Teachers on the margins: How low-income public schools burden our teachers

2.1 Study design and recruitment

A cross-sectional study was conducted with K-12 teachers from four public school districts in southeast Michigan from March through May 2017. School districts were selected by the median household income level of the district, to include high and low-income school districts. Low-income school districts were classified as those with a median household income below the qualification level for the U.S. National School Breakfast and Lunch Program for Michigan for a family of four ($46,435) in 2017, which provides free or low-cost meals for low-income students [22]. High-income school districts were classified as those with a median household income approximately double this level (at least $90,000).

No more than 25 teachers were recruited per school district. All teachers were eligible to participate if they had been in their current positions since at least the start of the 2016-2017 academic year and employed by the school district for at least one year. Institutional Review Board approval (ID: 907079) was obtained through Oakland University. Permission to work in each school district was granted through the Superintendent or Assistant Superintendent of each school district. All participants received a $50 Amazon gift card at the completion of sample collection.

The authors employed two convenience sampling recruitment methods for this study. For three school districts, the Superintendent or Assistant Superintendent forwarded a recruitment email to teachers asking for participation in this study. One of the school districts invited research personnel to recruit participants through a table at an annual school district event. Teachers were contacted individually to schedule a meeting time with a member of the research team during a break, prep hour, or lunch period, or prior to the start or at the end of their school day. During each meeting, teachers were surveyed and their primary classrooms and schools were assessed.

2.2 Surveys and assessment

Each participating teacher completed a survey on teaching roles, other employment, education level, sex, race, ethnicity, age, income, average hours of sleep per night, weight, height, and overall health; and smoking, alcohol consumption, asthma, and hypertensive history. Demographic, smoking, alcohol consumption, asthma, and hypertensive history questions were adapted from the U.S. National Health and Nutrition Examination Survey (NHANES) [23].

Occupational stressors were assessed through questions from the National Union of Teachers’ Teacher Stress Study, which were modified to add questions on school and classroom environments and to US American English [24]. Teachers rated how significantly they agreed with 39 occupational teaching stressor statements (e.g., strongly disagree, disagree, agree, strongly agree) on six categories (demands, control, support, relationships, roles, and work environment) [24]. Agreement to each statement was given a numeric value (1, 2, 3, or 4), with the highest numeric value (4) associated with the strongest level of concern/stress (i.e., either strong agreement or strong disagreement, depending on which aligned with strongest perceived stress for each statement). The numeric scores were summed and standardized out of four points into a standardized composite score for each of the six stressor categories. A standardized composite score for all teaching stressors was calculated to assess overall teaching stressors.

Participants’ body mass index (BMI), used to assess obesity status, was calculated by dividing the participant’s self-reported weight (kg) by her/his squared height (m). Smoking history was assessed to classify participants as current, past, and never smokers (i.e., smoked less than 100 cigarettes in their lifetime) [25]. Smoking pack-years was calculated from the average packs of cigarettes smoked per day for the duration of a participant’s smoking years. Alcohol consumption was estimated as the typical number of alcoholic drinks consumed per week.

2.3 Stress response measures

Participant’s blood pressure, heart rate, and salivary cortisol were measured to assess physiological stress response [26–28]. Salivary cortisol is a hormonal biomarker that is used to assess psychological stress [27, 29], as an indicator of biological stress reactivity [30]. Cortisol levels follow a diurnal pattern typically characterized by high levels upon waking and then declines throughout the day [27, 30]. Average salivary cortisol levels are 15.4 nmol/L in the morning and 2.8 nmol/L in the late evening [31]. Cortisol levels have been associated with stress levels [27], acute smoking [32, 33], alcohol consumption [34, 35], physical activity [33, 36], seasonality [33], age [37], and food and caffeine consumption [35].

Cortisol was measured through saliva samples [38]. Saliva samples were collected with Salimetrics Saliva Bio Oral Swabs (SOS), following the Salimetrics SOS collection methods (Salimetrics, LLC, Carlsbad, CA, USA) [39]. Each participant was coached on proper collection protocol and provided with a handout of protocols and the materials for sample collection. They were advised to avoid consuming any dairy products, acidic foods, or beverages for at least 20 minutes prior to collecting a sample; refrain from eating a large meal within one hour; and abstaining from alcohol for the duration of the sample collection period. The swab was placed under the tongue for at least 60 seconds until the swab was saturated in saliva and then placed into a labeled collection tube. Each sample tube was placed in a labeled Ziploc bag, where participants were directed to record the time and date of the sample.

Each participating teacher collected SOS samples twice a day for three days during the school week (Monday through Friday). Participants were advised to collect each sample at approximately the same time each day. They were encouraged to collect the first sample of the day prior to lunch when possible, which was either in the late morning or early afternoon. Three additional samples were collected about 30 minutes before the participants went to sleep. Teachers stored SOS samples in a refrigerator until a research team member was able to collect them, usually up to three days later. SOS samples were stored at –80°C before analysis.

SOS samples were analyzed for salivary cortisol through a DPC Coat-A-Count Cortisol modified protocol for saliva at the University of Michigan Core Assay Facility. The intra-assay coefficient of variation (CV) for salivary cortisol samples was 8.9%. The majority (89%) of samples had a CV below 15%. The inter-assay CV for salivary cortisol samples were 8.3–14.4%. SOS samples with too little saliva for analysis or with cortisol levels below the detection limit were excluded from analysis. Samples with missing time points or if they were taken more than four hours before/after other samples in the same time frame were excluded from analysis. Three individual salivary cortisol samples that had cortisol levels at more than five times the standard deviation of the average cortisol level of all participants (11.9 nmol/L) were excluded as outliers. Afternoon and evening salivary cortisol levels for all SOS samples were averaged for each participant (one to four samples per participant at each time point).

Blood pressure (BP) and hypertensive status are also used to assess chronic stress and biological stress response [26]. Teachers’ resting systolic and diastolic BP and heart rate were measured three times according to American Heart Association (AHA) guidelines [40], using an electronic sphygmomanometer (Omron Healthcare, Inc., Lake Forest, IL, USA) about 2 cm above the right elbow, above the brachial artery. Participants were asked to sit on a chair with back support for the duration of the survey and had the first measurement at least five minutes after sitting. The three BP and heart rate measurements were collected at three-minute intervals and were averaged. BP values were classified by the 2018 AHA standards for hypertensive status [41]. Under the 2018 AHA standards, apparent hypertension is defined as systolic BP (SBP) ≥130 mmHg or diastolic BP (DBP) ≥80 mmHg. Participants with elevated BP according to the 2017 AHA standards or who expressed interest were provided with AHA information sheets on high blood pressure [42]. Participants’ SBP and DBP were adjusted if they were currently taking anti-hypertensive medications to control their BP by adding 10 mmHg to SBP and DBP [43, 44].

2.4 School and classroom assessment

Participating teachers’ primary school and classroom were evaluated for environmental stressors including nearby industrial facilities and roadways, outdoor sound level, overall building maintenance, temperature, and classroom maintenance [45]. The classroom assessments were conducted at the time of the survey or immediately after if the survey took place outside of the teacher’s primary classroom. Classroom assessments evaluated the overall maintenance of the floor, ceiling, and walls; the presence of windows/doors and odors; room temperature; and humidity. School assessments of overall exterior maintenance and classroom assessments of interior maintenance were evaluated qualitatively by trained researchers.

Environmental assessments of a one (1) km area surrounding each school were conducted using online mapping tools by the research team. Traffic volume on federally-funded roadways within 1 km of each school were assessed using the Southeast Michigan Council of Governments (SEMCOG) database on the average annual daily traffic (AADT) [46]. The number of moderate AADT volume (5,000–10,000 vehicles/day) and high AADT volume (> 10,000 vehicles/day) roadways within a 1 km radius were tallied for each school. The number of sites of environmental contamination, designated by the Michigan Department of Environmental Quality (MDEQ), within 1 km of each school were assessed through the MDEQ’s Environmental Mapper program [47]. The U.S. Environmental Protection Agency’s (EPA’s) Toxic Release Inventory (TRI) data via the My Right-to-Know program was utilized to assess if there were any active or inactive facilities within 1 km of the schools that meet the qualifications to report to the Toxic Release Inventory [48] under the Emergency Planning and Community Right to Know Act (EPCRA, 1986). Schools were also designated as a neighborhood school if they were located in or surrounded by a residential neighborhood. Outdoor sound was measured with a Pulsar Nova sound level meter (Pulsar Instruments, Plc., North Yorkshire, UK).

2.5 Analysis

Data were analyzed with SPSS Statistical Software (v.24; IBM Corp., Armonk, NY, USA). Descriptive analyses, correlations, and statistical tests were performed on key variables, including demographics, reported stressors, salivary cortisol, BP, classroom characteristics, and school environmental hazards. Statistical significance was evaluated at p < 0.05. Cortisol levels were not normally distributed and therefore comparisons of means were assessed through the Mann-Whitney U test and bivariate analyses utilized Spearman’s rho. Bivariate correlation analyses were performed for standardized composite scores for teaching stressor categories and overall, salivary cortisol, SBP, DBP, heart rate, BMI, age, sex, cigarette pack-years, alcoholic drinks per week, average hours of sleep per night, average hours working at a second job, classroom temperature, environmental contamination and traffic volume within 1 km of the school, and the year school was built. All linear regression models included all participants with valid cortisol measurements. Linear regressions were performed with backward stepwise regressions and adjusted for district income level, teaching stressor scores, environmentally-contaminated sites, moderate and high vehicle AADT, age, sex, BMI, smoking pack-years, and alcoholic drinks per week.

3.1 School districts

There were four participating public school districts in southeast Michigan; three were low-income and one was high-income. The high-income school district (HISD) had a median household income that was almost double the Michigan state median ($50,800 from the 2012–2016 U.S. Census Bureau American Community Survey estimates), while the three low-income school districts (LISDs) had an average median household income of approximately $33,000 from 2012–2016 [49]. The average percentage of families with children in the school district living below the poverty line was 43% in the LISDs and less than 10% in the HISD, compared to 12% of families in Michigan overall from 2012–2016 [49]. School district rankings on student achievement and improvement values were also consistently lower in LISDs (12th percentile) compared to the HISD (∼85th percentile) [50].

3.2 Demographics

Fifty-nine (n = 59) K-12 teachers were recruited from 26 schools at four public school districts in southeast Michigan (Table 1). There were 39 participating teachers from 15 schools in the LISDs and 20 participating teachers from 11 schools in the HISD. The majority of the participating school teachers taught at elementary schools (n = 28) and high schools (n = 21), although some taught at middle (n = 6) and alternative schools (n = 4; schools that generally serve K-12 grades for students that are at risk of not graduating or have specific needs) [51]. The mean interview date was 31 days later in LISD than HISD schools.

The majority of teachers were female (86%), white (90%), Christian (67%), married (69%), and held a graduate degree (88%). LISD teachers were more likely to be divorced and have lower household incomes (< $60,000), and were significantly older (44.3 years) than HISD teachers (38.9 years). Fifty percent of HISD and 39% of LISD teachers held a second job during the school year, and worked an average of 8.4 and 9.7 additional hours each week, respectively.

Teachers from the HISD reported significantly better overall self-reported health, received more sleep, and had lower body mass index (BMI) and afternoon cortisol levels compared to LISD teachers (Table 2). Systolic and diastolic BP levels and hypertensive status (51%) were not significantly different among participants, although all of the participants currently taking anti-hypertensive medications were LISD teachers (n = 8).

Ever-smoking status was higher among LISD teachers than HISD teachers, driven primarily by past-smokers (31% and 5%, respectively). The mean cigarette pack-years, however, was low among all teachers (2.4). The prevalence of alcohol consumption (69%) and the number of drinks per week (0.8) was lower for LISD teachers compared to HISD teachers (85% and 1.4 drinks per week). The prevalence of respiratory infections in the past month was similar in LISD and HISD (∼30%) (data not shown). Asthma rates were not significantly different between LISD and HISD teachers, although the prevalence of ever experiencing asthma symptoms and diagnosed asthma were higher in HISD teachers (data not shown). BMI was positively correlated to SBP and DBP for all teachers (Pearson coefficient = 0.500 and 0.493, respectively, p < 0.001), and cigarette pack-years for all teachers (Pearson: 0.432, p = 0.001) and LISD teachers (Pearson: 0.392, p = 0.016).

Salivary cortisol samples, on average, were collected by teachers at the LISDs and HISD at a similar time in the evening (7 minutes later in LISDs), but 21 minutes earlier among LISD teachers in the early afternoon (Table 2). Average salivary cortisol levels from the early afternoon were significantly higher among LISD teachers (4.33±2.44 nmol/L) than HISD teachers (3.21±1.70 nmol/L). Neither the evening salivary cortisol levels nor the change in afternoon to evening (A-E) cortisol levels were significantly different by district type. There was a decrease in cortisol levels from the early afternoon to evening samples among all teachers. There was no significant difference in cortisol levels in teachers with second jobs than those without a second job. Afternoon and evening cortisol levels were not significantly correlated to age, cigarette pack-years, BMI, BP, or hours of sleep per night for LISD or HISD teachers. Alcoholic drinks per week were negatively correlated with evening cortisol levels for all participants (Spearman’s rho: –0.291, p = 0.043).

The occupational stressors that at least 60% of teachers expressed being strongly concerned or concerned about (“agree”/“strongly agree” and “disagree”/“strongly disagree”, as appropriate) included: unsuitable classroom temperature (92%), lack of consultation for significant changes (83%), lack of school support staff (81%), disruptive students (80%), lack of appropriate training and support for required changes (78%), lack of time to complete tasks (76%), uncertain job security and school funding (71%), lack of value for work done at home (68%), over-burdensome lesson planning requirements (66%), inadequate classroom maintenance (63%), unreasonable deadlines and time pressures (61%), too many students per class (61%), lack of work-life balance (61%), and the inability to take a proper break during the school day (60%). The occupational stressors that 30% or less of teachers expressed being strongly concerned or concerned about included: poor relationships with colleagues (3%), poor relationships with a supervisor (10%), unclear work expectations (12%), violence from aggressive parents (15%), lack of opportunities to share ideas and perspectives (19%), lack of encouragement to use skills and initiatives in their work (20%), unsupportive supervisors (24%), and unacceptable total working hours (27%). A full list of teaching stressor questions and responses are in listed in the Appendix.

The ten occupational stressors that LISD and HISD teachers rated most differently are presented in Table 3 (additional data are in Table A1). Concerns about job security and funding, disruptive and violent students, aggressive parents, acceptability of physical working conditions, and effective leadership in the school were all significantly different between the two district types. LISD teachers “agreed” or “strongly agreed” more frequently that they deal with disruptive (95% vs. 50% in HISD) and violent students (67% vs. 5% in HISD), and are concerned about violence from aggressive parents (23% vs. 0% in HISD) and job security (87% vs. 40% in HISD). LISD and HISD teachers largely “agreed” and “strongly agreed” that their classrooms were often too cold or warm (95% and 85%, respectively).

The overall standardized composite stressor score and four of the six stressor categories (demands, relationships, role, and environment) showed significant differences between LISD and HISD teachers (Table 4). The overall standardized composite stressor score and individual stressor categories were all significantly correlated with each other (p < 0.005) for all participants, except for demands and control composite scores with the environment composite score.

School characteristics varied for the 15 LISD schools and 11 HISD schools (Table 5). Schools were built significantly more recently in the HISD (1978) than in the LISDs (1953). All HISD schools were observed as having a well maintained exterior, while only 67% of LISD schools had a similar designation. The mean number of roadways within 1 km of each school with greater than 5,000 vehicles per day AADT was similar at LISDs (3.1) and at HISD (2.9). The mean number of high AADT volume roads (> 10,000 AADT) within 1 km of the school were more common in HISD schools than LISD schools (2.6 vs. 2.1, respectively) (data not shown). There were more schools within 1 km of Toxic Release Inventory-reporting facilities at LISD schools (27%) than at the HISD schools (9%), although this was not a significant difference. The number of environmentally-contaminated sites as reported by the MDEQ was significantly higher (p = 0.013) within 1 km of LISD schools (3.8) than HISD schools (0.8). There was no significant difference between outdoor sound measurements at each school (LAeq 55.6±5.5 at LISDs and LAeq 53.4±4.7 at HISD).

Most classrooms had windows or doors that could be opened in both school district types (83%). Mean classroom temperatures were significantly higher in LISD classrooms (72.8±4.1 °F) compared to the HISD (69.4±3.6 °F; p = 0.003). Relative humidity was only slightly higher in LISD classrooms (41% vs. 37% in HISD). Classroom flooring, ceiling, and wall maintenance quality were lower in LISDclassrooms.

Bivariate analyses of standardized composite scores for teaching stressors overall and by categories were performed with health, occupational, and environmental measures. The teaching stressors standardized composite score for demands was negatively correlated with average hours of sleep per night for all teachers (Pearson coefficient: –0.332, p = 0.013) and HISD teachers (–0.467, p = 0.044), and the year a school was built for all teachers (–0.342, p = 0.019). The year a school was built was negatively correlated with the standardized composite score for all teaching stressors (–0.386, p = 0.007) for all teachers. The teaching standardized composite score for relationships was positively correlated to BMI for all teachers (0.266, p = 0.046) and negatively correlated to hours spent working in a second job for LISD teachers (–0.668, p = 0.012). Cortisol levels and BP were not significantly correlated with teaching stressors, for all participants or by HISD or LISD separately.

Bivariate analyses were also performed for environmental measures with health, occupational, and other environmental measures. The number of environmentally-contaminated sites within 1 km of the school for all teachers was positively correlated with heart rate (Pearson coefficient = 0.316, p = 0.015), the number of moderate and high volume AADT roadways within 1 km (0.392, p < 0.002), and classroom temperature (0.286, p = 0.028). Among LISD teachers, the year a school was built was negatively correlated with average afternoon cortisol levels (Spearman’s rho: –0.328, p = 0.043), years in the position (Pearson coefficient: –0.419, p = 0.030), average hours of sleep per night (–0.449, p = 0.021) and positively correlated to the number of moderate and high volume AADT roadways within 1 km (0.879, p < 0.001) and the number of environmentally-contaminated sites within 1 km (0.538, p = 0.003).

In linear regressions of SBP and DBP for all teachers adjusting for the school district type, standardized teaching stressor score, environmentally contaminated sites near the school, moderate and high vehicle AADT volume, age, sex, and cigarette pack-years, BMI was significantly associated with increased SBP (p < 0.001) and DBP (p < 0.001; adjusted for BP medication usage). The number of alcoholic drinks per week was associated with an increase in DBP (p = 0.049) but not SBP (medication usage-adjusted) when adjusting for the same variables. Other demographic, health, stressor, and environmental quality variables were not significantly associated with SBP or DBP.

In linear regressions of cortisol levels for all teachers adjusting for the school district type, standardized teaching stressor score, environmentally contaminated sites near the school, age, sex, and cigarette pack-years, moderate and high vehicle AADT volume was significantly associated with a decrease in afternoon cortisol levels (p = 0.025). Age and the number of alcoholic drinks per week were significantly associated with a decrease in evening cortisol levels when adjusting for the same variables (p = 0.043 and 0.038, respectively). Other demographic, health, stressor, and environmental quality variables were not significantly associated with cortisol levels. Overall standardized teach stressor scores were only significantly associated with the district income level (p = 0.013) in models adjusted for the same demographic, health, and environmental quality variables.

4.1 Limitations

As a pilot study with a relatively low sample size, these results are limited and a larger sample size is needed to elucidate further relationships. Robust conclusions from linear regressions to assess associations while adjusting for covariates are limited given the small sample size. Owing to the use of convenience sampling, there is a possibility of selection bias, as teachers that are under significant stress or with significant health problems may have disproportionately participated in the study. At the recruitment event or through other contact with teachers before participation, all teachers were encouraged to participate regardless of any stated stress level or health concern. We relied on self-reported data for the majority of our measures, which could have influenced participants’ responses (e.g., smoking or drinking history, weight, income). The survey was completed by participants directly, to afford additional privacy and reduce pressures to modify answers to social norms.

Participants provided information on demographics, behavior, occupation, and perceived stress from occupational factors. It is noteworthy, however, that there are a multitude of non-occupational factors that may influence stress that were not assessed, including other physical health concerns, mental health, relationships, and financial concerns. Household income, the number of children living at home, and hours of sleep per night were assessed to have a rudimentary reflection of some of these factors. The impact of these other stressors is unknown and may have a significant role in stress response for participants.

For most participants, we have six salivary cortisol measurements, which aids our ability to capture a more accurate cortisol reading. Participants were advised to collect saliva samples shortly prior to lunch and before going to bed, however, there was some variability in timing between each of the three samples at each time point, as sleeping times varied and some samples during the day were delayed due to unforeseen demands at work. Additionally, collecting cortisol samples in the morning would have improved our ability to interpret the data relative to other research. Another limitation includes the risk for improper saliva collection methods. If the SOS was not fully saturated, cortisol measurements could have been skewed. Despite these limitations, to our knowledge, this is one of the only studies of its kind to assess perceived stress related to teaching, blood pressure, and salivary cortisol with public school teachers by income level in the U.S.