Cost-utility analysis of a consensus and evidence-based medication review to optimize and potentially reduce psychotropic drug prescription in institutionalized dementia patients

The details of the actual intervention analyzed in this study are described elsewhere [31]. In short, the analyzed intervention was performed in the metropolitan area of Barcelona Catalonia (Spain) (MN) between 2012 and 2014. The patients included were institutionalized dementia patients with pharmacological treatment with one or more psychotropic drugs from the Anatomical Therapeutic Chemical Classification System (ATC) code N of the World Health Organization (WHO) [32] for 3 months or longer. It was a structured, patient-centered medication review by a multidisciplinary team (primary care pharmacists, GPs, nursing home physicians and nurses) aimed at reducing inappropriate psychotropic drug prescriptions based on a therapeutic guideline for treating BPSD (created by a multidisciplinary team) as compared to standard care. Before implementing the medication review, one general practitioner (GP) and one primary care pharmacist underwent a training phase focused on managing patients with BPSD. Then, before each scheduled visit with the nursing home physician and nurse, the medication the patient was receiving was evaluated by the GP and pharmacist to detect any incidents related to the prescription (e.g., duplicates, inappropriate drugs) and to determine which prescription-related aspects required evaluation of the patient for decision-making (continue or discontinue a drug). As for the working session in the nursing home, before conducting the medication review (of all drugs included in the therapeutic plan), the GP and pharmacist met with the nursing home physician and nurse to establish the patient’s current status with regard to his/her prognosis, level of dependence, and frailty. This was done to facilitate decisions on the need to adapt the intensity of treatment, change treatment, or discontinue treatment based on the risk-benefit ratio for the patient. Patient assessment included the severity of dementia (Global Dementia Scale [GDS]), level of dependence (Barthel index score), cognitive state (Pfeiffer test), and prognosis (in end-of-life patients). Treatment assessment included the indication, effectiveness, and safety (duplicates, interactions, contraindications) of the drugs prescribed. Based on these evaluations, changes in the patient’s treatment plan considered to be appropriate and relevant were proposed. Follow-up of the consensus changes was carried out at 1 and 6 months to determine whether any psychotropic drugs that had been discontinued after the interview should be restarted, the dosage of drugs that had been reduced should be increased, or a new psychotropic drug should be initiated. Over the course of the intervention, the mean number of drugs prescribed per patient decreased from 8.04 at baseline to 5.86 at 6 months and from 2.71 to 2.01 for psychotropic drugs.

The analysis was conducted using the Monitoring and Assessment Framework for the European Innovation Partnership on Active and Healthy Ageing (MAFEIP) tool. In MAFEIP Users Guide Manual [33] describes the basic principles of the tool to better understand the tool and how it works. In this manual is described the MAFEIP tool is founded on the principles of Decision Analytic Modelling (DAM), an approach commonly used to evaluate the health and economic impact of healthcare innovations. More precisely, MAFEIP is based, by default, on a generic three-state Markov model, which provides the flexibility required to be adaptable to a large number of studies that focus on a variety of objectives, implement different interventions and target different cohorts of individuals with different demographic or disease characteristics.

The outcomes for the intervention and control group are calculated by simulating the health status of the target population as it transitions through the three health states defined in the Markov model: baseline health, deteriorated health, and dead (Fig. 1). Each health state is defined by an amount of resource use and quality of life (utility). This represents the average resource use and quality of life of a patient in that health state.

In the simulation, the population moves through the Markov states based on four transition probabilities: 1) Transition from the baseline health state to the deteriorated health state represents a patient becoming ill (i.e., the incidence of the deteriorated health condition); 2) Patients move from the deteriorated health state to the baseline health state when their illness is cured (i.e., the rate of recovery back to baseline health); 3) Transition to the dead state from the baseline state (i.e., baseline mortality in the target population); 4) Transition to the dead state from the deteriorated health state (i.e., excess mortality in the population with deteriorated health).

The MAFEIP tool builds up from a variety of surrogate endpoints commonly used across different studies in order to estimate health and economic outcomes in terms of incremental changes in quality adjusted life years (QALYs), as well as health and social care utilization.

A highly adaptable Markov model with initially three mutually exclusive health states (baseline health, deteriorated health and death) forms the basis of the tool, which draws from an extensive database of epidemiological, economic and effectiveness data. It also allows for further customization through remote data entry, thus enabling more accurate and context-specific estimation of intervention impact.

In our study, the three-state Markov model of the MAFEIP tool was used to describe patients without dementia (baseline health state), patients with dementia (deteriorated health state), and death. Nonetheless, considering there were no patients without dementia, and the probability that patients with dementia would improve and go to the baseline health state was zero. In the MAFEIP-tool it is not possible to dismiss this state. The study combined data from nursing homes with data collected from external sources to populate the model with values for transition probabilities and relative risk of mortality. Incidence rates were derived from data on the prevalence of dementia in institutionalized patients in our reference area. Moreover, estimates on mortality rates were derived from the findings of the DART-AD [2] clinical trial, which reported higher survival rates in patients who were given placebo drugs instead of neuroleptics.

Cost estimates were based on scientific literature, the eSalud database [34] and expert opinion. The cost estimates are direct costs, including medical visits, hospital care, medical tests and exams and drugs administered, among other concepts.

The healthcare costs of patients using the intervention were calculated by means of a medication review that compared drug-related costs of the patients before, during, and after the use of the intervention.

Intervention cost estimates were based on the amount of time spent by the physician, pharmacist, and/or nurse with the patients.

Regarding utility values, the study did not focus on utility calculations but used secondary sources to provide general and averaged estimates [35]. Utility values of the baseline group describe data collected on health-related quality of life (HR-QoL) [36] for people from the same country and of the same age group who do not have dementia. Since the intervention has been shown to have an impact on mortality rates rather than HR-QoL, utility values for the control group and intervention group are similar. (Utility for control and intervention group: baseline 0.70 and deteriorated 0.27).

The cost-utility analysis was performed from the perspective of a health care system with a time horizon of 12 months. It was focused on calculating the incremental cost-effectiveness ratio (ICER) of the intervention and comparing it to the standard of care. In addition to the analyses on the outcomes; we have included the willingness to pay as a threshold to show the relation between this willingness to pay and the cost-effectiveness of the outcomes.

As previously mentioned, the use case takes into account the three-state Markov model depicted in Fig. 1.

Each health state describes institutionalized patients between 65 and 90 years of age. The baseline health state describes patients without dementia whereas the deteriorated health state describes patients with dementia. Records show that the prevalence of dementia in institutionalized patients in our reference area is 40%. This is adopted as the incidence rate for both the preintervention group and postintervention group since the intervention does not focus on reducing the rate of incidence of dementia and therefore does not affect it. In this use case, it was assumed that patients cannot completely recover once they already have dementia and, thus, recovery rates were set to 0 for both groups. (Transition probabilities: intervention and control group, incidence rate: 40 and recovery rate: 0).

Mortality rate estimates were derived from the findings of the DART-AD [2] clinical trial, which reported higher survival rates in patients who were given placebo drugs instead of neuroleptics. Based on the clinical trial, the cumulative probability of survival at 12 months was 70% in the preinetrvention group (those who continued treatment with neuroleptics) vs. 77% in the postintervention group (those who withdrew neuroleptics). Moreover, the results show that the difference between the groups was greater after longer periods: at 24 months, survival rate was 46% vs. 71% and at 36 months, it was 30% vs. 59%. The mortality rates were obtained by first calculating the mean of all the survival rates given (49% vs. 69% or 0.49 vs. 0.69) and then subtracting the mean survival rates from 1. (Calculated mortality rate: 0.51 vs. 0.31).

The MAFEIP tool uses another measure for estimating mortality in a given population: the relative risk of mortality (RR). This is calculated by dividing the mortality rates in the use case by the mortality for a reference condition. In this particular use case, the reference condition is the age-dependent, all-cause mortality of the Spanish population as recorded in the Human Mortality Database (https://​www.​mortality.​org/​). The RR values were a baseline health of 1 and deteriorated health of 1.31 for the postintervention group, and 1 and 1.51 for the preintervention group, respectively.

Patients in the baseline health state had an RR of 1. This is the default value and it prompts the tool to simply use data that is stored in the Human Mortality Database. Those in the deteriorated health state, however, were given an RR greater than 1. Choosing the value RR > 1 leads the tool to interpret this as excess mortality to be added to the existing data from the Human Mortality Database (i.e., mortality rate from database + excess mortality from use of neuroleptics).

The drug-costs were calculated considering the drug prescribed, dose and frequency for then calculate the number of boxes according to the length of treatment or if the length was more than a year it was considered a maximum length of 1 year period time. This was calculated for each drug from patient pharmacotherapy. The final cost was the sum of the number of boxes multiplied by its cost.

Intervention cost estimates were based on the amount of time spent by the physician, pharmacist, and/or nurse with the patients. The total sum of intervention costs calculated from the staff spending 30 min with each patient, in accordance with the therapeutic plan, came to €11,654.40 or €48.56 per patient. Since the patient does not have to pay for the therapeutic guide itself before it is used, the intervention one-off costs were left at 0.

Cost estimates described the average healthcare costs per dementia patient per year as ranging from approximately €3596.88 to €5130.90, adapted from social-economic studies of Alzheimer disease and vascular dementia patients based in Spain [35, 37].

However, for this use case, the intervention specifically aimed to reduce the amount of neuroleptic drugs given to patients. A study of the costs of this intervention included a medication review that compared drug-related costs of the patients before, during, and after the use of the intervention. For this reason, only drug-related costs were considered in the MAFEIP tool.

Drug-related costs per patient were calculated to be €2265.68 before the intervention, €1720.77 at 1 month after intervention and €1539.90 at 6 months after the intervention. All other costs were considered constants. The Table 1 summarizes the results of the medication review and the rest of model parameters. These are transferred to the MAFEIP tool as follows: