Every publicly funded hospital stay in NZ longer than 3 hours is required to be collected by DHBs to form a national dataset. This includes people treated in ED for more than 3 hours and/or admitted as inpatients. A range of information is recorded with each eligible stay including information about the facility, length of stay, cost, procedures and diagnoses. Outside the IDI the data are known as the National Minimum Dataset (NMDS). Within the IDI the data are known as ‘hospital discharge data’. Hospital discharge data are a valuable and detailed resource of importance to a broad range of research areas.

The aim of this guide is to:

  • Describe the hospital discharge data currently available in the IDI
  • Discuss some of the ‘tips, trips and traps’ in using hospital discharge data in the IDI


Hospital discharge data is a key resource for health researchers. Nationally and internationally hospital discharge data is used for a range of purposes including health services research & planning, public safety, surveillance, informing policy, public health and disease registries (Schoenman, 2005).

In New Zealand, routinely collected hospital discharge data goes back to 1988 and has been available for use by researchers and policy makers for decades. The Ministry of Health (MoH) are the custodians of this data; as such they facilitate access to this data directly. The variables contained within the NMDS, their coding and limitations for use are well-documented in the NMDS Data Dictionary (National Health Board, 2014).

What hospital discharge data can I find in the IDI?

There are four tables in the IDI that relate to hospital discharges:

  • Publicly funded hospital discharge events (pub_fund_hosp_discharges_event)
  • Publicly funded hospital discharge diagnoses (pub_fund_hosp_discharges_diag)
  • Privately funded hospital discharge events (priv_fund_hosp_discharges_event)
  • Privately funded hospital discharge diagnoses (priv_fund_hosp_discharges_diag)

The public/private distinction is made on the basis of funding, rather than the type of facility. For example, surgery that is undertaken at a private hospital with funding from ACC will be included in the publicly funded hospital discharge dataset.


  • In the events tables there is one row per discharge from hospital. Since an individual can be discharged from hospital more than once in their lifetime, there can be multiple rows for the same person that can be ordered by ‘Event ID’.
  • The events table contains a range of information about the event including start and end date of visit, cost weight, and the hospital department the patient was discharged from.


  • In the diagnoses tables there may be two rows per diagnosis/procedure: one for ICD-9 coding and another for ICD-10 coding.
  • Each row includes a clinical code describing the diagnosis and a clinical coding system variable that indicates which ICD coding system was used to generate the clinical code.
  • If there are multiple diagnoses per person per discharge (event), they can be ordered by ‘Event ID’, ‘clinical coding system variable’, and ‘diagnosis sequence’.
  • Every discharge must have one and only one ‘principal diagnosis’ that reflects the primary reason for admission. Other relevant diagnoses may also be included. Researchers should consider whether restricting to patients with a principal diagnosis of interest is more appropriate than including all patients with a diagnosis of interest.

The events and diagnoses tables can be linked using the event_id variable.

Users that have previously obtained hospital discharge data directly from the Ministry of Health may or may not find the format of hospital discharge data in the IDI unfamiliar, depending on what kind of extract they requested from MoH.

Hospital discharge data from 1988 is available in the IDI. As at September 2019, the most recent data available was June 2018. As this is ‘discharge data’, patients admitted to hospital prior to this who have yet to be discharged are not included.

For more details about what specific variables are included in these tables, The IDI Data Dictionary for ‘Publicly funded hospital discharges – event and diagnosis/procedure information’ is very informative (Statistics NZ, 2015). Although there is not currently a data dictionary available for privately funded discharges, the variable names and coding are the same although the number of variables are reduced.

Diagnosis and procedural descriptions for each ICD code are available from the Ministry of Health, and may require you to sign a confidentiality agreement.


  • Publicly funded hospital discharge data goes back to 1988 giving researchers 30 years of data to work with.
  • Each discharge is accompanied by diagnostic information coded by professional clinical coders.
  • Patients can have multiple discharge events over time. Multiple discharge events per person can occur either for the same or separate injury /diseases. Depending on the reason for admission, it may be possible to estimate person-based incidence from hospital discharge data.  For example, the incidence of injury events can be estimated by excluding readmissions for the same injury event where readmissions are identified using: a person identifier and dates of admission, discharge and injury (Davie, 2011).


Funding source

  • While it is mandatory for all publicly funded hospital discharges to be reported to the Ministry of Health, this is not the case for privately funded discharges, which may be less complete although this has improved over time. Privately funded hospital discharge data are stored in a separate table in the IDI.
  • The quality of coding for publicly funded discharges from private hospitals may not be of as high a standard as that from public hospitals.

Changes over time

  • What is included (or not) in hospital discharge data has undergone many changes over the years. This is true for both which discharges are included and what variables are collected. Further details are available in the NMDS data dictionary. (National Health Board, 2014).
  • Short stay Emergency Department events (SSED) have been reported inconsistently from District Health Boards (DHBs) across time. Patterns over time and by DHB have been shown to vary considerably for hospitalised injury depending on whether SSED events are included or not. (Davie, 2019) The Ministry of Health recommends that SSED events be excluded from any regional or longitudinal analyses prior to July 2012. (Ministry of Health, 2015). The publically funded hospital discharge dataset in the IDI contains a SSED flag to allow you to do this.


  • Researchers should be aware that, on average, the number of diagnoses recorded per discharge event has increased over time which likely indicates changes in coding practices.
  • In mid-1999 New Zealand hospitals switched from using the International Classification of Diseases Ninth revision (ICD-9) coding scheme to using the Australian Modification of the Tenth revision (ICD10-AM) to summarise the injury(s)/disease(s) of patients.
  • The same diagnosis for a patient may occur multiple times in the table under different coding systems, so researchers should be wary of double counting diagnoses. You may need to remove duplicates or one set of codes.
  • Sometimes individuals may be transferred between hospitals or services and be recorded as separate events. These transfers can be identified through ‘event end type’ codes (researchers wanting to do this should contact MoH for details). Depending on the aims of the research, users may want to count transfers as separate events, or combine them with the original admission.
  • Sometimes ICD procedural and diagnosis codes may overlap. Researchers who are trying to select ICD diagnosis codes should restrict to diagnosis_type_code ‘A’ and ‘B’ (procedures are recorded under diagnosis_type_code ‘O’).
  • Hospital discharge data is an events based dataset. If there are missed links when discharge data is linked to other IDI datasets there is the potential to underestimate hospitalisation rates. Differential missed links by key study variables should be considered as a potential source of study bias.
  • To avoid double counting with deaths recorded in the Mortality Collection, for research that presents deaths and hospital discharge data, researchers should consider excluding patients that died in hospital (i.e present only non-fatal hospital discharges). For example if you wish to look into the prior hospitalisations of those in the mortality collection, you probably want to exclude the hospitalisations that end in patient’s actual death. The variable ‘Event end type’ can be used to identify patients that died hospital or in the Emergency Department acute facility, and those that were discharged for organ donation (eg codes DD, DO, and ED).


New Zealand’s hospital discharge data is of high quality and it is well-documented. Having hospital discharge data in the IDI greatly increases the value of this dataset by enabling linkage to other administrative and survey data available in the IDI. That said just like for any administrative dataset, users of the hospital discharge data should be mindful that research use is not the primary purpose of this data and as such, care should be taken with understanding the structure of the data and its interpretation. Interpreting trends over time should be done cautiously as changes may reflect coding practices, changes in hospital admission or treatment policies (service delivery issues) or funding.


Davie G., Samaranayaka A., Langley JD, Barson D. (2011) Estimating person-based injury incidence: accuracy of an algorithm to identify readmissions from hospital discharge data. Injury Prevention 17:338-342.

Davie G, Barson D, Simpson JC, Lilley R, Gulliver P, Cryer C. (2019) Using hospital discharge data for injury research or surveillance? An observational study illustrating the impact of administrative change. Injury Prevention 25:540-545.

Ministry of Health. 2015. Factsheet: Short stay emergency department events. Wellington: Ministry of Health. Available from https://www.health.govt.nz/publication/factsheet-short-stay-emergency-department-events

National Health Board. (2014) National Minimum Dataset (Hospital Events) Data Dictionary. Wellington: Ministry of Health. Published in 2014 by the Ministry of Health PO Box 5013, Wellington, New Zealand. Available from https://www.health.govt.nz/system/files/documents/publications/nmds_data_dictionary_v7.9.pdf

Schoenman JA, Sutton JP, Kintala S, Love D, Maw R. (2005) The value of hospital discharge databases. Healthcare Cost and Utilization Project (HCUP) report, Agency for Healthcare Research and Quality (AHRQ), Maryland, United States. Available from https://www.hcup-us.ahrq.gov/reports/final_report.pdf

Statistics New Zealand. (2015) IDI Data Dictionary: Publicly funded hospital discharges –event and diagnosis/procedure information (November 2015 edition). Available from http://archive.stats.govt.nz/browse_for_stats/snapshots-of-nz/integrated-data-infrastructure/idi-data/publ-fund-hosp-disch.aspx



Original 30/8/2019, by Gabrielle Davie, June Atkinson, Dave Barson, Vivienne Rijnberg, Mathu Shanthakumar, Andrea Teng, Sheree Gibb


This work is licensed under a Creative Commons Attribution 4.0 International License.