New Zealand Quality of Healthcare Study
reportposted on 2021-06-16, 00:13 authored by Peter DavisPeter Davis, Robin Briant, Stephan A Schug, Alastair ScottAlastair Scott, Wendy Bingley, Julie Fitzjohn, Roy Lay-YeeRoy Lay-Yee, Lisa Fellowes, Sandra Johnson
Downloadable files (above)
NZ Quality of Healthcare Study – Nurse Review Form
This was the first stage review form, undertaken by Registered Nurses on the research team, to ascertain for each record if it met any of 18 screening criteria selected as potentially indicative of an adverse event.
NZ Quality of Healthcare Study – Medical Review Form
Records that were marked as potentially indicative of an adverse event were then forwarded for full review by trained and experienced Medical Officers, also on the research team. This was an instrument relying on structured implicit review (guided exercise of professional judgment), and aimed to determine whether each marked record was indeed associated with an adverse event, and then if so, to characterise that adverse event according to key clinical criteria.
NZ Quality of Healthcare Study – Data Access Request Form
At each stage of review, both quantitative data and free-text summaries were collected, so the data collected by this project are highly sensitive. Hence we cannot have them freely accessible here, and instead provide this form to initiate contact with anyone interested in accessing any of these data sets. Please fill in the form and email it to firstname.lastname@example.org to begin this process.
For information on the publications from this study, see https://www.auckland.ac.nz/en/arts/our-research/research-institutes-centres-groups/compass/surveys/nzqhs.html
The main report is still available at https://www.health.govt.nz/system/files/documents/publications/adverseevents.pdf
AbstractThe objective was to assess the occurrence, impact and preventability of adverse events recorded in New Zealand public hospitals.
A two-stage retrospective review was carried out on 6,579 medical records. These were selected by systematic list sample from admissions for 1998 occurring in 13 public hospitals throughout New Zealand providing acute care and with over 100 beds, excluding specialist institutions. Following initial screening, medical records were subject to structured implicit review (that is, the guided exercise of professional judgement) by a team of trained medical officers, using a standardised protocol.
The information available in the sampled medical records was of a quality that permitted the adequate identification and analysis of adverse events. The processes and instruments used in comparator studies internationally were applied in the New Zealand setting with little difficulty. Reliability and validity measures displayed only moderate levels of agreement, however.
Analysis of the 850 adverse events identified revealed a distribution, impact, and clinical context comparable with other studies. Adverse events (which may have occurred either within or outside public hospitals) were associated with 12.9 percent of admissions. Approximately 35 percent of adverse events were classified as highly preventable. Although less than 15 percent of adverse events resulted in permanent disability or death, an average of over nine days per event was added to hospital stay.
Nearly a fifth of events originated from outside public hospitals, only a quarter of which arose in another institutional context. Patient age was an important risk factor for an adverse event. There were distinct patterns according to clinical and administrative context. Systems errors featured prominently in the analysis of areas for the prevention of recurrence.
This study provides the base parameters necessary to inform our understanding of patient safety and the quality of care in New Zealand public hospitals. These data have important managerial and clinical implications. Further work could be done on subgroups of patients and on the clinical detail available in the data. The investigation provides a baseline for more targeted studies and for quality improvement interventions. It also points to the importance of similar research on the sources and characteristics of adverse events outside public hospitals.
Fieldwork was conducted over the period from the beginning of July 1999 to the end of May 2000, by a team of four Registered Nurse (RN) screeners and three or four Medical Officer (MO) reviewers overseen by the Project Manager. An Expert Reviewer arbitrated on discrepant judgements (where an RN and an MO disagreed) and carried out an independent review of a subsample of selected medical records. Before the commencement of data collection, fieldworkers undertook an intensive training course.
Data collection took place over a period of three weeks at each hospital. Each sampled case was allocated a study identification number so that identifiers allowing linkage to hospital records could be deleted to maintain patient anonymity.
A stratified two-stage cluster sample design was used:
1. A nationally representative sample of 13 was generated from the 20 public hospitals with more than 100 beds.
2. A random sample of admissions was drawn within each hospital.
Sampling of hospitals followed stratification by hospital type and geographic area across New Zealand. The 3 strata were:
1. Six large tertiary service facilities.
2. Seven secondary service facilities with more than 300 beds.
3. Seven secondary service facilities with fewer than 300 beds.
The national sample comprised all six hospitals from the first stratum, probability proportional to size samples of four hospitals from the second stratum and three hospitals from the third stratum.
The New Zealand Health Information Service (NZHIS) selected a random sample of 575 admissions from each of the sampled 13 hospitals for the year 1998. The selected time of admission for sampled cases signalled an index admission (the sampled admission) and provided the point of reference in adverse event (AE) determination.
The sampling frame for each hospital was the list of all eligible admissions in that hospital. This number was divided by the 575 to be selected in order to come up with a systematic sampling interval, following from a random starting point between 1 and 100. This generated samples of the required size according to standard principles of systematic list sampling.
In order to assess the representativeness of the sample, the distributions of key patient characteristics were compared with the patterns for all New Zealand public hospital admissions in 1998. It should be noted that the data in this study represent approximately a 1 in 100 sample of all publicly funded hospitalisations. The sample medical records were closely representative of all New Zealand public hospital admissions in a number of key demographic and clinical characteristics, including age, gender, ethnic group, discharge status and mortality. Length of stay in the sample appeared to be shorter than the average for all publicly funded hospitalisations in New Zealand for 1998.
Characteristics of data collection situation
Routine quality checks were carried out to improve the quality of information gathered. During both the screening and review stages, forms were checked for completeness and adequacy. During the review stage, RN and MO discrepancies in judgement of criteria presence or AE determination were checked and if necessary forwarded for adjudication to the Expert Reviewer.
At the stage of data entry from the completed forms, standard checks for invalid, out-of-range, inconsistent and missing data were used to identify errors. Where medical knowledge was necessary, the Expert Reviewer was consulted.
Actions to minimise losses
In the event of a home being unoccupied at the time of the initial visit, at least 2 other visits were made at different times of the day, in an attempt to make contact.
The cases needed to be weighted to account for unequal selection probabilities in the sample design. Each hospital was given a weight inversely proportional to its selection probability when calculating estimates of rates, proportions and means. In actuality, weighting the cases gave similar results. Standard errors also needed to be adjusted for the two-stage cluster design.