Workability and Requests for Flexible Work Arrangements Among Older Adults: The Role of a Time and Place Management Intervention

Abstract

This article investigates the effect of an intervention on the workability of older adults (i.e., the competence, health, and other mental and physical characteristics that workers need to meet the demands of their jobs). We used data from health care workers (N = 437) who participated in a “time and place management” (TPM) intervention. Although related to flexible work options that aim to give workers more choice and control over the time and place of their work, TPM is conceptually distinct in that it focuses on the processes and guidelines necessary to the successful management of choice and control rather than the options alone. We focused on how the TPM intervention moderated the relationship between age and workability over time, with a particular focus on variation by baseline workability. Our results indicated that the intervention can benefit older workers with low workability.

Keywords

workability older workers workplace flexibility job demands

The labor force participation rate of workers aged 55 and older is expected to rise over the next decade (Toossi, 2013), as many older adults do want to continue working, both for enjoyment and for financial reasons (AARP, 2014). Unfortunately, many may lack the skills, competencies, or health needed to continue working in the jobs that are available to them. Workplace policies (i.e., the formal guidelines about how the workplace functions), to the extent that they directly or indirectly decrease the mismatch between workers and their jobs, may influence whether older adults are really able to work longer.

There are numerous ways to evaluate the potential role of workplace policies on the well-being of older workers. In this article, we argue that the concept of “workability” is a particularly promising way to understand the potential effects of workplace policies on older adults. We define workability—a term that Finnish researchers coined in the 1980s as a response to what they perceived as an overemphasis on disability (Ilmarinen, 2006, 2009)—as “the occupational competence, the health required for the competence, and the occupational virtues that are required for managing the work tasks” (Tengland’s 2011 definition for skilled occupations, p. 275).

Generally, from a public policy perspective, workability may be a useful measure of the consequences of workplace policies, because it has a clear link to whether the older adult can continue to work. Workability has an important advantage over more specific outcome measures (e.g., mental health) in that it allows researchers and human resource professionals to compare the results of various workplace interventions (e.g., introducing an exercise program or a training program) directly. Although a health-focused intervention such as an exercise program might improve the physical and mental health of older adults, it may insufficiently address the issue of worker skills and competencies. Similarly, a training program might increase their skills and competencies, but does not address their physical and mental health. Workability is a shared metric that includes considerations of health as well as of competencies, and as such is of immediate and practical interest to both employers and policy makers alike. To increase workability, an intervention might address any of several factors, such as occupational competencies and skills, mental and physical health, or the time and place fit between work and other aspects of life. As it is seldom feasible to collect longitudinal data that would follow workers years after an intervention to evaluate the ultimate effects on how long they were able to work, workability also provides a shorter term measure that may serve as a basis for comparison between interventions.

Existing intervention research suggests that workplace policies can shape workability. For instance, researchers have found that interventions designed to improve health and lifestyle behavior (Pohjonen & Ranta, 2001) and job restructuring (Marqueze, Voltz, Borges, & Moreno, 2008) are associated with higher workability. However, to our knowledge no existing study has used a randomized design to examine the effect of a time and place management (TPM) intervention on workability. In the present study, we assess the relationship between a randomized TPM intervention and workability among workers aged 50 and older at a series of hospitals run by a single provider (referred to here by the pseudonym “ModernMedical”). From an employee perspective, TPM introduces some specific flexible work options, and also provides structure and encouragement to aid in the use of those options. Previous research has found flexible work options to decrease employee turnover (Pavalko & Henderson, 2006) and absenteeism (Council of Economic Advisers, 2010) and to increase organizational performance (Combs, Liu, Hall, & Ketchen, 2006) and workers’ productivity and organizational commitment (Eaton, 2003). Through its link to greater life satisfaction, better mental health, and increased ability to negotiate the different demands of work and family (Bond, Thompson, Galinsky, & Prottas, 2002; Clark, 2001), flexible work options might also lead to better workability. From a management perspective, TPM is more than just introducing flexible work options. It is also a comprehensive approach providing structure and encouragement for managers and employees to communicate effectively, facilitating the functional implementation of those options. We explain the TPM intervention development process in greater detail in the “Method” section of this article.

In addition to its substantive findings about the relationship between the TPM intervention and workability, this article contributes to the literature in two related ways. First, it applies the person–environment (PE) fit framework (Edwards, 2008) to the concept of workability. There is an implicit conceptual match between PE fit and workability, in that the workability construct allows us to examine the extent to which characteristics of the person and characteristics of the job fit together. Although little research on workability has directly used this framework, the related research that does exist suggests this may be a promising approach. For instance, Merecz and Andysz (2012) found that optimal levels of person–organization fit help to maintain workers’ mental health (one facet of workability). Second, methodologically our study suggests a statistical approach for dealing with the typically skewed nature of workability in the employed population. Linear regression is often ill suited for studies of workability because of the highly skewed distribution, as people with low workability tend to exit the labor force early. Our study suggests that quantile regression, which fits an adjusted percentile (such as the median) rather than fitting the mean as in ordinary least squares (OLS) regression, may be a useful approach.

Background and Hypotheses

In this article, we use PE fit as a framework for understanding the potential relationship between the TPM intervention and workability among older adults. The PE fit framework focuses on the match between an individual and his or her context, based on both personal and environmental attributes (Kristof-Brown & Guay, 2011). According to this framework, behavior results from the dynamic interplay between individuals and their environments, with optimal behavior tied to optimal matching between individual and environment (Edwards, 2008). PE fit has the advantage of being more inclusive than a focus on the individual or the environment alone. This inclusiveness allows researchers to better capture the real complexity of subjects to which they apply the PE fit framework, and to provide many alternative solutions to problematic situations. In our study of older adults and workability, using PE fit allows us to meaningfully take account both of individual attributes such as age and health, while proposing a solution that focuses primarily on how to increase the fit between worker needs and job characteristics.

Arguably one of the most dominant and longest standing theoretical frameworks in the literature about the workplace (Kristof-Brown & Guay, 2011), PE fit has roots in Parsons’s (1909) theory of career selection, which emphasized the benefits of appropriately matching the traits of the workers with the nature of their vocation; Murray’s (1938) needs typology, which indicated that psychological needs vary by individual; and Lewin’s (1935) insight that the person and the environment interact to determine behavior. Due both to its long history and the general nature of its insights, PE fit encompasses a wide variety of subtheories. For instance, Muchinsky and Monahan (1987) characterized PE fit as either supplementary (the matching of person to environment on the basis of shared characteristics) or complementary (matching based on differences that are mutually beneficial when combined). In the present study, we consider workability as one facet of person–job fit, a variety of complementary fit that focuses on how a person’s individual traits fit with the requirements of a specific occupation or job (Kristof, 1996).

In addition to our theoretical perspective (PE fit), we take a practice perspective in which we evaluate each of our hypotheses for more vulnerable and less vulnerable workers. One of the major criticisms of workplace policies is that they can be perceived as mainly benefiting workers who are already advantaged, that is, if primarily highly paid professionals have access to schedule flexibility, does it really help to delay their retirements? What is potentially more useful from a practice perspective is whether the more vulnerable workers, those at risk of early retirements due to low workability and for subsequent financially vulnerable retirements, benefitted. In this study, we tested for differences in effect between more vulnerable workers (defined as those with low baseline workability) and less vulnerable workers (defined as those with high baseline workability).

Age and Declining Workability

Based on existing research, to the degree that the demands of the workplace environment exceed the resources of the worker, PE fit worsens (Edwards, 2008; Johnson, Mermin, & Resseger, 2011; Zacher, Feldman, & Schulz, 2014). Previous research has identified age as an important antecedent of workability (van den Berg, Elders, de Zwart, & Burdorf, 2009). Numerous studies, such as Monteiro, Ilmarinen, and Corraa Filho (2006), have indicated that older workers have lower workability than younger workers. Other research suggests that these declines in workability are attributable to a variety of factors. First, as the prevalence of physical limitations increases steadily with age (Holmes, Powell-Griner, Lethbridge-Cejku, & Heyman, 2009), older workers may be less suited to meeting the physical requirements of some jobs. Second, family caregiving demands can intensify for middle-aged and older adults, adding stress to the negotiation of work and family roles (Sterns, Barrett, Czaja, & Barr, 1994) and hence decreasing workability to the extent that they sap mental resources. Third, older workers have historically placed less emphasis on participating in training, because they are at later stages in their careers (Ng & Feldman, 2012). Although career stage theorists identify different numbers of stages (Hall, 2002; Super, 1990), most tie the stages loosely to chronological age. To the extent that older workers perceive themselves as late career, they may invest less in updating skills, which will in turn lower their workability. Due to changing physical health, mental health, and skills and competencies, we posit that overall,

Hypothesis 1: Age and workability are negatively associated.

As described above, we evaluate the hypothesized relationship for both those who are more vulnerable (low baseline workability) and those who are less vulnerable (high baseline workability). For the purpose of this analysis, we look at workers aged 50 and older. Many of these workers are near the ceiling of the workability measure at the beginning of the study, as this is the age around which we might expect to see declines, based on previous literature.

The Moderating Role of TPM Policies

Although on the basis of the existing studies we might expect workability to decrease with age, other studies have found a weak or nonexistent relationship between age and workability (van den Berg et al., 2009). These inconsistent findings point to the underlying complexity of workability, which encompasses both personal and environmental attributes (Zacher et al., 2014). Under unfavorable job conditions, such as heavy physical demands, older workers may experience marked declines in workability. However, under favorable job conditions, some older workers may have similar or higher workability compared with their younger counterparts (Gould, Ilmarinen, Järvisalo, & Koskinen, 2008). The TPM intervention sought to increase the favorability of job conditions for workers of all ages, but we would expect the intervention to be particularly important to older workers for two reasons. First, to the extent that older workers are at the greatest risk of declining workability (relative to physical health, mental health, and skills and competencies), a small schedule accommodation could make a big practical difference in workability. Although scheduling accommodations do not directly address all aspects of workability, the benefits of TPM could have a variety of ramifications. For example, an older worker with declining physical health might negotiate with management to work on a schedule that is less physically demanding, allowing him to do his work in shifts that let him perform at his best physical capacity. Second, in light of the relatively high importance of caregiving for this population (Sterns et al., 1994), a successful TPM program would increase workability by lessening the stress and mental demands placed on caregivers. We hypothesize as follows:

Hypothesis 2: The TPM intervention will moderate the association of workability and age, such that it is less negative for those in the treatment group.

Again, we evaluate the hypothesized relationship for both those who are more vulnerable (low baseline workability) and those who are less vulnerable (high baseline workability).

Method

In our study, we discuss the potential association between a randomized workplace intervention in a large hospital system and the workability of the employees in that system. For the intervention, work groups were randomly allocated either to a treatment group or to a control group. The treatment group received access to a structured learning module and follow-up materials designed for employees and managers to better understand and implement TPM processes. Below, we describe the development process, intervention design, and analysis in more detail.

Development Process

Discovery process

The study team used a four-strand discovery process to gather information with which to design the intervention. The strands included (a) a review of academic literature relevant to TPM interventions, with a focus on research documenting workplace flexibility initiatives in award-winning hospitals; (b) a review of survey data from ModernMedical employees, most of which ModernMedical had gathered as part of their internal monitoring and assessment; (c) 30-min telephone interviews with individual managers (n = 15) randomly selected from the 11 job categories at ModernMedical; and (d) 1-hr telephone focus groups with employees (n = 40) selected randomly from 10 of the 11 job categories. Sample questions from telephone interviews included the following: “What is the process for how schedules are created with your staff?” “What about that process works well for you as a manager?” “Is there anything about the process that doesn’t work well?” and “If something is not working well, do you feel you have any leeway to change the process?” Sample focus group questions included the following: “What is working well for you right now in terms of how your work hours are scheduled?” “What is not working so well?” and “Please rate the following statements about your work hours/shift using a scale from 1, strongly disagree to 5, strongly agree; we will offer three statements, and then stop for discussion [‘my work hours are too long’; ‘I don’t have enough work hours’; ‘my work hours are reasonably predictable’].”

Collaborative design

In part because TPM depends so strongly on organizational resources and constraints, effective TPM interventions must be designed collaboratively, with input from employees and managers. This ensures TPM interventions are suited to each organization. Using the information gathered as part of the discovery process, the research team collaborated with senior personnel at ModernMedical to design the TPM intervention. The senior vice president of human resources and ModernMedical’s chief executive officer headed an oversight committee that met with the research team weekly. Although this collaborative form of design is time-consuming, it was indispensable in designing a TPM intervention consistent with existing organizational norms and procedures.

Intervention Design

Learning module

The research team and ModernMedical personnel developed a 30-min learning module. The learning module encouraged greater discussion among managers and employees about making TPM fit requests, or requests to change schedules for a better fit between work and personal responsibilities. Examples of fit requests included changing a shift to deal with a caregiving issue, requesting a reduction in responsibilities, and requesting job sharing. This system was designed to take into consideration the needs of employees to attain good work–life balance and the needs of the organization to meet desired business outcomes, while also keeping front and center the need to provide quality care to patients. The learning module provided a systematic framework for discussing TPM fit requests, which encompassed (a) a structured classification of TPM fit request types, (b) a standardized procedure for employees to deliver requests, and (c) a standardized procedure for managers to process requests.

A number of the components of the learning module were forms and guides available to employees and managers on an ongoing basis. These included a flowchart of processes involved in discussing, requesting, evaluating, and implementing TPM fit changes; a table outlining a typology of request options; a self-assessment tool regarding appropriate TPM options; a request form; conversation guides for managers and employees in the case that either one broaches the topic of TPM fit requests; a follow-up conversation guide for managers and employees; and a tracking form for managers to note TPM-related conversations and outline the contents and outcomes of those discussions. The learning module provided managers and employees with the necessary understanding of these tools.

Supporting material

To continue education and maintain focus and motivation concerning TPM fit requests, the study team featured posters and videos on an ongoing basis following the launch of the learning module. The eight posters, which contained information about work–life fit, provided motivational reminders for employees and managers to continue thinking about and using TPM options. Example titles include “Most employers offer time and place management to their employees” and “Compressed workweeks [are] available at over half of workplaces.” The posters, designed to be eye catching, were successively displayed in a common space for a period of 3 weeks. Following the sequence of posters, a series of three 3-min videos were introduced via a link over email. The videos were called “Finding ‘Fit!,’” “Getting to ‘Yes!,’” and “No Happens.” Each video and poster emphasized key aspects of the intervention. For instance, “No Happens” clarified that not all TPM fit requests would be approved, a particularly important point as the hospital needed to balance employee and patient needs (e.g., the need for 24-hr coverage). The videos served the same purpose as the posters: to provide information and motivation related to TPM and to work–life fit and to keep employees and managers actively engaged in exploring options and utilizing the training materials from the online module.

Randomization

The TPM intervention occurred at the work group level and included both managers and employees at a system of hospitals run by a single medical provider. Because reactions to a TPM intervention may be influenced by work group culture and individual beliefs in ways that are difficult or impossible to measure, randomization—one of the most reliable ways to control for bias attributed to unmeasured differences between treatment and control groups—was particularly desirable in this case. Randomization by work group, rather than by individual, was necessary due to the structure of the intervention, such as the introduction of procedures and materials that employees and managers used for TPM fit requests.

This was a complicated process because, like many organizations, ModernMedical’s work groups did not always have clearly defined boundaries. For instance, ModernMedical contained many work units with more than one manager, managers with more than one work unit, and work units comprised entirely of managers. For the purpose of randomization, managers of managers were omitted. This allowed for the identification of 172 independent manager-work unit clusters (i.e., linked work groups or units), which included 439 work groups or units. The clusters were used in the randomization process to limit cross-contamination, as groups in which there were high linkages (e.g., a shared manager) were allocated to the same cluster. Based on cluster and a randomly assigned number, the eligible work units were assigned to either the treatment group (60%) or the control group (40%).

Data collection and sample

The study team invited nearly all employees, including 8,270 employees and 646 managers, to complete the baseline survey (referred to as baseline/Wave 1), prior to the intervention. The baseline survey was fielded from September to October of 2012. The intervention began in December of 2012. In all, there were six additional surveys, but only three included employees. The three follow-up surveys including employees occurred in March-April 2013, September-October 2013, and January 2014. It is important to note that only a small subset of employees (e.g., physicians) were excluded from the survey, but not all of these employees were randomized due to the complications noted above.

Although 63% of invited employees responded to at least one survey, our starting sample included only respondents who answered the baseline survey (n = 3,255, or 39% of the invited employees). Additional sample criteria included that respondents were in the treatment or control group, as some work groups could not be randomized due to irregular structures (remaining n = 2,818) and that respondents were employees rather than managers, so that they were presented with the workability items (remaining n = 2,538). Of these, similar percentages were aged 50 or older (36%) and female (84%), when compared with ModernMedical’s administrative data (35% and 80%). In addition, a small number of respondents were omitted due to item nonresponse on the dependent variable (workability), leaving a sample of 2,362 respondents. Of these, 752 were aged 50 or older (n = 337 control and n = 415 treatment), and 437 responded to at least one follow-up time point (n = 437).

Due to the small number of respondents answering the baseline and any given follow-up time point (e.g., baseline and Time 2, baseline and Time 3), we use the latest possible follow-up for each of the 437 respondents. We used multiple imputation (25 imputed datasets, generated using chained equations) to account for missing data. This approach increases the standard error of coefficients by an amount attributable to the variation between imputations, producing significance estimates that are both more conservative and less vulnerable to item nonresponse than simpler methods such as listwise deletion. For the purposes of this analysis, we consider only employee responses as workability was an employee-level variable.

Analysis

Intent to treat

The analysis used in this article used an intent-to-treat approach. Even if a given work unit had incomplete or low participation in the intervention, we considered them part of the treatment group. From a practice perspective, this provides more useful information than the alternative approach (i.e., considering only those with full or close to full participation as having received the treatment) because it more accurately represents what would happen in a real-world application of this type of intervention. In any actual application of an intervention, there is a substantial rate of noncompliance, yet excluding the noncompliant from analysis can dramatically overstate the actual effect of the intervention. Hence, the effect sizes for intent-to-treat analyses tend to be quite small due to the high rates of noncompliance (Gross & Fogg, 2004), much smaller than the effect size if only those who fully participated were considered.

Measures

Table 1 lists the variables used in this study. The dependent variable was workability, which asked respondents to rate their ability to continue working at their jobs for the next 5 years relative to a variety of demands and resources. Mean baseline and outcome workability were high (8.49 and 8.37, respectively), as were the individual components of baseline workability and physical and mental health. Key predictors included treatment, age, and workability at baseline, as well as two- and three-way interactions between these variables. Slightly over half of the respondents were in the treatment group (54.69%). Their median age was 56.65 and 86.50% were female. The control variables included whether female, occupation, physical health, mental health, whether eldercare responsibilities, and setting. Most respondents were either in the managerial, informational technology (IT), and clerical category (30.31%) or the nurses, pharmacists, technicians, and assistants category (42.33%). Approximately half were in direct care settings (i.e., ambulatory/outpatient, acute care/inpatient, clinic, or other direct care).

Table 1.

Description of Study Variables.

Outcome variable
Workability	Workability was the mean of six items that asked respondents to rate their ability to “continue working at [their] current job[s] for the next 5 years” relative to (a) their “lifetime best ability,” (b) the “physical demands of [their] job,” (c) the “mental demands of [their] job,” (d) their “skills, the competencies required,” (e) their “physical health,” and (f) their “mental health.” Possible responses for all items ranged from 0 (“unlikely”) to 10 (“highly likely”). Exploratory factor analysis indicated that, for the survey respondents overall, the six items loaded on a single factor. Confirmatory factor analysis indicated that the model fit was acceptable (TLI = 0.932, CFI = 0.959). The mean outcome workability was 8.37 (SD = 1.89).
Key predictors
Treatment	Treatment was a dichotomous variable. We coded respondents in work groups allocated to the treatment group as “1” and those in the work groups allocated to the control group as “0.” Potential respondents were allocated approximately evenly to the treatment and control groups. However, due primarily to higher nonresponse in the control group, 54.69% of the respondents were in the treatment group.
Age	We calculated age based on respondents’ reported birth year. We treated their age in 2012 (the time of the baseline data collection) as their age for all waves of data collection, and included age squared to account for quadratic nonlinearity. Mean age was 56.65 in this sample, which is limited to the 50 and older population due to the focus on workability.
Baseline workability	Baseline workability was a lagged dependent variable (i.e., the dependent variable at baseline) indicating the workability of the respondent before the intervention. Two- and three-way interactions between these variables were included. Baseline workability was 8.49 (SD = 1.73), and all components of baseline workability were similarly high (means ranging from 8.42 to 8.43).
Control variables
Whether female	Whether female was a dichotomous variable coded 1 for female or 0 for male. 86.50% of this sample was female, a gender distribution attributable to the occupations and industry represented in this study.
Occupation	Occupation was categorical, dummy coded for inclusion in multivariate models, including “managerial, IT, and clerical” (30.31%), “nurses, pharmacists, technicians, and assistants” (42.33%), “therapy and social work” (2.79%), “physical plant” (0.7%), and “other” (23.87%). These categories were developed with the help of ModernMedical to map on to the internal characteristics of the occupations (e.g., in terms of physical demands, mental demands, and schedule constraints).
Physical health	Self-rated physical health ranged from 0 (“worst possible health”) to 10 (“perfect health). The mean was 7.70 (SD = 1.49).
Mental health	Self-rated mental health ranged ranging from 0 (“worst possible health”) to 10 (“perfect health”). The mean was 8.28 (SD = 1.60).
Eldercare	We based eldercare responsibilities on an item asking respondents, “Do you provide care for or support to a family member age 65 or older who needs assistance?” coded 1 if “yes” and “0” otherwise. 22.65% of the sample had eldercare responsibilities.
Setting	Setting was a categorical variable, dummy coded for inclusion in multivariate models, including “No direct care to patients” (51.95%, reference group), “ambulatory/outpatient” (10.76%), “acute care/inpatient” (20.59%), “clinic” (9.61%), and “other” (7.09%).

Note. TLI = Tucker–Lewis index; CFI = comparative fit index.

We also tested a number of predictors that the previous literature has demonstrated as positively associated with workability; including marital status (Gould et al., 2008) and educational attainment (Monteiro et al., 2006), as well as additional control variables related to family responsibilities (childcare), socioeconomic status (income), and time of follow-up data collection. These are omitted from the final models due to non-significance.

Results

Table 2 shows the means, standard deviations, and Pearson correlations for the variables in our study. Baseline workability and outcome workability are positively correlated at r = .507 (p < .05). As we would expect, age is negatively correlated with outcome workability, at r = −.143 (p < .05). This would seem to lend support to the contention that age and workability are negatively related.

Table 2.

Means, Standard Deviations, and Correlations (N = 437).

		M	SD	Bivariate correlations
		M	SD	1	2	3	4	5	6a	6b	6c	6d	6e	7	8	9
1	Outcome workability	8.373	1.892	1.000
2	Treatment	0.547	0.498	−.014	1.000
3	Age	56.650	4.743	−.143	.043	1.000
4	Baseline workability	8.489	1.727	.507	.023	−.054	1.000
5	Whether female	0.865	0.342	.058	−.010	−.008	.004	1.000
6	Occupation
	a. Managerial, IT, and clerical	0.307	0.461	.121	.037	−.052	.060	.118	1.000
	b. Nurses, pharmacists, technicians, assistants	0.419	0.493	−.128	.027	.046	−.116	.118	—	1.000
	c. Therapy and social work	0.027	0.163	−.017	−.016	−.035	.025	−.057	—	—	1.000
	d. Physical plant	0.005	0.067	−.019	−.074	−.002	.004	−.172	—	—	—	1.000
	e. Other	0.243	0.429	.027	−.053	.017	.058	−.214	—	—	—	—	1.000
7	Physical health	7.700	1.491	.239	−.001	.099	.378	.091	−.016	−.060	−.023	−.032	.099	1.000
8	Mental health	8.283	1.599	.305	.038	.096	.558	−.010	−.022	−.014	.005	.009	.036	.511	1.000
9	Whether eldercare responsibilities	0.227	0.419	−.105	−.078	.047	−.051	−.074	−.052	.039	.043	.044	−.013	.054	.027	1.000
10	Setting
	a. No direct care to patients	0.519	0.500	.106	.045	−.044	.110	−.058	.520	−.650	−.175	.065	.245	.040	.045	−.037
	b. Ambulatory/outpatient	0.108	0.310	.056	−.025	−.049	.046	.051	−.199	.304	.032	−.024	−.145	.040	.077	.006
	c. Acute care/inpatient	0.206	0.404	−.216	−.037	−.017	−.206	.019	−.302	.348	.191	−.035	−.143	−.099	−.111	.008
	d. Clinic	0.096	0.295	.058	−.078	.037	.008	.038	−.099	.132	.040	−.022	−.058	−.033	−.082	.083
	e. Other	0.071	0.257	.000	.090	.129	.045	−.021	−.184	.199	−.046	−.019	−.011	.068	.090	−.043

Note. Correlations greater than .095 in absolute value are significant at p < .05. Correlations for dummy variables generated from the same categorical variable are not shown. Results are based on 25 imputed datasets. IT = information technology.

Table 3 shows coefficients from the models predicting median workability. Consistent with prior studies of workability among working adults (McGonagle, Fisher, Barnes-Farrell, & Grosch, 2015; Palermo, Webber, Smith, & Khor, 2012), the distribution of our workability index was negatively skewed, with most workers reporting high workability. The existing literature has attempted to address the skewed distribution of workability in a number of ways, such as defining poor workability as workability below a certain threshold. In our analysis we used quantile regression (Koenker & Bassett, 1978; Koenker & Hallock, 2001), focusing on the median workability. Quantile regression coefficients can be interpreted analogously to OLS regression coefficients, but the underlying point of reference differs. Whereas OLS regression estimates the conditional mean by minimizing the sum of least squared residuals, quantile regression estimates the conditional percentile by minimizing the sum of absolute residuals. Because the focus in quantile regression is on a percentile (in the case of this article, the median), it is useful when a distribution is skewed and the median is a more useful representation of the “average” than the mean.

Table 3.

Multivariate Models Predicting Median Workability (N = 437).

	b	(SE)
Main effects
Treatment	−0.451	(.161)**
Age	−0.033	(.022)
Age squared	−0.007	(.004)
Baseline workability	0.436	(.079)***
Two-way interactions
Treatment × Baseline Workability	0.276	(.096)**
Age × Baseline Workability	0.026	(.013)*
Age Squared × Baseline Workability	0.005	(.002)
Treatment × Age	0.022	(.030)
Treatment × Age Squared	0.008	(.005)
Three-way interactions
Treatment × Age × BW	−0.022	(.016)
Treatment × Age Squared × BW	−0.006	(.003)*
Control variables
Whether female	0.193	(.189)
Occupation
Nurses, pharmacists, technicians, assistants	−0.062	(.201)
Therapy and social work	0.174	(.416)
Physical plant	1.591	(.693)*
Other	−0.180	(.179)
Physical health	0.080	(.050)
Mental health	0.041	(.051)
Whether eldercare responsibilities	−0.268	(.150)
Setting
Ambulatory/outpatient	−0.086	(.245)
Acute care/inpatient	−0.576	(.211)**
Clinic	0.136	(.234)
Other	−0.011	(.280)
Constant	9.018	(.244)***

Note. Results are from a quantile regression model predicting median workability. Categorical variables are dummy coded, with the following reference groups: control group (for treatment); male (for whether female); managerial, information technology, and clerical positions (for occupation); no eldercare (for whether eldercare responsibilities); no direct care of patients (for setting). Interval-level predictors are centered at the mean for each model. Results are based on 25 imputed datasets. BW = baseline workability.

p < .05. **p < .01. ***p < .001.

Very few of the control variables were significant predictors. There was slight evidence that occupation was a factor in workability in this sample. Several main effects (treatment, b = −.451, p < .01; baseline workability, b = 0.436, p < .001), two-way interactions (Treatment × Baseline Workability, b = 0.276, p < .01; age and baseline workability, and b = 0.026, p < .05), and three-way interactions (Treatment × Age Squared × Baseline Workability, b = −0.006, p < .05) have significant effects. These coefficients indicate that the treatment and outcome workability were related, but not in a straightforward way.

To illustrate these results, we have shown median workability by age in Figure 1, for four different categories according to baseline workability (low = 1 SD below the mean, high = 1 SD above the mean) and treatment status. Among those with high baseline workability, outcome workability is high for both those in the treatment and control groups. Any difference in outcome workability between the two groups is modest.

Figure 1.

Predicted median workability for respondents aged 50 and older.

Among those with low starting workability, however, the relationship between baseline workability and age differs substantially depending on whether they were in the treatment group. For those in the treatment group, their workability remains low but does not decrease. For those in the control group, there is a noticeable nonlinear relationship between age and workability. Workability is highest for those in the mid-50s and then decreases, with the largest gap between the treatment and control noticeable for workers in their mid to late 60s. This indicates that, the relatively immediate benefits of the intervention were isolated to the oldest workers in our sample.

Both hypotheses received partial support. Regarding Hypothesis 1 (that age and workability were negatively associated), age and workability are not negatively associated for our sample as a whole, but they are for a specific subpopulation: workers with low starting workability (i.e., vulnerable workers) who are in the control group. Similarly, Hypothesis 2 (“The TPM intervention moderates the association of workability and age, such that it is less negative for those in the treatment group”) is only supported for workers with low starting workability. The implications of these results are discussed below.

Discussion

In this article, we argued that the PE fit framework can be useful for understanding the relationship between workplace policies and workability, and can help to elucidate some of the conceptual problems in the workability literature. Our hypotheses drawn from this framework received mixed support, indicating a distinct difference between less vulnerable workers (those with high starting workability) and more vulnerable workers (those with low starting workability). For those workers in our sample who had high starting workability, outcome workability was high even at the oldest ages included (into the mid to late 60s). However, for those with low baseline workability in the control group, outcome workability was lower for the oldest workers in the sample. Similarly, the treatment did moderate a negative association between age and workability, but only for those who were low in workability at the beginning of the study. Our results suggest that TPM might stem age-related declines in workability, but primarily for those workers who were already vulnerable.

The concept of workability includes a variety of factors, both individual and environmental, such as working conditions, mental health, social aspects of the job and the worker, and various types of aptitude that may come into play in weighing whether a particular person can be reasonably expected to work well in a particular job. The literature on workability supports the idea that a workplace intervention not focused on health may still have substantial positive effects on workability. High mental work demands and perceived job strain are associated with low workability (Gould et al., 2008; van den Berg et al., 2009). One of the most appealing aspects of targeting workability is that practitioners can choose workplace interventions based on a desired time horizon. For instance, workplace interventions for TPM may yield relatively quick improvements, as the present study (with a 1-year time frame) demonstrates. Because workability focuses on the fit between the person’s attributes (including their mental health, physical health, skills and competencies) and the job’s attributes (including the mental demands, physical demands, and the skills and competencies needed), workplace interventions could improve workability by focusing on the environment, the individual, or both.

Limitations and Suggestions for Future Research

Like all studies, this study must be understood in light of its limitations. First, the effect sizes were relatively small. However, the study used an intent-to-treat analysis, which generally will produce small effect sizes in studies of this type. In this case, the treatment group included workers who did not participate in the training or did not transfer the lessons learned to the workplace context. We do not have access to data on whether individual workers and managers participated. Although the effect sizes in an intent-to-treat analysis are susceptible to shrinkage (Gross & Fogg, 2004), they provide a relatively accurate portrayal of the real-world effects of such an intervention. Second, this sample cannot be understood as the representative of U.S. workers. However, as workability has more often been studied in the European context (e.g., many studies deal with Finnish public sector employees; see van den Berg et al., 2009), the U.S. private sector context provides a complementary source of information, that is, it provides complementary data on a different population. Third, though the study was longitudinal, the 1-year duration was relatively short. This study is of the most relevance to the degree that it can provide evidence that adopting more of a TPM approach in the workplace could be beneficial in general, and not just temporarily. If future TPM initiative were to last for several years, we might discover the gains we discovered were short lived, due to what would prove to be passing excitements and expectations regarding the initiative itself. However, we might find that, given enough time, median workability would improve throughout the age spectrum of older workers with low baseline workability; perhaps as the workplace culture adapts over time to the TPM policies, gains become even more substantial. Only a longer term study could address these questions. The majority of workability studies are cross-sectional, however (see van den Berg et al., 2009), and the results of this study regarding baseline workability highlight the importance of a longitudinal perspective. Fourth, due to a skewed distribution with clustering at the upper end (i.e., large numbers of workers reporting workability at 9 or 10), it was not possible to use quantile regression to compare the relative effects of predictors at different levels of the dependent variable. Hence, though quantile regression is a useful analytic strategy, for our data, it was not possible to select higher quantiles (e.g., the 75th percentile) to evaluate potential differences in the effect of the intervention. Fifth, although our data were nested within work groups, we could not evaluate the effect of clustering due to a small number of respondents in each cluster. We often had only one or two survey respondents 50 or older per work group. Sixth, because we had a design with one treatment rather than a multi-arm intervention study, this study cannot address only one aspect of PE fit. Another consequence of having only one treatment arm is that it is not possible to compare multiple arms to estimate the potential influence of respondents knowing they were in an intervention. Finally, due to limitations on survey length, some of the control measures were particularly abbreviated. Notably, mental health is measured by a single item ranging from 0 to 10.

The results of this study indicate many directions for future research. First, though this study involved a one-arm intervention aimed at schedule flexibility, future studies could investigate interventions with more than one arm. Second, this study is specific to hospital system workers, but different work settings in different industries and occupations will vary concerning the feasibility of implementing TPM. Thus, future research might investigate similar interventions in other industries. Third, though this study used Tengland’s (2011) definition of workability for skilled occupations, workability may take on a different form in unskilled occupations. Fourth, future longitudinal studies might involve a longer time frame with which information is obtained, to better ascertain longer term impacts of TPM interventions. Fifth, studies could scrutinize whether different measurement models of workability, when applied to the same sample, deliver comparable results. Potential measures include the standard Work Ability Index (WAI) (Tuomi, Ilmarinen, Jahkola, Katajarinne, & Tulkki, 1998), the concise measure that Thorsen, Burr, Diderichsen, and Bjorner (2013) suggest, and the measure used in this study. In particular, comparison with more objective measures (e.g., rates of hospitalization) could be helpful. Sixth, future research could investigate any psychosocial basis for changes in workability. For example, the TPM intervention may have caused a relatively fast boost to workability because it acted as a signal to employees that their employer cared about them, not because of specific tangible benefits. Finally, future research could investigate the potential lack of effect of interventions on the workability of workers who already have high workability. We found that the intervention was associated with higher outcome workability, but only for workers who began with lower workability. Although potential ceiling effects cannot be ruled out, this also raises questions about whether a small deficit in workability (e.g., workers who rate their workability as a 9) typically represents a single issue, such as occasional back pain, whereas a larger deficit in workability (e.g., workers who rate their workability as 4) represents a complex of issues (including poor health, family concerns, job demands). If this is the case, a larger range of interventions may be effective for those with lower workability, leading to a larger overall effect.

Conclusion

This study has several implications for organizational leaders. First, a key component of the intervention was the initial discovery process and the collaborative approach. Because of the importance of organization rules regarding overtime, scheduling, and other aspects of TPM, TPM interventions must be tailored for each individual organization. The development process described in this study provides a general framework. Second, especially important for organizations with rapid employee turnover, TPM initiatives may result in improvements in workability on a short time horizon. Although organizational leaders may be reluctant to provide interventions focused on other aspects of workability (such as physical or mental health) because of the perception that many employees may leave the organization before real gains can be garnered, TPM initiatives may offer relatively quick (although not necessarily large, when compared with other interventions) returns on investment. Third, it is also important for organizational leaders to take into consideration the subpopulation that benefitted from these interventions: older workers nearing retirement age who were already struggling with low workability. This suggests TPM interventions may be most effective as a way to extend retirement age for workers who are already having difficulties. It is not, however, a miracle cure, as evidenced by the low effect sizes. Those who benefitted did not develop high workability, but the intervention helped to stem the decline in workability from baseline to the subsequent survey.

From a public policy and societal perspective, workplace programs to improve workability are of critical importance. Workers with low workability may move to jobs with lower job stressors but also lower pay and benefits, or may retire sooner than they otherwise would have because jobs with more manageable levels of demands are not available to them. In many cases, these workers may not be able to afford such a move, leading to increased risk of financial insecurity in retirement. Given its critical importance, workability has received less than adequate attention, in terms of both research and general awareness rising, in the United States to date. In part, this is because many of the existing studies focus on samples from Finland specifically and the European Union as a whole, due to a combination of factors including role of political awareness raising through government programs (Ilmarinen, 2011). However, as the U.S. population continues to age, awareness rising around workability may be a key factor in encouraging older workers to remain in the workforce longer than they otherwise would have.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research study described in this article was funded by the Alfred P. Sloan foundation.

Author Biographies

Jeremiah C. Morelock is a PhD student in the Department of Sociology at Boston College. His current research interests include aging, health, inequalities, and organizations.

Tay K. McNamara is co-director of the Center on Aging & Work at Boston College. Her previous work has dealt with a wide range of issues related to older adults and quality of employment.

Jacquelyn B. James is co-director of the Center on Aging & Work at Boston College. Her previous work has looked at the effect of time and place management practices on a variety of worker outcomes, including engagement and work–family balance.

References

American Association of Retired Persons (AARP). (2014). Staying ahead of the curve: The AARP work and career study. Washington, DC: Author. Retrieved from http://www.aarp.org/content/dam/aarp/research/surveys_statistics/general/2014/Staying-Ahead-of-the-Curve-2013-The-Work-and-Career-Study-AARP-res-gen.pdf

Bond

J. T.

Thompson

Galinsky

Prottas

(2002). Highlights of the national study of the changing workforce: Executive summary. New York, NY: Families and Work Institute. Retrieved from http://familiesandwork.org/site/research/summary/nscw2002summ.pdf

Clark

S. C.

(2001). Work cultures and work/family balance. Journal of Vocational Behavior, 58, 348-365. doi:10.1006/jvbe.2000.1759

Combs

Liu

Hall

Ketchen

(2006). How much do high performance work practices matter? A meta-analysis of their effects on organizational performance. Personnel Psychology, 59, 501-528. doi:10.1111/j.1744-6570.2006.00045.x

Council of Economic Advisers. (2010). Work-life balance and the economics of workplace flexibility. Washington, DC: Executive Office of the President. Retrieved from https://www.whitehouse.gov/files/documents/100331-cea-economics-workplace-flexibility.pdf

Eaton

S. C.

(2003). If you can use them: Flexibility policies, organizational commitment, and perceived performance. Industrial Relations, 42, 145-167. doi:10.1111/1468-232X.00285

Edwards

J. R.

(2008). Person-environment fit in organizations: An assessment of theoretical progress. The Academy of Management Annals, 2, 167-230.

Gould

Ilmarinen

Järvisalo

Koskinen

(Eds.). (2008). Dimensions of work ability: Results of the Health 2000 Survey. Helsinki, Finland. Retrieved from http://www.etk.fi/fi/gateway/PTARGS_0_2712_459_440_3034_43/http%3B/content.etk.fi%3B7087/publishedcontent/publish/etkfi/fi/julkaisut/tutkimusjulkaisut/erillisjulkaisut/dimensions_of_work_ability_7.pdf

Gross

Fogg

(2004). A critical analysis of the intent-to-treat principle in prevention research. The Journal of Primary Prevention, 25, 475-489. doi:10.1023/B:JOPP.0000048113.77939.44

10.

Hall

D. T.

(2002). Careers in and out of organizations. Thousand Oaks, CA: SAGE.

11.

Holmes

Powell-Griner

Lethbridge-Cejku

Heyman

(2009). Aging differently: Physical limitations among adults ages 50 years and over: United States, 2001-2007 (NCHS Data Brief No. 20). Hyattsville, MD: National Center for Health Statistics. Retrieved from http://www.globalaging.org/health/us/2009/physicallimitations.pdf

12.

Ilmarinen

(2006). The ageing workforce—Challenges for occupational health. Occupational Medicine, 56, 362-364. doi:10.1093/occmed/kql046

13.

Ilmarinen

(2009). Work ability—A comprehensive concept for occupational health research and prevention. Scandinavian Journal of Work Environment & Health, 35, 1-5. doi:10.5271/sjweh.1304

14.

Ilmarinen

(2011). 30 years work ability and 20 years’ age management. In Nygard

C.-H.

Savinainen

T. K.

Lumme-Sandt

(Eds.), Age management during the life course: Proceedings of the 4th Symposium on Work Ability (pp. 12-22). Retrieved from http://www.arbeitsleben.at/cms/images/stories/age_management_2011.pdf

15.

Johnson

R. W.

Mermin

G. B.

Resseger

(2011). Job demands and workability at older ages. Journal of Aging & Social Policy, 23, 101-118. doi:10.1080/08959420.2011.551465

16.

Koenker

Bassett

Jr. (1978). Regression quantiles. Econometrica: Journal of the Econometric Society, 46, 33-50.

17.

Koenker

Hallock

(2001). Quantile regression: An introduction. Journal of Economic Perspectives, 15(4), 43-56. doi:10.1257/jep.15.4.143

18.

Kristof

A. L.

(1996). Person-organization fit: An integrative review of its conceptualizations, measurement, and implications. Personnel Psychology, 49, 1-49. doi:10.1111/j.1744-6570.1996.tb01790.x

19.

Kristof-Brown

Guay

R. P.

(2011). Person-environment fit. In Zedeck

(Ed.), APA handbook of industrial and organizational psychology, Vol. 3: Maintaining, expanding, and contracting the organization (pp. 3-50). Washington, DC: American Psychological Association.

20.

Lewin

(1935). A dynamic theory of personality: Selected papers of Kurt Lewin. New York, NY: McGraw-Hill.

21.

Marqueze

E. C.

Voltz

G. P.

Borges

F. N.

Moreno

C. R.

(2008). A 2-year follow-up study of work ability among college educators. Applied Ergonomics, 39, 640-645. doi:10.1016/j.apergo.2008.02.002

22.

McGonagle

A. K.

Fisher

G. G.

Barnes-Farrell

J. L.

Grosch

J. W.

(2015). Individual and work factors related to perceived work ability and labor force outcomes. Journal of Applied Psychology, 100, 376-398. doi:10.1037/a0037974

23.

Merecz

Andysz

(2012). Relationship between person-organization fit and objective and subjective health status (Person-Organization fit and health). International Journal of Occupational Medicine and Environmental Health, 25, 166-177. doi:10.2478/s13382-012-0020-z

24.

Monteiro

M. S.

Ilmarinen

Corraa Filho

H. R.

(2006). Workability of workers in different age groups in public health institutions in Brazil. International Journal of Occupational Safety and Ergonomics, 12, 417-427.

25.

Muchinsky

P. M.

Monahan

C. J.

(1987). What is person-environment congruence? Supplementary versus complementary models of fit. Journal of Vocational Behavior, 31, 268-277. doi:10.1016/0001-8791(87)90043-1

26.

Murray

H. A.

(1938). Explorations in personality. Oxford, UK: Oxford University Press.

27.

T. W.

Feldman

D. C.

(2012). Evaluating six common stereotypes about older workers with meta-analytical data. Personnel Psychology, 65, 821-858. doi:10.1111/peps.12003

28.

Palermo

Webber

Smith

Khor

(2012). Factors that predict work ability: Incorporating a measure of organisational values towards ageing. In Masaharu

(Ed.), Promotion of work ability towards productive aging: Selected papers of the 3rd International Symposium on Work Ability, Hanoi, Vietnam, 22-24 October, 2007 (pp. 45-58). Boca Raton, FL: CRC Press.

29.

Parsons

(1909). Choosing a vocation. Boston, MA: Houghton Mifflin.

30.

Pavalko

Henderson

(2006). Combining care work and paid work: Do workplace policies make a difference? Research on Aging, 28, 359-374. doi:10.1177/0164027505285848

31.

Pohjonen

Ranta

(2001). Effects of worksite physical exercise intervention on physical fitness, perceived health status, and work ability among home care workers: Five-year follow-up. Preventive Medicine, 32, 465-475. doi:10.1006/pmed.2001.0837

32.

Sterns

H. L.

Barrett

G. V.

Czaja

S. J.

Barr

J. K.

(1994). Issues in work and aging. Journal of Applied Gerontology, 13, 7-19. doi:10.1177/073346489401300102

33.

Super

D. E.

(1990). A life-span, life-space approach to career development. In Brown

Brooks

(Eds.), Career choice and development (pp. 197-261). San Francisco, CA: Jossey-Bass.

34.

Tengland

(2011). The concept of work ability. Journal of Occupational Rehabilitation, 21, 275-285. doi:10.1007/s10926-010-9269-x

35.

Thorsen

S. V.

Burr

Diderichsen

Bjorner

J. B.

(2013). A one-item workability measure mediates work demands, individual resources and health in the prediction of sickness absence. International Archives of Occupational and Environmental Health, 86, 755-766. doi:10.1007/s00420-012-0807-z

36.

Toossi

(2013). Labor force projections to 2022: The labor force participation rate continues to fall. Monthly Labor Review, 136, 1. Retrieved from: http://www.bls.gov/opub/mlr/2013/article/labor-force-projections-to-2022-the-labor-force-participation-rate-continues-to-fall-1.htm

37.

Tuomi

Ilmarinen

Jahkola

Katajarinne

Tulkki

(1998). Work Ability Index (2nd ed.). Helsinki: Finnish Institute of Occupational Health.

38.

van den Berg

T. I.

Elders

L. A.

de Zwart

B. C.

Burdorf

(2009). The effects of work-related and individual factors on the workability index: A systematic review. Occupational & Environmental Medicine, 66, 211-220. doi:10.1136/oem.2008.039883

39.

Zacher

Feldman

D. C.

Schulz

(2014). Age, occupational strain, and well-being: A person-environment fit perspective. In Perrewé

P. L.

Rosen

C. C.

Halbesleben

J. R. B.

(Eds.), Research in occupational stress and well-being, volume 12: The role of demographics in occupational stress and well being (pp. 83-111). Bingley, UK: Emerald Insight.