Cost-effectiveness of dedicated dietitians for hyperphosphatemia management among hemodialysis patients in Lebanon: results from the Nutrition Education for Management of Osteodystrophy trial

Abstract

Aim: To assess the cost-effectiveness of nutrition education by dedicated dietitians (DD) for hyperphosphatemia management among hemodialysis patients.

Materials and methods: This was a trial-based economic evaluation in 12 Lebanese hospital-based units. In total, 545 prevalent patients were cluster randomized to DD, trained hospital dietitian (THD), and existing practice (EP) groups. During Phase I (6 months), DD (n = 116) received intensive education by DD trained on renal nutrition, THD (n = 299) received care from trained hospital dietitians, and EP (n = 130) received usual care from untrained hospital dietitians. Patients were followed-up during Phase II (6 months).

Results: At baseline, EP had the lowest weekly hemodialysis time, and DD had the highest serum phosphorus and malnutrition-inflammation score. The additional costs of the intervention were low compared with the societal costs (DD: $76.7, $21,007.7; EP: $4.6, $18,675.4; THD: $17.4, $20,078.6, respectively). Between Phases I and II, DD showed the greatest decline in services use and societal costs (DD: –$2,364.0; EP: –$1,727.7; THD: –$1,105.7). At endline, DD experienced the highest decrease in adjusted serum phosphorus (DD: –0.32; EP: +0.16; THD: +0.04 mg/dL), no difference in quality-adjusted life-years (QALY), and the highest societal costs. DD had a cost-effectiveness ratio of $7,853.6 per 1 mg decrease in phosphorus, compared with EP; and was dominated by THD. Regarding QALY, DD was dominated by EP and THD. The results were sensitive to changes in key parameters.

Limitations: The analysis depended on numerous assumptions. Interpreting the results is limited by the significant baseline differences in key parameters, suggestive of higher baseline societal costs in DD.

Conclusions: DD yielded the greatest effectiveness and decrease in societal costs, but did not affect QALY. Regarding serum phosphorus, DD was likely to be cost-effective compared with EP, but had a low cost-effectiveness probability compared with THD. Regarding QALY, DD was not likely to be cost-effective. Assessing the long-term cost-effectiveness of DD, on similar groups, is recommended.

Keywords:

• Cost-effectiveness analysis
• Dietitians
• Patient education
• Hyperphosphatemia
• Hemodialysis

Introduction

Elevated serum phosphorus is a serious consequence of chronic kidney disease. It is associated with increased patient morbidity¹, mortality^2–5, and high healthcare costs⁶. International clinical practice guidelines recommend “normalizing” serum phosphorus of hemodialysis patients through dialysis, phosphate-binding medications, and low phosphorus diet^7,8. However, more than a decade after the first guidelines were issued, hyperphosphatemia remains a common condition among renal patients, especially those treated with hemodialysis, with nearly one in two patients being hyperphosphatemic^9,10.

Economic evaluations are a useful tool to identify the economic value of an intervention and guide decision-making in healthcare. Up-to-now, efforts were focused on one intervention in the field of hyperphosphatemia management, which is phosphate binders¹¹. Despite the mounting evidence on the clinical effectiveness of nutrition education¹², the cost-effectiveness of this intervention has not been addressed¹¹.

The most recent local evidence on the effectiveness of dietetic interventions for hyperphosphatemia management among prevalent hemodialysis patients emanates from the Nutrition Education for Management of Osteodystrophy (NEMO) trial¹³ Results from the NEMO trial showed that intensive nutrition education by dedicated dietitians targeting hyperphosphatemia management was superior to the other protocols in reducing serum phosphorus, without compromising the nutritional status of the patients^14–16. Assessing the value for money of dedicated dietitians would be very important for helping decision-makers efficiently allocate scarce resources devoted to hemodialysis patients.

The aim of this study was, therefore, to conduct an economic evaluation (including a cost-effectiveness analysis (CEA) and a cost-utility analysis (CUA)), from the societal perspective of the nutrition education provided by dedicate dietitians for hemodialysis patients in Lebanon, in comparison with the existing practice in the country and to another proposed alternative.

Methods

We used individual patient-level data from the NEMO trial to perform the CEA and CUA, and followed international guidelines for the analysis and reporting of economic evaluations^17,18. A societal perspective was used, with a time horizon of 1 year.

Trial

Design

The design, methods, and clinical results of the NEMO trial are detailed elsewhere^13–16. The NEMO trial was conducted in Lebanon, where all hemodialysis units are hospital-based. The NEMO trial compared three protocols of nutrition education for hyperphosphatemia management among hemodialysis patients. In brief, 12 hospital-based hemodialysis units (n = 570) were randomly assigned to cluster A (6 units; n = 271) and cluster B (6 units; n = 299). Cluster A patients were then equally assigned according to their hemodialysis shifts into two protocols: dedicated dietitian (DD) (n = 133) and existing practice (EP) (n = 138); cluster B patients were assigned to the trained hospital dietitian (THD) protocol. Participants were followed for 12 months, and measures were collected at three time-points: t-0 (beginning of month 1), t-1 (end of month 6) and t-2 (end of month 12). The NEMO trial received ethical approval from the ethical committees of participating hospital-based hemodialysis units, and was conducted according to the Declaration of Helsinki.

Subjects

Consenting patients, adult, receiving dialysis for at least 6 months, free of cancer, infection with human immunodeficiency virus, and hepatitis C were considered.

Interventions

Phase I (t-0 to t-1: 6 months):

• The DD group received intensive nutrition education during hemodialysis sessions by dietitians who were trained on renal nutrition and dedicated to the hemodialysis units. The nutrition education in this group targeted hyperphosphatemia management using a renal-oriented culturally-validated educational tool based on the transtheoretical model of behavioral change¹³. It was delivered on a basis of 2 h per month. The existing practice was not compromised in DD units, and hospital dietitians were left to provide their usual care.

• The EP group received the usual care by their hospital dietitians, i.e. patient consultations delivered only following nephrologists’ consult requests. This group represented the existing practice in Lebanon, where dietitians have limited knowledge about renal nutrition, and are not dedicated to hemodialysis patients.

• The THD group received care by their hospital dietitians, as per the latter’s availability; however, in this group, the dietitians were equally trained on renal nutrition as the dedicated dietitians, but were not devoted to hemodialysis patients in particular. This group represented an alternative to having a dedicated dietitian; whereby the hospital dietitians’ education is ensured, but the dietitian-to-patient frequency of consultations or contact time was not set.

Phase II (t-1 to t-2: 6 months) was a follow-up period.

Economic evaluation

Outcomes

The outcome of the CEA was serum phosphorus, and that of the CUA was quality-adjusted life-year (QALY).

Serum phosphorus was retrieved from the patients’ medical charts, and its 6-month mean values were calculated at t-0, t-1, and t-2.

QALY incorporates the quality of a health state (quality-of-life: QoL) with the duration of survival (life-years: LYs) using a multiplicative formula. It is the preferred and most common outcome in economic evaluation¹⁹. QoL was measured with the Short Form (SF)-36 questionnaire, assessing the two dimensions of physical and mental health over the past 30 days²⁰. Patients completed the validated culturally-specific pilot-tested version of the SF-36^13,21 at the three time-points.

Costs

Cost data were collected at t-1 and t-2, following the three steps of costs identification, measurements, and valuation, as detailed in Rizk et al.²².

Costs identification: included costs were categorized as (1) healthcare sector costs (cost of the intervention (nutrition education), costs associated with hemodialysis, emergency hemodialysis, healthcare professional consultation, hospitalization, medications, and integrated home care); (2) costs to patient and family (caregiver costs and productivity losses); and (3) costs in other sectors (travel costs)¹⁷.

Costs measurement: using a pilot-tested resource utilization questionnaire (Appendix 1) adapted to Lebanese hemodialysis patients.

Costs valuation: the costs were gathered and calculated in Lebanese pounds (LBP), converted to US$(1 US$ = 1,507.5 LBP; year of reference: 2011)²³ and uprated to 2015US$ using Consumer Price Indices (Index, 2010 = 100)²⁴. A macro-costing valuation was applied, whereby composite intermediate resources were identified and measured. The mean reported costs incurred by patients were used to value the costs of emergency hemodialysis, consultations with healthcare professionals, hospitalizations, and professional care. The costs of drugs were derived from the Lebanese National Drug Index²⁵. The valuation of informal care was based on the proxy good method²⁶, and the valuation of productivity losses was based on the human capital approach²⁷.

The costing of the intervention (nutrition education) in each group is detailed in Appendix 2.

The reported quantity/frequency of use of each service was multiplied by its respective mean cost to obtain the total costs. Discounting was not applied, because the analysis was limited to 1 year.

Preparation of data

Serum phosphorus was adjusted for baseline differences between the three groups, following a regression method controlling for baseline serum phosphorus, age, gender, weekly hemodialysis time, and malnutrition-inflammation score (MIS), as detailed by van Mastrigt et al.²⁸. For the base-case CEA, missing values of patients who withdrew during Phase I were replaced by their baseline value minus half of the difference in serum phosphorus between t-0 and t-1 of participants from the same unit and study group (peers), assuming that the patients attrited at mid-point of the phase and that evolution of serum phosphorus was linear. For Phase II, missing values were replaced by the patients’ values at t-1 minus the difference in serum phosphorus between t-1 and t-2 of their peers.

For the CUA, SF-36 QoL data were converted to utility²⁹. Mean utility between t-0 and t-1; and between t-1 and t-2 was retained for Phase I and Phase II, respectively. Missing QoL data were imputed by last observation carried forward, when available, or by the mean of the peers. Regarding LYs, for each phase survived, patients were allocated 0.5 LYs. Patients not surviving the phase due to mortality or extended hospitalization were allocated 0.25 LYs, assuming that they withdrew at the mid-point of the study phase. The LYs of those who were transferred, got a transplant, or withdrawal were imputed by the mean of their peers. QALYs were the product of utility and LYs. Total QALYs were obtained by summing QALYs at each phase.

As for the costs, the missing values of patients who survived the phase were imputed by the mean of their peers. For those who died or were extensively hospitalized, their values were imputed by half of the mean costs of their peers, for the same above-mentioned reasoning. Those who attrited for other reasons were attributed the mean costs of their peers in parallel to their LYs.

Analyses

Analyses were conducted using SPSS. Descriptive statistics were generated. Chi-square and Student’s T-test were used to determine the baseline between-group differences for categorical and continuous data, respectively. One-way ANOVA followed by Bonferroni Post-Hoc were used to assess the between-group differences in costs at t-1 and t-2; p < .05 was used for significance.

To determine the cost-effectiveness and cost-utility of DD, incremental cost-effectiveness and cost-utility ratios (ICERs, ICURs) were calculated. The ICER or ICUR is a ratio that compares the additional costs and effects of the assessed intervention with the control. The cost-effectiveness and cost-utility of DD vs EP/THD were calculated as the difference in mean costs (C) divided by the difference in mean effects (E) (serum phosphorus and QALY, respectively) at t-2:

DD vs EP = (C_DD – C_EP)/(E_DD – E_EP); DD vs THD = (C_DD – C_THD)/(E_DD – E_THD).

Sensitivity analyses

Three one-way sensitivity analyses were performed to test the robustness of the results of the base-case. The first analysis assessed the effects of excluding the costs of the maintenance hemodialysis and transportation, following the reasoning of Grima et al.³⁰. The second analysis examined the impact of imputation, thus included complete cases only. The third analysis was carried out only on hyperphosphatemic patients (phosphorus >5.5 mg/dL) at baseline, as they are at higher risks for morbi-mortality, and are suggested to have higher costs than normophosphatemic patients. In the final analysis, the mean drop in unadjusted serum phosphorus between t-2 and t-0 (delta adjustment) was used for the CEA. For the CUA, we used the regression-based adjusted QALYs. In this analysis, QALYs were adjusted for baseline differences in utility between the three groups. We resorted to this adjustment because QoL is associated with clinical outcomes among hemodialysis patients, including hospitalization and mortality^20,31.

Non-parametric bootstraps were conducted to assess the stochastic uncertainty in the data using the bootstrapping technique in Excel, where the original sample was re-sampled, resulting in 5,000 simulated ICERs/ICURs per scenario. Cost-effectiveness/utility acceptability curves (CEACs/CUACs) were plotted using the probability estimates of DD’s cost-effectiveness (compared with each of the other interventions) over a range of willingness-to-pay (WTP) thresholds. The latter were defined as the amount of money the society is willing to pay to gain one unit of effect. In Lebanon, the value the society is willing to pay to gain one QALY is not defined. Accordingly, we used several thresholds ranging from three times the Gross Domestic Product in Lebanon ($31,272.84)—threshold suggested by the World Health Organization^32,33, to £30,000 ($43,612.49)—explicit threshold adopted in the UK³⁴, and $50,000—implicit threshold adopted in the US³⁵.

Results

Patients’ characteristics

The flow diagram of the trial is detailed in Appendix 3. At baseline, the three groups were comparable with respect to sociodemographic characteristics, except for employment status. Weekly hemodialysis time was significantly lower in the EP group. The DD group had the highest mean serum phosphorus, which was above the recommended range⁷. Also, the DD group had the highest mean MIS, denoting the worst malnutrition-inflammation status (Table 1).

Table 1. Participants’ characteristics (n = 545).

	DD (n = 116)	EP (n = 130)	THD (n = 299)
Male (%)	66 (56.9)	72 (55.4)	182 (60.9)
Social status (%)
Single	26 (22.4)	18 (13.8)	44 (14.7)
Married	83 (71.6)	102 (78.5)	242 (80.9)
Other	7 (6.0)	10 (7.7)	13 (4.3)
Employed (%)*	38 (32.8)	24 (18.5)	99 (33.1)
Educational level (%)
Illiterate	23 (19.8)	28 (21.5)	79 (26.4)
Read and write	17 (14.7)	22 (16.9)	33 (11.0)
Elementary	38 (32.8)	36 (27.7)	108 (36.1)
High school	24 (20.7)	24 (18.5)	43 (14.4)
University	14 (12.1)	20 (15.4)	36 (12.0)
Co-morbidities (%)
Diabetes	39 (36.1)	50 (38.5)	99 (33.2)
Hypertension	77 (69.4)	83 (63.8)	203 (68.1)
Cardiovascular diseases	30 (27.8)	27 (20.8)	62 (20.9)
Other*	29 (40.3)	22 (25.6)	17 (16.3)
Age (years) (mean (SD))	57.56 (15.18)	59.96 (15.22)	60.51 (14.94)
Utility (mean (SD))	0.56 (0.06)	0.56 (0.06)	0.55 (0.06)
Vintage (months) (mean (SD))	61.02 (64.39)	46.94 (48.62)	57.30 (53.06)
HD time (hours/week) (mean (SD))	10.64 (2.49)‡	9.53 (2.77)‡§	11.14 (1.45)§
Sodium^† (mEq/L) (mean (SD))	59.61 (2.41)	58.88 (4.71)‡	59.88 (0.64)‡
Potassium^† (mEq/L) (mean (SD))	1.34 (0.22)	1.35 (0.22)	1.33 (0.20)
Phosphorus^† (mg/dL) (mean (SD))	5.57 (1.52)‡	5.39 (1.49)	5.16 (1.44)‡
Calcium^† (mg/dL) (mean (SD))	8.66 (0.80)‡	8.62 (0.97)	9.11 (1.57)‡
CaP^† (mg^2/dL^2) (mean (SD))	48.37 (14.43)	46.65 (13.56)	47.11 (15.09)
PTH^† (pg/mL) (mean (SD))	400.85 (457.61)	377.52 (360.64)	344.89 (338.50)
Albumin^† (g/L) (mean (SD))	38.76 (3.77)	39.42 (4.28)	39.74 (4.36)
MIS (mean (SD))	7.30 (3.31)‡	6.55 (3.17)	6.07 (3.90)‡
BMI (Kg/m^2) (mean (SD))	25.10 (5.69)	25.43 (5.20)	24.31 (4.96)

† Pre-dialysis values;

* p < .05 indicates a between-group difference based on Chi square Test,

‡,§p < .05 indicates a between-group difference based on Bonferroni Post Hoc.

DD, dedicated dietitian; EP, existing practice; THD, trained hospital dietitian; SD, standard deviation; HD, hemodialysis; PTH, parathyroid hormone; CaP, calcium-phosphorus byproduct; MIS, malnutrition-inflammation score; BMI, body mass index.

Services use

During Phase I, the DD group consumed more services than other groups (especially emergency hemodialysis, specialist physician consults, several medications, and hospitalization). The THD group used more professional home care and lost most productivity hours. The EP group used more informal care. During Phase II, the DD group consumed a greater volume of emergency hemodialysis, specialist physician consults, sevelamer, and erythropoietin. The THD group consumed more calcium carbonate, professional home care, and lost most productivity hours (Table 2).

Table 2. Services use during the 12-month period.

	DD group (n = 107)			EP group (n = 112)			THD group (n = 257)
	Users	Mean	Median	Users	Mean	Median	Users	Mean	Median
PHASE I
Healthcare sector costs
HD
HD (session)	107 (100)	72.4	78.0	112 (100)	67.6	78.0	257 (100)	75.6	78.0
Emergency HD (session)	39 (36.4)	1.2	0.0	21 (18.6)	0.4	0.0	9 (3.5)	0.1	0.0
Healthcare professionals
Specialist physician (contact)	28 (26.2)	0.7	0.0	26 (23.2)	0.4	0.0	62 (24.1)	0.4	0.0
Dietitian (contact)	1 (0.9)	0.0	0.0	0 (0)	0.0	0.0	21 (8.2)	0.1	0.0
Psychologist (contact)	0 (0)	0.0	0.0	0 (0)	0.0	0.0	0 (0)	0.0	0.0
Medications
Calcium Carbonate (Tablet)	80 (74.8)	563.5	549.0	89 (79.5)	478.5	366.0	174 (67.7)	402.6	213.5
Calcium Acetate (Tablet)	43 (40.2)	288.0	0.0	39 (34.8)	236.0	0.0	108 (42.0)	299.6	0.0
Sevelamer (Tablet)	32 (30.0)	226.6	0.0	26 (23.2)	74.2	0.0	55 (21.4)	84.5	0.0
Cinacalcet (Tablet)	1 (0.9)	15.4	0.0	0 (0)	0.0	0.0	0 (0)	0.0	0.0
Active Vitamin D (Capsule 100 mL)	63 (58.9)	4.5	0.4	64 (57.1)	3.0	0.7	136 (52.9)	4.9	1.4
Vitamin D (Tablet)	33 (30.8)	15.4	0.0	26 (23.2)	9.2	0.0	55 (21.4)	8.7	0.0
Vitamin B complex (Injectable Solution)	71 (66.4)	47.9	72.0	60 (53.6)	26.1	15.1	97 (37.8)	19.1	0.0
Iron (Tablet)	36 (33.6)	51.3	0.0	36 (32.1)	36.4	0.0	59 (23.0)	40.0	0.0
Intravenous Iron (Injectable Ampoule)	72 (67.2)	13.6	12.0	78 (69.6)	13.7	12.0	146 (56.8)	9.2	6.0
Erythropoietin (Various)	85 (79.4)	170,247.5	192,000.0	80 (71.4)	138,900.0	144,000.0	198 (77.0)	131,901.4	96,000.0
Hospitalization (day)	48 (44.9)	4.6	0.0	39 (34.8)	2.8	0.0	82 (31.9)	2.3	0.0
Professional home care (hour)	3 (2.8)	0.1	0.0	3 (2.6)	0.0	0.0	5 (1.9)	12.4	0.0
Costs to patient and family
Informal care (Hour)	51 (47.7)	94.4	0.0	53 (47.3)	115.6	0.0	113 (44.0)	62.1	0.0
Productivity losses (Hour)	70 (65.4)	192.2	56.0	59 (52.7)	99.3	10.0	167 (65.0)	291.7	48.0
Costs in other sectors
Travel (Trip)	107 (100)	144.8	156.0	112 (100)	135.1	156.0	257 (100)	151.1	156.0
PHASE II	DD group (n = 93)			EP group (n = 93)			THD group (n = 217)
Healthcare sector costs
HD
HD (session)	93 (100)	72.7	78.0	93 (100)	67.7	78.0	217 (100)	75.5	78.0
Emergency HD (session)	27 (29.0)	0.7	0.0	17 (18.3)	0.3	0.0	9 (4.15)	0.1	0.0
Healthcare professionals
Specialist physician (contact)	35 (37.6)	0.8	0.0	26 (28.0)	0.4	0.0	60 (27.6)	0.5	0.0
Dietitian (contact)	0 (0)	0.0	0.0	1 (1.1)	0.0	0.0	1 (0.5)	0.0	0.0
Psychologist (contact)	0 (0)	0.0	0.0	0 (0)	0.0	0.0	0 (0)	0.0	0.0
Medications
Calcium Carbonate (Tablet)	58 (62.4)	479.0	366.0	65 (69.9)	491.8	366.0	102 (47.0)	339.0	0.0
Calcium Acetate (Tablet)	21 (22.6)	207.9	0.0	19 (20.4)	146.4	0.0	81 (37.3)	314.6	0.0
Sevelamer (Tablet)	12 (12.9)	110.7	0.0	6 (6.5)	50.3	0.0	29 (13.4)	109.6	0.0
Cinacalcet (Tablet)	0 (0)	0.0	0.0	2 (2.2)	6.9	0.0	5 (2.3)	14.3	0.0
Active Vitamin D (Capsule 100 mL)	46 (49.4)	4.5	0.0	38 (40.9)	3.4	0.0	99 (45.6)	3.8	0.0
Vitamin D (Tablet)	13 (14.0)	16.3	0.0	16 (17.2)	18.3	0.0	24 (11.1)	7.0	0.0
Vitamin B complex (Injectable Solution)	54 (58.1)	37.8	14.2	42 (45.2)	30.6	0.0	75 (34.6)	16.9	0.0
Iron (Tablet)	22 (23.7)	37.8	0.0	19 (20.1)	24.2	0.0	38 (17.5)	34.4	0.0
Intravenous Iron (Injectable Ampoule)	57 (61.3)	13.8	12.0	58 (62.4)	10.6	12.0	134 (61.8)	10.5	6.0
Erythropoietin (Various)	87 (93.5)	168,258.1	192,000.0	81 (87.1)	121,633.5	96,000.0	189 (87.9)	1,451,006.0	96,000.0
Hospitalization (Day)	38 (40.8)	2.4	0.0	20 (21.5)	2.5	0.0	58 (26.7)	2.2	0.0
Professional home care (Hour)	6 (6.5)	3.2	0.0	2 (2.2)	0.0	0.0	8 (3.7)	12.0	0.0
Costs to patient and family
Informal care (Hour)	53 (55.2)	322.3	144.0	49 (52.7)	302.2	30.5	84 (38.7)	186.3	0.0
Productivity losses (Hour)	56 (60.2)	206.5	16.0	55 (59.1)	156.4	12.0	146 (67.6)	349.0	208.0
Costs in other sectors
Travel (Trip)	93 (100)	145.4	156.0	93 (100)	135.3	156.0	217 (100)	151.0	156.0

DD, dedicated dietitian; EP, existing practice; THD, trained hospital dietitian; HD, hemodialysis.

Between Phase I and Phase II, the DD group showed the greatest decline in the use of the most costly healthcare services. The volume of emergency hemodialysis in this group decreased by 44% compared with 18.6% and 16.8% in the EP and THD groups, respectively. The consumption of sevelamer decreased in the DD group by 51.1%, relative to 32.2% in the EP group, and to an increase of 29.8% in THD. Hospitalization days decreased by 48.4% in the DD group, compared with 11.9% and 2% in the EP and THD groups, respectively. The DD group also exhibited the slowest increase in productivity losses (+7.44% vs +57.6% and +19.6% for DD, EP, and THD, respectively).

Costs

The cost distribution of the three groups is displayed in Appendix 4. As per Table 3, the societal, healthcare, patient, and family-related costs differed between groups. The mean cost of the nutrition education was highest in the DD group; yet the cost of this intervention represented less than 1% of the societal costs of each of the three groups during both study phases.

Table 3. Costs during the 12-month period (US$2015).

	DD group (n = 116)		EP group (n = 130)		THD group (n = 299)
	Mean	Median	Mean	Median	Mean	Median
PHASE I
Healthcare sector costs
HD	7,023.6*	7,636.6	6,512.8*†	7,636.6	7,182.7†	7,636.6
Maintenance HD	6,911.0*	7,636.6	6,477.3*†	7,636.6	7,173.6†	7,636.6
Emergency HD	112.6*†	0.0	35.4*	0.0	9.2†	0.0
Healthcare professionals	31.3	0.0	26.4	0.0	20.8	0.0
Medications	1,938.9*†	1,898.5	1,331.6*	1,362.3	1,339.7†	1,279.5
Hospitalization	1,738.7*	0.0	1,564.6^	0.0	801.0*†	0.0
Professional home care	2.0	0.0	0.6	0.0	210.1	0.0
Total	10,463.9*†	9,899.3	8,877.8*	8,153.9	9,014.8†	8,740.8
Costs to patient and family
Informal care	155.0	20.2	195.0	20.2	118.1	40.3
Productivity losses	364.3	122.3	194.0*	34.9	493.5*	134.4
Total	510.3	307.1	388.9*	147.4	611.6*	285.0
Costs in other sectors
Travel	395.9	436.0	376.3*	436.0	421.5*	436.0
Intervention costs	74.5*†	76.9	2.6*‡	2.6	13.3^†‡	13.7
Total societal costs	11,724.2*†	10,697.3	10,203.8*	9,544.8	10,600.9†	10,102.3
PHASE II
Healthcare sector costs
HD	6,015.9	7,484.3	5,718.2*	6,385.6	6,423.3*	7,484.3
Maintenance HD	5,954.3	7,484.3	5,694.3*	6,385.6	6,416.5*	7,484.3
Emergency HD	61.6*†	0.0	23.9*	0.0	6.8†	0.0
Healthcare professionals	53.7*	0.0	13.1*	0.0	24.9	0.0
Medications	1,519.0	1,762.3	1,270.6	1,431.4	1,281.2	1,028.1
Hospitalization	587.7	0.0	554.6	0.0	490.4	0.0
Professional home care	61.2	0.0	0.5	0.0	140.4	0.0
Total	8,237.5	9,220.6	7,557.0	7,382.7	8,360.2	8,571.7
Costs to patient and family
Informal care	429.7*	32.5	367.4	2.9	228.9*	0.0
Productivity losses	352.2	38.5	216.6*	14.0	526.8*	119.8
Total	781.9	328.2	584.0	135.8	755.7	237.0
Costs in other sectors
Travel	338.7	427.3	333.1*	364.6	375.1*	427.3
Intervention costs	2.2*	2.6	2.1†	2.6	4.2*†	5.2
Total societal costs	9,360.2	10,270.0	8,475.2	8,709.8	9,495.2	9,656.5
TOTAL COSTS
Healthcare sector costs
HD	1,309.5	15,121.0	12,231.0*	13,911.9	13,606.0*	15,121.0
Maintenance HD	12,865.3	15,121.0	12,171.7*	13,871.7	13,590.0*	15,121.0
Emergency HD	174.2*†	67.0	59.3*‡	0.0	16.0^†‡	0.0
Healthcare professionals	85.1	20.5	39.5	16.6	45.8	14.7
Medications	3,457.9*†	3,341.1	2,602.3*	2,555.5	2,620.9†	2,318.0
Hospitalization	2,326.3*	603.5	2,119.2	445.8	1,291.5*	199.4
Professional home care	63.2	0.0	1.1	0.0	350.5	0.0
Total	18,972.0*	19,070.3	16,993.1*	17,358.1	17,914.7	17,593.3
Costs to patient and family
Informal care	584.7*	251.5	552.3†	161.2	346.9*†	116.6
Productivity losses	716.5*	415.1	410.6†	144.0	1,020.3*†	581.4
Total	1,301.2	998.1	972.9*	526.8	1,367.3*	783.4
Costs in other sectors
Travel	734.6*	863.3	709.3†	750.3	796.6*†	863.3
Intervention costs	76.7*†	79.5	4.6*‡	5.2	17.4^†‡	18.8
Total societal costs	21,007.7*	21,548.0	18,675.4*	18,721.3	20,078.6	19,826.6

*,†,‡p < .05 indicates a between-group difference based on Bonferroni Post Hoc.

DD, dedicated dietitian; EP, existing practice; THD, trained hospital dietitian; HD, hemodialysis.

The mean societal costs were highest in the DD group. Similarly, the mean healthcare costs were highest in the DD group. The THD group had the highest mean costs to patients and family. During Phase I, the social costs were highest in the DD group, whereas, during Phase II, THD had the highest costs.

Between Phase I and Phase II, the DD group exhibited the highest decrease in societal costs (–$2,226.4) compared with the EP (–$1,320.8) and THD (–$654.6) groups. This was remarkably noted for the costs of emergency hemodialysis (–$51 vs –$11.5 vs –$2.4, for DD, EP, and THD, respectively), medications (–$419.9 vs –$61.0 vs –$58.5), and hospitalization (–$1,151.0 vs –$1,010 vs –$310.6) costs.

Study outcomes

During Phase I, the adjusted serum phosphorus decreased in the DD (mean drop: –0.58 mg/dL) and EP (mean drop: –0.09 mg/dL) groups; and remained unchanged in the THD group. During Phase II, mean serum phosphorus increased in all groups. This increase was accentuated in the DD and EP groups (+0.26 and +0.25 mg/L, respectively) more than the THD group (+0.04 mg/dL). At the end of the study, the EP group had the highest mean adjusted serum phosphorus. Moreover, mean adjusted serum phosphorus was higher than baseline values in the EP and THD groups, in contrast to the DD group. At each study phase, the EP group gained the most QALYs (Table 4).

Table 4. Study outcomes (mean (SD)).

	DD (n = 116)	EP (n = 130)	THD (n = 299)
PHASE I
Phosphorus†‡ (mg/dL)	4.99 (0.98)*	5.30 (1.17)*	5.16 (0.89)
QALY†	0.2781 (0.0436)	0.2809 (0.0419)	0.2759 (0.0441)
PHASE II
Phosphorus†‡ (mg/dL)	5.25 (1.39)	5.55 (1.75)*	5.20 (0.97)*
QALY†	0.2570 (0.0877)	0.2620 (0.0844)	0.2609 (0.0797)
Total
QALY†	0.5352 (0.1233)	0.5429 (0.1207)	0.5368 (0.1185)

† Imputed values;

‡ Regression-based adjusted values;

* p < .05 indicates a between-group difference based on Bonferroni Post Hoc.

DD, dedicated dietitian; EP, existing practice; THD, trained hospital dietitian; SD, standard deviation.

Cost-effectiveness and cost-utility analyses

In the base-case CEA (Table 5), in comparison with the EP protocol, the ICER of DD for societal costs per 1 mg decrease in serum phosphorus was $7,853.6. On the other hand, the THD protocol dominated the DD protocol, i.e. the DD group had higher serum phosphorus and higher societal costs. The THD protocol was likely to be the most cost-effective intervention (Figure 1).

Figure 1. Cost-effectiveness acceptability curve presenting the probability that the protocol is cost-effective (y-axis) with respect to serum phosphorus, given various ceiling ratios for willingness-to-pay (x-axis). The ICER was represented in absolute values, given that the beneficial outcome is the decrease in serum phosphorus, resulting in a negative ICER. DD, dedicated dietitian; EP, existing practice; THD, trained hospital dietitian; ICER, incremental cost-effectiveness ratio.

Table 5. Incremental cost-effectiveness and cost-utility ratios.

	DD vs EP			DD vs THD
	Incremental effects	Incremental costs	ICER	Incremental effects	Incremental costs	ICER
Serum phosphorus (mg/dL)
Base-case*	–0.31	+2,404.5	7,853.6†	+0.05	+988.4	Dominated
Complete cases*	–0.41	+2,555.2	6,248.5†	–0.02	+1,049.8	59,425.9†
HD costs* excluded	–0.31	+1,685.6	5,505.5†	+0.05	+1,775.2	Dominated
p > 5.5 mg/dL*	–0.73	+4,295.6	5,887.4†	–0.004	+1,175.2	26,5243.2†
Decrease in p (Delta adjustment)	–0.40	+2,404.5	5,935.9†	–0.11	+988.4	9,212.8†
QALY
Base-case	–0.01	+2,404.5	Dominated	–0.002	+988.42	Dominated
Complete cases	–0.01	+2,555.2	Dominated	+0.002	+1,049.8	449,603.7
HD costs excluded	–0.01	+1,685.6	Dominated	–0.002	+1,775.2	Dominated
p > 5.5 mg/dL	+0.02	+4,295.6	192,938.1	–0.01	+1,175.2	Dominated
Adjusted QALY*	–0.01	+2,404.5	Dominated	–0.003	+988.4	Dominated

* Regression-based adjusted values;

† Absolute values are presented, given that the beneficial outcome is the decrease in serum phosphorus, resulting in a negative ICER.

DD, dedicated dietitian; EP, existing practice; THD, trained hospital dietitian; ICER, incremental cost-effectiveness ratio; HD, hemodialysis; P, serum phosphorus; QALY, quality-adjusted life-year.

In the CUA, the DD protocol was dominated by the EP and THD protocols. Even for the highest WTP threshold, the probability of the DD protocol being cost-effective was almost null. The EP protocol had the highest probability of being cost-effective (Figure 2).

Figure 2. Cost-utility acceptability curve presenting the probability the protocol is cost-effective (y-axis) with respect to quality-adjusted life-year gain, given various ceiling ratios for willingness-to-pay (x-axis). DD, dedicated dietitian; EP, existing practice; THD, trained hospital dietitian; ICUR, incremental cost-utility ratio.

Sensitivity analyses

The results of the cost-effectiveness and cost-utility of the three protocols were inconclusive (Table 5). Regarding serum phosphorus, for all the scenarios, the ICER of DD vs EP was lower than the base-case result, thus more cost-effective. When compared with THD, the ICER of DD ranged between $9,212.8 for the delta adjustment, and $265,243.2 for hyperphosphatemic cases. The THD protocol had the highest probability of being cost-effective for hyperphosphatemic patients and when hemodialysis and transportation costs were excluded. For the delta adjustment, and for complete cases, the probability of DD and THD being cost-effective tied at $8,000 and $60,000, respectively. Above these values, DD was the most cost-effective protocol (Appendix 5).

Regarding QALY, the DD protocol was dominated by EP for all scenarios, expect for hyperphosphatemic cases, and by THD for all scenarios, except for complete cases. Even for the highest suggested WTP threshold, the probability of the DD protocol being cost-effective was low. EP was the most cost-effective protocol (Appendix 6).

Discussion

In this study, we explored the cost-effectiveness of intensive nutrition education provided by trained dedicated dietitians (DD), compared with the existing practice (EP) in Lebanon—where dietitians have limited knowledge about renal nutrition and no dietitian-to-patient contact time is set, and to another protocol (THD)—where dietitian education is ensured, yet the contact time with the patients is not established.

The results showed that the direct costs of the nutrition education are extremely low compared with the costs of other interventions incurred as part of the management of hemodialysis patients. At the end of the study, in comparison with the EP and THD groups, the DD group exhibited the greatest decrease in serum phosphorus, no difference in QALY, and the highest societal costs.

Regarding serum phosphorus management, the DD protocol was likely to be cost-effective compared with the EP protocol. However, it is very difficult to draw a firm conclusion of this analysis, since we lack a societal WTP threshold for a decrease in serum phosphorus. On the other hand, the results were inconclusive for the comparison of the DD protocol vs THD protocol. Regarding QALY, the DD protocol was dominated by both of the other protocols, in most of the cases. When DD was not dominated (the case of hyperphosphatemic patients in comparison with EP, and complete cases in comparison with THD), it was associated with a cost per QALY gained higher than what is considered to be an efficient use of finite healthcare resources^32,34,35. The probability of the DD protocol being cost-effective was low for the base-case and sensitivity analyses.

Several factors might explain these results. At baseline, the DD group had the highest MIS and serum phosphorus. High MIS predicts increased mortality, days, and frequency of hospitalization, as well as lower QoL^36–39. High serum phosphorus is associated with the same deleterious consequences^2–5,40. This suggests higher baseline costs of DD compared with the other groups. This might also explain the greater decline in resources use and costs (notably for hospitalization and medications, especially costly phosphate binders) after the implementation of the intervention (Phase II). Within this scope, it is worthy to note that, despite achieving a greater serum phosphorus decrease in comparison with THD (–0.32 vs +0.04 mg/dL), mean serum phosphorus of the DD group remained higher, causing it to be dominated in the base-case analysis. The contradictory results between this analysis and the one using the delta adjustment illustrate this observation. Moreover, patients in the DD group might have been more critically ill than those in the other groups, as exhibited by the high mortality rate even before enrollment in the study (Appendix 3). In addition, as mentioned in Karavetian et al.¹⁴, the effectiveness of the DD protocol might have been partially “masked” by the improvement noted in the EP group, where contamination of information through patients and nurses and passive education through posters and distribution of educational material took place; and in THD group, where the intervention was sustained during Phase II, in contrast to the DD group, where the intervention was stopped.

The sensitivity analysis demonstrated uncertainty in the cost-effectiveness of the DD protocol. In particular, we were not able to identify whether the nutrition education is an economically attractive intervention among hyperphosphatemic patients, and contradictory results were obtained regarding serum phosphorus and QALY. This finding is not in line with results from a previous economic evaluation of phosphate binders⁴¹, which revealed better cost-effectiveness with increasing serum phosphorus—taking into account the accentuated morbi-mortality risk with hyperphosphatemia. The limited time horizon in our study (1 year vs lifetime in Brennan et al.⁴¹) might explain this contradictory finding, as it did not allow us to explore whether there is a future decrease in mortality and morbidity with better phosphorus control due to the nutrition education.

Our analysis depended on numerous assumptions and has several limitations. First, although we resorted to adjustments for baseline differences in key parameters of the economic evaluation (serum phosphorus and utility), we were not able to adjust our analysis for all baseline differences between the three groups, notably the malnutrition-inflammation status. Second, we did not collect cost data at baseline, assuming that the cluster randomization would results in similar groups; however, this was not the case and groups ended having different baseline key variables. As these baseline differences suggest higher baseline use of resources and costs in the DD group, it would have been wise to collect costs data at baseline, and resort to a regression-based adjustment of patient-level cost data⁴². Although it is uncommon to report baseline costs in trial-based economic evaluations, van Asselt et al.⁴² argued that costs at baseline would influence costs during the trial and presented the case for reporting these costs and investigating their influence, to be able to attribute the difference that is found at endline to the intervention. Third, the most important assumptions in our analysis related to attrited cases. We used patient-specific values when available from a previous time-point, and followed the imputation by the mean using the most detailed values, where data from a previous time-point were unavailable. We assumed that is the most feasible method, in light of the limited number of patients in each group in each unit, although it might not be the most robust one. The results were sensitive to this parameter, and ICERs/ICURs varied between the base-case analysis and when only complete cases were included. Potentially, attrited patients had different resource-consuming characteristics than the ones who finished the trial. A post-hoc analysis addressing this issue might provide an answer to this question. In addition, we assumed that death or hospitalization took place at the mid-point of each phase, that costs and serum phosphorus followed a linear evolution throughout time, and that patients maintained the same QoL at withdrawal, and applied our analyses accordingly. We adopted this method because costs, serum phosphorus, and QoL values were not available at time of attrition; and due to the lack of studies describing the evolution of costs, serum phosphorus, and QoL, mainly before death, transfer, and transplant in hemodialysis patients. Fourth, as discussed in Rizk et al.²², the estimation of resources consumption used a self-reported questionnaire, rather than documented sources (patient or facility records). This potentially leads to recall bias or poor understanding of the questions among patients with limited cognitive skills. Valuation of some resources costs also relied on information collected from patients, due to the lack of a manual for cost analysis in healthcare research in Lebanon²². In an attempt to overcome these two limitations, the resource utilization questionnaire used in this study was designed following good practices for improved accuracy⁴³. Fifth, economic evaluations of interventions targeting hyperphosphatemia management in hemodialysis patients usually adopt a time horizon longer than the one adopted in this study, in order to capture the long-term effect of reducing serum phosphorus¹¹. In fact, the time horizon should be long enough to include all relevant costs and outcomes of the intervention⁴⁴. This was not the case of our analysis, and exploring the cost-effectiveness of the DD protocol through a model-based analysis adopting a lifetime horizon is needed, especially that this protocol was associated with the best clinical outcomes and greatest decrease in societal costs in post-implementation. Additionally, it is worthy to note that the cost of the intervention is expected to further decrease in the long run, taken that the initial training of the dietitians (≈10% of the cost of DD intervention) is a one-time intervention. Finally, differences in healthcare systems funding, costs of healthcare and other resources, and societal WTP for health interventions, among other factors, limit our ability to directly generalize our results to other countries.

The primary strength of this analysis is that it used patient-level data from a randomized controlled trial; and tried to compare a novel intervention with the existing practice, and to another alternative. This could be considered as a first step towards implementing evidence- and value-based care for hemodialysis patients in Lebanon. In addition, the NEMO trial was conducted following a practical fashion, increasing the likelihood of the generalizability of its results to real-world practice; and its economic evaluation was performed from the preferred (societal) perspective.

According to the National Kidney Registry, nearly 3,300 patients receive hemodialysis in Lebanon⁴⁵. The monthly budget implications of making dedicated dietitians available for all patients would be substantial (≈$40,000). Yet, a closer look at the cost savings of this intervention are also likely to be quite substantial, making this intervention worthy of an exhaustive evaluation and possibly of its consideration as part of the management of hemodialysis patients. For instance, the difference in the mean decrease in societal costs between DD and EP ($636.3) and between DD and THD ($1,258.3) in post-implementation would offset more than 8- and 16-times the cost of the 6-month nutrition education, respectively.

Finally, this CEA/CUA assessed dedicated dietitians providing education targeting serum phosphorus management. This is one aspect of the medical nutrition therapy of renal patients; and exploring other outcomes of the dietetic management is recommended, especially that previous evidence suggests substantial cost savings of nutrition interventions among renal patients through malnutrition prevention and management⁴⁶ and improved QoL^47,48. Future clinical studies should investigate the impact of nutrition education on relevant clinical end-points (i.e. morbidity and mortality), and CEA should assess equitable groups at baseline using an extended time horizon, to provide a more realistic evaluation of the economic attractiveness of this intervention. While healthcare funding decisions must be made using the best currently available data, it is possible that future estimates of the cost-effectiveness of the nutrition education may differ.

Conclusions

The additional costs of the nutrition education were low as compared with the societal costs of hemodialysis patients. Regarding serum phosphorus, the DD protocol was likely to be cost-effective compared with EP, yet the probability of DD being cost-effective compared with the THD protocol was low. No effect of the DD protocol was noted on QALY; this protocol was, therefore, dominated by the EP and THD protocols. Although our results did not confirm the cost-effectiveness of the intensive nutrition education, a closer look at the low cost of the nutrition education, coupled with the greater decrease in the use of services and costs during the post-implementation phase, would potentially reveal the true added economic value of the DD protocol.

Transparency

Declaration of funding

None reported.

Declaration of financial/other relationships

None reported.

Acknowledgments

None reported.

Appendix 1: Resource utilization questionnaire

Appendix 2: Costing of the nutrition education

Cost components and mean costs per patient of the three protocols (LBP).

Table

	#	Unit	Cost	Unit	Phase I	Phase II
Dedicated Dietitian (DD) protocol
(1) Dietitian training^a					12,055.34	0
(2) Educational material^b	78	papers	50.00	LBP/paper	3,900.00	0
(3) Nutrition education	12	hours (2 h/month/6 months)	7,526.88	LBP/hour	90,322.58	0^c
(4) Standard dietetic care	0.5	hour (1 h/year)^d	7,527.88	LBP/hour	3,763.94	3,763.94
Total					110,041.87	3,763.94
Existing Practice (EP) protocol
Standard dietetic care	0.5	hour (1 h/year)^d	7,527.88	LBP/hour	3,763.94	3,763.94
Trained Hospital Dietitian (THD) protocol
(1) Dietitian training^a					12,055.34	0
(2) Nutrition education	1	hour (2 h/year)^e	7,526.88	LBP/hour	7,526.88	7,526.88
Total					19,582.23	7,526.88

^a Including training cost and cost of dietitian time for attending the training. Rational and conduct of the training was described in Karavetian and Rizk (2016)⁴⁹. Training cost per dietitian, obtained from its provider (MK who is a co-author of this article), was multiplied by the number of dietitians needed to provide the nutrition education and divided by the number of patients. Dietitian cost per hour was obtained from the Syndicate of Dietitians in Lebanon (personal communication) and from a national survey among Lebanese hospital dietitians (Karavetian et al., 2013)⁵⁰.

^b Obtained from the provider of the training (MK, who is a co-author of this article).

^c Intervention was stopped during Phase II.

^d Mean number of hours as per the standard care in Lebanon (Karavetian et al., 2013)⁵⁰.

^e Mean number of hours allocated to patients in the group (unpublished data obtained from MK, who is a co-author of this article).

The costs gathered and calculated in LBP were converted to US$(1 US$ = 1,507.5 L.L.; year of reference: 2011)²³ and uprated to 2015US$using Consumer Price Indices (Index, 2010 = 100)²⁴. The costs of the dietetic interventions for the DD, EP, and THD groups for Phase I and Phase II were as follows: $74.5, $2.2; $2.6, $2.1; and $13.3, $4.2, respectively.

Appendix 3: Flow diagram of the trial

A total of 720 patients participated in the study; 570 of those met the inclusion criteria.

Appendix 4: Cost distribution of the study groups

^aDD (dedicated dietitian): 2.5th percentile: $19,799.49; 97.5th percentile: $22,377.03.

^bEP (existing practice): 2.5th percentile: $17,556.73; 97.5th percentile: $19,843.65.

^cTHD (trained hospital dietitian): 2.5th percentile: $19,325.09; 97.5th percentile: $20,871.02.

Appendix 5: Cost-effectiveness acceptability curves for sensitivity analyses

Cost-effectiveness acceptability curve presenting the probability the protocol is cost-effective (y-axis), with respect to serum phosphorus, given various ceiling ratios for willingness-to-pay (x-axis). Sensitivity analyses performed for complete cases, hemodialysis and transportation costs excluded, hyperphosphatemia at baseline and difference in serum phosphorus as outcome. The ICER was represented in absolute values, given that the beneficial outcome is the decrease in serum phosphorus, resulting in a negative ICER. DD, dedicated dietitian, EP, existing practice; THD, trained hospital dietitian; P, serum phosphorus; HD, hemodialysis; ICER, incremental cost-effectiveness ratio.

Appendix 6: Cost-utility acceptability curves for sensitivity analyses

Cost-utility acceptability curves presenting the probability the protocol is cost-effective (y-axis), with respect to QALY gain, given various ceiling ratios for willingness-to-pay (x-axis). Sensitivity analyses performed for complete cases, hemodialysis, and transportation costs excluded, hyperphosphatemia at baseline and adjusted QALY. DD, dedicated dietitian, EP, existing practice; THD, trained hospital dietitian; P, serum phosphorus; HD, hemodialysis; QALY, quality-adjusted life-year; ICUR, incremental cost-utility ratio.