Predictors for Development of Pressure Ulcer in End-of-Life Care: A National Quality Register Study

Abstract

Background:

The scientific knowledge about pressure ulcers (PUs) is growing, but there is a shortage of studies of PUs at end of life. The recommendations regarding PU prevention in palliative care (PC) are based on consensus documents.

Aim:

To use data from a national register to identify predictors for development of PUs at the end of life.

Design:

A retrospective, descriptive, and comparative study design was used.

Setting/Participants:

All deceased patients over 17 years old (n = 60,319) and registered in the Swedish Register of Palliative Care (SRPC) during 2014 were included.

Statistical Analysis:

Logistic regression.

Results:

In the full model, all health units except general palliative home care had a significantly higher incidence of PUs than did the nursing homes. The well-known predictors of PUs in general, diabetes, postfracture state, infections, and multiple sicknesses, are predictors even in dying patients. Dementia was significantly associated with lower likelihood of PUs, while pain was associated with more PUs. Intravenous drip or enteral feeding was associated with a significantly decreased likelihood of developing PUs.

Conclusions:

The SRPC could be a unique resource for quality improvement and research. The present study cannot prove causation, but it can report correlations between background variables and PU prevalence. More studies, with different designs, are warranted to establish the roles of risk factors for PU in end-of-life care.

Introduction

The scientific knowledge about pressure ulcers (PUs) is growing, and research on the subject is being conducted, but there is a shortage of studies of skin changes at life's end (SCALE).¹ One of the conclusions of a literature review² was that the quality of PU prevention studies in patients with advanced illness has been poor. The recommendation regarding PU prevention in palliative care (PC) is based on consensus documents.^1,3,4 The 2014 National State of the Science Conference in the United States explored the issue of unavoidable PUs.⁵ The group reached consensus that unavoidable PUs occur, and one risk factor is terminal illness.

National quality registries

A system of national quality registries has been established in the health and medical services in Sweden in the last decades. There are about 100 registries that receive central funding and the Swedish Register of Palliative Care (SRPC)⁶ is one of them.⁷ These registries contain individualized data concerning patient problems, medical interventions, and outcomes after treatment in all healthcare settings. National quality registries are used in an integrated and active way for continuous learning, improvement, research, and management to create the best possible health and care together with the individual.

The National Board of Health and Welfare's guidelines for PC⁴ in Sweden proposes that the presence or absence of PUs serves as a qualitative indicator for PC and that the data source for this indicator is the SRPC.⁸ The use of pressure-relieving mattresses in end-of-life care has high priority in these guidelines. During 2014, a total of 60,353 deaths were reported to the register, representing 68.2% of all deaths in that year in Sweden.⁹ A study by Martinsson et al.¹⁰ found that participating in the SRPC improved end-of-life care in a number of ways. They report an increase in the prevalence of PUs of Category 1, but no change in the prevalence of more severe PU categories over time.¹⁰

PUs cause suffering and pain to the patient.¹¹ There is a clinical paradox to take into account in PC that turning a patient can both relieve and cause pain.¹² One consideration in the decision whether or not to turn the patient is prevention or exacerbation of PUs. A qualitative study by Searle and McInerney¹² on nurses' decision making about PU prevention during the last 48 hours of the patient's life showed that nurses' intention to give their patient a good death led to unsystematic practice, influenced by the feelings of the nurse, colleagues, and relatives as well as by the prevailing ward culture.

The goal in PC is to increase the quality of life and relieve suffering.¹³ Differences in the prevalence of PUs at death between different healthcare units raise the question whether they reflect a dissimilar risk for PUs or quality of care at the end of life.⁹ Furthermore, knowing whether patients dying with PUs have more pain compared with patients without PUs is clinically highly relevant knowledge that could guide nurses' decisions about turning or not turning a dying patient. Scientifically founded knowledge related to PUs at the end of life is therefore of huge importance.

The aim of this study is to use data from a national register to identify predictors for development of PU at end of life. Research questions were as follows:

1. Is place of death a predictor for PUs in end-of-life care?

2. Are age and gender predictors for PUs in end-of-life care?

3. Are different medical conditions predictors for PUs in end-of-life care?

4. Are different symptoms and treatments predictors for PUs in end-of-life care?

5. Is there a correlation between symptom control and PUs in end-of-life care?

Method

Design

A retrospective, descriptive, and comparative study design was used with data from a national quality register, the SRPC.

Sample

All deceased patients over 17 years old (n = 60,319) and registered in the SRPC during 2014 were included in this study.

Data collection and variables

Data were collected through an online questionnaire⁸ with items based on different essential aspects of end-of-life care. Questionnaires were answered by the registered nurse and/or physician in charge, as soon as the patient died. The questions are mandatory and have to be answered before submission of the questionnaire. Questions used in the present study related to place of death, expected/unexpected death, length of stay, sex, age, and PUs at admission and during the last week of life. PUs were classified as follows: Category 1 = nonblanchable erythema; Category 2 = partial-thickness skin loss; Category 3 = full-thickness skin loss; and Category 4 = full-thickness tissue loss.¹ The question regarding conditions that caused the death consisted of nine defined diagnosis categories and one “other” alternative, where the registering person could add an open-ended response. The questions relating to symptoms, symptom control, how long before death the patient lost decision-making capacity, and provision of parenteral/enteral feeding were only answered for expected deaths.

Data processing and statistical analysis

In the data processing of the question regarding “conditions that caused the death” (henceforth called “medical conditions”), answers under the answer alternative “other” have been placed in the right category wherever possible. A “new” category of infection has been added as some of the longer responses described infections. The deaths from “other” places of death are only reported in Table 1, and then excluded from the analyses because the open-ended responses showed that they had nothing in common.

Table 1.

Characteristics of the Patients Reported to the Swedish Register of Palliative Care in 2014 (n = 60,319)

Category	n	%
Sex
Male	27,536	45.7
Female	32,783	54.3
Death
Expected	50,960	84.5
Unexpected	7179	12.3
Missing	2180	3.6
Place of death
Nursing home	21,400	35.5
Hospital	21,272	35.3
Specialized PC inpatient unit	6009	10.0
Short stay at nursing home	4506	7.5
General palliative home care	3295	5.5
Specialized palliative home care	3109	5.2
Other (unspecified unit)^a	728	1.2
Medical conditions^b
Cancer	20,778	34.4
Heart disease	18,536	30.7
Respiratory disease	6211	10.3
Dementia	11,644	19.3
Stroke	6402	10.6
Other neurological diseases	2146	3.6
Diabetes	2937	4.9
Postfracture state	1818	3.0
Multiple diseases	14,425	23.9
Infection	2196	3.6

This category was excluded from further analysis.

The figures here do not add up to 60,319 because several patients had several of the listed diseases/conditions.

A logistic regression analysis was used to investigate the possibility of predicting risk for PUs in the expected death group. The outcome measure was existing PUs (Categories 1–4), or not, during the last week in life. The logistic regression analyses were theory-driven and in the first step, we entered type of healthcare unit into the model. Next, we entered age and gender, followed by medical conditions. The last step in the regression analysis was to add symptoms, length of stay, and the treatment variable into the model.

Spearman's rho was performed to assess the correlation between existing PU (Categories 0–4) and symptom control (0 = none at all; 1 = partial; 2 = complete). A p-value of 0.05 was considered significant. We used SPSS version 22 to analyze the data (SPSS, Inc., Chicago, IL).

Ethical considerations

The study was approved by the Ethics Review Board in Uppsala, Sweden (Dnr 2014/357/1). The principles set out in the Declaration of Helsinki as well as the national and local guidelines for research were followed.¹⁴ As this was a quality register study, it was not possible to obtain informed consent from the patients. However, all patients had been informed about the national quality register, as well as about voluntary participation and confidentiality, as part of routine healthcare. Under current legislation, all research projects approved by the ethics review board can access the variables in a quality registry that are related to the topic, provided there are no confidentiality issues, in accordance with the Public Access and Secrecy Act (2009:400).¹⁴ The study was approved by the register management group.

Role of the funding source

The funders of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. The authors had full access to all the data in the study, and the corresponding author had final responsibility for the decision to submit for publication.

Results

The mean age in this population was 81.7 (18–113) years. Patient characteristics are detailed in Table 1. Cancer and heart disease were the most common diagnoses. Nearly 19% of the deaths in the heart disease group were unexpected in contrast to not quite 4% in the cancer group. Most deaths occurred in hospitals or nursing homes. As previously mentioned, those whose death occurred in “other” locations have been excluded from further analysis.

The prevalence of PUs at admission varied between 6.9% (nursing homes) and 19.0% (specialized PC inpatient units) (Table 2). The highest prevalence of PUs at death was seen in specialized PC inpatient units (29.7%). Symptoms such as pain, anxiety, and death rattles were common, and intravenous/enteral feeding was most frequently used in hospitals. The mean length of stay varied between the different care units as follows: from 18 days in hospitals and 25 days at specialized PC inpatient units to 77 days for what was referred to as a short stay at nursing homes, 150 days with specialized PC home care, 726 days with general palliative home care, and 1018 days at nursing homes.

Table 2.

Pressure Ulcers and Symptoms at Different Healthcare Units

	Nursing home	Short stay at nursing home %	Specialized PC inpatient unit %	Hospital %	General PC home care %	Specialized PC home care %
PU at admission	6.9	16.2	19.0	13.8	11.0	10.2
Category 1	2.7	6.8	8.6	6.4	5.6	4.6
Category 2	1.9	4.8	5.4	4.0	2.8	3.3
Category 3	1.1	2.4	2.5	2.0	1.5	1.3
Category 4	1.2	2.2	1.6	1.5	1.1	1.1
PU at death	16.8	20.8	29.7	19.6	18.6	23.5
Category 1	7.8	9.0	13.3	10.0	9.4	11.4
Category 2	4.0	6.3	10.2	5.7	4.6	7.0
Category 3	2.5	2.9	3.8	2.3	2.4	3.1
Category 4	2.4	2.7	2.4	1.6	2.2	2.0
Symptoms
Pain	70.8	78.1	83.6	68.2	75.8	78.8
Confusion	25.9	28.2	29.4	31.7	22.6	27.9
Nausea	11.5	19.5	19.5	16.0	22.1	22.3
Shortness of breath	15.4	17.8	28.3	42.1	17.5	24.4
Anxiety	44.3	54.5	59.9	50.0	49.6	52.4
Rattling	50.0	47.4	54.0	58.3	48.3	50.9
Intravenous/enteral feeding	2.1	3.7	8.7	41.0	2.6	7.2

PC, palliative care; PU, pressure ulcer.

In the first step of the model (Table 3), all health units in the variable place of death had significantly more PUs than did the nursing homes (the reference value). When adding age and gender into the model (step 2), higher age was a predictor for PU prevalence. In step 3, when adding medical conditions into the model, the only change in prevalence of PUs, related to the place of death, was that general palliative home care was no longer significantly related to the development of PUs. The medical conditions that predicted higher risk for PUs were cancer, other neurological diseases, diabetes, postfracture state, multiple diseases, and infection. In the fourth step, when adding length of stay, symptoms, time of loss of decision-making capacity, and intravenous/enteral feeding into the model, cancer, infections, and other neurological diseases were no longer significantly associated with PU development. Pain was the only symptom associated with higher likelihood, while nausea was associated with less likelihood of PUs. Patients who lost their decision-making capacity weeks before death were at higher risk of developing PUs compared with patients who were decision-competent until death. Intravenous drip or enteral feeding was associated with a significantly decreased likelihood of developing PUs.

Table 3.

Risk of Pressure Ulcer at the End of Life (Expected Death Patients), Expressed in Odds Ratio with 95% Confidence Intervals

Category	Model 1 ^a OR (CI)	Model 2 ^b OR (CI)	Model 3 ^c OR (CI)	Model 4 ^d OR (CI)
Place of death
Nursing home (ref)
Hospital	1.201 (1.142–1.236)^***	1.227 (1.164–1.294)^***	1.125 (1.060–1.193)^***	1.278 (1.087–1.503)^**
Specialized PC inpatient units	2.090 (1.957–2.234)^***	2.175 (2.023–2.338)^***	2.010 (1.851–2.183)^***	2.088 (1.701–0.264)^***
Short stay at nursing homes	1.298 (1.197–1.407)^***	1.314 (1.211–1.426)^***	1.208 (1.109–1.316)^***	1.251 (1.034–1.513)^*
General palliative home care	1.129 (1.026–1.243)^*	1.145 (1.039–1.261)^**	1.055 (0.954–1.166)	1.106 (0.888–1.376)
Specialized PC home care	1.517 (1.385–1.662)^***	1.584 (1.438–1.744)^***	1.613 (1.313–1.982)^***	1.450 (1.143–1.838)^**
Age		1.003 (1.001–1.005)^**	1.005 (1.001–1.009)^*	1.004 (0.999–1.009)
Sex
Male		0.998 (0.957–1.040)	0.986 (0.919–1.068)	1.007 (0.910–1.115)
Female (ref)
Medical conditions
Cancer			1.161 (1.096–1.230)^***	1.102 (0.954–1.272)
Heart disease			0.944 (0.898–0.993)^*	0.990 (0.873–1.124)
Respiratory disease			1.015 (0.946–1.089)	1.074 (0.900–1.281)
Dementia			0.776 (0.727–0.828)^***	0.800 (0.676–0.947)^**
Stroke			0.846 (0.786–0.910)^***	0.841 (0.700–1.011)
Other neurological disease			1.161 (1.038–1.299)^**	1.206 (0.908–1.602)
Diabetes			1.447 (1.318–1.588)^***	1.318 (1.045–1.662)^*
Postfracture state			1.513 (1.351–1.694)^***	1.606 (1.201–2.148)^**
Multiple diseases			1.429 (1.358–1.504)^***	1.322 (1.163–1.502)^***
Infection			1.461 (1.314–1.624)^***	1.377 (1.027–1.848)^*
Length of stay, days
0–30				1.038 (0.923–1.167)
31–90				0.879 (0.741–1.087)
91–360				1.065 (0.888–1.276)
361–3664 (ref)
Symptoms
Pain				1.410 (1.245–1.597)^***
Confusion				1.111 (0.990–1.247)
Nausea				0.859 (0.745–0.992)^*
Shortness of breath				0.956 (0.843–1.084)
Anxiety				0.988 (0.888–1.100)
Death rattles				1.080 (0.975–1.197)
Lost decision-making capacity
Never (ref)
Hours before death				1.104 (0.915–1.331)
Days before death				1.189 (0.995–1.421)
Weeks before death				1.407 (1.118–1.770)^**
Months or more before death				1.010 (0.790–1.291)
Intravenous/enteral feeding
No (ref)
Yes				0.829 (0.703–0.976)^*

Adjustment for place of death.

Adjustment for place of death, age, and sex.

Adjustment for place of death, age, sex, and medical condition.

Adjustment for place of death, age, sex, medical condition, length of stay, symptoms, lost decision-making capacity, and intravenous/enteral feeding.

p < 0.05.

p < 0.01.

***

p < 0.001.

CI, confidence intervals; OR, odds ratio.

The only significant negative correlation between PU (Categories 0–4) and symptom control (Table 4) involved pain control as follows: the better the pain control the lesser the PU development (and the lower the PU category) and vice versa. Relief of nausea was positively significant with PU prevalence.

Table 4.

Correlation between Symptom Control and Pressure Ulcers, Using Spearman's Rho

Variable	r_xy	p
Symptom control^a
Pain	−0.047	<0.001
Confusion	−0.011	NS
Anxiety	0.00	NS
Death rattles	−0.010	NS
Shortness of breath	−0.06	NS
Nausea	0.29	<0.05

0 = none at all; 1 = partial; 2 = complete.

NS, not significant.

Discussion

The data from the SRPC used in this study cannot prove causation, but they can report correlations between background variables and PU prevalence. The analysis gave no clear answers regarding whether the large differences in PU prevalence at death between healthcare units were due to differences in the quality of care or to differences in risk factors related to PU. Compared with nursing homes, the risks for PUs at death were higher in all other healthcare units, except general palliative home care. Diagnosis did not explain the big difference in PU prevalence between nursing homes and specialized PC inpatient units. In the full model, cancer, which was the main diagnosis in specialized PC inpatient units, no longer predicted risk for developing PUs. On the contrary, one of the most common diagnoses in nursing homes in Sweden, dementia, had the lowest prediction for developing PUs. That the well-known risk factors for PUs in general, diabetes, postfracture state, infections, and multiple sickness, were predictors of PU development even in dying patients was expected.¹

An important consideration is that quite a large group of patients already had PUs at admission to the reported unit. A comparison of development of PUs from admission to death is not appropriate in this study because of the large differences, between the different healthcare units, in the length of stay. Many PUs present at admission to a nursing home may have time to heal before death, but the same is hardly possible in the hospital setting. Consequently, in our study, many of the PUs present at death in healthcare units with short length of stay were the same as at admission. These factors are important to take into consideration when PU prevalence is used as a quality indicator for good PC.

The strong correlation between pain and PUs was expected,¹¹ even though this study cannot clarify the direction of the correlation, that is, determine whether the patients with PUs had more pain or whether more pain is a risk factor for PU prevalence as patients with severe pain are more immobile and sometimes reluctant to reposition. It is important that patients receive analgesics before scheduled position change; often, micro-repositioning is sufficient to reduce pressure.¹⁶

The negative correlation we found between PUs and pain control gave an indication that the pain caused by higher categories of PUs is more difficult to relieve. That shortness of breath was not a predictor for PUs was surprising as short-of-breath patients often have the head of the bed elevated, which causes shear and pressure on the sacrum.¹ Elevating the head of the bed by more than 30° means that a PU may be unavoidable.⁵ This result is not in accordance with Brink et al.,¹⁷ who report an association between shortness of breath and PU in a palliative home care setting.

The results that patients with nausea had a decreased risk for PUs and that treatment for nausea was positively correlated with PU prevalence are difficult to explain. The most plausible explanation is that this was a false correlation.

That intravenous drip or enteral feeding in late palliative phase significantly decreased the likelihood of developing a PU was a little surprising. A common view is that in the dying phase, intravenous drip or enteral feeding can do more harm than good. However, in the present study, quite a few patients had intravenous or enteral feeding in the last week of life.

Many of the conditions for unavoidable PUs, according to the National Pressure Ulcer Advisory Panel consensus statements,⁵ are common in dying patients. Terminally ill patients who become immobile, malnourished, and/or cachexic are at increased risk for unavoidable PUs. Whether dying per se is a high risk factor for PUs is not possible to clarify in this study, as all our participants had died. Also, there was a lack of data on preventive interventions.

The Swedish national quality registries should be and are increasingly being used for research that could improve the quality of care. The registries are also an important source for generating new research questions. The SRPC is a “living” register, which means that temporary questions can be added to the questionnaire for a short period.⁹ Process indicators about preventive interventions for PUs should be included, both for quality evaluation and for research reasons. Some research has been undertaken to validate different aspects of the end-of-life questionnaire,^10,18,19 but more studies are needed to guarantee the quality of the data, especially for scientific purposes.

Strengths and limitations of the study

The main limitation of this study is that the questions in the end-of-life questionnaire were not specifically designed to answer our research questions. Moreover, all data were retrospectively reported by staff. The study's strengths are that the SRPC covers the majority of deaths across Sweden and all types of care facilities engaging in end-of-life care in Sweden. The large number of included patients is a further strength.

The question concerning “conditions that caused the death” was difficult for the staff to answer, which became apparent when we scrutinized the comments connected to the variable “other.” A large number of comments were added, even where appropriate categories existed. In some cases, no suitable category was available, for example, for kidney or liver disease or for bleeding. Another category that was missing was an infection category; 2198 comments were about different forms of infection. A plausible explanation as to why so many filled in different forms of infections is that this category existed in a former SRPC end-of-life questionnaire. The decision to add infections as a category was motivated by a possible link to PUs as well as by the large number of infection-related comments. However, the data on infections are probably underestimated. Processing the data in this way may be questioned, but we argue that the advantages of doing so outweigh the disadvantages. It is obvious that better instructions are needed for this question and, furthermore, that more categories are needed to prevent respondents from giving unnecessarily long responses.

Conclusion

The well-known predictors of PUs in general, diabetes, postfracture state, infections, and multiple diseases, are predictors even in dying patients. Dementia was significantly associated with lower likelihood of PUs, and pain was associated with more PUs. Intravenous drip or enteral feeding was associated with a significantly decreased likelihood of developing PUs. The Swedish national quality registers could be a unique resource for quality improvement and research.

Footnotes

Acknowledgments

The authors thank Greger Fransson for help with the register data and Ronnie Pingel for help with the statistics. They also thank the National Board of Health and Welfare and the Freemasonry organization for financial support.

Author Disclosure Statement

No competing financial interests exist.

References

National Pressure Ulcer Advisory Panel, European Pressure Ulcer Advisory Panel and Pan Pacific Pressure Injury Alliance: Prevention and Treatment of Pressure Ulcers: Clinical Practice Guideline. Haesler

(ed). Perth, Australia: Cambridge Media, 2014.

White-Chu

, Reddy

: Pressure ulcer prevention in patients with advanced illness. Curr Opin Support Palliat Care, 2013; 7:111–115.

Sibbald

, Krasner

, Lutz

, et al.: Scale: skin changes at life's end. Final consensus statement. Adv Skin Wound Care, 2010; 23:225–236.

Socialstyrelsen: Nationellt kunskapsstöd för god palliativ vård. Rekomendationer. www.socialstyrelsen.se/publikationer 2012/2012-6-17 (The National Board of Health and Welfare. Guidelines for palliative care. Recommendations) (last accessed November 28, 2015).

Edsberg

, Langemo

, Baharestani

, et al.: Unavoidable pressure injury. State of the science and consensus outcomes. J Wound Ostomy Continence Nurs, 2014; 41:1–22.

The Swedish Register of Palliative Care (SRPC): www.palliativ.se (last accessed September 16, 2016).

The National Quality Registries in Sweden: www.kvalitetsregister.se (last accessed September 16, 2016).

Lundström

, Axelsson

, Hedman

P-A

, et al.: Developing a national quality register in end-of-life care: The Swedish experience. Palliat Med, 2012; 26:313–321.

The Swedish Register of Palliative Care: Annual report from the Swedish Register of Palliative Care fiscal year 2014. http://media.palliativ.se/2015/08/Svenska-palliativregistret-årsrapport-2015_2.pdf (last accessed September 16, 2016).

10.

Martinsson

, Fürst

, Lundström

, et al.: Registration in a quality register: A method to improve end-of-life care-a cross-sectional study. BMJ Open, 2012; 2:e001328.

11.

Gorecki

, Brown

, Nelson

, et al.; European Quality of Life PU Project group: Impact of PUs on quality of life in older patients: A systematic review. J Am Geriatr Soc, 2009; 57:1175–1183.

12.

Searle

, McInerney

: Nurses' decision-making in pressure area management in the last 48 hours of life. Int J Palliat Nurs, 2008; 14:432–438.

13.

WHO: National cancer control programmes: Policies and managerial Guidelines. Geneva, Switzerland: World Health Organisation, 2002.

14.

Codex 2015: Rules and guidelines for research. http://www.codex.uu.se (last accessed November 28, 2015).

15.

http://www.kvalitetsregister.se/sekundarnavigering/inenglish/guidanceondisclosureofregistrydata/datadisclosedforresearch.1417.htlm (last accessed September 16, 2016).

16.

Gunningberg

, Carli

: Reduced pressure for fewer pressure ulcers: Can real-time feedback of interface pressure optimize repositioning in bed?. Int Wound J, 2014 [Epub ahead of print]; DOI: 10.1111/iwj.12374.

17.

Brink

, Smith

, Linkewich

: Factors associated with pressure ulcers in palliative home care. J Palliat Med, 2006; 9:1369–1375.

18.

Ghohlia

, Fransson

, Furst

, Heedman

, et al.: Stora luckor i journaler vid vård i livets slutskede. Journaluppgifter saknas för data inrapporterade till palliativ registret. Läkartidningen, 2011; 108:918–921. (In Swedish).

19.

Nilsson

, Tavelin

, Axelsson

: A Study of patients not registered in the Swedish cancer register but reported to the Swedish Register of Palliative Care 2009 as deceased due to cancer. Acta Oncol, 2014; 53:414–419.