At 1 month of isolation, all those in low- and middle-income groups were significantly more likely than those in the highest income group to say that they would experience serious financial problems if kept from work. There was a fine income gradient; those earning <$30,000 per year were 3.29× more likely, those earning $30,000–$49,000 per year were 2.93× more likely, and those earning $50,000–$74,000 per year were 1.89× more likely than those earning >$75,000 per year to say that staying away from work for 1 month would pose serious financial problems. At 3 months of isolation, the trend shifted somewhat. Low-income workers were still significantly more likely than high-income workers to say that they would have serious financial problems if isolated from work (OR 3.52), indicating disproportionate vulnerability for low-income populations across all durations of isolation. However, at 3 months, middle-income workers would not be more or less likely than those earning >$75,000 to say that they would experience serious financial problems, indicating that a 3-month period of isolation would likely be difficult for those in all income groups, including those in the highest income categories.