Closing Schools in Response to the 2009 Pandemic Influenza A H1N1 Virus in New York City: Economic Impact on Households

Abstract

Understanding the effectiveness of a school closure in limiting social interaction and the economic impact of school closure on households is critical when developing guidelines to prevent spread of pandemic influenza. A New York City survey conducted in June 2009 in 554 households affected by the 2009 pandemic influenza H1N1–related school closures showed that, during closure, 30% of students visited at least 1 locale outside their homes. If all the adults in the home were employed, an ill child was less likely to leave home. In 17% of the households, at least 1 adult missed some work because of the closure. If all adults in the home were employed, someone was more likely to take time off work. If other children were in the household, it was less likely that an adult took time off work. The findings of our study will be important when developing future pandemic school-closure guidance.

On 23 April 2009, an outbreak of a novel strain of pandemic influenza A (H1N1) (pH1N1) was identified in a New York City high school[1, 2]. This strain spread quickly throughout the city. In spring 2009 (schools were closed from the end of April through the end of June), and 57 schools in NYC were closed [3] for 5–7 days because high numbers of children were identified with influenza-like illnesses (ILI) (Influenza-like illness is an illness presenting as a fever and either a cough or sore throat.) [4]. School closure has been promoted as a nonpharmacological intervention that can mitigate the impact of an influenza pandemic [5]. However, the costs and potential benefits of such an intervention have not been fully studied [6, 7]. We measured the extent to which children stayed isolated, thus potentially slowing disease transmission during the closure, and the extent to which household inhabitants missed work because of the closure.

METHODS

We conducted a telephone survey among households whose children attended 1 of 7 public elementary schools closed during the first week of June 2009 in response to outbreaks of pH1N1. The survey asked adult respondents if the student had an ILI in the week before or during the closure, if someone had to take time off from work to look after the child, what the primary location of the child was during the closure, and whether the child left the primary location at all during the school closure and if so, where he or she went. The survey also collected information regarding household composition, employment, and income.

We used an online survey platform (mrInterview) to administer the survey, which were made available in 3 languages: English, Spanish, and Mandarin. We made calls from the 11th to the 25th of June 2009. The New York City Department of Education provided a student roster for the 7 schools. Using this list, we identified and attempted to contact a total of 3,343 households with ≥1 children in the affected schools.

We conducted bivariate analyses comparing the health of the child (ILI) to each location visited, in order to identify any potential patterns. We also analyzed the data using the following general (logistic) regression models: 1) Was household time lost from work due to school closures? (no: 0, yes: 1); 2) were all adults in household employed? (no: 0, yes: 1); 3) what was the school grade of child (children) (kindergarten-fourth: 0, fifth-sixth: 1); 4) Did child (children) have ILI a week before or during closure? (no: 0, yes: 1); 5) Are other children in the home? (no: 0, yes: 1); What is the household income? (<$25,000: 0, ≥$25,000: 1).

Children left primary place of residence during closure (no: 0, yes: 1); = All adults in household employed (no: 0, yes: 1); school grade of child (kindergarten-fourth: 0, fifth-sixth: 1), child had ILI week before or during closure (no: 0, yes: 1); other children in home (no: 0, yes: 1); household income (<$25,000: 0, ≥$25,000: 1).

We ran 2 regression models for each of the above equations: one that included income as a predictor variable and one that did not include this variable (Table 3). We then converted the predictor values to probabilities (Table 2). We used Robust Huber/White standard error estimates to adjust for school-level clustering. Statistical analysis was performed using Stata, version 11.0 (StataCorp).

Table 1.

Characteristics of 554 households surveyed: School closures due to the 2009 pandemic H1N1 influenza A virus outbreaks, New York City, Spring 2009

Variable (total, missing)	No.	%
Grade of childa (550, 4)
Pre-K to Fourth	472	86%
Fifth	78	14%
Child sick with ILI (551, 3)
Ill with an influenza-like illnessb in the week before or during the school closure	308	56%
Not sick with an influenza-like illnessb	243	44%
Other children (554, 0)
No additional children in household	206	37%
At least one additional child in household	348	63%
Income (303, 251)
Less than $25,000	219	72%
Greater than or equal to $25,000	84	28%
Employment (554, 0)
Households with at least one adult not employed	311	56%
Households with all adults partially or fully employed	243	44%
Time off work (521, 33)
No one took time off work	433	83%
At least one adult took some time off work due to closure	88	17%
Visiting other venues (554, 0)
Child did not leave the primary location during closure	389	70%
The child left the primary location at least once during the closure	165	30%

a

Grade of child (a cut off at fifth grade was used, as there was a clear change in parental behavior with children in the fifth grades versus K - fourth grades, as opposed to a gradual change, when the variables were individually examined as dummy variables).

b

An influenza-like illness was defined as having a fever and either a cough or sore throat.

Table 2.

Probabilities of Time Lost from work and children leaving their primary residence during school closures due to the 2009 pandemic H1N1 influenza A virus outbreak: New York City, Spring, 2009a

	Household lost time from work due to school closuresc						Children left primary residencea,d
	Including income (n = 290, R2 = .1090)			Excluding income (n = 514, R2 = .1182)			Including income (n = 299, R2 = .0331)			Excluding income (n = 547, R2 = .0149)
Variable	Coef.b	Robust Std. Error	P	Coef.b	Robust Std. Error	P	Coef.b	Robust Std. Error	P	Coef.b	Robust Std. Error	P
Baselinee	.11	.02	.00	.08	.01	.00	.40	.03	.36	.34	.02	.00
All adults in home employed	.24	.03	.00	.25	.02	.00	−.14	.02	.02	−.10	.01	.00
Grade of child in school > 4th	−.06	.02	.16	−.03	.01	.14	−.05	.01	.62	−.04	.01	.55
Child sick with ILIf	.06	.01	.06	.04	.01	.27	−.08	.02	.32	−.06	.01	.08
Other children in household	−.04	.01	.07	−.03	.01	.01	.01	.01	.88	.04	.01	.40
Household income ≥ $25,000	.02	.01	.52				−.10	.02	.25

a

The coefficients are expressed as probabilities of occurrence. Thus, the probability of a baseline household (see note above) losing time from work due to the school closures is 9% (.09). The impact of the other variables is relative to that probability. For example, having all adults in household employed increases the probability of work time lost by 33%, so that households with all adults employed have a 42% (9% + 33%) probability of reporting work time lost.

b

where ⁠, we calculated the coefficient by subtracting the difference between for each coefficient and the baseline (see Table 3 for the original values)

c

Logit (p (Household lost time from work due to school closures)) = Employment of adults + Grade of child + Health of Child + Other children in home + Income

d

Logit (p (Children leave primary location)) = Employment of adults + Grade of child + Health of Child + Other children in home + Income

e

The baseline (constant) represents a home where at least one adult is not employed and the income is less than $25,000, with an only child in the home who is in kindergarten through fourth grade and who did not have an influenza-like illness before or during school closure.

f

ILI = influenza-like illness, defined as having a fever and either a cough or sore throat.

Table 3.

Logistic models (original values) of time lost from work and children leaving their primary residence during school closures due to the 2009 pandemic H1N1 influenza A virus outbreak: New York City, Spring, 2009

	Household lost time from work due to school closuresa						Children left primary residenceb
	Including income (n = 290, R2 = .1090)			Excluding income (n = 514, R2 = .1182)			Including income (n = 299, R2 = .0331)			Excluding income (n = 547, R2 = .0149)
Variable	Coef	Robust Std. Error	P	Coef	Robust Std. Error	P	Coef.	Robust Std. Error	P	Coef.	Robust Std. Error	P
Baselinec	−2.10	.28	.00	−2.42	.22	.00	−.40	.43	.36	−.65	.20	.00
All adults in home employed	1.49	.28	.00	1.72	.33	.00	−.66	.27	.02	−.48	.12	.00
Grade of child in school >4th	−.92	.65	.16	−.58	.39	.14	−.20	.41	.62	−.18	.30	.55
Child sick with ILId	.48	.25	.06	.46	.41	.27	−.33	.33	.32	−.26	.15	.08
Other children in household	−.47	.26	.07	−.46	.18	.01	.05	.34	.88	.19	.23	.40
Household income ≥ $25,000	.15	.23	.52				−.46	.40	.25

a

Logit (p (Household lost time from work due to school closures)) = Employment of adults + Grade of child + Health of Child + Other children in home + Income.

b

Logit (p (Children left primary location)) = Employment of adults + Grade of child + Health of Child + Other children in home + Income.

c

The baseline (constant) represents a home in which at least one adult is not employed and income is less than $25,000, where there is a child in kindergarten through fourth grade who is an only child in the home and who did not have an influenza-like illness before or during the school closure.

d

ILI, influenza-like illness, defined as having a fever and either a cough or sore throat.

RESULTS

Response Rate

Of the 3,343 households, surveyors attempted to call each of these households at least once. A total of 1,034 (31%) of the households were successfully contacted. Of those, 587 (56.8%) agreed to participate; 554 completed the survey.

Characteristics of the Sample

The median age of the children was 8 years (second grade). Fifty-six percent were sick with an ILI in the week before or during the closure week. Forty-two percent of the children had missed some school in the week before the closure, and 50% of those absences were attributable to an ILI. Among the 55% of respondents who reported household income, 28% made ≥$25,000 annually.

Work Lost because of School Closure

Eighty-three percent of households reported losing no time from work because of school closure (Table 1). Overall, 44% of households reported that all adults in the household were employed. A baseline household, which had at least 1 adult not employed, an annual income less than $25,000, a child in kindergarten through fourth grade, only 1 child in the household, and no ILI in the child before or during the school closure, had an 8% (.08) probability of reporting time lost from work (Table 2). Having all adults employed increased the probability of work time lost by 25%, as compared to the baseline. Therefore, we concluded that households with all adults employed had a 33% (.08 + .25) probability of reporting work time lost. Having other children in the household decreased the probability that a household would miss work by 3%, to 5% (Table 2).

Location of Children During School Closure

Most students claimed that their home was their primary location during the full closure (72%). During school closure, 30% of children left their primary location at some point, with the most common locale being a playground (Figure 1). A baseline household had a 34% probability of a child visiting any other venue. If all adults were employed, the probability of children leaving the home decreased by 10% , to 24% (Table 2). We ran the regression with and without (not shown) playground visits due to concerns that an outdoor environment was very different than going to an indoor environment. However, except for 2 children, all children who visited a playground also visited an indoor environment other than their primary locale; therefore, the results of the 2 analyses are nearly identical.

Figure 1.

Percentage of school children visiting locations during the weekdays when schools were closed due to outbreaks of 2009 pandemic H1N1 influenza A virus outbreaks:New York City, Spring, 2009.

DISCUSSION

Compared with similar studies in suburban and rural settings (ie, North Carolina in 2006 and Pennsylvania in 2009), we found that 40%–59% fewer households in New York City reported their children going outside the home during the school closure [8, 9]. This may reflect differences among urban, suburban, or rural households. However, it may also suggest that school closures may be more effective in isolating children in urban areas. The findings on lost work time are similar to the study in Pennsylvania [9]. In both cases, the majority of households reported no lost time from work because of the school closures. This is not surprising, as more than half of the households surveyed had at least 1 nonworking adult at both locales. As expected, those households where all adults worked outside the home had a higher probability of reporting time lost from work.

This study has several limitations. First, it does not examine how many households missed work because of their child's ILI in the previous week to allow for a true impact analysis. It is possible that just as many household may have missed work because of their child's or their own, actual or potentially averted ILI. We know that 36% of the children had an ILI before and 24% had the illness during school closure. In 14% of homes, at least 1 adult was ill before school closures, and in 10% of the homes at least 1 adult was ill with an ILI during the closure. Second, the sample from New York City may not be typical of many US cities. Parents in New York City may have had heightened anxiety because the outbreak hit there early with intense media coverage. In addition, in order to limit recall bias, the study was limited to those schools that closed in the week before the survey was conducted. Therefore, this survey is not a random-population sample. Third, representation of socioeconomic status was difficult to assess, because many of those surveyed refused to provide income data. However, among those who did respond, income levels were remarkably low, and it is possible that the results may not be applicable to other communities with different socioeconomic status. Finally, many households did not answer the phone, potentially over-representing the survey toward those households with a non-working adult. However, among those who did participate, respondents were representative in terms of grade and school distribution. In addition, 35.3% of the households in our study are single-parent households (16% are single-adult households), which is between the national mean of 29.5% [10] and New York City’s mean of 41.5%. (This number represents single-parent households as a percentage of family households with children younger than 18 years old.) [11].

Many models and studies have examined the impact of school closures on a pandemic [12–24]. However, we found only a handful of studies that examined this issue from the household perspective [8, 9, 25–27]. This study provides insights on how urban households respond to school closure. Results from this, as well as other studies (eg North Carolina and Pennsylvania studies [8, 9]), should be used to update the models to determine a more realistic estimate of school closure costs to the household. This is critical to policy planning, as preliminary data have shown that school outbreaks have accounted for 16% of the cases of pH1N1 in the United States [28]. Because the effectiveness of this intervention continues to be debated, we hope that these data will help public health decision makers, school officials, and parents of school children evaluate the costs and benefits of future school closures as a means to manage epidemics in a community.

We thank the New York City Department of Education, for its tremendous assistance and cooperation, as well as Shane Grebel and his team for their data collection support.

Financial support. This information was collected and paid for by CDC; it was cleared by the human subjects’ board as a public health, non-research study to assess an existing public health intervention in response to an immediate public health threat.

Supplement Sponsorship. Published as part of a supplement entitled “The 2009 H1N1 Influenza Pandemic: Field and Epidemiologic Investigations,” sponsored by the Centers for Disease Control and Prevention.

Potential conflicts of interest. All authors: none reported.

Author notes

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.