Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Sep 1.
Published in final edited form as: J Clin Psychol. 2021 Apr 26;77(9):2041–2056. doi: 10.1002/jclp.23139

The effect of time outdoors on veterans receiving treatment for PTSD

Joanna E Bettmann 1, Kort C Prince 1, Kamala Ganesh 1, Kelsi F Rugo 1, AnnaBelle O Bryan 1, Craig J Bryan 1, David C Rozek 1, Feea R Leifker 1
PMCID: PMC8405544  NIHMSID: NIHMS1702267  PMID: 33899932

Abstract

Objectives:

Duration, frequency, and intensity of nature exposure link to different physical and psychological benefits. The present study aimed to determine how time outdoors affected military veterans’ posttraumatic stress disorder (PTSD) symptomology during PTSD treatment. Method: Hypotheses regarding time outdoors and the effect of program duration on PTSD symptoms were examined using multilevel models. The authors hypothesized that hours outdoors, both within‐ and between‐persons, would predict reduced PTSD symptomology, program duration would predict reduced PTSD symptomology, and that hours outdoors and program duration would be significant when accounting for the other.

Results:

The present study found that time outdoors correlated with participants’ decreased PTSD symptomology: the more time participants spent outdoors, the greater the reduction in their PTSD symptoms.

Conclusion:

The effect of time outdoors was significant within‐person, not between persons, suggesting that nature exposure may be used as an adjunct to traditional mental health treatment where exposure or dosage should be person‐specific.

Keywords: mental health, nature, outdoor, PTSD, veteran

1 |. INTRODUCTION

Humans receive physical and psychological benefits from nature exposure which include decrease in blood pressure, increase in physical activity, improved recovery from stress and attention fatigue, and relief of depressive symptoms (Mayer et al., 2009; Shanahan et al., 2016). Nature exposure also links to neuroendocrine and affective restoration from stress (Van den Berg & Custers, 2011). While researchers have explored a range of hypotheses to explain these findings, Stress Reduction Theory offers one lens for understanding nature’s impact on humans. Stress Reduction Theory posits that, because the human brain evolved in natural environments, processing natural stimuli puts less strain on sensory resources than processing urban stimuli. Therefore, the greater processing demands of urban environments may impede stress recovery (Ulrich et al., 1991). For example, physiological measurements in one study indicated a parasympathetic response to viewing nature during stress recovery which was not present when participants viewed urban environments (Ulrich et al., 1991). The stress‐reducing power of nature may link to the amount of vegetation in the nature setting (Ulrich et al., 1991).

Stress Reduction Theory is particularly useful when thinking about posttraumatic stress disorder (PTSD) because PTSD symptoms involve activation of arousal and emotion networks in the brain, increasing demand on cognitive, and attentional control networks (Falconer et al., 2008). In one study, research participants with PTSD conducting tasks requiring sustained attention were more vulnerable to distractions than participants without PTSD (Aupperle et al., 2012). In another study, researchers found that participants with PTSD committed more inhibition‐related errors on a study task than participants without previous trauma exposure. Furthermore, the researchers found a significant positive correlation between severity of PTSD symptomology and number of inhibition‐related errors (Falconer et al., 2008). Stress Reduction Theory offers one lens for understanding these findings, suggesting why those with PTSD might recovery better in natural environments.

Attention Restoration Theory offers another lens for considering these findings. Attention Restoration Theory suggests that voluntary, effortful attention is susceptible to fatigue while nature settings provide opportunities for soft fascination, thereby restoring the capacity for effortful attention (Kaplan, 1995). Thus, the capacity of nature to restore attentional capacity following attention fatigue may improve executive functioning and inhibitory control capacities in individuals with PTSD (Mayer et al., 2009). Given these findings, natural environments may provide optimal treatment settings for veterans with PTSD.

1.2 |. Nature‐based interventions for veterans

Numerous programs offer nature‐based therapies for veterans with PTSD (Poulsen et al., 2015). One study of such a program found no significant reductions in PTSD symptoms following a 5‐day Outward Bound course, but veterans self‐reported significant qualitative improvements in quality of life and ability to enjoy life (Hyer et al., 1996). Another study found significant reductions in PTSD symptomatology among veterans who participated in a year‐long, weekly sailing intervention compared with a waitlist control group (Gelkopf et al., 2013). A review of nature‐assisted therapies for veterans with PTSD found that nature exposure may mitigate automatic physical reactions that accompany PTSD, allowing veterans to respond more effectively to previously distressing stimuli (Poulsen et al., 2015). This review suggested that participation in nature‐assisted therapies links to veterans’ improved sense of well‐being and ability to cope with PTSD symptoms (Poulsen et al., 2015).

Researchers studying the restorative effects of nature on veterans found significant improvement in veterans’ psychological well‐being, social functioning, and life outlook 1 week after group‐based wilderness experiences with the Sierra Club’s Military Families and Veterans Initiative (Duvall & Kaplan, 2014). Veteran participants showed significant improvement in attentional functioning, positive affect, and tranquility. Another study of military veterans found that physical activity in nature produced statistically significant increases in perceived quality of life, vigor, and perceived personal competence, while decreasing tension, depression, and anger (Lundberg et al., 2011). Together, these findings indicate a link between nature‐based activities and positive psychological effects on veteran participants.

1.3 |. Activity type and dosage

Research indicates that nature activity type may not matter in terms of producing psychological benefit. Both an extended wilderness backpacking trip and a nature walk produce increased positive affect and recovery from attention fatigue (Hartig et al., 1991). Researchers found one disparity in affective outcomes: many wilderness backpacking participants expressed mixed feelings after returning from the trip, characterized by rejuvenation mixed with negative thoughts related to returning to their usual settings and situations. However, after the initial period post‐return, participants in one study experienced a sustained period of greater happiness than nature walk participants. Participants in the nature walk study did not report similar lowered affect (Hartig et al., 1991). In another study examining the psychological effects of 10 different activities in nature, including walking, cycling, fishing, and horseback riding, participants showed improvements in self‐esteem and mood across all activities, regardless of duration or intensity of exercise in nature (Pretty et al., 2007).

However, duration, frequency, and intensity of nature exposure link to different physical and psychological benefits (Shanahan et al., 2016). For example, individuals who visited a park for longer lengths of time reported lower stress levels compared with those with shorter park visits (Payne et al., 1998). Results of another study showed that nature exposure which lasted 30 min or more produced improvements in participants’ depressive symptoms and blood pressure (Shanahan et al., 2016), indicating that nature exposure duration may require a minimum threshold level to produce significant health benefits. Another study found that only four minutes of passively viewing a nature scene was required for stress recovery approaching baseline physiological levels (Ulrich et al., 1991). A meta‐analysis concluded that exercise in natural environments showed significant effects on participants’ self‐esteem and mood, even after only short periods of exercise (Barton & Pretty, 2010). Other research suggests that even brief nature exposure is associated with improvement in attentional control in children and adults (Faber Taylor & Kuo, 2009; Ohly et al., 2016; Stevenson et al., 2018). Collectively, these findings indicate that even short‐term nature exposure may facilitate stress recovery and cognitive functioning (Ulrich et al., 1991).

Given the positive benefits of nature exposure, it holds promise as an adjunct to traditional mental health treatments. Is it possible that nature exposure could enhance any benefits gained during traditional mental health treatment? Thus, the present study aimed to answer the question: does time spent outdoors during the course of outpatient trauma‐focused therapy impact military veterans’ PTSD symptomology?

2 |. METHODS

The present study investigated two hypotheses: that time outdoors during a 2‐week treatment program, both between and within‐persons, would predict participants’ reduced PTSD symptomology and that program duration would predict participants’ reduced PTSD symptomology. The present study also investigated whether both time outdoors and program duration would be significant when accounting for each other.

2.1 |. Study procedures

Veterans and military personnel were recruited via online advertisements, flyers, and referrals from community agencies, nonprofits, and mental health professionals. Participants were enrolled in an observational treatment outcome study examining the effectiveness of Cognitive Processing Therapy (CPT; Resick et al., 2016), an empirically supported psychological treatment for PTSD (Watts et al., 2013), delivered on a daily basis rather than the typical once or twice per week format. Participants traveled to an adaptive recreational facility in Park City, UT, to receive 1 h of individual daily sessions of CPT in the mornings. They completed either the full 12‐session version of CPT (Resick et al., 2016) or a modified seven‐session version (Chard, personal communication).

All participants received one hour of CPT per day. CPT is a trauma‐focused therapy that focuses on the identification and restructuring of maladaptive trauma‐related beliefs and assumptions that contribute to and sustain PTSD symptomatology. Through CPT, participants are taught how to recognize and change these beliefs using a series of worksheets that are introduced and practiced during each session. Participants then complete multiple worksheets each day to practice these new skills and concepts. In addition to CPT, all participants attended a 1‐h class focused on improving sleep quality (based on cognitive behavioral therapy for insomnia) (Taylor & Pruiksma, 2014), a 1‐h class focused on relaxation and mindfulness skills training, and a 1‐h relapse prevention group.

In the afternoons, program staff offered one or more optional activities (e.g., hiking, cycling, and rock climbing) to participants. The present study includes data across eight cohorts of participants. Before each day’s morning therapy session, participants self‐reported their PTSD symptoms and mood, as well as which afternoon activities they participated in the previous day. For more detailed information on study procedures, see Bryan et al. (2018). Participants recorded in minutes the daily time they spent engaging in different activities. Researchers coded minutes spent engaging in outdoor activities (including mountain biking, hiking, walking outside, and outdoor team building activities) as time outdoors.

2.2 |. Sample

Of the 49 study participants, 37 were male (75.5%) and 12 were female (24.5%). Forty two participants identified as White, five identified as Hispanic/Latino, two as Black, and three as American Indian/Alaskan Native. Participants had the option of choosing more than one race/ethnicity. Forty‐six supplied their categorical age: 13 (26.5%) were 25–34, 21 (42.9%) were 35–44, eight (16.3%) were 45–54, and four (8.2%) were 55 or older. Eight (16.3%) identified as active duty military, eight (16.3%) as National Guard/Reserve, and 32 (65.3%) as military veterans. Of the 20 who supplied a response to whether they had seen people being killed or wounded, 15 (75%) indicated they had. Similarly, 15 of 20 people who provided responses indicated they had seen a dead body. All 20 of the people who provided responses indicated they were in danger of being killed during at least one deployment.

Participants were eligible to participate if they had ever served in the U.S. military, were at least 18 years old, and met full diagnostic criteria for PTSD, as assessed by the Clinician Administered PTSD Scale for DSM‐5 (CAPS‐5: Weathers et al., 2013), a structured diagnostic interview with high reliability and validity. Participants were excluded if they had severe alcohol dependence that required medical monitoring, had made a suicide attempt during the preceding 3 months, or were unable to complete the informed consent process. Eligibility interviews with the CAPS‐5 were conducted by trained interviewers via phone and were audio‐recorded for review by trained interviewers and a licensed clinical psychologist to ensure reliability and consistency across interviewers. The researchers received approval from their university institutional review board before data collection. Informed consent was obtained from all participants before participation in the research.

2.3 |. Measures

The PTSD Checklist for Diagnostic and Statistical Manual of Mental Disorders – Fifth Edition (PCL‐5: Weathers et al., 2013) is a 20‐item self‐report Likert‐type questionnaire assessing an individual’s experience of PTSD symptoms in accordance with DSM‐V criteria on a scale of 0 (“Not at all”) to 4 (“Extremely”). Researchers administered the version of the PCL‐5 excluding DSM‐V Criterion A, with items assessing Criteria B through E. PCL item scores were summed to yield a total symptom severity score, which provides a continuous measure of PTSD symptom severity (Blevins et al., 2015). A study of the PCL‐5 in veterans indicated high internal consistency, test-retest reliability, and convergent and discriminant validity (Bovin et al., 2016).

The Visual Analog Mood Scale (VAMS) includes nine mood scales using horizontal lines and simple schematic faces and/or words displaying the following mood states: depressed, calm, agitated, hopeful, desire to kill self, anxious, ashamed, happy, and tired. The construct of “desire to kill self” did not include schematic faces, but rather was anchored on the left side with the word “none” and on the right with the word “extreme” (Bryan, 2019). VAMS responses are scored by dividing the number the participant reports by the total length of the line, and then multiplying this by 100 to get the final score. Research indicates Visual Analog Mood Scales have excellent convergent and discriminant validity, and high content validity (Stern et al., 1997). For the present study, the researchers coded the VAMS dichotomously in terms of participants’ positive emotions (hopeful, calm, and happy) or negative emotions (depressed, agitated, suicidal, anxious, ashamed, and tired).

2.4 |. Analyses

The study question regarding time outdoors and the effect of program duration on PCL scores was examined using multilevel models (Raudenbush & Bryk, 2002). To test the effect of time outdoors (hours) on PCL scores, the time outdoors variable was decomposed into a within‐ and between‐persons component (Curran & Bauer, 2011; Raudenbush & Bryk, 2002; Wang & Maxwell, 2015). To properly disaggregate the variable, time outdoors was centered within‐person to create a level‐1 variable representing person‐specific variation in time outdoors by day. Grand mean centered person means were then entered as a level‐2 predictor; these person means represent between‐person variance in the average amount of time spent outdoors across people (Raudenbush & Bryk, 2002).

Once time outdoors was properly decomposed into two orthogonal variables, a contextual effect was examined. A contextual effect is present when the within‐person (i.e., person‐level) relationship between a correlate and the outcome is different from the between‐person relationship for that same variable. In this context, a contextual effect hypothesized a level‐1 association between time outdoors and PCL scores wherein, on days where people spend more time outdoors, their PCL scores were predicted to be lower. The level‐2 association hypothesized that, among those who spend more time outdoors on average, PCL scores were expected to be lower relative to peers who spent less time outdoors on average, across all days.

The researchers also hypothesized an additive effect for program duration. The researchers predicted that, as one spent more days in the program, PCL scores would decline after accounting for time outdoors owing to the program’s cognitive processing therapeutic approach. The researchers predicted that time outdoors and program duration would have complimentary effects on PCL. Though this model is identical to a de‐trended model (Wang & Maxwell, 2015), the researchers’ intent here was different. The authors hypothesized that hours outdoors would predict reduced PCL scores, program duration would predict reduced PCL scores, and that both hours outdoors and program duration would be significant when accounting for the other. Accordingly, models (a through d) were built in the following sequence: (a) a random intercept only (null) model to examine the amount of variability in PCL scores within and between persons, (b) a model that added the level‐1 variable for hours outdoors (centered within‐person), (c) a model that added the level‐2 variable for hours outdoors (grand mean centered person means) to examine a contextual effect, and (d) a model that added the level‐1 variable for program duration (days, centered within‐person).

Note that the level‐1 and level‐2 variables representing within‐ and between‐persons time outdoors are orthogonal: the presence of one in the model had no impact on the coefficient of the other variable. The model fit, however, was expected to differ if the additional predictor improved predictive accuracy. Because the level‐1 association was uncorrelated with the level‐2 association, the level‐2 coefficient was not interpreted directly. Instead, a custom hypothesis test was used below. This approach tested whether the level‐1 coefficient differed from the level‐2 coefficient (Curran & Bauer, 2011; Raudenbush & Bryk, 2002; Wang & Maxwell, 2015).

2.5 |. Time lag

Because PCL scores (and other assessments) occurred each morning of the program, the modeling process utilized a time lag. Specifically, time was lagged such that time outdoors on one day predicted PCL score the next day. Without the lag, PCL scores would reflect PTSD symptomology before outdoor time occurred, which would negate a possible temporal precedence. This lag reduced the number of assessments by one; assessments on the first day of the program were dropped as a result of the lag. Removing the first day of the program from consideration, 49 participants contributed 446 assessments (one per day) with an average program duration of 9.3 days, a minimum of 5 days, and a maximum of 11 days.

2.6 |. Multiple imputation and analytic approach

An assessment of time outdoors was missing for 82 of the 446 assessments (18.4%). To address the missing data on hours outdoors, Bayesian imputation was used. Unlike typical multiple imputation, which creates a fixed number of complete data sets (van Buuren, 2018), Bayesian imputation creates an entire distribution of plausible values for the missing data (McElreath, 2015) based on the predictors of the missing values.

Imputation proceeded under the assumption that the data were missing at random (MAR: Rubin & Little, 2002; Rubin, 1976). Because the MAR assumption cannot be tested statistically, the analysis proceeds based on theoretical reasoning (Rubin & Little, 2002). The MAR assumption dictates that observed variables in the data can explain the systematically missing values (van Buuren, 2018). Under the empirically demonstrated assumption that mood influences how much time one spends outdoors (Christensen et al., 2013), the aforementioned VAMS’ scores (positive and negative) were used in the present study only as predictors of the missing values for hours outdoors. Day in the program was also used as a predictor of missing values for time outdoors. As mentioned above, the decision to include day in the program as a predictor of time outdoors was observationally based. As individuals become more familiar with the program and its available outdoor resources (e.g., local trails and access to bikes), the opportunity to spend time outdoors across a diverse range of activities increases. Finally, the person’s own random intercept was used as a predictor under the assumption that imputation of hours outdoors for each person, i, should consider his or her own overall average hours outdoors in addition to the overall average of all cases.

2.7 |. Model and priors

The full model (Model d above) can be represented by the following series of equations (including the composite form):

L1: Yti=π0i+π1imiHoursti+π2iDayti+σti
 L2: π0i=α00+β01Hoursi+σ0i
Yti=α00+β10miHoursti+β20Dayti+β01Hoursi+σ0i+σti

In the equations, Yti represents the PCL score for individual i at time t, where time is indexed by day in the program. α00 represents the grand intercept, β10 and β20 represent the coefficients for the time‐varying predictors of hours (partially imputed) and day in the program at level‐1 (centered within person), and β01 represents the time‐invariant between‐persons predictor of hours at level‐2 (grand mean centered person means). Finally, σij represents the error variance at level‐1 and σ0i represents variation in the random intercepts across people, i. Note that the model contains the predictor for which values have been imputed, miHoursti. The “mi” prefix denotes the multiple imputation.

In a Bayesian analysis, one needs to specify the beliefs about parameters, and their uncertainty, before seeing the data; these beliefs are called priors and they are updated in combination with the observed data to form a posterior distribution (McElreath, 2015). A detailed explanation of the selection of priors for the likelihood and each model parameter is available in the supplementary material. For model parameters, the researchers made use of regularizing priors. This type of prior argues, a priori, that the most likely effect of the predictor is no effect all, but regularizing priors can be easily overwhelmed by the data if, for example, the data suggest a larger effect; these priors also produce better out‐of‐sample predictions (McElreath, 2015).

2.8 |. Software and model comparison

Analyses were conducted using R 3.6.1 (R Core Team, 2019). Prospective power analysis was conducted using simulation-based methods and the r package “simr” (Green & MacLeod, 2016). Researchers conducted the power analysis to identify the sample needed to detect a small effect for the level‐1 variables and a medium‐sized effect for the contextual test at level‐2. At an N of 45 (but with multiple timepoints within‐persons, which increases power), the two level‐1 effects were found to have power point estimates of 1.00 (confidence interval [CI] = 0.99–1.00) for hours outside and 1.00 (CI = 0.99–1.00) for day in the program. The analysis indicated a medium‐sized contextual effect could be detected with 0.82 power (CI = 0.80–0.84) with an N of 45.

Bayesian multilevel models and Bayesian imputation were performed using the “brms” package (Bürkner, 2018). Each model presented below was run with four chains, each with 5000 iterations and 1000 warmup iterations, yielding 16,000 total post‐warmup iterations (greater than 10,000 is recommended to compute 95% highest density intervals [HDIs]; Kruschke, 2015). Examination of trace plots for chain convergence and auto‐correlation plots were performed using the “bayesplot” package (Gabry et al., 2019).

Bayesian statistics do not typically consider p values; however, HDIs can be used to understand whether no effect is likely given a posterior distribution. HDIs that do not overlap zero indicate that zero is not in the credible range given the selected probability. Because some readers are likely more familiar with a null hypothesis significance testing approach, the researchers elected to provide a statistic—probability of direction (PD)—that is sometimes used in Bayesian hypothesis testing (Makowski et al., 2019). The PD is directly comparable with the frequentist p value, but is robust against the distributional form of the outcome and predictors (Makowski et al., 2019). PDP values are only available for the fixed effects; one can use the HDIs to assess “significance” of random effects.

Model comparison was performed using Pareto‐smoothed importance sampling leave one out cross‐validation (LOOCV); Vehtari et al., 2016). Relative model fit is compared using both model weights and the difference between models in the expected log predicted density (“elpd_diff”). Finally, for each model, the researchers also provide the conditional LOO‐adjusted Bayes’ R2, a measure of the proportion of variance that would be explained using new data (Gelman et al., 2019). The LOO adjusted Bayes’ R2 is comparable to an adjusted R2 from frequentist statistics.

3 |. RESULTS

A three‐level model with time nested within people nested within cohort and a model with cohort as a factor revealed no significant variation at the cohort level, indicating the cohorts were not different in terms of the relationship between predictors and PCL scores. Accordingly, cohort as a random factor was dropped from the models. Analyses of differences as a function of cohort length revealed no significant difference on the PCL and so the effect of cohort length is not discussed further. Hours outdoors per day varied considerably, with a range of 0–9, but a relatively low mean number of hours outdoors per day (1.7). As seen in Table 1, mean PCL scores were higher on day one relative to other days.

TABLE 1.

Descriptive statistics for the sample (N = 49)

Variable Min. Max. Mean SD
Days in program 5 11 9.3 2.2
Hours outdoors per day 0 9 1.7 1.4
PCL score (baseline, Day 1) 0 75 47.1 20.7
PCL score (all other) 0 80 35.8 19.6
Open in a new tab

Note: Values for hours outdoors per day are calculated before imputation and, therefore, have removed missing values.

Abbreviations: PCL, posttraumatic stress disorder checklist; SD, standard deviation.

3.1 |. Modeling outcomes

Outcomes from the modeling process are provided in Table 2. For each model, the point estimate (Est.) is provided as the median. The deviation around the median (MAD) is also provided, as well as the HDIs and PDP‐values, when applicable. HDIs are available for random effects, but, owing to software limitations, PDP‐values are not. For fit statistics, only Bayes R2 has intervals around the point estimate, again owing to software limitations. “elpd difference” refers to the difference between the best fitting model, Model 4, and the other models in their elpd values. Here, larger negative values indicate greater departure from the best‐fitting model. The “Final weight” row provides model weights.

The first model, Model 1, is a null, or intercept only model, intended to assess the amount of variance that exists within‐ and between‐participants on the PCL outcome. The intraclass correlation coefficient (ICC) indicated substantial variance existed between participants. The ICC of 0.74 can also be interpreted as indicating the average correlation between any two assessments of the PCL within a single person is 0.74.

Model 2 added the centered within‐person level‐1 predictor of hours outdoors, Hoursti. The coefficient of −1.7 indicates that, for every one‐unit change in hours outdoors within‐persons, PCL scores would be expected to drop by 1.7. Other b‐coefficients can be interpreted similarly. The HDIs for the coefficient (−2.5 to −0.9) do not contain zero, suggesting that no effect is not among the most credible values given a 95% HDI. Similarly, the PDP value of 0.000 suggests the effect is significant in traditional terms.

In Model 3, the test of the contextual effect is not provided in the table. Because a test of a contextual effect is not determined by whether a coefficient differs from zero, a custom hypothesis test is needed to ascertain whether the within‐person coefficient, Hoursti, is different from the between‐persons coefficient, Hoursi. This is achieved by testing whether, from the coefficients above, Hoursti − Hoursi = 0 is plausible, indicating the coefficients do not differ from one another. The point estimate of the difference between the coefficients is 2.1 (or −1.7 to 0.4; effect not tabled). Because the HDIs from the test included 0 (−1.5 to 5.8), 0 is in the plausible range and a contextual effect is not supported by the data.

The final model, Model 4, added the person‐specific program duration variable, Dayti, to the model to examine whether the effect of program duration was significant after accounting for the person‐specific hours outdoors per day, Hoursti, and vice versa. This model revealed the lowest deviance (−1289.7) and also received the entirety of the model weights (given the candidate set). The effect of program duration was significant (PDP = 0.000). Also, once duration was added to the model, the effect of hours outdoors within‐person was reduced, but the variable remained significant (PDP = 0.011), suggesting that, while they share some variance in the outcome, both variables are important predictors of reductions in PCL scores.

3.2 |. Prior sensitivity and assumptions

In a Bayesian analysis, it is important to check the sensitivity of the posterior to different prior specifications (Kruschke, 2015). The researchers analyzed the final model using both software‐default and uninformative priors and, as one might expect given the reasonable sample size, the posterior distributions changed little under different priors. Importantly, the inferences derived from the model parameters did not change.

Another issue to consider is the sensitivity of the model regarding coefficients estimated with and without use of Bayesian imputation. The coefficients changed little in a model that used a complete cases approach and, notably, inferences from the models remained the same under imputation relative to a complete cases approach. Despite a lack of any inferential differences, the tabled coefficients and PDP values obtained under imputation would be considered more accurate under the assumption that data were missing at random (McElreath, 2015; van Buuren, 2018).

Assumptions regarding normality of the residuals were examined using a QQ‐plot. Only mild deviation from normality was observed for the level‐1 residuals, and the level‐2 best linear unbiased predictors (BLUPs) conformed to normality well. Posterior predictive checks indicated samples from the posterior converged well with the observed data. To examine whether the relationships were adequately linear, overall and individual trajectories were plotted using loess smoothers; no evidence of nonlinearity was observed. The issue was further investigated by a plot of the fitted values against the residuals, again using a loess smoother. No notable deviations from linearity were found.

The Rhat statistic is a measure of the convergence of the Markov chains (Gelman & Rubin, 1992). Generally, convergence of the chains is deemed to have occurred if all Rhat values are less than 1.1 (though lower is better). The largest Rhat value was 1.003, well below the critical criterion. Trace plots were also examined for each model parameter and revealed no evidence of lack of convergence.

Pareto‐k values, which are similar to Cook’s distance in frequentist models in that they identify cases that have an inordinate impact on the posterior fit and are outliers, were examined for each case. Pareto‐k values are classified as: 0.0–0.5 (good), 0.5–0.7 (ok), 0.7–1.0 (bad), and 1.0 to infinity (very bad). Values greater than 0.7 may indicate misspecification of the model (Gabry et al., 2019). For the final model, 443 cases had Pareto‐k values less than 0.5, three had a value between 0.5 and 0.7, and no cases had values greater than the critical value of 0.7.

4 |. DISCUSSION

The present study investigated the hypothesis that time outdoors, both between and within‐persons, would predict reduced PTSD symptomology and that program duration would predict reduced PTSD symptomology. The present study also investigated whether both time outdoors and program duration would be significant when accounting for each other. Previous studies have not investigated the effects of duration of nature exposure on PTSD symptomology. Instead, previous research has primarily focused on nature exposure and its effects on stress and mental health domains, such as depressive symptoms (Shanahan et al., 2016), mood (Van den Berg & Custers, 2011), psychological well‐being, social functioning, and life outlook (Duvall & Kaplan, 2014), and reduced stigma in psychological help‐seeking (Harper et al., 2014).

A small number of studies have explored outdoor recreation’s effect on veterans’ PTSD symptomatology (Gelkopf et al., 2013; Poulsen et al., 2015; Vella et al., 2013), but these studies have not examined specifically how the duration of nature exposure affects PTSD symptomology. Therefore, the present study is unique in examining links between duration of nature exposure and PTSD symptomatology. Findings from the present study may be understood in the context of Stress Reduction Theory. This theory posits that the processing demands of urban environments may impede stress recovery (Ulrich et al., 1991). Thus, veterans with PTSD may receive particular benefits from outdoor activity as an adjunct to the treatment given the stressors of PTSD symptomology, as well as the stressors associated with its treatment. In the present study, veterans may have experienced afternoons of outdoor activity as particularly restorative and supportive following their mornings of intensive PTSD treatment.

Attention Restoration Theory offers a different lens for understanding the present study’s findings. Attention Restoration Theory suggests that effortful attention is susceptible to fatigue while nature settings provide restoration which improves humans’ capacity for effortful attention (Kaplan, 1995). Using this theoretical lens, veterans with PTSD who spent more time outdoors during the 2‐week treatment period in the present study may have experienced stronger executive functioning and inhibitory control capacity due to nature’s capacity to restore attentional abilities. Such nature exposure may optimize veterans’ PTSD treatment by providing restorative opportunities, thereby improving their psychological capacities for intensive treatment.

The present study also found that, within‐persons, time outdoors correlated with participants’ decreased PTSD symptomology: the more time participants spent outdoors on a given day, the greater the reduction in their PTSD symptomology the subsequent day. The fact that the effect of time outdoors was significant within‐persons, not between persons, suggests that an individual participant who spent more time outdoors overall (i.e., across days) did not experience a greater reduction in PTSD symptomology than another participant who spent less time outdoors on average. However, within each individual participant, more time spent outdoors on a given day linked with reduced PTSD symptomology. This finding indicates that the dosage of time outdoors necessary for improving PTSD symptomology is person‐specific. For example, one individual may see reduced PTSD symptomology from spending 2 h outdoors daily, while another may see equal benefit from spending only one half‐hour outside. Furthermore, the findings indicate that both types of individuals may see greater symptom reduction by increasing their person‐specific amounts of time outdoors. Therefore, practitioners should gauge the effect of nature exposure on each individual client before recommending particular nature dosages as mental health interventions.

Furthermore, the study found that program duration correlated with participants’ decreased PTSD symptomatology: the more days a participant spent in the program, the greater the reduction in their PTSD symptomology even after accounting for person‐specific variability in time spent outdoors each day. These findings indicate that outdoor interventions could be useful as adjuncts to treatment for specific veterans with PTSD. Spending more time outdoors while receiving therapy appears correlated with more rapid reduction in symptoms for some, but not all, veterans receiving treatment for PTSD. The present study’s findings corroborate previous research in this area indicating the mental health benefits of time outdoors (Mayer et al., 2009; Shanahan et al., 2016). Collectively, these findings support the viability of nature exposure as an adjunct to mental health treatment. However, future research must elucidate which specific individuals might benefit from nature as an adjunct to traditional mental health treatment, in what dosage, for what conditions, and for which severity of illness.

4.1 |. Implications for practice

Providers working with veterans who have PTSD should consider prescribing nature exposure as an adjunct to traditional mental health treatment because research, including the current study, indicates benefits of nature exposure for veterans. Nature exposure presents no significant side effects and offers a promising low‐stigma means for veterans to access psychological benefits including PTSD symptom reduction (Poulsen et al., 2015). In prescribing nature exposure as an adjunct to traditional health care, providers should consider an individual’s barriers to recreational nature access such as location, safety, and disability access (Hartig et al., 2014; Von Benzon, 2010; Weiss et al., 2011). These factors may limit an individual’s ability to access nature or their ability to access nature in a way that is helpful rather than stressful.

In addition, the present study suggests that the dosage of nature exposure necessary to impact mental health may vary person‐to‐person. Thus, practitioners should gauge the effect of nature exposure on each individual client, developing a specific plan for each client to meet the client’s specific needs. For example, a clinician might prescribe one client short doses of nature exposure (such as 20 min outdoor walks) on successive days and then ask the client to track affect and symptomology following each walk. If the short outdoor walks do not appear to affect mental health, then the clinician might consider increasing the dose of nature exposure to 30 or 40‐min outdoor walks, slowly titrating up to achieve individual benefits in terms of reductions in mental health symptomology. The clinician should also make sure to identify with each client the location for such walks: for example, the nearest, most convenient, and accessible nature area to the client’s home. This area could be a city park with trees and other foliage, a country road, a local arboretum, or comparable location.

There are numerous efficacious programs for veterans that utilize nature‐based interventions already. These programs include horticulture therapy (Detweiler et al., 2015; Lehmann et al., 2018), fishing (Bennett et al., 2017; Bennett et al., 2014), and sailing interventions (Gelkopf et al., 2013; Marchand et al., 2018). In one study, veterans participating in horticulture therapy showed lower cortisol levels and depressive symptoms compared with veterans participating in other occupational therapy activities such as ceramic painting, flower arranging, or leather belt construction (Detweiler et al., 2015). Notably, a systematic review of nature‐assisted therapies for veterans with PTSD found only positive effects of these therapies, noting some benefits as “[C]ommunity with others, the ability to plan and perform a task, as well as a way back to work” (Poulsen et al., 2015, p. 444).

Those designing or delivering treatment for veterans should consider how such outdoor interventions can be integrated into existing programming, both in outpatient and inpatient settings. In one study of horticulture therapy for veterans, the therapeutic intervention was provided in a courtyard at the VA (Detweiler et al., 2015), simplifying the complex logistics which can accompany group outdoor expeditions. Similarly, hour‐long outdoor walking interventions could be delivered easily to veterans with a range of ability levels and in a range of settings.

Several interesting questions are raised by the present study, but not yet answered. Does the specific activity that veterans participate in mediate the benefits they receive from outdoor activity? Only a few studies have explored this question (Fraser et al., 2019; Hartig et al., 1991; Pretty et al., 2007), but none addressed veterans specifically. For example, one study comparing walking versus golfing suggests that participants in the walking condition experienced greater benefits in terms of directed attention, proposing that benefits from outdoor activity may be greater when the activities themselves require less concentration or focus (Fraser et al., 2019). A meta‐analysis of a range of outdoor activities (including walking, cycling, gardening, farming, sailing, and horseback riding) found robust effect on participants’ mood and self‐esteem in a range of outdoor environments, activities, and activity intensity (Barton & Pretty, 2010). However, the study did not explore whether the type of outdoor activity (e.g., gardening vs. walking) modified the benefits participants received. Thus, existing research suggests that adding outdoor interventions to veterans’ existing mental health programming is likely to offer substantial benefits.

4.2 |. Limitations

The present study has a number of limitations. First, the study used a sample size of 49 participants, limiting the generalizability of results. Study sample ethnic diversity may also limit result generalizability. Of the 31 participants who provided their ethnicity, 90.3% participants identified as White, 12.9% identified as Hispanic/Latino, 6.5% identified as Black, and 6.5% as American Indian/Alaskan Native. By comparison, US active duty service members are 69% White, 17.7% Black, 4.5% Asian, and 1.1% American Indian/Alaskan Native or other Pacific Islander (US Department of Defense, 2018). Incongruence between the study sample’s ethnic breakdown and the ethnic breakdown of the US military limit generalizability of the study’s findings to the general military population. Furthermore, of the 15 respondents who provided their military branch of service, 33.3% served in the Army, 6.7% in the Navy, 13.3% in the Marine Corps, and 46.7% in the Air Force. By contrast, 36.2% of active duty personnel are in the Army, 24.9% are in the Navy, 24.7% are in the Air Force, and 14.2% are in the Marine Corps (US Department of Defense, 2018). Differences between the study’s military branch representation and the US military’s branch representation may further limit generalizability of study results.

4.3 |. Future research

Future research should examine for which populations nature exposure is particularly impactful when implemented adjunct to traditional mental health treatment. Given the small sample of the present study, the researchers were unable to identify specific characteristics of veterans who benefited most from time outdoors. For instance, factors such as prior positive or negative associations with physical exercise or outdoor activity could impact the degree to which an individual participates in or benefits psychologically from time outdoors. A meta-analysis found that participants’ gender and age modified the degree of benefit individuals derived from exercise in nature (Barton & Pretty, 2010), so future research should investigate whether individual factors like gender, age, race, or socioeconomic status modify the benefit derived from nature exposure.

In addition, future research should explore whether the observed effects of outdoor activities are due to the outdoor settings or to the activities themselves. Furthermore, only limited research has investigated whether participants’ experiences of outdoor activities are affected by whether they are alone or with others (Rogerson et al., 2020). One study compared participants running alone outdoors to participants running with a small group of peers outdoors (Rogerson et al., 2020). The authors concluded that being alone or with others appeared not to modify in any way the positive psychological outcomes of the outdoor exercise.

In addition, future research should explore the specific effects of PTSD treatment components. For example, the present study included, during the 2‐week CPT training period, a 1‐h class focused on improving sleep quality, a 1‐h class focused on relaxation and mindfulness skills training, and a 1‐h relapse prevention group. How might these specific interventions have modified the study findings? Future research should aim to explore the specific effects of these interventions on participant outcome so that PTSD treatment can be optimized most effectively for all participants.

The moderate sample size of the present study meant that some theoretically interesting variables were not considered in the present study. Future research should consider whether variables such as psychotropic medications, psychiatric or medical diagnoses (including comorbidity), or therapist effects augment, attenuate, or moderate the relationship between time outdoors and PTSD symptomology. Future research should also explore whether specific outdoor activities or different intensities of outdoor activities modify participants’ experiences: at present, research appears to address this question only partially (Barton & Pretty, 2010). It will also be important to explore the mechanisms underlying clients’ choices to participate in outdoor activities. For example, a client’s choice to participate in an outdoor activity might reflect reduced avoidance, an important outcome of the PTSD therapy in which participants were engaged. Might experiencing benefit from PTSD therapy lead to more engagement in activities which might lead to more rapid reduction in PTSD symptoms, which reinforces engagement in these activities? Future research should explore this important question. However, the present study clearly illuminates how time outdoors links to improved PTSD symptomology among military veterans receiving treatment.

Supplementary Material

supinfo

TABLE 2.

Parameter estimates for bayesian multilevel models (NID = 49, Obs = 446)

Model 1
 Model 2
 Model 3
 Model 4
Fixed effects Est. (mad) HDI95 PDP Est. (mad) HDI95 PDP Est. (mad) HDI95 PDP Est. (mad) HDI95 PDP
Intercept 36.0 (2.4) 31.2–40.8 0.000 35.7 (2.4) 30.8–40.5 0.000 35.9 (2.4) 30.9–40.5 0.000 35.8 (2.3) 31.1–40.4 0.000
 Hoursti −1.7 (0.4) −2.5 to −0.9 0.000 −1.7 (0.4) −2.5 to −0.9 0.000 −0.9 (0.3) −1.2 to −0.2 0.011
 Hoursi 0.4 (1.8) −3.1 to 4.0 0.813 0.4 (1.8) −3.0 to 4.0 0.805
 Dayti −1.9 (0.1) −2.2 to −1.7 0.000
Random effects Est. (mad) HDI95 Est. (mad) HDI95 Est. (mad) HDI95 Est. (mad) HDI95
 σ 17.0 (1.9) 13.8–21.1 17.0 (1.9) 13.8–21.0 17.1 (1.8) 14.0–21.0 17.1 (1.8) 14.0–21.1
 τ00 10.0 (0.4) 9.4–10.8 9.8 (0.4) 9.1–10.5 9.8 (0.4) 9.1–10.5 7.9 (0.3) 7.4–8.5
Fit statistics Est. (mad) CI95 Est. (mad) CI95 Est. (mad) CI95 Est. (mad) CI95
  Bayes R2 0.74 (0.01) 0.71 – 0.76 0.75 (0.01) 0.72 – 0.77 0.75 (0.01) 0.73 – 0.77 0.84 (0.01) 0.82 – 0.85
 elpd loo −1366.7 −1361.2 −1361.3 −1289.7
 elpd difference −77.1 −71.5 −71.6 0.0
 Final weight 0.00 0.00 0.00 1.00
Open in a new tab

Abbreviations: CI, confidence interval; HDI, highest density interval; mad, deviation around the median; PDP, probability of direction p value.

ACKNOWLEDGMENTS

Grant funding supported the data collection for this study. Grants which supported this study were from the Kendeda Fund and the Bob Wodruff Foundation. Additional funding was received from the University of Utah’s KL2 program and the National Center for Advancing Translational Sciences of the National Institutes of Health (Award No. UL1TR002538 and KL2TR002539).

Funding information

University of Utah’s KL2 program; Bob Woodruff Foundation; Kendeda Fund; National Center for Advancing Translational Sciences of the National Institutes of Health, Grant/Award Number: UL1TR002538 and KL2TR002539

Footnotes

CONFLICT OF INTERESTS

The authors declare that there are no conflicts of interest.

PEER REVIEW

The peer review history for this article is available at https://publons.com/publon/10.1002/jclp.23139.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the sixth author upon reasonable request.

SUPPORTING INFORMATION

Additional Supporting Information may be found online in the supporting information tab for this article.

REFERENCES

  1. Aupperle RL, Melrose AJ, Stein MB, & Paulus MP (2012). Executive function and PTSD: Disengaging from trauma. Neuropharmacology, 62, 686–694. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Barton J, & Pretty J. (2010). What is the best dose of nature and green exercise for improving mental health? A multi-study analysis. Environmental Science & Technology, 44, 3947–3955. [DOI] [PubMed] [Google Scholar]
  3. Bennett JL, Piatt JA, & Van Puymbroeck M. (2017). Outcomes of a therapeutic fly‐fishing program for veterans with combat‐related disabilities: A community‐based rehabilitation initiative. Community Mental Health Journal, 53(7), 756–765. [DOI] [PubMed] [Google Scholar]
  4. Bennett JL, Van Puymbroeck M, Piatt JA, & Rydell RJ (2014). Veterans’ perceptions of benefits and important program components of a therapeutic fly‐fishing program. Therapeutic Recreation Journal, 48(2), 169. [Google Scholar]
  5. Blevins CA, Weathers FW, Davis MT, Witte TK, & Domino JL (2015). The posttraumatic stress disorder checklist for DSM‐5 (PCL‐5): Development and initial psychometric evaluation. Journal of Traumatic Stress, 28, 489–498. 10.1002/jts.22059 [DOI] [PubMed] [Google Scholar]
  6. Bovin MJ, Marx BP, Weathers FW, Gallagher MW, Rodriguez P, Schnurr PP, & Keane TM (2016). Psychometric properties of the PTSD checklist for diagnostic and statistical manual of mental disorders–fifth edition (PCL‐5) in veterans. Psychological Assessment, 28, 1379. [DOI] [PubMed] [Google Scholar]
  7. Bryan CJ (2019). A preliminary validation study of two ultra‐brief measures of suicide risk: The suicide and perceived burdensomeness visual analog scales. Suicide and Life‐Threatening Behavior, 49, 343–352. [DOI] [PubMed] [Google Scholar]
  8. Bryan CJ, Leifker FR, Rozek DC, Bryan AO, Reynolds ML, Oakey DN, & Roberge E. (2018). Examining the effectiveness of an intensive, 2‐week treatment program for military personnel and veterans with PTSD: Results of a pilot, open‐label, prospective cohort trial. Journal of Clinical Psychology, 74(12), 2070–2081. 10.1002/jclp.22651 [DOI] [PubMed] [Google Scholar]
  9. Bürkner PC (2018). Advanced Bayesian multilevel modeling with the R Package brms. The R Journal, 10, 395–411. [Google Scholar]
  10. van Buuren S. (2018). Flexible imputation of missing data. CRC Press. [Google Scholar]
  11. Christensen KM, Holt JM, & Wilson JF (2013). The relationship between outdoor recreation and depression among older adults. World Leisure Journal, 55(1), 72–82. [Google Scholar]
  12. Curran PJ, & Bauer DJ (2011). The disaggregation of within‐person and between‐Person effects in longitudinal models of change. Annual Review of Psychology, 62, 583–619. 10.1146/annurev.psych.093008.100356 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. R Core Team. (2019). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. [Google Scholar]
  14. Detweiler MB, Self JA, Lane S, Spencer L, Lutgens B, Kim D‐Y, Halling MH, Rudder TC, & Lehmann LP (2015). Horticultural therapy: A pilot study on modulating cortisol levels and indices of substance craving, posttraumatic stress disorder, depression, and quality of life in veterans. Alternative Therapies in Health and Medicine, 21(4), 36–41. [PubMed] [Google Scholar]
  15. Duvall J, & Kaplan R. (2014). Enhancing the well‐being of veterans using extended group‐based nature recreation experiences. Journal of Rehabilitation Research and Development, 51, 685–696. [DOI] [PubMed] [Google Scholar]
  16. Faber Taylor A, & Kuo FE (2009). Children with attention deficits concentrate better after walk in the park. Journal of attention disorders, 12(5), 402–409. [DOI] [PubMed] [Google Scholar]
  17. Falconer EM, Felmingham KL, Allen A, Clark CR, McFarlane AC, Williams LM, & Bryant RA (2008). Developing an integrated brain, behavior and biological response profile in posttraumatic stress disorder (PTSD). Journal of Integrative Neuroscience, 7(3), 439–456. [DOI] [PubMed] [Google Scholar]
  18. Fraser M, Munoz SA, & MacRury S. (2019). Does the mode of exercise influence the benefits obtained by green exercise? International Journal of Environmental Research and Public Health, 16(16), 3004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Gabry J, Simpson D, Vehtari A, Betancourt M, & Gelman A. (2019). Visualization in Bayesian workflow. Journal of the Royal Statistical Society: Series A (Statistics in Society), 182, 389–402. 10.1111/rssa.12378 [DOI] [Google Scholar]
  20. Gelkopf M, Hasson‐Ohayon I, Bikman M, & Kravetz S. (2013). Nature adventure rehabilitation for combat‐related posttraumatic chronic stress disorder: A randomized control trial. Psychiatry Research, 209, 485–493. [DOI] [PubMed] [Google Scholar]
  21. Gelman A, & Rubin DB (1992). Inference from iterative simulation using multiple sequences. Statistical Science, 7, 457–472. 10.1214/ss/1177011136 [DOI] [Google Scholar]
  22. Gelman A, Goodrich B, Gabry J, & Vehtari A. (2019). R‐squared for Bayesian regression models. The American Statistician, 73, 307–309. 10.1080/00031305.2018.1549100 [DOI] [Google Scholar]
  23. Green P, & MacLeod CJ (2016). Simr: An R package for power analysis of generalised linear mixed models by simulation. Methods in Ecology and Evolution, 7(4), 493–498. 10.1111/2041-210X.12504 [DOI] [Google Scholar]
  24. Harper NJ, Norris J, & D’astous M. (2014). Veterans and the Outward Bound experience: An evaluation of impact and meaning. Ecopsychology, 6, 165–173. [Google Scholar]
  25. Hartig T, Mang M, & Evans GW (1991). Restorative effects of natural environment experience. Environment and Behavior, 23, 3–26. [Google Scholar]
  26. Hartig T, Mitchell R, De Vries S, & Frumkin H. (2014). Nature and health. Annual Review of Public Health, 35, 207–228. [DOI] [PubMed] [Google Scholar]
  27. Hyer L, Boyd S, Scurfield R, Smith D, & Burke J. (1996). Effects of outward bound experience as an adjunct to inpatient PTSD treatment of war veterans. Journal of Clinical Psychology, 52, 263–278. [DOI] [PubMed] [Google Scholar]
  28. Kaplan S. (1995). The restorative benefits of nature: Toward an integrative framework. Journal of Environmental Psychology, 15, 169–182. [Google Scholar]
  29. Kruschke JK (2015). Doing bayesian data analysis: A tutorial with R and Bugs. Elsevier Academic Press. [Google Scholar]
  30. Lehmann LP, Detweiler JG, & Detweiler MB (2018). Veterans in substance abuse treatment program self‐initiate box gardening as a stress reducing therapeutic modality. Complementary Therapies in Medicine, 36, 50–53. [DOI] [PubMed] [Google Scholar]
  31. Lundberg N, Bennett J, & Smith S. (2011). Outcomes of adaptive sports and recreation participation among veterans returning from combat with acquired disability. Therapeutic Recreation Journal, 45, 105–120. [Google Scholar]
  32. Makowski D, Ben‐Shachar M, & Lüdecke D. (2019). bayestestR: Describing effects and their uncertainty, existence and significance within the Bayesian framework. Journal of Open Source Software, 4, 1–8. 10.21105/joss.01541 [DOI] [Google Scholar]
  33. Marchand WR, Klinger W, Block K, VerMerris S, Herrmann TS, Johnson C, & Sheppard S. (2018). Safety and psychological impact of sailing adventure therapy among Veterans with substance use disorders. Complementary Therapies in Medicine, 40, 42–47. [DOI] [PubMed] [Google Scholar]
  34. Mayer FS, Frantz CM, Bruehlman‐Senecal E, & Dolliver K. (2009). Why is nature beneficial? The role of connectedness to nature. Environment and Behavior, 41, 607–643. [Google Scholar]
  35. McElreath R. (2015). Statistical rethinking: a Bayesian course with examples in R and Stan. Chapman and Hall/CRC. [Google Scholar]
  36. Ohly H, White MP, Wheeler BW, Bethel A, Ukoumunne OC, Nikolaou V, & Garside R. (2016). Attention Restoration Theory: A systematic review of the attention restoration potential of exposure to natural environments. Journal of Toxicology and Environmental Health, Part B, 19(7), 305–343. [DOI] [PubMed] [Google Scholar]
  37. Payne L, Orsega‐Smith B, Godbey G, & Roy M. (1998). Local parks and the health of older adults: Results of an exploratory study. Parks & Recreation, 33, 64–70. [Google Scholar]
  38. Poulsen DV, Stigsdotter UK, & Refshage AD (2015). Whatever happened to the soldiers? Nature‐assisted therapies for veterans diagnosed with post‐traumatic stress disorder: A literature review. Urban Forestry & Urban Greening, 14, 438–445. [Google Scholar]
  39. Pretty J, Peacock J, Hine R, Sellens M, South N, & Griffin M. (2007). Green exercise in the UK countryside: Effects on health and psychological well‐being, and implications for policy and planning. Journal of Environmental Planning and Management, 50, 211–231. [Google Scholar]
  40. Raudenbush SW, & Bryk AS (2002). Hierarchical linear models: Applications and data analysis methods. Sage Publications. [Google Scholar]
  41. Resick PA, Monson CM, & Chard KM (2016). Cognitive processing therapy for PTSD: A comprehensive manual. Guilford Publications. [Google Scholar]
  42. Rogerson M, Colbeck I, Bragg R, Dosumu A, & Griffin M. (2020). Affective outcomes of group versus lone green exercise participation. International Journal of Environmental Research and Public Health, 17, 624. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Rubin DB (1976). Inference and missing data. Biometrika, 63, 581–592. [Google Scholar]
  44. Rubin DB, & Little RJA (2002). Statistical analysis with missing data (2nd ed.). Wiley. [Google Scholar]
  45. Shanahan DF, Bush R, Gaston KJ, Lin BB, Dean J, Barber E, & Fuller RA (2016). Health benefits from nature experiences depend on dose. Scientific Reports, 6, 28551. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Stern RA, Arruda JE, Hooper CR, Wolfner GD, & Morey CE (1997). Visual analogue mood scales to measure internal mood state in neurologically impaired patients: Description and initial validity evidence. Aphasiology, 11, 59–71. [Google Scholar]
  47. Stevenson MP, Schilhab T, & Bentsen P. (2018). Attention Restoration Theory II: A systematic review to clarify attention processes affected by exposure to natural environments. Journal of Toxicology and Environmental Health, Part B, 21(4), 227–268. [DOI] [PubMed] [Google Scholar]
  48. Taylor DJ, & Pruiksma KE (2014). Cognitive and behavioural therapy for insomnia (CBT‐I) in psychiatric populations: A systematic review. International Review of Psychiatry, 26(2), 205–213. [DOI] [PubMed] [Google Scholar]
  49. Ulrich RS, Simons RF, Losito BD, Fiorito E, Miles MA, & Zelson M. (1991). Stress recovery during exposure to natural and urban environments. Journal of Environmental Psychology, 11, 201–230. [Google Scholar]
  50. US Department of Defense. (2018). 2018 Demographics profile of the military community. Retrieved from https://download.militaryonesource.mil/12038/MOS/Reports/2018-demographics-report.pdf
  51. Van Den Berg AE, & Custers MH (2011). Gardening promotes neuroendocrine and affective restoration from stress. Journal of Health Psychology, 16, 3–11. [DOI] [PubMed] [Google Scholar]
  52. Vehtari A, Gelman A, & Gabry J. (2016). Practical Bayesian model evaluation using leave‐one‐out cross‐validation and WAIC. Statistics and Computing, 27, 1413–1432. 10.1007/s11222-016-9696-4 [DOI] [Google Scholar]
  53. Vella EJ, Milligan B, & Bennett JL (2013). Participation in outdoor recreation program predicts improved psychosocial well‐being among veterans with post‐traumatic stress disorder: A pilot study. Military Medicine, 178, 254–260. [DOI] [PubMed] [Google Scholar]
  54. Von Benzon N. (2010). Moving on from ramps? The utility of the social model of disability for facilitating experiences of nature for disabled children. Disability & Society, 25, 617–626. [Google Scholar]
  55. Wang LP, & Maxwell SE (2015). On disaggregating between‐person and within‐person effects with longitudinal data using multilevel models. Psychological Methods, 20, 63–83. 10.1037/met0000030 [DOI] [PubMed] [Google Scholar]
  56. Watts BV, Schnurr PP, Mayo L, Young‐Xu Y, Weeks WB, & Friedman MJ (2013). Meta‐analysis of the efficacy of treatments for posttraumatic stress disorder. The Journal of Clinical Psychiatry, 74(6, 541–550. [DOI] [PubMed] [Google Scholar]
  57. Weathers FW, Litz BT, Keane TM, Palmieri PA, Marx BP, & Schnurr PP (2013). The PTSD checklist for DSM‐5 (PCL‐5). Retrieved from www.ptsd.va.gov
  58. Weiss CC, Purciel M, Bader M, Quinn JW, Lovasi G, Neckerman KM, & Rundle AG (2011). Reconsidering access: Park facilities and neighborhood disamenities in New York City. Journal of Urban Health, 88, 297–310. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

supinfo
NIHMS1702267-supplement-supinfo.docx (14.6KB, docx)

RESOURCES