Health sciences: overview and applications

¶ … meticulous construction of the data analysis, statistical tabulation, and interpretation is provided in the following pages.

SPSS was used to manage and calculate the researcher-designed data.

Researcher-Designed Questions

Questions numbered 1-11 were administered as a part of the SF-36 mental inventory. As stated earlier, these questions provide a standard assessment means to assess clinical outcome and mental health. However, they are inadequate to assist in assessing quality of life in patients undergoing on-pump and off-pump by-pass surgery.

Questions 12-30 were research developed and designed to give a more accurate assessment of patient satisfaction with the two procedures. These questions were divided into categories which are representative of factors commonly associated with quality of life. The categories are Mental health, Cognitive ability, Social support, General health, and Bodily pain. Demographic data is contained in questions 21-26 and will be used to identify confounding factors. For purposes of data analysis, the questions will be analyzed according to category and results compared for the two process controls (type of surgery) using descriptive analysis. Confounding factors will be identified and reliability tests conducted using Cronbach's alpha reliability.

Mental Health

Mental health is perhaps the most defining concept of this study. The following tables and charts represent the descriptive statistics of mental health question twelve.

Table 2: Descriptives of Mental Health Question 12 igure 9: Chart of Mental Health Question

As can be seen from the data presented above, the mean for all three researcher-designed mental health questions was slightly higher in the off-pump patients than it was in the on-pump patients, but with a level of significance set at =0.05 the two groups are virtually identical.

This statistical analysis for individual question groups holds true, however these results must be compared across the sample population to determine if results will vary when various demographic data is compared. For instance, women reported four times more than men -- 10.8% of men and 41.1% of women -- that they had, in the past year, "experienced four or more weeks during which they felt sad, blue or depressed or when they lost all interest or pleasure in things that they enjoyed and cared about," as can be seen in the cross-tabulation below. Women who received on-pump CABG were slightly more depressed (5%) than women who received off-pump CABG. A previous study confirmed that women are more likely than men are to experience postoperative depression after CABG surgery attributable to their poor health (Ai, 1997). Several factors were not considered in this evaluation. It is not know if depression is a result of the type of CABG surgery or other factors.

Figure 10: Cross-Tabulation with Gender, Depression, and Surgery Type

Interestingly enough, however, one of the subjects of this study who was reported dead to the researcher by his daughter was a man who committed suicide. He experienced first mild and then severe depression after undergoing on-pump CABG and was admitted to a psychiatric unit for treatment. He spent two months there. The day he was released, he took a kitchen knife and pierced through his chest to his heart. There had been no previous history of depression or any kind of mental illness. He had been a socially and physically active physician prior to his CABG surgery. This case is isolated and it cannot be determined ex-post facto whether or not the suicide was a result of a decrease in quality of life after CABG surgery, or if there may have been other factors unknown to others before surgery that may have contributed to his actions.

Condition-Specific Questions

Questions 13 and 14 refer to cognitive ability and memory. Questions 15 and 16 refer to bodily pain. Question 17 refers to Health transition. The questions regarding surgical pain, current pain and the frequency of arrythmia are from the SF-36 portion of the survey. The tables and charts below pertain to the five condition specific questions, Cognitive Ability, Memory, Surgical Pain, Current Pain, and frequency of Arrhythmia. The scores range from worse to better for the Cognitive Ability and Memory, from better to worse for the two Pain questions, and from very frequently to never for the Arrhythmia question.

Table 3: Independent Samples T-Test Condition Specifics & CAGB

As can be seen from Table 3 above and Figure 11 and Table 4 below, the off-pump patients faired slightly better in every category -- except pain which was evenly split as would be expected because the incisions are identical -- but again, with a level of significance set at =0.05 the two groups are virtually indistinguishable and what little differences are noted, are not enough to be clinically significant.

Figure 11: Chart of Independent Sample T-Test Condition-Specific Questions

Table 4: Descriptive Frequencies of Condition-Specific Questions

The analysis of variance of the condition specific pain further supports that the two groups are virtually indistinguishable. As has been explained above, there is no difference in the incision to the sternum between the on-pump CABG procedure and the off-pump CABG procedure. The alpha reliability coefficients for the two researcher-developed pain questions are.987 and.937.

Table 5: Analysis of Variances for Post-Operative Pain and Current Pain

Figure 12: Post-Surgical Pain and Current Surgical Pain

Social Support Questions

Questions 18, 19, and 20 related to the level of social support that the patients received. The social support data showed that all respondents had either family members or pets that lived with them. Using Cronbach's alpha reliability, coefficients were calculated for the Mental Health Component Score and the Family-Pet support system. The reliability score was 0.977.

The figures on the next page show the scaled frequency in the family-pet support questions. One-hundred percent of the respondents had family members and/or pets living with them. Not a single respondent lived alone. Since all respondents had a social support system in place, it is doubtful that lack of a social support system contributed to any depression found, but a multivariate analysis is run in a later section to see if this is or is not the case. The question regarding social support, in this case, serves to rule out a confounding factor to the research question. This data serves to demonstrate that social support was not a factor in the overall mental health of the patients.

One patient actually wrote on her questionnaire that her family and friends would not let her get depressed, but she did admit that she had been depressed for a short period after surgery and felt that her sex life to this day, had been damaged due to her surgery. CABG surgery is particularly hard on women if the Saphenous vein is taken instead of one of the arteries. In addition to having a scar down the middle of their chest, taking the vein requires an incision down the length of the leg and results in a scar on the leg that is about two feet long. The patient had only been married a couple of years when the surgery was preformed. Depression in women may have more to do with self-image than declining health as the studies show (Ai, 1997). However, this data was not a part of this study and would be interesting material for further study.

The additional retrospective Likert scale question, which asked, "The way I would have answered the questions in this survey before my heart operation would be," was virtually split. The on-pump patients had a mean of 2.48 and the off-pump patients had a mean of 2.53, once again slightly better, but not clinically significant with an alpha coefficient reliability score of 0.794. Level of significance was set a =0.05.

Figure 13: Chart of Family Members in the Home

Figure 14: Cart of Pets in the Home

Figure15: General Linear Multivariate Chart of all Mental Health, Condition Specific, and Social Support Researcher-Designed Questions

Summary of Descriptive Statistics for Individual Traits and Characteristics.

Mental Health: On Question 12a, the mean for on-pump was 1.83 and for off pump it was 1.97, on Question 12b, the mean for on-pump was 1.80 and for off pump it was 1.86, and on Question 12c, the mean for on-pump was 1.79 and for off pump it was 1.89. Since these three questions are dichotomous, yes (1) or no (2) questions, the higher the mean the stronger the score.

Condition Specific: For Question 13 the mean for on-pump was 2.83 and for off pump it was 3.06. For Question 14 the mean for on-pump was 2.74 and for off pump it was 2.97. For Question 17 the mean for on-pump was 4.23 and for off pump it was 4.33. Since these three questions were Likert scale questions that moved from worse to better the higher the mean the stronger the score. For Question 15 the mean for on-pump was 2.56 and for off pump it was 2.67. For Question 16 the mean for on-pump was 2.32 and for off pump it was 2.15. Since these two questions were Likert Scale questions that moved from better to worse the lower the mean the stronger the score.

Social Support: For Question 18 the mean for on-pump was 2.20 and for off pump it was 1.97. For Question 19 the mean for on-pump was 1.84 and for off pump it was 2.02. These two questions were Likert scale questions asking about family members and pets in the home. Since these three questions moved from less to more, the higher the mean the stronger the score.

Thus as can be seen from Figure 15 above, with a level of significance set at =0.05 the two groups are virtually identical in all categories. Descriptive statistics on quality of life questions showed slight differences in some cases. However, none of these results were clinically significant. Virtually no differences were observed regarding quality of life for the two surgical procedures.

Demographics

Descriptive statistics indicate that no significant differences exist between the two CABG surgery patients. These results must be compared to various demographic groups in order to validate sampling procedures and determine if the minor differences that were reported may be due to demographic features, rather than surgical procedures.

The following summarizes the demographic findings. Less than 1% of patients reported that they smoke; 17.2% reported that they do not exercise at all; 57.6% reported that they were overweight; 61.4% reported that their income was below fifty thousand dollars annually, with 31.2% reporting an income below twenty-five thousand; education levels ranged from 8.8% completing graduate school to 4.2% completing grade school or below, the highest level reported were high school graduates at 31.2%; 7% reported that they followed their cardiac diet every day of every month and 47.4% reported that they followed it most days of every month, but 17.7% reported that they never followed their prescribed diet.

The most noteworthy demographic statistics came from the ethnicity question. Patients reported that their ethnicity was 3.7% African-American, 1.4% Asian, 70.7% Caucasian, 19.1% Native American, 0% Hispanic, 3.7% Middle Eastern or Arab-American, and 0.5% other. The participating physicians estimated that Hispanic surnames accounted for about 25% of the entire patient group and yet, not a single person claimed to be Hispanic or Mexican-American. A surname does not necessarily indicate a persons ethnicity, but in light of the fact that the bulk of the patients (70%) came from Louisiana and Texas and these states have large Hispanic population, perhaps this was an error in the research design and questionnaires should have been sent in English and Spanish, and perhaps, the return rate would have been higher from the Latin community, thus giving the study a better basis for comparisons. Cultural differences may contribute a person's ability to cope with stress and recover from an illness.

The following pages show the actual percentages of all demographics, which are important when looking at the multivariate data analysis for all dependent variables, co-variables, and fixed variables.

Figure 16: Chart of Gender Distribution

Figure 17: Chart of Age Distribution

Figure 18: Chart of Cardiac Diet Frequencies

Figure 19: Chart of Weight Frequencies

Figure 20: Chart of Exercise Frequencies

Figure 21: Chart of Income Frequencies

Figure 22: Chart of Education Frequencies

Figure 23: General Linear Model of All Demographics

Multivariate Data Analysis

Demographics can effect the outcome of the research question. If demographics are found to effect the outcome of the research question significantly, this could indicate sampling error and serve to cause sampling bias in the test results. Severe sample bias can make the results of the research inapplicable for the general population. To determine if demographics have affected the results multivariate data analysis is used. This technique examines the pattern of relationships between several variables simultaneously. Multivariate statistics helps the researcher to summarize data and reduce the number of variables necessary to describe it. Most commonly multivariate statistics are employed:

For developing taxonomies or systems of classification;

To investigate useful ways to conceptualize or group items;

To generate hypotheses; and finally,

To test hypotheses.

There are certain assumptions about multivariate analyses. The first assumption is that all of the models require that input data be in the form of interrelationships -- this means correlations for factor analysis. Multidimensional scaling and cluster analysis can use a variety of different input data -- distances, or measures of similarity or proximity. This means that multidimensional scaling and cluster analysis can be somewhat more flexible than factor analysis. The second assumption of these methods is that the data itself is valid. Because these methods do not use the same logic of statistical inference that dependence methods do, there are no robust measures that can overcome problems in the data. These methods, therefore, are only as good as the input one has.

In each case of multivariate data analysis, the output will look somewhat different, but in all of the techniques, the researcher is required to look at the results and make some determination of how many factors, dimensions or clusters to use in further analysis in order to represent the data. What the researcher should not forget is that each case or variable used in the analysis is simultaneously classified on all the dimensions. While this is most apparent in multidimensional scaling, it applies equally well to the other techniques.

The first task was to determine if the researcher-designed mental health questions were affected by any of the demographics, followed by the condition-specific questions and finally the social support questions. The six tables below are multivariate analysis and analysis between subjects. The tests between-subject effects is an analysis of variance table. The column labeled Source lists the effects in the model. The second column displays the sum of squares for each effect. The degrees of freedom for each sum of squares is displayed in the column labeled df. The mean square of each effect is calculated by dividing the sum of squares by its degrees of freedom. The F. statistic and its significance value are displayed in the next columns. The F. statistic is calculated by dividing the mean square by the mean square error. Effects with a small significance value (smaller than 0.05) are significant. The multivariate tests table displays four multivariate tests of significance of each effect in the model. Pillai's trace is the first multivariate test listed. Wilks' lambda is sometimes called the U. statistic. Lambda ranges between 0 and 1, with values close to 0 indicating the group means are different and values close to 1 indicating the group means are not different (equal to 1 indicates all means are the same). Hotelling's trace is based on the sum of eigenvalues. Roy's largest root is the largest eigenvalue. Of the four test statistics, Wilks' lambda is convenient and related to the likelihood-ratio criterion. For some practical situations, however, Pillai's trace may be the most robust and powerful criterion among the others. Choice of these multivariate statistics depends on the situation. The value of the test statistic is displayed followed by the F. statistic, which is a transformed value of the corresponding test statistic and has an approximate F. distribution. The hypothesis and error degrees of freedom of the F. distribution are shown. When the significance level is relatively small (less than 0.05) for the effect being tested, then it can be concluded that the effect is significant.

On the between-subjects tests, with the level of significance set at =0.05 Question 12b has a significant effect of 0.016 with gender and Question 12c has a significant effect of 0.037 with education. Question 12b asked, "Have you had two (2) years or more in your life when you felt depressed or sad most days, even if you felt okay sometimes?" Question 12c asked, "Have you felt depressed or sad much of the time in the past year?" The multivariate tests for the mental health questions had no significant effects.

On the multivariate test, with the level of significance set at =0.05 the researcher-developed condition-specific questions had several significant effects. All four tests -- Pillai's trace, Wilks' lambda, Hotelling's trace, and Roy's Largest Root -- had effects between education (0.025), ethnicity (0.006), exercise (0.050), the surgeon (0.003) and the type of surgery (0.019). On the between-subjects tests, ethnicity had significant effects between: Question 13 (0.024), which asked, "My cognitive (thinking) ability since I had my heart operation is?"; Question 14 (0.026) which asked, "My memory since I had my heart operation is?"; Question 15 (0.005) which asked, "The discomfort (pain) that I had during the first four weeks following my heart operation was?"; and finally, Question 17 (0.041), which asked, "Since my cardiac surgery, I have had cardiac arrhythmia?" There was also a significant effect between exercise and Questions 14 (0.026). The effect between education and Question 15(0.041), and Question 16 (0.004), which asked, "The discomfort (pain) I have now is?" There was also a significant effect between Question 16 (0.004) and weight and Question 15 (0.018) and gender. Questions 13 (0.001), 14 (0.025), 16 (0.002), and 17 (0.008) had significant effects between the surgeons, while Questions 13 (0.001), 14 (0.014), and 17 (0.008) had significant effects between the type of surgery.

On the multivariate test, with the level of significance set at =0.05 the researcher-developed social support questions had three significant effects. All four tests -- Pillai's trace, Wilks' lambda, Hotelling's trace, and Roy's Largest Root -- had effects on education (0.009), age (0.018), and exercise (0.000). On between-subject tests, with the level of significance set at =0.05 Question 18, which asked "Please choose the best answer for the number of family members who live with you," had significant effects between education (0.010), age (0.040), exercise (0.002), and weight (0.041). Question 19, which asked, "Please choose the best answer for the number of pets who live with you," had significant effects between education (0.016), age (0.012), and exercise (0.000).

Table 6: Analysis of Between-Subject Effects Mental Health Questions

Table 7: Multivariate Tests of Mental Health Questions

Table 8: Analysis of Between-Subject Effects Condition-Specific Questions

Table 9: Multivariate Tests of Condition-Specific Questions

Table 10: Analysis of Between-Subject Effects Social Support Questions

Table 11: Multivariate Tests of Social Support Questions

Data Analysis: The SF-36

The SF-36 Health Survey was initiated to achieve minimum standards of precision necessary for group comparisons in eight conceptual areas. It was also constructed to yield a profile of scores that would be useful in understanding population differences in physical and mental health status, the burden of chronic disease and other medical conditions, and the effects of treatments, such as CABG surgery, on general health status (Ware, 2000).

The SF-36 orders profiles from left to right, from the best physical health measure -- Physical Functioning -- to the best mental health measure -- Mental Health. ifferences on the left side of profiles reflect physical health status and differences on the right side reflect mental health status.

Five of the eight scales, Physical Functioning, Role-Physical, Bodily Pain, Social Functioning, and Role-Emotional define health status as the absence of limitation or disability (Ware, 2000). For these scales, the highest possible score of 100 is achieved when no limitations or disabilities are observed (Ware, 2000).

Three of the eight scales, General Health, Vitality, and Mental Health, are bipolar in nature and measure a much wider range of negative and positive health states. For these scales, a score in the mid-range is earned when respondents report no limitations or disability (Ware, 2000). A score of 100 on these bipolar scales is only earned when respondents report positive states and evaluate their health favorably on the survey (Ware, 2000).

SF-36 Sub-Scale Scores

The normative data collected here from the SF-36 makes it possible for the researcher to interpret the scale score for an individual respondent or as an average score for a group of respondents by comparison with scores from other individuals or other groups. Norm-based comparisons require valid norms for a well-defined and representative population of interest. Accepted guidelines for the standardization, scoring, and documentation of widely used and psychological tests emphasize the publication of norms prior to their widespread use (APA, 1985). All norms have been published for the SF-36 (Ware, 2000).

Multivariate Data Analysis

The following tables describe the multivariate data analysis for the SF-36. On the between-subjects tests, with the level of significance set at =0.05 the MCS Scores has a significant effect of 0.032 with education. The PCS had no significant effects. On the multivariate test, with the level of significance set at =0.05 the MCS and PCS had one effect. All four tests -- Pillai's trace, Wilks' lambda, Hotelling's trace, and Roy's Largest Root -- had effects between education at 0.019.

Because the MCS and PCS account for 84% of the eight subscales in the SF-36, it is logical to conclude that similar findings will be found on the multivariate data analysis for the eight sub-scales as for the PCS and MCS. On the multivariate test, with the level of significance set at =0.05 the eight concepts also had one effect. All four tests -- Pillai's trace, Wilks' lambda, Hotelling's trace, and Roy's Largest Root -- had effects between education at 0.040.

Table 13: Analysis of Between-Subject Effects MCS and PCS

Table 14: Multivariate Tests of MCS and PCS

Table 15: Multivariate Tests of SF-36 Concepts

Figure 33: Ribbon Graph of Tests of Between-Subjects Effects

The table for the analysis of Between-Subject Effects SF-36 Concepts was too large to present here. Figure 33 above, however, shows that there are two significant effects. On the between-subjects tests, with the level of significance set at =0.05 the Sub-Scale Scores for MH has a significant effect of 0.010 with education and the BP has a significant effect of 0.026 with ethnicity.

These multivariate statistics show that there are factors that affect the measurement of the SF-36 that may not be linked to the surgical procedure as much as other factors, in this instance, education and mental health and ethnicity and bodily pain.

Descriptive Statistics SF-36

The last results of this analysis are presented in ANOVA and descriptive tables. In one-way ANOVA, the total variation is partitioned into two components. Between Groups represents variation of the group means around the overall mean. Within Groups represents variation of the individual scores around their respective group means. The between groups variation is partitioned into linear trends. The significance level for the linear trend is higher than 0.05 in all groups but one.

If the groups do not have equal sample sizes, as is the case here, the trends are computed as both weighted and unweighted. Weighted takes the varying sample sizes into account and is the recommended approach for an unbalanced design. Sig indicates the significance level of the F-test. Small significance values (

Table 16: ANOVA Table of SF-36

Figure 34: ANOVA SF-36 Ribbon Graph

Finally, in Table 17 below, the descriptive statistics for the entire database of the SF-36 respondents are presented. As with all of the data that has been assessable, the two groups are virtually identical.

Table 17: Descriptive Statistics of the SF-36 Database

Data Reliability

Using Cronbach's alpha reliability, coefficients were calculated for each of the eight subscales of the SF-36. Depicted below are a table of the reliability coefficients of the survey and charts of the results, in percentages, of the SF-36 eight scales plus the physical component scores and the mental component scores.

Alphas for the SF-36 ranged from 0.81 to 0.93, which is in line with reported values on national reliability coefficients using Cronbach's alpha that ranged from 0.78 to 0.93 across the eight subscales (McHorney, Ware, Lu & Sherborune, 1994; Ware, 2000).

SF-36

Subscale Reliability Coefficients Alpha Standardized Item Alpha

Physical Functioning (PF) 0.9374 0.9386

Role-Performance (RP) 0.8277 0.8279

Bodily Pain (BP) 0.8193 0.8196

General Health (GH) 0.8478 0.8479

Vitality (VT) 0.8471 0.8472

Social Functioning (SF) 0.8150 0.8152

Role-Emotional (RE) 0.8268 0.8330

Mental Health (MH) 0.8399 0.8478

Table 12: Reliability Coefficients of SF-36

Reliability examines the consistency of results from different measures designed to evaluate the same variable. A reliability of 0.50 indicates that 50% of the measured variance is reliable. Reliability coefficients are thus merely proportions. As can be seen, the reliability in this study is between 81-93%. The data obtained on the SF-36 is therefore very reliable.