A listing of the characteristics of the 44 control group patients is presented in Appendix 6, together with the subject's original prospective Rosser-Kind estimate (rating) of his/her health-related quality of life, the derived prospective Rosser-Kind estimate (rating) of his/her health-related quality of life, the subject's Rosser-Kind estimate (rating) after three months, the derived Rosser-Kind estimate (rating) after three months, the subject's retrospective Rosser-Kind estimate (rating) of his/her health-related quality of life prior to the operation and the derived retrospective Rosser-Kind estimate (rating) of his/her health-related quality of life prior to the operation. Thus, there were six health-related quality of life scores for each control group patient. Three of the scores were derived from the subject's own positioning on the Rosser-Kind Classification at the preoperative, postoperative and retrospective stages. The other three were derived from an assessment of each patient's health-related quality of life at the same stages, based on their answers in the interviews.
3.31 Control Group: Comparison of Rosser Classification with Questionnaires
The first aim of the study was to compare the use of the Rosser-Kind Classification of Illness States with the questionnaires as tools of measurement in health-related quality of life estimates. The most common method of comparing two such measurements is to calculate the correlation coefficient (r) between them. Considering the detailed nature of the questionnaires, one might the expect the two approaches to yield different results and the correlation coefficient between them to be quite low. However, the correlation coefficients between the patient estimates of their
health-related quality of life (based on their positioning on the Rosser-Kind Disability/Distress Scale) and the derived estimates of their health-related quality of life (based on their responses to the questionnaires) were very high at every stage of the process: 0.899 at the preoperative stage, 0.972 at the postoperative stage and 0.931 at the retrospective stage. In addition, the correlation coefficients between the patient estimates and the derived estimates were high in every category of the operation site and sex subgroups. At the preoperative stage, the correlation coefficients were 0.930 for the hip subgroup, 0.788 for the knee subgroup, 0.893 for the male patients and 0.902 for the female patients. At the postoperative stage, the correlation coefficients were 0.953 for the hip subgroup, 0.989 for the knee subgroup, 0.961 for the male patients and 0.989 for the female patients. At the retrospective stage, the correlation coefficients were 0.947 for the hip subgroup, 0.908 for the knee subgroup, 0.938 for the male patients and 0.944 for the female patients. Though the correlation coefficient measures the strength of a relation between two variables, it does not tell us whether we can reliably use both variables interchangeably. As Bland and Altman (1986) inform us, a more appropriate method would be to test the agreement between the two variables by plotting the difference between them against their mean. Therefore, tests of agreement were conducted between the two variables at each of the three stages of interviews.
(a) Preoperative Stage
Figure 3.1 (page 52) displays a simple scatter diagram of the subjects' prospective rating scores and the derived prospective rating scores, with the line of agreement for the two sets of data. Figure 3.2 (page 52) plots the difference between the two sets of data against their mean.
Fig. 3.1: Control Group: Relationship between subjects* o o « S •■§ £ 2 > |l 8 S 0.4- 03-
Fig. 3.2: Control Group: Test of Agreement between subjects’ prospective (SUBPRRKR) and the derived prospective
(MYPRRKR) rating scores
Mean + 2SD pa □ cn -0.2 -03 -0.4 - 2— <2 t
%
°
c c UU --- o o --- -—---,--- p. Mean --- , 0.0 Mean 0.2 0.4 0.6 0.8 1.0Average rating score: (SUBPRRKR+MYPRRKR)/2
Q
Figure 3.2 shows no clear relationship between the difference and the mean; the difference between the two sets of prospective rating scores neither increases nor decreases as the average rating score increases. In other words, the Test of Agreement does not display any consistent bias over the range of average rating scores. The mean difference between the subjects’ prospective and the derived prospective rating scores was extremely small, 0.0034, with a 95% confidence interval of -0.0206 to 0.0273. The limits of agreement were
0.153, with a 95% confidence interval of 0.1115 to 0.1945, and -0.1462, with a 95% confidence interval of -0.1877 to -0.1047. The width of these intervals is the result of two outliers. The line of best fit between the two sets of data, y = -0.21153 + 1.2546x (R2=0.809), conforms quite strongly to the line of agreement. Moreover, the mann-whitney test found that the probability that there is no difference between the medians of the two sets of prospective scores is 0.2582.
(b) Postoperative Stage
Figures 3.3 and 3.4 (page 54) carry out the same exercise for the relationship between the subjects' postoperative and the derived postoperative rating scores.
O o 5 5ns Ch >- > 12 3 g qj (V -Q 8 8 & C CO I s .& & 0.05 0.04 0.03 0.02 0.01 0.00 -0.01 -0.02 -0.03 -0.04 -0.05
Fig. 3.4: Control Group: Test of Agreement between subjects’ postoperative (SUBPORKR) and derived
postoperative (MYPORKR) rating scores
Mean + 2SD
■ii11" innffTi Mean
Mean - 2SD
--- ,--- ,--- ,--- ,--- ,
0.0 0.2 0.4 0.6 0.8 1.0
Average rating score: (SUBPORKR+MYPORKR)/2
o
Figure 3.4 displays a very strong agreement between the subjects' postoperative and the derived postoperative rating scores. The mean difference between the two set of postoperative rating scores was 0.00124, with a 95% confidence interval of -0.0009 to 0.0034. The limits of agreement were an extremely narrow 0.0138, with a 95% confidence interval of 0.0101 to 0.0175, and -0.0113, with a 95% confidence interval of -0.0150 to -0.0076. The slope and intercept of the line of best fit between the two sets of data (y = -8.2909e-2 + 1.0880x, R2= 0.946) and the mann-whitney test of the difference between their medians (significance level = 0.5089) illustrate the strength of the agreement between them.
(c) Retrospective Stage
The two diagrams on page 55 illustrate the relationship between the subjects' retrospective and the derived retrospective rating scores.
Fig. 3.5: Control Group: Relationship between subjects’
0.4 n
Fig. 3.6: Control Group: Test of Agreement between subjects’ retrospective (SUBRERKR) and derived
retrospective (MYRERKR) rating scores
0.3 i 0.2- Mean + 2SD sg 2 > s S & u CQ C □ 0.1- 0.0 -0.1- -0.2 -0.3- -0.4
-e---9 9 «co<P Mean
o
Mean - 2SD
---,--- !--- ,--- ,--- ,
0.0 0.2 0.4 0.6 0.8 1.0
Average rating score: (SUBRERKR+MYRERKR)/2
No consistent bias is obvious over the range of average retrospective rating scores. The mean difference between the subject's retrospective and the derived retrospective rating scores was -0.0014, with a 95% confidence interval of -0.0264 to 0.0236. The limits of agreement were 0.1442, with a 95% confidence interval of 0.1009 to 0.1875, and -0.147, with a 95% confidence interval of -0.1903 to -0.1037. Moreover, the equation of the line of best fit between the two sets of data (y = -0.24529 + 1.3082x, R2 = 0.868) and the mann-
whitney test of the differences between their medians (significance level = 0.8322) again illustrates the agreement between them.
(d) All Stages
Figures 3.1-3.6 illustrate the strength of the agreement between the subjects’ estimates and the derived estimates of their health-related quality of life at each stage of the process. It seems therefore that the Rosser-Kind Classification is an excellent indicator of the health-related quality of life of patients who have undergone joint replacement surgery. However, certain points of caution should be introduced at this stage. It may be the case that the Rosser-Kind Classification provides a comprehensive measure of the health- related quality of life of lower limb joint replacement patients. Alternatively, it may be that its crude generic nature may be too insensitive to pick up subtle changes in the health-related quality of life of these patients. For example, most of the patients in this study were elderly and as such, the emphasis in the Classification placed on paid employment may have been superfluous. In addition, though most patients found it relatively straightforward to identify their position on the disability and distress scales, some ambiguity did surround the position of working-age patients who felt unable to undertake any paid employment but could perform all housework except heavy tasks. In such cases, the patient would be placed in the intermediate disability category (IV).
The Rosser-Kind Classification of Illness States was used rather than a disease-specific scale, because it allows us to compare the improvements in health status resulting from joint replacement surgery with those which result from other types of medical treatment. Hence, we can use it to calculate the relative cost effectiveness of joint replacement surgery vis-a-vis other medical procedures. However, its categories may be too broad and general to
pick up small changes in the health-related quality of life of our groups of patients. It may not be surprising therefore that in the conversion process, the information in the questionnaires should conform so strongly to the patients' own positioning in the Classification. Using an alternative disease-specific scale to measure the health-related quality of life of the patients would forgo the cross-programme comparison quality of the Rosser Classification. Moreover, there is little evidence that the assumptions and value judgments underlying alternative scales are any more acceptable than those underlying the Rosser Classification. Based on the information available in our study however, it does seem that the Rosser-Kind Classification is an acceptable basis for measuring the health-related quality of life of patients who have undergone hip and knee joint replacement surgery.