Study design
This is a cost-utility study, corresponding to a full economic evaluation comparing both costs and effectivity in patients undergoing MD in the outpatient versus in the inpatient setting in a Portuguese National Healthcare System hospital. Effectivity is presented as quality-adjusted life-years (QALYs), with the number of QALYs calculated by the product between life years and utilities. We followed the hospital perspective, considering direct hospital costs.
Uncertainty was explored via an one-way deterministic sensitivity analysis and probability sensitivity analyses.
The study was approved by a hospital ethics committee in May 25th 2017.
Costs
Costs were assessed from two cohorts of patients treated in the same spine center of a central Portuguese hospital. Accordingly, specific data from 20 outpatients and 20 inpatients undergoing MD with single excision of herniated intervertebral disk was gathered. To be included in either group, patients had to present clinical complaints compatible with lumbar disc herniation and with confirmation of clinical findings by radiological studies (computed tomography and/or magnetic resonance imaging). Patients were excluded if they presented: (1) comorbidities precluding outpatient surgery; (2) social conditions precluding outpatient surgery (i.e. living alone or far from the hospital, psychiatric conditions); (3) need for additional spine surgical procedures other than single excision of herniated intervertebral disk; or (4) previous lumbar spine surgery. Upon inclusion, all patients were submitted to a lumbar MD by the same surgical team.
Costs were defined as the sum of direct hospital costs related with inpatient and outpatient procedures. For outpatients, we quantified operatory room (OR) costs, including costs related with (1) staff and OR occupation; (2) used drugs; (3) supplies used in that particular intervention; and (3) other costs. In addition, we quantified costs related with eventual 30-days readmissions. Such costs were also quantified for inpatients, among whom costs related to hospital stay were also added. The latter include staff-, drug- and supplies- (i.e., bandages, disposable wearing, etc.) related costs. For both inpatients and outpatients, we retrieved other costs related with water supply, electricity, telephone services, administrative issues, etc. Both groups of patients had the first post-operative appointment 2 weeks after the surgery and follow a similar medical follow-up.
Regarding outpatients, we prospectively analyzed a consecutive sample of 20 patients, presenting to our spine center between 2017 and 2018, with clinical pain and disability due to radiologically-identified lumbar disc herniation that fulfilled the above-mentioned criteria. A pre-defined outpatient protocol was followed, with patients being submitted to a pre-operative anesthetic evaluation and provided with aseptic sponges to bath in the morning before the procedure. After surgery, all patients were discharged in the same day, less than 12 h after the procedure and received a pre-defined analgesic protocol. To assess complications in the immediate post-operative period, a physician performed a telephone call up to 24 h after discharge, with the patient being directed to an emergency appointment if any complication was suspected.
Assessed inpatients consisted of a sample of 20 individuals, fulfilling the above-mentioned criteria, with similar age and gender to those of outpatients, and who were retrospectively selected from patients submitted to MD in the same spine center.
Despite the literature describing a variable length of stay among MD patients in real life scenarios, there is an overall agreement among spine surgeons that an uncomplicated inpatient MD would only need a one-day admission [12]. As a result, we not only performed this economic evaluation study estimating inpatient costs as observed (irrespective of the admission time), but also performed a sensitivity analyses considering the scenario of all patients being only admitted for one day. To do so, costs for inpatients that stayed for longer periods were re-calculated for those expected in a one-day admission period.
Utilities
Utilities were estimated from the Oswestry Disability Index (ODI)
Outpatients were prospectively evaluated pre-operatively and three and six months post-operatively, with ODI being assessed in each evaluation, along with the overall visual analogue scale of pain (VAS), back pain VAS (BP-VAS), and leg pain VAS (LP-VAS).
Since inpatient data from our center were collected retrospectively, ODI data were retrieved from the literature. To do so, we performed a comprehensive search on MEDLINE from 2018 to 2020 (limited to humans and systematic reviews), using a combination of the search terms: “lumbar”, “hernia”, “protrusion”, “extrusion”, “discectomy” and “microdiscectomy”. We specifically searched for studies on lumbar MD, displaying ODI data on pre- and post-operative assessments at 3 and/or 6 months assessments after the surgical procedure. Of a total of 110 retrieved references, we identified one systematic review with meta-analysis fulfilling all eligibility criteria and utilities were estimated from its data on ODI [13].
QALYs were estimated based on three and 6-months utilities, adjusted for baseline values, using two different approaches—the area under the curve (AUC) and change from baseline (CfB) approach [14]. For outpatients, average and standard-deviation values for QALYs based on each approach were estimated using patient-level data. For inpatients, such values were estimated following Bayesian methods—a random-effects Bayesian meta-analysis was performed to obtain pooled baseline utilities and mean utilities differences, which were then used in the same Bayesian model to estimate the average and standard-deviation values for QALYs (via assessment of the posterior distributions) following the AUC and CfB approaches. Uninformative prior distributions were used in Bayesian models both for the effect size measures and for the tau parameters (dnorm (0,0.00001) and dunif (0,10), respectively).
Data analysis
Categorical variables were described using absolute and relative frequencies, while continuous variables were described using means and standard-deviations. Categorical variables were compared using the chi-square test, while continuous variables were compared using the independent samples t-test and its non-parametric counterparts.
To assess for cost-effectiveness, we estimated incremental cost-effectiveness ratios (ICER), consisting of the difference between costs (i.e., outpatient minus inpatient costs) dividing by the difference in QALYs (i.e., outpatient minus inpatient QALYs). To account for uncertainty, we performed one-way deterministic sensitivity analysis, testing the effect of changing one variable at each time according to a prespecified range of values—observed minimum and maximum values were used for costs, while for QALYs (which were estimated by Bayesian values), the minimum and maximum values used for sensitivity analyses were obtained after 10,000 simulations based on their distributions. In addition, to explore uncertainty we conducted probabilistic sensitivity analysis via Monte Carlo simulation methods—we ran 10,000 simulations in which we allowed each input variable to vary according to a probability distribution. A treatment choice was regarded as cost-effective if its ICER was lower than the defined willingness to pay (WTP) per gained QALY. As indicated by WHO, The WTP was defined at 3 times the Portuguese per capita gross domestic product (GDP) [15]. Using the last available International Monetary Fund values (2019), this corresponds to a WTP value of €60,000 [16]. This probability sensitivity analysis was performed for both inpatient’s observed admission time and for one day only. Frequentist statistical analysis was performed using SPSS v26 (IBM SPSS Statistics, NY. Bayesian models were performed using rjags package for software R (version 4.0). Probabilistic sensitivity analysis was performed using TreeAgePro 2019 (TreeAge Software, Williamstown, MA).