Evaluating the effects of a total joint education class on patient outcomes

Abstract

As the population ages and as people live longer, there is a growing demand for total hip and total knee procedures. Possible outcomes for these procedures is a postoperative joint infection (PJI) that can cause long postoperative lengths of stay (LOS) in the hospital. The PJIs can also negatively impact the quality of life for the patient. Using the roadmap of the continuous quality improvement theory, the purpose of this quantitative study was to examine the relationship between the independent variables (joint education class participation, body mass index [BMI], A1c, and smoking) and dependent variables (PJI and LOS). To evaluate the relationship with PJI, a logistical regression analyzed the sample population of 1216 patients and indicated a relationship between joint education class attendance and PJI among total hip patients, but not total knee patients when controlling for the other variables. The regression also indicated a significant relationship between BMI and smoking and PJIs, but it did not show a relationship between A1c/diabetes and PJI. To evaluate the relationship between joint class education and LOS a Poisson regression indicated that those who did not attended the joint education class, whether they had total hips or total knees, had a longer postoperative LOS. The implications for positive social change involve providing information to physicians and administrators regarding the effectiveness of the total joint education class in improving outcomes. This information could be used to justify the need for patient compliance with the class and/or the possible need for additional resources to support the total joint education program.

1. Objective

The purpose this study was to determine the effect of a joint education class on reducing (PJI) rates and (LOS) amongst patients who received total hip or total knee arthroplasty procedures at a hospital system in South Carolina. The dependent variables were PJI rates and LOS, and the independent variables were components related to the joint education class (A1c, BMI, and smoking). The results may provide administration, orthopedic surgeons and other staff with a better understanding of the influence the independent variables have on the outcomes of total hip and total knee arthroplasty cases.

2. Methods

The data collected included all total knee and total hip patients between January 1, 2017 and October 31, 2018, at Palmetto Health. Follow-up data on the patients was provided through April 30, 2019, to document if a PJI occurred. The sample that I used for this study consisted of 1216 patients. Of these patients, there were 752 females and 464 males. The procedural mix included 406 total hip procedures and 810 total knee joint procedures. Of these patients, 535 did not attend the preoperative total joint class and 681 did attend. Also, of the 1216 patients who had a total joint procedure, there were 41 who had a PJI and 1175 who did not have a joint infection within the timeframe of the study. With regard to smoking status, the sample consisted of 31 current smokers and 1185 noncurrent smokers. The sample had 84 recent (within the year) former smokers and 1101 who stated that they have never smoked. The A1c patients were grouped into three categories. These categories were “diabetes not well controlled,” “diabetes with reasonable control,” and “not diabetic.” The distribution in these categories were as follows. There were 46 patients in the “diabetes not well controlled.” The patients in this category had a recorded A1c of 7 or higher. The next category, “diabetes with reasonable control,” included 397 patients and were classified in this category because they had a recorded A1c and the A1c value was less than 7. Finally, the last group was “not diabetic” and included 773 patients who did not have an A1c recorded due to them being identified by their physician as not needing an A1c due to not being diabetic or having a medical history or indication of diabetes. Finally, the last variable was BMI. Patients were categorized into four categories for BMI. The first category was “underweight,” which consisted of 3 patients. The “underweight” category included any patient with a BMI less than 18.5. The next category was “normal” and consisted of 141 patients. The “normal” category included patients with a BMI greater than or equal to 18.5 and less than 25. The “overweight” category included 311 patients, which included any patient with a BMI greater than or equal to 25 and less than 30. Finally, the fourth category was obese and included 761 patients and this category was defined as anyone with a BMI greater than 30. However, testing for linearity in BMI resulted in treating this variable as continuous for the logistical regression and merging the groups as they proved to be linear.

I used four research questions to guide this study. The first is: Among total hip arthroplasty patients, is there a relationship with attendance of a joint education class; independent variables (smoking, BMI, A1c levels); on reducing postoperative prosthetic joint infection (PJI) rates (dependent variable) within a health care facility in South Carolina? Therefore, this question focuses solely on total hip arthroplasty patients and the relationship between the independent variables and one of the dependent variables (PJI).

The hypothesis associated with this research questions is as follows:

Ho: There is no relationship with attendance of a joint education class; independent variables (smoking, BMI, A1c levels); on reducing postoperative prosthetic joint infection (PJI) rates among total hip arthroplasty patients within a health care facility in South Carolina.

H1: There is a relationship with attendance of a joint education class; independent variables (smoking, BMI, A1c levels); on reducing postoperative prosthetic joint infection (PJI) rates among total hip arthroplasty patients within a health care facility in South Carolina.

The next research question focuses on the total knee arthroplasty patients and is as is as follows: Among total knee arthroplasty patients, is there a relationship with attendance of a joint education class; independent variables (smoking, BMI, A1c levels); on reducing postoperative prosthetic joint infection (PJI) rates (dependent variable) within a health care facility in South Carolina?

The hypothesis associated with this research questions is as follows:

Ho: There is no relationship with attendance of a joint education class; independent variables (smoking, BMI, A1c levels); on reducing postoperative prosthetic joint infection (PJI) rates among total knee arthroplasty patients within a health care facility in South Carolina.

H1: There is a relationship with attendance of a joint education class; independent variables (smoking, BMI, A1c levels); on reducing postoperative prosthetic joint infection (PJI) rates among total knee arthroplasty patients within a health care facility in South Carolina.

This research question focuses on total hip patients and the relationship between the independent variables and the second dependent variable (LOS) in this study and is as follows: Among total hip arthroplasty patients, is there a relationship with attendance of a joint education class; independent variables (smoking, BMI, A1c levels); on reducing overall patient LOS (dependent variable) within a health care facility in South Carolina?

The hypothesis associated with this research questions is as follows:

H0: There is no relationship with attendance of a joint education class; independent variables (smoking, BMI, A1c levels); on reducing overall patient LOS (dependent variable) among total hip arthroplasty patients within a health care facility in South Carolina.

H1: There is a relationship with attendance of a joint education class; independent variables (smoking, BMI, A1c levels); on reducing overall patient LOS (dependent variable) among total hip arthroplasty patients within a health care facility in South Carolina.

Finally, the fourth research question in this study evaluated the total knee arthroplasty patients and the relationship the independent variables have with LOS and is as follows: Among total knee arthroplasty patients, is there a relationship with attendance of a joint education class; independent variables (smoking, BMI, A1c levels); on reducing overall patient LOS (dependent variable) within a health care facility in South Carolina?

The hypothesis associated with this research questions is as follows:

H0: There is no relationship with attendance of a joint education class; independent variables (smoking, BMI, A1c levels); on reducing overall patient LOS (dependent variable) among total knee arthroplasty patients within a health care facility in South Carolina.

H1: There is a relationship with attendance of a joint education class; independent variables (smoking, BMI, A1c levels); on reducing overall patient LOS (dependent variable) among total knee arthroplasty patients within a health care facility in South Carolina.

Now that the research questions and hypothesis have been reviewed, the research design and rationale will be evaluated. This next section will describe the variables involved in the study, the type of study, the design of the analysis and some of the assumptions involved. It will then discuss some of the assumptions and tests needed to ensure the accuracy of the analysis.

3. Results

The goal in this study was to determine the effectiveness of the total joint class on postoperative outcomes. The first analysis performed used a logistic regression to evaluate the effect of the total joint class, as well as smoking status, A1c/diabetes, and BMI on PJIs. See Table 1 for the results.

Table 1.

Logistic Regression Model – Logistic GEE Shows O.R. for joint class in hip patients and joint in class patients, outcome is readmitted for PJI.

	OR	95% CI	p value
BMI Merged (Linear)	1.085	(1.045, 1.126)	<.001
Current Smoker	10.383	(2.800, 38.497)	<.001
Recent-Former Smoker	15.653	(5.834, 41.995)	<.001
A1c Diabetes with reasonable control	0.620	(0.448, 0.858)	<.004
A1c Diabetes not well controlled	1.523	(0.925, 2.507)	<.098
Hip - Total Joint Class (no attendance)	4.448	(1.958, 10.103)	<.001
Knee - Total Joint Class (no attendance)	0.665	(0.536, 0.825)	<.001
Knee - Total Joint Class (attended)	1.298	(1.013, 1.663)	<.039

Estimate of common correlation parameter is 0.0086. Calculated dispersion parameter = 0.972. CIC = 8.89. For the outcome variable, Y = 1 when Readmitted for PJI = 1.

Table 1 illustrates the results from the logistic regression performed on my data. I will first address the results directly related to my research questions and will then discuss other significant results related to the independent variables. To determine the relationship between the total joint class and PJIs in total hip patients (RQ1), I needed to create a reference variable as a value of 1. The reference value I used were the patients that had a total hip procedure, attended the joint class and had a PJI. Using this as the reference value and evaluating the total hip patients that did not attend class, I found a p value < 0.001, so therefore there was significance. In evaluating the odds ratio, the regression indicated an odds ratio of 4.45, so for patients that had a total hip procedure and did not attend the class, they were 4.45 times more likely to develop a PJI than those that did attend. Therefore, for my first research question evaluating the relationship of total joint class attendance for total hip patients and PJI, I would reject the null hypothesis as the results indicate there is a relationship between total joint class attendance and reduced PJI rates. Another notable result from Table 4 is that among patients attending the joint class, the odds of a total knee patient being readmitted for a surgical site infection are 1.30 times the same odds in a hip patient, with the remaining predictor values held constant. This odds ratio is statistically significantly different from 1 (the null value), with a p value of 0.039. In Table 5 are results of another logistic regression that focuses on the relationship among patients that did not attend the joint class and were readmitted with a PJI.

Table 4.

Categorical Variables by Readmitted for PJI – (Columns add to 100%).

	All Subjects	False	True	Estimate	p value
Gender - Female	752(61.8%)	729(62.0%)	23(56.1%)	0.593	0.441c
Gender – Male	464 (38.2%)	446(38.0%)	18(43.9%)
Procedure-Hip	406(33.4%)	383 (32,6%	23(56.1%)	9.839	0.002c
Procedure Knee	810(66.6%)	792(67.4%)
Joint Class – No	535(44.0%)	510(43.4%)	25(61.0%)	4.964	0.026c
Joint Class-Yes	681(56.0%)	665(56.5%	16(39.0%)
Current Smoker-False	1185(97.5%)	1149(97.8%)	36(87.8%)	NA	0.003f
Current Smoker-True	31(2.5%)	26(2.2%)	5(12.2%)
Recent/Former Smoker-False	1132(93.2%)	1108(94.3%)	24(58.5%)	NA	0.000f
Recent/Former Smoker-True	84(6.9%)	67(5.7%)	17(41.5%)
A1c Not Diabetic	773(63.6%)	746(63.5%)	27(65.9%)	−0.038	0.810 g
A1c Diabetes w/reasonable control	397(32.6%)	385(32.8%)	12(29.3%
A1c Diabetes not well controlled	46(3.8%)	44(3.7%)	2(4.9%)
BMI underweight	3(0.2%)	3(0.3%)	0(0.0%)	0.214	0.161 g
BMI normal	141(11.6%)	137(11.7%)	4(9.8%)
BMI overweight	311(25.6%)	304(25.9%)	7(17.1%)
BMI obese	761(62.6%)	731(62.2%)	30(73.2%)

f - Fisher's exact test; descriptive statistics are count (percent). The estimate is the odds ratio in the case of a 2 × 2 table.

c - Chi-square test; descriptive statistics are count (percent). The estimate is the chi-square statistic.

g - Goodman-Kruskal gamma test; descriptive statistics are count (percent). The estimate is gamma.

Table 5.

Categorical Variables by Joint Class Attendance – (Rows add to 100%).

	No	Yes	Estimate	p value
Gender - Female	322(42.8%)	430(57.2%)	1.109	0.292
Gender – Male	213(45.9%)	251(54.1%)
Procedure-Hip	183(45.1%)	223(54.9%)	0.287	0.592c
Procedure Knee	352(43.5%)	458(56.5%)
Current Smoker-False	511(43.1%)	674(56.9%)	14.422	<.001c
Current Smoker-True	24(77.4%)	7(22.6%)
Recent/Former Smoker-False	491(43.4%)	641(56.6%)	2.574	0.109c
Recent/Former Smoker-True	44(52.4%)	40(47.6%)
A1c Not Diabetic	335(43.3%)	438(56.7%)	−0.048	0.407 g
A1c Diabetes w/reasonable control	174(43.8%)	223(56.2%)
A1c Diabetes not well controlled	26(56.5%)	20(43.5%)
BMI underweight	2(66.7%)	1(33.3%)	−0/129	0.161 g
BMI normal	54(38.3%)	87(61.7%)
BMI overweight	124(39.9%)	187(60.1%)
BMI obese	355(46.6%)	406(53.4%)

f - Fisher's exact test; descriptive statistics are count (percent). The estimate is the odds ratio in the case of a 2 × 2 table.

c - Chi-square test; descriptive statistics are count (percent). The estimate is the chi-square statistic.

g - Goodman-Kruskal gamma test; descriptive statistics are count (percent). The estimate is gamma.

The results from the logistic regression in Table 2 indicates that among patients not attending the joint class, the odds of a hip patient being readmitted for a surgical site infection are 6.69 times the same odds in a knee patient, with the remaining predictor values held constant. This odds ratio is statistically significantly different from 1 (the null value), with a p value < 0.001. Next, I will continue to evaluate the results of the logistic regression as it pertains to my research questions, particularly the question (RQ2) related to patients that had total knee procedures and the relationship between joint class attendance and reduced PJI rates.

Table 2.

Logistic Regression Model – Logistic GEE Shows O.R. for joint procedures that did not attend joint class, outcome is readmitted for PJI. Estimate of common correlation parameter is 0.0086. Calculated dispersion parameter = 0.972.

	OR	95% CI	p value
BMI Merged (Linear)	1.085	(1.045, 1.126)	<0.001
Current Smoker	10.383	(2.800, 38.497)	<0.001
Recent-Former Smoker	15.653	(5.834, 41.995)	<0.001
A1c Diabetes with reasonable control	0.620	(0.448, 0.858)	0.004
A1c Diabetes not well controlled	1.523	(0.925, 2.507)	0.098
Hip - Total Joint Class (no attendance)	6.69	(2.383, 18.789)	<0.001
Knee - Total Joint Class (attended)	1.95	(1.257, 3.033)	0.003
Hip - Total Joint Class (attended)	1.50	(1.213, 1.866)	0.003

CIC = 8.89. For the outcome variable, Y = 1 when Readmitted for PJI = 1.

Referring back to Table 1, among patients attending the joint class, the odds of a total knee patient being readmitted for a surgical site infection are 1.30 times the same odds in a hip patient, with the remaining predictor values held constant. This odds ratio is statistically significant different from 1 (the null value), with a p value of 0.039. The next table displays data that looks into the total knee patients further and the relationship between joint class attendance and PJIs. It uses the reference variable of knee patients that attended the joint class.

In Table 3, the results can be interpreted as among knee patients, the odds of those who did not attend the total joint class being readmitted for a surgical site infection are 0.51 times the same odds in a patient who did attend the joint class, with the remaining predictor values held constant. This odds ratio is statistically significantly different from 1 (the null value), with a p value of 0.003. Therefore, for my second research question that focused on total knee patients sought to determine if there is a relationship with attendance of a joint education class; independent variables (smoking, BMI, A1c levels); on reducing postoperative prosthetic joint infection (PJI) rates (dependent variable) within a health care facility in South Carolina? The answer would be that I would fail to reject the null hypothesis, as there was no significance in the relationship that joint class attendance reduced PJI rates. These results somewhat tie back to literature reviewed. As mentioned previously, Jordan et al. (2014)² performed two studies that analyzed preoperative education and found a decrease in preoperative expectation and improvement in knowledge, flexibility and regularity of exercise among the patients, but alone there was no significant differences in validated joint specific patient reported outcomes. However, when the study was combined with treatment by a physiotherapist, there was a reduction in costs and LOS. While my results did indicate a relationship between the class and reduced PJI rates for total hip patients, the relationship for total knee patients was the inverse of what may have been expected. Now, that I have addressed the results directly related to the research questions, I will review the other results as they related to my independent variables, BMI, smoking, and A1c.

Table 3.

Logistic Regression Model – Logistic GEE Shows O.R. for joint class in knee patients and joint in class patients, outcome is readmitted for PJI.

	OR	95% CI	p value
BMI Merged (Linear)	1.085	(1.045, 1.126)	<0.001
Current Smoker	10.383	(2.800, 38.497)	<0.001
Recent-Former Smoker	15.653	(5.834, 41.995)	<0.001
A1c Diabetes with reasonable control	0.620	(0.448, 0.858)	0.004
A1c Diabetes not well controlled	1.523	(0.925, 2.507)	0.098
Hip - Total Joint Class (no attendance)	3.427	(2.383, 18.789)	<0.001
Knee - Total Joint Class (attended)	0.512	(1.257, 3.033)	0.003
Hip - Total Joint Class (attendance)	0.771	(1.213, 1.866)	<0.001

Estimate of common correlation parameter is 0.0086.

Calculated dispersion parameter = 0.972.

CIC = 8.89. For the outcome variable, Y = 1 when Readmitted for PJI = 1.

In Table 1, BMI was determined to have linearity. Therefore, in evaluating the results and treating BMI as a continuous variable, there was a p value of 0,003, which was less than 0.05, indicating significance. The odds ratio for BMI was 1.085. This indicates that for every 1 unit increase in BMI, the patient had an 8.5% increased likelihood of having a PJI. These results did coincide with results in other research reviewed. Meller et al. (2016)⁴ found that morbidly obese patients were more likely to have a PJI following a total hip or total knee procedure compared to those of normal weight. These findings indicate a focus on BMI could influence PJI and subsequently LOS. Now that the results of the independent variable BMI have been reviewed, the results for smoking will be evaluated.

In evaluating the results of smoking, whether the patient was a current smoker, recent smoker, or a nonsmoker, the regression once again used a reference variable and it was placed at a value of 1. The reference variable used for smoking was the nonsmoker. Evaluating the results of the current smoker, the p value of <0.001 indicated the results were significant. Therefore, with an odds ratio of 10.383, the results showed that for those patients that are current smokers, they are 10.4 times more likely to develop a PJI than those that are nonsmokers. As for the recent-former smokers, the p value < 0,001 indicated significance as well, and with an odds ratio of 15.653, these patients have 15.7 times the likelihood of having a PJI than those that are nonsmokers. These results not only agree with literature supporting smoking as being a factor in PJI, but the odds ratios are actually much higher than research reviewed. Gonzalez et al. (2017)¹ found that patients that had a total hip or knee procedure and smoked were 1.8 times more likely to develop a PJI than nonsmokers. This ratio is much less than the 10.4 and 15.7 odds ratios my results found for current and former smokers respectively. Next, I will review the results of the A1c data.

The A1c patients were categorized into 3 categories. The first category, which was the reference variable for this regression was nondiabetic patients. The other two categories include A1c/diabetic patients well controlled and A1c/diabetic patients that are not well controlled. With nondiabetics as the reference variable, the A1c/diabetic patients with reasonable control had a p value of 0.004. Therefore, there was significance in the relationship between nondiabetic patients and diabetic patients with reasonable control and the development of a PJI. The odds ratio of 0.620, however indicates that the A1c/diabetic patients well controlled, actually have less of a chance of developing a PJI than nondiabetics. As for the A1c/diabetics not well controlled, there was a p value of 0.098, therefore indicating there is no significance in the relationship of nondiabetic patients and diabetic patients not well controlled and the development of a PJI. These results did not relate to the results I found in my research. Other studies have found that A1c levels and diabetic status have contributed to PJIs. Kremers et al. (2015)³ found that patients with diabetes and perioperative hyperglycemia (elevated blood glucose and A1cs) were more likely to develop PJI when compared to patients with normal glucose levels. In order to further explore the variables and to test for Type 1 and Type 2 errors, the following analysis was also performed.

Table 4 shows the evaluation of the independent variables independent of one another; therefore, there were no controls. The analysis evaluated each group of categorical variables and tested for a relationship with readmission for PJI. The notable categorical variables are the procedural variables (hip and knee) and both smoking variables (current and recent/former). The p value for the procedural variables was 0.002 and had an odds ratio of 9.839, which was a significant relationship and indicated that patients that had total hip procedures were 9.8 times more likely to have a PJI than total knee patients. As for the current smokers, the p value of 0.003 indicated a significant relationship between current smokers and nonsmokers and whether or not they were readmitted with a PJI. The categorical variables of recent/former smokers and non-recent/former smokers was 0.000, which also indicated a significant relationship between recent/former smokers, non-recent/former smokers and whether or not they were readmitted with a PJI.

The second table (Table 5) of categorical variables analyzed the independent variables and whether or not they were predictors of joint class attendance. The analysis evaluated each group of categorical variables and tested for a relationship with joint class attendance. The notable categorical variables are the current smokers and noncurrent smokers and the BMI categorical variables. The p value for the current smokers and noncurrent smokers was <0.001 and had an odds ratio of 14.422, which was a significant relationship and indicated that patients that were current smokers were 14.4 times more likely not to attend joint class. As for the BMI categorical variables, the p value of 0.016 indicated a significant relationship between what BMI category the patient was categorized and whether or not they attended joint class. Now that the questions related to the relationship between joint class attendance and reduced PJI rates with the other independent variables have been reviewed, I will now move on to my last two research questions. The following Poisson regression was used to evaluate the relationship between joint class attendance, including the independent variables of smoking, A1c/diabetes and BMI and reducing postoperative LOS of the patients that had a total hip or total knee arthroplasty.

The following paragraphs will describe the results of the data listed in Table 6. First, I will review the results that are associated with my 3rd and 4th research questions. The results from the Poisson, showed that patients not attending the total joint class are expected to stay in the hospital 16.5% more days than those attending the class, with the remaining predictor values held constant. This is true whether the patient had hip or knee surgery. This association is not likely due to chance (p = 0.043). Therefore, for research questions 3 and 4 regarding the relationship between joint class attendance and reducing postoperative LOS, I determined that there was a relationship and therefore reject the null hypothesis. Therefore, whether the patient had a total hip or total knee procedure, there was a relationship between joint class attendance and a reduced postoperative LOS. As with the logistic regression I also evaluated the impact of the other variables on postoperative LOS. The first results are regarding smoking and the relationship with postoperative LOS.

Table 6.

Poisson GEE, outcome is LOS.

	I.R.R.	95% CI	p value
Total Joint Class (no attendance)	1.165	(1.005, 1.350)	0.043
Recent-Former Smoker	0.849	(0.785, 0.918)	<0.001
Current Smoker	1.152	(0.886, 1.497)	0.291
A1c Diabetes with reasonable control	1.031	(0.936, 1.137)	0.535
A1c Diabetes not well controlled	1.129	(1.093, 1.167)	0.098
BMI - Underweight	0.561	(0.507, 0.621)	<0.001
BMI - Overweight	0.998	(0.969, 1.009)	0.258
BMI - Obese	1.049	(1.000, 1.101)	0.049
Procedure - Hip	1.091	(0.998, 1.192)	0.054

The correlation structure is exchangeable, and the ID variable is Surgeon.

The estimate of the common correlation parameter is 0.0286.

Calculated dispersion parameter = 1.147.

CIC = 7.63.

The results for smoking were somewhat conflicting with what I have seen in other research. Recent/former smokers are expected to stay in the hospital 15.1% fewer days than nonsmokers, with the remaining predictor values held constant. This association is not likely due to chance (p < 0.001). As for current smokers, they are expected to stay in the hospital 15.2% more days than nonsmokers, with the remaining predictor values held constant. This association may be due to chance (p = 0.291). The results for the next variable discussed will be for A1c/diabetes.

The results of the relationship between patients who had their diabetes under control and nondiabetics, was also different from what I've seen in other research. Patients with diabetes under reasonable control (HgbA1c ≤ 7.0) are expected to stay in the hospital 3.1% more days than nondiabetics (no record of HgbA1c), with the remaining predictor values held constant. This association may be due to chance (p = 0.535). However, patients with diabetes not well controlled (HgbA1c > 7.0) are expected to stay in the hospital 12.9% more days than nondiabetics (no record of HgbA1c), with the remaining predictor values held constant. This association is not likely due to chance (p < 0.001). The next and final variable results were for BMI, and for the Poisson regression, this variable was not evaluated in linearity, but instead evaluated from a categorical standpoint of underweight, overweight and obese, with normal weight set as the reference variable. Underweight patients (BMI < 18.5) are expected to stay in the hospital 43.9% fewer days than patients of normal weight (18.5 ≤ BMI < 25), with the remaining predictor values held constant. This association is not likely due to chance (p < 0.001). Overweight patients (25 ≤ BMI < 30) are expected to stay in the hospital 1.1% fewer days than patients of normal weight (18.5 ≤ BMI < 25), with the remaining predictor values held constant. This association may be due to chance (p = 0.258). The obese patients (BMI ≥ 30) are expected to stay in the hospital 4.9% more days than patients of normal weight (18.5 ≤ BMI < 25), with the remaining predictor values held constant. This association is not likely due to chance (p = 0.049). One last notable result indicated that total hip patients are expected to stay in the hospital 9.1% more days than knee replacement patients, with the remaining predictor values held constant. This association may be due to chance (p = 0.054).

Based on the analysis there was a statistically significant relationship between total hip patients and their joint education class attendance and reduced PJI rates, but results did not show a significant relationship among total knee patients. There was also a relationship between BMI and smoking, on PJI rates. However, there was no relationship between patients A1c levels and PJI. As for the last two research questions, a Poisson regression was used to determine the relationship between the independent variables (joint class attendance, BMI, smoking, and A1c/diabetes) and the dependent variable LOS. This regression indicated that for those total hip or total knee patients that did not attend the joint education class, they were expected to stay in the hospital 16.5% more days than those attending the class, with the remaining predictor values held constant. This indicated that for both research questions 3 and 4, there was a relationship between joint class attendance and reduced postoperative LOS, so therefore I rejected the null hypothesis.

4. Conclusion

This study evaluated the effect of a total joint education class on patient outcomes to include PJI and postoperative LOS. The total joint class had a statistically significant effect on total hip patients in reducing PJIs as well as postoperative LOS. The class attendance also had a statistically significant relationship among total knee patients and reducing postoperative LOS. Although these results were the most significant, there was also useful information provided regarding the other independent variables such as BMI, smoking and A1c/diabetes. The results of this study will be used by administrators and physicians at the hospital from which the data was obtained. They will either use this information for further support and funding of the total joint class or use it in support of making the total joint class mandatory for all potential patients having these procedures. These decisions could impact social change at my hospital and in my community by improving future outcomes for the total hip and total knee patient population.