Scottish Inpatient Patient Experience Survey 2014 Volume 2: Technical Report

4 Sample

Introduction

• This section of the report presents detail on the design of the inpatient survey sample, the sampling frame used, sample selection, procedures and checks undertaken.
• It also provides detail on the procedures for data transfer of the agreed sample and arrangements around access to data.

Design of sample

• The sampling approach for the Inpatient Patient Experience Survey was designed to
  • meet the needs of each NHS Board
  • inform local improvement work
  • meaningfully compare findings across Boards and
  • provide national level results.
• In order to meet these requirements, a stratified sampling strategy was developed through consultation between Analytical Services Division (ASD) within the Scottish Government, Analysts with the Information Services Division of National Services Scotland (ISD) and representatives of each NHS Board.
• Sample strata were agreed with each NHS Board to meet their needs in providing local results at the most useful level. These strata were defined either at site (hospital) level, sub-site (directorate or grouped specialties) level or as groupings of smaller hospitals.
• Table 4 provides detail on the number of eligible patients within each NHS Board.

Sampling frame

• Eligible patients were identified from an extract from ISD's Scottish Morbidity Register database (SMR01) of hospital admissions.
• Eligible patients were defined as adults (aged 16 years old and above on discharge from hospital) who had an overnight stay in hospital as an inpatient (defined as one night or more where the individual was in hospital at midnight in an inpatient bed) between April 2013 and September 2013. A sampling period of 6 months was chosen for this survey rather than the one year period used in previous surveys. This period was chosen to shorten the maximum possible delay between a patient's discharge and receiving a questionnaire to 9 months from 15 months if a year sampling period was used.
• Certain patients were excluded from the survey, as follows:
  • Privately funded patients receiving care in NHS or private hospitals
  • Scottish NHS patients treated in hospitals outside of Scotland but whose care was commissioned by an NHS Board
  • Patients who were not resident in Scotland, with the exception of Northumberland residents treated in NHS Borders
  • Patients who were receiving care as an outpatient or day case which did not result in an overnight stay
  • Patients who were expected to have an overnight stay at admission but did not
  • Patients who stayed in hospital for termination of pregnancy
  • Patients who stayed in a hospital maternity unit
  • Patients who were known to be deceased, i.e. those who had death recorded in SMR01 records
  • Patients treated in a private hospital
  • Patients treated in a hospice
  • Inpatients who were being treated for a mental health condition in a mental illness hospital
  • Inpatients who were resident in a long-stay hospital
  • Inpatients who were being treated in a learning disabilities unit, and
  • Inpatients aged 15 or less on discharge
  • Patients within NHS Greater Glasgow & Clyde where the diagnosis was of a sensitive nature or which suggested that they were incapable of completing the questionnaire satisfactorily. Note that this does represent an inconsistency in the sampling between NHS boards however, time constraints were such that it was not possible to either apply this consistently across all NHS Boards or hold discussions with NHS Greater Glasgow & Clyde about removing this exclusion.
• Based on the selection and exclusion criteria, the total number of inpatients who were eligible to take part in the study was 177,690^[4].

Table 4 Table 4 summarises the total number of eligible inpatients within each NHS Board.

Table 4 Total number of eligible patients within each NHS Board including deceased patients

Sampling method

• The sample size calculation was based on a 95% confidence level. The margin of error was set as +/- 5% which yielded results which would be accurate to within +/- 5 percentage points (increased to +/-7 percentage points in sites with less than 1,000 patients annually to avoid sampling too many patients in smaller sites).
• Expected response rates within each NHS Board were calculated based on the response rates achieved by the 2012 Survey. These were used to adjust the required sample size to allow for non-response.
• In previous years the proportion of inpatients sampled at any hospital was limited to 60% to avoid surveying too many patients from smaller hospitals, however as the sample for this survey was based upon a 6 month period, rather than one year in previous surveys, this limit was removed.
• Table 5 shows that the total sample size allowing for non-response was 57,132, based on the requirement to achieve 29,485 completed returns for the stated accuracy.

Table 5 Required sample and total selected sample

Sampling procedures

• ISD compiled a list of eligible patients appropriate for inclusion in the inpatient survey by combining a list extracted from ISD's Scottish Morbidity Register database (SMR01) with a list of patients provided by NHS Greater Glasgow & Clyde. This was necessary as ISD do not hold data on the specific hospital of treatment for patients treated within the group of hospitals including Beatson West of Scotland Cancer Centre, Drumchapel Hospital, Gartnavel General Hospital, Glasgow Homeopathic Hospital and Glasgow Western Infirmary. This resulted in the production of a list of eligible patients.
• Once the list of eligible patients was compiled, ISD eliminated duplications, incomplete records, deceased patients and ineligible patients to determine a final sample frame.
• ISD then randomly selected the required number of patients for each strata from the sample frame, based on a random number generated on the uniform distribution U(0,N), where N is the total number of patients within that sampling strata, and allocated each patient a unique number.

Data transfer

• ISD and the contractor ensured that all data was transferred securely via secure File Transfer Protocol (sFTP) with secure login. Once files were received they were immediately deleted from this system. The data transferred to the contractor included patient name and address, hospital code and name, NHS Board of treatment and a unique identifier, generated for the specific purposes of this survey.
• The fields included in the data transfer were limited to those required to mail the survey pack to the individual patient and to identify which episode of care they should complete the survey in relation to.
• On completion of the survey fieldwork and data capture (described in more detail in Chapters 5 and 6) the contractors sent the complete anonymised data files electronically to ISD through secure FTP. This enabled ISD to undertake analysis for the reporting.

Access to data

• In accordance with the NHS ethical approval for the survey, contractors were given access to the names and addresses of the people who were to be sent the survey. All personnel were governed by the NHS Research Governance Framework for Health and Community Care. The survey guidance manual detailed the research governance responsibilities and accountabilities of all those involved in the survey.