Pressure injury data in Australian acute care settings: A comparison of three data sets

Abstract

Hospital‐acquired pressure injuries (HAPIs) represent a serious clinical and economic problem. The cost of treating HAPIs in Australian public hospitals was recently reported at AUS$983 million per annum. There are three main sources of data for documenting pressure injury (PI) occurrence in Australian hospitals: incident reporting, medical record coded data, and real‐time surveys of pressure injury. PI data reported at hospital level and to external agencies using these three different sources are variable. This reporting issue leads to inaccurate data interpretation and hinders improvement in accuracy of PI identification and PI prevention. This study involved a comparison of the three different data sources in selected Australian hospitals, to improve the accuracy and comparability of data. Findings from this study provide benchmark areas for improvement in PI documenting and reporting. Better understanding the agreement between the three data sets could lead to a more efficient and effective sharing of data sources.

1. INTRODUCTION

1.1. Background

Pressure injury (PI) is defined as a “localised injury to the skin and/or underlying tissue usually over a bony prominence, because of pressure, or pressure in combination with shear and/or friction.”1 A hospital‐acquired pressure injury (HAPI) is a localised injury to the skin and/or underlying tissue that develops during an inpatient hospital stay.2 However, there is no agreement on the timeframe (8, 24, or 48 hours) from admission to determine if PI was acquired in care—this needs further research and agreement.3, 4 According to the National Pressure Ulcer Advisory Panel (NPUAP), the European Pressure Advisory Panel (EPUAP) and Pan Pacific Pressure Injury Alliance (PPPIA),5 there are four stages of PIs, including stage I—non‐blanchable erythema, stage II—partial thickness skin loss, stage III—full thickness skin loss, and stage IV—partial thickness tissue loss. When the true depth of the injury is unknown, PI is classified as either unstageable PI eschar or slough obscuring the assessor's ability to view and therefore stage the true due or suspected deep tissue injury (SDTI) that is defined as a localised area of discoloured purple or maroon skin.

The most common sites of PI are the sacrum/coccyx (20%‐41%) and heels (16%‐27%).6, 7 The independent predictors of PI development are (a) mobility/activity; (b) peripheral perfusion, diabetes; and (c) skin/pressure injury status.8 These predictors vary in some groups, for example, in critical care patients, the main predictors were: age, mobility/activity, perfusion, and vasopressor infusion.9 The development of HAPI in critically ill and palliative patients may also be related to skin failure secondary to hypoperfusion.10, 11

The consequences of HAPI range from skin redness to full skin and tissue loss, exposing tendons and bones.12, 13 If a person's injury advances, they face the prospect of osteomyelitis14, 15 and sepsis.16, 17 Issues such as pain, immobility, lack of independence, wound exudate, and odour impact patient's self‐image and emotional state, influence a person's quality of life.18, 19, 20, 21, 22

In addition to individual patient impact, PI is an economic burden to Australian health care system23, 24 and exert high economic burden to health care systems worldwide25, 26, 27, 28 that is associated with increasing length of hospital stay, noting also that there is a vicious circle linking increased hospital stay with acquiring PI.29, 30, 31, 32 The cost of treating PIs in Australian public hospitals was recently reported at AUS$983 million per annum (approximately 1.9% of all public hospital expenditure).23 The associated opportunity cost, bed days lost, was valued at AUS$820 million per annum. Severe cases of PI (stages III and IV) represented 30% of the total costs.23 It is difficult to comment on less severe PI stages (unstageable and stage I) as the evidence for this group is lacking. Contemporary Australian data on PI prevention and management became a priority to justify the allocation of funds in this field.33 There is a consensus that PI is largely preventable,23, 34 particularly if multiple intervention programmes are implemented in acute care settings.35 Harm from PI is a high risk on most hospital risk registers.

The three main sources of data for PI identification include: (a) incident reporting system, (b) administrative/coded data extracted from documentation in the patient's medical record, and (c) data generated from real time surveys, such as Pressure Ulcer/Injury Point Prevalence Surveys (PUPPS/PIPPS). The extent to that prevalence is documented in each of these three data sources are unknown. Given the human and financial burden of HAPIs and the imperative to have reliable documentation of HAPIs to support and monitor interventions, the aim of this project was to assess agreement between the three methods of reporting PIs in an acute hospital setting and determine factors associated with reporting of HAPIs. Our primary objective was to compare stage 2 or higher PIs captured via a point prevalence study (gold standard) with that captured through administrative data (the International Classification of Diseases 10th edition‐Australian modified [ICD‐10‐AM] coding system completed on discharge) and the hospital's voluntary incidence reporting system. Our secondary objective was to assess overall PI prevalence and stage; including whether noted on admission or acquired in an acute hospital setting.

2. METHODS

2.1. Study design

This retrospective cohort study was conducted in a metropolitan health service in Melbourne, Alfred Health, that comprises three hospitals. All inpatient cases in the recruiting hospital on the date, when the PUPPS was undertaken, were included in the study. Data collected for the annual PUPPS/PIPPS for the 3 years, 2015 to 2017, were used as the benchmark for comparison. These data were collected on a single day and the retrospective data for the previous 2 years were used to assess the comparability, accuracy, and reliability of using each data set independently or conjointly.

2.2. Study setting

Alfred Health is one of Australia's largest health services, which has 846 multi‐day‐inpatient‐beds (based on budgeted multiday bed numbers), including a 43‐bed intensive care unit, admitting more than 3000 patients per year, and a large emergency department, which receives approximately 66 000 emergency presentations per year.36 Alfred Health is a core member of Monash Partners.37 Monash Partners is the Commonwealth‐recognised accredited academic health sciences centre that includes seven independent providers of health services, health research, and health education. Monash Partners provides an opportunity to work collaboratively in designing, developing, implementing, evaluating, refining, and ultimately scaling‐up novel evidence‐based programmes to drive direct improvement in health care outcomes,37 including in wound care. The findings of this study will be used for co‐designing a large scope research project on PI surveillance and reporting across Monash Partners core members with the ultimate goal of establishing a uniform PI reporting system across partnering organisations and nationally.

2.3. Data sources

Data were sourced from three hospital data sets for the 3 years 2015 to 2017.

2.3.1. Point prevalence study (PUPPS/PIPPS) data

PUPPS/PIPPS is an audit of PIs in hospitalised patients and provides a snapshot of the prevalence of PIs on a given day. PUPPS/PIPPS capture the presence of PI, the patient's PI risk and appropriateness of prevention strategies in place. PUPPS/PIPPS data collectors are appointed across the health service and are usually Division 1 registered nurses with the support of allied health. Prior to the survey day, data collectors receive education on PI recognition and classification. PUPPS/PIPPS data are captured on either paper or electronically, which generates the annual PUPPS/PIPPS database. Initially established as a state‐wide point prevalence study, since 2006 it has been undertaken voluntarily by some Victorian hospitals, without discrete funding. Some health services include a skin survey of other hospital acquired skin injuries, such as skin tears and incontinence‐associated dermatitis (IAD).

2.3.2. Incident reporting data

Any patient who is admitted with, or who develops a HAPI in hospital at stage II or greater, should have the injury recorded in the Victorian Hospital Incident Management System (VHIMS)38; some health services also report stage I. Risk Management (RiskMan) (http://www.riskman.net.au/) is the VHIMS software used by Australian hospitals to report PI. Incident reports may be lodged by any health care worker, but traditionally PIs are reported into the VHIMS by nursing staff.39 Personnel completing the incident report in RiskMan are required to document the type, and severity of injury, and a summary of the incident including information on the currency of the risk assessment, and PI prevention plan. RiskMan allows the following PI documentation: (a) PI stage, (b) PI location, (c) admission source, and (d) present on arrival/acquired/worsened. RiskMan is used by the local clinical area to explore opportunities for learning and improvement.

2.3.3. Administrative coding data

Each inpatient has their inpatient PI episode coded by qualified clinical coders within Health Informatics Services.40 Coders use the International Classification of Diseases 10th edition‐Australian modified (ICD‐10‐AM/ACHI/ACS)41 coding system. Patients admitted during the time period of the point prevalence study coded with an ICD‐10‐AM code of L89 (including any sub‐categories) as either a primary or a secondary diagnosis during the time period were extracted. The ICD‐10‐AM/ACHI/ACS codes are classified as follows: L89.1 stage II decubitus ulcer and pressure area, L89.2 stage III decubitus ulcer and pressure area, L89.3 stage IV decubitus ulcer and pressure area, and L89.9 decubitus ulcer and pressure area, unspecified.41

2.3.4. Data analysis

Using the PUPPS report taken on the single day (gold standard), unit record numbers were obtained for mapping against the incident reporting data set and the administrative data set. Day‐stay areas were excluded from PUPPS. Identifiable information was removed following data linkage of patients across the three data sets. Descriptive characteristics used numbers and percentages for categorical data and means, and standard deviations for normally distributed continuous data or medians with interquartile range for skewed continuous data. The binomial proportion of PIs was estimated using the lower and upper exact binomial confidence limit (95% confidence interval [CI]). True positive fraction of ICD‐10‐AM codes and incident reports for PI, greater than stage II, was calculated using PUPPS as the gold standard. The effect size representing the degree of difference in the reporting of PI was calculated as the odds ratio (OR), where an OR greater than one represented more likelihood of a particular factor in the comparison data set (data set 1 or data set 2) compared with the PUPPS (gold standard) data set. Agreement was determined using kappa statistics,42 where a kappa in the range of 0.21 to 0.40 was considered fair agreement, 0.41 to 0.60 – moderate agreement, 0.61 to 0.80 – substantial agreement, and kappa >0.81 – almost perfect agreement.43

3. RESULTS

Table 1 provides a summary of the characteristics of patients included in the study from the three contributing hospitals. PI data was extracted from 2302 cases reported in PUPPS conducted in 2015, 2016, and 2017.

Table 1.

Details of cases with pressure injuries as recorded in the Pressure Ulcer Point Prevalence Survey (PUPPS) 2015 to 2017^{a , b} (PUPPS data was used as the gold standard)

Characteristic	Year = 2015	Year = 2016	Year = 2017
Number of cases	752	760	790
Sex
Female	324 (43.1)	366 (48.2)	342 (43.3)
Male	428 (56.9)	394 (51.8)	448 (56.7)
Age (years)	65.1 ± 19.4	64.9 ± 20.0	63.7 ± 19.6
Hospital
Hospital 1	249 (33.1)	252 (33.2)	255 (32.3)
Hospital 2	42 (5.6)	52 (6.9)	34 (4.3)
Hospital 3	461 (61.3)	455 (59.9)	501 (63.4)
LOS (days)	14.0 [6.0–29.0]	15.0 [7.0–33.0]	14.0 [7.0–31.0]
LOS, exclude Hospital 1 (days)	9.0 [4.0–19.0]	10.0 [5.0–22.0]	11.0 [5.0–21.0]

^aData presented as mean ± SD, median, and [interquartile range] and n (%).

^bDay‐stay areas were excluded from PUPPS.

As captured in PUPPS, there were 342 (67%) cases with more than one PI, and 175 (35%) PIs were recorded on admission (Table 2). Of the 301 cases of PIs (excluded stage I) recorded in PUPPS, 182 (61%) cases had stage II PI, 19 (6%)—stage III, 2 (<1%)—stage IV, 38 (13%)—SDTI, 59 (20%)—unstageable, and 1 (<1%)—unspecified. Greater proportion of PIs were located on the lower back, including sacrum (N = 111; 37%), and heel (N = 58; 19%). Of the 301 cases of PIs (excluded stage I) recorded in PUPPS, there were 109 (36.2%, 95% CI: 30.8‐41.9) recorded in RiskMan and 158 (52.5%, 95% CI: 46.7‐58.3) coded in the medical records. Of the 123 cases of PIs (excluded stage I) present on admission, 32 (26.0%, 95% CI: 18.5%‐34.7%) were recorded in RiskMan and 55 coded in the medical records (44.7%, 95% CI: 35.7%‐53.9%) (Table 2).

Table 2.

Agreement between pressure injuries (PI) reporting in Pressure Ulcer Point Prevalence Survey (PUPPS) (gold standard) and incident reporting data set (Risk Management [RiskMan]), and PIs reporting in PUPPS and administrative coding data set

	PUPPS data set (gold standard)	Incident reporting data set (RiskMan) TPF	Binomial exact 95% CI	Administrative coding data set TPF	Binomial exact 95% CI
Number of PIa	301	109 (36.2)	30.8–41.9	158 (52.5)	46.7–58.3
Cases with ≥1 PI	211	81 (38.4)	31.8–45.3	111 (52.1)	45.6–59.5
PI on admission	123	32 (26.0)	18.5–34.7	55 (44.7)	35.7–53.9
Site
Head	20	3 (15.0)	3.2–37.9	2 (10.0)	1.2–31.7
Upper extremity	12	3 (25.0)	5.5–57.2	3 (25.0)	5.5–57.2
Upper back	2	1 (50.0)	1.3–98.7	0 (0)
Lower back (includes sacrum)	111	32 (28.8)	20.6–38.2	47 (42.3)	33.0–52.1
Ischium/buttocks	24	8 (33.3)	15.6–55.3	8 (33.3)	15.6–55.3
Trochanter	1	0 (0)		1 (100)
Heel	58	16 (27.6)	16.7–40.9	27 (46.6)	33.3–60.1
Toe	16	2 (12.5)	1.6–38.3	3 (18.8)	4.0–45.6
Other site of lower extremity	42	4 (9.5)	2.7–22.6	9 (21.4)	10.3–36.8
Site unspecified	15	1 (6.7)	0.2–31.9	2 (13.3)	1.7–40.5

Abbreviations: CI, confidence interval.

^aStaging not performed as time varying factor.

Table 3 provides an overview of characteristics associated with reporting PI. PI was more likely to be reported in RiskMan if the patient was male (OR = 1.57; 95% CI: 0.98‐2.53), or if the PI was located on the heel (OR = 2.18; 95% CI: 1.07‐4.44), or sacrum/hip/ischium/trochanter (OR = 2.15; 95% CI: 1.20‐3.87). PI location on sacrum/hip/ischium/trochanter (OR 3.0, 95% CI: 1.56–5.76) was associated with a significantly higher reporting of the PUPPS PI in the medical records.

Table 3.

The effect size representing the degree of difference in the pressure injuries (PI) reporting (excluded stage I) was calculated as the odds ratio (OR)

Characteristic	PI reporting in PUPPS and incident reporting data set (RiskMan)		PI reporting in PUPPS and administrative coding data set
	OR (95% CI)	P‐value	OR (95% CI)	P‐value
Age	0.99 (0.98–1.01)	.26	1.01 (1.00–1.02)	.26
Male	1.57 (0.98–2.53)	.06	0.85 (0.54–1.35)	.50
LOS	1.00 (0.995–1.0)	.77	1.01 (1.00–1.02)	.002
Acquired	0.92 (0.57–1.48)	.72	0.78 (0.49–1.24)	.30
Stage of PI (PUPPS)
II	1		1
III	0.63 (0.24–1.85)	.44	0.79 (0.31–2.03)	.62
IV	—		—
SDTI	0.33 (0.14–0.79)	.01	0.71 (0.35–1.43)	.34
Unstageable	0.87 (0.47–1.59)	.65	1.11 (0.62–2.01)	.72
Site of PI (PUPPS)
Limb	1		1
Heel	2.18 (1.07–4.44)	.03	1.92 (0.88–4.15)	.10
Sacrum/hip/ischium/trochanter	2.15 (1.20–3.87)	.01	3.0 (1.56–5.76)	.001
Other	0.37 (0.15–0.92)	.03	0.93 (0.36–2.43)	.88

Note: For example, people with PI were significantly more likely to be male (OR of 1.57 [95% confidence interval (CI); 0.98‐2.53]) in incident reporting data set (Risk Management [RiskMan]) compared with the Pressure Ulcer Point Prevalence Survey (PUPPS) (gold standard) data set.

Kappa statistics to determine agreement between PIs reported in PUPPS and the incident reporting and administrative data sets are presented in Table 4. Cohen/Conger's Kappa and Fleiss' Kappa statistics indicated only moderate agreement between the gold standard and the incident reports 0.50 and 0.49, respectively for agreement between PUPPS and incident reports. Slightly better agreement, that is, substantial agreement was evident between the PUPPS survey (gold standard) and the administrative data coding 0.66 (Cohen/Conger's Kappa) and 0.66 (Fleiss' Kappa). The high level of agreement according to the Gwet's AS reflects a strong level of agreement for the absence of PIs.

Table 4.

Kappa statistics to determine agreement between pressure injuries (PIs) reporting in Pressure Ulcer Point Prevalence Survey (PUPPS) and incident reporting data set (Risk Management [RiskMan]), and PIs reporting in PUPPS and administrative coding data set based on review of 2356 cases (2 ratings per subject, 2 rating categories)

Interrater agreement	Agreement between PIs reporting in PUPPS and incident reporting data set (RiskMan)					Agreement between PIs reporting in PUPPS and administrative coding data set
	Coef.	SE	t	P > t	95% CI	Coef.	SE	t	P > t	95% CI
Percent agreement	0.92	0.01	162.92	<.001	0.91–0.93	0.94	0.01	190.91	<.001	0.93–0.95
Brennan and Prediger	0.84	0.01	74.23	<.001	0.82–0.86	0.88	0.01	89.28	<.001	0.86–0.90
Cohen/Conger's kappa	0.50	0.03	16.56	<.001	0.44–0.56	0.66	0.03	25.30	<.001	0.61–0.71
Fleiss' kappa	0.49	0.03	15.30	<.001	0.43–0.55	0.66	0.03	24.43	<.001	0.60–0.71
Gwet's AC	0.90	0.01	124.88	<.001	0.89–0.92	0.93	0.01	146.64	<.001	0.91–0.94
Krippendorff's alpha	0.49	0.03	15.31	<.001	0.43–0.55	0.66	0.03	24.43	<.001	0.60–0.71

Abbreviations: CI, confidence interval.

4. DISCUSSION

The aim of this data analysis research was to compare the agreement between three data sources reporting PI in inpatient samples. The findings of this study indicate that approximately half of the PI (excluded stage I) captured in PUPPS were coded in the medical records and approximately one third of them were recorded in RiskMan. There was a lesser agreement for PI present on admission. PIs located on heel, sacrum, hip, ischium, and trochanter were better reported in RiskMan and coded in the medical records. This difference in reporting has previously been reported in the literature, and can be explained by the fact that these sites of PIs are commonly known locations while PIs located behind the ear, particularly medical device‐related, are less common.44 Clinician education may increase awareness of device‐related PIs and expand their skills on PI surveillance and prevention.45 Gender‐specific differences in PI reporting warrant further research and gender‐targeted interventions.

The worldwide PI prevalence rate reported in acute care settings range from 6.0% to 18.5%46 and are higher in intensive care patients 11.5% to 32.7%.7 In public hospitals, the global point prevalence of PI was 14.8%, period prevalence – of 11.6%, and a mean incidence of 6.3%.47 In Australia, these PI indicators are lower. In Queensland, the findings from the secondary data analysis of the Queensland Bedside Audits 2012 to 2014 data indicate that overall HAPI prevalence from 2012 to 2014 was 11% for intensive care patients and 3.0% for non‐intensive.48 The latest 2017 audit in New South Wales indicated that HAPI prevalence was 4.0%.49 When comparing data sources, it is often difficult to determine the gold standard because all have their flaws. However, the strength of this study is that we were able to use as a gold standard the high‐quality point prevalence study because of the rigour by which the audit was undertaken. We chose to show stage 2 or higher PIs because these should be clearly documented and visible to nursing staff. The main limitations were related to PI coding, that is, PI captured in PUPPS may not be the same as documented in RiskMan, as stage I PIs are not included in RiskMan. There is a chance of underreporting PI on discharge due limited number of the documented entries.

Our findings indicate that improvement is required to optimise staff reporting of PI into the incident reporting system. The patient record should be the most accurate record of the care provided to the patient. Support for clinical staff to more accurately record and code PI, therefore, enables improved benchmarking. Acute care services need to improve capacity to collect high quality PI incidence data over time, thereby facilitating capacity to benchmark performance with other hospitals and reduce incidence of preventable HAPI.50 Quality improvement activities used by acute care settings that rely on quality data help to ensure better health outcomes at lower cost.51 Best practice standards could be translated into routine patient care in specific clinical areas in a short period of time, if multihospital, data‐driven collaborations were established.52 Collaboration helps to ensure that uniform data collection methods are used, complete real time data are compiled and a uniform data platform is used by collaborating hospitals.51, 53

An example of multihospital collaborations and data‐driven quality improvement in wound care is the Swedish national quality registry of ulcer treatment (RUT).54 RUT originated from registrations started as Blekinge Wound Healing Centre in 2003 and over time spread throughout the country, which resulted in a significant reduction in wound healing time.54 Focus on uniform risk assessment tools, harmonised variables, uniform healing outcomes, standardised statistical information, and validation based upon experiences from RUT could contribute to the development of national wound registry in Australia and an international registry.54 Comprehensive computer‐generated data aggregated in a wound registry will facilitate rapid identification of slow healing wounds55 and enable clinicians to guide clinical decision‐making and provide optimised care based upon these data.56 Currently, reliable data on the prevalence and costs of chronic wounds, including PIs, in Australia is limited.50 Strengthening data collection on chronic wounds, and building surveillance, and providing support to a national wound prevalence survey were identified as a priority action in wound care reform in Australia.50

To facilitate future data‐driven quality improvement interventions on PI surveillance and prevention we recommend further research to scope and capture data harmonisation activities across partnering health services at Monash Partners, as well as on state and national levels. This recommendation is well aligned with the aims of the Australian Health Research Alliance57 data‐driven healthcare improvement initiative and global initiatives,58 including in the field of PI prevention.27, 59

CONFLICT OF INTEREST

The authors declare no potential conflict of interest.

AUTHOR CONTRIBUTIONS

All persons designated as authors qualify for authorship. Each author have participated sufficiently in the work to take public responsibility for the content.

Team V, Tuck M, Reeves J, et al. Pressure injury data in Australian acute care settings: A comparison of three data sets. Int Wound J. 2020;17:578–586. 10.1111/iwj.13320