People admitted to hospital with physical disability have increased length of stay: implications for diagnosis related group re-imbursement in England
1 University of Kent, Centre for Health Services Studies, Canterbury, Kent, UK
2 Health and Social Care Information Centre, Leeds, UK
3 Imperial College, Statistical Advisory Service, London, UK
4 University of Aberdeen, Department of Medicine and Therapeutics, Aberdeen, UK
Address correspondence to: I. Carpenter. Tel: +44 (0)1227 827760 Fax: +44 (0)1227 827868. Email: g.i.carpenter{at}ukc.ac.uk
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsMeasuresResultsDiscussionKey pointsConflicting interestsReferences |
|---|
Objectives to assess whether measures of cognitive and physical function can explain differences in observed and healthcare resource group (HRG) predicted length of stay for patients presenting with six target conditions at admission.
Design prospective observational study.
Setting three East Kent district general hospitals.
Participants One thousand nine hundred and forty-two consecutive emergency admissions, from March to July2004, with ne or more of six presenting conditions (stroke, fracture neck of femur, myocardial infarction, acute respiratory infection, chronic obstructive airways disease and falls).
Main Outcome Measures length of stay by physical and cognitive function score adjusted for HRG allocated at discharge and other covariates. Physical function was defined using Activities of Daily Living Hierarchy Scale and cognitive function using the Cognitive Performance Scale.
Results median difference between observed and HRG predicted length of stay was 1.2 days (25th percentile estimate, 3.9; 75th percentile estimate, 10.1) for patients with high physical dependency. They stayed 40% longer (95% confidence interval 26-56%) than patients with lower physical dependency after excluding effects of HRG and other covariates. Results are not consistent for cognitive function scores, mainly because most patients had no cognitive impairment.
Conclusions these patients, presenting with conditions common in older patients, would have incurred estimated annual costs of £1.9 million in excess of their HRG tariff-based re-imbursement. Physical function, defined by the degree of dependency in activities of daily living, should be incorporated into HRGs to reduce the financial risk faced by acute hospital services under Payment by Results, the UK diagnosis related group re-imbursement system.
Keywords: case-mix, length of stay, disability, health care costs, diagnosis related groups, elderly
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsMeasuresResultsDiscussionKey pointsConflicting interestsReferences |
|---|
Payment by Results, the UK National Health Service's new diagnosis related group re-imbursement system for England, will be rolled out to cover all hospital and community services from April 2008. The aim of this programme is to improve transparency of funding and hospital efficiency [1]. Essentially, primary care trusts (the commissioners or purchasers of secondary health care in England) will agree specialty level cost and volume contracts with providers, where the cost element will already be determined by a case-mix adjusted national tariff, based on national average costs. Case-mix will be defined by Healthcare Resource Groups ([HRGs]—Diagnosis Related Groups as developed for use in England and Wales), which aim, as far as possible, to group together patients with comparable levels of resource use. The fitness for purpose, or otherwise, of HRGs in defining these iso-resource groups will therefore be an essential element in the new payment system.
There are, however, a number of problems with the current version of HRG particularly for older patients with chronic disabling disease. Patients are allocated to a group based on diagnosis, procedures, age and gender. HRGs use average length of stay as a proxy for resource use rather than actual costs, therefore there is likely to be a significant amount of variation within groups if important factors that drive length of stay have not been used to define the groups. A recent systematic review of older patients who are admitted to hospital found that in addition to diagnosis, age and gender, a number of other factors affect length of stay including physical and cognitive impairment, illness severity, poor nutrition, co-morbidity and poly-pharmacy [2]. Furthermore, inadequate recognition of more complex and more costly cases within HRGs could fuel the concerns that the private healthcare providers could ‘cherry pick’ easier cases, leaving the more expensive ones for the National Health Service without sufficient financial re-imbursement [3].
This study, commissioned by the Health and Social Care Information Centre of the Department of Health, examined whether measures of cognitive and physical function can explain differences in observed and HRG predicted length of stay for six target conditions.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsMeasuresResultsDiscussionKey pointsConflicting interestsReferences |
|---|
Subjects
We conducted a prospective observational study of all consecutive emergency admissions presenting with one or more of six target conditions from March to July 2004 to three district general hospitals. The three hospitals constitute the East Kent Hospitals NHS Trust. The Trust has a catchment population of around 600,000 (of whom 20% are aged > 65 years), an annual budget of just over £310 million and treated 74,500 in-patients during 2004–2005. The six target presenting conditions chosen were stroke, fracture neck of femur, myocardial infarction, acute respiratory infection, chronic obstructive airways disease and falls.
Data collection
Trained nurse assessors visited the admissions unit and identified patients with the target conditions by reviewing the medical admissions notes within 24 h of admission and then followed them until discharge or death. They recorded the patients' admission cognitive and physical function, source of admission, discharge destination and information on informal care status. They also recorded pre-morbid physical and cognitive function on alternate days throughout the patient's hospital stay. Patients who had not been discharged by the end of the study period were excluded from the analysis.
Hospital information departments extracted patient's ICD-10 diagnostic codes and OPCS 4 procedure codes from the Trust's Patient Information System. The Health and Social Care Information Centre used these data to derive the spell-based healthcare resource group for each patient that will be used for re-imbursement purposes. Spell is defined as the package of care from patient admission to discharge, regardless of the hospital specialist who is ultimately responsible for the treatment. The untrimmed national average length of stay for all spell-based non-elective HRGs identified in the study were derived from the National Case-mix Statistics for England 2002–2003 [4].
|
Measures |
|---|
|
TopAbstractIntroductionMethodsMeasuresResultsDiscussionKey pointsConflicting interestsReferences |
|---|
We measured physical function using the Activities of Daily Living (ADL) Hierarchy Scale [5]. This scale is based on four ADL items: indoor locomotion; eating; usage of toilets and personal hygiene. Items are grouped according to the stage of the disablement process in which they occur. Lower scores are assigned for early loss items (e.g. dressing) than late loss items (e.g. eating). The scale ranges between zero (independent) and six (totally dependent). Cognitive function was measured using the Cognitive Performance Scale (CPS) [6], which combines information on decision-making, memory, making self understood and eating into a single categorical scale. Scores range from zero (intact) to six (very severe impairment). The scale correlates highly with long established scales, including Folstein's Mini-Mental Status Examination [7, 8]. The ADL hierarchy and CPS have demonstrated inter-observer reliability for pre-morbid admission status [9].
Data analysis
Prior to analysis, all patients who had more than one target condition at admission were assigned to a single condition group by application of the following rules: Fall plus fractured neck of the femur was treated as fractured neck of the femur; chronic obstructive airways disease plus respiratory infection was treated as chronic obstructive airways disease; fall plus respiratory infection was treated as fall; stroke plus fall was treated as stroke. A small proportion of patients could not be assigned using the allocation rules and were treated as a separate group for analysis.
Cognitive and physical function scores were grouped into 3 subgroups for analysis. We defined these subgroups after prior analyses of the relationship between the length of stay and function scores. Cognitive function was divided into the following groups: low score, defining no impairment (CPS score = 0); medium score, defining mild/moderate impairment (score = 1–3); high score, defining severe impairment (score = 4–6). Physical function was divided into low score, defining independence in ADL (ADL hierarchy scale score = 0); medium score, defining limited/extensive dependence (score = 1–3); high score, defining high level of dependence (score = 4–6).
The association between the observed length of stay and physical and cognitive function scores was explored using descriptive analyses. Length of stay data was log-transformed as the data was skewed to the right. All statistical analyses of length of stay were performed on transformed data. Normality of distribution of log-transformed data was confirmed using Liliefors and Shapiro-Wilks tests.
The observed length of stay for each patient was compared to the predicted length of stay as described by the national average for each patient's allocated spell-based HRG. Mean and median (25th and 75th percentiles) differences are presented. The ratio of observed to predicted length of stay with associated 95% confidence intervals was estimated for each physical and cognitive function subgroup.
The relationship between HRG adjusted lengths of stay of patients by levels of physical and cognitive function for each target condition group was tested. Analysis of covariance (ANCOVA) was used to control the effect that other factors have on length of stay. Covariates included hospital, discharge destination (private dwelling, nursing home, death), admission source (private dwelling, nursing home), and age. Ratios of adjusted length of stay between the groups and associated P-values are presented. Analyses were repeated using trimmed national average length of stay to examine whether trimmed data would change the overall results of the study. (Trimmed length of stay for each HRG is defined as the predicted median length of stay for that HRG where all patients admitted with that HRG (in the definition sample) and who had a length of stay greater than Q3 + 1.5x (Q3–Q1) (the third quartile + 1.5x the inter-quartile range) have been ‘trimmed’ out of the dataset [10].
|
Results |
|---|
|
TopAbstractIntroductionMethodsMeasuresResultsDiscussionKey pointsConflicting interestsReferences |
|---|
Patient characteristics
One thousand nine hundred and forty-two patients from the total of 19,700 emergency admissions (all specialties) during the 4 month study period were presented with one or more of the six target conditions. One thousand six hundred and eight-five patients (86%) were grouped into 257 different HRGs, based on their discharge data. (The medical records of the three hospitals in East Kent Hospitals Trust had recently been merged, which required reconciliation of the ‘hospital number’ of medical records from the three hospitals into a single record with a trust wide ‘hospital number’. The missing records are explained by the inability to track the correct record for these patients after an attempt to reconcile differences by manual examination of the Patient Information System records. This is considered unlikely to have introduced a selection bias into the study sample). One thousand six hundred and seventy-three (99%) of these had a valid physical function score on admission and 1,577 (94%) had a valid cognitive function score. Table 1 details the patient characteristics. Patients were generally elderly, with more than half being female.
|
Analysis of pre-morbid, admission and subsequent physical and cognitive function demonstrated that function at admission was the best predictor of length of stay. All results therefore relate to physical and cognitive function at admission.
The prevalence of cognitive impairment was low, however, most patients had some physical impairment. Mean length of stay was 14 days for all patients, which varied by 12 days between the condition groups with the highest and lowest lengths of stay.
Relationship between physical and cognitive function scores and difference in length of stay
Table 2 presents the means and medians for the difference between observed length of stay and HRG predicted length of stay by physical and cognitive function score at admission for all patients. The ratio of observed to predicted lengths of stay is also presented.
|
For physical function, the results show that overall median difference from expected length of stay was relatively small. However, the range was very different between the ADL groups. Twenty-five percent of the patients with a high ADL score (the upper quartile in the high score group) stayed 10 or more days longer than expected when compared with just 1.8 days in the upper quartile of those in the low ADL hierarchy group.
For cognitive function, the results show a different distribution. Patients with ‘medium’ cognitive impairment (cps score 1–3) had a length of stay greater than predicted by HRG whereas those with low or high scores did not. The range, however, was much greater in both the medium and high groups.
Table 3 presents the ratio of length of stay of higher physical function scores to lower physical function scores by target presenting condition, after controlling for the HRG that is allocated at discharge and other covariates (hospital, discharge destination, admission source and age) by means of ANCOVA. A high ADL hierarchy score, indicating higher dependency, is associated with significantly longer lengths of stay for some of the presenting conditions. Fractured neck of femur and myocardial infarction are not included in the table as most patients were very dependent (94% and 72% in the high ADL score group, respectively) and there were therefore, too few patients in the lower groups for comparison. The results show that level of physical dependency is a strong predictor of length of stay in patients presenting with stroke, acute respiratory infection and falls at admission. For those patients in the respiratory infection group, length of stay increased by 40% between those with low and medium ADL scores and a further 40% between those with medium and high ADL scores. Patients presenting with stroke or falls and a high ADL score had a length of stay 70% greater than those with medium or low ADL scores.
|
With regard to cognitive function, after controlling for HRG and confounding variables, the ratio of length of stay of those with any cognitive impairment when compared with those who had none, was only significant for patients presenting with hip fracture. Patients in the medium to high CPS score group (CPS = 1–6) stayed in hospital, 20% longer than patients who had no cognitive impairment (low, score = 0).
Use of trimmed rather than untrimmed national average length of stay did not change the overall results of the study.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsMeasuresResultsDiscussionKey pointsConflicting interestsReferences |
|---|
Our study shows that physical function, defined by the degree of dependency in ADL, was able to predict the differences in length of stay over and above that which is accounted for by the current version of HRGs. Patients with a high physical function score, indicating high dependency, had a length of stay 40% longer than those of lower dependency groups, even after controlling for HRG and other covariates. The results for cognitive function showed marginally increased length of stay in the intermediate group, but the numbers of patients with cognitive impairment was small.
The study targeted six conditions that are common presentations in older people at admission to hospital. These patients were allocated to 257 different HRGs at discharge, reflecting the fact that during a hospital stay, identification of co-morbidity, different final diagnosis, and almost certainly some inaccuracies in coding resulted in a wide range of case-mix groups, which determine the level of re-imbursement for the care. Sixty-five percent of the patients had an increased level of physical disability, which resulted in a hospital stay 4 days longer than that predicted by the HRG. If it is assumed that the cost of an in-patient day is around £155 for these patients (based on proposed per day long-stay payment for days exceeding trimpoint [11]), then in very crude cash terms, each of these patients could cost the hospitals an additional £600 over and above what is provided for in the HRG tariff, even after adjusting for savings on those with a shorter than predicted length of stay. So for this 4 month period, this group of patients alone will cost the acute hospital Trust, £650,000 more than they will be paid for their care. This amounts to £1.9 million over the course of a year, a significant sum when the Department of Health requires hospitals to ensure that expenditure does not exceed income.
These findings have implications for the future development of HRGs. It is important that physical function is used to better define HRGs, given that those patients with high physical impairment have longer lengths of stay, as highlighted in the present study, and higher costs of care. Two studies in the United States have reported higher hospital costs for patients dependent in ADL even after controlling for diagnostic related group (DRG) payments [12, 13]. Introducing measures of functionality would reduce the variation in actual care costs of individual HRGs and make payment by results (DRG re-imbursement) a fairer financial system. It would also serve to protect the National Health Service against cherry picking by private healthcare providers, as costs incurred to treat these more complex cases will be remunerated at the appropriate rate.
We did not identify which HRGs would benefit from the addition of a physical function measure because the numbers of many HRGs were too small for meaningful analysis. We suggest that a larger study is now needed to test whether physical and cognitive function measures recorded on admission, improve HRG prediction of length of stay across a wider range of conditions (including elective admissions), and to test whether it is feasible to record and use measures of physical and cognitive function in a systematic manner for all hospital admissions.
Physical function is an important factor, which is a driving cost variability within some HRGs. Use of a physical and possibly a cognitive function measure to better define groups would ensure that payments made to providers would reflect the actual costs of care.
|
Key points |
|---|
|
TopAbstractIntroductionMethodsMeasuresResultsDiscussionKey pointsConflicting interestsReferences |
|---|
- Different degrees of physical and cognitive impairment of patients in some Healthcare Resource Groups (the basis of tariff-based re-imbursement in England), results in wide variation in the actual costs of their care.
- Patients with high dependency in activities of daily living had longer lengths of stay when compared to those with lower dependency after excluding effects of HRG and other covariates.
- Physical function measures should be used to better define HRGs and reduce financial risk under case-mix based re-imbursement.
|
Acknowledgements |
|---|
The project was funded by the NHS Health and Social Care Information Centre as part of the HRG v 4.0 revision programme. We are grateful to Sue Langham whose contribution as independent medical writer was funded by CHSS at the University of Kent.
|
Conflicting interests |
|---|
|
TopAbstractIntroductionMethodsMeasuresResultsDiscussionKey pointsConflicting interestsReferences |
|---|
The authors declare that they have no conflicting interests.
|
References |
|---|
|
TopAbstractIntroductionMethodsMeasuresResultsDiscussionKey pointsConflicting interestsReferences |
|---|
- Department of Health. (2002) Reforming NHS Financial Flows: Introducing Payment by Results. London DOH.
- Campbell SE, Seymour DG, Primrose WR, et al. (2004) A systematic literature review of factors affecting outcome in older medical patients admitted to hospital. Age Ageing 33 110–5.
[Abstract/Free Full Text] - Lane R and Paton A. (2005) Bevan betrayed: the demise of the NHS. BMJ 331 853.
[Free Full Text] - National Casemix Office. (2003) National Casemix Statistics (England), 2002–2003. Winchester, MA The National Casemix Office, NHS Headquarters.
- Morris JN, Fries BE, Morris SA. (1999) Scaling ADLs within the MDS. J Gerontol A Biol Sci Med Sci 54 M546–53.[Abstract]
- Morris JN, Fries BE, Mehr DR, Hawes C, Phillips C, Mor V, et al. (1994) MDS cognitive performance scale. J Gerontol 49 M174–82.[ISI][Medline]
- Lawton MP, Casten R, Parmelee PA, Van Haitsma K, Corn J, Kleban MH. (1998) Psychometric characteristics of the minimum data set II: validity. J Am Geriatr Soc 46 736–44.[ISI][Medline]
- Snowden M, McCormick W, Russo J, Srebnik D, Comtois K, Bowen J, et al. (1999) Validity and responsiveness of the Minimum Data Set. J Am Geriatr Soc 47 1000–4.[ISI][Medline]
- Carpenter GI, Teare GF, Steel K, et al. (2001) A new assessment for elders admitted to acute care: reliability of the MDS-AC. Aging Clin. Exp. Res 13 316–30.
- Tukey JW. (1977) Exploratory Data Analysis. Reading, MA Addison-Wiley.
- Department of Health. (2006) National Tariff 2005/2006. Department of Health. http://www.dh.gov.uk/PublicationsAndStatistics/Publications/PublicationsPolicyAndGuidance/PublicationsPolicyAndGuidanceArticle/fs/en?CONTENT_ID=4127649&chk=YwGemZ.
- Chuang KH, Covinsky KE, Sands LP, Fortinsky RH, Palmer RM, Landefeld CS. (2003) Diagnosis-related group-adjusted hospital costs are higher in older medical patients with lower functional status. J Am Geriatr Soc 51 1729–34.[CrossRef][ISI][Medline]
- Covinsky KE, Justice AC, Rosenthal GE, Palmer RM, Landefeld CS. (1997) Measuring prognosis and case mix in hospitalized elders. The importance of functional status. J Gen Intern Med 12 203–8.[CrossRef][ISI][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  M. Espallargues, I. Philp, D.G. Seymour, S.E. Campbell, W. Primrose, S. Arino, E. Dunstan, G. Lamura, P. Lawson, E. Mestheneos, et al. Measuring case-mix and outcome for older people in acute hospital care across Europe: the development and potential of the ACMEplus instrument QJM, February 1, 2008; 101(2): 99 - 109. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. Rozzini, T. Sabatini, A. Ranhoff, and M. Trabucchi Disability, clinical complexity, and diagnosis-related groups in old patients Age Ageing, July 1, 2007; 36(4): 472 - 473. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||