
Study design and study population
In this time-controlled intervention study, a historical control group was used to evaluate the effects of the interventions. Therefore, two fire recruit populations were enrolled; (a) the control group comprised of two classes of recruits going through the academy training with existing, pre-pandemic training curricula and (b) the intervention group consisted of two classes of recruits enrolled during the pandemic and also receiving an HLS intervention.
The control group was recruited in early 2019 from two fire academies (academy A and B) in the New England area19,29. Both academies provide a 15–16 week training program that meets National Fire Protection Association (NFPA)’s standards, NFPA 1001: Standard for Fire Fighter Professional Qualifications. Except for minor differences between the two (for example, academy B requires recruits to stay overnight in the academy on training days while recruits at academy A go home every day after training, and academy B provides additional aquatic classes as part of their physical training), the recruits comprising the historical control group across academies were comparable according to our previous19 and current studies.
For the intervention group, we enrolled fire recruits from one fire academy in New England (academy B) and one in Florida (academy C) in late 2020. The training in academy C resembles that in academy B, regarding NFPA standards and overnight staying requirement, with similar training durations of 15 and 13 weeks for academies B and C, respectively.
All enrolled fire recruits who were older than 18 years old and provided informed consent were included. Those who did not consent to participate in the study or lacking essential demographic information (i.e., age and sex) were excluded. The current study is part of the “Fire Recruit Health Study” approved by the Institutional Review Board of Harvard T.H. Chan School of Public Health (IRB18-1902). We followed the Declaration of Helsinki throughout the study.
Selected health outcomes
The outcomes selected for study included body composition, blood pressure, physical fitness testing, mental health screens, and lifestyle behaviors. All academies, except academy C, had complete data collection, while academy C provided only subjective outcomes (i.e. questionnaire). The related measurements were described in our previous studies19,29 and are summarized below.
Recruits’ BMI and percent body fat were examined as body composition outcomes. A clinic stadiometer (Portable Stadiometer 213, SECA, Hamburg, Germany) and a Bioelectrical Impedance Analysis scale (BIA) (BC-418 Segmental Body Composition, Tanita, Tokyo, Japan or InBody 230, Seoul, South Korea) using athletic mode, operated by experienced physical trainers or medical personnel, were used to retrieve the parameters. The measurements were performed at entry to the academy, mid-training (i.e. 8th week for academy A and 7th week for academy B), and academy graduation. Body composition data were not available at academy C.
Blood pressures were measured using an automated and calibrated sphygmomanometer (10 series, Omron, Kyoto, Japan), following professional guidelines30. The measurements were done before recruits started their daily training or during rest break. Each recruit was asked to rest seated for at least 5 min before being measured in a sitting position. The automated sphygmomanometer would then take three readings, with 1-min interval between each, and record an average. Blood pressures measurements were conducted at entry to academy and graduation, and were not available at academy C.
Select physical fitness outcomes were push-ups, pull-ups, and 1.5-mile running time, with each measurement taken at entry to academy, mid-training, and at graduation. These are existing tests used by the academies to evaluate recruits’ physical performance over time. Push-ups were determined as the number a recruit performed continuously in one minute, without breaking the cadence. Pull-ups were counted as the number in single trial with good cadence and overhead grip. Running time for 1.5 miles was recorded in minutes. Physical fitness testing results were not available at academy C.
We used a questionnaire to examine participants’ mental health and lifestyle behaviors, administered at their entry to the academy and at graduation. The questionnaire was comprised of components derived from validated questionnaires, incorporating a modified version of Beck Depression Inventory for Primary Care (BDI-PC) (total scoring 0–18)31, Patient Health Questionnaire (PHQ-9) (total scoring 0–27)32, and a modified version of Posttraumatic Stress Disorder Checklist (PCL-5) (total scoring 0–76)33, with higher scores indicating worse mental health. As to lifestyle behaviors, the questionnaire contained items needed to calculate the MEDI-lifestyle score29, which is a 7-item healthy lifestyle score ranging from 0 to 7, embodying BMI, smoking history, dietary pattern (measured by the PREDIMED score, a 14-item Mediterranean Diet adherence screener34), physical activity35, sedentary behavior (measured by time spent watching television), daily sleep time, and afternoon naps. In particular, one point was given for each of the following: no smoking in the last 6 months, physical activity equivalent to greater than 16 METs-h/wk, PREDIMED score more than or equal to nine points, BMI less than or equal to 30 kg/m2, TV screen time less than 2 h/day, sleeping time between 7 and 8 h/day, and taking daytime naps; otherwise a value of 0 would be given to each item.
Interventions
Compared with the historical control group, the intervened classes underwent the following changes in the existing training materials implemented by the academies.
First, the academies adopted an HLS intervention throughout the 13- or 15-week training based on the firefighters’ Mediterranean pyramid24, which illustrates a healthy lifestyle combination of balanced nutrition, regular physical activity, restorative sleep, positive social and family connections with resiliency strategies, and the avoidance of tobacco and other toxic substances. Each participant was given (a) access to a web-based toolkit (https://www.hsph.harvard.edu/firefighters-study/feeding-americas-bravest/) that includes information and resources for “Survival Mediterranean Style”, (b) a half-hour talk on healthy lifestyle at the entry to the academy training, (c) a waterproof, plastic paper sheet illustrating the firefighters’ Mediterranean pyramid, (d) a refrigerator magnet with the Mediterranean pyramid on it, (e) weekly nutrition/lifestyle tips throughout the academy training, and (f) an introduction to meditation/breathing exercise apps (for example, the Calm app (San Francisco, USA)). Except for (e), all intervention materials were given at the beginning of the academy training. The participants were able to review the HLS contents via the provided measures throughout the training period. While the practice of the HLS is voluntary, sponsored olive oil was supplied to the central kitchen of academy B and consumed by the fire recruits when they stayed at the academy on weekdays, and academy C gave each recruit of the intervened class a WHOOP (Boston, USA) wearable device that tracked recruits’ fitness and physiological parameters. Notably, weekly homework such as practicing a healthy recipe was assigned along with the weekly tips to the recruits. With the collaboration with the academies, extra training credits were given as incentives if the recruits showed their adherence to the HLS outside regular training time.
Second, as the intervention classes were trained during the COVID-19 pandemic, some curriculum adjustments were made to conform to public health policies. These changes included face masking required at all times during the training, limited class size, and shifting large group activities (such as group running) to small group physical training to increase social distancing. Moreover, previously there was a weekly 1-h aquatic training in the swimming pool at academy B, but since the pool was closed, the aquatic classes were replaced by weekly 1-h joint mobility exercise, in which recruits conducted a whole-body slow paced, yoga-like stretching workout.
Statistical analysis
Baseline characteristics and select health outcomes were reported as mean ± standard deviation or median (Q1-Q3) for continuous variables after checking for normality, or number (%) for categorical variables, and compared between groups using the t-test or the Wilcoxon rank sum test, as appropriate, for continuous variables and the Pearson’s Chi-squared test with Yates’ continuity correction or the Fisher’s exact test, as appropriate, for categorical variables.
Furthermore, we computed the changes in select health outcomes over time during academy training by calculating the longitudinal difference “the measurement at graduation—the measurement at baseline”, and presenting them as mean ± standard deviation, after checking normality. The differences in temporal changes between the intervention group and the control group were compared using the t-test.
For multivariable adjustment, mixed effects models were built incorporating the interaction term “Intervention Group × Time” to examine whether the health changes over time during academy training differed between the two groups. Potential confounders based on our domain knowledge and the baseline characteristics comparisons were included into the models. These are age, sex, baseline percent body fat, baseline push-ups, and/or baseline BDI-PC scores.
Finally, we built multivariable adjusted linear models to regress the health changes on the change of MEDI-lifestyle score, in order to demonstrate the changes in health per unit changes of MEDI-lifestyle score. For these models, the health changes throughout training were defined as percent changes from baseline measurements, except for those variables with any values of zero at baseline (i.e. pull-ups, BDI-PC, PCL-5, and PHQ-9).
All P values reported are two-tailed and a P < 0.05 was considered statistically significant. We used the R software (version 3.6.3) to conduct the statistical analyses.
Sensitivity analysis
With regard to the differences in training across academies, we conducted further sensitivity analysis limiting to fire recruits at the academy with both historical control class and intervention class available, which is academy B. In fact, there was one more class at academy B that took place in early 2020, receiving the lifestyle intervention, but undergoing unexpected training interruption for 3 months due to the initial COVID-19 outbreak. By comparing the three classes at academy B (i.e. the historical control class, COVID-19 interrupted class, and the intervention class), we were able to examine the effects of the intervention as well as the impact of the 3-month training interruption on recruits’ health. Notably, only the body composition and fitness testing data are available for the COVID-19 interrupted class.
In addition, while there were differences in the intervention contents across the academies B and C, as described above, we further conducted secondary analyses to investigate if the health changes differed between the two populations (i.e., the fire recruits comprising the intervention group from academy B and C, respectively) throughout academy training. Since objective data were not available at academy C, only subjective measurements (i.e., behavioral and mental health outcomes) could be compared.
Ethics approval and consent to participate
The study is part of the “Fire Recruit Health Study”, which was approved by the Institutional Review Board of Harvard T.H. Chan School of Public Health (IRB18-1902), and we followed the Declaration of Helsinki throughout the study.