The economic evaluation consisting of a cost utility analysis (CUA) was undertaken alongside a cluster randomised controlled trial (cRCT). The cRCT was conducted in Valencia (Spain) for 6 months between December 2017 and May 2018. The full report of the study in Spanish language is available .
The Pharmaceutical Association of Valencia provided CPs with study information via phone or email. CPs within twenty-one municipalities agreed to participate. The municipalities were the clusters of the study to avoid contamination between groups. Municipalities with pharmacies who accepted to participate were randomised through a sequence of computer-generated random numbers to the control (UC) and the MAS arms applying a ratio of 1:1. Due to the nature of the intervention, pharmacists could not be blinded.
Patients aged ≥16 years or between 2 and 15 years of age if they were accompanied by a responsible adult, who were seeking care i.e. presenting symptoms or requesting a product (direct product request) for the minor ailments were included in the study. The minor ailments considered in the study were: dermatological problems (cold sore, foot fungi), gastrointestinal disturbance (diarrhoea, flatulence, heartburn or vomiting), pain (dysmenorrhea, headache, sore throat) and upper respiratory tract (cough, cold or nasal congestion).
Description of the intervention (MAS)
The intervention is described using the TIDieR  checklist (Additional file 1). The main components to the intervention being:
A standardised pharmacist–patient consultation  using: co-developed with doctors’ management protocols for each specific symptom , patient educational material, and a web-based data collection software  that guided pharmacists through the consultation with selected pop-ups such as referral criteria (i.e. “red flag symptoms”).
Practice change facilitation (PCF): PCFs made regular on-site visits to MAS CPs to identify and resolve barriers with service provision and check the fidelity of the intervention.
Educational training for MAS pharmacists was a twelve-hour training session delivered prior to the beginning of the trial. It covered service provision, good practice standards, service protocols, communication’s skills with the patient and other health professionals, web-based data collection software use, data collection and trial protocol.
Patients attending UC pharmacies received usual pharmacist practice. Normally, in Spain, usual pharmacist practice does not include the use protocols or an IT system. When a patient presents in CP with a minor ailment or requesting a product, a consultation is carried. However, the depth and breadth of this consultation does vary. Also, pharmacists in the UC group did not receive the support of a practice change facilitator. Pharmacists in the control group attended a three-hour training on data collection procedures and patient recruitment.
Incremental cost-utility ratio (ICUR) of the service and health related quality of life (HRQoL) were the main economic and humanistic outcomes. A health system and patient perspective was chosen.
Data collection was undertaken at the moment of the consultation in the CP (i.e. patient demographics, medicines supplied, HRQoL) using a web-based data collection software . Ten days after the consultation, patients were contacted by telephone by the research group to collect post consultation data (i.e. HRQOL, symptom resolution using a Likert scale from 1 “not at all” to 5 “completely”, reconsultation rates whenever the patient had to consult again for the same ailment during the 10 days after prior consultation in CP, type of reconsultation which could be CP, GP or emergency department). This follow-up time frame was considered appropriate given minor ailments are self-limiting conditions that should resolve within a short period of time. Patients self-reported HRQoL using EuroQoL 5D-5L self-complete version on paper at the time of the pharmacy consultation and 10 days later through telephone interview with a researcher using the EuroQoL 5D-5L telephone version .
In addition, the control group documented the consultation, which is not normally part of UC.
Sample size calculation was based on the clinical primary study outcomes. A 10% absolute increase in appropriate medical referral rate (85 to 95%)  and modification of direct product request (8 to 18%) [31, 32] were estimated. The sample size was calculated with ≥0.9 power, type I error rate of 5%, equal allocation ratio and assuming an intra-cluster correlation of 0.01. The larger of the two-estimated sample size calculations was used to determine the overall sample size, of 726 patients (allowing for 10% dropout).
Statistical analysis and economic evaluation description
A descriptive analysis was performed to analyse baseline characteristics by group. Continuous variables were described using mean and standard deviation (SD) and categorical variables were presented as counts and percentages. Comparison of continuous variables between groups was undertaken using t-Student test and Kruskal-Wallis or Mann-Whitney (when skewed). Comparison of categorical variables was undertaken using Pearson’s χ2 tests. The level of statistical significance was established as p < 0.05.
The research was reported in accordance with the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) checklist (Additional file 2) .
The effectiveness of the intervention was estimated as quality-adjusted life years (QALYs). QALYs were obtained from the cRCT  with the HRQoL questionnaire administered at the time of the pharmacy consultation using the EuroQoL 5D-5L self-complete version on paper and at 10-day follow-up via the telephone interview version by the research team . Utility indexes associated with each health state were taken from the published Spanish tariff . QALYs were calculated as area under the curve  considering a 10-day time horizon for patient follow-up. This method was implemented adding areas under the curve from geometrical forms obtained by lineal interpolation between utility values during the study period (10 days). Due to missing data in the sample in relation to the EQ-5D-5L data at follow-up, multiple imputation with chained equations was performed in order to allocate data. The multiple imputation missing data model included as predictive variables, EQ-5D-5L indices, at baseline and follow-up, and patient characteristics such as sex, age, and type of minor ailment (dermatological problems, gastrointestinal disturbance, pain and upper respiratory tract).
Firstly, the missing data were imputed under the missing at random (MAR) assumption; this was the main analysis (base-case scenario). Secondly, a completed case analysis was performed, assuming that patients who completed all follow-up were representative of the entire sample who initially agreed to participate. In this case, all participants who did not return the HRQoL questionnaire were removed, which assumes data is missing completely at random (MCAR). A simple pattern mixture model was implemented, following the approach recommended by Faria et al.  The utilities shown in both cases have been adjusted for age, gender, type of minor ailment, symptom duration and baseline utility.
Total direct costs included costs to the health system, CPs and direct patient costs. Direct costs included the cost of health professionals’ consultation time, medication costs, pharmacists’ training costs and investment of the pharmacy (infrastructure, etc.) and costs of patients’ re-consultation in the following 10 days after initial consultation in CP (including contacts with all health providers such as GP and ED consultations). To value each resource item in terms of its unit cost, additional data was used:
Medication costs, as out-of-pocket costs borne by the patient, were based on official prices listed in the Bot Plus® database at September 2018 .
CP consultation costs (for both groups) were calculated on time spent by pharmacists for the provision of the service, obtained from the pharmacists’ web-based data collection software for the cRCT. Considering pharmacist role (supervisor/regular pharmacist): time used for pharmacist-patient consultation was 8.00 min (SD = 2.45) for a supervisor in MAS group, 5.35 min (SD = 3.20) for a regular pharmacist in MAS group, 6.57 min (SD = 3.90) for a supervisor in UC group and 4.95 min (SD = 3.85) for a regular pharmacist in UC group. Pharmacists in UC group were asked to document the consultation which is not part of usual care.
The unit cost of labour of the community pharmacist was calculated using the pharmacist’ salary as determined by the Spanish community pharmacy agreement . Supervisor pharmacists’ salary was 0.315 €/minute and regular pharmacists’ salary was 0.293 €/minute.
To calculate the costs related to the infrastructure investment and maintenance costs for the pharmacy, estimates were used from a previous Spanish study .
Pharmacist training costs were for the twelve-hour course (classroom costs, presenter’s costs, travels and accommodation costs).
GP (56.95€) and ED (105.27€) consultation costs were based on Valencian Law 20/2017 .In order to allocate the costs proportional to each patient, the mean cost of investment per pharmacy was divided by the mean estimated number of patients included in a potential MAS per CP. A study conducted by the General Pharmaceutical Council of Spain stated that 15–20% of patients attending community pharmacy presented with minor ailment symptoms or were self-selecting a medication for treatment of their symptoms . Five thousand four hundred patients were estimated to potentially access a CP MAS annually with an estimated of approximately 120 patients per day per CP .
The investment of pharmacies in the UC group was assumed to be null.
Seemingly unrelated regression (SUR) was used to jointly model costs and QALYs.Cost and QALY were adjusted for sex, age, type of minor ailment, and QALYs were also adjusted for baseline utility to account for differences between groups at baseline. Coefficients were combined across the multiple imputed datasets using Rubin’s rules .
To assess the robustness of the results, a subgroup analysis and probability analysis was carried out alongside the missing data model described above. The two subgroups analysis accounting for presentation type were performed to evaluate possible differences. This included patients presenting to the CP with symptoms or patients directly requesting a product.
The probabilistic analysis assesses the uncertainty in cost and outcomes across both alternatives of the study (MAS and UC) and was conducted by bootstrapping cost and QALY pairs from each patient with 1000 replicates. The results of the replicates are presented in the cost-utility plane. The probability of the intervention being cost-effective was calculated assuming data was bivariate and normally distributed .
In addition, a cost-effectiveness acceptability curve (CEAC) was derived to estimate the probability of the intervention being cost-effective at different amounts of society’s willingness to pay for health outcomes. Since there is not a formal QALY monetary value assigned by the Valencian or National Spanish government, a study carried out by the government in Canary Island (Spain)  was used for the assignment of a monetary value to the QALY. They suggested a willingness to pay per QALY from €20,000 to €25,000. All analyses were performed using STATA 14 .
Ethics approval and consent to participate
The study was approved by the University of Granada Ethics Committee (approval number 331/CEIH/2017) and Xátiva-Ontinyent Ethics Committee “Lluís Alcanyís”. All methods were carried out in accordance with relevant guidelines and regulations. Pharmacists were informed of the study and provided written consent to participate. Informed consent was obtained from all subjects and/or their legal guardian(s).
ISRCTN, ISRCTN17235323. Registered 07/05/2021 - Retrospectively registered, https://www.isrctn.com/ISRCTN17235323