ADVERTISEMENT
Using Statistical Process Control for Monitoring the Prevalence of Hospital-acquired Pressure Ulcers
Abstract
Institutionally acquired pressure ulcers are used as outcome indicators to assess the quality of pressure ulcer prevention programs. Determining whether quality improvement projects that aim to decrease the proportions of institutionally acquired pressure ulcers lead to real changes in clinical practice depends on the measurement method and statistical analysis used.
To examine whether nosocomial pressure ulcer prevalence rates in hospitals in the Netherlands changed, a secondary data analysis using different statistical approaches was conducted of annual (1998–2008) nationwide nursing-sensitive health problem prevalence studies in the Netherlands. 
Institutions that participated regularly in all survey years were identified. Risk-adjusted nosocomial pressure ulcers prevalence rates, grade 2 to 4 (European Pressure Ulcer Advisory Panel system) were calculated per year and hospital. Descriptive statistics, chi-square trend tests, and P charts based on statistical process control (SPC) were applied and compared. Six of the 905 healthcare institutions participated in every survey year and 11,444 patients in these six hospitals were identified as being at risk for pressure ulcers. Prevalence rates per year ranged from 0.05 to 0.22. Chi-square trend tests revealed statistically significant downward trends in four hospitals but based on SPC methods, prevalence rates of five hospitals varied by chance only. Results of chi-square trend tests and SPC methods were not comparable, making it impossible to decide which approach is more appropriate. P charts provide more valuable information than single P values and are more helpful for monitoring institutional performance. Empirical evidence about the decrease of nosocomial pressure ulcer prevalence rates in the Netherlands is contradictory and limited.
Potential Conflicts of Interest: none disclosed
Measuring clinical practice performance is part of quality improvement and assurance in healthcare.1-3 Hospital-acquired pressure ulcers are used as outcome indicators for measuring the quality of pressure ulcer prevention in various healthcare settings.4-6 Pressure ulcer prevalence rates in European hospitals vary between 8% and 23%.7 In the past years, several national and local quality initiatives have been launched to improve pressure ulcer prevention and management. However, empirical evidence as to whether these initiatives and projects were successful is contradictory. Pressure ulcer prevalence and incidence rates in different settings decreased,8,9 remained stable,10,11 or increased.4 One reason might be that clinical practice improved, did not change, or became worse, subsequently affecting outcomes; assessing changes in clinical practice seems to be a factor of measurement methods and statistical analysis. Two key issues prevail: measurement frequency and trend detection.
Measurement frequency. Numerous improvement projects typically are evaluated by applying before-after measurements9,11,12; subsequent statistical tests are used to detect statistically significant differences. If the null hypothesis is rejected, it is concluded that a “real” change has taken place. This conclusion may be misleading because whether a relationship exists between the intervention and the measured change is uncertain. Prevalence or incidence rates could have increased or decreased even in the absence of the intervention. When only two measurements are used, detecting any trends or even long-term changes is hardly possible.2,13 Furthermore, the probability for detecting statistically significant differences increases with the sample size. In other words, to a certain extent the number of patients or residents determines whether improvement evaluations are considered successful. This is a clear disadvantage, especially for smaller healthcare institutions evaluating their own performance.
Detecting trends. Although it is more favorable for institutions to have more measurements through time, determining “real” trends is difficult.2,13 Descriptive statistics14-16 or chi-square tests for trends8,17 are frequently applied to address related concerns. Both approaches are limited to determining changes or trends and interpreting results remains subjective8 and sometimes elusive.18
In recent years in the Netherlands, there has been considerable effort to improve pressure ulcer management.19 Additionally, since 1998, annual nationwide pressure ulcer point prevalence surveys (LPZ) have been conducted. Once a year, hospitals, nursing homes, and home care institutions are invited to participate. According to a standardized study protocol, data regarding pressure ulcers, pressure ulcer risk, and preventive and therapeutic measures are obtained by trained nurses in all participating institutions; in general, pressure ulcer prevalence rates seem to have decreased over the survey years.20 However, this hypothesis has not been rigorously tested.
The aim of this secondary data analysis was to examine changes in nosocomial pressure ulcer prevalence rates in hospitals in the Netherlands between 1998 and 2008 using descriptive statistics, chi-square trend tests, and Statistical Process Control (SPC) methods and to evaluate which statistical approach might be more appropriate.
Methods
Design and measures. An analysis of the LPZ data from 1998 to 2008 was conducted. At the time of the analysis, this data base (owned by the University of Maastrich) contained information about several nursing-sensitive health problems (eg, falls, nutritional status) and about preventive and therapeutic measures implemented among approximately 320,000 individuals in 900 different healthcare institutions across the Netherlands. Confidentiality was ensured by the assignment of identification numbers to participating institutions. For this secondary data analysis, no identifying information (eg, names, addresses) was available.
Hospitals that participated regularly in all survey years were identified and data about pressure ulcer risk and pressure ulcer prevalence were extracted. To adjust for pressure ulcer risk, only patients with restricted mobility were considered for the analysis because immobility is the strongest predictor for pressure ulcer development. 21 The degree of immobility was rated from 1 (completely immobile) to 4 (no limitation) on the 4-point “Mobility” scale included in the Braden scale.22 Inter-rater reliability of the item scores of the Dutch version were ICC (1,1) 0.86 (95% CI 0.78–0.85) and 0.86 (0.83–0.89).23 Patients with item scores of 1, 2, or 3 were regarded as being at risk.
Prevalence rates for hospital-acquired pressure ulcers were defined as persons at risk for pressure ulcers with at least one pressure ulcer grade 2, 3 or 4 according to the European Pressure Ulcer Advisory Panel (EPUAP)24 that developed in the institution divided by all persons in the institution at risk at that point of time. Due to high degrees of measurement error, grade 1 pressure ulcers were excluded from all calculations.25,26
Statistical analysis. Prevalence rates were calculated per year and institution. Chi-square trend tests (Mantel-Haenzel chi-square) testing the significance of linear relationships27 between increasing years and prevalence were applied for every hospital and assumed that due to improved pressure ulcer prevention over years the prevalence changed. This approach is frequently used to investigate trends of prevalence rates over years.28-30 P values <0.05 (two sided) were considered statistically significant.
The basic assumption of SPC is that repeated measurements from a process will exhibit variation. If the process is stable, its variation is predictable within some limits and can be described by a statistical distribution. Natural random variation within a process over time is called common cause variation. Based on the statistical distribution, common cause variation is stable and predictable within upper and lower control limits. Special cause variation refers to unnatural variation due to events, changes, or circumstances (eg, quality improvement projects) not typical or inherent to the regular process.2,31,32
P charts were created for every hospital. Prevalence rates were plotted over the years in chronological order including the center line (mean) and upper and lower control limits. Subsequently, every P chart was screened for special cause variation. Special cause variation was assumed when 1) a single data point exceeded the upper or lower control limit, 2) eight consecutive data points were on the same side of the center line, or 3) a trend (six data points constantly increasing or decreasing) was exhibited.2,31 Finally, results of the descriptive statistics, chi-square tests for trends, were compared with the presence of special cause variations detected by SPC methods.
Results
In total, 905 different healthcare institutions throughout the Netherlands participated in at least one survey between 1998 and 2008. Some institutions participated only once, some participated regularly. Six hospitals (labeled A to F) participated in all years (see Table 1). A total of 11,444 patients across these six hospitals and 10 years were considered at risk for developing pressure ulcers. Prevalence rates per year ranged from 0.05 (hospitals B, E, F) to a maximum of 0.22 (hospital D). Mean prevalence rates between all six hospitals were comparable, ranging from 0.10 to 0.12. Results of chi-square trend tests suggest the prevalence rates changed significantly over time with increasing years in hospitals A, B, C and E (see Table 1). P charts are shown in Figures 1 through 6. Based on the predetermined rules, there was one case of special cause variation in hospital D (see Figure 4); one data point was above the upper control limit in 2001. Prevalence rates of the remaining hospitals exhibited common cause variation.
Discussion
Six hospitals participated annually in the national prevalence studies in the Netherlands between 1998 and 2008. Based on calculated proportions, the prevalence rates in the first half of the study period appear to be higher than in the second half. However, considerable variation between these point estimates was observed. Chi-square trend tests revealed statistically significant changes for four hospitals. When SPC methods were applied, no decrease in hospital-acquired pressure ulcers grade 2, 3 or 4 from 1998 to 2008 could be detected. Changes in prevalence rates of five hospitals were considered as common cause variation, indicating that a real change or shift in the process had not taken place. Assuming a stable process, the only significant increase of hospital-acquired pressure ulcers was observed in 2001 in one hospital.
Results of the chi-square trend tests and SPC methods were not comparable. In four cases, P values indicated downward trends; whereas, P charts showed common cause variation only. It is impossible to decide which approach is more appropriate or “correct,” but P charts display much more relevant information. Single P values are unable to characterize the ups and downs of prevalence rates per year.
Hospital B provides a valuable example: According to the chi-square trend test, pressure ulcer prevalence was reduced significantly over the study period but the corresponding P chart reveals a rather cyclic behavior within predictable limits. Hospital D provides another example. Based on the P value of 0.078, one must accept the null hypothesis and conclude no trend occurred over the years. However, the corresponding P chart (see Figure 4) reveals that in 2001 the prevalence rate was significantly higher statistically as could be expected by chance.
The examples provided indicate that P charts are more helpful for institutions monitoring their own performance. First, based on data from the past, future performance can be predicted assuming no fundamental change is instituted and the process remains stable. Second, P charts help evaluate the success of quality improvement projects because they indicate when and whether significant changes have taken place. After the change, P charts can be used as a monitoring aid to ensure improvements are sustained over time. SPC methods have been applied in various fields of quality improvement in healthcare such as infection control, blood sugar management in diabetes, medication errors, or mortality after certain procedures.2,33 The current results support the idea that control charts are also suitable and useful for monitoring clinical performance.32-35 However this approach has limitations. Even if processes are under statistical control, the values of the measured indicators are not automatically clinically acceptable or desired.33 This is especially true for this topic, considering the long-enduring debate whether all pressure ulcers are preventable.36 Measurement alone is not appropriate for improving care but the authors believe that application of SPC provides a valuable tool for measuring and evaluating processes in pressure ulcer management.
b>Limitations
Based on the results, the answer to the research question regarding whether nosocomial pressure ulcer prevalence rates in hospitals in the Netherlands changed is tempered by the fact that only a few institutions with annual participation in such efforts were identified. Therefore, a selection bias is most likely and findings are not generalizable to other hospitals in the Netherlands. Furthermore, more frequent point estimates (eg, more frequent incidence or prevalence studies) would have helped in detecting trends.
Conclusion
Evidence regarding a decrease of pressure ulcer prevalence in Dutch hospitals examined is contradictory. Based on SPC methods, no decrease in hospital-acquired pressure ulcer prevalence in at risk patients occurred from 1998 to 2008. SPC methods provide valuable tools to measure and monitor care processes and outcomes in pressure ulcer and wound care. Control charts help managers and practitioners distinguish between natural and special variation and to decide whether “real” changes have taken place over time.
Acknowledgements
The authors thank the nurses and patients who participated in the annual pressure ulcer prevalence surveys.
Please address correspondence to: Jan Kottner, PhD, RN, Fanningerstrasse 61, 10365 Berlin, Germany; email:jan.kottner@charite.de.
1. Øvretveit J. Evaluating Health Interventions. Maidenhead, UK: Open University Press;1998.
2. Lloyd R. Quality Health Care. A Guide to Developing and Using Indicators. Sudbury, Massachusetts: Jones and Bartlett;2004.
3. Kazandjian VA, Wicker K, Ogunbo S, Silverman N. Understanding safer practices in health care: a prologue for the role of indicators. J Eval Clin Pract. 2005;11:161–170.
4. Scott JR, Gibran NS, Engrav LH, Mack CD, Rivara FP. Incidence and characteristics of hospitalized patients with pressure ulcers: state of Washington, 1987 to 2000. Plast Reconstruct Surg. 2006;117:630–634.
5. Baharestani MM, Black JM, Carville K. Dilemmas in measuring and using pressure ulcer prevalence and incidence: an international consensus. Int J Wound Care. 2009;6:97–104.
6. Leffmann CJ. [Quality management in decubitus prevention]. Zeitschrift für Gerontologie and Geriatrie. 2004;37:100–108.
7. Vanderwee K, Clark M, Dealey C, Gunningberg L, Defloor T. Pressure ulcer prevalence in Europe: a pilot study. J Eval Clin Pract. 2007;13:227–235.
8. Kottner J, Wilborn D, Dassen T, Lahmann N. The trend of pressure ulcer prevalence rates in German hospitals: results of seven cross-sectional studies. J Tissue Viabil. 2009;18:36–46.
9. DeLaat EH, Schoonhoven L, Pickkers P, Verbeek A, van Achterberg T. Implementation of a new policy results in a decrease of pressure ulcer frequency. Int J Qual Health Care. 2006;18:107–112.
10. VanGilder C, MacFarlane G, Meyer S. Results of nine international pressure ulcer prevalence surveys. Ostomy Wound Manage. 2008;54:40–54.
11. Padula CA, Osborne E, Williams J. Prevention and early detection of pressure ulcers in hospitalized patients. J WOCN. 2008;35:65–75.
12. Gunninberg L. EPUAP pressure ulcer prevalence survey in Sweden. J WOCN. 2000;33:258–266.
13. Baumgarten M, Shardell M, Rich S. Methodological issues in studies of the effectiveness of pressure ulcer prevention interventions. Advances Skin Wound Care. 2009;22:180–188.
14. Milne CT, Trigilia D, Houle TL, DeLong S, Rosenblum D. Reducing pressure ulcer prevalence rates in the long-term acute care setting. Ostomy Wound Manage. 2009;55:50-59.
15. Hopkins B, Hanlon M, Yauk S. Reducing nosocomial pressure ulcers in an acute care facility. J Nurs Care Qual. 2000;14:28–36.
16. Young ZF, Evans A, Davis J. Nosocomial pressure ulcer prevention. J Nurs Admin. 2003;33:380–383.
17. Lahmann NA, Dassen T, Poehler A, Kottner J. Pressure ulcer prevalence rates from 2002 to 2008 in German long-term care facilities. Aging Clin Experimental Res. In press.
18. Elliot R, McKinley S, Fox V. Quality improvement program to reduce the prevalence of pressure ulcers in an intensive care unit. Am J Crit Care. 2008;17:328–334.
19. Tannen A, Bours G, Halfens R, Dassen T. A comparison of pressure ulcer prevalence rates in nursing homes in the Netherlands and Germany, adjusted for population characteristics. Res Nurs Health. 2006;29:588–596.
20. Halfens R, Meijers JMM, Neyens JCL, Schols JMGA. Rapportage resultaten landelijke prevalentiemeting zorgproblemen 2008 [Report of the results of the nationwide prevalence study of care problems]. Maastricht: Universiteit Maastricht; 2008. Available at: www.lpz-um.eu/media/text/LPZ%20rapport%202009.pdf. Accessed April 29, 2010.
21. Sharp C, McLaws M-L. Estimating the risk of pressure ulcer development: is it truly evidence based? Int Wound J. 2006;3:344–353.
22. Bergstrom N, Braden BJ, Laguzza A, Holman V. The Braden scale for predicting pressure sore risk. Nurs Res. 1987;36:205–210.
23. Kottner J, Halfens R, Dassen T. An interrater reliability study of the assessment of pressure ulcer risk using the Braden scale and the classification of pressure ulcers in a home care setting. Int J Nurs Stud. 2009;46:1307–1312.
24. Dealey C, Lindholm C. Pressure ulcer classification. In: Romanelli M (ed). Science and Practice of Pressure Ulcer Management. London, UK: Springer;2006:37–41.
25. Kottner J, Dassen T, Lahmann N. Comparison of two skin examination methods for grade 1 pressure ulcers. J Clin Nurs. 2009;18:2464–2469.
26. Kottner J, Dassen, T. [Pressure ulcer prevalence measurement and interrater reliability]. PR-Internet für die Pflege. 2008;10:499–503.
27. Peat J, Barton B. Medical Statistics: A Guide to Data Analysis and Critical Appraisal. Malden, MA: Blackwell;2005.
28. Anthracopoulos MB, Liolios E, Panagiotakos DB, Triantou K, Priftis N. Prevalence of asthma among school children in Patras, Greece: four questionnaire surveys during 1987-2003. Arch Dis Child. 2007;92:209–212.
29. Chiang S-C, Chen C-Y, Chang Y-Y, Sun H-J, Chen WJ. Prevalence of heroin and methamphetamine male users in the northern Taiwan, 1999-2002: capture-recapture estimates. BMC Public Health. 2007;7:292.
30. Creek TL, Thuku H, Rahman M, Kilmarx PH. Declining syphilis among pregnant women in northern Botswana: an encouraging sign for the HIV epidemic? Sexually Transmitted Infect. 2005;81:453–455.
31. Benneyan JC, Lloyd RC, Plsek PE. Statistical process control as a tool for research and healthcare improvement. Qual Safety Health Care. 2003;12:458–464.
32. Noyez L. Control charts, Cusum techniques and funnel plots. A review of methods for monitoring performance in healthcare. Interact Cardiovasc Thorac Surg. 2009;9:494–499.
33. Thor J, Lundberg J, Ask J, et al. Application of statistical process control in healthcare improvement: systematic review. Qual Safety Health Care. 2007;16:387–399.
34. Clark DE, Cushing BM, Bredenberg CE. Monitoring hospital trauma mortality using statistical process control methods. J Am Coll Surg. 1998;186:630–635.
35. Lim TO. Statistical process control tools for monitoring clinical performance. Int J Qual Health Care. 2003;15:3–4.
36. Thomas DR. Are all pressure ulcers avoidable? J Am Med Direct Assoc. 2003;4:S43–S48.
