ABSTRACT
Background and Objective: One of the most common complications in postoperative procedures is the surgical site infection (SSI). Comprehensive monitoring and recording of SSIs are regularly performed by a joint collaboration between nations and WHO. The present study aimed to use statistical process control (SPC) as guiding tools to provide insight into data analysis to spot current state of challenges and provide protective measures and actions for future improvements for control of SSI in healthcare facilities. Materials and Methods: WHO global records of SSIs in yearly rate basis were collected from WHO online Internet data. Four countries were selected for comparative study namely: Czech Republic (CZE), Sweden (SWE), Kazakhstan (KAZ) and Finland (FIN). The SSI data were arranged chronological and subjected to trending, statistical analysis and SPC monitoring using statistical software packages. Results: The SWE and FIN records showed a gradual slight increase in SSI numbers. On the other hand, SSI values of both CZE and KAZ a general decrease in the trend line with the former slope being more stepper due to higher initial SSI (%) of cases. Interestingly, strong significant correlation either positive and negative between data of the four nations. The initial 6 years of SSI rate for CZE were exceptionally out of the normal trend record of the following years. The lowest mean value for SSI (%) was in FIN, while the lowest fluctuation range was for SWE. I-MR and Laney attribute charts were in agreement in control limits and alarming points of SSI. Conclusion: Improvement measures are required for control of SSI to avoid excursions.
PDF Abstract XML References Citation
How to cite this article
DOI: 10.3923/ajaps.2019.76.84
URL: https://scialert.net/abstract/?doi=ajaps.2019.76.84
INTRODUCTION
One of the most challenges of concern during invasive medical procedures is the intrusion of micro-organisms into the surgical wound which results in infection with possible postoperative complications1. This type of infection is called surgical site infection (SSI)2. The SSIs prevention is related to the implementation of good practices and control measures associated with surgery and preparation of the operation theater2. Hence, continuous monitoring of the cases of SSI is mandatory to determine the suitability of the existing measures and to set preventive control actions for future improvements3.
Statistical process control (SPC) tools have been applied successfully in different sections of the healthcare industry to provide a monitor, diagnose, assess, improve and control of specifically dedicated processes4. One of the basic tools in SPC is the control (also called process-behavior or Shewhart) chart which is very useful mean for the visualization of the inspected properties under investigation5. Thus, SPC may provide a useful mean and guiding tool to study SSI levels. However, fulfillment of presumed distribution of data is prerequisite for application of a specific type of control chart, otherwise false interpretation of results may occur6.
There have been studies that demonstrated debatable issue about the application of SPC-including Individual-Moving Range (I-MR) chart for data that show a significant departure from normality unless prior modification on data was made. However, other researchers showed the validity of the use of the I-MR chart for normality-skewed data7-11. Moreover, Laney-corrected attribute control charts have shown significant agreement with (I) chart control limits (CL) and alarms in some studied cases12,13.
Accordingly, the current work aimed to study the general trends of SSI in four different selected nations through analysis of yearly reported data collected through the world health organization (WHO) website database. A comparative study between the four countries will determine the existing degree state of SSI control and retrospectively. Using SPC tools embedded in commercial statistical software packages, the risk from SSI could be assessed and the future improvement may be determined. Two types of trending charts (I-MR and Laney-attribute charts) will be used comparatively to assess their mutual outcome to elucidate the previously mentioned dilemma.
MATERIALS AND METHODS
Data source and collection: Retrospective data of SSI from four selected countries were collected from WHO European Health Information At Your Fingertips.Gateway.euro.who.int14. The four nations were: Czech Republic, Sweden, Finland and Kazakhstan abbreviated by ISO 3 as CZE, SWE, FIN and KAZ. Table 1 shows general information of each country based on WHO database14.
Trending chart: Data of European Health for All database (HFA-DB) was interpreted using Microsoft Office Excel 2007.
Histogram and non-parametric correlation (p<0.05%), box plot diagram and basic statistical analysis: Processing of European HFA-DB record was done using Minitab® V 17.1.0 and GraphPad Prism V 6.01, according to their electronic manual15,16.
Normality test and distribution fitting: Description and modeling of European HFA-DB record was done using GraphPad Prism V 6.01 and XLSTAT (add-in program) V 2014.5.03, according to their electronic manual15,17.
Data fitting test for ordinary attribute chart, construction of Laney-modified attribute and I-MR control chart: Test for fitting to Poisson distribution, construction of U’ and I-MR charts for the yearly record of SSI rate of the four countries: CZE, SWE, FIN and KAZ was done using Minitab® V 17.1.0 according to their electronic manual16. Alarms interpretation were done as previously reported in Six Sigma work18.
Table 1: | WHO data information of four selected nations for SSI rate study |
Source: WHO14 |
RESULTS AND DISCUSSION
Morbidity and mortality associated with SSI are worth close monitoring and continuous monitoring worldwide to contain it19. The SPC methods may provide crucial means to combat this threat by delivering guiding tools20. Thus, countries that showed a slight gradual increase in SSI rate with time are SWE and FIN in contrary to KAZ and CZE where the initial SSI (%) was higher with CZE slope more steeper due to exceptionally high initial rates of infections. This trend could be demonstrated as shown in Fig. 1. On the other hand, Fig. 2 showed that all SSI rate records are not unimodal indicating possibly more than one set of procedures are affecting SSI (%). Bimodal or multimodal distributions of infection in hospitals are not uncommon phenomenon21,22. It should be noted that SSI rates are strongly related with CZE and CAZ are negatively associated with SWE and FIN. The very low p-values are indicative of the exclusion of correlation (rs) possibility due to randomness and it is possible that SSI influential factors are related in different nations.
Boxplot diagram in Fig. 3 shows data dispersion visually with CZE demonstrating high spreading with six outlier results (indicated by asterisks) of the 1st years of SSI recording and SWE followed by FIN are showing low spreading of data. Figure 3 demonstrated greatest variability, range and standard error of mean (SEM) for CZE and lowest corresponding values for SWE indicating better SSI control during the study period23,24. The SSI rate of CZE is significantly different from that of SWE, FIN and KAZ at 95% confidence interval (CI) and α = 0.05. However, the mean figure of SSI rate was higher for SWE than FIN and KAZ values. All data failed to pass normality test at α = 0.05 except FIN by Kolmogorov-Smirnov (KS) normality test. However, at lower confidence values they may show positive Gaussian distribution test response.
The diagnostic test of data fitting for ordinary attribute process-behavior charts shows that SSI results for the four data sets did not meet the requirement which required consideration of Laney-modification for correction of over-dispersion or under-dispersion of data to avoid false alarms25 as could be shown in Fig. 4.
Fig. 1: | Time series plot of SSI rate of Czech Republic (CZE), Sweden (SWE), Kazakhstan (KAZ) and Finland (FIN) |
Source: European Health for All database (HFA-DB) interpreted using Microsoft Office Excel 2007 |
Fig. 2(a-d): | Histogram of four record sets showing the distribution of SSI rate data visually and numerical correlation matrix between them at 95% CI |
Source: European Health for All database (HFA-DB) using Minitab® V 17.1.0 and GraphPad Prism V 6.01 |
Fig. 3: | Box and whisker diagram visualizing data distribution pattern from four record sets with descriptive statistical information |
SEM: Standard error of mean. Source: European Health for All database (HFA-DB) using Minitab® V 17.1.0 and GraphPad Prism V 6.01 |
In the same line, distribution fitting test showed that the most fitting distribution for each data does not follow any assumed distribution requirements for the required Shewhart charts. Instead, the distribution is far than that presumed or not follow it at all at confidence level = 0.05%. Laney-solution for correction of data dispersion provided a solution previously in other healthcare researches without the need of complex, time-consuming and error prone transformation of the numbers of the record sets26,27.
Fig. 4(a-d): | Test for the determination of validity for direct application of attribute control charts (a) CZE SSI, (b) SWE SSI, (c) FIN SSI and (d) KAZ SSI of four data sets using Minitab® V 17.1.0 |
Fig. 5(a-d): | Laney-modified attribute control charts (a) CZE SSI, (b) SWE SSI, (c) FIN SSI and (d) KAZ SSI of the four data records of SSI rates |
Source: European Health for All database (HFA-DB) using Minitab® V 17.1.0 |
Figure 5 and 6 illustrated SSI rates for the four countries using two types of control charts: Laney-U’ and I-MR charts, respectively. Red dots are indicative of out-of-control years of SSI rate. Both Laney and (I) charts are in agreements in both alarm states, number and the control limits (CLs) which is in agreement with recently reported work28.
Fig. 6(a-h): | I-MR control charts for the four data records of SSI rates |
Source: European Health for All database (HFA-DB) using Minitab® V 17.1.0 |
However, it should be noted that MR chart may provide an additional advantage in the current case for monitoring the stability of SSI yearly variation before making a decision on the control of SSI mean rate. In such instance, while SWE and FIN showed stable process variations, while both CZE and KAZ were out-of-control with unstable variation shift. Nevertheless, the drift in CZE MR mean could be linked to the abnormal sudden switch from the high period of SSI rates to low values due to external factors (supported by two hump-shaped histogram), in contrast to the more gradual inflection between the two periods in KAZ. Accordingly, reconstruction of an updated control chart may be required for SSI of both countries to examine the new situation but after gathering more data to build-in sufficient confidence in the newly emerged change and to verify its stability28. Interestingly, despite the stability in SSI rate of change for both SWE and FIN, the general trend line of the process mean tends to increase progressively reaching a newly different state of SSI (%). Again, more data are required to generate statistically significant control parameters to re-examine the new situation.
CONCLUSION
The SSI is a global problem that impacts human health and challenging healthcare industry. The SPC provides an indispensable tool to control, monitor, improve and investigate SSI rates. It assesses the implemented measures adequacy to prevent infections from surgical procedures and evaluate changes made for improvements. The present investigation shows that further improvements are required to control and stabilize SSI rates. The current case provided evidence for equivalency between Laney-attribute and individual charts in alarms and CLs interpretation despite the failure of data to fulfill the prerequisite distribution at the significance level of the test.
SIGNIFICANCE STATEMENT
The present study will help healthcare practitioners to evaluate their measures to control SSI and the stability of the process over time. The SPC can relate data distribution, behavior and shape to the changes in the action taken, missed processes, event occurrence and/or system modification. Control charts provide insight into the area of defects to correct them and improvement spots to use them in the process enhancement. Early warning of the drift in inspection quality characteristic values could be visualized in the process-behavior chart before any excursion in SSI numbers would occur and if the shift is toward the favorable side (decrease percent of the cases) the applied good practices should be maintained, controlled and stabilized by the medical staff to be reassessed over relatively long time to judge its validity.
REFERENCES
- Clayton, J.L. and K.J. Miller, 2017. Professional and regulatory infection control guidelines: Collaboration to promote patient safety. AORN J., 106: 201-210.
CrossRefDirect Link - Eissa, M.E.A., 2018. Minimization of microbial infection risk from surgery. Open J. Surg., 1: 20-23.
Direct Link - Thor, J., J. Lundberg, J. Ask, J. Olsson, C. Carli, K.P. Harenstam and M. Brommels, 2007. Application of statistical process control in healthcare improvement: Systematic review. BMJ Qual. Saf., 16: 387-399.
CrossRefDirect Link - Chakraborti, S., 2014. Nonparametric (Distribution-free) quality control charts. Wiley StatsRef: Stat. Ref. Online.
CrossRefDirect Link - Aliverdi, R., L.M. Naeni and A. Salehipour, 2013. Monitoring project duration and cost in a construction project by applying statistical quality control charts. Intl. J. Project Manage., 31: 411-423.
CrossRefDirect Link - Salehipour, A., L.M. Naeni, R. Khanbabaei and A. Javaheri, 2015. Lessons learned from applying the individuals control charts to monitoring autocorrelated project performance data. J. Construct. Eng. Manage., Vol. 142.
CrossRefDirect Link - Eissa, M.E., 2018. Variable and attribute control charts in trend analysis of active pharmaceutical components: Process efficiency monitoring and comparative study. Exp. Med., 1: 31-44.
Direct Link - Eissa, M.E., M. Seif and M. Fares, 2015. Assessment of purified water quality in pharmaceutical facility using six sigma tools. Int. J. Pharm. Qual. Assur., 6: 54-72.
Direct Link - Sawyer, R.G. and H.L. Evans, 2018. Surgical site infection-the next frontier in global surgery. Lancet Infect. Dis., 18: 477-478.
CrossRefDirect Link - Baker, A.W., S. Haridy, J. Salem, I. Ilies and A.O. Ergai et al., 2018. Performance of statistical process control methods for regional surgical site infection surveillance: A 10-year multicentre pilot study. BMJ Qual. Saf., 27: 600-610.
CrossRefDirect Link - Kong, W., J. Wang, Q. Mao, L. Ren and S. Zhang et al., 2015. Early-versus late-onset prosthetic mesh infection: More than time alone. Indian J. Surg., 77: 1154-1158.
CrossRefPubMedDirect Link - Misiakos, E.P. and K.G. Tsalis, 2018. Editorial: Surgical infections. Front. Surg., Vol. 5.
CrossRefDirect Link - McGill, R., J.W. Tukey and W.A. Larsen, 1978. Variations of box plots. Am. Statistician, 32: 12-16.
CrossRefDirect Link - Frigge, M., D.C. Hoaglin and B. Iglewicz, 1989. Some implementations of the boxplot. Am. Statistician, 43: 50-54.
CrossRefDirect Link - Chhapola, V., A.G. Sharma, S.K. Kanwal, V. Kumar and B. Patra, 2018. Neonatal exchange transfusions at a tertiary care centre in north India: An investigation of historical trends using change-point analysis and statistical process control. Int. Health, 10: 451-456.
CrossRefDirect Link - Eissa, M., 2018. Evaluation of microbiological purified water trend using two types of control chart. Eur. Pharm. Rev., 23: 36-38.
Direct Link - Eissa, M.E., 2018. Adulterated pharmaceutical product detection using statistical process control. Bangladesh Pharm. J., 21: 7-15.
CrossRefDirect Link