Improving cancer pain management in communities: Main results from a randomized controlled trial

Improving cancer pain management in communities: Main results from a randomized controlled trial

Original Article Improving Cancer Pain Management in Communities: Main Results from a Randomized Controlled Trial Itim-m A large gap exist* brtwc~...

770KB Sizes 0 Downloads 52 Views

Original

Article

Improving Cancer Pain Management in Communities: Main Results from a Randomized Controlled Trial

Itim-m A large gap exist* brtwc~r~ pwbibll. ierld acn~al rrliri of cancrr paiti in clinici pwcti< ~1. Man! studies clwumr~~t th;tt lioY -!Nl!Y of civilccr patient\ wilt eaprt ictiw p.3in wmctimc~ during the cowx of their illncrs.1 Kecc~~t clinical rrpeticncr has dr~tir~tl\ti-atrrl that proper Irain aswwnent and twlati~rl~ *inipk ph.trmar~,ttleia()~ comhitwd with kiwi-drug lrcamrnt* c.rn prrnidc wti&c-t
~~duca,~~~r~al ac,ivi,icar ,ha, :,pproac bed imti\idwals and atTcc,c~I com,n,,ni,~ ,,orm~ as wll. III ,hir study. we evaluated the efle~,i\c,ws~ of a comn,uni,~basrd, mulrimodal educa,ionat program designed ,o improw CPY without many of ,he limitadons of prrviour studies. The aims of the study were ,o improw the knowledge, attitudlcx and clinicnl behaion of phyririanr and n,,rw% improw ,be knowlrdgr and attitudes aho,,, CPM of cancer patienrs and their lamil! mcmbcrs. dncl 10 reduce cancer-rrla,ed pain rxpericnccd in cancer parirntr. We hypo,hrGrd tha, rhe in,crwn,ion would yield impnncmr,,, i,, ,hew vxiahlm among thrse rargc, popul;,tions. A r.,ndomizrd communiw rrial wa conductrd i,,rr,l\ing six Yi,lnew,;, romm,,ni,ie\ wi,lt ,h,re awigned ,,a ron,rol and three assigned ,o expwimental condi,ions. L%cca,wc the intervention wa\ directed ,o whole cwnmunilics. i,,rluding all practicing physician\ and ,~urscs. and all cancrr patienb and theinfamilies. ,he co,,,muni,y was chosen as ,he uni, of s,udy. A dclailrd descrip,ion of the rewearct, design, methods. and i,,,ervcwio,l activi,ies has been published elsewhew.” Here, we preen, ,t,r main res,,l~~ of ,lw \,udv.

Thr Mi,i,w~,a (:a,,cer Pail, Project (.W:PP) is a randomized c-ornmuniw trial and m,plo!a both nrrtrd cohort and nrstrd cro~,wc,io,,al designs.“’ Six medium-sized communi,icr located a, leas, .50 milps from thr Yimlrapolis-S,. Paul me,rc~poli,an arc2 ucrc rrlrrred for the s,ud): Selec,ion nctuircmtnn included (I) more ,han 2.5 physicians. (2) a, least .50 but no more than 2% mile:. from Ihe Minneapolis-S,. Px! mc,rr3poli,an drca. (9 diwwel~ defined mrdical senirc arca withour owrlap with other srudy communi,ir\, (4) a\ailabte pool of rimita,- ci,ies ,o permit pairing. (5) geographic halancr across Xlinncw,a, (6) heal,h
Sur\ ~3 WCTC~c-ollrc,rd .U bawlinc and df ,‘,Ithe i,,,cncw,ion in CACTIof ,bc \I\ ~omm,~~~~. ,ics. Ph)slcianr and nurses acre s,lr\-r\ed uri,q a t,cwd rnhon dcs,g,,. ” In ,hiz design. intact wcial groups arc. all~a,rd 10 ,rva,mw, or co,~,roI cortdi,~ow and individuals arc lolIrwcd o~cr time ax a cohor, ,o ;,SU’W the impdc, of the intcrvrn!io,i. In this ;tpplic;,,i,m, whole c-ommuni,io WTP r;mdomi~rd ,,I ,~a,mea, or conlrol condi,i~mr, .,od ,ticir ptn\G(.ia,,s and mwws wrc followed .I< ., coho,, ,o aswss the impar, of the inrrrvcnrior, on ,hcir kmwl~dg~ and a,,i,,,dlrs towrrd canrrr pain lllallq(Cml~rl~. (Iatlccr pa,icri,\ and ,tirii fdmih rarcgiwrs wrrc wr\r!vd wing a nc\,rd cra)w rrriond drsign.X! In ,his drsign. in,x, VKial glo,lp~ arc allcaGa,rd 10 ,rra,n,r,,, or ccu,rrr,l c-ondi,ionr. ar,d srparatr cror\-arc,ion;tl sampk~ of individuals .,re nbwrwd owr ,imr ,o ;LEVSS,he impact ofthr in,crrm,ion. In ,hi* q~pl~ra,~cm. wi~rlr rommunirirr HCW randomilcd ,o ,rea,men, or comrol comti,irms, and cancer pa,irn,* and rhrir family carrgi\en wrre s,,rve)rd a, baseline. .4 new ample ~‘35 swvrwd 15 momhr lacr ,o aww thr impac, of ,hr in,cr\m,ion on the paiems repor, of c-ancrr pain and on ,ht. paarirnts’ and family members knostedge and attitudes ,o~~d CPY. A croswe~ ,ional wrvc~ of MO indrpenden, samplrs ofpairmr was ;,ec-rwrv btu,w of the high mor,ali,~ raw in tbir popr~lat~cu~.

A L’+ ,niniliWnvship wils Iwld a, the iniua,ion of the in,ervcn,ion per;,>d in .Minnrapolis-SL Paul ,o protide swr-of-the-x, CPM k,,owlcdpe .,nd skills. and ,o promote mm approp,ia,e CPM ;,,,i,,,dcs rind beharion .,mong ,hr 27 community opinion leadr, rlinirianr from the thrcr in,erv.w,ion co,n,n,r ni,io. l’hr c>pinion leaders were s&r,& b! ,heir rxprcswd willingnrs~ ,o par,icipa,r in the studv and b rcrommcndaions of Uwir comm&i,v peers a.5 pfxrntial~ rftrcrivr opinion laden using a method similar ,o tha, described by Hiss c, al.“” This double endorwemm, procrduw ~-as designed ,o idend@ the natural opinion leaders with regard 10 CP.54 and ,o dcwlup .I *msc of protqar~ owwnhip within, ,hc communi,v. The minifcllowship , onsistcd of didactic prcwmarions and clinical precrprorrhipr with experiential clinical rounds i,l o,~col~~ i,,pa,ien, units and ho+

lion.” This questiormailt cwtta~n(.d I.5 kmntledge items answcrrd with it live-point b(illc’ and nine attitude items dnswcred with 3” I I-point scale. Hlth two of tbr ltrnn omiurd. the total knowledRr score\ bad (:ronbach alpha scorch of 0.68 for physiri;ws arid 0.67 for ~S(IXC. The total knowledge ~-ore for pb!sicians and IIUISCS had a powible mngc 4 I.345 With two of the items omitred. the total atCrude .scorcs had Cronhacb alpha S(‘ICC?I 1~1 0.58 for physicians and 0.77 fur nurses. The total attitude score for physicians and niir.ws had a possible range of O-77. These tot.11 fores were reported as mean v&es Girt, low scores preferred. Further details of imtrw merits and their application in this study ha\<, been published else~hrrr.“.‘“~” C:opiw of the study inwumrnu arc xailahle on rc~qw~

The MCPP difTers from [be usual c-linical trial because intact social groups rahcr tbdn individuals were randomized to ~eatment and control conditions. Common to all such community rrials is the expectation that prrwnr within an intact .social group will be more sirnilar to one another than to persons in the other imax social group~.~ This positiw innclas correlation must be taken into a~-count both in the design and in the analysis of il community trial to avoid what can be a highI! inflawd Type I error rate: “’ The cl-irical step i% to account for the rxtn variation awiburablc~ to communities that is above and beyond thr. variation among perwnr witbio the co~n:~::i~~ties and is confounded with dir variation due to treatmmls.“’ We accounted for this extra variation with a mixed-model regression analysis wing SAS PROC MIXED,” a prognm particularl~~ wited IO analysis of data from community tri:d5.‘L) For the physicians’ and mwrs’ data. we conducted a repeated measures analytis of cowriante (ANCOVA) in which condition. rime, and their interdction were included as fixed effects, and community w&s included a a random effect nested within condition; variahlcs identified as potential confounders WCI‘C included as fixed covariatrs. This approach accounted for the conelation among respmdents within a single community and for the correlation in a single respondent’s data over time. Repeated measures ANCCY’A is Ed to

WC- also atnl)wd die physician and nurws’ data with analysis of cowrinncr in which condition w;n included as a lixcd effect. community was itlclodrd as a rutdon, ellect nes~d withit) condirion. and the rv~pwlde111~s ba.wlinv \aluc 011 1bc 01tlcc~me measure w:lli included ilc a fined Lvariate in addition to the potrtllial confounders. This appwach also ac-rountcd for the c orrelation among rerpow clcntr wiltiin a *in& commimi~ and allowed tlw data IO dictate lbr wgrc’ssicm wefficiem linking tbc bawliw and follow-up dam. For thr p”tivnl\ ;IIKI cxrgiwrs’ da& we condurtrd a tinw by condition ANC:OVA. witbout wpeawd mranw~, in which condition. lime. and their interaction WC~P included a\ fwd rflrcu. and community within condition wa\ included as a random effect nc5ted within condilion; po(enrial confounders were included as l&d covari:itrs. This approach acc-oimted for tbc correlation among respow denr\ within a single wmmttnity and reflccwd the fart that independent samples were inrltded in the baseline and follow-up surveys. FOI &I analyses. ;L Gaussian disoibution wac asstmwd for both the individual and community components of variance. In addition to these primary amdyws. an ellort was made to examine dw re!ationrhip 4 the IrveI of confirmed rapa,nure to the intervendon program 10 changes in knowledge and attitudes in nurses and physicians. Nurses and physicians wcw asigned a one on the expo. wre mc.~m if they had attended itt least olx intervention program. as confirmed by regi+ [ration records. Tbosc who had not attended, together with all norscs and physicians in the control communities, were ;Lwigned a aw) on rhr exposure n~+swe. The follow-up actitudr ;:nd knwlrdgr scores ucrc then ni,+.+ for differcows arsociatcd with the two exp>wre lewlr. after adjurtment for ba.sclir?r levels on the outcome and other powotial confounders. kkra~w SAS PRO(: MIXED employs maximum likelihood estimation, perwns mixGng either haxliw or the follow-up obcerwtion were retained tn the repeated meast~rcs ANOVA. All analyses were run using SAS 6.0!4.”

E&zz ,, al.

IO?

Main

I’d. I3 I% 4 A/nil 1997

Outcomes

P&-n/, md farnib mm&m. The rewl~s Tron~ thr rcpeatcd meas,~rcs ANCO\‘A for tbc patient and caqiwer data are wmmariwd in Table 2. For patients. the prc\alcnre ol pail] declined and tbr pain managrmrnt indrn improved in the trcatmcnt condition rrlatiu: to the control condition; this *a% also true L)r the pain management indrx when the anal!G was restricted to patirnlr reporting pain at the time of the interview. On thr other hand. tbr pain intensity score and the total attitude score in patients and the moral attitudr xow in caregivers deteriorated in the treatment condition relatiw to the control condition. How ever, none of these changes wcw statistically signilicant. Ph@ianr and nunrs. The results from the repeated measures ANCOVA for the physician and nurse data are summarized in Table 3. There were improvemenls for all rnri~~~nrs in tk treatmmt group relativr to thr control group. but none of these changes wcrc statistically sip@icant.

DiSCUUi~ M’c studied the vffccts of a romnlunit!. bawd. multimodal rcluca,ional intrnrntion on lancer patienu pain intcnsitv and tbrir attitodrs about (:PM. farnix mrmbcrs attitutlrs about CPM, and physi&’ and mtrws knowlrdgc and attitudes ahout CPM and tbrir clinical beharion ar mr;wtrcd by a pain managemmt index. .4lthough the inwrvention did not wsult in significant improvements in thcw variables, thr trends KC~C ir, the right direction for most of the mrawwd wtcom~s. Furthermorr. thr intervention rxposure an+ir of ph+cian xnd nurse knowledge and attitude vxiabter suggeswd that more intense intrrvention may have been herter. The haeline scores of cancer patients’ rrporttd pain and of their physicians’ and nunes’ knowledge and

m

t:lhJlt

in CPM and have many competing imclca for their lime and eITorts. .krording to llxri\. community hospitals wide les.. cobcsiw mcdiCal stafk, and especially solo wd small group practices with loose xffiliations. in a f~:e-fo~~service environmmt dppeal 1r.u amen;;blc to change.‘” Cliriical community setting is dn important determinant of which srl-awgy is likely to be cflecti\e. Mitmian et al. haw plw posed that the opinion leader eduralion model. along with *old! groups, and wciali~alion programs. may bc appropriate strawqicr to change clinical behaviors for community hospitals and loosely aftiliarcd ph!sicidns.” t3ecau.w of this anwrtion and tbc drmonstrated eflertivencrs of this model in otbc~ studies, dir cow of our study’s inw vrntion was the opinion leader rducatol: Variable\ rtia~ we beliew may impact on thr cfTectivrnc\s ol the opinion leader education model are local clinician culture. orgdnimlional strut-mrc, ano care setting. These variabler, plus the commw nication style and leadership quality of the opinion leader educator. further determine the process and eventual outcomes of this model. Second. the intervrntion as applied may not have heen implemented long enough or Itill! enough. The intervention zxposurc analysis suggests that more exporure to the interwntion would have been better. Third, our intervrntion did not include practice performance feedbach LO indi\idus;, or group practices, which may have enhanced ill effectiveness. However, an cflective opinion leader education trial in obsreuical c&c did no1 use practice performance fccdhark rither.‘lL Fourlh, secular trends toward better CPM practices and fawrahle baseline wiablcs may have diminished the cl~ances of any intcrvenlion having a significant effect. In this study, the baseline cancer patient pain scores were already good [mean, 10.52 (SE, 0.95): range. O-40. with low score preferred]. AI bawline, 84% Of all cancer pa&x& reporting pain in the 3 months prior to the interview and 59% of cancer patienti reporting pain at the dmr of the interview had positive pain managrmerit indices suggesting adequate CPM, rompared to 58% of canwr patienrs reporting pain or taking analgesic drugs daily during the week preceding the interview in the study by

d nl

I ;,/. I3 .\‘,I. 4 A

I I w7

Llrcland et al.“’ The bawliw knowledge and attitude wws of phyGcians and ~NY’P also WC’I‘Cg(wd. At tuwliw. their knowledge and artitudes mean wows were ,jw within the top quaile of both ualrs. Although there is nred IOI- improveme~~t. further innrments in thew vnd points may be difficult 10 achieve. Filth. (X’SI is a complex c-liniral activity wherein the clinical end point. pain, is diff% cult to arvx .md treat. Pain i\ a rut+cti\r rxpwience. and it* aw~sw~cnr can he rhallengiiig. (:I’!4 utilizes multiple trratmrnl modalitira and clinical dixiplines. controlled ~ubstancc~. and occasionally complicated wchnologies. Pain management responsibilities al-r dish ibutcd mmdtilatrrally w the patient. family, and several clinicians. Analgesic side eftrc~s can be rhalleuging to control and can contritniw to paw analgesic 011tcomes. Must experimental studies reporting improved practice beheviors and patient outcomes have focawd on simple clinical activities such as screening nests. ordering laboramry testc. or prescribing drugs in conditions with unambiguous indications and objcctiw end points. Our study has several strengths. The experimental design allowed a rigorous evaluation of the community-bawd interventions. WC found consistent results across the six communitie.. The primary outcome-patients’ report of cancer pain--was u~.:ain?d face-m&ace utiliring a rtaudard assessment tool. Pxticipant wtponr rates were high and essentially rrpresentadve of thr entire group of eligible sub jrcn, thus reducing external bias. WC believe there were no obvious methodological flaws. The primary results are based on actual ctiniral performance and patient outcomes. Tbr intervention strategies were built upon rhe reported experience of others and relevant theoretical models. E\,en so. this study has several limitations. First, only one follow-up survey w’as done shordy after the 15month interwndm ended. Change may have occurred later and hat been mind hy our study. Second. we did plot measure patients’ health-related quality of life by any standard tool and may have missed an imporunt outcome that could not hc asseswd hy thr pain variables. A modest drcremem in pain intensity may have major effects 011 patients’ quality of life.‘& Third, we did not measure the change in knowledge and attl-

tude about CPM of each participant in e,rn romponent of the rducational pwgmm. WC mrasurrd the changes in knowledge and auiwde of a reprcscnratiw sample of community clinicians, some of whom participated in thr inwrwntion. Substantial empirical evidencr exists that shows significant improvrmmt in knowledge and attitudes occurs from partiripadon in similar rducarional programs.“‘-“’ Our study focused beyond these traditional \-Gables and assessed the opinion leader education rffect in the community on practice behaiors and patient outcomer. Fourth. pub lit education and mrdia ewnh were minimal. Thcrrforc. lay community awxcnes~ trwnd the clin,ral rommunity WAS probably slight. Finally. dirc,t t rtient and farnil! education abou: ~CPM by study investigators was nr)t done. Although we pro\idcd conununily opinion trader clinicians with thcsc critical trchniqurs, WC depended upon them to implement this straw-a). Several of the wtcornc measur-r depend& on their imptemrnration. The -cwl& of this study may be applicable to olhrr comparable settings and similar clinicat act,vities. The clinical communities Ycrc compoud mostly of primary care clinicians in a fee-fc,r-srrvicr rnvironmcnl, working in Ioosety organized small group p,xticn with informal structures and systuns. Rerauw CPM is a complex clinical arCvity. other similar clinical rctiGtic3 may respond only msrginall! to thi* intervention model as well. Our findings have impoti,I implic-ation* fC,r faure work in this area. First educadonal interventions mav 1x4 to be more intense and po longed. Perhaps owr 3-5 yrars, to acbiew desired and durable res&s. Second. a comhinatin of rduracional nwbods seems nrccsw~ in order to increaw the rlfcct Third. clinical setting and practice behaior context are probabl) important determinants for sel~~dng potrndal~ effective edu~tion.4 intenrntions. Thhr opinion lrader educator model may nrcd rcxision to Ix effective on CPM in communit) pracliccs. Finally, we suggest that s~wral additional. unrested straregir\ should he ro,,sidrred for future imerwntions in ccn~trolled vial\ such a~ (a) continuous qualily impmwmrnt metbodotogy using group consensus building. decision making techniques, and rewards for desired behaviors; (h) public education and media campaigns; (c) feedback on individual and

-202

__--

/l/hll___-I-, ,,I.

tic ! i .Yr,. .I ., ,“/ l wi