A five year descriptive analysis of potentially preventable hospitalisations for Ear, Nose, and Throat conditions in regional Victoria, Australia, from 2015 to 2020 | BMC Public Health

Clinically meaningful subgroups of ENT PPHs

There were 4816 hospital separations in the Murray PHN between 2015 and 2020 with a primary diagnosis of ENT as defined by the PPH framework in the National Health Agreement. These could be classified into three clinically meaningful subgroups as follows: upper respiratory tract infections (URTI) (2328 separations, or 48.3%), acute tonsillitis (AT) (1832 separations, or 38.0%) and otitis media (OM) (656 separations, or 13.6%). Within each subgroup, the ‘unspecified’ ICD10-AM code for that subgroup was by far the most common code used (83.7%, 91.3% and 78.7%, respectively), preventing a more detailed classification of ENT hospitalisations into subgroups. Additional file 1 provides further details.

Postcodes of potentially preventable ENT hospitalisations

Of the 169 postcodes located within the Murray PHN catchment, 15 were identified as having higher than expected numbers of URTI hospitalisations, 14 were identified as having higher than expected numbers of AT hospitalisations and 12 were identified as having higher than expected numbers of OM hospitalisations. The relevant postcodes and the associated suburbs are depicted on Figs. 1, 2, and 3, respectively. The regional cities are named on the maps to provide context as to the towns distance from major health care resources.

Fig. 1
figure 1

Upper respiratory tract infection

Fig. 2
figure 2
Fig. 3
figure 3

The Murray PHN has four subregions: the North West, Central VIC, Goulburn Valley, and North East (see Fig. 4). Two postcodes overlapped in all three ENT subgroups, namely Ardmona (3629) and Koonoomoo (3644) (bordering with NSW), both of which are located in the Goulburn Valley. OM and URTI overlapped in three postcodes, namely Upper Ryans Creek (3673), Benalla (3672), and Wangaratta (3677), all located in the North East region. URTI and AT also overlapped in three postcodes, namely Narraport (3483) (located in the North West), Dunkirk (3630) (located in the Goulburn Valley), and Echuca South (3564) (located in Central VIC). There were no overlapping postcodes for OM and AT. Additionally, there were no postcodes with higher than expected numbers of OM hospitalisations in the North West and Central VIC regions, and no postcodes with higher than expected numbers of AT hospitalisations in the North East region.

Fig. 4
figure 4

Murray PHN subregions of Victoria

Selected patient characteristics

Table 1 summarises for each of the identified ENT subgroups (i.e. URTI, AT and OM) the number and percent of selected characteristics for Murray PHN patients residing in postcodes with higher than expected hospitalisations compared to those residing in other postcodes in the region. Results show that for OM and URTI, patients from ‘higher than expected’ postcodes were more likely than others to be aged between 0 and 9 years (75.3% vs 62.1%, p = 0.01; 60% vs 47.4%, p < 0.01 respectively). Conversely, for URTI and AT, patients from ‘higher than expected’ postcodes were more likely than others to be Indigenous (7.5% vs 5.1%, p = 0.02; 10.5% vs 5.2%, p < 0.01 respectively), require an interpreter (1.8% vs 0.3%, p < 0.01; 1.5% vs 0.4%, p = 0.01 respectively), and speak a language other than English at home (2.6% vs 1.1%, p = 0.01; 2.7% vs 1%, p = 0.01 respectively). Additionally for AT, patients from ‘higher than expected’ postcodes were more likely than others to require emergency treatment (99% vs 95.7%, p < 0.01). There was no difference between the two patient groups for the remaining combinations of ENT subgroups and patient characteristics.

Table 1 Demographic comparison of patients with potentially preventable hospitalisation, ear, nose, and throat conditions in the Murray Primary Health Network region

Results from the logistic regression analyses for each of the selected patient characteristics and ENT subgroups are shown in Table 2. The bivariate analyses confirmed the results of the Chi-squared tests above, with the exception of OM patients requiring an interpreter and OM patients who speak a language other than English at home, where unadjusted odds ratio could not be calculated using standard techniques because these characteristics perfectly predicted the likelihood of residing in a postcode with higher than expected hospitalisations due to OM.

Table 2 Odds ratios and adjusted odds ratios of patient characteristics

Adjusting for all patient characteristics simultaneously only slightly attenuated the likelihood amongst URTI patients of residing in a postcode with higher than expected hospitalisations if they were Indigenous (OR = 1.52, 95%CI 1.06–2.17 cf. AOR = 1.45, 95%CI 1.01–2.08) or required an interpreter (OR = 6.54, 95%CI 2.12–20.11 cf. AOR = 5.98, 95%CI 1.33–26.81). However, it did reduce the strength of the bivariate association with LOTE patients to non-significance (OR = 2.39, 95%CI 1.22–4.67 cf. AOR = 1.06, 95%CI 0.39–2.86) and increase the strength of the bivariate association with emergency patients to significance (OR = 0.79, 95%CI 0.55–1.11 cf. AOR = 0.65, 95%CI 0.45–0.93), suggesting a degree of confounding with these characteristics in the bivariate analyses.

The picture with AT patients was similar with only a slight attenuation in the likelihood of residing in a postcode with higher than expected hospitalisations if patients were Indigenous (OR = 2.13, 95%CI 1.47–3.09 cf. AOR = 2.11, 95%CI 1.45–3.06) or required emergency treatment (OR = 4.42, 95%CI 1.89–10.38 cf. OR = 5.98, 95%CI 1.33–26.81) but a reduction in the strength of the association to non-significance for patients requiring an interpreter (OR = 4.44, 95%CI 1.36–14.47 cf. AOR = 2.09, 95%CI 0.43–10.22) and LOTE patients (OR = 2.89, 95%CI 1.33–6.26 cf. AOR = 2.18, 95%CI 0.75–6.34), again suggesting a degree of confounding with these characteristics.

For OM patients, the strength of the association between residing in a postcode with higher than expected hospitalisations and being a transfer patient increased to significance (OR = 3.35, 95%CI 0.89–12.62 cf. AOR = 4.72 95%CI 1.20–18.54) after adjusting for all patient characteristics simultaneously, suggesting a degree of confounding with this characteristic in the bivariate analysis. In the univariate analysis, transfer patients were no more likely than others to reside in postcodes with higher than expected numbers of OM presentations. This patient-level attribute only became significant in the multivariate logistic regression for OM suggesting a degree of confounding in the bivariate analysis (we have no plausible explanation for this finding).

For both URTI and OM patients, the strength of the association between residing in a postcode with higher than expected hospitalisations and the 20 + year old age group relative to the 0–9 year old age group did not change after adjusting for all patient characteristics simultaneously (OR = 0.58, 95%CI 0.48–0.70 cf. AOR = 0.58, 95%CI 0.48–0.71 and OR = 0.55, 95%CI 0.35–0.86 cf. AOR = 0.55, 95%CI 0.34–0.88), suggesting no confounding due to age in the bivariate analyses.

link

Leave a Reply

Your email address will not be published. Required fields are marked *