Healthcare costs, productivity loss and quality of life after rehabilitation including a new approach to physical activity: A case study of ReSpAct.

1

Healthcare costs, productivity loss and quality of life after

rehabilitation including a new approach to physical activity: A

case study of ReSpAct.

Master thesis MSc. Economics

Author: Floris Groeneveld

– s2041782

Date: January 14, 2016

Supervisors:

J.O. Mierau

F. Hoekstra

1

_{Floris Groeneveld: MSc. Economics and MSc. Finance student at the Faculty of Economics}

and Business, University of Groningen, the Netherlands. The author can be contacted via

e-mail on floris_groeneveld@hote-mail.com

2

_{Jochen Mierau: Associate Professor, Department of Economics, Econometrics and Finance,}

University of Groningen, the Netherlands.

3

_{Femke Hoekstra: PhD student, Centre for Human Movement Sciences, University of}

2

Abstract

Objective:

Methods:

Results:

Conclusion:

3

Introduction

In high income countries, a chronic condition negatively affects personal wages, hours worked, total earnings, consumption, savings, workforce participation, retirement, job turnover and increases healthcare costs.1-3 In the Netherlands, only 26 percent of people with a chronic condition and/or physical disability have a paid job compared to 66 percent of the Dutch population.4 In addition, poor health is related to lower performance at work (presenteeism) and more absency.5 Therefore, chronic conditions and/or physical disabilities are a root cause of rising healthcare costs and productivity loss for society. Earlier studies have shown that physical activity can reduce

cardiovascular disease.6-9 From an economic perspective, it is important for the chronically ill and physically disabled to engage in physical activity to reduce the odds of multi-morbidity, which is associated with a significant multiplication of healthcare costs.10,11 Physical activity generate important medical and economic benefits which can be stimulated.

Recent studies conclude that physical activity promoting programmes are effective

post-rehabilitation treatment and could be more effective if incorporated in the post-rehabilitation.12-15 To realise considerable health and economic benefits, the Rehabilitation, Sports and Exercise (RSE) programme has been developed and aims to stimulate physical activity of patients with a chronic condition and/or physical disability. RSE is an evidence based programme incorporated in and after the rehabilitation treatment based on the study of Van der Ploeg.16 Currently, patients with a chronic condition and/or a physical disability engage less in physical activity relative to the Dutch

population.16 Therefore, patients with a chronic condition and/or physical disability constitute a good target group of physical activity promoting programmes.16,17 Despite the fact that there are signs that the RSE programme is effective, empirical evidence on the societal costs of the patients enrolled in the RSE programme have remained untouched until today. Therefore, this paper provides insight in societal costs, which include healthcare cost and productivity loss. It is important to provide an estimation of the relative size of costs for further economic evaluations.

4

In sum, the main objective of this paper is to provide insight in societal cost and wellbeing of patients enrolled in the RSE programme. Healthcare costs include all major healthcare services. Levels and differences in costs between discharge (T0) and 14 weeks after discharge from rehabilitation (T1) are estimated. It is evaluated whether healthcare costs are influenced by personal characteristics and type of medical condition. Personal characteristics include gender, age, being overweight, smoking behaviour and level of education. Earlier studies confirm that before mentioned personal

characteristics have an effect on healthcare costs. In general, women use more healthcare due to pregnancy and birth control, yet may play a limited role in this study sample.25 Further, healthcare costs rise with age as one becomes more prone to illnesses at higher age.26 Being overweight and smoking behaviour are also associated with higher healthcare costs.27,28 Education level is included to account for social economic class as literature concludes that the low (high) educated consults the GP (medical specialist) more frequently.29 The effect of being diagnosed with a brain disease, a movement related illness or chronic pain on healthcare costs at T0, T1 and its difference will be assessed. The second part of societal costs, productivity loss, plays a role in a societal perspective and includes paid and unpaid work loss. Productivity loss for a patient with a stroke has been found to be different than chronic pain patients and therefore productivity loss will also be subjected to a subgroup analysis.30,31 Thirdly, the HR-QoL is estimated using the RAND-12 form and its relation to healthcare costs or type of medical condition is evaluated. Main hypothesis is that total costs decline at T1 as patients have been diagnosed and discharged from rehabilitation at T0. It is expected that heterogeneity between patients exist in all outcome variables.

This paper is structured as follows. Following section features the design, data collection, study population and data analysis. Section 3 describes the results of the analysis. Section 4 presents a discussion of the results. Section 5 concludes the paper with a brief summary and recommendations for future research.

Methods

Design

5

productivity loss and Health Related Quality of Life (HR-QoL). See Alingh32 and Hoekstra33for more detailed sections on the design of the ReSpAct study.

RSE programme

RSE is an evidence based physical activity promoting programme for the chronically ill and/or physically disabled.16 During rehabilitation, patients were introduced to multiple types of exercise and are referred to a sports counselling centre. Between 6 and 3 weeks before discharge from rehabilitation, a face-to-face advice session at the sports counselling centres took place and addressed possible motivational issues for physical activity and any personal or environmental barriers. After discharge from rehabilitation, four counselling sessions (C.1, C.2, C.3 and C.4) took place by phone or e-mail to assess patient’s engagement in physical activity. Counselling sessions were targeted to provide information on the suitability of the patients’ condition with physical activity. An overview of the programme can be found in figure 1.

Figure 1: Overview of the RSE programme. The first face-to-face advice took place during rehabilitation. At discharge, the first questionnaire was filled in. 2, 5, 8 and 13 weeks after discharge

there were counselling sessions. Second questionnaire was filled in 14 weeks after discharge.

Study population

Patients who participated in the RSE programme were included in this paper if they met the following conditions a) agreed to participate, b) were at least 18 years of age, c) suffered from a physical disability and/or chronic condition and d) received inpatient or outpatient treatment at one of the participating centres or hospitals between 2012 and 2015. Patients were excluded if they were unable to complete the Dutch questionnaires or if they participated in another physical activity promoting programme. The study population consisted of inpatient (n=33) and outpatient (n=1114) patients who received a rehabilitation treatment. Patients’ characteristics can be found in table 1. Income level was reported to be either higher, lower or equal to the most common level (modal) of the population in the Netherlands. Medical conditions of patients were documented by the

6

TABLE 1

CHARACTERISTICS OF THE SAMPLE (n = 1147)

Variable Percent N Variable Percent N

Age High Body Mass Index > 25

18-44 25,3% 290 Yes 61,4% 704

45-64 53,3% 611 (missing) 0,7% 8

65-85 20,6% 236 Medical condition

(missing) 0,9% 10 Brain disease 25,4% 291

Gender Of which stroke 197

Male 45,8% 525 Movement related illness 17,5% 201

Female 53,9% 618 Chronic pain syndrome 16,1% 185

(missing) 0,3% 4 Neurological disease 14,6% 167

Education Organ disease 12,7% 146

Lower than secondary school 4,2% 48 Of which heart disease 115

Lower vocational eduction 32,9% 377 Amputation lower or upper 5,1% 58

Secondary education 37,4% 429 Paraplegia 3,7% 42

Higher professional or university 24,8% 284 Other disease or problem 3,7% 42

(missing) 0,8% 9 (missing) 1,3% 15

Income level Number of RSE counselling

Lower than biggest group 23,3% 267 C.1 11,2% 129

Biggest group 21,7% 249 C.2 12,0% 138

Higher than biggest group 30,8% 353 C.3 38,1% 437

Not revealed 22,5% 258 C.4 5,6% 64

(missing) 1,7% 20 (missing) 33,0% 379

Data collection

Two questionnaires were distributed via an online tool and through paper versions. Data was included from May 2013 until May 2015. Healthcare utilisation was measured by the non-disease specific Medical Consumption Questionnaire (MCQ) with a recall period of 13 weeks.35 The MCQ included a broad range of questions from non-referral required care (e.g. GP) to referral required care (e.g. medical specialists). An overview of all items can be found in the appendix. For this paper, the MCQ was extended with questions adapted to the chronically ill and/or physically disabled patients in the rehabilitation setting. Extensions included the level of informal care provided at home and medical examinations such as laboratory and imaging tests. One version of the questionnaire contained an error on the nursing part of informal care.

In addition, this paper included the Productivity Cost Questionnaire (PCQ) with questions on productivity loss.36 Productivity costs were assessed via three channels. Firstly, the worker

7

period of four weeks since a longer period would have had a negative influence on the accuracy of reporting37. The productivity loss was extrapolated to 13 weeks. Utilisation levels and productivity loss derived from the questionnaires were multiplied by their respective unit cost (see costing methods section) in order to arrive at final costs. Both MCQ and PCQ were tested in a feasibility study and were framed to ensure that 95% of the Dutch population understood its meaning35,36. In addition to monetary components, HR-QoL was estimated by the validated RAND-12 version 1 to indicate changes in the wellbeing of the patient. RAND-12 included questions on eight dimensions that feature in to a physical as well as mental HR-QoL score. RAND-12 dimensions consist of physical functioning, role limitations due to physical and emotional problems, energy/fatigue, emotional wellbeing, social functioning, pain and general health. HR-QoL was derived from scoring

aforementioned eight topics and resulted in a relative physical and mental health score.38 _{In theory,} physical and mental health scores can range from zero (worst health state) to 100 (best health state) where changes between T0 and T1 were considered. Furthermore, a visual analogue scale from zero (worst health) to 10 (perfect health) was added as a supplementary instrument how a patient assessed its own health.

Costing methods

Healthcare utilisation was measured via the MCQ and was multiplied by cost per resource unit in order to arrive at final costs. Currently, The Dutch Manual of Costing (2010) is the standard in healthcare research as these prices reflect opportunity costs while keeping a level of comparability with other studies.39 Secondly, an average of the first five public prices found was used when the Manual of Costing was too limited. Public pricing method was needed in merely a small amount of cases in terms of utilisation reported and number of healthcare services. An overview of costing sources can be found in the appendix.

8

consumer price index by Statistics Netherlands except for the services where the public prices of 2015 were needed.

Data analysis

Firstly, healthcare costs of the inpatients and outpatients were evaluated at T0 and T1. Analysis on healthcare costs included total costs, non-referral required care, (in)formal home care, medical specialists, medical scans and non-imaging tests, in/outpatient treatment in centres other than the hospital. Estimations were conducted on the cost levels at T0 and T1 and the difference were

reported in mean values and standard deviations. For the difference between T0 and T1, a two-tailed test was used and p-values were reported. After that, a multivariate Ordinary Least Squares (OLS) was conducted on the level of healthcare costs at T0 , T1 and its difference. In this way, insight is given in possible marginal effects of specific heterogeneity in patients’ personal and lifestyle

characteristics on healthcare cost. The OLS included age, a female dummy, being overweight, being a smoker and level of education. Earlier studies mentioned in the introduction concluded these

characteristics could have an effect on healthcare costs. A correlation matrix concerning the above characteristics can be found in the appendix. Secondly, productivity loss was calculated via

absenteeism, presenteeism and unpaid work loss. Productivity loss was reported by mean, standard error and 95% confidence intervals. A subgroup analysis was provided, since previous studies concluded that different medical conditions have an effect on a patient’s productivity loss. HR-QoL was described by the RAND-12 form which components were reported. RAND-12 components include physical and mental health score of which means and standard errors were reported. Possible changes in physical and/or mental health scores were compared to the minimum clinically important difference of “three to five points” of the more extensive RAND-36.40 A patient’s assessment on personal wellbeing through a visual analogue scale was provided. Possible

associations between healthcare costs and (changes in) HR-QoL were analyzed. Statistical significance was set at p < 0.05.

9

Results

Healthcare costs at T0 and T1

Mean total healthcare costs at T0 were € 2.595 and € 976 at T1.Total healthcare costs at T1 is significantly lower compared to T0 (95% CI € 1.160 to € 2.078, p<0,01). See table 2. Dividing the below costs with the per unit cost in the appendix will result in healthcare utilisation levels.

TABLE 2

HEALTHCARE COSTS

T0 T1 Difference

Costs in € (2014) Mean St. Dev. Mean St. Dev. Mean St. Dev. P-value

Non-referral required care

General practitioner 50 71 54 80 -4 85 0,12 Social counselor 97 205 28 119 67 206 <0,01 Physical therapist 301 528 238 392 63 624 <0,01 Ergo therapist 113 187 15 61 98 190 <0,01 Speech therapist 31 145 11 75 20 149 <0,01 Dietitian 10 84 5 22 4 101 0,11 Homoeopathist 20 151 16 89 5 162 0,21

Psychologist (incl. clinical) 206 536 118 344 76 598 <0,01

Company doctor 180 450 130 222 50 423 <0,01

Emergency department 26 77 15 55 11 85 <0,01

Ambulance transport 36 134 11 69 24 139 <0,01

Total non-referral required care* 1.018 1.456 311 401 707 1.377 <0,01

Formal care at home

Household chores 90 475 61 295 28 441 0,01

Care of patient 112 1.168 58 520 55 1.170 0,06

Nursing of patient 45 525 14 203 31 531 0,03

Total formal care at home* 161 838 176 847 -15 642 <0,01

Informal care at home

Household chores 340 1.058 318 1.288 22 1.450 0,36

Care of patient 162 718 82 400 80 724 0,01

Nursing of patient** 316 1.601 184 899 132 1.684 0,04

Total informal care at home* 435 1.907 350 1.813 85 2.320 0,24

Hospital care

Medical specialist* 374 447 248 219 126 450 <0,01

Medical scans and tests* 113 208 65 135 48 227 <0,01

Inpatient treatment hospital* 844 2.995 119 788 726 3.075 <0,01 Rehabilitation care

Outpatient rehabilitation centre 1.395 2.534 116 606 1.278 2.592 <0,01

10

A majority of healthcare costs decreased significantly between T0 and T1. Between T0 and T1, costs for the general practitioner, dietician and homoeopathist did not change. Costs of the social

counsellor, physical therapist, ergo therapist, psychologist and company doctor were significantly lower and relevant from an economic point of view. Total non-referable required care declined significantly by € 707, a reduction of 69 percent compared to the level at T0.

Formal care at home increased marginally, though household chores and nursing declined

marginally. Informal care was provided by relatives or friends and care of patient decreased by € 80. An item of Informal care (nursing of patient) was reported for consistency, however one paper version of the questionnaire contained an error in this part.

Medical specialist care was lower by € 126 compared to T0. Medical scans and non-imaging tests was lower. Inpatient treatment at the hospital was reduced very significantly by € 727. Outpatient

treatment costs at the rehabilitation centre were € 1.279 less compared to T0.

Subgroup analysis

Analysis for lifestyle and participants characteristics

Healthcare costs could have been impacted by patients’ age, gender, being overweight or a smoker and education level. Total healthcare costs at T0 were related to age and costs at T1 were not different from zero. The constant in healthcare costs at T0 was € 2.463 with an increase of € 25 each year above age 18 (95% CI € 4,77 to € 44,96, p=0.01). See table 3.

TABLE 3

MULTIVARIATE OLS TOTAL HEALTHCARE COSTS

T0 and T1

Variables Coefficient St. Error 95% CI P-value

11

Differences in healthcare costs between T0 and T1 could as well have been impacted by patient’s age, gender, being overweight or a smoker and education level. Multivariate OLS was used to assess the marginal effects of before mentioned patient aspects on cost components of healthcare costs (see table 4). Non-referable required care decreased significantly with € 1432 and an additional € 511 for the group of smokers in comparison to T0. For every year above 18 years of age, the

difference was decreased by € 14. For someone aged 50, the age difference is 50 minus 18 = 32 and the marginal effect of age 50 on the difference of non-referable required care costs between T0 and T1 was 32 x -14 = € - 448. The change in formal care at home costs was too marginal (€ - 15) to perform a meaningful multivariate analysis. In outpatient treatment centres excluding hospital, age was not relevant in explaining changes in costs between T0 and T1. Difference in medical specialist care cost was significantly lower for women. Decreases in medical scans and tests cost showed to be consistent by 120 euro while the gap is statistically significantly narrowed for women, yet not relevant from an economic point of view. Decline in costs for inpatient treatment in hospital was greater for higher-aged patients.

The subgroup analysis for non-referable required care costs showed the strongest marginal effects of age, gender, being overweight, being a smoker and education level. In two out of five cost models, age was an important determinant of healthcare cost difference. Being female was significant in explaining cost differences of both medical scans, tests and medical specialist care. Being a smoker resulted in a € 500 higher change in non-referable required care. In conclusion, characteristics such as age, being female and smoking affected differences between T0 and T1 in a few cost items, yet was not apparent enough to conclude heterogeneity in patient characteristics and lifestyle is relevant from a cost perspective.

Analysis for three most frequent medical conditions

12

TABLE 4

MULTIVARIATE OLS OF DIFFERENCE IN

HEALTHCARE COSTS PER ITEM*

Non-referable required care

Outpatient treatment centres

(excl. hospital) Medical specialist care

Variables Coefficient St. Error 95% CI P-value Coefficient St. Error 95% CI P-value Coefficient St. Error 95% CI P-value

Constant 1.432 383 [678 to 2.187] <0,01 336 337 [- 405 to 1.077] 0,38 116 116 [-112 to 345] 0,32 Age -14 6 [-26 to -3] 0,01 10 5 [-0,5 to 20] 0,06 -1 2 [-4 to 3] 0,72 Female -215 142 [495 to 66] 0,13 33 135 [-232 to 297] 0,81 -116 47 [-209 to -23] 0,02 Overweight -123 144 [-406 to 159] 0,39 -173 135 [-439 to 93] 0,21 -10 43 [-94 to 74] 0,82 Smoker 511 178 [160 to 861] <0,01 -103 176 [-448 to 242] 0,56 102 55 [5 to 210] 0,06 Education -78 82 [-239 to 84] 0,34 4 80 [-155 to 162] 0,97 33 25 [-17 to 82] 0,20

Medical scans and tests Inpatient hospital treatment

Variables Coefficient St. Error 95% CI P-value Coefficient St. Error 95% CI P-value

Constant 120 52 [18 to 223] 0,02 -494 484 [-1.641 to 653] 0,40 Age -1 1 [-2 to 1] 0,22 17 8 [1 to 33] 0,04 Female -43 20 [-82 to 4] 0,03 -399 217 [-824 to 27] 0,07 Overweight 1 20 [-38 to 41] 0,94 122 221 [-311 to 556] 0,58 Smoker -21 25 [-71 to 29] 0,42 188 281 [-363 to 739] 0,50 Education -8 11 [-30 to 15] 0,50 216 129 [-37 to 469] 0,09

13

Figure 1: Non-referable required healthcare costs per person per medical condition with one SD error bar

Productivity loss

Next to healthcare costs, societal costs also included productivity loss. Of the study population, 46% had a paid job at T0 compared to 44% at T1. Net income for paid workers was significantly lower at T1 (€ 110, 95 % CI € 37 to € 182, P<0,01). The friction period of 160 days absence was exceeded in 41% of cases. Of the paid workers, 26% did not report in being absent because of their medical condition. Productivity loss of paid work at T0 was € 21.299 per person (see table 5) with some outliers in the € 70.000s which were explained by higher wages. Presenteeism and unpaid work were significantly lower than the productivity loss of paid workers in both periods. Presenteeism did not change significantly between T0 and T1 while the productivity loss of unpaid work decreased significantly after T0.

Additionally, patients’ type of medical condition determined productivity loss. At T0, patients with a brain disease were related with € 28.721 (95 % CI € 25.357 to €32.086) absenteeism costs compared to movement related illness € 15.355 (95 % CI € 11.648 to € 19.062) and chronic pain € 18.286 (95 % CI € 13.856 to €22.725). Presenteeism and unpaid work loss were unrelated to the type of medical condition.

TABLE 5

PRODUCTIVITY LOSS

T0 T1

Costs in € (2014) Mean St. Error 95% CI Mean St. Error 95% CI

Paid work 21.299 910 [19.510 to 23.088] 5.037 488 [4.076 to 5.998]

Presenteeism 3.141 260 [2.629 to 3.653] 2.633 265 [2.108 to 3.157]

14

Health-related quality of life (HR-QoL)

The HR-QoL was assessed using the RAND-12 form (see table 6) between T0 and T1. Physical health score is low. The physical health score rose by 0,84 score points (95% CI 0,29 to 1,39, P<0,01). Mental health score did not change (95% CI -0,18 to 1,19, P=0,07). Changes in the physical and mental health scores were not relevant from a clinical point of view. The visual analogue scale was a reported scale from zero to 10 and did not change between T0 and T1. Pearson’s correlations between healthcare costs and physical or mental health score were not significant. The mental and physical health scores were also not different in other age groups.

TABLE 6

RESULTS RAND-12 AND VAS SCALE

Variable Mean St dev. Variable Mean St dev.

RAND12 - T0 RAND12 - T1

Physical health related score 35,9 9,0 Physical health related score 36,8 9,5 Mental health related score 47,1 10,7 Mental health related score 47,6 10,6

Visual analogue scale - T0 5,55 2,1 Visual analogue scale - T1 5,49 2,25

While a patient’s medical condition was previously not relevant for healthcare costs, it soundly influenced productivity costs at T0 and physical health score at T0 and T1. Patients with a brain disease had a significant higher mental health score than the movement related illness and chronic pain group in this study (see table 7). The mental health score was not different for other types of medical conditions and is reported in table 8.

TABLE 7

PHYSICAL HEALTH SCORE PER MEDICAL CONDITION

Variable Mean St. Error Variable Mean St. Error

RAND12 - T0 RAND12 - T1

Brain disease 39,1 0,50 Brain disease 39,7 0,67

Movement related illness 33,4 0,62 Movement related illness 35,5 0,81

Chronic pain 32,9 0,61 Chronic pain 33,6 0,82

RAND12 - T0 95 % CI RAND12 - T1 95 % CI

Brain disease 38,2 40,1 Brain disease 38,3 41,0

Movement related illness 32,2 34,6 Movement related illness 33,9 37,1

Chronic pain 31,7 34,1 Chronic pain 31,9 35,2

15

TABLE 8

MENTAL HEALTH SCORE PER MEDICAL CONDITION

Variable Mean St. Error Variable Mean St. Error

RAND12 - T0 RAND12 - T1

Brain disease 46,0 0,80 Brain disease 45,5 0,82

Movement related illness 46,3 0,95 Movement related illness 47,6 0,83

Chronic pain 45,5 1,00 Chronic pain 45,5 1,03

RAND12 - T0 95 % CI RAND12 - T1 95 % CI

Brain disease 44,4 47,6 Brain disease 43,9 47,1

Movement related illness 44,4 48,1 Movement related illness 46,0 49,3

Chronic pain 43,5 47,4 Chronic pain 43,4 47,5

Discussion

In this paper, levels and differences in healthcare costs and productivity loss are evaluated for the chronically ill and/or physically disabled who undergo a rehabilitation treatment and are referred to a sports counselling centre to participate in the RSE programme. Patients’ wellbeing is assessed using physical and mental health scores by scoring the RAND-12 form. A subgroup analysis determines whether heterogeneity in the study population is relevant in explaining healthcare costs, productivity loss or HR-QoL. Multivariate OLS models are provided to assess the marginal effects of patient characteristics and lifestyle on healthcare costs (differences).

Between T0 and T1, all healthcare costs decline in accordance to the hypothesis that patients use fewer healthcare when diagnosed and discharged from rehabilitation treatment. Home care costs do not decrease and a possible explanation could be due to a more fixed nature of home care. In the subgroup analysis, levels and changes of healthcare costs are evaluated and are found to be the same for patients with different characteristics or medical condition. Productivity loss declines at T1 and depends significantly on a patient’s medical condition. Changes in physical and mental health score are not relevant from a clinical point of view. Physical and mental health scores are different for patients with other medical conditions.

16

cost approach of rehabilitation treatment in the mentioned paper contributes to a majority

difference in healthcare costs (€ - 6852) between this and before mentioned paper. 31 In this paper, rehabilitation treatment cost involves a large share (54%) of total healthcare costs, therefore total healthcare costs are sensitive to unit cost of rehabilitation treatment. A similar percentage of

rehabilitation cost in total healthcare cost is found in the before mentioned paper (48%). 31 Secondly, current paper reports significantly higher absenteeism (€ 28.721 at T0) for patients with a brain disease, of which 68% incurred a stroke compared to 100% stroke patients in the study of van Eeden. 31

No conclusive answer can be found to explain the difference in productivity loss outcomes as both papers use the friction cost approach. This paper found that brain disease had higher productivity loss than movement related illness and chronic pain and can be explained by a lower cognitive and working ability when suffering from a brain disease, compared to movement related illness and chronic pain.

In the subgroup analysis, level of total healthcare cost at T0 is related to age, which is in accordance with the literature. 26 Level of total healthcare cost at T1 and difference compared to T0 is not influenced by age, gender, being overweight, smoking behaviour and education level which is in contrast with previous literature.26-29 A possible explanation might be that mentioned studies have more observatory power and variation in the patient groups to evaluate marginal effects more precisely. Difference in healthcare cost of non-referable required care between T0 and T1 was greater for smokers for which no satisfying explanation can be found. In this paper, further

confirmation is provided that chronic conditions are associated with less workforce participation and income. 1-3 In existing studies, a constant or marginal increase in HR-QoL after rehabilitation is found and current study concludes the same. 18-21 The unique setting of this paper is the incorporated physical activity promoting programme in rehabilitation treatment. Still, HR-QoL is positively associated with physical activity, yet indicators to investigate such an association was absent in this paper.

In this paper, changes in RAND-12 physical and mental health scores are not clinically relevant (<3). RAND-12 is a non-disease specific form of HR-QoL but is less sensitive to changes over time than disease specific HR-QoL forms. In this way, the RAND-12 is a better predictor of HR-QoL between different medical condition groups for a given intervention than within one single medical condition group. Possible improvements in HR-QoL might not have fully materialized yet in the follow-up period of 14 weeks. Further, adhering to absolute minimum clinical important changes might overlook important effects as quality of life improvement is ought to be compared to the

17

have a significantly better physical health score than movement related illness and chronic pain sufferers. While the literature on physical and mental health scores for multiple conditions is slim, it is concluded that the type of medical condition affects HR-QoL. Poorest HR-QoL is found in

neurological and musculoskeletal conditions. 42 This paper concludes that physical health scores of before mentioned conditions are indeed low, yet further lower for movement related illness and chronic pain in comparison to brain disease. Different results in HR-QoL may be found when

controlling for gender, education and living situation. While this paper suggests age does not explain different physical and mental health scores, controlling for education and living situation might result in other conclusions. 43

Current study also has some limitations. Patient questionnaires suffer from recall bias. However, recall bias is limited if the cognitive ability to remember stayed constant, which can be argued given the short follow-up period. In that case, changes in healthcare costs will be unbiased and levels at T0 and T1 will be biased by the systematic error. Overall, healthcare costs are close to the true costs as the patients report utilisation and prices were largely obtained via the Manual of Costing. Recently, a new Manual of Costing was published, yet there were no large differences in the costs used for this paper. Secondly, productivity cost at T0 is a good representation of the relative constant cost period post-diagnosis and rehabilitation period. Productivity costs at T1 may have been underestimated. Patients are discharged at T0 and a negative trend in productivity loss is expected throughout the period between T0 and T1, while extrapolation took the latter 4 weeks as a proxy for the period between T0 and T1. This paper confirms that productivity costs are an important aspect in a societal perspective. A majority of absenteeism cost is incurred at T0 and return to a lower level at T1. External validity may be a drawback in this paper. It was difficult for the sport counselling centres to include a good representation of the non-Dutch born population due to language barriers. Therefore, before mentioned could explain that current study population has a higher proportion of patients with a paid job (44% vs 26%) as insufficient knowledge on the Dutch language does not facilitate paid work jobs. 4 Yet, this paper also has strong aspects in the study population. This paper includes a multicentered approach to account for centre and regional specific differences. Another strong aspect is the amount of observations with heterogeneity among patients where the specific personal characteristics, lifestyle and medication condition allows to estimate marginal effects on outcome variables.

18

and results in no economical relevant difference. Yet, studies show that not imputing missing data lead to less efficient results compared to complete-case analysis (see appendix). 46

Improvements in this paper could have been made by a longer study period and constructing control groups on both receiving rehabilitation and RSE treatment or not in order to provide policy makers with a higher level of evidence44. In the current setting, this paper could have been improved by identifying patients that have higher exposure to the RSE programme but such indicators are debatable. However, the literature suggests that patients with a chronic condition and/or a physical disability are a good target group for physical activity. Yet, endogeneity may exist between physical activity and patient’s health conditions and is left for further research.

Overall, this paper provides important findings. Healthcare costs decline in a similar fashion after rehabilitation when controlled for patient characteristics, lifestyle and medical condition.

Heterogeneity is observed among different medical conditions in productivity loss and HR-QoL.

Conclusion

19

‘do nothing’ case in conducting a full cost and benefit analysis of the Sports and Exercise part of the Rehabilitation, Sports and Exercise programme.

20

References:

1. Suhrcke M, Nugent RA, Stuckler D, Rocco L. Chronic disease: An economic perspective. The Oxford

Health Alliance. 2006.http://www.sehn.org/tccpdf/Chronic%20disease%20economic%20persp

ective.pdf. Accessed January 13, 2016.

2. Currie J, Madrian BC. Health, health insurance and the labor market. Handbook of labor

economics. 1999; 3: 3309-3416.

3. Heijmans M, Spreeuwenberg P, Rijken M. Ontwikkelingen in de zorg voor chronisch zieken: Rapportage 2010. NIVEL. 2010. http://www.nivel.nl/sites/default/files/bestanden/Rapport-ontwikkelingen-in-zorg-chronisch-zieken.pdf. Accessed January 13, 2016.

4. Putter ID, Cozijnsen R, Rijken M. Het vergroten van arbeidsparticipatie onder mensen met een chronische ziekten of beperking; een werkwens alleen is niet voldoende. NIVEL. 2015.

http://www.nivel.nl/sites/default/files/bestanden/Facstheet-werkwens.pdf. Accessed January 13, 2016.

5. Van den Heuvel SG, Swenne G, Geuskens GA, Hooftman WE, Koppes LL, Van den Bossche, Seth NJ. Productivity loss at work; health-related and work-related factors. J Occup Rehabil.

2010;20(3):331-339.

6. Cecchini M, Sassi F, Lauer JA, Lee YY, Guajardo-Barron V, Chisholm D. Tackling of unhealthy diets, physical inactivity, and obesity: Health effects and cost-effectiveness. The Lancet.

2010;376(9754):1775-1784.

21

8. Sullivan PW, Morrato EH, Ghushchyan V, Wyatt HR, Hill JO. Obesity, inactivity, and the prevalence of diabetes and diabetes-related cardiovascular comorbidities in the U.S., 2000-2002. Diabetes Care. 2005;28(7):1599-1603.

9. Nocon M, Hiemann T, Muller-Riemenschneider F, Thalau F, Roll S, Willich SN. Association of physical activity with all-cause and cardiovascular mortality: A systematic review and meta-analysis.

Eur J Cardiovasc Prev Rehabil. 2008;15(3):239-246.

10. Yoon J, Zulman D, Scott JY, Maciejewski ML. Costs associated with multimorbidity among VA patients. Med Care. 2014;52 Suppl 3:S31-6.

11. Bähler C, Huber CA, Brüngger B, Reich O. Multimorbidity, health care utilization and costs in an elderly community-dwelling population: A claims data based observational study. BMC health

services research. 2015;15(1):23.

12. Rimmer JH. Getting beyond the plateau: Bridging the gap between rehabilitation and community-based exercise. PM&R. 2012;4(11):857-861.

13. Ang DC, Kaleth AS, Bigatti S, et al. Research to encourage exercise for fibromyalgia (REEF): Use of motivational interviewing, outcomes from a randomized-controlled trial. Clin J Pain. 2013;29(4):296-304.

14. Plotnikoff R, Karunamuni N, Courneya K, Sigal R, Johnson J, Johnson S. The alberta diabetes and physical activity trial (ADAPT): A randomized trial evaluating theory-based interventions to increase physical activity in adults with type 2 diabetes. Annals of Behavioral Medicine. 2013;45(1):45-56.

22

16. van der Ploeg HP. Promoting physical activity in the rehabilitation setting. Wageningen, the

Netherlands: Ponsen & Looijen BV; 2006.

17. van den Berg-Emons RJ, Rita J, Bussmann JB, Stam HJ. Accelerometry-based activity spectrum in persons with chronic physical conditions. Arch Phys Med Rehabil. 2010;91(12):1856-1861.

18. Yohannes AM, Doherty P, Bundy C, Yalfani A. The long-term benefits of cardiac rehabilitation on depression, anxiety, physical activity and quality of life. J Clin Nurs. 2010;19(19-20):2806-2813.

19. Schag C, Ganz PA, Heinrich RL. CAncer rehabilitation evaluation system–short form (CARES‐SF). A cancer specific rehabilitation and quality of life instrument. Cancer. 1991;68(6):1406-1413.

20. Oldridge N, Guyatt G, Jones N, et al. Effects on quality of life with comprehensive rehabilitation after acute myocardial infarction. Am J Cardiol. 1991;67(13):1084-1089.

21. Wijkstra PJ, Van Altena R, Kraan J, Otten V, Postma DS, Koeter GH. Quality of life in patients with chronic obstructive pulmonary disease improves after rehabilitation at home. Eur Respir J. 1994;7(2):269-273.

22. Rejeski WJ, Brawley LR, Shumaker SA. Physical activity and health-related quality of life. Exerc

Sport Sci Rev. 1996;24(1):71-108.

23. Anokye NK, Trueman P, Green C, Pavey TG, Taylor RS. Physical activity and health related quality of life. BMC Public Health. 2012;12:624-2458-12-624.

24. Bize R, Johnson JA, Plotnikoff RC. Physical activity level and health-related quality of life in the general adult population: A systematic review. Prev Med. 2007;45(6):401-415.

23

26. Schneider EL, Guralnik JM. The aging of america: Impact on health care costs. JAMA. 1990;263(17):2335-2340.

27. Wolf AM, Colditz GA. Current estimates of the economic cost of obesity in the united states.

Obes Res. 1998;6(2):97-106.

28. Barendregt JJ, Bonneux L, van der Maas PJ. The health care costs of smoking. N Engl J Med. 1997;337(15):1052-1057.

29. Hansen AH. A population-based study of health care utilisation according to care level, socio-economic group, and continuity of primary care. The Tromsø study. 2013. http://munin.uit.no/

bitstream/ handle/10037/5556/thesis.pdf?sequence=6. Accessed January 13, 2016.

30. van Tulder MW, Koes BW, Bouter LM. A cost-of-illness study of back pain in the Netherlands.

Pain. 1995;62(2):233-240.

31. van Eeden M, van Heugten C, van Mastrigt GA, van Mierlo M, Visser-Meily JM, Evers SM. The burden of stroke in the Netherlands: Estimating quality of life and costs for 1 year poststroke. BMJ

Open. 2015;5(11):e008220-2015-008220.

32. Alingh RA, Hoekstra F, van der Schans CP, Hettinga FJ, Dekker R, van der Woude LH. Protocol of a longitudinal cohort study on physical activity behaviour in physically disabled patients participating in a rehabilitation counselling programme: ReSpAct. BMJ Open. 2015;5(1):e007591-2015-007591.

33. Hoekstra F, Hettinga FJ, Alingh RA, et al. The current implementation status of the integration of sports and physical activity into dutch rehabilitation care. Disabil Rehabil. 2015:1-6.

24

35. Hakkaart L, Koopmanschap M, Krol M, Severens H, Brouwer W. Manual of the iMTA Medical Cost Questionnaire (iMCQ). Rotterdam: Institute for Medical Technology Assessment

(iMTA), Erasmus University Rotterdam. 2013.

36. Bouwmans C, Hakkaart-van Roijen L, Koopmanschap M, Krol M, Severens H, Brouwer W. Manual of the iMTA Productivity Cost Questionnaire (iPCQ). Rotterdam: Institute for Medical

Technology Assessment (iMTA), Erasmus University Rotterdam. 2013.

37. Severens JL, Mulder J, Laheij RJ, Verbeek AL. Precision and accuracy in measuring absence from work as a basis for calculating productivity costs in the netherlands. Soc Sci Med. 2000;51(2):243-249.

38. Ware J, Kosinski M, Keller S. SF-36 Physical and Mental Health Summary Scales: A User's Manual. Boston, MA: The Health Institute; 1994

39. Hakkaart L, Tan SS, Bouwmans CAM. Handleiding voor kostenonderzoek. Methoden en standaard kostprijzen voor economische evaluaties in de gezondheidszorg. 2010:30-11-2015.

http://www.scribd.com/doc/74251141/handleiding-kostenonderzoek-2010#scribd. Accessed January 13,2016.

40. Samsa G, Edelman D, Rothman ML, Williams GR, Lipscomb J, Matchar D. Determining clinically important differences in health status measures. Pharmacoeconomics. 1999;15(2):141-155.

41. Hays RD, Morales LS. The RAND-36 measure of health-related quality of life. Ann Med. 2001;33(5):350-357.

25

43. Sprangers M, Schwartz CE. Reflections on changeability versus stability of health-related quality of life: Distinguishing between its environmental and genetic components. Health and Quality of Life

Outcomes.2008;6:89

44. Krumholz HM. Outcomes research: Generating evidence for best practice and policies.

Circulation. 2008;118(3):309-318.

45. Rubin DB. Multiple imputation after 18 years. Journal of the American statistical Association. 1996;91(434):473-489.

46. White IR, Royston P, Wood AM. Multiple imputation using chained equations: Issues and guidance for practice. Stat Med. 2011;30(4):377-399.

47. Sterne JA, White IR, Carlin JB, et al. Multiple imputation for missing data in epidemiological and clinical research: Potential and pitfalls. BMJ. 2009;338:b2393.

48. Vroomen JM, Eekhout I, Dijkgraaf MG, et al. Multiple imputation strategies for zero-inflated cost data in economic evaluations: Which method works best? The European Journal of Health Economics. 2015:1-12.

26

Appendices

Appendix 1: Correlation matrix independent variables

Correlation matrix

Variable Age Gender Overweight Smoker Education

Age 1

Gender -0,257 1

Overweight 0,142 -0,07 1

Smoker -0,1249 -0,048 -0,0057 1

Education -0,1708 0,0638 -0,1547 -0,0847 1

Appendix 2: Introduction, application and discussion on multiple

imputation

Multiple imputation (MI) was firstly described in Rubin (1996) and dealt with the problem of missing data.45 Earlier studies concluded that using MI techniques led to a significant improvement of statistical inference compared to complete-case analysis.46 Complete-case analysis had serious drawbacks if the observed data was not representative for the entire study sample. This section was meant to introduce MI to the reader as well as an evaluation of the possible drawbacks of MI

reported by Sterne (2009).47

The key to a successful implementation of MI was to use proper imputation assumption and methods. When using MI, one should have made the assumption of data being missing at random. To clarify, neither an observed nor an unobserved variable possessed explanatory power in

predicting an observation being missing or not. In this paper, missing data of T1 was explained by not receiving the paper questionnaire (yet), loss of follow-up or missing one smaller part of total costs. It was assumed that no observed or unobserved variable could powerfully predict whether this is the case. In this paper, an average of 30% cost data of T1 was missing. Heterogeneity in the level of missing data per healthcare service costs was reported, mainly due to missing one or more parts of the total cost variable. Missing data of non-referable required care was close to 50% where the smaller parts of total costs were causing the additional missing data. Imputing on a lower cost level would have solved the additional missing data, though such an approach would have tripled the time and computational resources required by MI.

27

imputation regression model. In this paper, healthcare costs at T0 were fully observed and entered the imputation model in addition to age, gender, being overweight, smoking behaviour and

education level. It was concluded that before mentioned variables had predictory power on the value of the missing variable. A possible limitation of the current approach might be that a very recent study recommended using the natural logarithmic form of healthcare costs in the multiple imputation model when amount of missing data exceeded 35%.48 _{Unfortunately, this}

recommendation and study publication came late in the MI process to be taken into account. Secondly, predictive mean matching was used to account for the distribution of healthcare costs. Once a missing variable was imputed, the ten closest observed costs (called neighbours) were identified of which a random one was picked. In this way, all imputed costs were positive and took the distribution of the observed cost data into account. It was found that an ordinary multiple imputation method seriously biased the results by ignoring the observed cost distribution. Moreover, a comparison of the imputed results with the complete case analysis was done to confirm raw imputation validity. There was no significant economic difference between the multiple imputation and complete case method. This finding might suggest that 50% missing data was not always a serious obstacle for statistical inference.

28

Appendix 3: Costs and sources

COSTS PER UTILISATION UNIT AND

SOURCES

Healthcare service Price Source Healthcare service Price Source

Non-referral required care Hospital care

General practitioner 31,1 (1) Medical doctor 79,1 (1)

Social counselor 71,5 (1) Inpatient treatment hospital 502,3 (1)

Physical therapist 39,0 (1) Rehabilitation care

Ergo therapy 24,2 (1)

Outpatient rehabilitation

centre 120,9 (1)

Speech therapist 36,3 (1) Medical testing and scans

Dietitian 29,7 (1) X-Ray 55,0 (1)

Homoeopathist 69,1 (2) MRI 275,9 (2)

Psychologist 96,7 (1) CT 275,9 (2)

Company doctor 227,0 (2) PET 1.425,5 (2)

Emergency department 166,0 (1) Ultrasound 82,0 (2)

Ambulance transport 363,8 (1) Isotope scan 192,2 (2)

Formal care at home SPECT 284,4 (2)

Household chores 27,8 (1) Blood testing 13,9 (1)

Care of patient 48,9 (1) Urine sample test 13,9 (1)

Nursing of patient 72,3 (1) Activity test 110,0 (2)

Informal care at home Biopsy 75,0 (2)

Household chores 13,7 (1) Sources:

Care of patient 13,7 (1)

(1) Dutch Manual of Costing