CHAPTER 3

DATA AND METHODOLOGICAL CONSIDERATIONS

Introduction

This study examines reports of attending physician estimations of patient pain contrasted with patient self-reports of pain to measure the difference, if any, that may be the result of alterations in patient subjectivity. This chapter will concern itself with the data used and the methodology employed for the present study.

Computer run results were compared, where applicable, with data reported in previous research conducted with this data set. A closely-related study (Emmerson, 1982), provided a good basis for data comparison and data cleaning opportunities. 

The Field Work Site

This study employs data gathered in 1970 by Robert Gray of the University of Utah, and paid full-time professional interviewers (Gray, et al., 1982). Patients being treated at the University of Utah Medical Center Out-patient Arthritis Clinic form the case group. Controls from the community of Salt Lake City, many of whom also claim to have arthritis, were also interviewed. The case group had the additional advantage of having their attending physicians interviewed for comparative data on their condition.

The original study was concerned with factors that affect utilization of health services by rheumatoid arthritics. The data then collected was sufficiently comprehensive to allow this current study.

Four cases from the original data have been lost and deemed irretrievable. In comparing the 119 remaining cases with earlier data analysis, however, the four missing cases do not appear to be “outliers” that alter the data in any significant manner. Further reductions to 108 cases resulted from some patients or their physicians not answering the specific pain questions which were necessary to conduct comparisons for this analysis.

Procedures

The original interviews were conducted individually with each RA patient by a member of the research staff. Each patient’s attending physician clinically evaluated several aspects of his or her current condition as well as changes noted in the arthritic condition.

Subsequent coding of the data allows the patient anonymity yet allows comparisons of physician rating and patient rating of health condition.

With a series of singular and scaled questions, the original patient questionnaire required approximately 1 1/2 hours to complete. The questions covered a variety of independent variables as well as social-psychological characteristics, attitudinal measurements, individual arthritis knowledge, and behavioral aspects of patients and their social support network. Specific illness history and course were asked as well as subjective feelings about the disease.

The Rheumatoidologist Questionnaire indicated changes in patient conditions over a year, disease activity, pain as assessed by the doctors, and structural damage and deformity. Other specific questions of compliance and utilization of services were gathered simultaneously. 

Pain Measurements

Patient self-reported pain was defined as perceived pain when they received treatment for the disease. The physicians also rated the patients’ level of over-all physical pain. These measurements on pain have been used in earlier studies and the validity of their inclusion is documented in those studies (Emerson, 1982).

Since this study examined reports of the attending physician estimations on patient pain contrasted with patient self-reports of pain, the difference, if any, required a new variable be created to serve as the dependent variable of the study.

This new variable was constructed by comparing, case by case, the amount of agreement between physician and patient with regards to patient pain during treatment.

Depression Measurements

When depression is lowered and greater personal control over a chronic disease is felt, illness predictability increases (Affleck, Tennen, Pfeiffer, & Fifield, 1987). Internal locus of control is associated with lower levels of depression (Cwikel, Dielman, Kirscht, & Israel, 1988). Emotional stress, which can be often measured in depressive outcomes, has been shown to be positively related to RA disease activity (Crosby, 1987).

This study employed the Self-Rating Depression Scale (SDS Index) by William W. K. Zung, M.D., North Carolina, as a recognized diagnostic aid providing a measurement of severity of depression for clinically depressed patients and other emotional disorders. It is not a diagnostic tool in itself, but a measurement of the degree of depression present. Clinically the SDS Index is used to save time and help uncover “hidden depressions.”

The patients were given a list of 20 statements, each one reflecting one common characteristic of depressive disorder. Ten of the statements were worded symptomatically negative, ten were worded symptomatically positive. The negatives and positive were randomly mixed in the scale. The patients were asked to mark their agreement with each statement.

Measurement readings given by the completed SDS Index indicate (p — 959) normal controls: range: 25—45   with mean of  33
       transient situational  adjustment reactions: range 38-68 with mean of 53

       anxiety reaction: range 40-68 with mean of 53

       personality disorders: range 42-68 with mean of 53

      depressed (out-patient): range 50-78 with mean of 64

The definition of “depressed out-patients” given in this scale was that the scores indicated a clinical measurement of the need for out-patient psychiatric intervention. This scale purports to be valid in assessing such hidden clinical depressions.

       Besides the variables of gender, age, and illness duration, the variable of severity of patient RA illness was incorporated in the study. The operational definition of RA illness severity was observable structural damage caused by the disease as reported by the physician. 

Structural Damage 

The attending physician rated the patient on the basis of the current stage of structural damage due to the RA disease. The four choices were: early stage of damage, moderate stage of damage, severe stage of damage, and terminal stage of damage. This assessment was based upon observable damage using traditional methods, such as x—ray analysis and mobility distance test results.

Change in Rheumatoid Arthritis Condition

The attending physicians also rated the patients on the observable amount of change that had occurred in the afflicted joints in the past year. The four choices on this variable were: patient condition improved more than expected, no significant change, condition changed about as much as expected, more RA deterioration than expected f or the patient.

Other physician ratings, not used in this particular study, included patient compliance with doctors orders, utilization of services, utilization of prescribed drugs and appliances and patient attitudes towards their physicians.  

                                                 Statistical Analysis

Most of the data used in this study were nominal and ordinal level data. They were subjected to three way ANOVA. The dependant variable, which is the measure of agreement or disagreement with physician evaluations of pain, was entered into ANOVA against the variables of gender, depression, and severity of the disease. The severity measurement used in the ANOVA was the physician rating of RA structural damage.

To allow the ANOVA procedure, the Depression Scale was collapsed into three categories:

1)     normal controls: range 25—45 with mean of 33

2)     transient situational adjustment reactions, or

anxiety reactions, or personality disorders:

range 38-68 with mean of 53

3)     depressed(out-patient) :range 50-78 with mean of 64

In sum, by using multivariate (ANOVA) techniques, a testing of the hypothesis model was made in assessing the variables influencing pain evaluations by both the RA patients and attending physicians.

--------------------new page-----------------------

CHAPTER 4

FINDINGS

Patient Profiles

The present study consisted of 108 patients attending the University of Utah Arthritis Clinic as out”­patients during 1970.

As shown on Table 1, this sample was comprised of 73 females (67.6%) and 35 males (32.4%). The Arthritis Foundation (1980) statistics estimate that women suffer from Rheumatoid Arthritis two to three times more than men. This study sample was representative of the national estimation.

Of the sample, 66 patients (61.1%) were single, (having never married, or being separated, divorced or widowed), while 42 (38.9%) were married and living with spouse.

Of the sample, 96.3% were “white” respondents. The literature does not suggest any racial bias of RA disease onset or progression, other than culturally based definitions of sickness. Therefore, this sample is not cross-sectionally representative of the sample community at the date of data gathering.

Table 1

SAMPLE CHARACTERISTICS

Variables                                                               Percent                                     N

GENDER

Male

Female

MARITAL STATUS

Married, living with spouse

Single, Widowed, Divorced

PACE

White Other

RELIGION

Catholic

Protestant

L.D.S.

Other

EDUCATION

Less than 7th grade

Junior High School

Partial High School

High School Graduate

Partial College (1 yr or more)

Standard College

Graduate Degree

M = 12.056, SD = 3.769

EMPLOYMENT

Employed Full-Time

Employed Part-Time

Unemployed! Retired

INCOME (only 56 respondents replied to

            Below $10,000                                                 76.8
            $10,000 and Above                                           23.2

this question)

      Table 1 further defines the stated religion of the sample as 67.6% Latter-

day Saints (Mormon). Catholics represented 8.8% of the sample, Protestants

represented 14.8%, and the balance of the sample showed “other” as their

answer to the question. This sample was taken in Salt Lake City, Utah. At the

time of the gathering of the data this was close to being representative of the

sample community.

Educational level completed of the sample members is also shown on Table 1. The completed years of education ranged from 2 to 21 years with the mean being 12.056 years. Both the median and mode were 12.0 years. High School Graduates comprised 62.9% of the sample, with 24.1% completing college and 11.1% completing graduate degrees.

Current employment status of the sample shows the full-time workers numbered 21 (19.4%), with the part-time workers numbering 12 (11.1%). The remaining sample, 75 respondents (68.5%), reported that they were retired or not currently employed. Given the larger number of women in our study, the sample age averaging 51 years of age, and the time period of collecting the data, this seems fairly representative of the sample population.

Of those 56 patients replying to the income question, only 13 (23.2%) reported earning over $10,000 per year, leaving 43 (76.8%) who reported earning less than $10,000per year. Without the full 108 patients reporting their income, it is difficult to ascertain if this sample is representative of the general community at the time.

The health characteristics of the sample are shown in Table 2. It comprises variables of age, RA illness duration, and the depression scores found in this study.

The ages of the study sample ranged from 17 to 81 years, with the mean at 51.13 and the median at 55.0, and the mode falling at 54.0 years. The standard deviation of the age for this sample was 14.252. This sample tends to follow the expected trend, with the fifth decade of life being the most prevalent timing for RA to manifest itself.

Table 2 notes the length of duration of RA for this sample ranged from 1 year to 49 years, with the mean at 13.552 years, and both median and mode falling at 10.000. At a 95% confidence level, the Zung Depression Scale

(SDS Index) ranged from 22 to 67 with the mean of 39.981. 

Table 2

HEALTH CHARACTERISTICS OF SAMPLE

Variable                                    Range               Mean           Standard

                                                                                             Deviation

AGE, in years                          17       — 81           51.13           14.252
PA DURATION                       1       — 49           13.552         10.655
DEPRESSION SCORE           22       — 67           39.981         10.377

Table 3 gives the objective evaluations of attending physicians objective RA disease conditions. Severity, as measured by the physicians’ assessments of structural damage, was reported as 13% of the patients belonging to stage 1, or “early stage of damage,” with 46.3% in stage 2 called “moderate stage of damage,” along with 31.5% in stage 3 or “severe stage of damage.” As noted earlier, an additional 8.3% of the patients were rated in the “terminal stage of damage,” stage 4. Also shown in Table 3 is a summary of physician reports of changes in patient RA conditions over the past year.

Table 3

RHEUMATOID ARTHRITIS DISEASE SEVERITY

(ATTENDING PHYSICIAN EVALUATIONS)

STRUCTURAL DAMAGE

Early stage of damage                              13.0                                 14
Moderate stage of damage                        46.3                                 50          Severe stage of damage                            31.5                                 34
Terminal stage of damage                           8.3                                   9

M = 2.076 SD = .914

CHANGE IN CONDITION

Improved more than expected                   13.9                                     15
No significant change                               59.3                                     64
Changed about as expected                      16.7                                     18
More deterioration than expected               7.4                                       8

M = 2.120, ~ = .840

       Results show that 13.9% of patients were reported to have “improved more than expected” by the physician. “No significant change” was reported for the majority at 59.3%. The physicians reported only 16.7% of the patients had the RA disease progress “about as expected” with 7.4% having “more deterioration than expected.” 

Measures of Association

Having examined the frequencies of important variables, measures of association (chi-square) techniques were made. The gamma statistic was used as several of the measures did not have the statistically proper number of cells with frequencies of five or higher in them.

Table 4 represents the associations of the more significant variables of pain evaluations. First are findings of the entire sample, followed by the subset of findings based on gender. Table 4 shows that physician rating of observed RA structural damage and of disease activity had a substantial gamma of ±.91538. This suggests that physicians are likely to rank the patient’s disease as an objective measurement of the observed diagnosis in measuring actual structural damage.

In response to the subjective question of pain felt by the patient, based upon structural damage, the attending physicians recorded a gamma of .73382. Questioned about the pain the patient felt, based upon the change in condition noted over the year.

Table 4

EVALUATION ASSOCIATIONS

Variables                                                                                  Gamma

PHYSICIAN

            Disease activity by damage                                            .91538
            Pain by structure damage                                               .73382
            Pain by change in condition                                            .70510
            Damage by change in condition                                      .62479

PATIENT

            Depression by gender                                                     .17225
            Pain by structure damage                                               .30275
            Pain by change in condition                                             .14748

                                                                            female             male

PHYSICIAN

Pain by structure damage                                     .39241             .87013
Pain by change in condition                                   .21127             .89087

PATIENT

Pain by structure damage                                     .21708             .37796
Pain by change in condition                                  ~. 02041              .22629

      The physicians show a gamma of .70510. Both of these measures are more

substantial than the patient pain evaluations of .30275 and .14748, respectively.

They are further broken down in the gender subset, and show a marked gender

difference.

These substantial disparities between physicians and patients (including gender differences among the patients) led to the creation of a new variable in this study which addressed this disagreement problem. This problem will be treated as found in analysis that provided Table 5.

It is important to note other associations found in Table 4, including gender-specific disparities among the patients’ evaluations. The first important measurement shows a gender difference in depression scores as measured by the gamma of .17225.

After breaking down the pain associations by gender, more disparities appeared. In the patients’ subjective evaluations of pain with respect to structure damage and change in RA condition, the male patients recorded lesser degrees of gamma significance than did the female patients. This differential association created the need for ANOVA which was performed.

As shown in Table 5, patient answers to the question of the amount of pain in the treatment of arthritis was bimodal, with 28.7% answering that they had “very little pain,” 13.9% answering “little pain,” 30.6% answering “some pain,” 13.9% answering “much pain,” and 13% reporting “very much pain.”

Furthermore, in Table 5, physician evaluations of patient pain were reported as: 6.5% “very mild pain,” 36.1% “mild pain,” 53.7% “average pain,” 3.7% “severe pain.” No patients were reported in “very severe pain” despite nine patients being in the “terminal stage.”

Table 5

PAIN EVALUATIONS

Variable   

                                     Prercent            Number

PATI ENT

Very little pain                  28.7                     31

Little pain                         13.9                     15
Some pain                        30.6                     33
Much pain                        13.9                     15
Very much pain                13.0                     14

M = 2.685, SD = 1.365

PHYSICIAN

Very mild pain                    6.5                       7
Mild pain                          36.1                      39
Average pain                    53.7                      58
Severe pain                        3.7                        4
Very severe pain                0.0                        0

M = 2.546, .~fl = .675

COMB INED

4 steps LESS pain than Dr  2.8                        3
3 steps LESS pain than Dr  7.4                        8
2 steps LESS pain than Dr 11.1                      12
1 step LESS pain than Dr   9.4                       21
Same pain evaluation as Dr 20.4                    22
1 step MORE pain than Dr 17.6                     19
2 steps MORE pain than Dr 20.4                    22
3 steps MORE pain than Dr  0.9                       1
4 steps MORE pain than Dr  0.0                       0

H = 4.861 SD = 1.683

In Table 5, the evaluations of the patient and physician were, case by case, combined to form the dependant variable in this study. This resulted in 2.8% of patients reporting “4 steps less pain” than the physicians ratings, 7.4% at “3 steps less pain,” 11.1% claiming “2 steps less pain,” and 19.4% reporting only “1 step less pain.” Of this sample, 20.4% of the patients agreed with physician pain evaluations.

Of those reporting more pain than the physician evaluated, 17.6% claimed “1 step more pain”, 20.4% reported “2 steps more pain,” while 0.9% said they had “3 steps more pain” than the physician evaluations. There were no patients reporting “4 steps more pain.”

This comprised the first step in breaking down the disparity of pain evaluations between the patients and their attending physicians. This new variable, evaluation groups, was then subjected to the Zung depression scores (SDS Index) as well as variables of gender, age, illness duration, structural damage and changes in PA condition to create Table 6.

Relationship s Between Evaluation Group s As Table 6 indicates, SDS Index measurements are consistent with the hypothesis that higher depression scores result in differences in patient pain evaluation, using the physicians’ evaluation as a baseline. Both

Table 6

MEANS OF SELECTED VARIABLES CONCERNING

PAIN PERCEPTION DIFFERENTIAL BETWEEN

PHYSICIAN AND PATIENT EVALUATIONS

Patient:                    Feels LESS Pain                 Same Pain                        Feels MORE Pain

                                 = 44                  N* = 22                           N* = 42

Gender                  Female 70.5% Female 68.2% Female 64.3%

                             Male 29.5%                  Male 31.8%              Male 35.7%

Age                      54.886 years 51.136 years 47.190 years (Range 17-81)

RA Duration 14.595 years 13.952 years 12.310 years (Range 1-49)

Structural               2.682                         2.227                            2.024

RA Damage

(Range 1-4)

Change in RA        2.318                         2.045                            1.952

(Range 1-4)

Depression           42.250                       36.045                         39.667

(Range 22—67)

*Total N = 108

patient groups that did not agree with the physicians’ evaluations exhibited higher depression scores than did the group reporting agreement with the physicians’ evaluations.

Severity measurements varied over the three patient classifications and are linear. These findings are now discussed under the three evaluation groups.

Patients Reporting LESS Pain than

Physician Reports

Patients who reported LESS pain than corresponding physician evaluations, in comparison to the two other classifications, were shown to be older and to have had the illness longer. This classification included a higher percentage of female patients than the other two.

The physicians also reported more structural damage in this classification, with more changes in physical condition over the past year. Depression scores were higher for this group than for the other two groups.

In summary, the group reporting LESS pain than that perceived by their doctors, were more likely to be older patients who had suffered longer and who had more damage to affected joints. They were also more likely to have experienced recent changes in their condition, to be female, and to have a higher incidence of depression.

Patient Evaluations the Same as

Physician Evaluations

Findings of this study show that in comparison to the two other classifications, patients who matched physician­reported levels of pain were between the other two classifications in age, illness duration, gender, depression levels, structural damage and recent changes in their RA condition. They were found to be closer to the “feels LESS pain than doctor reports” group.

Patients Reporting MORE Pain than

Physician Reports

Patients who reported MORE pain than physician evaluations, in comparison to the two other classifications, were younger patients and had a short history of RA. This classification included a higher percentage of male patients than the other two.

The physicians also reported less structural damage in this classification with fewer changes in physical condition a year. Depression was also rated at a higher level than that of the “agreement” group.

In summary, the group reporting MORE pain than their doctors perceive are more likely to be younger patients who have suffered the illness for a shorter time, have had less damage to the affected joints, and fewer recent changes. They have an elevated level of depression and are principally male. 

Intercorrelations of Evaluation Groups

Under the earlier evaluation associations (Table 4), significant disparities necessitated the creation of a new variable termed “evaluation groups.” This was needed due to physician evaluations and the patient evaluations being substantially different. Gender differentials were also strongly noted.

Table 7 documents the intercorrelations of important variables with respect to the three new evaluation groups 

Table 7

INTERCORRELATI[ONS OF VARIABLES

BETWEEN EVALUATION GROUPS

Patient evaluates pain as LESS than physician

Age                      .1650             .1857             —.0196           —.1515
Illness Duration         .1724             .5502               .1659
Depression                  .1566            .0263
Structure Damage   .3752

Change in Condition

Patient evaluates pain as the SAME as physician

Age                      .2661             .0027                .2425                —.1256
Illness Duration       .3867            .2529             —.0765
Depress ion                  .3596            .1629
Structure Damage     .4729

Change in Condition 

Patient evaluates pain as MORE than physician

Age                                       .5939             .2048                .1630                   .1262
Illness Duration   —                                                                     .0331              .0603                .1626
Depression                                                                      —                 .1309                .1516
Structure Damage                           .4242
Change in Condition and it supports earlier statements of the relationship between the evaluation groups.

Correlations of illness duration with age as well as with structure damage were found to be linear in this table, and, with respect to age, the correlations range from .1650 for LESS pain group to .2661 for SAME pain group and finally to .5939 for MORE pain group.

In structure damage, the correlations with illness duration range from .5502 in the LESS pain group to .2529 in the SAME as physician evaluation group to .0603 in the MORE pain than the physician evaluation group. This suggests that more structure damage occurs over time.

Correlations of depression with structure damage and illness duration are curvilinear. This is noted when comparing the evaluations groups with respect to structure damage. For the evaluation group that reported LESS pain than did the physician, the correlation was .1566; for the group agreeing with the physician evaluations, it was .3596; and for the group feeling MORE pain than the physician thought, the correlation was -.1309. This again shows that the lower depression scores of patients are correlated to more agreement with the physicians on pain evaluations.

Illness duration and depression also appear to follow a similar curvilinear pattern with .1724, .3867 and .0331, respectively. This shows that depression is not linearly correlated to illness duration, suggesting another variable besides time interplays with this correlation.

Table 7 also shows an important correlation between change in condition and structure damage. Although the three groups did not vary widely (.3752, .4729, and .4242, respectively), it is notable that the group which had the MOST deterioration was the middle group who agreed with the physicians on the pain experienced. This suggests that, even though they had more illness progression, the lower depression scores of this group may have helped in keeping a more objective outlook on their condition, as based upon more agreement with physician evaluations.

To further test the statistical significance of the findings, ANOVA was performed. Table 8 documents that the dependant variable, which is the measure of agreement or disagreement with the physician evaluation of pain, was entered into ANOVA against the variables of gender, depression, and severity of the disease, as measured by the physician ratings of structural damage.

To allow the ANOVA procedure, the Depression Scale, as used in Table 6, was collapsed into three categories:

1)     normal controls: range 25-45 with mean of 33 

2)    transient situational adjustment reactions, or anxiety reactions, or personality disorders: range 38-68 with mean of 53

3)    depressed: range 50-78 with mean of 64.

Table 8

THREE-WAY ANOVA: EVALUATION GROUPS

BY GENDER, DEPRESSION, AND

RA STRUCTURE DAMAGE

MAIN EFFECTS

Gender                                1.606               1           1.606           .741
Depression                          8.017               2           4.009         1.851
Structure Damage              56.818               3         18.718         1.217*

TWO-WAY INTERACTION

Gender Depression                .568               2             .284           .131
Gender Struct Damage       19.330               3           6.443         2.975**
Depress Struct Damage     15.818               6           2.636         1.217

THREE-WAY INTERACTION

Gender Depres StrDamge 12.017                4          3.004             1.387

RESIDUAL                    184.083             85          2.166
TOTAL                          298.299            106          2.814
*p < .000                   **p < .036

The three evaluation groups used in the ANOVA resulted from collapsing the original nine pain evaluation variables. Gender and structural damage variables were already in the proper form to apply ANOVA.

Two important effects were established in the ANOVA procedure. One is a main effect, and the other is a two-way interactive effect:

1) The main effect of structural damage on agreement/disagreement of pain with the physician was measured at 1.217 (p = .000).

2) An important interactive effect between gender and structural damage was measured at 2.975 (p = .036).

Gender and depression were not significant main effects to influence the amount of agreement with physician pain evaluations. However, structure damage was a significant factor.

In the two-way interaction, gender and structural damage exerted a significant impact on patient agreement with physician pain evaluations; this analysis is consistent with the earlier findings. No other interactions were found to be significant, even though depression was documented in earlier findings.