Introduction

An essential component of systems used to describe populations of children admitted to intensive care is a method of recording and classifying diagnoses and reasons for admission. A method of coding reasons for admission is used in most paediatric intensive care units, however, to our knowledge none of the ICU-specific methods are published and there is little uniformity. If comparisons of intensive care workload and outcome are made between units, regions, or countries, ICUs should use the same diagnostic classification system as well as the same system of estimating mortality risk.

In Australia and New Zealand a uniform approach to diagnostic coding was arrived at by consensus when the Australian and New Zealand Paediatric Intensive Care Registry (ANZPIC Registry) was established in 1997. In this report we describe the coding system, highlight the benefits of a uniform approach, and promote the need to develop a uniform international approach to coding diagnosis and reasons for admitting children to intensive care.

Materials and methods

During the design phase of the ANZPIC Registry, the diagnostic coding systems in use in the region were reviewed. Questionnaires were used to obtain expert opinion and a consensus meeting was held in October 1996. This process identified the following system requirements: (a) to be simple enough to be used accurately at the bedside by intensive care physicians and nurses, (b) to specify the main reason for intensive care admission and to allow a number of associated conditions to be coded in addition, (c) to enable the common conditions in intensive care to be coded specifically and uncommon conditions to be classified under diagnostic groups, (d) to enable coding of the common procedures leading to admission when recovery from the procedure is the main reason for admission, and (e) to enable coding of aetiological factors for injury and infection.

The ANZPIC Registry diagnostic codes are listed in Appendix 1 (these can be downloaded in PDF and speadsheet format from http://www.anzics.com.au/paed/index.htm). The instructions for using the ANZPIC registry diagnostic codes are:

  1. 1.

    Code the reason most directly responsible for ICU admission as the principal diagnosis.

  2. 2.

    Code up to five associated diagnoses.

  3. 3.

    For patients admitted primarily for recovery after a procedure use a post-procedural diagnosis for principal diagnosis.

  4. 4.

    For patients having an operative procedure during the admission code the post-procedural diagnosis as an associated diagnosis.

  5. 5.

    Do not use injury mechanism or infection codes for the principal diagnosis (e.g. for respiratory syncytial virus bronchiolitis code bronchiolits as the principal diagnosis and respiratory syncytial virus as an associated diagnosis).

  6. 6.

    If new information (e.g. a test result) becomes available during the admission that allows more accurate coding, amend the original codes but ensure that the principal diagnosis still indicates the reason most directly responsible for ICU admission.

The following examples illustrate how we use the system. In example 1, a child with leukaemia and chemotherapy-induced neutropenia requires admission to intensive care for management of Escherichia coli septic shock. The principal diagnosis is "shock, septic" (832) and the associated diagnoses are "leukaemia or lymphoma" (821), "neutropenia" (823), and E. coli (730). In example 2, a pedestrian is struck by a car and sustains a severe head injury with raised intracranial pressure. The principal diagnosis is "trauma, head" (117) and the associated diagnoses are "MVA pedestrian" (156) and "intracranial hypertension" (316). If this patient is admitted to intensive care from the operating theatre following insertion of an intracranial pressure monitor, the principal diagnosis should still be coded as "trauma, head" rather than "ICP monitor or ventricular drain insertion". In this situation recovery from the procedure is not the main reason for intensive care admission.

To assess the use of the system we reviewed 19,249 patient records submitted to the ANZPIC Registry between 1997 and 2000. Eleven intensive care units submitted data. One unit continued to use a local system of diagnostic classification and map the local codes to the ANZPIC codes before submission. Records from this unit were excluded. The frequency of use of each of the diagnostic codes was calculated and the number of diagnoses per patient tabulated.

Results

A principal diagnosis had been coded in all the records that we reviewed. There were two diagnoses coded in 61% of the cases, three in 29%, four in 13%, five in 6%, and six in 3%. Table 1 lists the ten most frequent principal diagnoses. Within each diagnostic group there is a code for "other diagnosis" for conditions that cannot be otherwise classified. A total of 2,133 (11.1%) patient records had one of the "other diagnoses" used to code the principal diagnosis.

Table 1. The ten most frequent reasons for admitting children to intensive care
Full size table

Discussion

In this report we describe the uniform method used to classify diagnoses and reasons for admitting children to intensive care in Australia and New Zealand. Even though diagnostic classification systems are used in nearly all intensive care units, there is little uniformity and very little has been published in the field. Young et al. [1] recently described the ICNARC coding method, a five-tiered system specific to intensive care units and high-dependency units that was developed in the United Kingdom. The system was derived from data collected during the APACHE II study in the United Kingdom [2], and the applicability of the system to paediatric intensive care is unknown. The tenth edition of the International Classification of Disease (ICD 10) is commonly used to code hospital admission data, however, the applicability of ICD 10 to intensive care is limited. It is not user friendly for ICU staff collecting clinical data at the bedside, and the majority of conditions are not relevant to ICU.

The extremes in approach to coding reasons for admission to intensive care are to code only the primary organ system failing, on the one hand, or to code all possible conditions presenting on the other hand. Any system between the extremes inevitably requires compromise between simplicity of use and the amount of detail. Apart from ease of use, another major benefit of a simple system is that compliance and accuracy are likely to be enhanced. Although there are 370 codes in our system, they are printed on one double-sided A4 page for ease of use. Coders familiar with the system are able to code most admissions in less than 1 minute.

The ANZPIC codes have four levels of classification: non-operative and post-procedural admission, diagnostic group, specific condition and, for injury and infection, aetiological factor. Specific procedures, listed by organ system, are used for patients admitted for recovery after a procedure. An alternative approach is to code the underlying condition (e.g. ventricular septal defect) with a separate flag for post-operative admission. This approach, however, would not differentiate, for example, between patients admitted after pulmonary artery banding and patients admitted after complete repair of a ventricular septal defect. In many circumstances the procedure as well as the underlying condition influence mortality risk and workload. For this reason we designed the system to allow coding of specific procedures.

The advantage of the system is that it is relatively simple and user friendly. There are obvious limitations resulting from the need to limit the number of possible conditions. The analysis of the registry data indicated that all patient records could be classified by operative status and diagnostic group. Eleven percent of patients had "other diagnosis" used for the primary reason for intensive care admission, indicating that these children had specific conditions not included in the coding system. Future refinement of the system will require more detailed analysis of this group of patients. To validate the system further it will be important to assess inter-observer agreement in coding. This will address a limitation of this report.

A number of benefits have been obtained from using a common system of coding diagnosis in our region. Explanatory power has been added to comparisons of unit outcome and length of stay. Add hoc questions have been answered, for example, the incidence and outcome of invasive pneumococcal disease in children admitted to intensive care. The coding also enabled a detailed analysis of the relationship between the reason for admission and adjusted mortality risk. This analysis was central to the revision of the paediatric index of mortality reported elsewhere in this edition [3].

We acknowledge that the system is imperfect, and that there is potential for further improvement. Nevertheless, a pragmatic approach requires compromise and a perfect system is probably not a realistic goal. We believe the benefits obtained from using a uniform system across two countries outweigh the imperfections. For international comparisons of paediatric intensive care, similar reasoning can be applied. Consensus is required on a uniform international approach to coding diagnosis and reasons for admitting children to intensive care. The specifics of the system itself are less important than the agreement to use a uniform method. Until such a system is developed, we recommend that intensive care units consider using the system described here or, alternatively, investigate the feasibility of mapping the diagnostic codes in the system that they are using to the codes in this system. This will be particularly worthwhile for units planning to collaborate internationally with assessment of intensive care outcome using PIM2 [3]. We intend to update the system periodically. A number of minor modifications have occurred since the system was first designed. We welcome suggested improvements, particularly if modification would enhance the applicability of the system to other countries.