ArticlePDF AvailableLiterature Review

Does Monitoring Goal Progress Promote Goal Attainment? A Meta-Analysis of the Experimental Evidence

Authors:
  • European Food Information Council (EUFIC)

Abstract and Figures

Control theory and other frameworks for understanding self-regulation suggest that monitoring goal progress is a crucial process that intervenes between setting and attaining a goal, and helps to ensure that goals are translated into action. However, the impact of progress monitoring interventions on rates of behavioral performance and goal attainment has yet to be quantified. A systematic literature search identified 138 studies (N �= 19,951) that randomly allocated participants to an intervention designed to promote monitoring of goal progress versus a control condition. All studies reported the effects of the treatment on (a) the frequency of progress monitoring and (b) subsequent goal attainment. A random effects model revealed that, on average, interventions were successful at increasing the frequency of monitoring goal progress (d� �= 1.98, 95% CI [1.71, 2.24]) and promoted goal attainment (d� �= 0.40, 95% CI [0.32, 0.48]). Furthermore, changes in the frequency of progress monitoring mediated the effect of the interventions on goal attainment. Moderation tests revealed that progress monitoring had larger effects on goal attainment when the outcomes were reported or made public, and when the information was physically recorded. Taken together, the findings suggest that monitoring goal progress is an effective self-regulation strategy, and that interventions that increase the frequency of progress monitoring are likely to promote behavior change.
Content may be subject to copyright.
Does Monitoring Goal Progress Promote Goal Attainment?
A Meta-Analysis of the Experimental Evidence
Benjamin Harkin, Thomas L. Webb,
and Betty P. I. Chang
University of Sheffield
Andrew Prestwich, Mark Conner, and Ian Kellar
University of Leeds
Yael Benn
University of Sheffield Paschal Sheeran
University of North Carolina at Chapel Hill
Control theory and other frameworks for understanding self-regulation suggest that monitoring goal
progress is a crucial process that intervenes between setting and attaining a goal, and helps to ensure that
goals are translated into action. However, the impact of progress monitoring interventions on rates of
behavioral performance and goal attainment has yet to be quantified. A systematic literature search
identified 138 studies (N19,951) that randomly allocated participants to an intervention designed to
promote monitoring of goal progress versus a control condition. All studies reported the effects of the
treatment on (a) the frequency of progress monitoring and (b) subsequent goal attainment. A random
effects model revealed that, on average, interventions were successful at increasing the frequency of
monitoring goal progress (d
1.98, 95% CI [1.71, 2.24]) and promoted goal attainment (d
0.40,
95% CI [0.32, 0.48]). Furthermore, changes in the frequency of progress monitoring mediated the effect
of the interventions on goal attainment. Moderation tests revealed that progress monitoring had larger
effects on goal attainment when the outcomes were reported or made public, and when the information
was physically recorded. Taken together, the findings suggest that monitoring goal progress is an
effective self-regulation strategy, and that interventions that increase the frequency of progress moni-
toring are likely to promote behavior change.
Keywords: self-monitoring, self-recording, progress monitoring, self-regulation, behavior change
The present review investigates the impact of monitoring goal
progress on rates of goal attainment. Goals are mental representa-
tions of desired outcomes (Austin & Vancouver, 1996)—such as to
run a marathon or to be happy—and goal intentions are self-
instructions to act toward those outcomes (Sheeran & Webb, 2011;
Triandis, 1980). Goal intentions capture both the nature of the set
goal (e.g., the number of exercise sessions that one intends to
engage in this week) and how committed one is to attaining it (e.g.,
the strength of one’s intention to exercise five times this week).
Intentions are the starting point for the willful control of action
(Gollwitzer & Moskowitz, 1996). However, evidence indicates
that intentions have only a modest impact on performance. A
meta-analysis of 47 experimental studies found that a medium-to-
large-sized change in intentions had a small-to-medium-sized ef-
fect on subsequent behavior (Webb & Sheeran, 2006). Evidence
indicates that people who intend to exercise do not necessarily do
so (Rhodes & de Bruijn, 2013), that most people want to be
happier than they are (Oishi, Diener, & Lucas, 2007), and that it
has become almost as traditional to fail to achieve New Year’s
resolutions as it is to form them in the first place (Marlatt &
Kaplan, 1972; Norcross & Vangarelli, 1988). In short, forming a
goal intention is not, on its own, sufficient to ensure goal attain-
ment (for reviews, see Gollwitzer & Sheeran, 2006; Sheeran,
Milne, Webb, & Gollwitzer, 2005; Sheeran & Webb, 2011; Webb,
2006).
This “gap” between intention and action (Sheeran, 2002) has led
researchers to investigate which factors determine intention-
behavior consistency. For instance, properties of intentions such as
temporal stability (Cooke & Sheeran, 2004; Sheeran & Abraham,
2003), the extent of actual control over performance (Sheeran,
Trafimow, & Armitage, 2003), and the operation of habits (Neal,
Wood, Wu, & Kurlander, 2011; Ouellette & Wood, 1998) have
each been shown to moderate the relationship between intention
and behavior (for reviews, see Sheeran & Webb, 2011; Webb &
Sheeran, 2006). There is also evidence concerning the cognitive
processes that support the translation of goals into action. For
This article was published Online First October 19, 2015.
Benjamin Harkin, Thomas L. Webb, and Betty P. I. Chang, Department
of Psychology, University of Sheffield; Andrew Prestwich, Mark Conner,
and Ian Kellar, School of Psychology, University of Leeds; Yael Benn,
Department of Psychology, University of Sheffield; Paschal Sheeran, De-
partment of Psychology, University of North Carolina at Chapel Hill.
The research was supported by a grant from the European Research
Council (ERC-2011-StG-280515) to Thomas L. Webb. The first and sec-
ond authors contributed equally to this research.
Correspondence concerning this article should be addressed to Benjamin
Harkin or Thomas L. Webb, Department of Psychology, University of
Sheffield, Western Bank, Sheffield S10 2TN, United Kingdom. E-mail:
b.harkin@sheffield.ac.uk or t.webb@sheffield.ac.uk
Psychological Bulletin © 2015 American Psychological Association
2016, Vol. 142, No. 2, 198–229 0033-2909/16/$12.00 http://dx.doi.org/10.1037/bul0000025
198
example, adopting a goal heightens the activation of goal-relevant
information (Aarts, Dijksterhuis, & De Vries, 2001) and inhibits
alternative goals (Shah, Friedman, & Kruglanski, 2002; for a
review, see Johnson, Chang, & Lord, 2006). However, these
findings raise the question: What do people actually do between
setting and getting a goal?
Many theories in social and health psychology accord goal
intentions the key role in determining behavior, including the
Theory of Reasoned Action (Fishbein, 1980; Fishbein & Ajzen,
1975), the Theory of Planned Behavior (Ajzen, 1991), Social
Cognitive Theory (Bandura, 1986, 1991, 1999), the Model of
Interpersonal Behavior (Triandis, 1977, 1980), Protection Motiva-
tion Theory (Rogers, 1983), the Prototype–Willingness Model
(Gibbons, Gerrard, Blanton, & Russell, 1998; Gibbons, Gerrard, &
Lane, 2003), and Locke and Latham’s (1990) Theory of Goal
Setting. However, for the most part, these theories do not specify
the processes that intervene between intention formation and goal
attainment (de Bruin et al., 2012). An important exception is
Control Theory (Carver & Scheier, 1982; Powers, 1973). Accord-
ing to Control Theory, goal setting simply reflects the adoption of
a reference value or standard for performance. For example, some-
one who decides to try to lose weight might aim to lose 2 lbs a
week. The crucial activity of goal striving, however, is monitoring
goal progress—that is, evaluating one’s ongoing performance rel-
ative to the standard—and responding accordingly.
Monitoring goal progress involves periodically noting qualities
of the target behavior (e.g., how much one has eaten) and/or
outcome (e.g., how much weight one has lost) and comparing these
perceptions with the desired standard (e.g., lose 2 lbs; Baumeister
& Vohs, 2007; Carver & Scheier, 1982; Webb, Chang, & Benn,
2013). Progress monitoring should promote goal attainment be-
cause it serves to identify discrepancies between the current state
and the desired state, and thus enables people to recognize when
additional effort or self-control is needed (Fishbach, Touré-Tillery,
Carter, & Sheldon, 2012; Myrseth & Fishbach, 2009). For exam-
ple, dieters who monitor their intake of calories can better decide
whether they should allow themselves to have an extra helping of
food. Expending effort or exerting self-control serves to bring
behavior in line with a standard. However, progress monitoring
precedes efforts to reduce discrepancies—discrepancies must first
be identified before people can adjust their behavior appropriately.
A number of models posit a central role for progress monitoring,
including Feedback Intervention Theory (Kluger & DeNisi, 1998),
Goal Setting Theory (Latham & Locke, 1991), Field Theory
(Lewin, 1951), models of self-awareness (e.g., Duval & Wicklund,
1972), Kanfer and Karoly’s (1972) account of self-regulation, the
Test-Operate-Test-Exit system (Miller, Galanter, & Pribram,
1960), the “living systems perspective” (Ford, 1987), and the
Model of Multiple-Goal Pursuit (Louro, Pieters, & Zeelenberg,
2007). Like Control Theory, these models suggest that the real
“work” of goal striving involves monitoring goal progress and
acting on discrepancies. Prompting the self-monitoring of goal
progress is also frequently deployed as a technique for promoting
behavior change. A recent review reported that 38% of interven-
tions designed to promote healthy eating and physical activity
incorporated progress monitoring (Michie, Abraham, Whittington,
McAteer, & Gupta, 2009). Monitoring goal progress is also an
important component of clinical practice (for reviews, see Feb-
braro & Clum, 1998; Korotitsch & Nelson-Gray, 1999) and inter-
ventions designed to reduce energy usage (for a review, see Abra-
hamse, Steg, Vlek, & Rothengatter, 2005).
Despite the theoretical and empirical prominence of progress
monitoring, however, the field lacks an empirical synthesis of its
impact on goal attainment. There are numerous meta-analytic
reviews of the impact of goal intentions on goal attainment (e.g.,
Albarracín, Johnson, Fishbein, & Muellerleile, 2001; McEachan,
Conner, Taylor, & Lawton, 2011; Ouellette & Wood, 1998;
Sheeran, 2002; Webb & Sheeran, 2006) and the factors that
influence people’s ability to act on discrepancies such as trait
self-control (de Ridder, Lensvelt-Mulders, Finkenhauer, Stok, &
Baumeister, 2012), ego-depletion (Hagger, Wood, Stiff, & Chatz-
isasrantis, 2010), and if–then planning (Gollwitzer & Sheeran,
2006). However, it is not yet clear whether, or to what extent,
monitoring goal progress promotes goal attainment. The present
review therefore quantifies the impact of progress monitoring on
rates of behavioral performance and goal attainment. In so doing,
the review both Tests Control Theory (and related theories) and
assesses the utility of progress monitoring as a behavior change
technique (Abraham & Michie, 2008).
Available Evidence Concerning the Relation Between
Progress Monitoring and Goal Attainment
The available evidence offers a mixed picture of the impact of
progress monitoring on goal attainment. Some studies have ob-
served that progress monitoring promotes goal attainment. For
example, Polivy, Herman, Hackett, and Kuleshnyk (1986, Study 1)
investigated the effect of being able to monitor consumption on
unhealthy eating. Female dieters were asked to taste some choco-
lates and to eat as many as they needed to evaluate them accu-
rately. Polivy et al. (1986) manipulated how easy it was for
participants to monitor their consumption; some participants were
asked to leave their chocolate wrappers on the table, while others
were asked to place them in a wastebasket that was already half
full of wrappers. The main finding was that participants who were
asked to leave their wrappers on the table ate fewer chocolates than
participants who were asked to put their wrappers in the waste-
basket, presumably because leaving the wrappers on the table
made it easier for participants to monitor how many chocolates
they had eaten.
Evidence also points to a relationship between the ability to
identify discrepancies (between the current state and desired state)
and self-control. For example, Skoranski et al. (2013) found that,
relative to normal weight children, obese children were poorer at
monitoring their performance on a variant of the Stroop task, as
indicated by blunted error-related negativity in their neural acti-
vation. This finding suggests that problems identifying when ac-
tions deviate from goals may hamper self-regulation and could
have contributed to their obesity (Smith & Mattick, 2013, reported
a similar relationship among heavy drinkers). Similarly, Chambers
and Swanson (2012) found that people who were successful in
maintaining weight loss tended to monitor their weight and have a
clearly defined upper limit (a “trigger point,” such as an increase
in dress size or gaining 10 lbs) at which they would take action to
reduce their weight.
Other studies have observed no effects of progress monitoring
on outcomes, however. For example, DeWalt et al. (2006) ran-
domly allocated patients with heart failure to receive usual care or
199
PROGRESS MONITORING AND GOAL ATTAINMENT
an intervention emphasizing the importance of daily self-weighing.
Although patients who were exposed to the intervention reported
monitoring their weight daily, there was no difference in quality of
life at 12 months. A review by Michie, Abraham, Whittington,
McAteer, and Gupta (2009) also reported no significant bivariate
association between the use of progress monitoring as an interven-
tion technique and effect sizes obtained in physical activity and
dietary interventions (see Table S5 in the Supplementary Materials
of Michie et al., 2009; http://dx.doi.org/10.1037/a0016136.supp).
Some studies have even found that progress monitoring has det-
rimental outcomes. For instance, Carli et al. (2008) reported that
participants who were asked to monitor their sun exposure using a
UV-meter actually spent more time in the sun, used fewer sun
protective measures, and experienced greater sunburn than partic-
ipants who were not asked to use a UV-meter. In short, the impact
of progress monitoring on goal attainment differs across primary
studies.
Extant reviews of the literature also have not clarified the role of
progress monitoring in goal attainment. Narrative reviews have
been criticized as being subjective, scientifically unsound, and
inefficient (Light & Pillemer, 1984). Furthermore, these reviews
do not permit inferences about the magnitude of the effect of
progress monitoring on outcomes (e.g., Abrahamse et al., 2005;
Korotitsch & Nelson-Gray, 1999). Previous quantitative syntheses
of the impact of progress monitoring exhibit three important lim-
itations. First, some reviews include only a relatively small number
of studies (e.g., Richardson, Newton, Abraham, Sen, Jimbo, &
Swartz, 2008, could locate only nine studies examining the impact
of pedometers on weight loss) or focus on specific contexts (e.g.,
Febbraro & Clum[rquote]s, 1998, review focused on the effects of
self-monitoring on adult problem behaviors), and so preclude
generalizations. Second, previous reviews have not computed the
effect of interventions on the frequency of progress monitoring,
and so we do not know whether interventions designed to promote
progress monitoring actually succeeded in doing so. This is par-
ticularly problematic as many interventions incorporate progress
monitoring alongside other behavior change techniques (BCTs).
For example, Burke, Giangiulio, Gillam, Beilin, and Houghton
(2003) provided participants with a 16-week program designed to
promote physical activity and healthy eating. Completing activity
diaries was just a small part of the larger intervention program
(that, according to Michie et al., 2009, included 14 other BCTs).
Without examining the effects of such interventions on the fre-
quency of progress monitoring, it is difficult to isolate the effects
of progress monitoring on goal attainment. Finally, some reviews
have merely assessed the correlation between progress monitoring
and outcomes. For instance, Michie et al. (2009) regressed effect
sizes on the presence versus absence of a progress monitoring
component in relevant interventions (respectively coded 0 and 1 by
the researchers). Thus, a meta-analytic integration of the experi-
mental evidence is needed to draw firm conclusions about whether,
and to what extent, progress monitoring promotes rates of behav-
ioral performance and goal attainment.
Moderators of the Impact of Progress Monitoring on
Goal Attainment
Several variables could influence the impact of progress moni-
toring on goal attainment. The present review delineates three
broad classes of moderator variables pertaining to the characteris-
tics of the intervention, study methodology, and sample, respec-
tively.
Intervention characteristics. To answer both conceptual and
practical questions about when and how progress monitoring in-
fluences goal attainment, it is important to examine the nature of
progress monitoring prompted by the intervention. Drawing upon
conceptual frameworks for understanding the nature of progress
monitoring (e.g., Anseel, Beatty, Shen, Lievens, & Sackett, 2015;
Ashford & Cummings, 1983; Wilde & Garvin, 2007) and careful
examination of extant interventions designed to promote progress
monitoring (i.e., a combined “top-down” and “bottom-up” ap-
proach, as advocated by Koole, 2009; Skinner, Edge, Altman, &
Sherwood, 2003; Webb, Miles, & Sheeran, 2012), we identified
six key dimensions that could be used to code how goal progress
was monitored in each of the interventions identified in the present
review (see Table 1).
The first dimension is the focus of monitoring, which distin-
guishes between monitoring behavior versus monitoring the out-
comes of behavior (e.g., Michie et al., 2011; Michie et al., 2013).
For example, people seeking to lose weight could keep track of
their snacking behavior, or they could keep track of their weight (a
likely outcome of snacking behavior). We predict that a match
between the focus of monitoring (behavior vs. outcome) and the
dependent variable (behavior vs. outcome) will improve perfor-
mance. Thus, we expect that monitoring behavior (e.g., snack
intake) will have a larger impact on subsequent behavior (e.g., the
number of snacks consumed) than on outcomes (e.g., weight loss),
whereas focusing on outcomes will have a larger impact on sub-
sequent outcomes than on behavior. This is because behavioral
discrepancies are informative about the need to adjust the specific
behavior but may say little about outcomes that are likely deter-
mined by multiple behaviors. Outcome discrepancies, on the other
hand, may suggest the need to increase effort on multiple behav-
iors in order to reach the desired outcome but may say little about
any particular behavior (as substitute behaviors could serve the
same ends; Kruglanski et al., 2002).
The second dimension concerned whether interventions asked
participants to monitor their progress in public or in private.
Protocols that require participants to monitor their progress in
public (e.g., weigh themselves during a weight loss class, Samuel-
Hodge et al., 2009) or to submit reports on their goal progress (e.g.,
step counts, De Cocker, De Bourdeaudhuij, & Cardon, 2008;
diaries of peak flow or symptoms related to asthma, Buist,
Vollmer, Wilson, Frazier, & Hayward, 2006) may engender a
greater sense of public commitment to the goal (Cialdini, 2001;
Kiesler, 1971), accountability (e.g., Stuckey et al., 2011), presen-
tational concerns (Schienker, Dlugolecki, & Doherty, 1994), or
experimenter demand (Zizzo, 2010), each of which could serve to
promote goal attainment.
The third dimension involved whether or not participants were
asked to physically record the information obtained from monitor-
ing (e.g., write the information in a diary). This form of monitoring
has been termed “self-recording” (Korotitsch & Nelson-Gray,
1999). Physical logs, even if kept private, can provide the oppor-
tunity for the person to examine and reflect on their progress
toward the goal over time, and potentially identify actions that
promote or hamper goal progress. We therefore expected that
interventions that prompted participants to physically record the
200 HARKIN ET AL.
information that they obtain from monitoring their goal progress
would obtain larger effects on goal attainment than interventions
that did not have this requirement.
Assessing goal progress involves comparing the current state
with a reference value (Carver & Scheier, 1982, 1990). The effects
of progress monitoring on goal attainment might, therefore, also be
influenced by the nature of the reference value against which
participants evaluate their progress. The fourth dimension of prog-
ress monitoring examined here concerned whether reference val-
ues took the form of (a) a desired target or goal (e.g., target blood
sugar levels, Bell, Fonda, Walker, Schmidt, & Vigersky, 2012); (b)
a reference value in the past (e.g., with respect to previously
abnormal HBA1C levels; Farmer et al., 2007); or (c) comparison
with others (e.g., comparing level of exercise with that of others;
Hurling et al., 2007).
The fifth dimension on which approaches to progress monitor-
ing can differ distinguished between monitoring distance from a
goal versus rate of progress toward the goal. Goal Setting Theory
(Locke & Latham, 1990) proposed that the absolute size of the
discrepancy between current and desired states (i.e., the distance
from the goal) determines subsequent effort. According to Carver
and Scheier (1982, 1990), however, it is not only the absolute
Table 1
Dimensions of Progress Monitoring
Dimension Definition Example
Focus of progress monitoring
Monitor behavior The person monitors their behavior(s) A person uses a pedometer and records the number
of steps that they take
Monitor outcomes The person monitors the outcome(s) of their behavior
(including thoughts and feelings) A person weighs themselves and records their
weight on a graph
Public vs. private monitoring
Public monitoring Progress is monitored in a public context A person weighs themselves at a dieting group, in
front of others who are trying to lose weight
Private monitoring (reported) Progress is monitored privately, but the information
derived from progress monitoring is reported to at
least one other person
A person weighs themselves and telephones a
research assistant to report their weight
Private monitoring (not reported) Progress is monitored privately, and the information
derived from progress monitoring is not reported to
anyone else
A person weighs themselves, but does not report
their weight to anyone
Recording of monitoring
Monitoring is recorded The information obtained from monitoring is physically
recorded A person weighs themselves and writes this
information in their diary
Monitoring is not recorded The information obtained from monitoring is not
recorded in any way A person weighs themselves, but does not record or
report this information
Reference value
a
Past Goal progress is compared with a past state or previous
rate of progress A person compares how much they weigh now, with
how much they weighed previously
Desired (future) target Goal progress is compared with a desired future state
or goal A person compares how much they weigh now, with
how much they would like to weigh
Others Goal progress is compared with others progress or
states (close others or those striving for a similar
goal)
A person compares how much they weigh now, with
how much others around them weigh
Monitor rate of progress vs. distance from the goal
Monitor rate of progress toward
a goal The person monitors their rate of progress toward a
specified goal A person notes that they weigh 1 kg less each week
Monitor distance from the goal The person monitors how far they are away from a
goal or starting point A person notes that they weigh 6 kg more than
desired
Passive vs. active monitoring
Passive monitoring The person attends to information about progress that
can be accessed without deliberate effort; that is,
information that is readily available in the
environment
A person notices that clothes feel looser than before,
recognizes that a number of friends have
commented on their weight loss, or receives text
alerts with their weight
Active monitoring The person makes deliberate efforts to attend to goal-
related behavior, and/or seeks out information about
goal-related outcomes
A person steps on a set of weighing scales or
records the amount of exercise performed
a
Because participants asked to monitor their progress toward a specified goal might evaluate their progress with respect to different reference values (e.g.,
participants asked to walk 10,000 steps per day might compare the number of steps that they took on a particular day with this value or to the number of
steps that they took the previous day), we suggest that the nature of the reference value should only be coded if the intervention explicitly directs participants
to monitor their progress with respect to a particular reference value.
201
PROGRESS MONITORING AND GOAL ATTAINMENT
discrepancy between the current state and the reference value that
matters, but also the rate with which progress is being made (or not
made). For example, a dieter may be a long way from their goal of
losing 30lbs, but if they have lost 4 lbs over the preceding week
then they are likely to feel pleased with their rate of progress,
which could galvanize effort.
The final dimension of progress monitoring examined the dis-
tinction between passive versus active monitoring. This distinction
originated in organizational psychology (e.g., Anseel, Beatty,
Shen, Lievens, & Sackett, 2015; Ashford & Cummings, 1983) and
the literature on information seeking (e.g., Berger, 2002). Passive
monitoring involves obtaining information about progress without
having to make deliberate efforts to seek out and scrutinize that
information. For example, passive monitoring of progress toward
a weight loss goal could involve noticing that clothes feel looser
than before (Chambers & Swanson, 2012), or realizing that friends
have commented on how slim one looks. In contrast, active mon-
itoring involves actively seeking out and attending to information
about goal progress (e.g., deliberately weighing oneself).
In addition to the six dimensions of progress monitoring out-
lined in Table 1, we also examined the method used to promote
progress monitoring. For example, Acharya, Elci, Sereika, Styn,
and Burke (2011) compared the effects of monitoring dietary and
exercise behavior using a personal digital assistant (PDA) versus a
written diary and observed no significant differences in weight
loss. Given the range of methods that can be used to monitor goal
progress and the increasing availability of technology to support
self-monitoring (Conroy, Yang, & Maher, 2014), it is important to
compare the effects of different methods on both the frequency of
progress monitoring and goal attainment.
We also considered whether the source and the duration of the
intervention influenced effect sizes. Research on persuasion sug-
gests that the source of the message has an important influence on
its impact (Chaiken, 1980; Maddux & Rogers, 1980, see Wilson &
Sherrell, 1993, for a review). It is possible that interventions
designed to promote progress monitoring that are delivered by
credible or expert sources (e.g., health professionals) improve
adherence and goal attainment compared with interventions deliv-
ered by other parties (e.g., researchers; for reviews of the effects of
source credibility, see Eisend, 2004; Kumkale & Albarracin, 2004;
Latimer, Brawley, & Bassett, 2010; Pornpitakpan, 2004). It is less
clear how the duration of the intervention might be expected to
influence effect sizes. On the one hand, monitoring progress over
a longer period of time could provide more extensive and useful
information, and afford greater opportunity to change behavior and
outcomes. Thus, we might expect a “dose-response” relationship
such that the length of time over which participants monitor their
goal progress is related to the size of the change in goal attainment.
On the other hand, monitoring progress may become less infor-
mative over time, or the person may habituate to the information
obtained from progress monitoring (Ashford & Cummings, 1983;
Webb, Chang, & Benn, 2013). Therefore, longer periods of mon-
itoring may not confer additional benefit.
The final intervention characteristic concerns the use of addi-
tional BCTs alongside progress monitoring (Abraham & Michie,
2008). Michie et al. (2009) found that combining progress moni-
toring with one of four other BCTs (intention formation, specific
goal setting, feedback on performance, and review of behavioral
goals) was associated with larger effects of interventions on phys-
ical activity and diet. Similar findings have been reported in
systematic reviews of behavioral interventions for weight control
(Dombrowski et al., 2012), physical activity and healthy eating
(Greaves et al., 2011), and problem behaviors (Febbraro & Clum,
1998). The present meta-analysis therefore coded the use of addi-
tional BCTs alongside monitoring of behavior and/or outcomes.
We also separated immediate from delayed feedback (on behavior
vs. outcomes, respectively) because previous research has shown
that immediate feedback is more beneficial than delayed feedback
(e.g., in learning contexts, Dihoff, Brosvic, Epstein, & Cook, 2004;
Opitz, Ferdinand, & Mecklinger, 2011).
Methodological characteristics. The second category of
moderator variables relates to methodological characteristics of the
primary studies, and includes (a) The nature of the comparison
group; (b) the nature of the focal goal; (c) how key variables were
measured (e.g., self-report vs. objective assessment); (d) study
quality; (e) publication status; and (f) participant characteristics.
Progress monitoring interventions have been compared with con-
trol conditions that do not involve monitoring (e.g., Spence et al.,
2009, asked participants in the intervention condition, but not the
control condition, to return a diary of their step counts as recorded
by a pedometer), control conditions where participants monitor
their progress but in a different manner to the treatment condition
(e.g., Beasley et al., 2008, compared the use of PDAs with paper
diaries for monitoring food intake), and control conditions where
participants monitor their progress to a lesser extent than do
participants in the treatment condition (e.g., Gokee LaRose, Gorin,
& Wing, 2009, had participants in the control condition weigh
themselves only once a week, whereas participants in the treatment
were asked to weigh themselves daily). Reviews in other domains
have shown smaller effects for interventions when compared with
active control conditions than when compared with passive control
conditions (e.g., Portnoy, Scott-Sheldon, Johnson, & Carey, 2008),
and so we expected that the effect of interventions designed to
promote progress monitoring would be larger if the control con-
ditions did not involve progress monitoring.
We also anticipated that the effects of progress monitoring
might vary for different goals. For example, self-monitoring of
blood glucose levels could be an effective way to manage diabetes
(Allemann, Houriet, Diem, & Stettler, 2009; Coster, Gulliford,
Seed, Powrie, & Swaminathan, 2000), while self-weighing may
have a smaller impact on weight loss (Burke, Wang, & Sevick,
2011; VanWormer, French, Pereira, & Welsh, 2008). The meth-
odological rigor of the primary studies might also influence the
validity of estimated effect sizes (Juni, Altman, & Matthias, 2001;
Moher et al., 1999; Moja et al., 2005; Oxman & Guyatt, 1988). We
therefore rated aspects of study quality including participant blind-
ing, experimenter blinding, and the type, quality, and success of
randomization. As unpublished studies may use less rigorous
methods than published studies, publication status was also coded.
Finally, the type of sample (e.g., general public vs. people with
particular health conditions), and the mean age and gender com-
position of the sample was also coded.
The Present Review
The foregoing discussion indicates that progress monitoring
constitutes a key component of Control Theory and other leading
models of goal-directed behavior, and is a crucial process that
202 HARKIN ET AL.
intervenes between intention formation and goal attainment. How-
ever, despite the conceptual significance of progress monitoring,
and its increasing deployment as a technique for promoting be-
havior change, empirical evidence concerning the role of progress
monitoring is equivocal. A meta-analytic review is needed to
quantify the impact of progress monitoring on rates of behavioral
performance and goal attainment.
The present meta-analysis includes only studies that randomly
allocated participants to a treatment condition designed to promote
progress monitoring versus a control condition. The review as-
sesses the impact of interventions on both the frequency of prog-
ress monitoring and rates of goal attainment. We also test the
predictions that (a) The effect of the interventions on goal attain-
ment are mediated by changes in progress monitoring, and (b)
intervention effects on outcomes are mediated by changes in
behavior among participants who monitored their goal progress.
Finally, we assess whether dimensions of progress monitoring and
other intervention, methodological, and sample characteristics in-
fluence effect sizes.
Method
Selection of Studies
There were three inclusion criteria for the review. First, studies
had to randomly assign (adult, human) participants to a treatment
condition that received an intervention designed to promote mon-
itoring of goal progress, or a control condition that received either
an active, comparison intervention (e.g., an intervention that also
prompted progress monitoring, but in a different manner or to a
lesser degree than the treatment condition) or a no-intervention
group (e.g., a waiting list control group). Interventions were
deemed to have prompted progress monitoring if participants were
invited to monitor their behavior (e.g., to use a pedometer) and/or
the outcomes of their behavior (e.g., to weigh themselves). Second,
studies had to measure the frequency of progress monitoring
following the intervention. Finally, studies had to include a mea-
sure of behavior(s) (e.g., step count) or outcome(s) (e.g., levels of
glycated haemoglobin [HbA1c], weight) in the wake of the inter-
vention.
1
The sample of studies was generated via a computerized search
of social scientific databases (those accessed by Web of Knowl-
edge,
2
as well as UMI Dissertation Abstracts). Three search filters
were used, one for randomized control trials (random
AND in-
tervention or random
AND experiment
),
3
one for progress mon-
itoring designed to reflect different terms for self-monitoring and
methods that interventions might use to invoke progress mon-
itoring (monitor
OR progress OR track OR diary OR website
OR Personal Digital Assistant OR Phone OR pedometer OR
meter OR self-weigh
), and one filter for the dependent variable
(goal OR behav
OR outcome OR perform
OR consum
). Articles
had to include at least one term from each of the filters in the title,
abstract, or keywords. We also (a) considered all of the articles that
cited Carver and Scheier’s (1982) article on Control Theory; (b)
searched the reference lists of reviews of self-monitoring in other
domains (e.g., proenvironmental behavior, Abrahamse et al., 2005;
clinical practice, Korotitsch & Nelson-Gray, 1999; educational,
Webber, Scheuermann, McCall, & Coleman, 1993; and organiza-
tional settings, Anseel et al., 2015; Ashford, 2003); (c) examined
the reference lists in each article that was deemed suitable for
inclusion (ancestry approach; Johnson, 1993); and (d) sent e-mails
requesting published and unpublished data to the distribution lists
of the European Association of Social Psychology,European
Health Psychology Society,Society for Personality and Social
Psychology, and the British Psychological Society (Social Psychol-
ogy Section).
Figure 1 shows the flow of information through the review. Of
the 22,054 articles that were initially identified, 21,556 were
obtained from the database search and 498 from citations of Carver
and Scheier (1982); 9,753 duplicates were removed. During initial
screening, the title, abstract, and keywords were considered. The
majority of articles rejected at this stage did not randomly assign
participants to conditions. For example, Poirier and Cobb (2012)
examined engagement with a web-based intervention. However,
the impact of engagement (on the frequency of progress monitor-
ing and goal attainment) was not evaluated with respect to a
control condition, which did not receive the intervention. Of the
12,301 articles screened in this manner, 636 studies were identified
as potentially eligible for inclusion. These studies were then eval-
uated in detail. Two-hundred and 19 studies (34%) were excluded
because they did not include a measure of progress monitoring
following the intervention (e.g., Gillis, Lumley, Mosley-Williams,
Leisen, & Roehrs, 2006; Levy, Xu, Daly, & Ely, 2013). A further
141 studies (22%) did include a measure, but did not report
sufficient data for us to be able to compute an effect size and this
information could not be obtained by e-mailing the author(s) (e.g.,
Reijonsaari et al., 2012). Eighty-one studies (13%) were excluded
because they did not randomly assign adult participants to an
intervention condition designed to promote self-monitoring of goal
progress (e.g., Graham et al., 2013; Te Velde, Wind, Perez-
Rodrigo, Klepp, & Brug, 2008). Nineteen studies (3%) were
excluded as they outlined a method for a future study (e.g.,
described the protocol for an RCT; e.g., Focht et al., 2012; Ma et
al., 2013). Seventeen studies (3%) were excluded because they
reported additional effects of data already included in the review
(e.g., Farmer et al., 2009, and French, Wade, Yudkin, Neil, Kin-
month, & Farmer, 2008, reported findings from the same dataset as
1
The measures of progress monitoring and goal attainment needed to be
empirically distinct. For example, studies that prompted participants to
monitor their blood glucose levels and then used frequency of monitoring
as a measure of goal attainment were not suitable for inclusion. However,
studies that prompted participants to monitor their blood glucose levels
were included if an independent outcome measure like blood glucose levels
(e.g., HbA1c levels) was used as a measure of goal attainment.
2
Databases include the Science Citation Index Expanded (1900
present), Social Sciences Citation Index (1956–present), Arts & Humani-
ties Citation Index (1975–present), Conference Proceedings Citation In-
dex—for Science and for Social Science & Humanities (1990–present),
Book Citation Index—for Science and for Social Science & Humanities
(2005–present), Current Chemical Reactions (1985–present), Index
Chemicus (1993–present), BIOSIS Citation Index (1926–present), BIOSIS
Previews (1969–present), Current Contents Connect (1998-present), Data
Citation Index (1900–present), Derwent Innovations Index (1963–present),
MEDLINE (1950–present), SciELO Citation Index (1997–present), and
the Zoological Record (1864–present).
3
It was not possible to use random OR intervention/experiment
as
recommended by Haynes, McKibbon, Wilczynski, Walter, and Werre
(2005) as this combination of search terms produced over half a million
records in Web of Science, even when combined with the other search
filters.
203
PROGRESS MONITORING AND GOAL ATTAINMENT
Farmer et al., 2007). Nine studies (1%) were excluded because
they focused on children (e.g., Belzer et al., 2014; Brown, Dunn,
& Budney, 2014), and five studies (1%) were excluded as they did
not measure goal attainment in the wake of the intervention (e.g.,
Olson, Schmidt, Winkler, & Wipfli, 2011). Finally, we rejected
four duplicate studies, along with two studies where the measure of
progress monitoring was not sufficiently distinct from the measure
of goal attainment (Carr et al., 2013; Williams et al., 2006) and one
study that was not written in English (Wang, Kueffer, Wang, &
Maercker, 2014). Table 2 presents the characteristics and effect
sizes for each included study. (An asterisk precedes each of these
articles in the reference list.)
Data Extraction
Coding of study characteristics. For each study, we coded
the following methodological characteristics: (a) bibliographic in-
formation (e.g., publication status); (b) the nature of the focal
behavior or outcome (e.g., weight, HbA1c levels, steps taken); (c)
the nature of the measures of progress monitoring and goal attain-
ment (i.e., self-report or objective); and (d) aspects of study quality
as defined by Chalmers et al. 1990; e.g., participant and experi-
menter blinding, randomization success, method of randomization,
quality of randomization). We also coded the following character-
istics of the focal sample: (a) the type of sample (i.e., general
public, university students, specific sample); (b) the average age of
participants in the treatment condition; and (c) the proportion of
females in the treatment condition. Finally, we coded the following
characteristics of each intervention: (a) whether participants were
prompted to monitor behavior or the outcomes of behavior; (b)
whether progress was monitored in public or in private (the latter
category was further divided into monitoring in private and not
reported vs. monitoring in private and the information was re-
IDENTIFICATION
SCREENING
ELIGIBILITY
INCLUDED
Records identified through
database search
(n = 21,556)
Records that cited
Carver & Scheier
(n = 498)
Records after duplicates (n = 9,753) removed (n = 12,301)
Records screened
(n = 12,301)
Records excluded
(n = 11,665)
Studies
deemed
potentially
eligible for
inclusion
(n = 636)
of progress monitoring following the
measure of progress monitoring they did not
81 rejected as they did not randomly assign
participants to an intervention designed to
promote self-monitoring of goal progress
1 rejected as not in English
138 studies
included in
meta-analysis
Figure 1. Flow of information through the review.
204 HARKIN ET AL.
Table 2
Effect Sizes for Progress Monitoring and Goal Attainment for Studies Included in the Meta-Analysis
Study Progress monitoring
method Focal behavior/outcome N
E
N
C
Effect size (d)
PM GA
Abrahams et al. (2010) Diary Prophylaxis use 125 128 .27
a
.15
c
Abraira et al. (1995) SMBG BG 75 78 2.41
ⴱⴱⴱa
.47
ⴱⴱc
Acharya et al. (2011) PDA Weight 129 62 .66
ⴱⴱⴱa
.12
de
Akers et al. (2012) Tracking sheets Weight, diet 18 21 .13
ⴱⴱⴱa
.02
d
Allen et al. (2013) Comp. 1
f
Phone Weight 11 4 2.76
ⴱⴱⴱb
.45
d
Allen et al. (2013) Comp. 2
g
Phone Weight 10 4 2.71
ⴱⴱb
.42
d
Allen et al. (2013) Comp. 3
h
Phone Weight 10 4 2.02
ⴱⴱb
.10
d
Amsberg et al. (2009) Blood sugar BG 36 38 .70
ⴱⴱa
.47
c
An et al. (2006) Website Smoking 257 260 10.72
ⴱⴱⴱb
.45
ⴱⴱⴱc
Anderson et al. (2011) Diary Weight 18 13 4.31
ⴱⴱⴱb
3.21
ⴱⴱⴱd
Andrews et al. (2011) Pedometer BG 240 246 4.11
ⴱⴱⴱb
.02
de
Antypas & Wangberg (2014) Website PA 27 37 .17
a
.84
c
Arbour & Martin Ginis (2008) Log book PA 25 17 .10
a
.50
de
Aronson (2006) Diary Medication adherence 19 19 .61
a
.11
c
Atienza et al. (2008) PDA Diet 16 11 2.95
ⴱⴱⴱb
.51
d
Beasley et al. (2008) Diary Weight 71 78 .00
a
.25
d
Bell et al. (2012) Log/diary BG 31 33 .36
a
.32
d
Berg et al. (1997) Diary Asthma 31 24 4.32
ⴱⴱⴱa
.12
de
Blasco et al. (2012) Internet CAD 87 83 4.07
ⴱⴱⴱb
.09
d
Boutelle et al. (1999) Diary Weight 26 31 .47
a
1.64
Brindal et al. (2013) Phone Weight 21 23 .95
ⴱⴱa
.37
de
Buist et al. (2006) Diary Asthma 149 147 .00
a
.16
de
Caldwell et al. (2005) Diary Heart failure 20 16 .62
a
1.01
ⴱⴱde
Carli et al. (2008) UV meter Sun protection 46 40 3.39
ⴱⴱⴱb
.51
c
Carter et al. (2013) Comp. 1
i
Phone Weight, diet, PA 40 20 .94
ⴱⴱa
.30
d
Carter et al. (2013) Comp. 2
j
Phone Weight, diet, PA 27 20 .14
a
.33
d
Chambliss et al. (2011) PDA Weight 34 33 .59
a
.08
c
Chao (2010) Step log PA 20 20 3.52
ⴱⴱⴱb
.27
d
Chau et al. (2012) PDA COPD 22 18 3.64
ⴱⴱⴱb
.05
d
Cho et al. (2006) SMBG BG 35 36 .50
a
.82
ⴱⴱde
Clarke et al. (2009) Website Depression 58 58 2.58
ⴱⴱⴱb
.81
ⴱⴱⴱde
Coughlin et al. (2013) Comp. 1
k
Diary Diet 292 144 .26
ⴱⴱa
.13
d
Coughlin et al. (2013) Comp. 2
l
Diary Diet 301 144 .38
ⴱⴱⴱa
.27
ⴱⴱdl
Cussler et al. (2008) Website Weight, diet, PA 38 40 4.17
ⴱⴱⴱb
.01
c
D’Eram (1987) Comp. 1
m
SMBG Weight 15 6 .42
a
.54
de
D’Eramo (1987) Comp. 2
n
SMBG Weight 19 6 .51
a
.53
de
De Blok et al. (2006) Diary PA 8 8 4.61
ⴱⴱⴱb
1.50
ⴱⴱde
De Cocker et al. (2008) Pedometer PA 51 52 .38
a
.11
de
De Cocker et al. (2012) Diary PA 32 37 .19
a
.51
d
Dennis et al. (2012) Website Weight, diet, PA 18 21 3.39
ⴱⴱⴱb
.06
de
Dennison et al. (2014) Comp. 1
o
Website Weight, diet, PA 247 138 2.35
ⴱⴱⴱb
.91
ⴱⴱⴱd
Dennison et al. (2014) Comp. 2
p
Website Weight, diet, PA 264 138 2.09
ⴱⴱⴱb
.77
ⴱⴱd
DeWalt et al. (2006) Diary Healthcare use 52 59 1.22
ⴱⴱⴱa
.19
c
Domingo et al. (2011) Website/television Healthcare use 44 42 3.48
ⴱⴱⴱb
.11
c
Duran et al. (2010) SMBG BG 99 62 2.28
ⴱⴱⴱa
2.70
ⴱⴱⴱde
Farmer et al. (2007) Comp. 1
q
SMBP BG 150 76 3.31
ⴱⴱⴱb
.16
dq
Farmer et al. (2007) Comp. 2
r
SMBP BG 151 76 2.97
ⴱⴱⴱb
.20
dr
Gajecki et al. (2014) Phone Alcohol 341 489 3.50
ⴱⴱⴱb
.19
de
Gokee LaRose et al. (2010) Diary Weight 21 23 .96
ⴱⴱⴱb
1.22
ⴱⴱd
Gokee LaRose et al. (2009) Digital scales Weight 20 17 1.12
ⴱⴱⴱa
.10
d
Gold et al. (2007) Website Diet 51 50 .58
ⴱⴱⴱa
.46
d
Goto et al. (2014) Phone PA, blood coagulation 16 16 4.11
ⴱⴱⴱb
.05
de
Goulis et al. (2004) Phone Weight, BP, Physiol. 45 77 3.08
ⴱⴱⴱb
.57
ⴱⴱde
Haapala et al. (2009) Phone Weight, diet, PA 45 40 2.43
ⴱⴱⴱb
.42
de
Haddock et al. (2014) Website Weight, diet, PA 229 253 .91
ⴱⴱⴱa
.25
ⴱⴱd
Hannum et al. (2004) Diary Weight, diet 26 27 .16
a
.38
d
Hellerstedt & Jeffrey (1997) Comp. 1
s
Phone Weight 20 11 4.32
ⴱⴱⴱb
1.48
ⴱⴱd
Hellerstedt & Jeffrey (1997) Comp. 2
t
Phone Diet, PA 17 11 3.70
ⴱⴱⴱb
.83
d
Helsel et al. (2007) Diary Weight 21 21 .23
a
.07
d
Homko et al. (2012) Website/phone BG 40 40 .04
a
.21
c
(table continues)
205
PROGRESS MONITORING AND GOAL ATTAINMENT
Table 2 (continued)
Study Progress monitoring
method Focal behavior/outcome N
E
N
C
Effect size (d)
PM GA
Hurling et al. (2007) Website Weight, PA 47 30 3.71
ⴱⴱⴱb
2.86
ⴱⴱⴱd
Hyman et al. (1998) Diary Cholesterol 65 58 3.44
ⴱⴱⴱb
.10
d
Janson et al. (2003) Diary Asthma 31 27 .09
a
.72
ⴱⴱde
Janson et al. (2009) Diaries Asthma 45 39 .17
a
.12
d
Jefferson (2005) Diary Mood, weight 21 29 3.82
ⴱⴱⴱb
.09
de
Jennings et al. (2014) Website PA 77 77 3.79
ⴱⴱⴱb
.06
de
Jurgens et al. (2013) Diary Heart failure 48 51 .77
ⴱⴱⴱa
.13
c
Kempf et al. (2013) Diary Diabetes 62 60 2.75
ⴱⴱⴱb
.43
de
Kim et al. (2012) Website Weight, BG 19 23 3.73
ⴱⴱⴱb
.29
d
King et al. (2008) PDA PA 19 18 3.34
ⴱⴱⴱb
.71
c
Kirwan et al. (2013) Diary Diabetes 32 36 .43
a
3.78
ⴱⴱⴱd
Kobulnicky (2002) Diary Effects of chemotherapy 42 29 3.95
ⴱⴱⴱb
.05
d
Kraschnewski et al. (2011) Website Weight, diet, PA 43 45 2.16
ⴱⴱⴱb
.28
d
Kristal et al. (2000) Diary Diet 601 604 2.87
ⴱⴱⴱb
.21
ⴱⴱⴱd
Kroenke et al. (2010) Phone Depression 202 203 4.50
ⴱⴱⴱb
.54
ⴱⴱⴱde
Kwon et al. (2004) SMBG BG 51 50 5.23
ⴱⴱⴱa
.89
ⴱⴱⴱde
Ligibel et al. (2012) Diary PA 48 51 2.80
ⴱⴱⴱb
.18
d
Linde & Jeffrey (2011) Diary Weight, diet, PA 22 26 1.16
ⴱⴱⴱa
.20
d
Logan et al. (2012) SMBP BP 55 55 3.26
ⴱⴱⴱb
.38
d
Maljanian et al. (2005) Diary Diabetes 181 162 .13
a
.15
de
Marquez-Contreras et al. (2006) MEMS BP 100 100 1.11
ⴱⴱⴱa
.24
de
Maruyama et al. (2010) Website PA 48 39 .61
ⴱⴱa
.44
c
McKinstry et al. (2013) PDA BP 182 177 4.17
ⴱⴱⴱb
.41
ⴱⴱⴱd
McManus et al. (2010) SMBP BP 234 246 3.49
ⴱⴱⴱb
3.80
ⴱⴱⴱd
McMurdo et al. (2010) Comp. 1
u
Diary PA 60 33 4.61
ⴱⴱⴱb
1.32
ⴱⴱⴱd
McMurdo et al. (2010) Comp. 2
v
Diary PA 53 33 4.50
ⴱⴱⴱb
3.31
ⴱⴱⴱd
Mehos et al. (2000) SMBP BP 18 18 4.00
ⴱⴱⴱb
.56
d
Moreland et al. (2006) Comp. 1
w
SMBG BG 50 49 .44
a
.14
d
Moreland et al. (2006) Comp. 2
x
SMBG BG 50 49 .07
a
.07
d
Morgan et al. (2009) Website Weight, PA 24 31 .51
b
1.21
ⴱⴱⴱd
Muchmore et al. (1994) SMBG BG 12 11 3.61
ⴱⴱⴱb
1.86
ⴱⴱⴱde
Nanchahal et al. (2009) Pedometer Weight 48 55 3.24
ⴱⴱⴱb
.07
d
Nguyen et al. (2009) Phone PA 9 8 .68
a
.54
de
O’Kane et al. (2008) SMBG BG 96 88 3.68
ⴱⴱⴱb
.09
de
Ornes (2006) Diary PA 30 29 3.41
ⴱⴱⴱb
.73
ⴱⴱde
Orsama et al. (2013) PDA Blood glucose 24 24 3.02
ⴱⴱⴱb
.47
d
Oshima et al. (2013) PDA Weight 28 28 1.51
ⴱⴱⴱa
.30
de
Pellegrini et al. (2012) Website Weight, diet, PA 17 13 .29
a
.13
d
Petersen et al. (2012) Pedometer PA 192 173 3.87
ⴱⴱⴱb
.07
d
Petrella et al. (2014) Phone PA 67 60 3.37
ⴱⴱⴱb
.04
de
Phelan et al. (2014) Phone Weight, diet, PA 128 133 .28
a
.16
d
Piette et al. (2011) Diary Physical activity 145 146 .47
ⴱⴱⴱa
.40
ⴱⴱde
Polonsky et al. (2011) SMBG BG 188 187 .37
ⴱⴱⴱa
.33
ⴱⴱd
Polzien et al. (2007) Diary Weight, diet, PA 16 16 .58
a
.07
d
Proudfoot et al. (2013) Website Depression 126 185 1.14
ⴱⴱⴱb
.24
de
Quinn et al. (2008) Log Book Diabetes 13 13 4.26
ⴱⴱⴱa
.72
c
Ralston et al. (2014) Diary/Survey BP 186 197 .29
ⴱⴱa
.29
ⴱⴱd
Raynor et al. (2012) Diary Weight 94 96 .00
a
.42
ⴱⴱc
Richardson et al. (2010) Website PA 254 70 .44
ⴱⴱa
.16
d
Rosal et al. (2011) SMBG BG 124 128 .58
ⴱⴱⴱa
.19
d
Rosal et al. (2005) SMBG BG 15 10 .81
a
1.04
d
Rote (2013) Website PA 27 26 .41
a
1.34
ⴱⴱⴱde
Runyan et al. (2013) Phone Time management 41 20 2.70
ⴱⴱⴱb
.68
c
Samuel-Hodge et al. (2009) Diary Weight 64 62 3.64
ⴱⴱⴱb
.38
d
Sengpiel et al. (2010) PDA Lung function 28 28 .30
a
.23
c
Seto et al. (2012) Phone Heart function 44 50 3.82
ⴱⴱⴱb
.05
de
Shapiro et al. (2012) Phone Weight, PA 64 79 3.17
ⴱⴱⴱb
.19
d
Sheldon (1996) Diary Diet 8 6 1.55
a
.45
de
Sherwood et al. (2013) Diary/Survey Weight 178 186 .45
ⴱⴱⴱa
.16
d
Smith et al. (1997) Diary Weight, diet, BG 6 10 1.29
a
.65
d
Spence et al. (2009) Log sheets Physical activity 16 16 3.52
ⴱⴱⴱb
1.07
ⴱⴱde
Steinberg et al. (2013) Phone Weight, diet, PA 45 44 3.82
ⴱⴱⴱa
1.60
ⴱⴱⴱd
Suffoletto et al. (2012) Phone Antibiotic adherence 72 72 3.31
ⴱⴱⴱb
.27
c
Suffoletto et al. (2013) Phone Symptom assessment 14 22 4.06
ⴱⴱⴱb
.38
d
206 HARKIN ET AL.
ported to at least one other person); (c) whether the information
obtained from monitoring was physically recorded or not; (d) the
nature of the reference value against which the information derived
from monitoring was compared (e.g., past performance, a desired
target, or others’ performance); (e) whether participants were
prompted to monitor their rate of progress or their distance from
the reference value; (f) whether monitoring was active or passive;
(g) the method used to promote progress monitoring (e.g., diary,
personal digital assistant, pedometer); (h) the source of the inter-
vention (i.e., health professionals, researchers/study intervention-
ists, or a mixed team); and (i) the duration of the intervention (in
days).
We also coded whether the interventions included any BCTs in
addition to progress monitoring. We coded for the presence versus
absence of eight BCTs identified by Michie et al. (2009) using the
definitions provided by Michie et al. (2013): (a) goal setting
(behavior); (b) goal setting (outcome); (c) review of behavioral
goals; (d) review of outcome goals; (e) action planning; (f) prompt
identification of a discrepancy between current behavior and goal;
(g) feedback on behavior (immediate vs. delayed); and (h) feed-
back on outcomes (immediate vs. delayed). Immediate feedback
was defined as that provided immediately following the perfor-
mance of a behavior. Where there was a gap between the behavior
and the feedback, the feedback was defined as delayed (e.g.,
participants posted information on their dietary behaviors to which
a dietitian returned handwritten feedback). Only BCTs that dif-
fered between the treatment and control conditions (and so could
account for differences between the conditions) were coded.
All of the studies were coded by the first and third authors.
There was a high level of agreement (for categorical characteris-
tics, median ␬⫽0.95, range 0.48 to 1.00; for continuous
characteristics, median r.99, range 0.94 to 1.00) and dis-
agreements were resolved jointly by discussion.
Computing effect sizes. Effect sizes (representing the effect
of interventions on the frequency of progress monitoring and
behavior and/or outcomes) were calculated as the standardized
mean difference between the treatment and comparison conditions
divided by their pooled standard deviation (Hedges & Olkin,
1985). Whenever possible, effect sizes were calculated using the
means and standard deviations. However, if the means and stan-
dard deviations were not reported, then the metric that was avail-
able (e.g., Fratio, chi-square) was converted to an effect size.
When effect sizes could not be computed precisely on the basis of
information in the report or correspondence with authors (10
effects on progress monitoring, 7%; 13 effects on goal attainment,
9%), then we estimated values based on the significance levels
Table 2 (continued)
Study Progress monitoring
method Focal behavior/outcome N
E
N
C
Effect size (d)
PM GA
Sugden et al. (2008) Diary PA 27 18 .88
ⴱⴱa
.04
d
Talbot et al. (2003) Pedometer PA 17 17 3.55
ⴱⴱⴱb
.29
de
Tan et al. (2011) SMBG BG 82 82 14.46
ⴱⴱⴱa
.49
ⴱⴱc
Tate et al. (2001) Website Weight 33 32 1.09
ⴱⴱⴱa
.65
ⴱⴱd
Thorndike et al. (2012) Website Weight 145 130 3.36
ⴱⴱⴱb
.16
d
Turner-McGrievy & Tate (2011) Phone Weight 42 45 .22
a
.05
d
Van der Meer et al. (2009) Website Asthma 91 92 2.46
ⴱⴱⴱb
3.80
ⴱⴱⴱd
Wang et al. (2012) PDA Weight 59 60 .29
a
.46
d
Webber et al. (2008) Diary Weight 33 33 .03
a
.31
d
Wing et al. (2006) Comp. 1
y
Phone Weight 103 49 3.82
ⴱⴱⴱb
.37
d
Wing et al. (2006) Comp. 2
z
Website Weight 100 49 3.75
ⴱⴱⴱb
.04
d
Wing et al. (2010) Website Weight 74 76 .16
a
.61
d
Wing et al. (1996) Study 1 Diary Weight 23 27 3.51
ⴱⴱⴱb
.44
d
Young, Medic, & Starkes (2009) Swim log PA 15 11 .71
a
1.00
ⴱⴱⴱde
Note.N
E
number of participants in treatment group; N
C
number of participants in comparison group; Comp. comparison; PM progress
monitoring; GA goal attainment; Monitor BG monitoring of blood glucose; Monitor BP monitoring of blood pressure; PA physical activity;
BG blood glucose; BP blood pressure; PDA personal digital assistant; MEMS medication event monitoring system (a product developed by the
Aardex Group); Physiol. physiological measure(s) (e.g., cholesterol, HDL).
a
Effect size calculated by comparing the frequency of progress monitoring in the treatment group and comparison conditions.
b
Effect size calculated by
comparing the frequency of progress monitoring in the treatment group to zero (i.e., studies where the frequency of progress monitoring was not reported
for the comparison condition).
c
Effect size calculated using follow-up measures.
d
Effect size calculated using change scores from baseline.
e
Effect
size calculated by converting a follow-up measure to change score.
f
Comparison 1 from Allen et al. (2013): Intensive counseling smartphone vs.
intensive counseling.
g
Comparison 2 from Allen et al. (2013): Less intensive counseling smartphone vs. intensive counseling.
h
Comparison 3 from
Allen et al. (2013): Smartphone vs. intensive counseling.
i
Comparison 1 from Carter et al. (2013): Smartphone vs. diary.
j
Comparison 2 from Carter
et al. (2013): Website vs. diary.
k
Comparison 1 from Coughlin et al. (2013): Personal contact vs. self-directed.
l
Comparison 2 from Coughlin et al.
(2013): Interactive technology vs. self-directed.
m
Comparison 1 from D’Eramo (1987): Diabetes skills instruction 11 week diabetes education vs. skills
instruction.
n
Comparison 2 from D’Eramo (1987): Diabetes skills instruction 11 week diabetes education follow-up counseling vs. skills
instruction.
o
Comparison 1 from Dennison et al. (2014): Power coaching vs. control.
p
Comparison 2 from Dennison et al. (2014): Power only vs.
control.
q
Comparison 1 from Farmer et al. (2007): Less intensive blood glucose monitoring vs. control.
r
Comparison 2 from Farmer et al. (2007): More
intensive blood glucose monitoring vs. control.
s
Comparison 1 from Hellerstedt and Jeffrey (1997): Weight focused phone group vs. minimal
contact.
t
Comparison 2 from Hellerstedt and Jeffrey (1997): Behavior focused phone group vs. minimal contact.
u
Comparison 1 from McMurdo et al.
(2010): Behavior change pedometer vs. usual care.
v
Comparison 2 from McMurdo et al. (2010): Behavior change vs. usual care.
w
Comparison 1
from Moreland et al. (2006): Blood glucose monitoring manual vs. usual care.
x
Comparison 2 from Moreland et al. (2006): Blood glucose monitoring
vs. usual care.
y
Comparison 1 from Wing et al. (2006): Face to face vs. newsletter control.
z
Comparison 2 from Wing et al. (2006): Internet vs.
newsletter control.
207
PROGRESS MONITORING AND GOAL ATTAINMENT
reported. For example, if the effect was nonsignificant, then we
assumed zero difference (d0.00). If the effect was significant at
p.05, then we used the smallest value of d(given the sample
size) that was significant at this level of alpha.
4
When multiple intervention conditions used the same method to
promote progress monitoring, and there were no differences in the
frequency of progress monitoring between the conditions, the
conditions were combined and compared to relevant comparison
condition(s) (three studies or 2%, e.g., Nanchahal et al., 2009).
Where there were multiple intervention conditions that used the
same method to promote progress monitoring, but differed in the
frequency of progress monitoring (10 studies or 7%, e.g., Heller-
stedt & Jeffery, 1997), the intervention conditions were treated as
separate tests and the sample size for the comparison condition
was divided by the number of intervention conditions (as recom-
mended by Higgins & Green, 2011). When there were multiple
comparison conditions (e.g., in Andrews et al., 2011, either usual
care or the intervention without a pedometer could be treated as the
comparison condition), we selected the comparison condition that
most closely matched the treatment condition, in an effort to
isolate the effect of progress monitoring (six studies or 4%). If
studies did not clearly define which conditions were the treatment
versus control (e.g., Pellegrini, Verba, Otto, Helsel, Davis, &
Jakicic, 2012, compared standard behavioral weight loss, a
technology-based system, and a combined intervention), we pri-
oritized conditions for which there was information on the fre-
quency of progress monitoring, and treated the condition that
reported the most frequent progress monitoring as the treatment
condition, and the condition that reported the least frequent prog-
ress monitoring as the comparison condition (seven studies or 5%).
This strategy was designed to maximize our ability to test the
effect of changes in the frequency of progress monitoring on goal
attainment.
When there were multiple measures of behavior and/or out-
comes, effect sizes were computed separately for each measure
and then meta-analyzed in their own right before inclusion in the
main dataset. Where studies reported baseline and follow-up mea-
sures of behavior or outcomes, we computed effect sizes based on
change scores (83 studies or 60%). If change scores were not
reported (45 studies or 33%), then follow-up scores were con-
verted to change scores, using the method described by Higgins
and Green (2008).
Meta-Analytic Strategy
Effect size computations were undertaken using STATA version
11 and the revised metan command (StataCorp, 2009). This pro-
vided effect sizes, weighted by sample size, with a 95% confidence
interval, and an estimate of heterogeneity. A random effects model
was employed as we expected that effect sizes from the primary
studies were likely to be too complex to be accurately captured by
a few study factors (Cooper, 1986). Three studies used cluster
randomization and effect sizes were corrected using the procedures
described by Higgins and Green (2011).
5
Outlying effect sizes
(defined as effect sizes that were three standard deviations larger
or smaller than the mean) were winsorized and replaced with the
next most extreme value (Dixon, 1960; Tukey, 1962). Following
Cohen’s (1992) recommendations, d0.20 was taken to represent
a “small” effect size; d0.50 a “medium” effect size; and d
0.80 a “large” effect size and we used these qualitative indexes to
interpret the findings.
Results
Effect of the Interventions on Frequency of Progress
Monitoring and Goal Attainment
We first computed the effect size for the difference in the
frequency of progress monitoring between the treatment and con-
trol conditions following the intervention (see Figure 2). The
sample-weighted average effect size was d
1.98 with a 95%
confidence interval from 1.72 to 2.24, based on 138 studies and a
total sample size of 19,951. This indicates that the interventions
had, on average, a (very) large effect on the frequency of progress
monitoring according to Cohen’s (1992) criteria. Our sample of
studies, therefore, is suitable for testing whether progress moni-
toring promotes rates of behavioral performance and goal attain-
ment. There was, however, significant heterogeneity in effect sizes
across the primary studies, Q(137) 7490.15, p.001, and it is
worth noting that interventions had a larger effect on the frequency
of progress monitoring when the comparison condition involved
no monitoring (d
3.34) than when the comparison condition
involved some progress monitoring (d
0.68), Q(1) 5252.05,
p.001 (see Table 3).
Next, we computed the average effect of the interventions on
goal attainment (see Figure 3). The sample-weighted average
effect size was d
0.40 with a 95% confidence interval from
4
To examine the impact of these estimation procedures, we compared
the effect sizes for progress monitoring and goal attainment when esti-
mated values were included versus excluded from respective computations.
Findings showed that the effect size for goal attainment did not differ when
effect sizes based on estimated values were included (d
0.40) versus
excluded (d
0.40), Q(1) 0.00, p.96. However, the effect size for
progress monitoring was slightly smaller when effect sizes based on
estimated values were included (d
1.98) versus excluded (d
2.06),
Q(1) 10.74, p.01. This is not surprising as the estimation procedures
are conservative in assuming the smallest possible effect size that would
produce a given significance value and that the effect size is zero when the
effect is reported as nonsignificant. However, the fact that estimation
procedures were used to compute a relatively small proportion of the effect
sizes (7% of the effect sizes for progress monitoring, 9% for goal attain-
ment) and the difference in the sample-weighted effect sizes for progress
monitoring is small suggests that these procedures did not unduly influence
our findings.
5
Effect sizes for the three studies employing cluster randomization were
adjusted using the design effect equation of: 1 (M 1ICC), where
M and ICC refer to the average cluster size and interclass correlation
coefficient, respectively. In the absence of an ICC, it was estimated to be
0.05. The design effect was then used to calculate the corrected sample size
for the treatment and control conditions (Higgins & Green, 2011). In order
to check that the inclusion of studies with cluster-randomized designs did
not bias effect sizes, we also conducted a sensitivity analysis removing
studies where the unit of analysis was the group. The effect sizes did not
differ significantly when studies with cluster randomized designs were
included (d
1.98 and 0.40 for effects on progress monitoring and goal
attainment, respectively) versus excluded (d
1.96 and 0.40, respec-
tively), Q(1) 0.33 and 0.02, p.57 and 0.89.
208 HARKIN ET AL.
NOTE: Weights are from random effects analysis
Overall (I-squared = 98.2%, p = 0.000)
Raynor et al.
Phelan et al.
Proudfoot et al.
Helsel et al.
Smith et al.
Wing et al. - Comparison 1 A
Thorndike et al.
Caldwell et al.
Sengpiel et al.
Moreland et al. - Comparison 2
Brindal et al.
McMurdo et al. - Comparison 2
Chao et al.
Tan et al.
Blasco et al.
McMurdo et al. - Comparison 1
Sherwood et al.
Beasley et al
Young & Starkes
Bell et al.
DeWalt et al.
Maljanian et al.
Cussler et al.
Talbot et al.
Authors
Duran et al.
Farmer et al. - Comparison 1
Dennison et al. - Comparison 1
Pellegrini et al.
Wang et al.
Moreland et al. - Comparison 1
Anderson et al.
Jefferson
Rosal et al. B
Kirwan et al.
Kroenke et al.
O'Kane et al.
De Blok et al.
Maruyama et al.
Nanchahal et al.
Shapiro et al.
Wing et al. B
Haddock et al.
Farmer et al. - Comparison 2
Gold et al.
Petrella et al.
Coughlin et al. - Comparison 2
Kraschnewski et al.
Runyan et al.
Morgan et al.
McManus et al.
De Cocker et al. A
Mehos et al.
Steinberg et al.
Domingo et al.
Orsama et al.
Acharya et al.
Haapala et al.
Gokee et al. A
Allen et al. - Comparison 3
Logan et al.
McKinstry et al.
Aronson et al.
Hyman et al.
Ligibel et al.
Hurling et al.
Quinn et al.
Linde & Jeffrey
Rote
King et al.
Carter et al. - Comparison 2
Kempf et al.
Andrews et al.
Dennis et al.
Gokee et al. B
De Cocker et al. B
Abraira et al.
Van der Meer et al.
Hannum et al.
Homko et al.
Dennison et al. - Comparison 2
Wing et al. C
Atienza et al.
Ornes
Buist et al.
Hellerstedt & Jeffrey - Comparison 2
Gajecki et al.
Piette et al.
Kim et al.
Suffoletto et al. B
Petersen et al.
Sugden et al.
Goto et al.
Tate et al.
Arbour & Ginis
D'Eramo - Comparison 2
Suffoletto et al. A
Coughlin et al. - Comparison 1
Chau et al.
Samuel-Hodge et al.
Amsberg et al.
Rosal et al. A
Sheldon
Allen et al. - Comparison 1
Webber et al.
Oshima et al.
Spence et al. - Comparison 1
Turner-McGrievy & Tate
Cho et al.
Ralston et al.
D'Eramo - Comparison 1
Chambliss et al.
Seto et al.
Carter et al. - Comparison 1
Antypas & Wangberg
Goulis et al.
Jurgens et al.
Nguyen et al.
Boutelle et al.
Marquez-Contreras et al.
Berg et al.
Hellerstedt & Jeffrey - Comparison 1
An et al.
Jennings et al.
Kristal et al.
Janson et al. B
Wing et al. - Comparison 2 A
Clarke et al.
Polzien et al.
Allen et al. - Comparison 2
Richardson et al.
Muchmore et al.
Kobulnicky
Kwon et al.
Akers et al.
Polonsky et al.
Janson et al. A
Abrahams et al.
Carli et al.
0-6.49 0 6.49
Figure 2. Forest plot showing the effect of interventions on the frequency of progress monitoring.
209
PROGRESS MONITORING AND GOAL ATTAINMENT
Table 3
Categorical Moderators of the Effect of Interventions on Progress Monitoring and Goal Attainment
Progress monitoring Goal attainment
Moderator Nk Q 95% CI d
Nk Q 95% CI d
Focus of PM
a
Monitor behavior 12,624 78 4,510.44
ⴱⴱⴱ
[1.84, 2.54] 2.19 11,461 78 383.73
ⴱⴱⴱ
[.33, .52] .43
Monitor outcomes 12,390 83 4,433.71
ⴱⴱⴱ
[2.00, 2.66] 2.33 11,360 83 571.64
ⴱⴱⴱ
[.31, .52] .42
Public vs. private monitoring
Public monitoring 218 3 71.16
ⴱⴱⴱ
[.18, 4.76] 2.47
a
214 3 11.35
ⴱⴱ
[.16, 1.26] .55
a
Private (reported) 13,417 95 1,359.49
ⴱⴱⴱ
[2.17, 2.85] 2.51
a
12,155 95 665.24
ⴱⴱⴱ
[.37, .58] .47
a
Private (not reported) 3,251 14 973.15
ⴱⴱⴱ
[.40, 1.78] 1.09
b
3,177 14 42.77
ⴱⴱⴱ
[.05, .33] .19
b
1,039.30
ⴱⴱⴱ
49.86
ⴱⴱⴱ
Recorded vs. not recorded monitoring
Recorded 16,931 106 6,362.60
ⴱⴱⴱ
[2.08, 2.70] 2.39
a
15,589 106 748.26
ⴱⴱⴱ
[.34, .53] .43
a
Not recorded 3,020 32 295.99
ⴱⴱⴱ
[.36, .85] .60
b
2,809 32 86.48
ⴱⴱⴱ
[.15, .42] .29
b
1,784.65
ⴱⴱⴱ
12.71
ⴱⴱⴱ
Reference value
Past 2,491 12 631.54
ⴱⴱⴱ
[1.98, 3.79] 2.88
a
2,019 12 68.15
ⴱⴱⴱ
[.20, .66] .43
Desired (future) target 5,740 44 1,874.19
ⴱⴱⴱ
[1.87, 2.80] 2.33
b
5,480 44 333.38
ⴱⴱⴱ
[.25, .58] .41
Others 479 2 176 2
67.70
ⴱⴱⴱ
.14
Monitor rate vs. distance
Monitor rate of progress 293 3 91.89
ⴱⴱⴱ
[.18, 3.85] 1.84
b
286 3 2.85 [.10, .68] .39
Monitor distance from
goal 8,593 44 3,017.50
ⴱⴱⴱ
[1.72, 2.67] 2.20
b
8,172 44 449.23
ⴱⴱⴱ
[.29, .59] .44
6.68
ⴱⴱ
.19
Passive vs. active monitoring
Passive monitoring 2,063 13 436.12
ⴱⴱⴱ
[1.35, 2.71] 2.03
b
1,426 13 18.90 [.24, .54] .39
Active monitoring 17,462 111 6,612.62
ⴱⴱⴱ
[2.02, 2.62] 2.32
a
16,105 111 787.97
ⴱⴱⴱ
[.34, .52] .43
24.90
ⴱⴱⴱ
.49
Method used to promote PM
BP monitor 1,126 5 5.95 [3.07, 3.56] 3.31
a
1,074 5 136.40
ⴱⴱⴱ
[.18, 1.45] .64
a
BG monitor 1,886 15 737.08
ⴱⴱⴱ
[.90, 2.50] 1.70
e
1,726 15 88.15
ⴱⴱⴱ
[.34, .86] .60
a
Website 5,576 30 1,632.30
ⴱⴱⴱ
[1.40, 2.40] 1.90
d
4,787 30 185.10
ⴱⴱⴱ
[.31, .63] .47
b
Written diary 5,815 46 1,964.24
ⴱⴱⴱ
[1.11, 1.92] 1.51
f
5,626 46 183.01
ⴱⴱⴱ
[.29, .54] .42
b
Phone 2,934 24 873.23
ⴱⴱⴱ
[1.99, 3.34] 2.67
c
2,775 23 102.07
ⴱⴱⴱ
[.06, .43] .25
c
PDA 1,156 12 449.60
ⴱⴱⴱ
[.98, 2.93] 1.96
d
1,140 12 21.70
[.02, .39] .21
c
Pedometer 1,258 5 248.60
ⴱⴱⴱ
[1.52, 4.52] 3.02
b
1,070 5 1.36 [.10, .14] .02
d
MEMS 200 1 200 1
826.86
ⴱⴱⴱ
102.38
ⴱⴱⴱ
Source of the Intervention
Health professionals 3,944 32 1,827.89
ⴱⴱⴱ
[1.67, 2.96] 2.31
a
3,800 32 137.85
ⴱⴱⴱ
[.22, .52] .37
Researchers 6,215 42 2,515.50
ⴱⴱⴱ
[1.13, 2.13] 1.63
c
5,864 42 121.52
ⴱⴱⴱ
[.25, .46] .35
b
Mixed team 1,790 16 567.26
ⴱⴱⴱ
[1.29, 2.66] 1.97
b
1,662 16 64.02
ⴱⴱⴱ
[.22, .66] .44
Not face-to-face 8,002 48 2,341.74
ⴱⴱⴱ
[1.65, 2.46] 2.06
b
7,072 48 509.45
ⴱⴱⴱ
[.27, .61] .54
a
213.40
ⴱⴱⴱ
7.13
Focal behavior or target outcome
Depression 832 3 226.75
ⴱⴱⴱ
[.56, 4.91] 2.74
a
658 3 18.21
ⴱⴱⴱ
[.15, 1.16] .66
a
Blood pressure 1,615 6 590.07
ⴱⴱⴱ
[1.21, 4.19] 2.70
a
1,563 6 148.48
ⴱⴱⴱ
[.03, 1.23] .63
a
Asthma management 700 5 189.69
ⴱⴱⴱ
[.05, 2.59] 1.32
d
657 5 69.36
ⴱⴱⴱ
[.14, 1.34] .60
a
Physical activity 2,808 25 1,312.18
ⴱⴱⴱ
[1.28, 2.75] 2.02
b
2,347 25 142.85
ⴱⴱⴱ
[.37, .82] .59
a
Blood glucose 3,385 24 1,426.23
ⴱⴱⴱ
[1.27, 2.61] 1.94
b
3,378 24 145.91
ⴱⴱⴱ
[.32, .70] .51
a
Weight 6,255 50 1,549.38
ⴱⴱⴱ
[1.32, 2.00] 1.66
c
5,487 50 138.27
ⴱⴱⴱ
[.20, .40] .30
b
Diet 1,446 6 331.15
ⴱⴱⴱ
[.42, 3.01] 1.71
c
1,419 6 5.61 [.12, .34] .23
b
Heart care/cholesterol 757 10 251.03
ⴱⴱⴱ
[1.75, 3.94] 2.85
a
757 10 7.36 [.00, .28] .14
b
Use of healthcare systems 363 3 59.85
ⴱⴱⴱ
[1.08, 4.23] 2.66
a
347 3 3.77 [.29, .30] .01
b
Alcohol 830 1 830 1
Smoking cessation 517 1 517 1
Prophylaxis use 253 1 253 1
210 HARKIN ET AL.
Table 3 (continued)
Progress monitoring Goal attainment
Moderator Nk Q 95% CI d
Nk Q 95% CI d
Sun protection 91 1 86 1
Time management 61 1 61 1
Medication adherence 38 1 38 1
452.48
ⴱⴱⴱ
114.43
ⴱⴱⴱ
Measure
Objective 10,667 79 3,831.55
ⴱⴱⴱ
[1.96, 2.67] 2.31
a
10,867 80 547.54
ⴱⴱⴱ
[.33, .55] .44
a
Self-report 9,203 58 3,328.98
ⴱⴱⴱ
[1.12, 1.88] 1.50
b
4,658 24 116.90
ⴱⴱ
[.19, .50] .34
b
Combination 2,873 34 158.80
ⴱⴱⴱ
[.17, .52] .34
b
564.86
ⴱⴱⴱ
10.33
ⴱⴱ
Nature of the comparison group
No monitoring 10,864 67 968.41
ⴱⴱⴱ
[3.10, 3.57] 3.34
a
9,883 67 598.04
ⴱⴱⴱ
[.28, .55] .42
Some monitoring 9,087 71 1,077.50
ⴱⴱⴱ
[.50, .86] .68
b
8,565 71 239.73
ⴱⴱⴱ
[.29, .47] .38
5,252.05
ⴱⴱⴱ
1.62
Study quality
Participant blinding
Blind 2,159 14 876.54
ⴱⴱⴱ
[.72, 2.44] 1.58
b
1,911 14 38.85
ⴱⴱⴱ
[.14, .49] .32
b
Not blind 8,154 34 2,684.96
ⴱⴱⴱ
[2.12, 3.24] 2.68
a
7,285 34 479.14
ⴱⴱⴱ
[.34, .73] .53
a
360.58
ⴱⴱⴱ
17.37
ⴱⴱⴱ
Experimenter blinding
Blind 6,616 44 2,808.16
ⴱⴱⴱ
[1.43, 2.41] 1.92
b
6,027 44 164.89
ⴱⴱⴱ
[.26, .48] .37
Not blind 7,449 33 2,564.17
ⴱⴱⴱ
[1.97, 3.09] 2.53
a
6,864 33 401.73
ⴱⴱⴱ
[.23, .60] .41
202.39
ⴱⴱⴱ
1.39
Randomization success
Successful 12,340 88 4,464.83
ⴱⴱⴱ
[1.62, 2.28] 1.95
b
11,229 88 286.77
ⴱⴱⴱ
[.25, .40] .33
c
US not controlled 2,591 18 1,000.74
ⴱⴱⴱ
[1.98, 3.61] 2.80
a
2,513 18 140.53
ⴱⴱⴱ
[.23, .73] .48
b
US controlled 2,276 16 784.73
ⴱⴱⴱ
[1.22, 2.69] 1.95
b
2,007 16 163.55
ⴱⴱⴱ
[.29, .94] .62
a
Not assessed 1,725 11 648.59
ⴱⴱⴱ
[.61, 2.68] 1.65
c
1,664 11 183.54
ⴱⴱⴱ
[.04, .98] .51
ab
252.31
ⴱⴱⴱ
46.03
ⴱⴱⴱ
Type of randomization
Individual 9,653 88 3,971.82
ⴱⴱⴱ
[1.63, 2.31] 1.97
c
9,003 88 403.54
ⴱⴱⴱ
[.28, .48] .38
b
Cluster 417 3 124.60
ⴱⴱⴱ
[.19, 4.92] 2.56
a
306 3 3.11 [.12, .54] .21
c
Stratified 375 1 291 1
Minimization 466 3 168.03
ⴱⴱⴱ
[.91, 4.42] 2.22
b
466 3 5.37 [.23, .61] .19
c
Combined 8,067 39 2,512.03
ⴱⴱⴱ
[1.58, 2.50] 2.04
c
7,357 39 413.29
ⴱⴱⴱ
[.31, .64] .47
a
25.80
ⴱⴱⴱ
17.79
ⴱⴱⴱ
Quality of randomization
High
b
5,429 35 2,128.99
ⴱⴱⴱ
[1.40, 2.44] 1.92
b
5,184 35 133.76
ⴱⴱⴱ
[.23, .47] .35
b
Medium
c
11,567 73 4,057.69
ⴱⴱⴱ
[1.73, 2.46] 2.09
a
10,379 73 562.79
ⴱⴱⴱ
[.30, .54] .42
a
Low
d
2,955 30 1,055.83
ⴱⴱⴱ
[1.21, 2.30] 1.75
c
2,835 30 137.61
ⴱⴱⴱ
[.21, .58] .40
54.96
ⴱⴱⴱ
4.29
Type of participant
General public 3,420 14 1,136.43
ⴱⴱⴱ
[1.32, 3.08] 2.20
b
2,969 14 103.82
ⴱⴱⴱ
[.27, .76] .52
a
Specific samples 14,780 114 5,040.47
ⴱⴱⴱ
[1.64, 2.17] 1.91
c
13,688 114 616.01
ⴱⴱⴱ
[.30, .47] .39
b
Diabetes 4,026 30 1,759.84
ⴱⴱⴱ
[1.26, 2.46] 1.86
e
4,038 30 152.14
ⴱⴱⴱ
[.32, .63] .48
e
Overweight 5,406 47 1,304.96
ⴱⴱⴱ
[1.24, 1.93] 1.59
e
4,724 47 145.41
ⴱⴱⴱ
[.22, .45] .33
e
Psychological illness 467 3 43.35
ⴱⴱⴱ
[1.02, 3.68] 2.18
e
446 3 18.51
ⴱⴱⴱ
[.12, 1.33] .60
e
Other conditions 4,881 34 1,846.29
ⴱⴱⴱ
[1.78, 2.89] 2.34
e
4,480 34 311.73
ⴱⴱⴱ
[.16, .55] .35
e
University students 1,751 10 475.50
ⴱⴱⴱ
[1.20, 3.65] 2.42
a
1,741 10 77.83
ⴱⴱⴱ
[.06, .77] .41
88.41
ⴱⴱⴱ
10.15
ⴱⴱ
Note. Effect sizes with different subscripts (within each moderator) differ significantly (p.05). CI confidence interval; PM progress monitoring;
BCT behavior change technique; BP blood pressure; BG blood glucose; PDA personal digital assistant; MEMs medication event monitoring
system.
a
Levels of the moderator were not compared as categories are not mutually exclusive (i.e., interventions could prompt participants to monitor their behavior
and the outcomes of their behavior, see, for example, Farmer et al., 2007).
b
Truly randomized (and method described) and experimenter unlikely to know
condition or described as randomized and double blinded (but method not described) and no significant differences in relevant pretest mea-
sures.
c
Randomized, but method not described and experimenter blinded or randomization described but it is possible that the experimenter may have
known the condition.
d
Randomized, but method not described and experimenter was not blinded to condition.
e
Effect sizes for different specific
samples were not statistically compared.
p.05.
ⴱⴱ
p.01.
ⴱⴱⴱ
p.001.
211
PROGRESS MONITORING AND GOAL ATTAINMENT
NOTE: Weights are from random effects analysis
Overall (I-squared = 83.6%, p = 0.000)
De Cocker et al. A
Petersen et al.
Hellerstedt & Jeffrey - Comparison 2
Ligibel et al.
Linde & Jeffrey
Spence et al. - Comparison 1
Orsama et al.
Hurling et al.
Berg et al.
Talbot et al.
Farmer et al. - Comparison 2
Steinberg et al.
Polzien et al.
Farmer et al. - Comparison 1
Bell et al.
Rosal et al. B
Wing et al. B
Runyan et al.
Nanchahal et al.
Oshima et al.
Janson et al. B
Haapala et al.
Dennison et al. - Comparison 1
Caldwell et al.
Carter et al. - Comparison 1
Kroenke et al.
An et al.
Polonsky et al.
Sengpiel et al.
McKinstry et al.
Jefferson
Nguyen et al.
Moreland et al. - Comparison 2
Rote
Goto et al.
Antypas & Wangberg
Wing et al. C
Logan et al.
De Cocker et al. B
Haddock et al.
Cussler et al.
Rosal et al. A
Dennison et al. - Comparison 2
Coughlin et al. - Comparison 1
Kwon et al.
Maruyama et al.
Smith et al.
Chao et al.
King et al.
D'Eramo - Comparison 2
Richardson et al.
Gajecki et al.
Kim et al.
Tan et al.
Amsberg et al.
Thorndike et al.
Sugden et al.
Arbour & Ginis
Wang et al.
Webber et al.
Aronson et al.
Domingo et al.
Gokee et al. B
Samuel-Hodge et al.
Young & Starkes
Tate et al.
Kobulnicky
Kraschnewski et al.
Morgan et al.
Abrahams et al.
Kempf et al.
Jurgens et al.
Suffoletto et al. B
D'Eramo - Comparison 1
Muchmore et al.
Janson et al. A
Buist et al.
Wing et al. - Comparison 1 A
Petrella et al.
Kristal et al.
Brindal et al.
Hellerstedt & Jeffrey - Comparison 1
Jennings et al.
Ralston et al.
Wing et al. - Comparison 2 A
Goulis et al.
Quinn et al.
Gold et al.
Ornes
Gokee et al. A
Helsel et al.
McManus et al.
Phelan et al.
Duran et al.
Allen et al. - Comparison 3
Hannum et al.
De Blok et al.
Authors
Dennis et al.
Marquez-Contreras et al.
Homko et al.
Suffoletto et al. A
Proudfoot et al.
Kirwan et al.
Chau et al.
Moreland et al. - Comparison 1
O'Kane et al.
Clarke et al.
Carter et al. - Comparison 2
Van der Meer et al.
Shapiro et al.
Maljanian et al.
Cho et al.
Pellegrini et al.
Raynor et al.
Mehos et al.
Turner-McGrievy & Tate
Carli et al.
Allen et al. - Comparison 1
Blasco et al.
McMurdo et al. - Comparison 1
Coughlin et al. - Comparison 2
Sheldon
Andrews et al.
Atienza et al.
Boutelle et al.
McMurdo et al. - Comparison 2
Abraira et al.
Anderson et al.
Allen et al. - Comparison 2
DeWalt et al.
Akers et al.
Hyman et al.
Beasley et al
Piette et al.
Acharya et al.
Chambliss et al.
Sherwood et al.
Seto et al.
0-2.84 0 2.84
Figure 3. Forest plot showing the effect of the interventions on goal attainment.
212 HARKIN ET AL.
0.32 to 0.48 (k138; N18,398). Effect sizes were heteroge-
neous, Q(137) 837.77, p.001. The sample-weighted average
effect of the interventions on studies that only measured behavior
was d
0.37 (k35, N5,518, 95% CI [0.25, 0.49]), and
among those that focused on outcomes was d
0.44 (k86,
N10,593, 95% CI [0.33, 0.55]). There was significant variabil-
ity in effect sizes in both cases, Q(34) 116.77, p.001 and
Q(85) 581.76, p.001, respectively. In summary, interventions
that engendered large-sized changes in the frequency of progress
monitoring, on average, led to small-to-medium-sized changes in
goal attainment.
To check this conclusion, we conducted a mediation analysis
using data from the 21 studies (N1,995) where the correlation
between the frequency of progress monitoring and goal attainment
could be retrieved.
6,7
In line with Kenny, Kashy, and Bolger’s
(1998) recommendations, four multiple regressions were con-
ducted to establish mediation (dvalues were converted to effect
size rfor this purpose, and the sample-weighted average correla-
tions between intervention, frequency of progress monitoring, and
goal attainment were used in the matrix input function for multiple
regression). Regression analyses showed that intervention (the
independent variable) predicted both changes in goal attainment
(the dependent variable) and changes in the frequency of progress
monitoring (the proposed mediator; see Figure 4). Changes in the
frequency of progress monitoring also predicted changes in goal
attainment. Most important, however, changes in the frequency of
progress monitoring attenuated the effect of the interventions on
goal attainment in a simultaneous regression analysis. This con-
clusion was confirmed by a significant value on Kenny et al.’s
(1998) modification of the Sobel (1982) test (Z13.09, p
.001), which shows that changes in the frequency of progress
monitoring significantly reduced the association between interven-
tions and goal attainment.
Tests for Potential Bias
Three analyses tested for publication and small sample bias.
First, we compared effect sizes for published (k129, 93%)
versus unpublished studies (k9, 7%). No significant difference
in effect sizes was observed for the frequency of progress moni-
toring (d
1.99 vs. 1.86, for published and unpublished studies,
respectively) or goal attainment (d
0.40 vs. 0.42, respectively)
suggesting a lack of publication bias. Second, we computed Egg-
er’s regression (Egger, Davey Smith, Schneider, & Minder, 1997)
to test for asymmetry in the distribution of effect sizes. The
regression coefficients were significant for progress monitoring
(p.001) and goal attainment (p.01). We therefore used Duval
and Tweedie’s (2000) trim and fill analysis to estimate adjusted
effect sizes. The trim and fill analysis imputed 37 and 31 additional
effect sizes and produced adjusted estimates of d
1.16 (95% CI
[0.89, 1.44]) and d
0.19 (95% CI [0.10, 0.28]) for the fre-
quency of progress monitoring and goal attainment, respectively.
Peters, Sutton, Jones, Abrams and Rushton (2007) pointed out that
“[w]hen there is large between-study heterogeneity the trim and
fill method can underestimate the true positive effect when there is
no publication bias” (p. 4544). Both of these conditions were met
in the present data. Our interpretation, therefore, is that (a) The
influence of publication bias in the current meta-analysis is modest
rather than severe (Rothstein, Sutton, & Borenstein, 2005); and (b)
the magnitude of effects on frequency of progress monitoring and
goal attainment can be deemed large and small-to-medium, respec-
tively. This interpretation is also supported by findings from tests
for small sample bias. Coyne, Thombs, and Hagedoorn (2010)
recommended that researchers compute effect sizes separately for
studies that have at least 55% power to detect a medium-sized
effect (i.e., n35 per condition). Seventy studies in the present
review (51%) met this criterion. The effect sizes among these,
adequately powered, interventions was d
2.05 for frequency of
progress monitoring and d
0.33 for goal attainment.
6
The 20 studies used in the mediation analysis were Akers, Cornett,
Savla, Davy, and Davy (2012); Boutelle, Kirschenbaum, Baker, and Mitch-
ell (1999); Chambliss et al. (2011); Cussler et al. (2008); Duran et al.
(2010); Gokee LaRose, Gorin, and Wing (2009); Hellerstedt and Jeffery
(1997, behavior-focused phone group), Helsel et al. (2007); Kempf,
Tankova, and Martin (2013); Kraschnewski et al. (2011); Morgan et al.
(2009); Nguyen, Gill, Wolpin, Steele, and Benditt (2009); Pellegrini et al.
(2012); Runyan, Steenbergh, Bainbridge, Daugherty, Oke, and Fry (2013);
Samuel-Hodge et al. (2009); Tate et al. (2001); Tan, Maganee, Chee, Lee,
and Tan (2011); Wang, Sereika, Chasens, Ewing, Matthews, and Burke
(2012); Webber, Tate, Ward, and Bowling (2010), and Wing, Crane,
Thomas, Kumar, and Weinberg (2010). These 20 studies did not differ
significantly from excluded studies in terms of their reported effect on goal
attainment (d
0.44 and 0.39, respectively), Q(1) 0.71, p.40, but
did tend to report smaller effects on the frequency of progress monitoring
(d
1.51) than excluded studies (d
2.06), Q(1) 83.47, p.001.
A potential explanation of these differences is that studies may have been
more likely to report the correlation between progress monitoring and goal
attainment (and therefore could be included in the mediation analysis) if
participants in the comparison condition were also asked to monitor their
progress—a methodological feature that led to smaller effect sizes for
progress monitoring.
7
For the purposes of this analysis we recomputed the effect size for goal
attainment using only the measures for which the primary studies reported
the correlation between the frequency of progress monitoring and goal
attainment. For example, Cussler et al. (2008) reported the effect of their
intervention on two behaviors (energy expenditure and energy intake) and
five outcomes (weight, BMI, percentage body fat, total body fat, and
fat-free mass). However, Cussler et al. (2008) only reported the correlation
between the frequency of progress monitoring and three of these measures
of goal attainment (weight, energy expenditure, and energy intake). For the
purposes of the mediation analyses, we therefore recomputed the effect of
this intervention on the three measures of goal attainment for which
correlations were reported.
Intervention
Goal Attainment
Frequency of
progress
monitoring
β = 0.33***
(β = 0.19***)
β = 0.55*** β = 0.36***
Figure 4. Mediation of the effect of interventions on goal attainment by
changes in the frequency of progress monitoring. Note: The value in
parentheses represents the effect of interventions on goal attainment, con-
trolling for changes in the frequency of progress monitoring.
ⴱⴱⴱ
p.001.
213
PROGRESS MONITORING AND GOAL ATTAINMENT
Does Behavior Change Mediate the Impact of Progress
Monitoring on Outcomes?
Next, we investigated whether changes in behavior mediated the
effect of interventions designed to promote progress monitoring on
outcomes (e.g., whether interventions improved dietary and exer-
cise behavior which, in turn, explained weight loss). Mediation
analyses were undertaken using data from the six studies (N
473) where the correlation between changes in behavior and
changes in outcomes could be retrieved.
8
Intervention (the inde-
pendent variable) predicted changes in outcomes (the dependent
variable), and in behavior (the proposed mediator; see Figure 5).
Changes in behavior also predicted changes in outcomes. Most
important, however, simultaneous regression analysis showed that
changes in behavior attenuated the effect of intervention on out-
comes. This conclusion was confirmed by a significant value on
Kenny et al.’s (1998) modification of the Sobel (1982) test (Z
3.54, p.01). Thus, changes in behavior mediated the relation-
ship between interventions and outcomes.
Moderators of Intervention Effects on Progress
Monitoring and Goal Attainment
The effects of the interventions on the frequency of progress
monitoring and goal attainment were heterogeneous, which en-
courages the search for moderator variables. The sample-weighted
effect size (d
) and homogeneity statistic (Q) were therefore
calculated separately for each level of the moderator, and Schwar-
zer’s (1988) META program was used to test whether effect sizes
differed significantly (see Table 3). The impact of continuous
moderators (i.e., duration of the intervention) on effect sizes was
examined using metaregression (via the metareg command in
STATA, see Table 4).
Intervention characteristics. We began by examining mod-
eration by intervention characteristics. Several dimensions of prog-
ress monitoring influenced the frequency with which participants
monitored their progress and the effects of so doing on goal
attainment. Below, we focus on the effects on goal attainment, but
the effects on the frequency of progress monitoring were broadly
similar (see Table 3). The focus of progress monitoring (behavior
vs. outcomes) did not appear to influence effect sizes for goal
attainment (see Table 3). However, we observed the predicted
“matching effect,” such that monitoring behavior had a large,
reliable effect on behavior (d
0.79, 95% CI [0.50, 1.07], k
17, N2,565), but no reliable effect on outcomes (d
0.14,
95% CI [0.18, 0.46], k8, N1,175), Qfor comparison
82.91, p.001. In contrast, monitoring outcomes had a medium-
to-large, reliable effect on outcomes (d
0.62, 95% CI [0.26,
0.98], k30, N4,199), but did not reliably affect behavior
(d
0.17, 95% CI [0.01, 0.36], k4, N975), Qfor
comparison 39.59, p.001.
9
Prompting participants to monitor their progress in public or to
report the information that they obtained via monitoring had larger
effects on goal attainment (d
0.55 and 0.47, respectively) than
did monitoring in private (d
0.19), Q(1) 6.17 and 48.91,
respectively, p.05 and .001. Physically recording the infor-
mation derived from monitoring led to larger effects on goal
attainment (d
0.43) compared with not recording this infor-
mation (d
0.29), Q(1) 12.71, p.001), and this was also
the case when goal attainment was measured objectively (d
0.57 vs. 0.23). The nature of the reference value did not influence
effect sizes. Comparing the current state to a desired (future) target
had comparable effects on goal attainment (d
0.41) as com-
paring the current state with a reference value in the past (d
0.43), Q(1) 0.14, p.71. Finally, whether participants moni-
tored their rate of goal progress or distance from the goal, or used
passive versus active forms of monitoring, did not influence the
impact of monitoring on goal attainment, Q(1) 0.19 and 0.49 for
the two comparisons, respectively.
The method used to promote progress monitoring influenced the
frequency with which participants monitored their progress,
Q(6) 826.86, p.001, and the effect of the interventions on
goal attainment, Q(6) 102.38, p.001. Interventions that asked
participants to monitor their progress using a phone (d
2.67),
blood pressure monitor (d
3.31), or pedometer (d
3.02)
showed the largest differences in the frequency of progress mon-
itoring (relative to comparison conditions). The largest effects on
goal attainment were observed among participants using a blood
pressure monitor or blood glucose monitor to assess their goal
progress (d
0.64 and 0.60, respectively). The source of the
intervention also significantly influenced the frequency of progress
monitoring, Q(3) 213.40, p.001, and (marginally) goal
attainment, Q(3) 7.13, p.07. Pairwise comparisons revealed
that interventions that were delivered by health professionals were
associated with larger changes in the frequency of progress mon-
itoring (d
2.31) than interventions delivered by researchers
(d
1.63), mixed teams (d
1.97), or interventions that were
not delivered face-to-face (d
2.06). For goal attainment, the
only significant difference was that interventions delivered by
researchers tended to have smaller effects (d
0.35) than inter-
ventions that were not delivered face-to-face (d
0.54). The
8
The six studies used in the mediation analysis were Arbour and Martin
Ginis (2008); Haapala, Barengo, Biggs, Surakka, and Manninen (2009);
Janson, Fahy, Covington, Paul, Gold, and Boushey (2003); Tan et al.
(2011); Tate et al. (2001); and Wing et al. (2010). These six studies did not
differ significantly from excluded studies in terms of their reported effect
on goal attainment (d
0.55 and 0.39, for included vs. excluded studies,
respectively), p.08, but did tend to report smaller effects on the
frequency of progress monitoring (d
1.48) than excluded studies (d
2.00), Q(1) 30.24, p.001.
9
Where studies measured both behavior and outcomes, only the mea-
sures relevant to the nature of progress monitoring (i.e., behavioral mea-
sures when participants were prompted to monitor their behavior, outcome
measures when participants were prompted to monitor outcomes) were
included in this analysis.
Intervention
Changes in
outcomes
Changes in
behavior
β = 0.24***
(β = 0.20***)
β = 0.24*** β = 0.21***
Figure 5. Mediation of the effect of Interventions on outcomes by
changes in behavior. Note: The value in parentheses represents the effect
of interventions on outcomes, controlling for changes in behavior.
ⴱⴱⴱ
p
.001.
214 HARKIN ET AL.
duration of the intervention had no impact on the frequency of
progress monitoring, ␤⫽0.00, t0.08, p.94, or on goal
attainment, ␤⫽⫺0.00, t⫽⫺0.15, p.88 (see Table 4).
The inclusion of additional BCTs—notably, goal setting, high-
lighting the discrepancy between current behavior and the goal,
immediate feedback on behavior, delayed feedback on behavior or
outcomes, or action planning—increased the effect of the inter-
ventions designed to promote progress monitoring on goal attain-
ment, relative to interventions that did not incorporate these BCTs
(see Table 5). Interestingly, providing immediate feedback on
behavior alongside progress monitoring engendered larger effects
on goal attainment than each of the other types of feedback (p
.05 for all comparisons). Prompting review of behavioral or out-
come goals was not associated with a significant increase in the
impact of interventions on goal attainment (see Table 5).
Methodological characteristics. Finally, we examined the
impact of methodological characteristics on effect sizes. The na-
ture of the focal behavior or target outcome had a significant
impact on the effect of interventions on the frequency of progress
monitoring, Q(1) 452.48, p.001, and goal attainment, Q(1)
114.43, p.001. As Table 3 shows, prompting progress moni-
toring had medium-sized effects on goal attainment among studies
focusing on depression (d
0.66), blood pressure (d
0.63),
the management of asthma (d
0.60), physical activity (d
0.59), and blood glucose levels (d
0.51), and small effects
among studies focusing on weight (d
0.30) and diet (d
0.23). Prompting progress monitoring did not promote goal attain-
ment among studies focusing on heart care behaviors (d
0.14)
or the use of health care systems (d
0.01). Effect sizes also
differed as a function of the measure of progress monitoring,
Table 4
Continuous Moderators of the Effect of Interventions on Progress Monitoring and Goal Attainment
Progress monitoring Goal attainment
Moderator MSDNkI
2
95% CI Adj-R
2
NkI
2
95% CI Adj-R
2
Age 48.00 13.44 19,951 138 98.18 .02 [.00, .04] 1.61 18,398 138 83.48 .00 [.01, .01] .82
% female 66.31 24.47 19,951 138 98.18 .00 [.01, .01] .73 18,398 138 83.73 .00 [.00, .00] 1.03
Duration
a
178.78 163.36 19,951 138 98.18 .00 [.00, .00] .77 18,398 138 83.68 .00 [.00, .00] 1.16
Note. CI confidence interval; Adj adjusted.
a
Duration was coded as the number of days over which participants were asked to monitor their progress.
p.05.
Table 5
Effect of Additional Behavior Change Techniques (BCTs) on Goal Attainment
Behavior change technique Nk Q 95% CI d
Goal setting—behavior
Included 6,073 46 215.82
ⴱⴱⴱ
[.36, .61] .48
a
Not included 12,325 92 621.28
ⴱⴱⴱ
[.25, .46] .35
b
Goal setting—outcome
Included 3,345 22 206.59
ⴱⴱⴱ
[.30, .78] .54
a
Not included 15,053 116 595.94
ⴱⴱⴱ
[.29, .45] .37
b
Review behavioral goals
Included 4,046 31 130.27
ⴱⴱⴱ
[.24, .53] .38
Not included 14,352 107 704.35
ⴱⴱⴱ
[.31, .50] .40
Review outcome goals
Included 1,000 10 20.53
[.15, .56] .36
Not included 17,398 128 817.21
ⴱⴱⴱ
[.32, .49] .40
Action planning
Included 5,757 43 445.32
ⴱⴱⴱ
[.32, .68] .50
a
Not included 12,641 95 364.88
ⴱⴱⴱ
[.27, .42] .34
b
Prompt identification of discrepancy
Included 5,053 45 321.63
ⴱⴱⴱ
[.28, .61] .45
a
Not included 13,345 38 489.64
ⴱⴱⴱ
[.29, .46] .38
b
Feedback on behavior—immediate
Included 1,316 6 61.10
ⴱⴱⴱ
[.29, 1.07] .68
a
Feedback on behavior—delayed
Included 2,420 11 45.28
ⴱⴱⴱ
[.22, .76] .49
a
Feedback on outcome(s) of behavior—immediate
Included 2,830 17 104.28
ⴱⴱⴱ
[.18, .62] .40
Feedback on outcome(s) of behavior—delayed
Included 1,632 12 36.89
ⴱⴱⴱ
[.14, .91] .52
a
No feedback 7,392 66 435.60
ⴱⴱ
[.22, .48] .35
b
Note. CI confidence interval.
215
PROGRESS MONITORING AND GOAL ATTAINMENT
Q(1) 564.86, p.001, and goal attainment, Q(1) 10.33, p
.01. Interventions had larger effects when the frequency of prog-
ress monitoring and goal attainment were measured objectively
(d
2.32 and 0.44, respectively) rather than by self-reports
(d
1.50 and 0.34, respectively).
Effect sizes were influenced by indicators of the quality of the
primary studies such as the type, success, and quality of random-
ization procedures, and whether participants and experimenters
were blind to condition (see Table 3). In general, and as might be
expected, smaller effects tended to be observed in better quality
studies. The type of sample also influenced effect sizes for fre-
quency of progress monitoring, Q(2) 88.41, p.001, and goal
attainment, Q(2) 10.15, p.01. Interventions had smaller
effects on the frequency of progress monitoring and goal attain-
ment among participants with particular medical conditions (d
1.91 and 0.39) than among the general public (d
2.20 and
0.52). Participants’ age or gender was not associated with effect
sizes (see Table 4).
Discussion
Control Theory and other frameworks for understanding self-
regulation propose that monitoring goal progress is crucial for
effective goal striving and promotes goal attainment. Whereas
other “core” self-regulatory processes such as goal setting and
responding to discrepancies have been the subject of meta-analytic
reviews (e.g., De Ridder et al., 2012; Gollwitzer & Sheeran, 2006;
Hagger et al., 2010; Locke, Shaw, Saari, & Latham, 1981;
McEachan et al., 2011; Ouellette & Wood, 1998; Sheeran, 2002;
Webb & Sheeran, 2006), the impact of interventions on the fre-
quency of progress monitoring and rates of goal attainment has not
been quantified. As a result, it has been difficult to evaluate the
role of progress monitoring in shaping goal attainment. The pres-
ent review provided this evaluation and observed a large-sized
effect of interventions on the frequency of progress monitoring and
a small-to-medium-sized effect on goal attainment.
Interventions designed to promote progress monitoring were
highly effective at increasing monitoring frequency, and generated
an effect size that was more than twice the magnitude of a
conventional “large” effect (d
1.98). This finding raises the
question, why were interventions designed to promote progress
monitoring so effective? One answer may be “the ostrich prob-
lem,” or peoples’ motivated avoidance of information concerning
goal progress (Webb et al., 2013). Webb et al. suggested that
relatively few people spontaneously monitor their household en-
ergy consumption, check their bank balance, keep track of their
food intake, or generally take stock of their current standing
relative to their goals (see also Liberman & Dar, 2009). The
present findings thus indicate that there is considerable scope for
improving monitoring frequency.
Prompting progress monitoring had a small-to-medium-sized
effect on rates of goal attainment. Furthermore, changes in fre-
quency of progress monitoring mediated the relationship between
interventions and goal attainment. These findings confirm the
importance of progress monitoring as a key mechanism by which
people strive for goals (Burnette, O’Boyle, VanEpps, Pollack, &
Finkel, 2013; Carver & Scheier, 1982; Carver, Johnson, Joormann,
& Scheier, 2015; de Bruin et al., 2012; Ford, 1987; Louro et al.,
2007; Miller et al., 1960; Powers, 1973; Powers, Clark, & McFar-
land, 1960a, 1960b), and have both conceptual and practical im-
portance. At the conceptual level, the findings suggest that models
of behavior that posit a direct relationship between intentions and
behavior (e.g., the Theory of Planned Behavior, Protection Moti-
vation Theory) neglect a key volitional process that intervenes
between goal setting and goal attainment—namely, monitoring
goal progress (for reviews, see Gollwitzer & Sheeran, 2006;
Sheeran et al., 2005; Sheeran & Webb, 2011). It is notable that
progress monitoring had an impact on goal attainment (d
0.40)
that is comparable with that reported for goal intentions (d
0.36 according to Webb & Sheeran, 2006), suggesting that effec-
tive goal striving requires that people not only decide upon an
appropriate goal (e.g., “What is it that I want to achieve?”), but
also that they compare ongoing behavior or the current status of
the outcome to that goal (e.g., “Where do I currently stand with
respect to this goal?”). Monitoring goal progress serves to identify
discrepancies between the current and desired state, which enables
people to decide how best to allocate effort among salient goals
(Carver & Scheier, 1982; Louro et al., 2007), and when and how
to exercise restraint or initiate corrective action (Fishbach et al.,
2012; Myrseth & Fishbach, 2009). In light of the present review,
we contend that models concerned with specifying the determi-
nants of intentions such as the Theory of Planned Behavior might
profitably be extended to integrate the important role of monitor-
ing goal progress. Such integration holds the promise of a more
complete understanding of goal-directed behavior. At the practical
level too, the present findings could serve to improve behavior
change interventions by affording new targets for intervention
beyond behavioral intentions (see also de Bruin et al., 2012).
The Impact of Dimensions of Progress Monitoring on
Goal Attainment
By identifying the key dimensions on which efforts to monitor
progress may differ (see Table 1), we were able to code these
features of interventions and compute associations with both the
frequency of monitoring and goal attainment. These analyses re-
vealed support for our hypotheses concerning the match between
the focus of progress monitoring and the dependent variable.
Specifically, prompting participants to monitor their behavior had
a significant impact on rates of behavioral performance but not on
outcomes, whereas prompting participants to monitor outcomes
had a significant impact on outcomes, but not on behavior. This
finding can be explained by a goal systems perspective (Kruglan-
ski et al., 2002), which suggests that goals can be achieved via a
range of behavioral means. For example, the goal to reduce house-
hold energy bills could be achieved by taking shorter showers, by
replacing light bulbs with low energy alternatives, or by fitting
solar panels. Therefore, monitoring outcomes could prompt a
range of corrective actions, and so is more likely to influence
outcomes than the performance of any specific behavior. In con-
trast, monitoring behavior (e.g., the length of a shower) is likely to
influence the performance of that behavior, but may not influence
the outcome, particularly if the outcome can be influenced by a
variety of behaviors. Monitoring behavior versus outcomes could
also differentially influence commitment such that people who
monitor outcomes become more committed to the goal and are
prepared to substitute different means to attain relevant outcomes,
whereas people who monitor a particular behavior become com-
216 HARKIN ET AL.
mitted only to that particular means of goal attainment (Kruglan-
ski, Pierro, & Sheveland, 2011).
Progress monitoring had larger effects on goal attainment when
the information gleaned from monitoring was reported or made
public, than when it was kept private. This finding may indicate
that monitoring progress in public increases the amount of effort
that people put into striving for the goal—due to a sense of public
commitment (Cialdini, 2001; Kiesler, 1971), personal accountabil-
ity (e.g., Stuckey et al., 2011), presentational concerns (Schienker
et al., 1994), or experimenter demand (Zizzo, 2010). Future re-
search might directly compare reported versus not reported forms
of progress monitoring in order to assess whether these mecha-
nisms mediate the effects of monitoring in public on outcomes.
We also observed larger effects of progress monitoring on goal
attainment when the information obtained from monitoring was
physically recorded than when it was not. There are a number of
possible explanations for this effect. First, recording progress may
increase the likelihood that the information is remembered, both in
terms of strengthening the encoding of information and also facil-
itating retrieval. Second, given that information on goal progress
may reflect badly on the self (Carlson, 2013; Karlsson, Loewen-
stein, & Seppi, 2009; Northcraft & Ashford, 1990; Tuckey,
Brewer, & Williamson, 2002; Zuckerman, Brown, Fox, Lathin, &
Minasian, 1979) or demand undesired action (Sweeny, Melnyk,
Miller, & Shepperd, 2010), people may ignore or reject such
information (for a review, see Webb et al., 2013). Thus, it is not
enough merely to monitor progress—the person must also face up
to what the information shows (akin to self-confrontation, Bailey
& Sowder, 1970; Schoutrop, Lange, Hanewald, Davidovich, &
Salomon, 2002). Information may be more difficult to ignore or
reject when it has been recorded (Roggeveen & Johar, 2002),
thereby reducing the scope for self-deception (Greenwald, 1997).
Finally, recording information may increase goal commitment
because evidence suggests that people feel more committed and
certain about decisions that are expressed via action (Cioffi &
Garner, 1996). Future research should examine the mechanisms
that underlie the utility of the recording information on progress,
and the circumstances in which such recording is likely to be
particularly beneficial.
The nature of the reference value generally did not influence the
effect of interventions on goal attainment. Although there was
some evidence that participants prompted to evaluate their prog-
ress with respect to a past state did so more frequently than those
prompted to evaluate their progress with respect to a desired future
state, both effect sizes were very large and the use of these
different reference values did not influence the effect of the inter-
ventions on goal attainment. One intriguing hypothesis that we
were unable to test here is that different reference values are suited
to different stages of goal striving. Research by Bonezzi, Brendl,
and De Angelis (2011) suggests that people tend to adopt their past
state as a reference value in the early stages of goal pursuit (i.e.,
people ask themselves “How far have I gone?”) and adopt the
desired end state as their reference point when nearing the goal
(i.e., people ask themselves “How far do I have to go?”). It was
also the case that there were insufficient studies to examine the use
of others’ performance as a reference value. Given the pervasive-
ness of social comparison (e.g., Collins, 1996; Pinkus, Lockwood,
Schimmack, & Fournier, 2008; Suls, Martin, & Wheeler, 2002)
and evidence attesting to the substantive impact that others’ per-
formance can have on self-regulatory processes (e.g., Aarts, Goll-
witzer, & Hassin, 2004; Fitzsimons & Finkel, 2010; Shah, 2003a,
2003b), studies investigating the effects of monitoring goal prog-
ress with respect to others’ performance are a priority for future
research.
Although participants prompted to actively monitor their prog-
ress did so more frequently than those who passively monitored
progress, both active and passive forms of monitoring influenced
goal attainment, and there was no difference in their relative
efficacy. Similarly, although participants who were prompted to
monitor distance from the goal did so more frequently than those
prompted to monitor their rate of progress toward their goal, both
forms of monitoring were equally effective in promoting goal
attainment. However, only three primary studies prompted partic-
ipants to consider their rate of goal progress and so further tests are
needed to draw firm conclusions, especially as small samples tend
to bias the effect size upward (Coyne et al., 2010). Indeed, few
empirical studies have explicitly investigated whether people are
sensitive to the rate of discrepancy reduction (see, however, Goll-
witzer & Rohloff, 1999; Hsee & Abelson, 1991, for notable
exceptions).
Taken together, our findings provide some of the first tests of
theoretical distinctions that have been drawn between different
types of progress monitoring (e.g., Anseel et al., 2015; Ashford &
Cummings, 1983; Wilde & Garvin, 2007) and information seeking
(e.g., Berger, 2002), and make it clear that monitoring is not a
unitary process. Rather, there are multiple ways in which people
can assess their goal progress. The dimensions identified here may
provide a useful impetus for examining the impact of specific
forms of progress monitoring on goal attainment. In particular,
there is a need for studies that directly compare the efficacy of
different forms of progress monitoring and identify the mecha-
nisms by which they influence goal attainment.
The present review found that, while all of the techniques and
tools for promoting progress monitoring were effective, some were
more effective than others. Ideally, studies should compare the
effects of different methods of progress monitoring for the same
goal (e.g., Helsel et al., 2007, compared the impact of completing
detailed diaries vs. abbreviated diaries on monitoring food intake).
Indeed, we intended to conduct such analyses in the present
review; however, there were insufficient studies to permit mean-
ingful comparisons. Even for the most frequently studied goal
(weight loss; k50, 36% of studies), only three methods of
progress monitoring (written diaries, websites, or PDAs) were
used in at least three studies. Thus, caution is warranted in drawing
conclusions about the effectiveness of different methods of prog-
ress monitoring. Research that explicitly compares different meth-
ods of monitoring goal progress will lead to more conclusive
findings.
Previous reviews have found that interventions that incorporated
additional BCTs alongside monitoring goal progress tended to
have larger effects than interventions that prompted progress mon-
itoring alone (e.g., Dombrowski et al., 2012; Febbraro & Clum,
1998; Greaves et al., 2011; Michie et al., 2009). Our findings
support this idea—interventions that included goal setting, action
planning, and some forms of feedback (namely, immediate feed-
back on behavior) alongside progress monitoring engendered
larger effects than interventions that did not incorporate these
additional BCTs. These findings could arise because additional
217
PROGRESS MONITORING AND GOAL ATTAINMENT
BCTs target different self-regulatory processes that serve to bolster
the impact of progress monitoring. That is, goal setting may help
people to set appropriate reference values (Bandura, 1991; Carver
& Scheier, 1982), immediate feedback on behavior facilitates
attention to and reinforces ongoing performance (Ashford, 1986;
Della Libera & Chelazzi, 2006; Kluger & DeNisi, 1998), and
planning helps people to act on discrepancies (for reviews, see
Carraro & Gaudreau, 2013; Gollwitzer & Sheeran, 2006). More
generally, these findings underline the idea that theoretically sup-
ported combinations of BCTs can be particularly effective in
promoting goal attainment (Michie et al., 2009; Prestwich, Webb,
& Conner, 2015).
The Role of Methodological Factors
The nature of the focal goal had a substantial impact on the size
of the effects observed in the present review. Progress monitoring
appeared to have larger effects on goal attainment when it was
used to manage specific medical conditions (e.g., asthma, Bateman
et al., 2008; blood pressure, Imai et al., 2003; diabetes, Allemann
et al., 2009; Norris, Engelgau, & Narayan, 2001), compared with
other health goals (e.g., weight loss or dieting). It was notable that
most of the studies that met our inclusion criteria focused on health
goals. In fact, only one study could be included in the present
review that asked participants to monitor in a domain unrelated to
health (time spent doing different activities, Runyan et al., 2013),
despite reviews attesting to the benefits of self-monitoring in
clinical, educational, and environmental domains (e.g., Abrahamse
et al., 2005; Febbraro & Clum, 1998; Korotitsch & Nelson-Gray,
1999). One reason why studies in these domains could not be
included in the present review is that they tended not to examine
the impact of interventions on the frequency of progress monitor-
ing (one of the key criteria for inclusion in the present review)
meaning that it is difficult to attribute the effects of such interven-
tions to changes in the frequency of progress monitoring. Such
measures should be included in future studies in these domains.
Effect sizes in the present review were not influenced by the age
or gender of the sample, but were influenced by the type of sample.
Specifically, effect sizes tended to be smaller for participants with
particular medical conditions compared with members of the gen-
eral public. It is possible that chronic health conditions make it
harder to monitor progress or change behavior, or alter the impact
of behavior on health outcomes because of genetic or physiolog-
ical factors. In either case, it is worth noting that the interventions
still had substantive effects on progress monitoring and goal at-
tainment even for participants with chronic health conditions.
Measurement features and indicators of study quality also in-
fluenced effect sizes. Effect sizes were larger when progress mon-
itoring and goal attainment were measured objectively, rather than
by self-report. This finding may suggest caution in using self-
report measures as interventions can influence outcomes in ways
that are not amenable to self-report (cf. Maidment, Jones, Webb,
Hathway, & Gilbertson, 2014). Consistent with previous meta-
analyses (e.g., Wood et al., 2008), interventions had smaller effects
when participants and experimenters were blind to conditions. This
finding suggests that expectations about the benefits of progress
monitoring can influence both the frequency of monitoring and
goal attainment. Success and quality of randomization were asso-
ciated with larger effects. Fortunately, studies with poor random-
ization procedures were in the minority and effect sizes remained
robust across different types of randomization.
10
Finally, publica-
tion bias had a modest influence in the present review and effect
sizes remained substantively unaltered in studies that had adequate
power according to Coyne et al.’s (2010) criterion.
Limitations
Conclusions drawn from the present meta-analysis must be
mindful of the evidence base upon which it stands. After a search
that started with over 22,000 records, 138 tests provided data on
the effect of interventions prompting progress monitoring on goal
attainment and could be included in the meta-analysis. These tests
provided a robust evidence base for answering our key research
questions, but we acknowledge the paucity of data concerning
effects in particular domains (e.g., behaviors not related to health),
the impact of particular types of monitoring (e.g., only three
studies examined the effect of monitoring the rate of progress), and
how moderators combine to influence effect sizes (e.g., how dif-
ferent dimensions of progress monitoring can best be combined to
promote goal attainment). It is also worth noting that our analyses
of moderators did not correct for the increased Type I error rate
associated with conducting multiple tests. This was because most
of the effects did not derive from the same sample and our focus
was on determining the magnitude of effects, rather than signifi-
cance testing. Finally, we acknowledge that some moderators
examined here are likely to be correlated (e.g., studies that prompt
participants to monitor the outcomes of their behavior may also be
more likely to also ask participants to physically record this
information). Future research might address such potential multi-
collinearity by independently manipulating features of progress
monitoring and examining the effects on goal attainment.
We also recognize that the nature of the control condition had an
important influence on the effect size observed for the frequency
of progress monitoring. Arguably, control conditions in which
participants are not prompted to monitor their progress provide the
clearest test of the impact of monitoring goal progress on out-
comes. However, such studies rarely measured frequency of prog-
ress monitoring among participants in the control condition, and so
we had to assume zero progress monitoring when computing effect
sizes. An alternative approach would have been to substitute the
mean frequency of progress monitoring from studies where these
data were available. Unfortunately, the primary studies differed in
too many substantive respects to permit this imputation strategy.
We acknowledge that assuming zero levels of progress monitoring
in the no-progress-monitoring control conditions may be subopti-
mal. However, in the absence of a viable alternative strategy, and
in the light of evidence that people rarely monitor their progress
unless prompted to do so (Liberman & Dar, 2009; Webb et al.,
2013), we consider that the approach adopted here best captures
the nature of the target processes. Further observational