Recruiting & HR

The Natural Progression of Health-Related Quality of Life: Results of a Five-Year Prospective Study of SF-36 Scores in a Normative Population

Description
The Natural Progression of Health-Related Quality of Life: Results of a Five-Year Prospective Study of SF-36 Scores in a Normative Population
Published
of 10
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  The natural progression of health-related quality of life: Results of a five-yearprospective study of SF-36 scores in a normative population Wilma M. Hopman 1 , Claudie Berger 2 , Lawrence Joseph 3 , Tanveer Towheed 4 , ElizabethVandenKerkhof  5 , Tassos Anastassiades 6 , Jonathan D. Adachi 7 , George Ioannidis 7 , Jacques P. Brown 8 ,David A. Hanley 9 , Emmanuel A. Papadimitropoulos 10 & The CaMos Research Group 111 Clinical Research Centre, Kingston General Hospital and Department of Community Health and Epidemiology, Queen’s University, Kingston, Ontario, K7L 2V7, Canada (E-mail: hopman@kgh.kari.net); 2 CaMos Methods Centre, McGill University, Montreal, Quebec, Canada;  3 Department of Epidemiology and Biostatistics, McGill University, Montreal, Quebec, Canada;  4 Division of Rheumatology, Department of Medicine, Department of Community Health and Epidemiology, Queen’s University, Kingston, Ontario,Canada;  5 Department of Anesthesiology; School of Nursing; Department of Community Health and Epidemiology, Queen’s University, Kingston, Ontario, Canada;  6 Division of Rheumatology, Department of Medicine, Queen’s University, Kingston, Ontario, Canada;  7 Department of Medicine, McMaster University,Hamilton, Ontario, Canada;  8 Laval University, Ste-Foy, Quebec, Canada;  9 University of Calgary, Calgary,Alberta, Canada;  10 Eli Lilly Canada Inc, Toronto, Canada and University of Toronto, Faculty of Pharmacy,Toronto, Canada;  11 See Acknowledgements for complete list Accepted in revised form 14 August 2005 Abstract Background  : Limited information exists regarding the natural progression of health-related quality of life (HRQOL) in the general population, as most research has been cross-sectional or has followedpopulations with specific medical conditions. Such norms are important to establish, because the effectof any intervention may be confounded by changes due to the natural progression of HRQOL overtime.  Methods : Participants were randomly selected from 9 Canadian cities and surrounding rural areas.Changes in the eight domains and 2 summary component scores of the Medical Outcomes Study36-item short form (SF-36) were examined over a 5 year period (1996/1997–2001/2002). Mean changeswere calculated for men and women within 10 year age categories. Multiple imputation was used toadjust for potential selection bias due to missing data.  Results : The baseline sample included 6539women and 2884 men. Loss to follow-up was 17% for women and 23% for men. Mean changes tendedto be small, but there was an overall trend towards decreasing HRQOL over time. Changes were morepronounced in the older age groups and in the physically oriented domains. Younger age groups tendedtowards small mean improvements, particularly in the mentally oriented domains. Large standard errorssuggest that on an individual level, large improvements in some participants are balanced by largedeclines in others.  Conclusion : In general, the HRQOL of Canadians appears relatively stable over a5 year period. However, care should be taken when assessing HRQOL longitudinally in certain age orgender groups, as changes associated with an intervention can potentially be confounded by the naturalprogression of HRQOL. Key words:  Longitudinal, Normative, Prospective, Quality of life, SF-36 Quality of Life Research (2006) 15: 527–536   Springer 2006DOI 10.1007/s11136-005-2096-4  Introduction The widespread use of the Medical OutcomesStudy 36-item short form (SF-36) [1, 2] for theassessment of health-related quality of life(HRQOL) has resulted in a large body of literatureestimating the burden of a variety of diseases andassessing the benefits of a wide range of treatmentsand interventions [1–6]. The content of the SF-36is based on earlier work of the Health InsuranceExperiment (HIE), which had aimed to constructthe best possible scales for measuring a broad ar-ray of functional status and well being concepts.The HIE clearly demonstrated that such tools werevalid and reliable for assessing changes in healthstatus in the general population [1]. However, littleinformation is available regarding the naturalprogression of HRQOL in the general population,as most research using the SF-36 has been cross-sectional or has followed populations with specificmedical conditions [3, 7, 8].One longitudinal, population-based studyexamined British residents between 35 and55 years of age at baseline, with a mean follow-upof 36 months [7]. They noted that cross-sectionaldata underestimated the within-person declinesassociated with increasing age, and emphasized theneed for repeated measures of the SF-36 in pop-ulation-based studies. A secondary analysis of asubset of civil servants between the ages of 54 and59 at baseline noted small declines in physicalfunctioning in both men and women, althoughmental functioning remained stable [9].A second population-based study assessed par-ticipants in the Canadian Multicentre Osteoporo-sis Study (CaMos) [8] who were between the agesof 40 and 59 years at baseline, also over a 3 yearperiod. In general, it appeared that while the meanHRQOL of the participants remained fairly stableover 3 years, declines were more evident in thephysical than the mental domains, as in the Britishstudy. Although mean HRQOL changes weresmall, the standard deviations of the observationswere generally high, suggesting that a number of participants had large improvements while othershad large declines, essentially canceling each otherout. This suggests that changes in HRQOL fol-lowing an intervention are probably not con-founded by declines due to population aging, aslong as the follow-up period is of relatively shortduration, but whether this holds over longerperiods of time is unknown. Moreover, little isknown about the stability of SF-36 scores inyounger (under 35 years) and older (over 60 years)men and women.The initiation of CaMos in 1996 provided theopportunity to develop age- and sex-standardizednorms for Canadians [10], assess regional differ-ences [11], and more recently, examine changesover a 3 year follow-up for a subset of participants[8]. A follow-up of the full cohort took place5 years after the baseline assessment, providing theopportunity to lengthen the follow-up periodcompared to the 3 year assessment, and expandthe age range to include all adults ages 25 yearsand over. This information will provide the re-search community and health care professionalswith some guidance as to how much change inHRQOL over time can be attributed to a naturalevolution of HRQOL, and what can be attributedto an intervention or the progression of a chronicillness. Methods CaMos is an on-going, prospective cohort study of 9423 non-institutionalized, randomly selected menand women 25 years of age and older at baseline,drawn from a 50 km radius of nine Canadian cities(St John’s, Halifax, Quebec City, Toronto,Hamilton, Kingston, Saskatoon, Calgary andVancouver). A detailed description of the purpose,methodology and sampling framework for CaMosis available elsewhere [10, 12]. In brief, householdswithin each region were selected by random drawsof listed telephone numbers, and one randomlyselected household member 25 years of age orolder was asked to participate. Ethics approvalwas obtained through the Review Boards of eachparticipating centre.Baseline assessments took place betweenFebruary 1996 and September 1997, and the5 year follow-up took place between February2001 and September 2002. The interviews wereconducted in person at both time points, and in-cluded sociodemographic, family history, riskfactors and lifestyle questions, as well as theMedical Outcomes Trust 36-item short form(SF-36) [1, 2]. The SF-36 yields 8 domain scores528  including physical functioning (PF), role physical(RP), role emotional (RE), bodily pain (BP), gen-eral health perceptions (GH), vitality (VT), socialfunction (SF) and mental health (MH) [1], as wellas a Physical Component Summary (PCS) and aMental Component Summary (MCS) [2]. Al-though a more recent SF-36 version is now avail-able [3], version 1.0 was used to increase thevalidity of comparisons between the two timepoints. In order to assess change over time, base-line scores were subtracted from the year 5 scores,so that positive values indicate improvement inHRQOL.Multiple imputation [13] was used to adjust forpossible selection bias, as not all subjects whoprovided baseline data also participated at year 5.Participants who had died were omitted from theanalysis, because change scores are not defined forthis population. Imputed values were derived fromregression models predicting SF-36 changes over5 years, using data from respondents with com-plete data at both time points. We developedseparate models for each age and gender category.Variables considered for the models are presentedin Appendix A, and were selected on the basis of empirical evidence and/or demonstrated associa-tion with HRQOL in other CaMos analyses.Variables entered into each of the models differed,depending on the age and gender category.The BIC model selection criterion [14] was usedto select the best predictive model for each domainand summary score. We then entered baseline datafrom those missing year 5 data into the models toimpute the changes in scores. Simultaneousregression parameter estimations and multipleimputations for all missing data were carried outvia a Gibbs sampler algorithm as implemented inBugs software (version 0.603, Cambridge, Instituteof Public Health, MRC Biostatistics Unit, 1995).We ran up to 20,000 iterations of the Gibbs sam-pler to ensure accurate estimation. Multiple valueswere imputed for each missing item, and final re-sults were averaged across all imputations. Themultiple values account for the fact that themissing data predicted from the regression modelsare not known with certainty. Results Table 1 outlines the loss to follow-up over the5-year period. More women were lost to follow-up, but the percentage of men was greater (1109 or17.0% women; 662 or 23.0% men). However, asubstantial number of these preferred to do a shortquestionnaire over the telephone (282 women; 147men), and as a result they did not have SF-36 dataat the time of the 5-year follow-up. The otherreasons for loss to follow-up include death (401women; 243 men); unable to contact (149 women,128 men), not interested (92 women, 25 men), toosick (69 women, 29 men), moved away (64 womenand 59 men) and other (52 women, 31 men) whichincluded reasons such as cancelled, no time or noreason given. Comparison of those with and without 5-year data The baseline scores of those with and without5-year follow-up data differed for a number of age-domain and age-component summary categories.For the women, those with a 5-year follow-up hadhigher scores in 40 of 48 (83.3%) age-domain cellsand 11 of 12 (91.7%) age-component summary Table 1.  Sample size and loss to follow-up at year 5, by age group and genderAge group(years)Women MenBaselinetotal 1996/19975 yeartotal 2001/2002Lost tofollow-upBaseline total1996/19975 yeartotal 2001/2002Lost tofollow-up25–34 200 170 30 (15.0%) 200 152 48 (24.0%)35–44 286 261 25 (8.7%) 212 173 39 (18.4%)45–54 1111 1024 87 (7.8%) 587 486 101 (17.2%)55–64 1639 1461 178 (10.9%) 640 554 86 (13.4%)65–74 2134 1771 363 (17.0%) 802 625 177 (22.1%) P 75 1169 743 426 (36.4%) 443 232 211 (47.6%)Total 6539 5430 1109 (17.0%) 2884 2222 662 (23.0%) 529  cells. For the men, those with follow-up data hadhigher scores in 41 of 47 (87.2%) age-domain cells(the mean score for one cell was identical) and 11of 12 (91.7%) age-component summary cells.Higher scores represent better health status, so ingeneral, those with follow-up had a better meanhealth status at baseline than those lost to follow-up. Results from those who provided data at bothtime points may be biased, and therefore the re-sults in Tables 2–5 are based on the multipleimputation adjusted mean differences in scoresfrom baseline to 5 years.Variables selected for the best models for impu-tation of missing datadiffered for men and women,across the age strata and for each domain andsummary score, and therefore are not described indetail. Appendix A contains the list of variablesthat contributed to change in one or more domainsof the SF-36. In general, the baseline values of theSF-36 domain scores or component summaryscores were included in the models, and these werealways the strongest predictors. Mean changes The estimated mean changes for women and menare presented in Tables 2 and 3, respectively, alongwith 95% credible intervals (CrIs). The CrI is theBayesian analogue to frequentist confidenceintervals, and provides an interval within whichthe true parameter value lies with 95% probability,given the data and any prior information includedin the model. As we used very low informationprior distributions for all analyses, the 95% CrIsare essentially based on information from the dataalone. While the mean changes were not large, thestandard deviations (SDs) of the observations weregenerally high, suggesting that large HRQOLimprovements in some participants were cancelledout by large declines in others. Table 2.  Estimated mean SF-36 change for womenAge group at baseline SF-36 Domains/Summary scoresPF RP BP GH VT SF RE MH PCS MCS25–34 ( n  = 198–200)*Mean difference  ) 0.6 2.8 0.8 1.3 2.0 2.9 5.0 2.5  ) 0.1 1.795% CrI of mean difference  ) 1.3;0.1 1.1;4.5  ) 0.3;1.8 0.5;2.1 1.1;2.9 1.9;3.9 3.4;6.6 1.7;3.1  ) 0.4;0.3 1.3;2.235–44 ( n  = 268–285)Mean difference  ) 1.7 1.7  ) 1.9  ) 0.6  ) 0.4 2.9 2.8 0.9  ) 0.7 1.195% CrI of mean difference  ) 2.2; ) 1.1 0.5;3.0  ) 2.7; ) 1.1  ) 1.1; ) 0.1  ) 1.0;0.2 2.1;3.6 1.6;4.0 0.4;1.4  ) 1.0; ) 0.4 0.7;1.445–54 ( n  = 766–1101)Mean difference  ) 3.0 0.0  ) 1.4  ) 0.5 0.7 1.5 2.4 1.6 1.4 1.395% CrI of mean difference  ) 3.3; ) 2.8  ) 0.6;0.6  ) 1.8; ) 1.1  ) 0.7; ) 0.2 0.5;1.0 1.2;1.8 2.0;2.9 1.3;1.8  ) 1.5; ) 1.2 1.2;1.555–64 ( n  = 1221–1604)Mean difference  ) 3.7  ) 2.0  ) 2.0  ) 0.9  ) 0.3 0.9 2.1 1.6  ) 1.6 1.395% CrI of mean difference  ) 3.9; ) 3.4  ) 2.5; ) 1.5  ) 2.3; ) 1.7  ) 1.2; ) 0.7  ) 0.5; ) 0.1 0.6;1.2 1.6;2.5 1.3;1.8  ) 1.7; ) 1.4 1.2;1.465–74 ( n  = 1865–2013)Mean difference  ) 7.1  ) 4.0  ) 3.2  ) 2.0  ) 4.1  ) 0.7 1.2 0.4  ) 2.5 0.795% CrI of mean difference  ) 7.4; ) 6.8  ) 4.6; ) 3.4  ) 3.6; ) 2.9  ) 2.2; ) 1.8  ) 4.4; ) 3.8  ) 1.0; ) 0.4 0.6;1.7 0.2;0.6  ) 2.6; ) 2.4 0.6;0.875+ ( n  = 555–999)Mean difference  ) 11.8  ) 2.9  ) 2.7  ) 4.0  ) 5.8  ) 3.6 1.7  ) 0.5  ) 3.5 0.695% CrI of mean difference  ) 12.5; ) 11.1  ) 4.7; ) 1.2  ) 3.4; ) 2.0  ) 4.5; ) 3.4  ) 6.4; ) 5.3  ) 4.5; ) 2.7 0.6;2.8  ) 1.0;0.1  ) 3.8; ) 3.1 0.3;0.9* These values represent the range of the sample size for each age category. The sample size varies due to the differing amounts of missing data available for the imputation models.CrI = Credible Interval.Bias adjusted through multiple imputation.PF = Physical function, RP = Role physical, BP = Bodily pain, GH = General health perceptions, VT = Vitality, SF = Socialfunction, RE = Role emotional, MH = Mental health, PCS = Physical component summary, MCS = Mental componentsummary. 530  Overall, for the 48 age-domain cells, the numberof mean scores that declined outnumbered thenumber of mean scores that improved, for bothmen and women. Women reported HRQOLimprovements in 21 of 48 (43.8%) age-domaincells; declines were noted in 27 of 48 (56.3%) age-domain cells. Men reported improvements in 19 of 48 (39.6%) age-domain cells, and declines in 29 of 48 (60.4%) age-domain cells.For the component summary scores, improve-ments and declines were close to equal. Womenimproved in 7 of 12 (58.3%) and declined in 5 of 12 (41.7%) of the age-component summary cells.Men improved in 5 of 12 (41.7%), declined in 6 of 12 (50.0%) and stayed the same in 1 of 12 (8.3%)of the age-component summary cells. Declineswere much more apparent in the physically ori-ented domains of PF, RP, BP and GH and thePCS, for both men and women. By contrast, thementally oriented domains of VT, SF, RE andMH were more likely to improve, which was morepronounced in women. Size of mean changes There is debate in the literature about whether fiveor ten point changes are required for clinicalrelevance. A minimum change of five points isconsidered clinically and socially relevant by some[1], while others consider a larger change of tenpoints as important [15, 16]. For both men andwomen, older age groupings showed much moredecline than the younger age groupings, particu-larly in the physically oriented domains. Forwomen, the PF for those 65–74 and 75+ years hada mean decline that exceeded five points. The meanchange in VT also exceeded five points for the agegroup of 75+ years.For men, mean changes greater than five pointswere seen for the PF domain in those 65–74 and Table 3.  Estimated mean SF-36 change for menAge group at baseline SF-36 Domains/Summary scoresPF RP BP GH VT SF RE MH PCS MCS25–34 ( n  = 53–200)*Mean difference  ) 2.7  ) 1.0 2.4  ) 1.8  ) 3.7 1.3 0.7 0.4  ) 0.7 0.095% CrI of mean difference  ) 5.0; ) 0.4  ) 3.3;1.1 0.9;3.9  ) 2.9; ) 0.6  ) 4.8; ) 2.5  ) 0.1;2.8  ) 1.5;3.0  ) 0.9;1.6  ) 1.2; ) 0.2  ) 0.6;0.735–44 ( n  = 117–212)Mean difference  ) 0.3 2.3 0.6  ) 0.3  ) 0.7 1.3 8.0 0.3 0.0 0.595% CrI of mean difference  ) 1.2;0.5 0.6;4.0  ) 0.8;2.0  ) 1.2;0.6  ) 1.6;0.2 0.3;2.4 6.3;9.8  ) 0.6;1.2  ) 0.4;0.5  ) 0.0;1.045–54 ( n  = 544–575)Mean difference  ) 1.3 0.4  ) 1.0  ) 1.1 0.5 3.4 5.4 2.6  ) 1.1 2.095% CrI of mean difference  ) 1.8; ) 0.8  ) 0.5;1.3  ) 1.6; ) 0.3  ) 1.5; ) 0.6  ) 0.0;1.0 2.9;4.0 4.7;6.2 2.1;3.0  ) 1.3; ) 0.9 1.8;2.255–64 ( n  = 454–627)Mean difference  ) 1.6 0.1 0.6  ) 1.0  ) 0.5 1.1  ) 0.2 1.5  ) 0.4 0.595% CrI of mean difference  ) 2.0; ) 1.2  ) 0.7;1.0 0.1;1.2  ) 1.4; ) 0.6  ) 0.8; ) 0.1 0.7;1.6  ) 0.8;0.4 1.2;1.9  ) 0.6; ) 0.2 0.3;0.765–74 ( n  = 703–735)Mean difference  ) 7.8  ) 4.4  ) 2.4  ) 3.2  ) 4.4  ) 2.7  ) 1.6  ) 0.8  ) 2.6  ) 0.295% CrI of mean difference  ) 8.4; ) 7.2  ) 5.4; ) 3.2  ) 3.1; ) 1.8  ) 3.7; ) 2.7  ) 4.8; ) 3.9  ) 3.3; ) 2.1  ) 2.5; ) 0.7  ) 1.2; ) 0.4  ) 2.9; ) 2.3  ) 0.4;0.175+ ( n  = 319–350)Mean difference  ) 12.5  ) 7.2  ) 2.8  ) 4.4  ) 8.6  ) 4.3 2.9  ) 0.5  ) 4.4 0.595% CrI of mean difference  ) 14.3; ) 10.7  ) 10.2; ) 4.2  ) 4.5; ) 1.1  ) 5.6; ) 3.2  ) 9.9; ) 7.2  ) 6.0; ) 2.7 0.6;5.2  ) 1.6;0.6  ) 5.2; ) 3.7  ) 0.1;1.1* These values represent the range of the sample size for each age category. The sample size varies due to the differing amounts of missing data available for the imputation models.CrI = Credible interval.Bias adjusted through multiple imputation.PF = Physical function, RP = Role physical, BP = Bodily pain, GH = General health perceptions, VT = Vitality, SF = Socialfunction, RE = Role emotional, MH = Mental health, PCS = Physical component summary, MCS = Mental Component sum-mary. 531
Search
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks