Economic evaluation of a collaborative model of pharmaceutical care in an Irish hospital: cost-utility analysis

Introduction

Hospitalised patients are at increased risk of medication misadventure relative to the general population, with a minority of medication errors leading to severe patient harm^1,2. It is now well recognised that pharmaceutical care of hospitalised patients should include medication review and medication reconciliation at care transitions, facilitating integrated care within and between care settings³. However, the economic value of such protective services to prevent medication error or mitigate the associated harm is uncertain^4,5. Methodological variability is a feature of studies examining the prevalence and economic impact of medication error or the benefit of pharmaceutical care in hospitalised patients^6,7. Furthermore, research to date tends to be undertaken within either the primary or secondary care setting, with relatively less work examining the transitional care journey^6,7.

A collaborative model of pharmaceutical care developed and examined in an Irish hospital through an uncontrolled before-after study, demonstrated a beneficial effect on discharge medication error rates and quality of prescribing⁸. Medication reconciliation and medication review were integral component of this intervention. Evidence of its cost-effectiveness was required to support decision-making regarding implementation.

Methods

The economic evaluation was based on the primary study and the decision problem to be addressed was whether the intervention was more cost-effective than standard care (SC)⁸. The evaluation was framed within the context of the study’s primary outcome measure, prevalence of discharge medication error.

The evaluation was undertaken from the perspective of the Health Service Executive (HSE), the payer for public healthcare in Ireland. The timeframe was one-year post-discharge. The weeks following hospital discharge have been identified as the highest risk for medication-related harm (MRH)^4,9. However, a time horizon of one year was specified to consider manifestations of harm of longer latency or sequelae requiring lengthy remediation. As the time horizon did not exceed one year, discounting was not applied¹⁰. The evaluation complied with the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) and Irish national guidelines^10–12. The intervention study was conducted between 2011 and 2012, published in 2014 and the cost-utility analysis study was undertaken between 2015 and 2017.

The setting is the Irish healthcare system. Study participants were adult inpatients acutely admitted to hospital who were taking at least three regular medicines and receiving medical, rather than surgical, care. There were no statistically significant differences in baseline population characteristics between the study groups in terms of age, gender, number of medicines, length of stay, or comorbidity burden⁸.

The main differences between the intervention and comparator service have previously been reviewed and reported, and in brief, were as follows⁸: PACT involved team-based working with pharmacist-led medication history taking and reconciliation on admission, pharmacist amending the prescription chart (collaborative prescribing) where necessary during admission; and pharmacist-led medication reconciliation at discharge, including discharge prescription amendments (collaborative prescribing) as required. Standard care (SC) comprised pharmacist admission medication history taking and documentation, and a routine inpatient clinical pharmacy service aligned with wards/clinical areas (Table 1). SC pharmacists did not lead admission medication reconciliation, were not empowered to collaboratively prescribe, or amend the prescription and did not provide input at discharge.

Model selection, structure, and assumptions

The evaluation was undertaken within a decision-analytic modelling framework. There was no interaction between individuals in the study on which the evaluation is based, while the patient pathways did not involve recurring events. The analysis was structured within a decision tree model built in Microsoft Excel® 2010 which describes the patient sequence from admission to discharge (Figure 1).

Figure 1. Decision tree model structure for the cost-effectiveness analysis of PACT versus Standard Care.

The decision-analytic model reflected the study primary outcome measure: prevalence of discharge medication error. While budget allocations in Ireland for hospitals and general practitioners are administered by separate instruments of the HSE, the overall payer for the provision of healthcare to public patients in both settings is the HSE. The model assumes that the healthcare system, and budget allocation within, operates in cohesion, with no delineation between primary and secondary care.

All patients in the analysis were assumed to be entitled to full State endowed healthcare privileges and to not hold private health insurance and receive all aspects of pharmaceutical care, either PACT or SC. The accrual of costs due to post-discharge healthcare utilisation consequent to discharge medication error is reflected by the mean cost of error per patient in either group. The change in patient health-related quality of life (HrQoL) was identified as that caused by any discharge medication error(s) only. Therefore, patients who did not experience discharge medication error were assumed to have no change in their HrQoL.

Results

Four experts provided judgement on 81 cases that involved one or more discharge medication error, as identified in the intervention study and reported in the expert elicitation study^13,24. Of these, 75 were judged to have potential clinical consequences. Between 56 and 69 of the 81 cases were variably judged to require remedial healthcare utilisation. The mean calculated cost per case (representing an individual patient), based on all 81 cases, was €1009.58, 95% CI 726.64 to 1585.67. The mean QALY loss was 0.03 (95% CI 0.01 to 0.05). The bootstrap mean cost of discharge medication error per patient was calculated for PACT as €638.73 (95%CI 248.41-1074.17) and SC as €1091.12 (95%CI 702.51-1551). The mean (SD) utility decrement was -0.023 (SD 0.075) for PACT and -0.029 (SD 0.072) for SC.

Twenty-nine PACT patients and 63 SC patients were enrolled in the time-and-motion study, exclusion criteria were applied, and data were collected for 19 PACT patients and 30 SC patients. There were no significant differences between study groups in terms of patient gender, age, comorbidity burden, or number of admission medications¹³. The mean pharmacist time spent per patient providing the PACT intervention was 72.8 (SD 52.47) and SC was 40.6 (SD 29.17) minutes.

An hourly difference in salary of approximately €17 between SC and PACT pharmacists was calculated, driven by differential labour costs between staff grades (Table 2). Multiplying the cost per minute values obtained by the mean time to deliver the relevant pharmacy service resulted in a per patient cost of €87.91 (median €62.02; range €29.92-235.84) for PACT and €37.22 (median €32.08; range €13.75-159.51) for SC. The difference in mean cost between the groups was statistically significant (p = 0.002). Bootstrap mean costs were used in the base-case analysis (Table 3). Bootstrapping identified the 95%CI for the mean cost of pharmacist time for PACT as €58.84-119.77 and for SC as €29.18-47.39.

The expected values for costs and utilities were identified for PACT and SC, from which the incremental costs and effects were calculated. The ICER for PACT relative to SC in the base-case was negative (-€36, 537.24/QALY) indicating that the PACT intervention dominates SC, i.e., cost-effective.

The DSA showed that the model was robust to variation in individual parameters (Figure 2); all deterministic ICERs indicated that PACT dominated SC. The parameters most sensitive to variation were the cost of SC discharge medication errors and the QALY decrement for SC. The distributions and their associated hyperparameters for each input within the PSA are summarised (Table 3). The PSA-derived cost-effectiveness plane for PACT versus SC produced illustrates that cost-effect pairs generated are mostly located in the south-east quadrant, confirming that PACT dominates SC, and is cost-effective (Figure 3). A CEAC illustrated the probability of cost-effectiveness of PACT at several ceiling willingness-to-pay values (Figure 4). PACT was certain to be cost-effective (probability of 1) across all cost-effectiveness threshold values from €0-50,000. The intervention NMB was positive for each PSA produced incremental cost and effect pairing, even when the cost-effectiveness threshold was set at zero. Therefore, this is a cost saving intervention.

Figure 2. Tornado diagram of one-way sensitivity analysis.

PACT = Collaborative PhArmaceutical Care at Tallaght Hospital (the intervention); SC = Standard care.

Figure 3. Scatter plot of the incremental cost/QALY pairs of PACT vs SC.

PACT = Collaborative Pharmaceutical Care at Tallaght Hospital (the intervention), QALY = Quality Adjusted Life Years, SC = Standard Care.

Figure 4. Cost-effectiveness Acceptability Curve.

Discussion

The key study findings are that a complex, multi-faceted, collaborative pharmaceutical care intervention demonstrated cost-effectiveness compared to standard ward-based clinical pharmacy care. The sensitivity analyses confirmed that PACT dominated SC. The model was relatively robust to variation in input parameters through one-way sensitivity analysis; however, it was noted that the cost of discharge medication error and effect parameters relating to SC were most sensitive to change. This study is the first CUA of a complex pharmaceutical care intervention that includes medication reconciliation and medication review, delivered by hospital pharmacists in a collaborative pharmaceutical care model in Ireland and adds to the scant cost-effectiveness evidence available in the field.

A strength of this study was the use of national, rather than hospital specific, data to estimate labour costs and the economic burden of post-discharge management of discharge medication errors, potentially extending generalisability to similar national contexts. However, the intervention, PACT, was developed in the study hospital, and the economic findings are specific to the investigated intervention and standard care models. Given the considerable practice variation in the delivery of clinical pharmacy services in Ireland, the nature of the comparator is likely to vary widely between hospitals, with implications for the generalisability and likelihood of enhanced cost effectiveness in hospitals with less intensive service²⁵. Variation between pharmacists delivering the intervention due to differences in training and experience may also influence outcomes. Consideration should therefore be given to post-registration professional development to harmonise and credential hospital pharmacist competence²⁶.

The assumptions made in structuring the model, and the sources of data used to populate it, result in several potential limitations. While the analysis was based on a published study, it was not a trial-based economic evaluation, as direct costs associated with patient care were not available and were established by expert elicitation. It was not possible to capture inter-personal variation in the time-and-motion study of the simulated PACT service as only one pharmacist was involved, a limitation addressed through sensitivity analysis. The assumption that all patients were entitled to full statutory healthcare privileges meant that potential cost offsets due to patient co-payments or health insurance contributions were not captured, limiting generalisability beyond the specific state-assisted cohort. Other costs not included in the analysis include: litigation costs; opportunity costs arising from time spent by healthcare professionals resolving discharge medication error; the impact on medical/surgical or nursing service costs²⁷. It is speculated that the PACT model likely resulted in a net benefit in terms of physician/surgeon time, however, the model may have had a deleterious impact on nursing time²⁸.

The results are comparable to economic analyses of broadly similar interventions undertaken in Sweden, the United Kingdom and United States (US), all demonstrating the net economic benefit of pharmacist-led interventions with a medication reconciliation component^4,29,30. The US study identified a further potential economic benefit by targeting the discharge medication reconciliation service to patients at increased risk of medication-related rehospitalisation⁴.

The findings from our time-and-motion study provide an estimate of the time required to deliver end-to-end pharmaceutical care, inclusive of elements at care interfaces and during the inpatient care period. This information will be useful to policy makers and managers planning service delivery and future cost-effectiveness analyses. It is difficult to directly compare our findings with previous studies, given that most reported estimates assess the time for admission or discharge medication reconciliation delivery, or both, without other pharmaceutical care service elements^31–33. The combined evidence suggests that medication reconciliation consumes most of the pharmacist’s time delivering end-to-end hospital inpatient care and that investment in admission medication reconciliation complements and reduces the time required to perform the process well at discharge.

A significant proportion of healthcare resource utilisation consequent to discharge medication error (which often derives from admission medication error) is likely to occur post-discharge in primary care^34,35. Although the costs associated with outpatient resource use are relatively small compared with the inpatient costs, in the context of large volumes of patient discharge activity and opportunity costs, the burden of remediating consequences of discharge medication error may be substantial. The data which informed the value of the study model parameters considered this, as did the study perspective of the health service provider being both the payer and the monetary beneficiary of the intervention. This approach was employed previously and concords with Irish policy to achieve integrated care^4,36.

The paucity of good quality data on the manifestations of harm consequent to discharge medication error has been noted elsewhere as a reason for diminished certainty in CUA findings³⁷. The Expected Value of Perfect Information approach may support identifying the upper bound on the value of undertaking additional research to reduce uncertainty in decision-making³⁸. Although this study established the cost-effectiveness of PACT, a Budget Impact Analysis would forecast the impact of intervention adoption and diffusion within the Irish health service context³⁹.

Conclusion

This CUA demonstrates that the PACT intervention is cost saving and cost-effective compared to standard care, notwithstanding that confidence in the results is diminished by the lack of data on the actual costs and health impact of discharge medication errors. The ICERs generated indicate that the PACT intervention dominated standard care, and sensitivity analyses assured the results were robust to parameter variation. The findings may support decision making regarding PACT implementation and contribute to the economic evidence-base regarding pharmaceutical care, collaborative models of care and medication reconciliation-based interventions.