Associations Between Procrastination and Subsequent Health Outcomes Among University Students in Sweden

Key Points Question Is procrastination associated with subsequent health outcomes among university students? Findings In this cohort study of 3525 Swedish university students, procrastination was associated with worse subsequent mental health (depression, anxiety, and stress symptom levels), having disabling pain in the upper extremities, unhealthy lifestyle behaviors (poor sleep quality and physical inactivity), and worse levels of psychosocial health factors (higher loneliness and more economic difficulties). Meaning This study suggests that procrastination may be associated with a range of health outcomes.


Introduction
Procrastination is defined as voluntarily delaying an intended course of action despite expecting to be worse off because of the delay 1 and is common, especially among younger people. 2,3It is estimated that at least half of university students engage in consistent and problematic procrastination, such as postponing studying for examinations or writing papers. 1,4Procrastination is described as a form of self-regulatory failure linked to personality traits such as impulsiveness, 5,6 distractibility, and low conscientiousness. 1,7An individual's tendency to procrastinate is relatively stable over time, but specific procrastination behaviors are influenced by contextual factors such as task aversiveness. 1For some students, procrastination is occasional and related to specific academic tasks, while for others it is more of a general disposition, potentially affecting academic achievements 8 and health.
0][11] Procrastination is also associated with prevalent general physical health problems, 12 cardiovascular disease, 13 and unhealthy lifestyle behaviors. 12,14,15udents engaged in university studies have high levels of freedom and low structure, which places high demands on their capacity to self-regulate. 16These high demands on self-regulation may explain the high prevalence of procrastination among university students and make persons who are prone to procrastinate more vulnerable to the negative consequences of procrastination while at the university. 16e procrastination health model 12,17 suggests that a general tendency to procrastinate is associated with negative health outcomes by increasing levels of stress, reducing healthy behaviors, and delaying treatment. 12,17,18However, the causal direction between procrastination and health outcomes is not well understood and could be bidirectional, with poor physical or mental health reducing energy levels and motivation, potentially leading to more procrastination.Some of the strongest support for the causal claims of the procrastination health model comes from studies on the effects of treating procrastination.Results from randomized clinical trials suggest that intervening on procrastination with psychological treatments can reduce subsequent levels of depression and anxiety and improve quality of life. 19However, the evidence is limited as there are only a few studies, with a focus solely on mental health outcomes. 19e procrastination health model provides a useful theoretical foundation for how procrastination could affect different health outcomes. 12,171][22][23] Furthermore, most studies have limited control for important confounders, including other health problems, age, and sociodemographic factors.Also, no study has, to our knowledge, controlled for the potential reverse causation that may arise if health problems increase procrastination.
In this study, we investigated the associations between procrastination and a range of subsequent health outcomes while controlling for a large set of potential confounders, including prior levels of the outcomes.This method allows for stronger conclusions regarding the causal direction 24 between procrastination and health outcomes and can provide a broader picture of the potential associations between procrastination and health.
We aimed to evaluate the associations between procrastination and 16 health outcomes at the end of a 9-month period.Following the procrastination health model, 12,17 we hypothesized that higher levels of procrastination would be associated with worse health outcomes at follow-up.

Design and Study Population
The cohort for this study came from the Sustainable University Life study, 25  Eligible students were informed about the study during an in-class presentation and/or by an email with a link to the survey.Students who agreed to participate were followed up every 3 months for 1 year at 5 time points.In the present data analysis, the first time point was defined as prebaseline and was used for measures of the covariates.The second time point was defined as the baseline and was used for the procrastination measure and the fifth time point was defined as the 9-month follow-up and used for measurement of the health outcomes (Figure 1).The follow-up period of 9 months represents the length of an academic year, which we believe is adequate for procrastination to manifest its potential associations with different health outcomes.The sample was restricted to those responding at baseline and included 3525 participants; the follow-up rate was 73% (n = 2587) at the 9-month follow-up (Figure 2).

Measures
Exposure: Procrastination Procrastination was measured using 5 items from the Swedish version of the Pure Procrastination Scale (PPS). 26The items were rated on a Likert scale from 1 ("very rarely or does not represent me") to 5 ("very often or always represents me") and summed to give a total procrastination score ranging from 5 to 25.The short version of the PPS includes items 4 through 8 from the full PPS and has shown adequate psychometric properties in nonclinical samples. 26,27In this sample, the scale had a Cronbach α of 0.92 at baseline and a test-retest reliability (Pearson r) of 0.79 over 3 months and of 0.75 over 9 months.

Outcomes
We assessed a range of health outcomes at the 9-month follow-up, including mental health problems (symptoms of depression, anxiety, and stress), disabling pain (neck and/or upper back, lower back, upper extremities, and lower extremities), unhealthy lifestyle behaviors (poor sleep quality, physical inactivity, weekly tobacco use, monthly cannabis use, daily alcohol use, and breakfast skipping), psychosocial health factors (loneliness and economic difficulties), and general health.For more information on the outcome measures, see eMethods 1 in Supplement 1.

Confounders
Potential confounders were selected using the modified disjunctive cause criterion. 28This approach is a feasible alternative to directed acyclic graphs when the causal knowledge needed to produce a reliable directed acyclic graph is unavailable. 28We used the same set of covariates in all the outcome models, as has been suggested for outcome-wide studies. 29The covariate set included all outcome variables at prebaseline to limit the risk of reverse causation, 30 as well as age, gender, highest parental level of education, previous physical and psychiatric diagnoses, civil status, place of birth, and university education type.All covariates were measured at prebaseline to ensure that none were mediators between procrastination and the outcomes.Information on covariate measurement and coding is provided in eMethods 1 in Supplement 1.

Statistical Analysis
The characteristics of the sample are presented in Table 1, both for the full sample and across quartiles of procrastination (procrastination score range, 5-25; first quartile, 5-9; second quartile, 10-12; third quartile, 13-17; and fourth quartile, 18-25).To assess whether procrastination was associated with the health outcomes at follow-up, we used the outcome-wide approach, 29 a newly developed analytic approach to study associations between an exposure and multiple outcomes.We built separate multivariable regression models for each of the outcomes to assess the association between baseline procrastination and the different outcomes at the 9-month follow-up.For continuous outcomes, linear regression models were used.For binary outcomes, modified Poisson regressions 31 were used (rather than logistic regression models because some outcomes were relatively common).
Estimates for continuous outcomes are thus interpreted as the difference in the outcome as measured in SDs associated with a 1-SD increase in procrastination, and estimates for the binary outcomes are interpreted as the risk ratio (RR) of the outcome associated with a 1-SD increase in procrastination.All models were adjusted for the set of prebaseline confounders described.For The definition of each time point (T) in the present study (T1, T2, and T5) refers to the time point in the Sustainable University Life (SUN) Cohort (Figure 1).Note that T3 and T4 are not shown here, because data from these time points were not used for the main analyses.
unadjusted estimates, see eTable 1 in Supplement 1. Visual inspection of the residuals from all models indicated fairly linear associations between procrastination and the outcomes.The homoscedasticity and distribution of the residuals were not assessed because they typically have minor impact on the regression estimates, 32 but the modified Poisson models still applies robust SEs.This analytic strategy deviates from the ones prespecified in our study protocol 25 but was selected as it was deemed most appropriate for the research questions.Analyses were conducted using R, version 4.1.2(R Group for Statistical Computing).

Missing Data
Outcome data were missing for 938 participants not responding at the 9-month follow-up, and the main analyses were performed as complete-case analyses (n = 2587).Prebaseline characteristics stratified by missingness at the 9-month follow-up are presented in eTable 2 in Supplement 1.We had complete data on the exposure and all covariates, except for the sleep quality variable.The Pittsburgh Sleep Quality Index (PSQI) had missing data on 3 items for sleep disturbances for 9% (n = 225) of the respondents in the analytic sample due to a technical error at the beginning of data collection.These missing data were handled by imputing the person-mean of the items for sleep disturbances.

Sensitivity Analyses
First, we computed the E-value 33,34 that assesses how strongly unmeasured and residual confounding would need to be associated with the exposure and the outcomes on the RR scale to reduce the point estimates to the null.When the 95% CIs did not include the null, we also computed the E-value of the 95% CI bound closest to the null, to assess the strength of unmeasured and residual confounding needed to shift the 95% CI to include the null.Second, we calculated the pointbiserial correlation between procrastination at baseline and missingness at follow-up to investigate the possibility of selection bias in relation to the internal validity.No association between the exposure and loss to follow-up indicates that selection bias cannot affect the internal validity, but only the external validity. 35Third, we conducted analyses controlling for prior levels of procrastination, as has been suggested for outcome-wide studies. 29These analyses had to be performed with a 6-month follow-up because we did not have data on procrastination at prebaseline (eMethods 2 and eTable 3 in Supplement 1).Fourth, the robustness of our results to the personmean imputation of values on the PSQI was assessed by performing multiple imputation for the 3 missing items (eMethods 3 in Supplement 1).

Results
At prebaseline, the study included 3525 participants (2229 women [63%] and 1274 men (36%); mean [SD] age, 24.8 [6.2] years) (Table 1).The mean (SD) procrastination score at baseline was 12.9 (5.4).Gender and age were similar across levels of procrastination, but participants with higher levels of procrastination tended to be more likely to study technical sciences, be single, and have been born outside of Europe.All outcome variables showed higher prevalence or mean levels at prebaseline among participants with higher procrastination levels at baseline.
For estimates with a 95% CI that excluded the null, the E-values suggested that unmeasured confounders would need to increase the risk of both procrastination and the outcomes by 35% to 86%, depending on the outcome, to move the point estimates to the null (Table 2).The point-biserial correlation between procrastination at baseline and missingness at the follow-up was 0.05.Outcome

JAMA Network Open | Public Health
Associations Between Procrastination and Health Outcomes Among University Students Downloaded From: https://jamanetwork.com/ on 03/10/2023 levels at prebaseline were generally somewhat higher in the group that was missing at the 9-month follow-up (eTable 2 in Supplement 1).The associations between procrastination and health outcomes 6 months later, while controlling for prior procrastination levels, were similar to the results in the main analysis (eMethods 2 and eTable 3 in Supplement 1).Using multiple imputation for the 3 missing items on the PSQI gave results identical to the main analysis when rounded to the second decimal place (eMethods 3 in Supplement 1).

Discussion
In this cohort of Swedish university students, higher levels of procrastination were associated with worse subsequent mental health (depression, anxiety, and stress symptom levels), having disabling pain in the upper extremities, unhealthy lifestyle behaviors (poor sleep quality and physical inactivity), and worse levels of psychosocial health factors (higher loneliness and more economic difficulties) 9 months later.We found no clear associations between procrastination and subsequent disabling pain in other body regions (neck and/or upper back, lower back, or lower extremities), other unhealthy lifestyle behaviors (alcohol, tobacco, or cannabis use and breakfast skipping), or general health.
The identified associations are in the same direction, but to a lesser magnitude, than those reported in most previous studies on procrastination and health outcomes. 1,2,9,10,12There are several potential reasons for this difference in magnitude, such as the use of different scales to measure procrastination and the health outcomes.However, the arguably largest difference between this study and most previous studies is that we controlled for an extensive set of potential confounders.
To our knowledge, this is the first study of procrastination and health outcomes to adjust for prior levels of the outcomes.As shown in eTable 1 in Supplement 1, when controlling for prior levels of the outcome variables, the unadjusted estimates of associations between procrastination and the health outcomes corresponded more closely to those of some previous studies. 2,9,10By adjusting for prebaseline levels of the outcome variables, we reduced the risk of reverse causality. 30Still, interpreting these association as causal effects requires the assumptions of no unmeasured or a All models were adjusted for the set of prebaseline covariates described, including all outcomes prebaseline.Procrastination and all continuous outcomes were standardized (mean = 0 and SD = 1), and β is the standardized effect size.
residual confounding, selection bias, or other biases affecting the associations, 36 assumptions that are unlikely to be fully met in the present study.We do believe, however, that our results may correspond more closely to the consequences that interventions targeting procrastination would have for different health outcomes, than those of earlier cross-sectional studies.
Our estimates indicate that the associations between procrastination and subsequent health outcomes are weak.For instance, a 1-SD increase in procrastination was associated with a mean increase in subsequent depression symptoms of only 0.13 SDs, and many of the associations were weaker than this.The follow-up time of 9 months represents 1 academic year and was chosen because we believe this is sufficient induction time for procrastination to lead to health problems.It is possible, however, that these estimates would be stronger for a longer follow-up because the potential negative associations of procrastination with health outcomes could accumulate over time.
Still, even though the associations are not very strong, it seems that procrastination could have associations with many different aspects of health, including mental health, physical pain, lifestyle behaviors, and psychosocial health factors.Thus, although it seems that intervening on procrastination is unlikely to produce large associations with any specific health outcome, it could possibly produce small associations with a diverse set of different health outcomes.

Limitations
This study has some limitations.The E-value analysis suggests that unmeasured and residual confounding (from, for instance, neuroticism, conscientiousness, or genetic factors) could potentially explain away the observed associations.This potential unmeasured or residual confounding would, however, need to be moderately associated with procrastination and the outcomes, independent of all measured covariates 29 (ie, associated with procrastination and the outcomes even after adjusting for all covariates).Given the large number of confounders adjusted for in these analyses, it is likely that a substantial proportion of any unmeasured confounding is already accounted for in our analyses.Controlling for prior procrastination levels gave results similar to those in the main analysis, limiting the risk of reverse causality and unmeasured confounding 29 (eMethods 2 and eTable 3 in Supplement 1).
We also have the risk of misclassification of exposure and outcomes, which is probably not an issue for the procrastination measure given that the PPS showed excellent reliability.The outcomes, however, differ somewhat in their psychometric properties, and misclassification may have attenuated the observed associations for some of the outcomes.
We found that procrastination at baseline had a near-null association with loss to follow-up, so it is unlikely that the internal validity of observed associations is subject to selection bias. 35There were, however, small differences in prebaseline characteristics, especially prior outcome levels, between responders and nonresponders, which could affect external validity (eTable 2 in Supplement 1).Furthermore, the sample is not fully representative of the overall Swedish student population (for instance, we have more medical and health and technical students than average).
Thus, it is uncertain whether our estimates are generalizable to Swedish students overall or to other populations.Furthermore, our measures were collected during the COVID-19 pandemic, which could affect generalizability to other time periods.

Figure 1 .
Figure 1.Timeline of Data Collection in the Sustainable University Life (SUN) Cohort and How Data Were Organized for the Present Study Time point in the SUN cohort (time since enrollment)

Figure 2 .
Figure 2. Flowchart of the Inclusion of Participants

Table 1 .
Characteristics of Participants Prebaseline for the Full Sample and by Quartiles of Procrastination Levels a a Procrastination was measured at baseline, while all covariates and outcomes were measured prebaseline.Percentages of some categorical variables do not add up to 100% due to rounding.bTheprocrastination score ranged from 5 to 25, with allergies; respiratory diseases; cardiovascular diseases; gastrointestinal diseases; diabetes; diseases in urinary tracts or internal or external genitals; diseases in the nervous system, eyes, or ears; mental disorders; tumors or cancers; and attention deficits or learning disabilities.

Table 2 .
Procrastination and Subsequent Health Outcomes a Assessment of the Outcome and Confounder Variables eTable 1. Comparisons of Estimates Under Different Adjustments eTable 2. Characteristics of Participants at Pre-baseline Stratified by Missingness at the Nine-Month Follow-up eMethods 2. Sensitivity Analysis Controlling for Prior Levels of Procrastination eTable 3. Procrastination (T3) and Subsequent Health Outcomes Six Months Later (T5), Adjusted for Prior Levels of Procrastination (T2) eMethods 3. Sensitivity Analysis for the Imputation of the Three Missing Items on the Pittsburgh Sleep Quality