Acid suppressive therapy (AST) is used in a variety of indications in prevention and treatment of gastric acid related disorders. In recent years, an increase in the use of acid suppressive therapy has been documented. It has generally been accepted that these agents are effective and safe. However, very recently, reports describing hypomagnesaemia, hypocalcemia, vitamin B12 deficiency and enteric infections, were published.

The goal of this study is to evaluate 1) how many cases of long-term ambulatory treatment with AST are initiated during hospital stay (= ASTRA-1), and 2) assess the appropriateness of initiation of AST and the appropriateness of continuation of AST at discharge in a subset of the hospitalized patient population (= ASTRA-2). Finally, the pharmacoeconomic impact of inappropriate AST during hospitalisation will be calculated from these data.

The primary endpoint is defined as the number of patients on long-term ambulatory AST after initiation in the hospital. Secondary endpoints are the appropriateness check of AST both at the level of initiation during hospitalization and at discharge (appropriate continuation); also the impact on the hospital budget (lump sum budget) will be calculated, and potential cost avoidance for long-term ambulatory inappropriate AST will be estimated from a healthcare payer perspective for a period of 1, 3, 6 and 12 months.

ASTRA 1

A retrospective, multi-centric, observational study will be carried out in 9 Belgian hospitals. Patients will be recruited hospital wide in all participating hospitals.

All hospitalized adult patients (defined as 18 years or older at the moment of discharge) discharged from participating hospitals from in 01-07-2012 until 01-07-2013 will be included, if they were treated at any moment during hospitalization with AST (both H2A, defined as all drugs belonging to ATC A02BA, and PPI, defined as all drugs belonging to ATC A02BC). Younger patients are excluded as they are often treated with compounded suspensions and these preparations are impossible to match with the dispensed medication. Fulfillment of these criteria will be based on queries retrieving data from the hospital information system. The query will generate, for each participating hospital, a list of unique hospitalizations. In order to guaranty the patients privacy, datasets will be made anonymous by a trusted third party (see further in protocol). Reporting will be anonymous, never on an individual patient, physician or hospital.

For this list, all potentially interesting covariates will be documented by adding coded patient-related data, in order to allow determination of associated risk factors by uni- and multivariate analysis. The following patient-related information will be requested from the HIS: number of participating hospital, date of admission and discharge, age, weight, sex, region of living, length of stay, diagnosis on admission, hospital outcome in terms of mortality, comorbidity details (based on ICD-9 coding), presence of gastro(duodenal) investigations during hospital stay (eg. gastroscopy), discipline of treating clinician and/or hospital ward.

Patients with multiple hospitalizations during the inclusion period will be included. Each separate hospitalization will be counted as a separate line in the database. Bias is expected to be minimal as long-term use of AST will exclude the subsequent hospitalization periods because the patient would not be naive to AST on admission.

The sample size of the dataset is not limited for ASTRA-1 in order to provide the maximum power to the study, but will include a minimum of 5000 patients.

The datasets, along with covariates and related data, will be anonymized by Custodix, a Trusted Third Party, experienced in anonymizing hospital data. All anonymized datasets will be linked by the trusted third party over the anonymized patient number with the anonymized IMS-database of ambulatory dispensed medication. All further queries will be performed on this anonymized set of data which combines patients covariates and ambulatory consumption of relevant medication.

Patients on long-term ambulatory AST (defined as continued use during > 3 months) without ambulatory use before hospitalization are the ones we would like to investigate in the primary endpoint analysis of this study.

We will determine the number of those patients vs. all discharged adult patients and also vs. the number of adult patients with AST treatment at any moment during hospitalization.

More in detail, descriptive statistics will be carried out using appropriate parametric or non- parametric tests (Student t-test or Mann-Whitney U-test, based on distribution of data) to describe the demographics of the included patients.

Patients will be divided in 3 groups: 1) patients on AST before admission (whether therapy is stopped or continued during hospitalization), 2) patients started on AST during hospitalization but stopped before or at discharge and 3) patients started on AST during hospitalization and continued use after discharge. For patients in group 3, we will describe what the proportion of patients is with use of AST at +1, 3, 6 and 12 months after discharge, therefore group 3 will be subdivided along with these proportions.

To investigate potential risk factors correlating with long-term use, groups 2 and 3 will be compared using uni- and multivariate analysis evaluating associated risk factors such as age, sex, weight, comorbidities, type of wards, and concomitant medication. A subgroup comparison between groups 2 and 3(b+c+d) will be carried out to discuss risk factors associated with continued long-term use.

Statistical analysis will be carried out in SPSS or SAS by the investigator in collaboration with L-Biostat (KU Leuven).

Based on this dataset an estimate on cost of AST from a healthcare payer perspective will be calculated using the list price (= public price) at 01-07-2013 of the involved medication. In combination with the results of ASTRA-2 the impact of parameters like percent of appropriateness, duration of therapy, diagnosis and comorbidities will be studied in a multivariate analysis. As the main dataset from ASTRA-2 comes from UZ Leuven, the other participating hospitals are each requested to review 20 medical files from their hospital in order to investigate possible extrapolation bias or to support the findings.

ASTRA 2

A retrospective, non-interventional, multi-centric study will be carried in the participating hospitals. Patients will be recruited hospital wide in all participating hospitals. The main number of inclusions will be in the University hospitals Leuven, inclusions in other participating hospitals will be used to investigate possible extrapolation bias and to support the findings.

All hospitalized patients discharged from the participating hospitals from 01-07-2012 until 01-07-2013 will be included. Exclusion of deceased patients, juvenile patients (defined as younger than 18 years at time of discharge), critically ill patients, patients without a startup of AST and patients without AST at discharge leads to the subgroup that is subject of this research. An overview of the inclusion process is given in figure 8.

Patients admitted at the ICU, during hospitalization, will not be included as literature concerning the use of AST (especially as SUP) in the ICU setting (for critically ill patients) is already exhaustive 3,15.

Startup or initiation of AST during hospitalization is defined as treatment with a H2A (ATC-classification A02BA) or a PPI (ATC-classification A02BC) starting from day 3 or later of the hospitalization period. Patients with H2A/PPI treatment before hospitalization will be excluded as their AST is not initiated in the hospital.

The patient inclusion will be performed based on queries retrieving data from the hospital information system (admission and discharge data) and from the CPOE (AST prescriptions) of the participating hospitals.

When patients were readmitted to the hospital during the study period, the different hospitalizations will be evaluated separately.

Assessment of appropriateness of AST will be performed by medical record review by trained staff pharmacists of the participating hospitals, who can be assisted under close supervision by master students of the Faculty of Pharmaceutical Sciences, and are led by a validated procedure.

The method of reviewing will be subject to an interrater validity evaluation using kappa-statistic. The sole intend of the record review is to gather data on the use and appropriateness of AST medication and will not intervene in any way in the therapy nor the patient - physicians relationship. Reporting will be anonymous, never on an individual patient, physician or hospital.

The estimated prevalence of inappropriate AST is unknown. Assuming a worst case scenario of 50 % inappropriate AST, 384 patients would need to be included in the trial for a confidence interval of 95% (see table below). Assuming a rate of 30 % of inappropriate AST, 323 patients would need to be included. To guarantee this sample size at least 500 medical files will be screened in order to compensate for potential dropouts. These 500 medical files will be randomly selected out of all the included patients using a random code generator. In UZ Leuven at least 500 medical records will be screened, supplementary other participating hospitals are requested to screen 20 medical files each.

To guarantee a representative sample of the studied population the sample will be checked for proportional allocation for the baseline variables sex (male/female), age (minus 65 years or older) and admission label (internal or surgical). In order to prevent selection bias we will not use random sampling with stratified allocation as we plan no intervention groups.

Factors justifying AST in non-critically ill hospitalized patients are based on a literature review an will be used for evaluation of appropriateness and pinpointing risk factors for inappropriate longterm use of AST after hospitalisation.

The impact on cost savings for the hospital population will be studied from the healthcare-payer perspective. Therefore the official list price for acquisition, which equals the reimbursement price for hospitalized patients, at 01-07-2013 for the hospital will be used. In case of inappropriate therapy or late IV/PO switch there is an opportunity for cost saving. This will be subject to a sensitivity analysis. Besides direct cost savings there may be a possible long term cost-avoidance, but this is more difficult to calculate because side-effects of long term therapy are just popping up in research and cannot be monetized yet.

Extrapolation of the percentage of patient on inappropriate AST in the hospital to inappropriate continued long-term use will be performed by combining the results of ASTRA-1 and ASTRA-2. In this way the impact on cost from the healthcare payer perspective can be calculated. In the analysis we will also look at patients compliance, switch to less expensive products and perform a sensitivity analysis on cost saving as function of timely stopping the inappropriate therapy.

Inclusion Criteria:

• Discharged from participating hospitals from July 2012 till July 2013.
• Specific for part 2: startup of AST in hospital and AST on time of discharge

Exclusion Criteria:

• Deceased at discharge
• Critically ill patients
• Younger then 18 years old at time of discharge
• Requested by treating physician (negative consent)

Part 1: All adult patients discharged from the participating hospitals in the period July 2012 - July 2013.

Part 2: Subset of Part 1 - random selection of 600 patients with startup of AST in the hospital and AST on date of discharge.