Applied and Preventive Psychology 11 (2005) 141–156
Modularity in the design and application of therapeutic interventions
Bruce F. Chorpitaa,∗,1, Eric L. Daleidenb,1, John R. Weiszc
a Department of Psychology, University of Hawai‘i at M¯anoa, 2430 Campus Road, Honolulu, HI 96822, USAb Child and Adolescent Mental Health Division, Hawaii Department of Health, USA
The promotion of human competencies and alleviationf human suffering are highly complex challenges.Simon1996)has argued that most complex problems are amenableo multiple representations that have different strengths andeaknesses in guiding problem solution. Thus, it is not sur-rising that the psychological interventions to address thesehallenges represent a diversity of forms. For example, earlyfforts to understand and organize therapeutic activities wereased on elaborating major psychological theories and princi-les (e.g., Psychodynamic, Humanistic, and Behavioral) thattherapist used to “design” treatments within the therapeuticetting (e.g.,Freedheim, 1992). In this approach, the therapistses theoretical principles to develop strategies or responses
o the client as events occur in the therapy session. Alterna-ively, some contemporary interventions have tended to focusn the codification of therapeutic activities into standardizedrotocols (e.g., manuals), which are designed prior to ther-py and tested in tightly controlled settings (Weisz, 2004).hese traditions represent two extremes regarding the locus
of treatment design (i.e., treatment setting or researchratory).
Designing treatment based on theoretical knowlewithin the treatment setting has the advantages of prommore highly individualized services and providing the thpist with an expert role, but it may be excessively suscepto clinical judgment biases, inconsistency, and limitatin generating cumulative knowledge. Designing treatmin the research laboratory supports stronger empiricalof efficacy, promotes generalizable knowledge and theof actuarial decision-making, but may not be as relevaindividual clinical situations and is less preferred by mclinicians (Addis & Krasnow, 2000).
Along with these contrasting approaches to psychothedesign, a third genre has evolved representing somewcenter point on this continuum. The majority of approacwithin this genre fall under the heading of “prescriptiapproaches, and they are partially designed in the laborbut also allow for systematic design decisions to be mduring the course of therapy. Examples include protocolscan be applied across multiple theoretical orientations u
1 These authors contributed equally in the preparation of this manuscript.
a set of general principles to match particular strategies orstyles to client characteristics (Beutler & Harwood, 2000;cf. Norcross & Beutler, 2000), methods for systematically
142 B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156
developing interventions based on a cognitive behavioral caseformulation (Persons, 1991; Persons & Tompkins, 1997) orbased on individual client characteristics—e.g., for depressedteens (Curry et al., 2000), children with autism (Durand,1990; Durand & Crimmins, 1988), anxious school refusalin youth (Burke & Silverman, 1987; Kearney & Silverman,1990), and childhood generalized anxiety disorder (Eisen &Silverman, 1998).
The above list is necessarily a partial sampling of the widevariety of approaches that use a prescriptive or matching strat-egy as a core principle. The collective ideas that inspire suchapproaches are promising and represent a movement towardreconciling earlier traditions of psychotherapy design (e.g.,individualization within the clinic setting) with more recentdesign trends (e.g., manualization). Although some of theliterature is quite positive regarding some prescriptive strate-gies (e.g.,Eisen & Silverman, 1998), there is also evidencethat flexibility in interventions implicitly harbors possiblerisks (e.g.,Schulte, Kunzel, Pepping, & Schute-Bahrenberg,1992), and debates cause one to ponder the challenges asso-ciated with prescription at the level of specific individuals(e.g.,Jacobson et al., 1989; Persons, 1991; Wilson, 1996).
Despite their different locations on the design continuum,the various traditions of psychotherapy approaches all pre-sumably have highly similar goals in mind (e.g., reducedhuman suffering, improved functioning, higher quality ofl niten Thisr mentd ongt of theb odelc signt es-s tory)a itera-t tive-n nP ures,1 es,1 theq
ncyb effi-c d orr rs tot eal-w colt
esigna f them rityi f thec l con-t ingo od-
ularity begins with a more precise definition and illustrationof modular treatment design. Modularity is then evaluated interms of its efficiency, effectiveness, and efficacy as a treat-ment design principle.
1. What is modularity?
1.1. Defining principles
Modularity is not a new concept in design or in psy-chotherapy. Some sophisticated examples have emerged ofpsychotherapy protocols that describe themselves as consist-ing of modules or being modular (e.g.,Carroll, 1998; Clarke,Lewinsohn, & Hops, 1990; Curry et al., 2000; Wells & Curry,2000). Aside from an author’s proclamation that a proto-col is modular, what characteristics are central to a “truly”modular protocol? Generally, modularity refers to break-ing complex activities into simpler parts that may functionindependently. More specifically, modules are self-containedfunctional units that connect with other units, but do not relyon those other units for their own stable operations. Modulardesigns have been described as consisting of visible designrules (i.e., standardized guidelines for how modules interactwith each other) and hidden design parameters (i.e., fea-t ons;B n-t rtsiI nter-d Fore as ah sid-e grald iento
1par-d-iallysions,er-tc.).iththeg.,
2 chndedtion
elax-tedust
ision
ife), and although diverse, presumably rely upon a fiumber of techniques in pursuit of these outcomes.aises the question of whether a general model of treatesign exists that allows for “continuous scaling” am
hese specific approaches, to take advantage of someest features of each. In other words, a single design mould perhaps yield a protocol at either extreme of the deraditions (i.e., highly individualized, designed in the sion versus highly standardized, designed in the laboras well as anywhere in between. The psychotherapy l
ure has outlined in some detail the dimensions of effecess and efficacy of specific treatments (e.g.,Task Force oromotion and Dissemination of Psychological Proced995; Task Force on Psychological Intervention Guidelin995), and these dimensions can be suitably applied touestion of psychological intervention design as well.
We also introduce the additional dimension of efficiey which to evaluate modular design. In this context,iency refers to the ability of a protocol to be designee-designed with reduced time or cost, effectiveness refehe ability of a protocol to be generalizable or feasible in rorld contexts, and efficacy refers to the ability of a proto
o achieve its desired effect or outcome.The present paper proposes that one specific d
pproach, modularity, represents a potential unifier oultiple design traditions noted above. Although modula
n psychological intervention design cannot address all ohallenges raised so far, it appears to provide an optimaext for their eventual empirical resolution while conferrther potential advantages as well. Our exploration of m
ures within modules that govern their internal operatialdwin & Clark, 1997). Modular designs have been co
rasted with so-calledintegral designs that combine panto a single functional whole (cf.Ulrich & Ellison, 1999).ntegral designs are characterized by a high level of iependence and minimal differentiation among parts.xample, a traditional therapy treatment manual writtenighly interconnected, cumulative narrative would be conred integral in nature. Removal of one piece of that inteesign might render the remainder of the protocol deficr unusable.
We define modularity to include four key properties:
. Partial decomposability(cf. Simon, 1996) refers to thenotion that a complex system may be at leasttially divided into meaningful functional units (i.e., moules). For example, a treatment protocol may be partdecomposed into various types of units such as seswithin-session activities (e.g., homework review), or thapeutic practices (e.g., relaxation, problem-solving, eAlthough some division of this nature is possible wintegral designs, the difference with modularity is thatsubdivision results in units with highly similar form (e.sessions, skills, paragraphs, exercises).
. Proper functioningsignifies that the operation of eamodule in the design is expected to produce the interesult. For example, if a therapy protocol uses relaxaas an intervention to reduce autonomic arousal, a ration module with proper functioning would be expecto reduce client arousal. This implies that modules mhave a specified purpose, and are not simply a subdiv
B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156 143
based on other considerations. For example, a therapy pro-tocol designed to span four months might be divided intofour units, involving similar content, with Unit 1 corre-sponding to the first month, Unit 2 to the second month,etc. This type of subdivision would not be modular in thatthe proper functioning of each module is not specified.Rather, in this example, the “modules” are simply subdi-visions that structure how long to engage in each aspectof the protocol.
3. Standardized interfacedenotes that modules within thedesign connect or communicate with each other in astructured fashion. In the most basic sense, this propertyis similar to the property of children’s “lego” buildingblocks—pieces are designed so that one can plug into thenext. This property thus highlights that although indepen-dently structured, modules interact to produce a wholewith better functioning than the sum of the parts, just asbuilding blocks can make a house. Among other things,the standardized interface allows the ability to rearrangemodules without problems regarding how they connect(e.g., performing “relaxation” before “exposure” versus“exposure” before “relaxation”).
The “connection” of therapy modules involves morethan just their sequencing. It also involves the neededinformation exchange from one module to the next. Forexample, a module might begin with a homework review
nizeedign-eriazedinte-
t is,(a)
fully,r and
ofmore.
4 stailsher-duleany
olve, deeplect-uldtheherennerif a
ax-uire, but
the specific nature of those skills would not be outlinedin the public speaking module (i.e., that information is“hidden” in the relaxation module). Keeping informationself-contained in this manner allows great flexibility inthe arrangement or interchange of modules. Thus, in thisexample, assuming information hiding, one could substi-tute a breathing module for a muscle relaxation module,without affecting the subsequent module for practice ofpublic speaking.
When defined in terms of these properties, modularity isnot an all-or-nothing feature of designs but can be described indegrees (Mikkola & Gassmann, 2003). Although variabilityexists, the current “industry standard” for an evidence-basedtreatment protocol tends toward a highly integral design.1 Inparticular, most current designs commonly lack the proper-ties of information hiding and standardized interfaces. Thatis, although many protocols can be sub-divided by sessions(i.e., possess partial decomposability), sometimes with well-defined purposes for each subdivision (i.e., possess properfunctioning), these sessions usually contain details aboutother sessions (i.e., lack information hiding), and sessionsare rarely designed to allow them to immediately precede orfollow any other session (i.e., lack standardized interfaces).
1.2. Modular protocol components: content andc
tionic nr urald con-t ist tot hatm ypi-c pists lientt lar-i tiona o bed thec ss as fol-l llown tc. tob ard-i tailsf nt ofo tionh s of
in am is usem ons ina
procedure. Upon completion, that module might orgathe following information: (a) the module has concludsuccessfully, (b) whether the module involved the assment of homework, and if so, (c) the content of and critfor reviewing that homework. If there is a standardiinterface, the subsequent module should be able tograte this information into its own procedures. Thanext module would prompt the therapist to check onwhether the previous module was completed success(b) whether homework was assigned, and (c) whethehow to review that homework. Thus, thestandardizedinterfacehelps to specify how much and what forminformation can pass among modules, and ensuresgenerally, that one module can connect with another
. Information hiding (Parnas, 1972) is also known a“encapsulation” and refers to keeping the specific deof operation entirely within a module. For example, a tapist using a protocol that includes a relaxation mowould not need to know how that module works to useof the other modules. The relaxation module could invany number of strategies, such as breathing exercisesmuscle relaxation, or pleasant imagery, but prior to seing the module to use with a client, the therapist woonly need to know that a module was available forpurpose of relaxation. Further, other modules elsewin the protocol would not be dependent upon the main which that relaxation was achieved. For example,module for practicing public speaking followed this relation module, the public speaking module might reqthat the client use previously learned relaxation skills
oordination modules
Given these four defining principles, a further distincs warranted in psychotherapy design betweencontentandoordinationmodules.Contentmodules contain informatioelated to therapeutic activities and are similar to procedescriptions in typical therapy manuals. For example, a
ent module might contain the procedures for a theraprain a client in how to be more assertive with others. Todule would therefore consist of specific instructions, t
al of many manualized protocols, detailing how the therahould perform various activities and exercises with the co achieve that goal. According to the principles of moduty, those instructions would need (a) to be a unit that funcs part of a larger system (partial decomposability), (b) tesigned to bring about their intended aim of traininglient to be assertive (proper functioning), (c) to possetandardized structure that would allow it to precede orow other modules in that larger system and would aeeded information about homework, goal attainment, ee carried along smoothly from module to module (stand
zed interface), and (d) to have all of its operational deully self-contained, so that the omission or rearrangemether modules in the system would not affect it (informaiding). Thus, content modules are the building block
1 Although practitioners may report that they use integral manualsodular fashion, doing so does not make integral manuals modular. Therely reflects the natural tendency of those clinicians to make decisi“modular” fashion, even when using integral protocols.
144 B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156
modular interventions that contain the detailed descriptionsof therapy procedures.
Coordination modules, on the other hand, act as thecement that binds those building blocks together. Coordina-tion has been described as “managing dependencies betweenactivities” (Malone & Crowston, 1994, p. 90). Thus,coordi-nation modules outline the algorithms for managing deci-sions about whether and when to use the variouscontentmodules. For example, a coordination module might instructa therapist to select a relaxation module if the client has ahigh degree of muscle tension or to select a problem-solvingtraining module if the client has poor problem-solving skills.These algorithms can be outlined in narrative form or can beoutlined in form of flowcharts, depending on the design struc-ture of the coordination module. Although a narrative detailof the clinical algorithm has the advantage of not requiring thedefinitions of symbols (i.e., a flowchart “legend”), a flowcharthas considerable advantages in terms of its ability to refer-ence other coordination modules and its ability to representcomplex algorithms efficiently.
Coordination modules adhere to the same four princi-ples of modularity outlined above. For example, they, too,should operate as part of a protocol that can be divided intoindependent functional units (i.e., partial decomposability).Coordination modules should specify a particular function(e.g., to manage the sequence and selection of content mod-u erf owst rt tol tan-d d bes (i.e.,i ed,a rga-n cog-n apyp y,1 oor-d izesm sonalT le,& notn rdi-n lient,a n fort
1
iallym volvec canl racto ectlyra
1.3.1. Shared resourcesThe first dependency involvesshared resourcesin therapy.
Whenever two activities have the potential to use the sameresource there is an implicit dependency that needs to becoordinated. In other words, if one activity needs a resource,and another activity requires the same resource, a decisionneeds to be made about sharing (e.g., whether the activitiesshould “take turns,” or whether one activity should be deniedaccess to the resource). In therapy, the types of resources to bemanaged include direct service time, client memory capac-ity, therapy cost, office space, etc. For example, engaging inmany therapeutic activities in a given hour might tax a client’scapacity to remember the information, require more time thanis available, etc. This can be managed by selecting and pri-oritizing modules for implementation (e.g., implementing asingle module now, another one later, and so forth).
1.3.2. Task–subtask dependenciesSecond,task to subtaskdependencies exist in therapy,
such that overall goals must often be decomposed into activ-ities or subgoals. Any time one therapeutic activity must beperformed as part of a larger set of activities, a dependencymust again be coordinated. For example, if the goal to pro-duce a decrease in anxious responding (i.e., thetask) requiresmodules for psychoeducation, relaxation, and exposure (i.e.,subtasks), and if knowledge of anxiety and relaxation skillsa thesem
1in
t at isu e fors rds,b ightb mod-u ring,w thec ordi-n ncingo s area
1in
a ther.F con-s reast ouldn
2
ocolc n D.
les for Attention Deficit Hyperactivity Disorder; i.e., propunctioning). They should ideally have a structure that allhem to reference each other (e.g., allowing one flowchaink to another one) as well as content modules (i.e., sardized interface). Finally, coordination modules shoulelf-contained such that they can operate independentlynformation hiding). For example, if properly self-contain
“depression” coordination module that selects and oizes content modules (“seeking alternative solutions,” “itive restructuring”) to produce a cognitive behavior therrotocol for depression (e.g.,Beck, Rush, Shaw, & Emer979) could be exchanged with a different “depression” cination module with an algorithm that selects and organodules representing the basic elements of Interperherapy for Depression (Klerman, Weisman, RounsavilChevron, 1984). The therapist using the system would
eed to know about the workings of the “depression” cooation module until that therapist treated a depressed cnd could simply assume that the necessary informatio
reating depression was contained therein if needed.
.3. Multiple dependencies in the modular approach
Coordination modules have more to manage than initeets the eye. Relationships among content modules in
ontingencies or “dependencies,” i.e., conditions thatimit or constrain the way modules are allowed to inter connect. At least four types of dependencies seem direlevant to therapeutic activities (Malone & Crowston, 1994),nd these are outlined here.
re components of effective exposure, then selectingodules and coordinating their order is essential.
he therapy, where one activity produces something thsed by a subsequent activity. For example, a modulelf-monitoring can yield such products as thought recoehavior records or narrative diaries. Such products me needed for proper implementation of subsequentles, such as self-reinforcement or cognitive restructuhich presumably involve interpreting and acting onontents of the records or diary. One procedure for coating producer–consumer relationships involves sequef modules to ensure that such prerequisite conditionlways satisfied.
.3.4. Simultaneity constraintsFourth,simultaneity constraintsmay exist when certa
ctivities must occur together or may never occur togeor example, a module for response prevention might betrained only to occur simultaneously with exposure, wheherapy engagement and therapy termination modules wot be allowed to co-occur.
. Example of a modular protocol
Our specific example of a modular psychotherapy protomes from an effectiveness trial sponsored by the Joh
B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156 145
and Catherine T. MacArthur Foundation (i.e., the “MATCH-ADC: A Modular Approach to Therapy for Children withAnxiety, Depression and Conduct Problems”). The MATCHprotocol is designed to target anxiety disorders, depression,and disruptive behavior in children aged 8–13. As such, itcontains modules that represent common cognitive behav-ioral and behavioral parent training strategies for these disor-ders. One part of the study design involves the comparison ofintegral, evidence-based interventions (e.g.,Barkley, 1997)to a modular intervention system (i.e., MATCH;Chorpita& Weisz, 2003). The design of the MATCH protocol wasintended to incorporate the principles of modularity, whileemploying the same basic therapeutic strategies as the com-parison interventions. Developing original intervention con-tent for the MATCH protocol involved a number of steps andconsiderations, which are outlined below.
2.1. Example: content modules
2.1.1. Addressing decomposability: defining andidentifying discrete practice elements
To construct thecontentmodules, we first developed alist of 55 discrete psychological procedures (i.e., “practiceelements,”Chorpita, Daleiden, & Weisz, 2005; Child andAdolescent Mental Health Division, 2003). These 55 practiceelements were nominated by several panels of practitioners,i weres asedp neda (e.g.,“ geri fory ump-t e.g.,u ithinp (c)d mon.B refert that“ , andi rate-g tifiedj cols[ nte-gt ulei ntaino f thed twop g”).
cols(Ka ET,”Ww ulted
in the identification of 29 practice elements. The 29 ele-ments included some that were appropriate across differ-ent treated conditions—e.g., rapport building appeared intwo protocols, and family engagement appeared in twoprotocols. Other practice elements were appropriate onlyfor anxiety—i.e., child psychoeducation for anxiety, par-ent psychoeducation for anxiety, self-monitoring for anxi-ety, exposure, cognitive/coping for anxiety, and maintenancefor anxiety skills. Other practice elements were appropriateonly for depression—i.e., child psychoeducation for depres-sion, parent psychoeducation for depression, problem solv-ing, activity selection, skill building, social skills training,cognitive/coping for depression, and maintenance for depres-sion skills. Another set of practice elements was appropri-ate only for treatment of conduct problems via behavioralparent training—i.e., parent psychoeducation for disruptivebehavior, parent monitoring for disruptive behavior, limitsetting, parent praise, selective attention, tangible rewards,time out, antecedent control, and maintenance for parentingskills.
Once these elements were identified, their functions weredesigned into discrete modules. For example, a “gettingacquainted” module was constructed to reflect the practiceelement of relationship/rapport building. This module wasapplicable to both anxiety and depression conditions in theMATCH protocol. In some instances, practice elements thatw theu n allo a sin-g ditiona
2f
formo stent“ e ofs rity,a lema dulef jec-t sts too ducen , per-f t asn od-u ts oft notb d ast ulefl neda
them self( the
ntervention developers, and other domain experts, andelected as those most likely to appear in evidence brotocols or in usual care services for youth. We defipractice element as a clinical technique or strategy
time out,” “relaxation”) that can be used as part of a larntervention plan (e.g., a manualized treatment programouth depression). This definition is based on the assions that (a) practice elements can be explicitly defined (sing a definition or coding manual), (b) their presence wsychological interventions can be reliably coded, andifferent treatments may have practice elements in comecause “modules” and “practice elements” can each
o discrete therapy procedures, it is important to clarifypractice elements” describe the strategies themselvesmply nothing about the design features of how those sties are codified. Thus, practice elements can be iden
ust as easily in integral protocols as in modular protoe.g., “cognitive restructuring” is a practice element in an iral CBT protocol for depression (Beck et al., 1979) and in
he modular MATCH protocol]. On the other hand, a mods best thought of as a structured “container” that can cone or more practice elements (e.g., the first module oepression portion of the MATCH protocol contains theractice elements of “engagement” and “rapport buildin
Next, the content of the integral comparison protoe.g., “Defiant Children,”Barkley, 1997; “Coping Cat,”endall, Kane, Howard, & Siqueland, 1990, “Primarynd Secondary Control Enhancement Training-PASCeisz, Weersing, Valeri, & McCarty, 1999) was codedith respect to the practice elements used. This res
ere highly integral in the comparison protocols (e.g.,se of rewards, which showed up in multiple sessions if the integral comparison protocols), were designed asle module that could be accessed for any treatment cons needed.
.1.2. Addressing standardized interface and properunctioning: using templates and meta-data
Next, templates were developed to standardize thef the modules and to provide the therapist with a consilook and feel” across all treatment conditions. This typtandardization is one of the core properties of moduland allows for the therapist to move freely from one probrea to another without experiencing a disruption in mo
ormat. The MATCH template included: a statement of obives, a list of needed materials, and prompts to therapibtain a measure of progress, review homework, introew material, rehearse new material, assign homework
orm a rapport-building activity, and brief the family. Jusot every treatment case was expected to involve all mles, not every module was expected to include all par
he template. For example, “review of homework” woulde an appropriate template entry for a module selecte
he first therapy session. A partial illustration of a modrom the MATCH protocol is presented inFig. 1, which fol-ows the template for many of the common activities outlibove.
A separate part of the template involved outliningodule “meta-data,” or information about the module it
ADL, 2003). Examples of meta-data include a list of
146 B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156
Fig. 1. An abbreviated content module for childhood depression.
template contents (including number and type of exercises,role plays, etc.), a statement of conditions of use and expectedfunctioning (see modularity property 2, above), sequencingconstraints (whether a module needs to co-occur with anothermodule, or be used at a particular point in a protocol), infor-mation transfer parameters (e.g., whether and what type ofhomework was assigned, as well as any other things thatwould require information to be obtained from prior mod-
ules or preserved for subsequent modules; see modularityproperty 3, above) application boundaries (e.g., age range,cultural limitations), ownership/authorship, references, andannotations. When catalogued, meta-data can be useful byallowing modules from different designers to be combinedin future design efforts, without a loss of understanding ofthe origins, intentions, and general characteristics of eachmodule.
B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156 147
2.1.3. Addressing information hiding: designingindependence into the modules
One challenge that had to be addressed involved design-ing protocol content that was similar in intended therapeuticfunction without the explicitly cumulative material commonto evidence-based manuals organized according to integraldesign principles. For example, a psychotherapy manualmight introduce cognitive restructuring skills early in the pro-tocol, and then review those skills repeatedly in all sessionsthat follow. Each session, therefore, might have traces of whatcould constitute many prior modules. One solution used in thedesign of the MATCH manual involved building conditionalchecks into modules that might reference other modules. Forexample, the module for in vivo exposure asks the therapist tocheck whether any cognitive modules have previously beencovered. If so, exposure is to be performed with some addi-tional enhancements, namely, cognitive exercises (from anearlier module) to precede or follow each exposure trial (e.g.,the child might make a list of negative predictions prior to theexposure trial, counter those predictions before and during theexposure trial, and review the accuracy of those predictionsafter the exposure trial). Otherwise, in vivo exposure is per-formed alone. Such obstacles might not be pronounced whendesigning in a modular format without having to match strate-gies with an existing integral protocol as in our effectiveness
anyhttly
e-
killthe
mndngofatd-toto
ive
ee.urenlete-nte,er-ntlyesof
Yet another issue encountered in the design of theMATCH manual involved how to effectively use charactersor mnemonic acronyms. For example, in the anxiety manualselected for the clinical trial (Kendall et al., 1990), the home-work review is described as a “Show That I Can” (STIC)Task. The homework review in the PASCET depression man-ual does not use such an acronym. Because any such acronymwould have to work universally across all modules, such anacronym was not designed for the MATCH manual.
Similarly, some sessions from the integral PASCETdepression protocol use an acronym that connects all of thesession materials together. Because modular delivery mightmean that some children do not receive every module, manyof the modules were instead given internal acronyms. Forexample, in the integral version, one of the cognitive therapymodules contained three techniques denoted by the letters“H-I-N” (“Help from a Friend,” “Identifying the Silver Lin-ing,” “No Replaying Bad Thoughts”), which fit together intothe work “THINK” when performed in sequence with othersessions. Consistent with the principle of information hiding,this acronym was changed to “F-U-N” (“Friends Who CanHelp,” “Understanding the Silver Lining,” and “No ReplayingBad Thoughts”), so that it would stand alone if that modulewere delivered in isolation or in a different sequence thanspecified in the integral PASCET manual.
Similar issues arose with themes and characters. For exam-p xietym lard ntoa usw thep efula trat-e nualw rac-t h thet ptedw fer-e tiger.E roto-c e toi
2
fort thera on ofc t con-v apyo tivet s&
CHm ulesd
le, the cat character and theme from the integral ananual (Kendall et al., 1990) created challenges for moduesign that prohibited building this type of explicit theme inxiety portion of the MATCH manual. Although an obvioay to handle this issue is to remove these devices fromrotocols altogether, that solution may forfeit some usnd engaging properties of the therapy. An alternative sgy, and one deployed in the design of the MATCH maas to encourage the selection of an individualized cha
er, whose consistent appearance and involvement witherapy material is managed by the therapists and promithin various modules. Thus, one child might have a prence for a dolphin, another for a dog, and another for aach would select or articulate a character early in the pol, and the therapist would use this character and themllustrate various aspects of the protocol.
.2. Example: coordination modules
A second aspect of the protocol that was designedhe MATCH manual was the procedure for deciding whend when to implement its content modules. The selectiontent modules was guided by algorithms that represenentional clinical applications of cognitive behavior therr behavioral parent training (akin to formalizing prescrip
reatment decisions; e.g.,Beutler & Harwood, 2000; PersonTompkins, 1997).The coordination modules designed for the MAT
anual have important parallels with the content modescribed above. Just as content modules may use atemplate
trial design. However, such challenges force designers incase to consider the full range of possibilities of what migcome before or after any module, and to specify expliciwhat information should transfer across modules (e.g., homwork content, skill sets to be repeated indefinitely, etc.).
Sometimes repeated review of skill sets requires senhancement upon each performance. For example, inMATCH module for “building talents and skills,” youth areasked to pick a skill and work on it as part of a long-terplan. This requires weekly action on the part of the child, athus weekly review of the plan by the therapist. Describithis cumulative process in the homework review sectionsall other modules would be prohibitive, especially given thmany modules besides the “building talents and skills” moule involve cumulative homework. The solution was thusmake each homework review in other modules refer backall previous homework and to consider whether cumulatassignments are active.
Another example of avoiding traditionally cumulativdesign in favor of modular format involved in vivo exposurRather than specify multiple sessions that involve exposof increasing intensity or complexity, all in vivo exposure ithe MATCH protocol is delivered through a single moduthat was to be repeated until an intensity or complexity cririon was met. This module was therefore written to accoufor the fact that it might be the child’s first exposure exercisthe last, or somewhere in between. The “notes for the thapist” at the end of the exposure module were consequedesigned to provide guidelines for the different circumstancunder which exposure would occur and for the selectionincreasingly challenging stimuli as part of the exercises.
148 B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156
to representpractice elements, coordination modules mayuse aflowchartto represent analgorithm. Thus, the processof coordination module design was similar to that of con-tent module design: in the same manner that we coded forpractice elements and then designed therapy content whichwas outlined in standardized templates, so too we identi-fied traditional therapy algorithms, and then outlined these instandardized flowcharts. For example, the MATCH coordina-tion module for depression contains a flowchart that selectsbehavioral modules first (e.g., “problem solving,” “activityscheduling”) and then progresses to cognitive modules (e.g.,“cognitive restructuring”).
The final result was that the MATCH manual containedone high-level coordination module, which linked to threeproblem-specific coordination modules (i.e., one each foranxiety, depression, and conduct problems). The therapisttherefore uses the initial high-level coordination module (seeFig. 2) to make an initial decision about the primary prob-lem of the youth. If the problem is anxiety, the therapistis instructed to reference the anxiety coordination module,which contains a flowchart outlining the algorithm for theanxiety protocol. If the problem is depression, the therapistis guided to the depression coordination module, which con-tains a flowchart for the depression algorithm, and so forth.
Similar to how we defined narrative headings for thetemplates in the content module templates (e.g., “home-w ofs les.T ep-r ond( isionm ngles( nt”)r tionsf Pri-m adet up-t n.F atet dule.
ules. In a classic article that helped launched the structuredprogramming movement in computer science,Bohm andJacopini (1966)showed that the three procedural constructsof sequencing (do A, then B, then C), alternation (either doA or do B), and iteration (repeat A until condition is satis-fied) are sufficient to represent almost all arrangements ofactivities. Accordingly, these constructs may serve as thecore elements in understanding coordination modules. Forexample, the first two modules inFig. 3 illustrate the pro-cedural construct of sequencing with “parent monitoring”starting upon completion of “self monitoring.” The “able toproceed” diamond is a yes or no decision representing analternation procedure. Specifically, the therapist either pro-ceeds to the primary therapeutic sequence or implementsprocedures for handling therapeutic interference. Finally, the“module complete” diamond represents an example of aniteration procedure, wherein the practice element continuesuntil a condition (i.e., “module gains complete” or “unableto proceed”) is met.
This illustration is far from an exhaustive display of allpossible combinations and arrangements of psychotherapycontent; indeed, it illustrates only a fraction of even thoseprocedural constructs found within the MATCH manual. Thepoint is that the algorithms for coordinating psychotherapycontent can be outlined according to a set of definable rules,just as is true of the narrative representation of practice ele-m
ork review,” “role-play”), we also outlined the legendymbols for flowcharts found in the coordination moduhe rectangles with rounded corners (“Begin,” “End”) resent entry and exit points into the algorithm. The diam“Already in treatment?”) represents that a yes or no decust be made regarding the question within. The recta
“Conduct Initial Assessment,” “Conduct Brief Assessmeepresent content modules that outline specific instrucor performing therapeutic procedures. The triangle (“ary problem area”) indicates that a decision must be m
o select one of multiple options and the circles (“Disrive Behavior,” “Anxiety”) represent criteria for the decisioinally, the chevrons (e.g., “Depression Flowchart”) indic
hat the therapist should refer to another coordination moCoordination module meta-data was also outlined,
ncluded a flowchart legend, a list of referenced moduletatement of conditions of use and expected functioningodularity property 2, above), information transfer parters (e.g., whether specific assessment information su
ype of anxiety disorder should be indexed in a subsequeferenced coordination module; see modularity prop, above), application boundaries for the algorithm (ge range, cultural limitations), ownership/authorship,rences, and annotations.
In terms of their algorithms, the problem-specific coination modules each specify the set of practice elemr content modules, along with their rationally determirrangement.Fig. 3shows the final algorithm for the MATCepression coordination module, and offers an illustraf the different possible arrangements of content m
ents in content modules.These algorithms provide an excellent illustration of
nformation-hiding quality of modular design. If a problepecific coordination module (e.g.,Fig. 3) was found to bawed (e.g., implementing relaxation was never associth clinical improvement beyond that produced by oontent modules for depression), a portion of that probpecific coordination module could be reordered or deompletely without affecting the other coordination mode.g., anxiety flowchart) or the other content modules (psychoeducation for child”). In that sense, coordinaodules follow the same rules regarding standardized
ace (modularity property 3, above) and information-hidmodularity property 4, above) as do the content modul
. Evaluation of modular design: efficiency,ffectiveness, and efficacy
Having reviewed a concrete illustration of a modular photherapy design, we will now evaluate the qualities of mlar design principles relative to integral treatment deshe relative strengths and weaknesses of integral vodular designs have already been discussed in some
n various business and engineering contexts (e.g.,BaldwinClark, 1997; Garud & Kumaraswamy, 1995; Mikkola &assmann, 2003; Parnas, 1972; Ulrich & Ellison, 1999), and
o provide background these are briefly reviewed here fiSome of the identified strengths of integral designs in
exts other than psychotherapy are as follows: (1) much o
B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156 149
Fig. 2. Example of a high-level coordination module.
design work need not be allocated to developing templatesor standardized guidelines for how modules should interactwith each other, thus saving time and materials; (2) inte-gral products can be streamlined, because there is no needto ensure that they can be disassembled into standardizedparts (e.g., compare a common mail truck with a tractor-trailer); (3) integral designs can also provide for superioraccess to information, because any time a feature is needed, itis designed in directly, as opposed to having that single featureindexed or accessed from multiple places (e.g., a how-to bookthat repeatedly puts both English and metric amounts in theinstructions, versus one that references a conversion table);(4) integral designs emphasize craftsmanship and form overfunction, given fewer constraints than would be required bymodularity (imagine a traditional versus a modular couch);(5) they promote systemic innovation rather than incremen-tal improvements by encouraging redesign of entire productsrather than innovation of parts (compare single-lens cameraswith those having interchangeable parts); (6) by that samelogic, integral designs can protect innovations from imitation,and create high barriers for competitors who wish to enterthe market (e.g., consider the two major traditions of desktopcomputer design); and (7) they similarly can increase marketvolume by making it difficult for consumers to select some
parts but not others (e.g., consider the “extra value meal”versus orderinga la carte).
Among the many strengths of modular design are short-ened development time and ease of incremental improve-ments, through the availability of standardized components.For example, a new product can be designed that uses manyreadily available modules in conjunction with one or twonew modules (e.g., much modern software, updated throughpatches rather than installation of a revised version). Simi-larly, modular design promotes increased product flexibilityand variety, by allowing for the rearrangement or combina-tion of the existing module set (e.g., a rolodex versus a boundaddress book). Modular design facilitates rapid comprehen-sibility because components can be studied one module at atime (e.g., a box of recipe cards versus a narrative cookbook).Modular design incorporates efficiency through the reuse ofmodules, and reduced inventory and logistic expenses, inthat many products can be built from a smaller number ofparts (e.g., consider the greater variety of storage and place-ment options with modular shelves versus a single all-in-oneshelf unit). Finally, modularity can promote improved sys-tem reliability due to higher production volume supportinggreater experience with components (e.g., compare assemblyline production with custom, start-to-finish designs). Many
150 B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156
Fig. 3. Example of a problem-specific coordination module for depression.
of these strengths of modular designs that have been noted inindustrial and business contexts may also be realized in thecontext of psychotherapy development.
To elaborate these benefits in a framework familiar to psy-chologists, we apply the notions of efficacy, effectiveness,and efficiency to the psychotherapy design context. As astarting point for this discussion, we offer the following rudi-ments.Efficientdesigns should be relatively parsimonious(include only the material that is necessary), comprehen-sive (apply to a broad class of psychotherapy problems),reusable (incorporate some interchangeable parts), fault tol-erant (tolerate interruptions and unanticipated events withoutcompromising progress), and cost-effective (i.e., benefits out-weighing costs).Effectiveinterventions are believed suitablefor real-world problems because of their demonstrations ofhigh feasibility and/or generalizability (Task Force on Psy-chological Intervention Guidelines, 1995), and the same istrue ofeffectivedesigns. Specifically, they should be scalable(integrate a framework for adaptation and generalization todifferent circumstances), transportable (readily implementedin new settings), and satisfying (elicit therapist or clientsatisfaction). Finally, much in the way that efficacious inter-
ventions should produce reliable, internally valid evidence oftheir ability to yield positive outcomes, efficacious designsshould similarly produce working solutions to targetedproblems.
3.1. Efficiency of modularity
3.1.1. ParsimonyA pilot demonstration byChorpita et al. (2005)illustrated
the potential efficiency of using modules to summarize treat-ment protocols. Specifically, a set of 49 protocols for child-hood disorders classified as evidence-based using structuredcriteria (APA Task Force on Dissemination of PsychologicalProcedures;Chorpita et al., 2002) could be reduced to a set ofapproximately 20 practice elements. Further, within particu-lar problem areas, multiple protocols could be reduced to aneven smaller set of common elements (e.g., coding across 25anxiety protocols showed a common set of approximately sixpractice elements). Were such practice elements representedin modular form, one could approximate the implementationof these 25 anxiety protocols through just a few differentcoordination modules.
B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156 151
Although modularity might show considerable potentialfor data reduction in this context, it carries the potential forgreater complexity when viewed in an individual protocol-to-protocol design comparison. A “typical” manualized treat-ment reads like a linear recipe for how to provide a therapyprogram. Although a modular protocol could replicate a lin-ear manual, our vision of a “typical” modular protocol doesnot. A modularized protocol reads like a tool kit of practiceoptions that are organized by a collection of flowcharts guid-ing the user through their selection. When the other potentialbenefits of modularity are ignored, the linear, integral recipecould seem more user-friendly to treatment professionalsThe efficiency of modular design becomes most pronouncedwhen multiple protocols are represented in a single treat-ment system, allowing for redundancies in content to beaddressed (e.g., both a depression protocol and anxiety pro-tocol could share a relaxation module—thus, the clinicianneed not learn a different relaxation strategy for each differ-ent protocol). In that sense, modularity is better viewed as aparsimonious approach to a treatmentsystemrather than asa parsimonious approach to any one treatment program orprotocol.
3.1.2. ComprehensivenessModular design yields the potential to address diverse vari-
ations in treatment targets. Using a counting rule for orderedc nt isi tionsd il-l le( dis-t 00[ ; i.e.,1 to-c ss,s ting1 cticee s oft xam-p er ina ores ers,t
crip-t ever,i adyp s nott entsr ord ch al ign,w velo B).B stingp thee hing
and design, and potentially yield advances in the compre-hensiveness of interventions more generally.
3.1.3. ReusabilityModular design can preserve particular elements of psy-
chotherapy techniques found to be highly useful, while revis-ing or designing new protocols. Some techniques are handledin this manner under current design strategies. For example,a variety of relaxation techniques (e.g.,Deffenbacher, Lynch,Oetting, & Kemper, 1996; Laxer & Walker, 1970) emanatefrom Jacobsen’s (1938)description of the technique. In amodular context, such a relaxation module could be insertedseamlessly into any protocol seeking to incorporate relax-ation, without requiring developers of each new protocol toconstruct a new description of relaxation (or even to rewritean old one). Over time, such development might allow for anaccumulation of particularly effective modules, whose com-binations and arrangements with newer techniques could becontinually tested and refined.
This reusability aspect also allows for the efficient incor-poration of therapy innovations. The psychotherapeuticknowledge base is rapidly expanding, and ongoing revisionto treatment protocols should be expected. However, giventhat treatment developments are likely to be incremental,modular treatment design provides a structure for integratingnew developments. Based on the specific modular frameworkw elop-m on ofn ion-m newc enti-fi o bed e cana odu-l tingt
3ne of
t -m ani ngedo igneda iza-t tiald nc-t rapyp okend om-b setst Thish ingc xist-i , andc ed, ori
ombinations (assuming that order and not simply contemportant) shows that the number of 10 session intervenrawn from a library of only 15 modules is nearly 11 b
ion (i.e., 15!/(15!− 10!)). If certain orders are not possibe.g., due to simultaneity constraints), the lower limit ofinct modular contents for an intervention is still over 30assuming a counting rule for non-ordered combinations5!/10!(15!− 10!)]. Of course, a large portion of these prools would likely be of little incremental value. Nevertheleuch design potential would be able to replicate all exis0-element integral protocols based on the same 15 pralements, and would likely yield some new combination
hat could address variations of treatment targets. For ele, the exchange of one new practice element for anothprotocol for eating disorders might yield a variation m
uitable for a certain subset of clients with eating disordhus increasing the versatility of existing protocols.
The potential challenges associated with such presive matching have already been discussed above. Howt should be noted that all current interventions are alrerescriptive at some level of abstraction, so prescription i
he issue at hand. For example, most manualized treatmequire matching of the intervention to a target problemiagnostic area (i.e., use manual X for disorder Y). Su
evel of matching can be maintained with a modular deshile allowing for a new level of prescription at the lef techniques (i.e., use module A for problem featurey making these prescriptive assumptions behind exirotocols more explicit, modularity can efficiently allowvaluation of more assumptions about treatment matc
e presented here, one can easily see how the devent of new techniques could be integrated by generatiew modular content. Similarly, improvements in decisaking knowledge could be integrated by generation of
oordination modules. When these innovations are ided, a whole new treatment system would not need teveloped and implemented. To the degree the literaturdvance psychotherapeutic activities and inferences, m
ar protocols can provide a means of rapidly incorporahese incremental innovations.
.1.4. Fault toleranceFrom the perspective of design and development, o
he major benefits of modularity is thepreservation of interediate states of development. This means that when
nstruction set such as a psychotherapy manual is char adapted, it need not be completely dissolved and desnew. In his articulation of a theory of design organ
ion,Simon (1996)emphasized the importance of this parecomposition of a full design into semi-independent fu
ional parts. Were one to apply this notion to psychotherocedures, it would mean that interventions could be brown into individual units or techniques that could be cined in different ways. Those combinations might create
hat could then be combined with other sets, and so forth.ierarchical “tree structure” of design is the basis for avoidomplete decomposition in efforts to adapt or innovate eng instruction sets. Rather, portions could be preservedomponents or sub-components could be removed, addndependently modified.
152 B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156
Simon (1996)illustrated this notion with the following“Watchmaker” parable:
There once were two watchmakers, named Hora and Tem-pus, who manufactured very fine watches. . . The watchesthe men made consisted of about 1,000 parts each. Tempushad so constructed his that if he had one partly assembled andhad to put it down. . . it immediately fell to pieces and hadto be reassembled from the elements. [Hora] had designedthem [his watches] so that he could put together subassem-blies of about 10 elements each. Ten of these subassemblies,again, could be put together into a larger subassembly; anda system of ten of the latter subassemblies constituted thewhole watch. It is rather easy to make a quantitative anal-ysis of the relative difficulty of the tasks of Tempus andHora: suppose the probability of an interruption will occur,while a part is being added to an incomplete assembly isp. Now if p is about 0.01. . . We arrive at the estimate asfollows:
1. Hora must make 111 times as many complete assembliesper watch.
2. Tempus will lose on the average 20 times as much workas Hora.
3. Tempus will complete an assembly only 44 times permillion attempts. . . Hora will complete nine out of ten.Tempus will have to make 20,000 as many attempts per
as as ier-a ablei plexs prin-c esigna
havel y timea anuan se oft ce.R ffectm atch,i e. Inm e out-s notr s. Inf tionm fromt at isn
3iated
w sts.M ts aree ears
additional costs that are not sufficiently offset by the otherefficiencies associated with modularity.
3.2. Effectiveness of modularity
3.2.1. ScalabilityScalability refers to the ability to increase or decrease a
particular property of the protocol as part of the design pro-cess. For example, treatment length could be scaled, suchthat the protocol dictates a course of treatment that is long orshort. Content can be scaled, such that the protocol can dic-tate which procedures (modules) are allowed and which arenot. Logic can be scaled, such that the protocol can allow formany conditional decisions (e.g., “if no progress is observed,skip to module B”) or few conditional decisions.
Scalability is not to be confused with flexibility. Scalabilityrefers to adaptability of a protocol during its design; whereasflexibility refers to adaptability of the protocol during itsdelivery. A modular treatment – which is implicitly scaleable– can be inflexible, so long as its coordination modules dic-tate a linear relation of content modules with a minimum ofchoice points. And similarly, flexible treatments need not bemodular and hence scalable. For example,Jacobson et al.’s(1989) protocol for flexible marital therapy does not pos-sess discrete units meeting the four principles of modularityoutlined above, nor does Multisystemic Therapy (Henggeler,Sb oni oft gnedt ro-t r orl
si-b gralPd est nt tot ge theo lgo-r em con-t asts itor-i tivitys ouldo beena rmi-n hus,F scalet tmentl
en-s ationmi hm
completed assembly as Hora (Simon, 1996; pp.188, 189).
Although this numerical example was not intendederious quantitative estimate of the relative efficiency of hrchical designs, it calls to light the powerful effect that st
ntermediate states can have on the evolution of comystems. Again, as will be argued below, such designiples have significant advantages for psychotherapy dnd innovation.
In the psychotherapy design context, integral designsess stable intermediate design states. For example, ann integral manual needs to be changed, the entire meeds to be examined from beginning to end, becau
he lack of information hiding and standardized interfaemoving one session from an integral protocol might aany later exercises in other sessions—so, like the w
t must be recreated from a much earlier design statodular design, these contingencies are handled at th
et of design, so that deleting a single module shouldequire re-examination of any of the rest of the moduleact, the deletion needs only to be noted in the coordina
odule—the content module need not even be omittedhe protocol, but can remain as an “orphan module” thever accessed.
.1.5. Cost-effectivenessAt present it is unclear whether the benefits assoc
ith modular treatment development will outweigh the coore research is needed on whether modular treatmenfficacious, on whether the complexity of the designs b
l
choenwald, Borduin, Rowland, & Cunningham, 1998), yetoth involve a great deal of flexibility and individualizati
n their delivery. Flexibility is best thought of as a propertyherapy that itself is scalable (i.e., protocols can be desio be highly flexible or highly inflexible), and modular pocols possess the ability to scale flexibility to be higheower during protocol design.
To further illustrate the notion of scalability, it is posle to represent the implicit algorithm underlying the inteASCET manual for depression (Weisz et al., 1999) in a coor-ination module (seeFig. 4). The PASCET manual specifi
hat the therapist proceed from the first practice elemehe last one, in straight sequence. No decisions to chanrder or content of treatment are part of the implicit aithm of the PASCET manual. InFig. 5, the algorithm of thanual can be rearranged to allow the protocol to omit
ent selectively following successful completion of at leix practice elements (i.e., self-monitoring, parent monng, psychoeducation, and parent psychoeducation, accheduling, and maintenance). Successful completion wccur upon determination that the treatment goal hasttained. This protocol also allows for unsuccessful teation after administration of at least four modules. Tig. 5demonstrates the use of a coordination module to
he same set of therapeutic content with respect to treaength.
Of course, one can go further to scale on the dimion of therapeutic content. The final depression coordinodule chosen for the MATCH manual (seeFig. 3) spec-
fied not only changes to the original PASCET algorit
B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156 153
Fig. 4. A problem-specific coordination module corresponding to the stan-dard PASCET algorithm.
in terms of treatment length, but also added a set of proce-dures for coping with therapeutic interference (e.g., a child’sbehavior becomes too disruptive at home to continue thecognitive behavioral regimen). The specific interference pro-cedures were selected from the set of modules developedfor problems in other evidence-based protocols (e.g., for dis-ruptive behavior or anxiety). This allowed for a conditional“mixing-and-matching” of supplemental modules necessaryto promote completion of the depression protocol (e.g., intro-ducing a token economy or a time out procedure).
Given this kind of scalability, modularity can be seen notso much as an adaptation of the integral manual itself, but
rather as a framework allowing adaptation across a variety ofdimensions (e.g., content, length, flexibility). Thus, an advan-tage of modularity in general is that it provides anexplicitframework for adaptation, without automatically dictatingthe adaptation of an existing protocol. Research strategiescould easily be deployed to test whether gross variations inorder, content, or logic of the intervention lead to differentialefficacy, and empirically informed revisions to the protocolcould be quickly deployed.
When moving beyond laboratory-based interventiondesign, this notion of adaptability becomes increasinglyimportant. In general, the literature supports the idea that coretechnologies in many fields are adapted when put into use inthe field (e.g.,Rogers, 1995). Weisz’s (2004)DeploymentFocused Model argues that optimum intervention designrequires evolution of a protocol based on interplay betweenthe intervention program and the context in which it is to bedeployed; this model makes a case for the idea that adaptationof psychological interventions is critical to their uptake andeffectiveness in clinical contexts. That said, minimal researchexists to date to suggest what types of adaptations are needed,and which adaptations might threaten the efficacy of a pro-tocol. Nevertheless, whether or not we agree that adaptationmay be of value, modularity provides a suitable context foradaptation if needed, and imposes few costs if adaptationsare not needed.
od-u olst e toe ove,t itht er,a com-m thep t thel iag-nF gicalP tiali thep thep ingfl pir-i blep anyp ever,i at ag treatd ildrenw usk nts– ucep ing.A it isl baseo oped.
In light of these properties, one challenge posed by mlarity is that of complexity. In particular, modular protoc
hat use a flexible algorithm might introduce errors duxcessive reliance on clinical judgment. As alluded to abhis is not a problem with modularity itself, but rather whe notion of prescription or flexibility in treatment. Furths mentioned above prescription is already a ratheron feature of existing psychotherapy protocols, withrincipal difference being that the prescription occurs a
evel of the main problem or disorder (i.e., matching dosis or primary problem area to treatment protocols;Taskorce on the Promotion and Dissemination of Psycholorocedures, 1995), and does not take into account poten
ndividual differences that might warrant modification ofrotocol. Thus, it is not even prescription per se that isroblem, but more likely the fact that decisions regardexible versions of modular treatments would require emcally informed and reliable decision rules, just as reliasychiatric diagnosis is currently required to prescribe msychotherapy manuals. The complexity remains, how
n that the evidence base required to inform decisionsreater level of specificity is largely absent (e.g., how toepressed men versus depressed women, or anxious chithout comorbidity versus anxious children with varioinds of comorbidity). Without such data, flexible treatmewhether modular or not – have the potential to introd
roblems related to the requisite clinical decision-maklthough such protocols could possibly still be effective,
ikely that they can be further improved as the evidencen client by treatment interactions becomes better devel
154 B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156
Fig. 5. A problem-specific coordination module corresponding to the PASCET algorithm, adapted for variable protocol length.
3.2.2. TransportabilityModularity can allow for rapid adaptation of a protocol
for new contexts, thus increasing transportability. For exam-ple, a change to the protocol to match the intervention to anew target population (e.g., children versus adults; males ver-sus females) could presumably tailor embedded behavioralrehearsal exercises or role-plays to the client by selectingfrom a library of possibilities, while preserving the other ele-ments of the modules or session structure. The potential fortransfer of protocols from one context to another is thereforegreatly enhanced.
In a similar manner, this transportability has importantimplications for research. As noted byKazdin, Bass, Ayers,& Rodgers, (1990), the notion of discovering which inter-ventions work for whom and under what conditions requiresa greater understanding ofwhy interventions work (Kiesler,1966). For example, does Panic Control Treatment (Craske& Barlow, 1993) for adults with panic disorder work becauseof its inclusion of breathing retraining, interoceptive expo-sure, or cognitive restructuring, or perhaps some combination(Barlow, Allen, & Choate, 2004)? Such research questionsare more easily addressed in the context of modularity, whichprovides adequate controls for content and ensures that min-imal problems will arise as a consequence of omission orreordering of particular pieces of the program.
One of the greatest potential benefits related to transporta-b t ofc thep mentp on ofp ol-l tocolo uldf thert ifica-t tracta ring,
and supervision. As noted above, the modular structure weillustrate here provides a framework to help users incorporatenew practice modules and new decision rules without need-ing to learn a whole new treatment system. The existence ofa schema to which incremental knowledge is added shouldfacilitate the transfer of practice knowledge and innovation(e.g.,Owens, Bower, & Black, 1979).
3.2.3. SatisfactionSome preliminary data from a study in progress (e.g.,
Francis & Chorpita, 2003) showed promising findings regard-ing therapist satisfaction with modular procedures. Thesample included 16 community therapists and 21 gradu-ate trainees participating in a study on modular applicationof therapy procedures. Specifically, 83.3% of the combinedsample agreed or strongly agreed that a modular approachto therapy was “very applicable to cases in clinical prac-tice,” whereas among that same group only 50% agreed orstrongly agreed that a traditional (i.e., integral) manualizedapproach to therapy was “very applicable to cases in clini-cal practice.” Similarly, 58.5% of the sample rated a modularapproach as “more applicable to clinical cases” than tradi-tional approaches, whereas 7.3% rated traditional manualsas more applicable to clinical cases than a modular approach.Obviously, more formal data regarding therapist satisfactionneed to be collected, particularly data that are based not justo withm
3
treat-m ioust thatm indi-v itudeo acy
ility of modular protocols can be seen in the contexlinical training. As we noted above when discussingarsimony of modular designs, because different treatrotocols share common practice elements, the collectiractice modules is likely to grow more slowly than the c
ection of protocols. In the extreme case where a new pronly involves a reordering of practice modules, training co
ocus on teaching a single new decision algorithm rahan say 16 new therapy sessions. The explicit idention of decision algorithms may make these rather absnd covert processes more amenable to training, monito
n first impression but rather on continued experienceodular design.
.3. Efficacy of modularity
Because modular designs can recreate manualizedent protocols, in principle, they should be no less efficac
han traditional integral designs. There is the possibilityodular design could enhance efficacy, assuming that
idualization in some contexts could increase the magnr the speed of the effects of a protocol. However, effic
B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156 155
likely results from a combination of the design structure (e.g.,modular versus integral) and the design content. The structurein and of itself will not solve any problem without efficaciouscontent.
That said, there are emerging data that modular designscan meet sufficient efficacy criteria to warrant further researchand development (e.g.,Chorpita, Taylor, Francis, Moffitt, &Austin, 2004). In this recent study,Chorpita et al. (2004)demonstrated positive outcomes for seven youth with anxi-ety disorders in an experimental, multiple-baseline design.Diagnoses represented in the sample included separationanxiety disorder, generalized anxiety disorder, panic disor-der with agoraphobia, specific phobia, anxiety disorder nototherwise specified, major depression and trichotillomania.For all seven youth, primary diagnoses as measured by asemi-structured interview were absent at post treatment and6-month follow-up assessments, and measures of anxietysymptoms and life functioning almost uniformly evidencedclinically significant improvements.
4. Conclusions
A final point worth noting is the argument concerningwhether module content is all that is important. For exam-ple, the modular approach to design – as described so far –e redi-e encee ra-tt ntenta lgo-r e oft therr erente pingm shipsr dingw mosti onalt eci-s thefi
pro-v fort rip-t longt , ande nt ofe uiva-l sion,i , if am nlye firstt s andp tentia
benefits in future efforts involving psychotherapy develop-ment and testing.
Acknowledgements
A project supported through a grant from the HawaiiDepartment of Health and carried out in conjunction withthe Research Network on Youth Mental Health, sponsoredby the John D. and Catherine T. MacArthur Foundation.
References
Addis, M. E., & Krasnow, A. D. (2000). A national survey of practic-ing psychologists’ attitudes toward psychotherapy treatment manuals.Journal of Consulting and Clinical Psychology, 68, 331–339.
Baldwin, C. Y., & Clark, K. B. (1997, September/October). Managing inan age of modularity.Harvard Business Review, 84–93.
Barkley, R. A. (1997).Defiant children: A clinician’s manual for assess-ment and parent training(2nd ed.). New York, NY: Guilford.
Barlow, D. H., Allen, L. B., & Choate, M. L. (2004). Toward a unifiedtreatment for emotional disorders.Behavior Therapy, 35, 205–230.
Beck, A. T., Rush, A. J., Shaw, B. F., & Emery, G. (1979).Cognitivetherapy of depression. New York: Guilford.
Beutler, L. E., & Harwood, T. M. (2000).Prescriptive psychotherapy: A
B ande
B t of
C ainem-an
C andaii
/
C andns: A
C in,for
C en,on ofrva-
C rand
C -,
C obia.rs
mphasizes specific clinical strategies as the “main ingnts” of psychotherapy, so to speak. Reasonable evidxists to challenge this notion, for example in the liteure on therapeutic relationship (Norcross, 2002). It shouldherefore be emphasized that representing therapy cos a division into practice elements and coordinating aithms is not the only way to conceptualize the enterprisherapeutic intervention and package it into modules. Oepresentations are possible that could yield a rather diffxample, yet could still be modular in nature (e.g., develoodules to represent therapeutic processes or relation
ather than specific clinical strategies). Decisions regarhich aspects of therapy (e.g., process, strategy) are
mportant for achieving its goals are somewhat orthogo the potential benefits of modularity in design. Such dions will likely remain the subject of great debate ineld.
That said, modularity as a design principle appears toide a promising approach for innovations in therapy andhe discovery of the optimal balance of flexibility, prescion, and structure. The potential benefits are organized ahe three broad dimensions of efficiency, effectivenessfficacy. Importantly, these dimensions are independeach other, such that if modular design does proves eq
ent but not superior to integral design along one diment can still offer advantages in the other two. For example
odular design is efficient and fits well into practice, but oquivalent in efficacy to traditional integral designs, the
wo advantages would still support its use. Researcherrotocol designers are encouraged to consider these po
l
practical guide to systematic treatment selection.New York: OxfordUniversity Press.
ohm, C., & Jacopini, G. (1966). Flow diagrams, Turing machines,languages with only two formation rules.Communications of thACM, 9, 366–371.
urke, A. E., & Silverman, W. K. (1987). The prescriptive treatmenschool refusal.Clinical Psychology Review, 7, 353–362.
arroll, K. M. (1998). A cognitive behavioral approach: Treating cocaddiction (Therapy Manual for Drug Addiction NIH Publication Nuber 98-4308). Rockville, MD: US Department of Health and HumServices, National Institutes of Health.
hild and Adolescent Mental Health Division. (2003). Instructionscodebook for provider monthly summaries. Honolulu, HI: HawDepartment of Health. Available via Internet athttp://www.hawaii.govdoh/camhd/reports/monthlysummaryinstructions.pdf.
horpita, B. C., Daleiden, E., & Weisz, J. R. (2005). Identifyingselecting the common elements of evidence based interventiodistillation and matching model.Mental Health Services Research, 7,5–20.
horpita, B. F., Taylor, A. A., Francis, S. E., Moffitt, C. E., & AustA. A. (2004). Efficacy of Modular Cognitive Behavior Therapychildhood anxiety disorders.Behavior Therapy, 35, 263–287.
horpita, B. F., Yim, L. M., Donkervoet, J. C., Arensdorf, A., AmundsM. J., McGee, C., et al. (2002). Toward large-scale implementatiempirically supported treatments for children: A review and obsetions by the Hawaii Empirical Basis to Services Task Force.ClinicalPsychology: Science and Practice, 9, 165–190.
horpita, B. F., & Weisz, J. R. (2003).MATCH-ADC: A ModulaApproach to Treatment for Children with Anxiety, Depression,Conduct Problems. Unpublished treatment manual.
larke, G., Lewinsohn, P., & Hops, H. (1990).Leader’s manual for adolescent groups: Adolescent coping with depression course. PortlandOR: Kaiser Permanente Center for Health Research.
raske, M. G., & Barlow, D. H. (1993). Panic disorder and agoraphIn D. H. Barlow (Ed.),Clinical handbook of psychological disorde(2nd ed.). New York: Guilford Press.
156 B.F. Chorpita et al. / Applied and Preventive Psychology 11 (2005) 141–156
Curry, J. F., Wells, K. C., Brent, D. A., Clarke, G. N., Rohde, P., Albano,A. M., et al. (2000).Treatment for Adolescents with Depression StudyCognitive Behavior Therapy Manual: Introduction, Rationale, andAdolescent Sessions. Unpublished manuscript, Duke University Med-ical Center.
Deffenbacher, J. L., Lynch, R. S., Oetting, E. R., & Kemper, C. C. (1996).Anger reduction in early adolescents.Journal of Counseling Psychol-ogy, 43, 149–157.
Durand, V. M. (1990).Severe behavior problems: A functional approachto communication training. New York: Guilford.
Durand, V. M., & Crimmins, D. B. (1988). Identifying the variables main-taining self-injurious behavior.Journal of Autism and DevelopmentalDisorders, 18, 99–117.
Eisen, A. R., & Silverman, W. K. (1998). Prescriptive treatment for gen-eralized anxiety disorder in children.Behavior Therapy, 29, 105–121.
Francis, S. E., & Chorpita, B. F. (2003, November).An examinationof clinical decision making strategies: Preliminary findings. Paperpresented at the annual meeting of the Child and Adolescent Anxi-ety Special Interest Group, Association for Advancement of BehaviorTherapy, Boston, MA.
Freedheim, D. K. (Ed.). (1992).History of psychotherapy: A century ofchange. Washington, DC: American Psychological Association.
Garud, R., & Kumaraswamy, A. (1995). Technological and organizationaldesigns for realizing economies of substitution.Strategic ManagementJournal, 16, 93–109.
Henggeler, S. W., Schoenwald, S. K., Borduin, M. C., Rowland, M. D.,& Cunningham, P. B. (1998).Multisystemic treatment of antisocialbehavior in children and adolescents. New York: Guilford.
Jacobson, E. (1938).Progressive relaxation. Chicago: University ofChicago Press.
Jacobson, N. S., Schmaling, K. B., Holtzworth-Munroe, A., Katt, J.vs.
apy.
K icalarch.
K f a
K
K search
K . S.
L tionl-
Malone, T. W., & Crowston, K. (1994). The interdisciplinary study ofcoordination.ACM Computing Surveys, 26, 87–119.
Mikkola, J. H., & Gassmann, O. (2003). Managing modularity of prod-uct architectures: Toward an integrated theory.IEEE Transactions onEngineering Management, 50, 204–218.
Norcross, J. C., & Beutler, L. E. (2000). Prescriptive eclectic approachto psychotherapy training.Journal of Psychotherapy Integration, 10,247–261.
Norcross, J. C. (Ed.). (2002).Psychotherapy relationships that work:Therapist contributions and responsiveness to patients. Oxford, UK:Oxford University Press.
Owens, J., Bower, G. H., & Black, J. B. (1979). The “soap-opera” effectin story recall.Memory and Cognition, 7, 185–191.
Parnas, D. L. (1972). On the criteria to be used in decomposing systemsinto modules.Communications of the ACM, 15, 1053–1058.
Persons, J. B. (1991). Psychotherapy outcome studies do not accuratelyrepresent current models of psychotherapy: A proposed remedy.Amer-ican Psychologist, 46, 99–106.
Persons, J. B., & Tompkins, M. A. (1997). Cognitive-behavioral caseformulation. In T. D. Eells (Ed.),Handbook of psychotherapy caseformulation (pp. 314–339). New York, NY, USA: Guilford Press.
Rogers, E. M. (1995).Diffusion of innovations(4th ed.). New York: TheFree Press.
Schulte, D., Kunzel, R., Pepping, G., & Schute-Bahrenberg, T. (1992).Tailor-made versus standardized therapy of phobic patients.Advancesin Behaviour Research and Therapy, 14, 67–92.
Simon, H. A. (1996).The sciences of the artificial(3rd ed.). Cambridge,MA: MIT Press.
Task Force on Promotion and Dissemination of Psychological Procedures,Division of Clinical Psychology, American Psychological Associa-tion. (1995). Training in and dissemination of empirically-validated
T Psy-ines:physi-tion.
U and
W ts:-
W A.trol.
W ithand
ke
W ation.
L., Wood, L. F., & Follette, V. M. (1989). Research-structuredclinically flexible versions of social learning-based marital therBehaviour Research and Therapy, 27, 173–180.
azdin, A. E., Bass, D., Ayers, W. A., & Rodgers, A. (1990). Empirand clinical focus of child and adolescent psychotherapy reseJournal of Consulting and Clinical Psychology, 58, 729–740.
earney, C. A., & Silverman, W. K. (1990). A preliminary analysis ofunctional model of assessment of school refusal behavior.BehaviorModification, 14, 340–366.
endall, P. C., Kane, M., Howard, B., & Siqueland, L. (1990).Cognitive-behavioral therapy for anxious children: Therapist manual. Ardmore,PA: Workbook Publishing.
iesler, D. (1966). Some myths of psychotherapy research and thefor a paradigm.Psychological Bulletin, 65, 110–136.
lerman, G. L., Weisman, M. M., Rounsaville, B. J., & Chevron, E(1984). Interpersonal psychotherapy of depression. New York: BasicBooks.
axer, R. M., & Walker, K. (1970). Counterconditioning versus relaxain the desensitization of test anxiety.Journal of Counseling Psychoogy, 17, 431–436.
psychological treatments: Report and recommendations.The ClinicalPsychologist, 48, 3–23.
ask Force on Psychological Intervention Guidelines, Americanchological Association. (1995). Template for developing guidelInterventions for mental disorders and psychosocial aspects ofcal disorders. Washington, DC: American Psychological Associa
lrich, K. T., & Ellison, D. (1999). Holistic customer requirementsthe design-select decision.Management Science, 45, 641–658.
eisz, J. R. (2004).Psychotherapy for children and adolescenEvidence-based treatments and case examples.Cambridge, UK: Cambridge University Press.
eisz, J. R., Weersing, V. R., Valeri, S. M., & McCarty, C.(1999).Therapist’s Manual PASCET: Primary and secondary conenhancement training program. Los Angeles: University of California
ells, K. C., & Curry, J. F. (2000).Treatment for Adolescents wDepression Study Cognitive Behavior Therapy Manual: ParentConjoint Parent–Adolescent Sessions. Unpublished manuscript, DuUniversity Medical Center.
ilson, G. T. (1996). Manual-based treatments: The clinical applicof research findings.Behaviour Research and Therapy, 34, 295–314
