NutritionDay in Worldwide Hospitals: An International Audit and Registry on Nutrition and Outcome

One day international cross-sectional audit in all types of hospital wards. Intensive care units are excluded. Data are collected with the help of questionnaires. Anonymised data entry into the audit database is done with individual and anonymous center and unit-codes.

The data collected consist of four parts:

1. Unit organisation and structures: Structural information about the unit (one sheet / unit) to be filled by the unit supervising physician together with the nursing head.
2. Patient's demographics \& medical information: Demographic profile, diagnostic category based on ICD 10 and nutritional interventions for all patients (one line / patient) to be filled by a responsible person from the medical staff. Oncological patients have an additional caregiver sheet to define precisely cancer stage and specific treatment.
3. Individual self-administered patient questionnaire: Each patient should document her/his nutritional intake during the study period. In addition patients may be asked to fill a questionnaire about changes in nutritional habits and reasons for decreased nutritional intake from the patient's perspective. Oncological patients have an additional optional patient sheet to determine the specific nutritional problems and supportive treatments used.
4. Individual patient outcome: hospital cohort: at hospital discharge or day 30, whatever comes first: date of unit discharge, date of hospital discharge, site of discharge and health status.

Participation rules:

1. A user needs to register to nutritionDay as a member to the nutritionDay network. Each may choose a personal user name. Each user needs to provide a valid email address. After responding to a validation email the user is registered with the chosen user name. User details are stored on a system that is not connected to the nutritionDay registry at any time. One user may serve as contact for several units and centers.
2. A user may order codes for participation for one or several centers and units. Access to the registry is only possible with a center code and unit code. These codes are selected from a list of random numbers.

Audit/registry recruitment plan:

Participation to the registry is voluntary. There is actually no participation fee. All necessary information can be obtained from a dedicated website (www.nutritionday.org). Participation can be promoted via international and national scientific societies, universities, health care organisations or governmental agencies as well as via advertisement at international and national congresses. The target would be all types of hospital wards within hospitals of different sizes and level of care.

Risk and benefit assessment:

The benefit for each patient is that awareness and knowledge about nutrition related factors and treatment options in the treating unit is increased. There is no individual risk since the audit is purely observational.

The benefit for the individual unit is to receive an extensive benchmarking report displaying the unit data in comparison with all units from the same specialty from the previous 3 years immediately after the end of data entry and a validation step. Units or groups of units may request specific reports that can be obtained only after a case by case agreement and financial coverage. All data used as a reference for benchmarking purpose are from units where a minimum of 60% of actually present patients have been recruited and the outcome at day 30 is available in more than 80% of these patients. There is no risk for the unit since anonymity of unit is structurally strictly maintained.

The benefit for the registry up-to-date data enabling benchmarking in pace with medical and care evolution. Moreover the registry data are used for research of the scientific community.

The aim is to recruit a minimum of 10-50 units with 20-30 beds per participating country. A minimum of 10 units per medical specialty will be necessary to allow specialty adjustment.

Data security:

On the datasheets the unit and the hospital/center are identified by a numeric code delivered after application to the nutritionDay coordinating center by an automatic system. The only requirement is a valid email address for direct communication with the unit.

Patients are usually identified on the locally used questionnaires by initials and age, but use of initials is not mandatory. Consecutive numbering is also possible. Only the participating unit has to trace patients identifier to be able to collect hospital outcome at day 30 after nutritionDay and to answer automatic requests for data clarification during the data quality feedback.

During data entry into the electronic registry only anonymous codes for center, unit and patients are possible. Thus the data handling centre cannot trace data back to an individual patient. The access to data entry is protected by anonymous center-code and unit-code.

Typically all data are collected via a dedicated website accessed via www.nutritionday.org .

The protected data server is run by the Center for Medical Statistics Informatics and Intelligent Systems (CEMSIS) of the Medical University Vienna. The data server is mirrored and backed up. The data server is protected within the University Firewall against external access.

Data feedback and individual unit report

All participating units are entitled to receive a bench-marking report from the registry. The report generator is started by the participating unit, when data entry has been completed. As a first step each unit receives a data feedback sheet that is based on an electronic data plausibility and missing data analysis. After stating that all data are correctly entered the final report can be generated. This final report offers complete descriptive statistics of the unit data compared to the reference data from the previous 4 years of the corresponding specialty. Only data from units fulfilling a high data quality standard are used for the reference. More than 60% of the patients present in the unit need to participate and 80% of these patients need to have their outcome recorded. Optionally the period used for comparison can be extended until 2006. All descriptive statistics represent prevalent data and are not corrected for cross-sectional sampling to allow direct data control and interpretation.

All downloaded reports of a unit are consecutively numbered and stored for documentation purpose.

Data analysis and modelling:

The first aim of the project is to serve local specialty specific and up-to-date benchmarking with regard nutrition and feeding status , risk profile and nutrition care. Given the typical cross-sectional sample of 20-30 patients frequent risk factors can be easily compared because such risk factors as recent weight loss or less than normal eating are observed in one third to half of the patients.

The second aim is scientific data analysis. The actual research questions are:

• Identifying the risk factors associated with decreased eating.
• Analysing the impact of risk factors for increased length of hospital stay
• Analysing time-trends in risk profile and nutrition care
• Analysing the economic impact of nutrition risk factors on the health care system
• Updating the PANDORA score if major change in performance is identified

The actual database includes 103 920 patients for the period 2006-2012 and allow the following precision for estimating the prevalence of risk factors:

Risk factor prevalence 1% precision as 95% CI 0.93-1.07%

Risk factor prevalence 5% precision as 95% CI 4.87-5.14%

Risk factor prevalence 10% precision as 95% CI 9.81-10.19%

Risk factor prevalence 20% precision as 95% CI 19.75-20.25%

Risk factor prevalence 30% precision as 95% CI 29.72-30.29%

Risk factor prevalence 40% precision as 95% CI 39.70-40.30%

Given the typical participation of 12 000-20 000 patients annually, the investigators expect to have a sufficiently precise database for benchmarking, when using the last 4 years data.

For the scientific data analysis the investigators use COX-regression or general linear model with proper weighting of observations to account for the effect of cross-sectional sampling for events such as discharge or death and linear regression models to analyse length of stay. Whenever several events can occur in a patient a competing risk analysis is performed.

All data analysis will be done at the Dept. for Medical Statistics, Medical University Vienna. After publication of the multinational results, all national datasets will be available for national publication based on a research plan, if the number of wards is large enough to ensure anonymity for the individual ward within the country.