- Email: [email protected]
Factors affecting hospital costs in lung cancer patients in the United Kingdom
Lung Cancer 97 (2016) 8–14
Contents lists available at ScienceDirect
Lung Cancer journal homepage: www.elsevier.com/locate/lungcan
Factors affecting hospital costs in lung cancer patients in the United Kingdom Martyn P.T. Kennedy a,∗ , Peter S. Hall b , Matthew E.J. Callister a a b
Leeds Teaching Hospitals NHS Trust, Beckett Street, Leeds, LS9 7TF, UK Edinburgh Cancer Research Centre, University of Edinburgh, Western General Hospital, Crewe Road South, EH4 2XR, UK
a r t i c l e
i n f o
Article history: Received 19 February 2016 Received in revised form 29 March 2016 Accepted 13 April 2016 Keywords: Lung cancer Cancer Costs Health economics Informatics
a b s t r a c t Introduction: Rising healthcare costs and financial constraints are increasing pressure on healthcare budgets. There is little published data on the healthcare costs of lung cancer in the UK, with international studies mostly small and limited by data collection methods. Accurate assessment of healthcare costs is essential for effective service planning. Methods: We conducted a retrospective, descriptive cohort study linking clinical data from a local electronic database of lung cancer patients at a large UK teaching hospital with recorded hospital income. Costs were adjusted to 2013–2014 prices. Results: The study analysed secondary care costs of 3274 patients. Mean cumulative costs were £5852 (95% CI, £5694 to £6027) at 90 days and £10,009 (95% CI, £9717 to £10,278) at one year. The majority of costs (58.5%) were accumulated within the first 90 days, with acute inpatient costs the largest contributor at one year (42.1%). The strongest predictor of costs was active treatment, especially surgery. Costs were also affected by age, route to diagnosis, clinical stage and cell type. Discussion: Successful early diagnosis initiatives that increase radical treatment rates and improve outcomes may significantly increase the secondary care costs of lung cancer management. The use of routine NHS clinical and financial data can enable efficient and effective analyses of large cohort health economic data. © 2016 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Lung cancer is the second most common cancer in England and is the most common cause of cancer death in the England and Wales [1,2]. Lung cancer survival in the UK is inferior to other developed countries [3]. Rising healthcare costs and financial constraints are increasing pressure on international healthcare budgets. In the UK, cancer and tumours are associated with the third largest NHS expenditure of all disease groups. During the three years from 2009/10 to 2012/13, the budget for all cancer services increased below the rate of inflation at 0.7% per year, while the budget for lung cancer services reduced by 11.5% per year. The diagnosis, treatment and follow-up of lung cancer predominantly occur within secondary care, which accounts for 82.6% of the lung cancer budget [4].
∗ Corresponding author at: c/o Dr M. Callister, Leeds Teaching Hospitals NHS Trust, Beckett Street, Leeds, LS9 7TF, UK. E-mail address: [email protected] (M.P.T. Kennedy). http://dx.doi.org/10.1016/j.lungcan.2016.04.009 0169-5002/© 2016 Elsevier Ireland Ltd. All rights reserved.
New developments to improve early detection, rates of radical treatment and outcomes in lung cancer are often associated with a significant financial cost. In order to provide the highest quality care in a system of scarce resources and financial constraint, it is imperative to have a detailed understanding of the factors that affect costs in the management of patients with lung cancer. Data for UK healthcare costs in lung cancer are limited. A study of factors affecting the costs of 724 patients in Northern Ireland in 2008 remains the only published data [5]. This study was performed using a manual case note review. It reported that inpatient stays were the largest resource cost and that costs were affected by lung cancer stage, co-morbidities, age and social deprivation. More recent studies of factors affecting costs in lung cancer have been published in Europe [6–9], Australia [10] and the USA [11]. The sources for cost data include manual case note review [8,9], insurance claims [10,11], hospital episode statistics [7] (cost data derived from the expected resource utilisation for patients based on primary diagnosis and comorbidities) and patient level statistics [6] (cost data derived from actual resources used by an individual patient). Most of these studies are small, with fewer than 250 patients [5,6,8–10], however two larger studies, conducted in
M.P.T. Kennedy et al. / Lung Cancer 97 (2016) 8–14
9
Europe and USA, have included more than 10,000 patients [7,11]. Inpatient admission and treatment costs commonly account for the greatest cost components [5–7,9–11], and costs are higher in comparable western European countries (such as Germany and France) than in the UK [7]. This study evaluated the direct costs of hospital care in the diagnosis and management of lung cancer in a single large UK teaching hospital using routine NHS data, and aimed to identify factors that were predictive of high costs. 2. Materials and methods Fig. 1. Flowchart of eligible patients.
2.1. Data collection The National Health Service (NHS) provides publicly funded healthcare in the United Kingdom. Services in England are commissioned locally by Clinical Commissioning Groups (CCGs) or centrally through specialist commissioning and are funded via an internal market established in the 1990s according to a national Payment by Results tariff. NHS trusts are required to regularly provide clinical coding data for care episodes, which is processed using the Payment by Results (PbR) grouper software to describe care spells for an individual patient and then coded by Human Resource Group (HRG). Each HRG spell is allocated a cost in pounds based on the national PbR tariff. HRG version 3.5 was used prior to April 2008 with HRG4 (core and unbundled) being used from April 2008 onwards. The HRG-coded data is returned to the NHS trust to guide income claims that are then reimbursed by CCGs. The recorded income (or ‘sold activity’) for Leeds Teaching Hospitals NHS Trust (LTHT) was used to represent direct costs per patient. These costs include all emergency, inpatient and outpatient services with the exception of some specially funded services (e.g. hospital palliative care and PET-CT), and included the cost of highvalue drugs, including tyrosine kinase inhibitors. Finance data was collected in December 2014 for all care spells from January 2008 to October 2014. HRG codes and tariffs are year-specific and are based on the HRG version and PbR tariff in use during that year. All costs were adjusted for inflation to a common base-year of 2013–2014 using the Personal Social Services Research Unit Hospital and Community Health Services Pay and Prices Index [12]. Costs assigned to each care spell were assumed to be incurred on the end date of that spell. Data was not collected on the use or costs of primary care or social services. Clinical staging, outcomes and demographic data were retrospectively collected from a local electronic database of all patients diagnosed with lung cancer at LTHT. This database is based on the National Lung Cancer Audit Database (LUCADA) and National Registry data. All patients who were first seen at LTHT between 01/01/08 to 31/10/13 were included. The clinical and healthcare costs databases were linked deterministically using the NHS number as a unique identifying reference with 99.9% (3274/3276) successful linkage of records. Day zero was defined as the date a patient was first seen by a member of the lung cancer team, and all patients had 12 months of healthcare costs data from the date they were first seen. 2.2. Statistical analysis All statistical analyses were carried out using the R statistical software package version 3.1 [13]. 90 day and 1 year cumulative costs were calculated. Ordinary least squares regression analysis was undertaken on log-costs which were approximately normally distributed. More complex models were avoided in an attempt to allow easy interpretation and back-calculation in future cost-effectiveness modelling. Data with
Fig. 2. Kaplan-Meier survival curve.
no associated cost was allocated a nominal £0.001 cost to permit analysis. Confidence intervals were calculated using the bootstrap method.
3. Results 3.1. Patient characteristics There were 3289 patients first seen between January 2008 and October 2013 at LTHT. Of these, 15 patients were excluded (13 second or recurrent lung cancers and 2 corrupted data); Fig. 1. The remaining 3274 patients were included for analysis. The mean age was 72.5 years (95% CI: 72.1–72.9 years). There were 1883 (57.5%) patients with non-small cell lung cancer, 406 (12.4%) with small cell lung cancer, 25 (0.8%) with carcinoid tumours, and 960 (29.3%) with an unknown cell type. All patients had at least one year of follow-up and one year survival was 38.6% (95% CI: 37.1–40.4%); Fig. 2. Table 1 describes characteristics of patients in the study. Treatment categories are not mutually exclusive and some patients may have received multiple treatment modalities. Epidermal growth factor receptor (EGFR) mutation analysis was performed on 465 (23.8%) patients with stage IIIB and IV lung cancer, with 45 (9.7% of tests) sensitising EGFR mutations detected.
10
M.P.T. Kennedy et al. / Lung Cancer 97 (2016) 8–14
Fig. 3. Mean monthly cost per patient.
Table 1 Patient characteristics.
Patient characteristics Age (median, range)
3.2. Hospital costs Number
Percentage
73 years
Range 22–101 years
Performance status 0 Performance status 1 Performance status 2 Performance status 3 Performance status 4 Performance status NK – Route to diagnosis Emergency admission GP or fast-track referral Inter-consultant referral Radiological follow-up – Cancer characteristics Stage IA Stage IB Stage IIA Stage IIB Stage IIIA Stage IIIB Stage IV
310 1031 740 832 350 11
9.4% 31.5% 22.6% 25.4% 10.7% 0.3%
1027 1389 704 154
31.4% 42.4% 21.5% 4.7%
348 299 116 160 395 383 1573
10.6% 9.1% 3.5% 4.9% 12.1% 11.7% 48.0%
Adenocarcinoma Large cell NOS Small cell Squamous Carcinoid Unknown – Treatment Surgery Chemotherapy Radical radiotherapy SABR Palliative radiotherapy
771 135 307 407 669 25 960
23.5% 4.1% 9.4% 12.4% 20.4% 0.8% 29.3%
484 894 241 130 405
14.8% 27.3% 7.4% 4.0% 12.4%
855 1198 1221
37.3% 36.6% 26.1%
Treatment Intent Radical Palliative Best supportive care
NK = not known; NOS = non-small cell lung cancer not otherwise specified; SABR = stereotactic ablative body radiotherapy.
The total direct cost of hospital care over 12 months for the 3274 patients included in the study was £32,768,229. The mean cumulative costs at 90 days and one year were £5852 (95% CI, £5694 to £6027) and £10,009 (95% CI, £9717 to £10,278) respectively. More than half the cost was accumulated within the first 90 days, and there was a progressive reduction in mean monthly costs over the 12 months following presentation; Fig. 3. Univariate analysis identified that advanced age (p < 0.001, p < 0.001), poor performance status (p < 0.001, p < 0.001), advanced stage (p = 0.737, p < 0.001) and small cell lung cancer (p = 0.070, p = 0.008) are all significant predictors for lower costs at either 90 days or one year respectively; Appendix A. Both palliative (p < 0.001, p < 0.001) and radical treatment (p < 0.001, p < 0.001) were associated with increased costs compared to best supportive care. Presentation via an emergency admission was associated with increased costs at 90 days (p = 0.013) and lower costs at one year (p < 0.001) compared to other routes to diagnosis. Sensitising EGFR mutations predict increased costs at one year (p < 0.001), but are not significant at 90 days (p = 0.367). Fig. 4 demonstrates factors that affect 90 day and one year costs. All factors in the univariate analysis remained significant independent predictors of costs on multivariate analysis, with the exception of performance status and presentation via emergency admission; Appendix B. Presentation via emergency admission was associated with increased 90 day (p < 0.001) and one year (p < 0.001) costs on multivariate analysis. Advanced performance status was no longer significant, being associated with stronger predictors of reduced costs including advanced stage at presentation and lower rates of active treatment Costs were grouped into acute inpatient (including unplanned admissions and emergency department presentations), elective inpatient (including planned diagnostic and therapeutic procedures), lung cancer treatment (including radiotherapy, chemotherapy and surgery) and outpatient (including appointments, investigations and radiology) costs. Data was not available on the costs of some specially funded services, including hospital palliative care and PET-CT. Acute hospital presentations were identified as the largest single contributor to available hospital costs,
M.P.T. Kennedy et al. / Lung Cancer 97 (2016) 8–14
11
Fig. 4. Factors affecting mean 90 day and one year costs.
accounting for 40.8% at 90 days and 42.1% at one year. The other contributors to 90 day and one year costs were elective inpatient (19.0%, 14.7%), lung cancer treatment (23.1%, 26.0%) and outpatient (17.0%, 17.2%) costs.
4. Discussion Previous methods of collecting cost data have primarily relied on manual case note review or insurance claims data. The use of manual case note review to assess cost data is limited due to a timeintensive process that precludes analysis of large populations, with
12
M.P.T. Kennedy et al. / Lung Cancer 97 (2016) 8–14
an associated potential for inconsistency and bias. Insurance claims provide electronic access to the allocated costs of large numbers of healthcare interactions and their use does not require manual data collection. However, interpretation of insurance claim costs is difficult as they are priced at market rates, which may not reflect true healthcare costs. The financial data used in this study provides electronic patient level individualised costs of secondary care. As these data are derived from the national tariff, they represent the true cost of lung cancer to healthcare commissioners. This use of hospital income data is suitable for automated use in large populations and provides costing data that more accurately reflects the actual healthcare expenditure. This study was designed to analyse secondary care costs. Data on the use or costs of primary care, palliative care or social services was not collected. Previous research has demonstrated an increased use of primary care and social services following a diagnosis of cancer [14], which persists even after curative treatment [15]. The absence of this data is a limitation. Although a diagnosis of lung cancer will undoubtedly have an impact on the income and economic productivity of the patient and their family, this study did not evaluate indirect or 3rd sector costs. This is in line with the NICE reference case on the evaluation of evidence on resource use and costs that only consider costs under the direct control of NHS, personal and social services [16]. In the UK, the NHS and associated organisations routinely collect audit data at a local and national level. We have successfully linked clinical data from a national audit database with financial data. There are, however, limitations to this approach related to the content of automated data as the clinical database did not contain information on sex, comorbidities or socioeconomic data. Our study has identified that acute hospital care is the greatest single contributor to available hospital costs at 90 days and one year. Inpatient stays are not only expensive, but are also associated with a poor patient experience. The results of this study were consistent with previous studies that have also identified acute inpatient costs, shortly followed by treatment costs, as the greatest contributors to the overall cost of secondary care lung cancer management [5–7,9–11]. We identified a number of factors that are significant predictors of 90 day and one year costs. The strongest factor associated with increased cost was active treatment, especially surgery or radical treatment. Previous studies also reported that costs were increased dependent on treatment received [6–8,10,11]. Other factors associated with increased costs included good performance status, younger age and outpatient diagnosis, are likely to be associated with an increased probability of receiving radical treatment. A separate univariate analysis identified that diagnosis via an emergency presentation predicts increased costs at one month, however was associated with decreased one year costs in the present analysis. This is, however, heavily influenced by the excess mortality in this poor prognosis group. The multivariate analysis identifies that, when considering the influence of other factors,
diagnosis via an emergency presentation is associated with an increase in one month and one year costs. Our study found that metastatic disease predicted reduced costs. This is likely due to the reduced rate of active treatment, reduced rate of surgery and shorter survival in this patient group. In stage IA–IIIB lung cancer, there was no significant association between stage and cost, however there was a trend to show reduced costs in stage IA. This reduction is likely due to the heterogenous nature of the management of stage IA lung cancer, with a lead-time bias in patients who elect for radiological surveillance and variation in the costs associated with radical treatment options. Reduced costs in stage IV lung cancer are likely to be due to reduced rates of active treatment and shorter survival. These findings reflect the results of previous studies that found that stage influenced costs [5], however some other studies have found that it did not [6,17]. This difference may be due to variability in patient selection and treatments. Finally, Fleming et al. [5] noted that socioeconomic status reduced lung cancer costs, but we were unable to evaluate this in this study. This analysis is of use to CCGs and the government, who are trying to identify drivers of excess cost with a view to manipulating health policy to reduce the cost burden of care. This analysis, as well as the successful use of routine NHS data to perform this analysis, is vital intelligence for that objective. This data could also be used to compare the costs to CCGs of two different centres. Early diagnosis initiatives aim to improve the stage of diagnosis and increase the rate of radical treatment in lung cancer. If these initiatives are successful, this study adds to the weight of published data that predicts they will lead to a significant increase in the direct hospital costs of lung cancer management. This will need to be reflected in an increase in budgets provided to hospitals by CCGs for lung cancer, which in recent years have been reducing. Strategies to reduce acute hospital admission costs may go some way to attenuating these cost rises. This is the largest and most recent analysis of direct costs of the investigation, diagnosis and management of lung cancer in the UK. It demonstrates the utility of routine NHS data from hospital income linked with a national audit database to enable large cohort analyses of health economic data. These methods could be used to investigate broader multicentre data, or alternatively investigate smaller patient groups in more detail. Conflicts of interest None of the authors have any conflicts of interest to declare. Funding No funding to declare. Acknowledgements We have no acknowledgements to declare.
M.P.T. Kennedy et al. / Lung Cancer 97 (2016) 8–14
13
Appendix A. . Univariate regression of 90 day and one year costs (log-scale). 90 day costs
Patient characteristics Age PS 3-4 or NK Referral Emergency Year of diagnosis (from 2008) Cancer characteristics Stage IIIA–IV Intercept: NSCLC Small cell Carcinoid Unknown EGFR + Treatment Surgery Chemotherapy All RT Radical RT SABR Palliative RT Treatment Intent Intercept: BSC Palliative Radical Survival One year survival
1 year costs
Int.
Est.
s.e.
p-value
Int.
Est.
s.e.
p-value
9.940 8.363
−0.024 −0.448
0.003 0.058
<0.001 <0.001
11.108 9.093
−0.032 −0.822
0.002 0.045
<0.001 <0.001
8.153 8.177
0.150 0.060
0.060 0.058
0.013 0.296
8.883 8.583
−0.284 0.072
0.049 0.014
<0.001 <0.001
8.185 8.449 8.449 8.449 8.449 8.197
0.021
0.737
−0.177 −0.992 0.048 0.888
0.050 0.028 0.067 0.049 0.247 0.197
<0.001
0.070 <0.001 0.145 0.367
9.061 9.106 9.106 9.106 9.106 8.782
−0.373
−0.155 −0.077 −0.461 0.220
0.062 0.036 0.086 0.062 0.316 0.244
8.131 8.052 8.166 8.171 8.203 8.198
0.470 0.544 0.142 0.396 −0.080 0.016
0.079 0.062 0.066 0.107 0.144 0.085
<0.001 <0.001 0.031 <0.001 0.579 0.850
8.661 8.574 8.753 8.750 8.788 8.811
0.897 0.803 0.173 0.598 0.146 −0.140
0.062 0.049 0.053 0.086 0.116 0.069
<0.001 <0.001 0.001 <0.001 0.211 0.043
7.748 7.748 7.748
0.685 0.770
0.045 0.064 0.070
<0.001 <0.001
8.056 8.056 8.056
1.053 1.347
0.033 0.047 0.052
<0.001 <0.001
8.177
0.060
0.058
0.296
8.177
0.060
0.058
0.296
0.008 <0.001 0.845 <0.001
PS = performance status; NK = not known; NSCLC = non-small cell lung cancer; EGFR = epidermal growth factor receptor; RT = radiotherapy; SABR = stereotactic ablative body radiotherapy; BSC = best supportive care.
Appendix B. . Multivariate regression of 90 day and one year costs (log-scale) 90 day costs
intercept Age PS 3-4/NK Surgery Chemotherapy Radiotherapy Emergency admission Stage IIIA–IV Small cell Histology NK Carcinoid EGFR+ Year of diagnosis (from 2008)
1 year costs
Estimate
Std. Error
8.095 −0.008 0.008 0.827 0.680 0.446 0.548 0.074 −0.274 −0.444 −0.622 −0.292 0.077
0.238 0.003 0.072 0.114 0.088 0.075 0.064 0.076 0.090 0.075 0.313 0.239 0.016
p value
Estimate
Std. Error
p value
0.004 0.916 <0.001 <0.001 <0.001 <0.001 0.327 0.002 <0.001 0.047 0.222 <0.001
9.040 −0.009 −0.057 1.127 1.059 0.561 0.303 −0.252 −0.235 −0.365 −0.405 0.344 0.030
0.175 0.002 0.053 0.084 0.065 0.055 0.047 0.056 0.066 0.055 0.230 0.175 0.012
<0.001 0.284 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.078 0.050 0.012
PS = performance status; NK = not known; EGFR = epidermal growth factor receptor. References [1] Office for National Statistics, 2014, Deaths registered in England and Wales (Series DR), 2013, London: Office for National Statistics, Online available from: http://www.ons.gov.uk/ons/publications/all-releases. html?definition=tcm%3A77-27475 (accessed 30.09.15.). [2] Office for National Statistics, 2015, Cancer registration statistics, England 2013, London: Office for National Statistics, Online available from: http:// www.ons.gov.uk/ons/rel/vsob1/cancer-statistics-registrations-englandseries-mb1-/index.html (accessed 30.09.15.). [3] S. Walters, S. Benitez-Majano, P. Muller, et al., Is England closing the international gap in cancer survival? Br. J. Cancer 113 (2015) 848–860, http:// dx.doi.org/10.1038/bjc.2015.265. [4] N.H.S. England, 2014, 2012/13CCG programme budgeting benchmarking tool, London, Department of Health, Online available from: https://www.england. nhs.uk/resources/resources-for-ccgs/prog-budgeting/ (accessed 17.02.15.). [5] I. Fleming, P. Monaghan, A. Gavin, et al., Factors influencing hospital costs of lung cancer patients in Northern Ireland, Eur. J. Health Econ. 9 (2008) 79–86, http://dx.doi.org/10.1007/s10198-007-0047-4.
[6] J. Corral, J.A. Espinàs, F. Cots, et al., Estimation of lung cancer diagnosis and treatment costs based on a patient-level analysis in Catalonia (Spain), BMC Health Serv. Res. 15 (2015) 70, http://dx.doi.org/10.1186/s12913-015-0725-3. [7] A. McGuire, M. Martin, C. Lenz, et al., Treatment cost of non-small cell lung cancer in three European countries: comparisons across France, Germany and England using administrative databases, J. Med. Econ. Mar. 24 (2015) 1–16. [8] V. Zarogoulidou, E. Panagopoulou, D. Papakosta, et al., Estimating the direct and indirect costs of lung cancer: a prospective analysis in a Greek University Pulmonary Department, J. Thorac. Dis. 7 (S1) (2015) S12–S19, http://dx.doi. org/10.3978/j.issn. 2072-1439.2015.01.57. [9] J.A. Arca, M.A. Ramos, R.G. de la Infanta, et al., Lung cancer diagnosis: hospitalization costs, Arch. Bronconeumol. 42 (11) (2006) 569–574. [10] S. Kang, E.S. Koh, S.K. Vinod, et al., Cost analysis of lung cancer management in South Western Sydney, J. Med. Imaging Radiat. Oncol. 56 (2) (2012) 235–241, http://dx.doi.org/10.1111/j.1754-9485.2012.02354.x. [11] M. Lanuti, H.J. Hong, S. Ali, et al., Observations in lung cancer over multiple decades: an analysis of outcomes and cost at a single high-volume institution, Eur. J. Cardiothorac. Surg. 46 (2) (2014) 254–261, http://dx.doi.org/10.1093/ ejcts/ezt611.
14
M.P.T. Kennedy et al. / Lung Cancer 97 (2016) 8–14
[12] L. Curtis, Unit Costs of Health and Social Care 2014, Personal Social Services Research Unit, Canterbury, UK, 2014 http://www.pssru.ac.uk. [13] R Development Core Team, R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria, 2014 http://www.2014R-project. [14] X. Chitnis, A. Steventon, A. Glaser, et al., 2014, Use of health and social care by people with cancer, London, Nuffield Trust, Online available from: http:// www.nuffieldtrust.org.uk/publications/use-health-and-social-care-peoplecancer (accessed 22.03.16.).
[15] N.F. Khan, E. Watson, P.W. Rose, Primary care consultation behaviours of long-term, adult survivors of cancer in the UK, Br. J. Gen. Pract. 61 (584) (2011) 197–199. [16] National Institute for Health and Care Excellence, 2013, Guide to the methods of technology appraisal 2013, London: NICE, Online available from: https:// www.nice.org.uk/article/pmg9/resources/non-guidance-guide-to-themethods-of-technology-appraisal-2013-pdf (accessed: 22.03.16.). [17] S. Demeter, P. Jacobs, C. Chmielowiec, et al., The cost of lung cancer in Alberta, Can. Respir. J. 14 (2007) 81–86.
Contents lists available at ScienceDirect
Lung Cancer journal homepage: www.elsevier.com/locate/lungcan
Factors affecting hospital costs in lung cancer patients in the United Kingdom Martyn P.T. Kennedy a,∗ , Peter S. Hall b , Matthew E.J. Callister a a b
Leeds Teaching Hospitals NHS Trust, Beckett Street, Leeds, LS9 7TF, UK Edinburgh Cancer Research Centre, University of Edinburgh, Western General Hospital, Crewe Road South, EH4 2XR, UK
a r t i c l e
i n f o
Article history: Received 19 February 2016 Received in revised form 29 March 2016 Accepted 13 April 2016 Keywords: Lung cancer Cancer Costs Health economics Informatics
a b s t r a c t Introduction: Rising healthcare costs and financial constraints are increasing pressure on healthcare budgets. There is little published data on the healthcare costs of lung cancer in the UK, with international studies mostly small and limited by data collection methods. Accurate assessment of healthcare costs is essential for effective service planning. Methods: We conducted a retrospective, descriptive cohort study linking clinical data from a local electronic database of lung cancer patients at a large UK teaching hospital with recorded hospital income. Costs were adjusted to 2013–2014 prices. Results: The study analysed secondary care costs of 3274 patients. Mean cumulative costs were £5852 (95% CI, £5694 to £6027) at 90 days and £10,009 (95% CI, £9717 to £10,278) at one year. The majority of costs (58.5%) were accumulated within the first 90 days, with acute inpatient costs the largest contributor at one year (42.1%). The strongest predictor of costs was active treatment, especially surgery. Costs were also affected by age, route to diagnosis, clinical stage and cell type. Discussion: Successful early diagnosis initiatives that increase radical treatment rates and improve outcomes may significantly increase the secondary care costs of lung cancer management. The use of routine NHS clinical and financial data can enable efficient and effective analyses of large cohort health economic data. © 2016 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Lung cancer is the second most common cancer in England and is the most common cause of cancer death in the England and Wales [1,2]. Lung cancer survival in the UK is inferior to other developed countries [3]. Rising healthcare costs and financial constraints are increasing pressure on international healthcare budgets. In the UK, cancer and tumours are associated with the third largest NHS expenditure of all disease groups. During the three years from 2009/10 to 2012/13, the budget for all cancer services increased below the rate of inflation at 0.7% per year, while the budget for lung cancer services reduced by 11.5% per year. The diagnosis, treatment and follow-up of lung cancer predominantly occur within secondary care, which accounts for 82.6% of the lung cancer budget [4].
∗ Corresponding author at: c/o Dr M. Callister, Leeds Teaching Hospitals NHS Trust, Beckett Street, Leeds, LS9 7TF, UK. E-mail address: [email protected] (M.P.T. Kennedy). http://dx.doi.org/10.1016/j.lungcan.2016.04.009 0169-5002/© 2016 Elsevier Ireland Ltd. All rights reserved.
New developments to improve early detection, rates of radical treatment and outcomes in lung cancer are often associated with a significant financial cost. In order to provide the highest quality care in a system of scarce resources and financial constraint, it is imperative to have a detailed understanding of the factors that affect costs in the management of patients with lung cancer. Data for UK healthcare costs in lung cancer are limited. A study of factors affecting the costs of 724 patients in Northern Ireland in 2008 remains the only published data [5]. This study was performed using a manual case note review. It reported that inpatient stays were the largest resource cost and that costs were affected by lung cancer stage, co-morbidities, age and social deprivation. More recent studies of factors affecting costs in lung cancer have been published in Europe [6–9], Australia [10] and the USA [11]. The sources for cost data include manual case note review [8,9], insurance claims [10,11], hospital episode statistics [7] (cost data derived from the expected resource utilisation for patients based on primary diagnosis and comorbidities) and patient level statistics [6] (cost data derived from actual resources used by an individual patient). Most of these studies are small, with fewer than 250 patients [5,6,8–10], however two larger studies, conducted in
M.P.T. Kennedy et al. / Lung Cancer 97 (2016) 8–14
9
Europe and USA, have included more than 10,000 patients [7,11]. Inpatient admission and treatment costs commonly account for the greatest cost components [5–7,9–11], and costs are higher in comparable western European countries (such as Germany and France) than in the UK [7]. This study evaluated the direct costs of hospital care in the diagnosis and management of lung cancer in a single large UK teaching hospital using routine NHS data, and aimed to identify factors that were predictive of high costs. 2. Materials and methods Fig. 1. Flowchart of eligible patients.
2.1. Data collection The National Health Service (NHS) provides publicly funded healthcare in the United Kingdom. Services in England are commissioned locally by Clinical Commissioning Groups (CCGs) or centrally through specialist commissioning and are funded via an internal market established in the 1990s according to a national Payment by Results tariff. NHS trusts are required to regularly provide clinical coding data for care episodes, which is processed using the Payment by Results (PbR) grouper software to describe care spells for an individual patient and then coded by Human Resource Group (HRG). Each HRG spell is allocated a cost in pounds based on the national PbR tariff. HRG version 3.5 was used prior to April 2008 with HRG4 (core and unbundled) being used from April 2008 onwards. The HRG-coded data is returned to the NHS trust to guide income claims that are then reimbursed by CCGs. The recorded income (or ‘sold activity’) for Leeds Teaching Hospitals NHS Trust (LTHT) was used to represent direct costs per patient. These costs include all emergency, inpatient and outpatient services with the exception of some specially funded services (e.g. hospital palliative care and PET-CT), and included the cost of highvalue drugs, including tyrosine kinase inhibitors. Finance data was collected in December 2014 for all care spells from January 2008 to October 2014. HRG codes and tariffs are year-specific and are based on the HRG version and PbR tariff in use during that year. All costs were adjusted for inflation to a common base-year of 2013–2014 using the Personal Social Services Research Unit Hospital and Community Health Services Pay and Prices Index [12]. Costs assigned to each care spell were assumed to be incurred on the end date of that spell. Data was not collected on the use or costs of primary care or social services. Clinical staging, outcomes and demographic data were retrospectively collected from a local electronic database of all patients diagnosed with lung cancer at LTHT. This database is based on the National Lung Cancer Audit Database (LUCADA) and National Registry data. All patients who were first seen at LTHT between 01/01/08 to 31/10/13 were included. The clinical and healthcare costs databases were linked deterministically using the NHS number as a unique identifying reference with 99.9% (3274/3276) successful linkage of records. Day zero was defined as the date a patient was first seen by a member of the lung cancer team, and all patients had 12 months of healthcare costs data from the date they were first seen. 2.2. Statistical analysis All statistical analyses were carried out using the R statistical software package version 3.1 [13]. 90 day and 1 year cumulative costs were calculated. Ordinary least squares regression analysis was undertaken on log-costs which were approximately normally distributed. More complex models were avoided in an attempt to allow easy interpretation and back-calculation in future cost-effectiveness modelling. Data with
Fig. 2. Kaplan-Meier survival curve.
no associated cost was allocated a nominal £0.001 cost to permit analysis. Confidence intervals were calculated using the bootstrap method.
3. Results 3.1. Patient characteristics There were 3289 patients first seen between January 2008 and October 2013 at LTHT. Of these, 15 patients were excluded (13 second or recurrent lung cancers and 2 corrupted data); Fig. 1. The remaining 3274 patients were included for analysis. The mean age was 72.5 years (95% CI: 72.1–72.9 years). There were 1883 (57.5%) patients with non-small cell lung cancer, 406 (12.4%) with small cell lung cancer, 25 (0.8%) with carcinoid tumours, and 960 (29.3%) with an unknown cell type. All patients had at least one year of follow-up and one year survival was 38.6% (95% CI: 37.1–40.4%); Fig. 2. Table 1 describes characteristics of patients in the study. Treatment categories are not mutually exclusive and some patients may have received multiple treatment modalities. Epidermal growth factor receptor (EGFR) mutation analysis was performed on 465 (23.8%) patients with stage IIIB and IV lung cancer, with 45 (9.7% of tests) sensitising EGFR mutations detected.
10
M.P.T. Kennedy et al. / Lung Cancer 97 (2016) 8–14
Fig. 3. Mean monthly cost per patient.
Table 1 Patient characteristics.
Patient characteristics Age (median, range)
3.2. Hospital costs Number
Percentage
73 years
Range 22–101 years
Performance status 0 Performance status 1 Performance status 2 Performance status 3 Performance status 4 Performance status NK – Route to diagnosis Emergency admission GP or fast-track referral Inter-consultant referral Radiological follow-up – Cancer characteristics Stage IA Stage IB Stage IIA Stage IIB Stage IIIA Stage IIIB Stage IV
310 1031 740 832 350 11
9.4% 31.5% 22.6% 25.4% 10.7% 0.3%
1027 1389 704 154
31.4% 42.4% 21.5% 4.7%
348 299 116 160 395 383 1573
10.6% 9.1% 3.5% 4.9% 12.1% 11.7% 48.0%
Adenocarcinoma Large cell NOS Small cell Squamous Carcinoid Unknown – Treatment Surgery Chemotherapy Radical radiotherapy SABR Palliative radiotherapy
771 135 307 407 669 25 960
23.5% 4.1% 9.4% 12.4% 20.4% 0.8% 29.3%
484 894 241 130 405
14.8% 27.3% 7.4% 4.0% 12.4%
855 1198 1221
37.3% 36.6% 26.1%
Treatment Intent Radical Palliative Best supportive care
NK = not known; NOS = non-small cell lung cancer not otherwise specified; SABR = stereotactic ablative body radiotherapy.
The total direct cost of hospital care over 12 months for the 3274 patients included in the study was £32,768,229. The mean cumulative costs at 90 days and one year were £5852 (95% CI, £5694 to £6027) and £10,009 (95% CI, £9717 to £10,278) respectively. More than half the cost was accumulated within the first 90 days, and there was a progressive reduction in mean monthly costs over the 12 months following presentation; Fig. 3. Univariate analysis identified that advanced age (p < 0.001, p < 0.001), poor performance status (p < 0.001, p < 0.001), advanced stage (p = 0.737, p < 0.001) and small cell lung cancer (p = 0.070, p = 0.008) are all significant predictors for lower costs at either 90 days or one year respectively; Appendix A. Both palliative (p < 0.001, p < 0.001) and radical treatment (p < 0.001, p < 0.001) were associated with increased costs compared to best supportive care. Presentation via an emergency admission was associated with increased costs at 90 days (p = 0.013) and lower costs at one year (p < 0.001) compared to other routes to diagnosis. Sensitising EGFR mutations predict increased costs at one year (p < 0.001), but are not significant at 90 days (p = 0.367). Fig. 4 demonstrates factors that affect 90 day and one year costs. All factors in the univariate analysis remained significant independent predictors of costs on multivariate analysis, with the exception of performance status and presentation via emergency admission; Appendix B. Presentation via emergency admission was associated with increased 90 day (p < 0.001) and one year (p < 0.001) costs on multivariate analysis. Advanced performance status was no longer significant, being associated with stronger predictors of reduced costs including advanced stage at presentation and lower rates of active treatment Costs were grouped into acute inpatient (including unplanned admissions and emergency department presentations), elective inpatient (including planned diagnostic and therapeutic procedures), lung cancer treatment (including radiotherapy, chemotherapy and surgery) and outpatient (including appointments, investigations and radiology) costs. Data was not available on the costs of some specially funded services, including hospital palliative care and PET-CT. Acute hospital presentations were identified as the largest single contributor to available hospital costs,
M.P.T. Kennedy et al. / Lung Cancer 97 (2016) 8–14
11
Fig. 4. Factors affecting mean 90 day and one year costs.
accounting for 40.8% at 90 days and 42.1% at one year. The other contributors to 90 day and one year costs were elective inpatient (19.0%, 14.7%), lung cancer treatment (23.1%, 26.0%) and outpatient (17.0%, 17.2%) costs.
4. Discussion Previous methods of collecting cost data have primarily relied on manual case note review or insurance claims data. The use of manual case note review to assess cost data is limited due to a timeintensive process that precludes analysis of large populations, with
12
M.P.T. Kennedy et al. / Lung Cancer 97 (2016) 8–14
an associated potential for inconsistency and bias. Insurance claims provide electronic access to the allocated costs of large numbers of healthcare interactions and their use does not require manual data collection. However, interpretation of insurance claim costs is difficult as they are priced at market rates, which may not reflect true healthcare costs. The financial data used in this study provides electronic patient level individualised costs of secondary care. As these data are derived from the national tariff, they represent the true cost of lung cancer to healthcare commissioners. This use of hospital income data is suitable for automated use in large populations and provides costing data that more accurately reflects the actual healthcare expenditure. This study was designed to analyse secondary care costs. Data on the use or costs of primary care, palliative care or social services was not collected. Previous research has demonstrated an increased use of primary care and social services following a diagnosis of cancer [14], which persists even after curative treatment [15]. The absence of this data is a limitation. Although a diagnosis of lung cancer will undoubtedly have an impact on the income and economic productivity of the patient and their family, this study did not evaluate indirect or 3rd sector costs. This is in line with the NICE reference case on the evaluation of evidence on resource use and costs that only consider costs under the direct control of NHS, personal and social services [16]. In the UK, the NHS and associated organisations routinely collect audit data at a local and national level. We have successfully linked clinical data from a national audit database with financial data. There are, however, limitations to this approach related to the content of automated data as the clinical database did not contain information on sex, comorbidities or socioeconomic data. Our study has identified that acute hospital care is the greatest single contributor to available hospital costs at 90 days and one year. Inpatient stays are not only expensive, but are also associated with a poor patient experience. The results of this study were consistent with previous studies that have also identified acute inpatient costs, shortly followed by treatment costs, as the greatest contributors to the overall cost of secondary care lung cancer management [5–7,9–11]. We identified a number of factors that are significant predictors of 90 day and one year costs. The strongest factor associated with increased cost was active treatment, especially surgery or radical treatment. Previous studies also reported that costs were increased dependent on treatment received [6–8,10,11]. Other factors associated with increased costs included good performance status, younger age and outpatient diagnosis, are likely to be associated with an increased probability of receiving radical treatment. A separate univariate analysis identified that diagnosis via an emergency presentation predicts increased costs at one month, however was associated with decreased one year costs in the present analysis. This is, however, heavily influenced by the excess mortality in this poor prognosis group. The multivariate analysis identifies that, when considering the influence of other factors,
diagnosis via an emergency presentation is associated with an increase in one month and one year costs. Our study found that metastatic disease predicted reduced costs. This is likely due to the reduced rate of active treatment, reduced rate of surgery and shorter survival in this patient group. In stage IA–IIIB lung cancer, there was no significant association between stage and cost, however there was a trend to show reduced costs in stage IA. This reduction is likely due to the heterogenous nature of the management of stage IA lung cancer, with a lead-time bias in patients who elect for radiological surveillance and variation in the costs associated with radical treatment options. Reduced costs in stage IV lung cancer are likely to be due to reduced rates of active treatment and shorter survival. These findings reflect the results of previous studies that found that stage influenced costs [5], however some other studies have found that it did not [6,17]. This difference may be due to variability in patient selection and treatments. Finally, Fleming et al. [5] noted that socioeconomic status reduced lung cancer costs, but we were unable to evaluate this in this study. This analysis is of use to CCGs and the government, who are trying to identify drivers of excess cost with a view to manipulating health policy to reduce the cost burden of care. This analysis, as well as the successful use of routine NHS data to perform this analysis, is vital intelligence for that objective. This data could also be used to compare the costs to CCGs of two different centres. Early diagnosis initiatives aim to improve the stage of diagnosis and increase the rate of radical treatment in lung cancer. If these initiatives are successful, this study adds to the weight of published data that predicts they will lead to a significant increase in the direct hospital costs of lung cancer management. This will need to be reflected in an increase in budgets provided to hospitals by CCGs for lung cancer, which in recent years have been reducing. Strategies to reduce acute hospital admission costs may go some way to attenuating these cost rises. This is the largest and most recent analysis of direct costs of the investigation, diagnosis and management of lung cancer in the UK. It demonstrates the utility of routine NHS data from hospital income linked with a national audit database to enable large cohort analyses of health economic data. These methods could be used to investigate broader multicentre data, or alternatively investigate smaller patient groups in more detail. Conflicts of interest None of the authors have any conflicts of interest to declare. Funding No funding to declare. Acknowledgements We have no acknowledgements to declare.
M.P.T. Kennedy et al. / Lung Cancer 97 (2016) 8–14
13
Appendix A. . Univariate regression of 90 day and one year costs (log-scale). 90 day costs
Patient characteristics Age PS 3-4 or NK Referral Emergency Year of diagnosis (from 2008) Cancer characteristics Stage IIIA–IV Intercept: NSCLC Small cell Carcinoid Unknown EGFR + Treatment Surgery Chemotherapy All RT Radical RT SABR Palliative RT Treatment Intent Intercept: BSC Palliative Radical Survival One year survival
1 year costs
Int.
Est.
s.e.
p-value
Int.
Est.
s.e.
p-value
9.940 8.363
−0.024 −0.448
0.003 0.058
<0.001 <0.001
11.108 9.093
−0.032 −0.822
0.002 0.045
<0.001 <0.001
8.153 8.177
0.150 0.060
0.060 0.058
0.013 0.296
8.883 8.583
−0.284 0.072
0.049 0.014
<0.001 <0.001
8.185 8.449 8.449 8.449 8.449 8.197
0.021
0.737
−0.177 −0.992 0.048 0.888
0.050 0.028 0.067 0.049 0.247 0.197
<0.001
0.070 <0.001 0.145 0.367
9.061 9.106 9.106 9.106 9.106 8.782
−0.373
−0.155 −0.077 −0.461 0.220
0.062 0.036 0.086 0.062 0.316 0.244
8.131 8.052 8.166 8.171 8.203 8.198
0.470 0.544 0.142 0.396 −0.080 0.016
0.079 0.062 0.066 0.107 0.144 0.085
<0.001 <0.001 0.031 <0.001 0.579 0.850
8.661 8.574 8.753 8.750 8.788 8.811
0.897 0.803 0.173 0.598 0.146 −0.140
0.062 0.049 0.053 0.086 0.116 0.069
<0.001 <0.001 0.001 <0.001 0.211 0.043
7.748 7.748 7.748
0.685 0.770
0.045 0.064 0.070
<0.001 <0.001
8.056 8.056 8.056
1.053 1.347
0.033 0.047 0.052
<0.001 <0.001
8.177
0.060
0.058
0.296
8.177
0.060
0.058
0.296
0.008 <0.001 0.845 <0.001
PS = performance status; NK = not known; NSCLC = non-small cell lung cancer; EGFR = epidermal growth factor receptor; RT = radiotherapy; SABR = stereotactic ablative body radiotherapy; BSC = best supportive care.
Appendix B. . Multivariate regression of 90 day and one year costs (log-scale) 90 day costs
intercept Age PS 3-4/NK Surgery Chemotherapy Radiotherapy Emergency admission Stage IIIA–IV Small cell Histology NK Carcinoid EGFR+ Year of diagnosis (from 2008)
1 year costs
Estimate
Std. Error
8.095 −0.008 0.008 0.827 0.680 0.446 0.548 0.074 −0.274 −0.444 −0.622 −0.292 0.077
0.238 0.003 0.072 0.114 0.088 0.075 0.064 0.076 0.090 0.075 0.313 0.239 0.016
p value
Estimate
Std. Error
p value
0.004 0.916 <0.001 <0.001 <0.001 <0.001 0.327 0.002 <0.001 0.047 0.222 <0.001
9.040 −0.009 −0.057 1.127 1.059 0.561 0.303 −0.252 −0.235 −0.365 −0.405 0.344 0.030
0.175 0.002 0.053 0.084 0.065 0.055 0.047 0.056 0.066 0.055 0.230 0.175 0.012
<0.001 0.284 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.078 0.050 0.012
PS = performance status; NK = not known; EGFR = epidermal growth factor receptor. References [1] Office for National Statistics, 2014, Deaths registered in England and Wales (Series DR), 2013, London: Office for National Statistics, Online available from: http://www.ons.gov.uk/ons/publications/all-releases. html?definition=tcm%3A77-27475 (accessed 30.09.15.). [2] Office for National Statistics, 2015, Cancer registration statistics, England 2013, London: Office for National Statistics, Online available from: http:// www.ons.gov.uk/ons/rel/vsob1/cancer-statistics-registrations-englandseries-mb1-/index.html (accessed 30.09.15.). [3] S. Walters, S. Benitez-Majano, P. Muller, et al., Is England closing the international gap in cancer survival? Br. J. Cancer 113 (2015) 848–860, http:// dx.doi.org/10.1038/bjc.2015.265. [4] N.H.S. England, 2014, 2012/13CCG programme budgeting benchmarking tool, London, Department of Health, Online available from: https://www.england. nhs.uk/resources/resources-for-ccgs/prog-budgeting/ (accessed 17.02.15.). [5] I. Fleming, P. Monaghan, A. Gavin, et al., Factors influencing hospital costs of lung cancer patients in Northern Ireland, Eur. J. Health Econ. 9 (2008) 79–86, http://dx.doi.org/10.1007/s10198-007-0047-4.
[6] J. Corral, J.A. Espinàs, F. Cots, et al., Estimation of lung cancer diagnosis and treatment costs based on a patient-level analysis in Catalonia (Spain), BMC Health Serv. Res. 15 (2015) 70, http://dx.doi.org/10.1186/s12913-015-0725-3. [7] A. McGuire, M. Martin, C. Lenz, et al., Treatment cost of non-small cell lung cancer in three European countries: comparisons across France, Germany and England using administrative databases, J. Med. Econ. Mar. 24 (2015) 1–16. [8] V. Zarogoulidou, E. Panagopoulou, D. Papakosta, et al., Estimating the direct and indirect costs of lung cancer: a prospective analysis in a Greek University Pulmonary Department, J. Thorac. Dis. 7 (S1) (2015) S12–S19, http://dx.doi. org/10.3978/j.issn. 2072-1439.2015.01.57. [9] J.A. Arca, M.A. Ramos, R.G. de la Infanta, et al., Lung cancer diagnosis: hospitalization costs, Arch. Bronconeumol. 42 (11) (2006) 569–574. [10] S. Kang, E.S. Koh, S.K. Vinod, et al., Cost analysis of lung cancer management in South Western Sydney, J. Med. Imaging Radiat. Oncol. 56 (2) (2012) 235–241, http://dx.doi.org/10.1111/j.1754-9485.2012.02354.x. [11] M. Lanuti, H.J. Hong, S. Ali, et al., Observations in lung cancer over multiple decades: an analysis of outcomes and cost at a single high-volume institution, Eur. J. Cardiothorac. Surg. 46 (2) (2014) 254–261, http://dx.doi.org/10.1093/ ejcts/ezt611.
14
M.P.T. Kennedy et al. / Lung Cancer 97 (2016) 8–14
[12] L. Curtis, Unit Costs of Health and Social Care 2014, Personal Social Services Research Unit, Canterbury, UK, 2014 http://www.pssru.ac.uk. [13] R Development Core Team, R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria, 2014 http://www.2014R-project. [14] X. Chitnis, A. Steventon, A. Glaser, et al., 2014, Use of health and social care by people with cancer, London, Nuffield Trust, Online available from: http:// www.nuffieldtrust.org.uk/publications/use-health-and-social-care-peoplecancer (accessed 22.03.16.).
[15] N.F. Khan, E. Watson, P.W. Rose, Primary care consultation behaviours of long-term, adult survivors of cancer in the UK, Br. J. Gen. Pract. 61 (584) (2011) 197–199. [16] National Institute for Health and Care Excellence, 2013, Guide to the methods of technology appraisal 2013, London: NICE, Online available from: https:// www.nice.org.uk/article/pmg9/resources/non-guidance-guide-to-themethods-of-technology-appraisal-2013-pdf (accessed: 22.03.16.). [17] S. Demeter, P. Jacobs, C. Chmielowiec, et al., The cost of lung cancer in Alberta, Can. Respir. J. 14 (2007) 81–86.